linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software
@ 2020-10-30 15:02 Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 01/10] fs/ntfs3: Add headers and misc files Konstantin Komarov
                   ` (12 more replies)
  0 siblings, 13 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This patch adds NTFS Read-Write driver to fs/ntfs3.

Having decades of expertise in commercial file systems development and huge
test coverage, we at Paragon Software GmbH want to make our contribution to
the Open Source Community by providing implementation of NTFS Read-Write
driver for the Linux Kernel.

This is fully functional NTFS Read-Write driver. Current version works with
NTFS(including v3.1) and normal/compressed/sparse files and supports journal replaying.

We plan to support this version after the codebase once merged, and add new
features and fix bugs. For example, full journaling support over JBD will be
added in later updates.

v2:
 - patch splitted to chunks (file-wise)
 - build issues fixed
 - sparse and checkpatch.pl errors fixed
 - NULL pointer dereference on mkfs.ntfs-formatted volume mount fixed
 - cosmetics + code cleanup

v3:
 - added acl, noatime, no_acs_rules, prealloc mount options
 - added fiemap support
 - fixed encodings support
 - removed typedefs
 - adapted Kernel-way logging mechanisms
 - fixed typos and corner-case issues

v4:
 - atomic_open() refactored
 - code style updated
 - bugfixes

v5:
- nls/nls_alt mount options added
- Unicode conversion fixes
- Improved very fragmented files operations
- logging cosmetics

v6:
- Security Descriptors processing changed
  added system.ntfs_security xattr to set
  SD
- atomic_open() optimized
- cosmetics

v7:
- Security Descriptors validity checks added (by Mark Harmstone)
- atomic_open() fixed for the compressed file creation with directio
  case
- remount support
- temporarily removed readahead usage
- cosmetics

v8:
- Compressed files operations fixed

v9:
- Further cosmetics applied as suggested
by Joe Perches

v10:
- operations with compressed/sparse files on very fragmented volumes improved
- reduced memory consumption for above cases

v11:
- further compressed files optimizations: reads/writes are now skipping bufferization
- journal wipe to the initial state optimized (bufferization is also skipped)
- optimized run storage (re-packing cluster metainformation)
- fixes based on Matthew Wilcox feedback to the v10
- compressed/sparse/normal could be set for empty files with 'system.ntfs_attrib' xattr

Konstantin Komarov (10):
  fs/ntfs3: Add headers and misc files
  fs/ntfs3: Add initialization of super block
  fs/ntfs3: Add bitmap
  fs/ntfs3: Add file operations and implementation
  fs/ntfs3: Add attrib operations
  fs/ntfs3: Add compression
  fs/ntfs3: Add NTFS journal
  fs/ntfs3: Add Kconfig, Makefile and doc
  fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile
  fs/ntfs3: Add MAINTAINERS

 Documentation/filesystems/ntfs3.rst |  112 +
 MAINTAINERS                         |    7 +
 fs/Kconfig                          |    1 +
 fs/Makefile                         |    1 +
 fs/ntfs3/Kconfig                    |   23 +
 fs/ntfs3/Makefile                   |   11 +
 fs/ntfs3/attrib.c                   | 1400 +++++++
 fs/ntfs3/attrlist.c                 |  463 +++
 fs/ntfs3/bitfunc.c                  |  135 +
 fs/ntfs3/bitmap.c                   | 1504 ++++++++
 fs/ntfs3/debug.h                    |   61 +
 fs/ntfs3/dir.c                      |  610 ++++
 fs/ntfs3/file.c                     | 1153 ++++++
 fs/ntfs3/frecord.c                  | 2696 ++++++++++++++
 fs/ntfs3/fslog.c                    | 5221 +++++++++++++++++++++++++++
 fs/ntfs3/fsntfs.c                   | 2565 +++++++++++++
 fs/ntfs3/index.c                    | 2665 ++++++++++++++
 fs/ntfs3/inode.c                    | 2030 +++++++++++
 fs/ntfs3/lznt.c                     |  452 +++
 fs/ntfs3/namei.c                    |  576 +++
 fs/ntfs3/ntfs.h                     | 1262 +++++++
 fs/ntfs3/ntfs_fs.h                  |  988 +++++
 fs/ntfs3/record.c                   |  613 ++++
 fs/ntfs3/run.c                      | 1190 ++++++
 fs/ntfs3/super.c                    | 1479 ++++++++
 fs/ntfs3/upcase.c                   |   77 +
 fs/ntfs3/xattr.c                    | 1069 ++++++
 27 files changed, 28364 insertions(+)
 create mode 100644 Documentation/filesystems/ntfs3.rst
 create mode 100644 fs/ntfs3/Kconfig
 create mode 100644 fs/ntfs3/Makefile
 create mode 100644 fs/ntfs3/attrib.c
 create mode 100644 fs/ntfs3/attrlist.c
 create mode 100644 fs/ntfs3/bitfunc.c
 create mode 100644 fs/ntfs3/bitmap.c
 create mode 100644 fs/ntfs3/debug.h
 create mode 100644 fs/ntfs3/dir.c
 create mode 100644 fs/ntfs3/file.c
 create mode 100644 fs/ntfs3/frecord.c
 create mode 100644 fs/ntfs3/fslog.c
 create mode 100644 fs/ntfs3/fsntfs.c
 create mode 100644 fs/ntfs3/index.c
 create mode 100644 fs/ntfs3/inode.c
 create mode 100644 fs/ntfs3/lznt.c
 create mode 100644 fs/ntfs3/namei.c
 create mode 100644 fs/ntfs3/ntfs.h
 create mode 100644 fs/ntfs3/ntfs_fs.h
 create mode 100644 fs/ntfs3/record.c
 create mode 100644 fs/ntfs3/run.c
 create mode 100644 fs/ntfs3/super.c
 create mode 100644 fs/ntfs3/upcase.c
 create mode 100644 fs/ntfs3/xattr.c

-- 
2.25.4


^ permalink raw reply	[flat|nested] 21+ messages in thread

* [PATCH v11 01/10] fs/ntfs3: Add headers and misc files
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
@ 2020-10-30 15:02 ` Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 02/10] fs/ntfs3: Add initialization of super block Konstantin Komarov
                   ` (11 subsequent siblings)
  12 siblings, 0 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This adds headers and misc files

Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
---
 fs/ntfs3/debug.h   |   61 +++
 fs/ntfs3/ntfs.h    | 1262 ++++++++++++++++++++++++++++++++++++++++++++
 fs/ntfs3/ntfs_fs.h |  988 ++++++++++++++++++++++++++++++++++
 fs/ntfs3/upcase.c  |   77 +++
 4 files changed, 2388 insertions(+)
 create mode 100644 fs/ntfs3/debug.h
 create mode 100644 fs/ntfs3/ntfs.h
 create mode 100644 fs/ntfs3/ntfs_fs.h
 create mode 100644 fs/ntfs3/upcase.c

diff --git a/fs/ntfs3/debug.h b/fs/ntfs3/debug.h
new file mode 100644
index 000000000000..9e57ef878b05
--- /dev/null
+++ b/fs/ntfs3/debug.h
@@ -0,0 +1,61 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ * useful functions for debuging
+ */
+
+// clang-format off
+#ifndef Add2Ptr
+#define Add2Ptr(P, I)		((void *)((u8 *)(P) + (I)))
+#define PtrOffset(B, O)		((size_t)((size_t)(O) - (size_t)(B)))
+#endif
+
+#define QuadAlign(n)		(((n) + 7u) & (~7u))
+#define IsQuadAligned(n)	(!((size_t)(n)&7u))
+#define Quad2Align(n)		(((n) + 15u) & (~15u))
+#define IsQuad2Aligned(n)	(!((size_t)(n)&15u))
+#define Quad4Align(n)		(((n) + 31u) & (~31u))
+#define IsSizeTAligned(n)	(!((size_t)(n) & (sizeof(size_t) - 1)))
+#define DwordAlign(n)		(((n) + 3u) & (~3u))
+#define IsDwordAligned(n)	(!((size_t)(n)&3u))
+#define WordAlign(n)		(((n) + 1u) & (~1u))
+#define IsWordAligned(n)	(!((size_t)(n)&1u))
+
+#ifdef CONFIG_PRINTK
+__printf(2, 3)
+void ntfs_printk(const struct super_block *sb, const char *fmt, ...);
+__printf(2, 3)
+void ntfs_inode_printk(struct inode *inode, const char *fmt, ...);
+#else
+static inline __printf(2, 3)
+void ntfs_printk(const struct super_block *sb, const char *fmt, ...)
+{
+}
+
+static inline __printf(2, 3)
+void ntfs_inode_printk(struct inode *inode, const char *fmt, ...)
+{
+}
+#endif
+
+/*
+ * Logging macros ( thanks Joe Perches <joe@perches.com> for implementation )
+ */
+
+#define ntfs_err(sb, fmt, ...)  ntfs_printk(sb, KERN_ERR fmt, ##__VA_ARGS__)
+#define ntfs_warn(sb, fmt, ...) ntfs_printk(sb, KERN_WARNING fmt, ##__VA_ARGS__)
+#define ntfs_info(sb, fmt, ...) ntfs_printk(sb, KERN_INFO fmt, ##__VA_ARGS__)
+#define ntfs_notice(sb, fmt, ...)                                              \
+	ntfs_printk(sb, KERN_NOTICE fmt, ##__VA_ARGS__)
+
+#define ntfs_inode_err(inode, fmt, ...)                                        \
+	ntfs_inode_printk(inode, KERN_ERR fmt, ##__VA_ARGS__)
+#define ntfs_inode_warn(inode, fmt, ...)                                       \
+	ntfs_inode_printk(inode, KERN_WARNING fmt, ##__VA_ARGS__)
+
+#define ntfs_alloc(s, z)	kmalloc(s, (z) ? (GFP_NOFS | __GFP_ZERO) : GFP_NOFS)
+#define ntfs_free(p)		kfree(p)
+#define ntfs_memdup(src, len)	kmemdup(src, len, GFP_NOFS)
+// clang-format on
diff --git a/fs/ntfs3/ntfs.h b/fs/ntfs3/ntfs.h
new file mode 100644
index 000000000000..8af2f8f489e5
--- /dev/null
+++ b/fs/ntfs3/ntfs.h
@@ -0,0 +1,1262 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ * on-disk ntfs structs
+ */
+
+// clang-format off
+
+/* TODO:
+ * - Check 4K mft record and 512 bytes cluster
+ */
+
+/*
+ * Activate this define to use binary search in indexes
+ */
+#define NTFS3_INDEX_BINARY_SEARCH
+
+/*
+ * Check each run for marked clusters
+ */
+#define NTFS3_CHECK_FREE_CLST
+
+#define NTFS_NAME_LEN 255
+
+/*
+ * ntfs.sys used 500 maximum links
+ * on-disk struct allows up to 0xffff
+ */
+#define NTFS_LINK_MAX 0x400
+//#define NTFS_LINK_MAX 0xffff
+
+/*
+ * Activate to use 64 bit clusters instead of 32 bits in ntfs.sys
+ * Logical and virtual cluster number
+ * If needed, may be redefined to use 64 bit value
+ */
+//#define NTFS3_64BIT_CLUSTER
+
+#define NTFS_LZNT_MAX_CLUSTER	4096
+#define NTFS_LZNT_CUNIT		4
+#define NTFS_LZNT_CLUSTERS	(1u<<NTFS_LZNT_CUNIT)
+
+struct GUID {
+	__le32 Data1;
+	__le16 Data2;
+	__le16 Data3;
+	u8 Data4[8];
+};
+
+/*
+ * this struct repeats layout of ATTR_FILE_NAME
+ * at offset 0x40
+ * it used to store global constants NAME_MFT/NAME_MIRROR...
+ * most constant names are shorter than 10
+ */
+struct cpu_str {
+	u8 len;
+	u8 unused;
+	u16 name[10];
+};
+
+struct le_str {
+	u8 len;
+	u8 unused;
+	__le16 name[1];
+};
+
+static_assert(SECTOR_SHIFT == 9);
+
+#ifdef NTFS3_64BIT_CLUSTER
+typedef u64 CLST;
+static_assert(sizeof(size_t) == 8);
+#else
+typedef u32 CLST;
+#endif
+
+#define SPARSE_LCN     ((CLST)-1)
+#define RESIDENT_LCN   ((CLST)-2)
+#define COMPRESSED_LCN ((CLST)-3)
+
+#define COMPRESSION_UNIT     4
+#define COMPRESS_MAX_CLUSTER 0x1000
+#define MFT_INCREASE_CHUNK   1024
+
+enum RECORD_NUM {
+	MFT_REC_MFT		= 0,
+	MFT_REC_MIRR		= 1,
+	MFT_REC_LOG		= 2,
+	MFT_REC_VOL		= 3,
+	MFT_REC_ATTR		= 4,
+	MFT_REC_ROOT		= 5,
+	MFT_REC_BITMAP		= 6,
+	MFT_REC_BOOT		= 7,
+	MFT_REC_BADCLUST	= 8,
+	//MFT_REC_QUOTA		= 9,
+	MFT_REC_SECURE		= 9, // NTFS 3.0
+	MFT_REC_UPCASE		= 10,
+	MFT_REC_EXTEND		= 11, // NTFS 3.0
+	MFT_REC_RESERVED	= 11,
+	MFT_REC_FREE		= 16,
+	MFT_REC_USER		= 24,
+};
+
+enum ATTR_TYPE {
+	ATTR_ZERO		= cpu_to_le32(0x00),
+	ATTR_STD		= cpu_to_le32(0x10),
+	ATTR_LIST		= cpu_to_le32(0x20),
+	ATTR_NAME		= cpu_to_le32(0x30),
+	// ATTR_VOLUME_VERSION on Nt4
+	ATTR_ID			= cpu_to_le32(0x40),
+	ATTR_SECURE		= cpu_to_le32(0x50),
+	ATTR_LABEL		= cpu_to_le32(0x60),
+	ATTR_VOL_INFO		= cpu_to_le32(0x70),
+	ATTR_DATA		= cpu_to_le32(0x80),
+	ATTR_ROOT		= cpu_to_le32(0x90),
+	ATTR_ALLOC		= cpu_to_le32(0xA0),
+	ATTR_BITMAP		= cpu_to_le32(0xB0),
+	// ATTR_SYMLINK on Nt4
+	ATTR_REPARSE		= cpu_to_le32(0xC0),
+	ATTR_EA_INFO		= cpu_to_le32(0xD0),
+	ATTR_EA			= cpu_to_le32(0xE0),
+	ATTR_PROPERTYSET	= cpu_to_le32(0xF0),
+	ATTR_LOGGED_UTILITY_STREAM = cpu_to_le32(0x100),
+	ATTR_END		= cpu_to_le32(0xFFFFFFFF)
+};
+
+static_assert(sizeof(enum ATTR_TYPE) == 4);
+
+enum FILE_ATTRIBUTE {
+	FILE_ATTRIBUTE_READONLY		= cpu_to_le32(0x00000001),
+	FILE_ATTRIBUTE_HIDDEN		= cpu_to_le32(0x00000002),
+	FILE_ATTRIBUTE_SYSTEM		= cpu_to_le32(0x00000004),
+	FILE_ATTRIBUTE_ARCHIVE		= cpu_to_le32(0x00000020),
+	FILE_ATTRIBUTE_DEVICE		= cpu_to_le32(0x00000040),
+	FILE_ATTRIBUTE_TEMPORARY	= cpu_to_le32(0x00000100),
+	FILE_ATTRIBUTE_SPARSE_FILE	= cpu_to_le32(0x00000200),
+	FILE_ATTRIBUTE_REPARSE_POINT	= cpu_to_le32(0x00000400),
+	FILE_ATTRIBUTE_COMPRESSED	= cpu_to_le32(0x00000800),
+	FILE_ATTRIBUTE_OFFLINE		= cpu_to_le32(0x00001000),
+	FILE_ATTRIBUTE_NOT_CONTENT_INDEXED = cpu_to_le32(0x00002000),
+	FILE_ATTRIBUTE_ENCRYPTED	= cpu_to_le32(0x00004000),
+	FILE_ATTRIBUTE_VALID_FLAGS	= cpu_to_le32(0x00007fb7),
+	FILE_ATTRIBUTE_DIRECTORY	= cpu_to_le32(0x10000000),
+};
+
+static_assert(sizeof(enum FILE_ATTRIBUTE) == 4);
+
+extern const struct cpu_str NAME_MFT;
+extern const struct cpu_str NAME_MIRROR;
+extern const struct cpu_str NAME_LOGFILE;
+extern const struct cpu_str NAME_VOLUME;
+extern const struct cpu_str NAME_ATTRDEF;
+extern const struct cpu_str NAME_ROOT;
+extern const struct cpu_str NAME_BITMAP;
+extern const struct cpu_str NAME_BOOT;
+extern const struct cpu_str NAME_BADCLUS;
+extern const struct cpu_str NAME_QUOTA;
+extern const struct cpu_str NAME_SECURE;
+extern const struct cpu_str NAME_UPCASE;
+extern const struct cpu_str NAME_EXTEND;
+extern const struct cpu_str NAME_OBJID;
+extern const struct cpu_str NAME_REPARSE;
+extern const struct cpu_str NAME_USNJRNL;
+
+extern const __le16 I30_NAME[4];
+extern const __le16 SII_NAME[4];
+extern const __le16 SDH_NAME[4];
+extern const __le16 SO_NAME[2];
+extern const __le16 SQ_NAME[2];
+extern const __le16 SR_NAME[2];
+
+extern const __le16 BAD_NAME[4];
+extern const __le16 SDS_NAME[4];
+extern const __le16 EFS_NAME[4];
+extern const __le16 WOF_NAME[17];
+extern const __le16 J_NAME[2];
+extern const __le16 MAX_NAME[4];
+
+/* MFT record number structure */
+struct MFT_REF {
+	__le32 low;	// The low part of the number
+	__le16 high;	// The high part of the number
+	__le16 seq;	// The sequence number of MFT record
+};
+
+static_assert(sizeof(__le64) == sizeof(struct MFT_REF));
+
+static inline CLST ino_get(const struct MFT_REF *ref)
+{
+#ifdef NTFS3_64BIT_CLUSTER
+	return le32_to_cpu(ref->low) | ((u64)le16_to_cpu(ref->high) << 32);
+#else
+	return le32_to_cpu(ref->low);
+#endif
+}
+
+struct NTFS_BOOT {
+	u8 jump_code[3];	// 0x00: Jump to boot code
+	u8 system_id[8];	// 0x03: System ID, equals "NTFS    "
+
+	// NOTE: this member is not aligned(!)
+	// bytes_per_sector[0] must be 0
+	// bytes_per_sector[1] must be multiplied by 256
+	u8 bytes_per_sector[2];	// 0x0B: Bytes per sector
+
+	u8 sectors_per_clusters;// 0x0D: Sectors per cluster
+	u8 unused1[7];
+	u8 media_type;		// 0x15: Media type (0xF8 - harddisk)
+	u8 unused2[2];
+	__le16 sct_per_track;	// 0x18: number of sectors per track
+	__le16 heads;		// 0x1A: number of heads per cylinder
+	__le32 hidden_sectors;	// 0x1C: number of 'hidden' sectors
+	u8 unused3[4];
+	u8 bios_drive_num;	// 0x24: BIOS drive number =0x80
+	u8 unused4;
+	u8 signature_ex;	// 0x26: Extended BOOT signature =0x80
+	u8 unused5;
+	__le64 sectors_per_volume;// 0x28: size of volume in sectors
+	__le64 mft_clst;	// 0x30: first cluster of $MFT
+	__le64 mft2_clst;	// 0x38: first cluster of $MFTMirr
+	s8 record_size;		// 0x40: size of MFT record in clusters(sectors)
+	u8 unused6[3];
+	s8 index_size;		// 0x44: size of INDX record in clusters(sectors)
+	u8 unused7[3];
+	__le64 serial_num;	// 0x48: Volume serial number
+	__le32 check_sum;	// 0x50: Simple additive checksum of all
+				// of the u32's which precede the 'check_sum'
+
+	u8 boot_code[0x200 - 0x50 - 2 - 4]; // 0x54:
+	u8 boot_magic[2];	// 0x1FE: Boot signature =0x55 + 0xAA
+};
+
+static_assert(sizeof(struct NTFS_BOOT) == 0x200);
+
+enum NTFS_SIGNATURE {
+	NTFS_FILE_SIGNATURE = cpu_to_le32(0x454C4946), // 'FILE'
+	NTFS_INDX_SIGNATURE = cpu_to_le32(0x58444E49), // 'INDX'
+	NTFS_CHKD_SIGNATURE = cpu_to_le32(0x444B4843), // 'CHKD'
+	NTFS_RSTR_SIGNATURE = cpu_to_le32(0x52545352), // 'RSTR'
+	NTFS_RCRD_SIGNATURE = cpu_to_le32(0x44524352), // 'RCRD'
+	NTFS_BAAD_SIGNATURE = cpu_to_le32(0x44414142), // 'BAAD'
+	NTFS_HOLE_SIGNATURE = cpu_to_le32(0x454C4F48), // 'HOLE'
+	NTFS_FFFF_SIGNATURE = cpu_to_le32(0xffffffff),
+};
+
+static_assert(sizeof(enum NTFS_SIGNATURE) == 4);
+
+/* MFT Record header structure */
+struct NTFS_RECORD_HEADER {
+	/* Record magic number, equals 'FILE'/'INDX'/'RSTR'/'RCRD' */
+	enum NTFS_SIGNATURE sign; // 0x00:
+	__le16 fix_off;		// 0x04:
+	__le16 fix_num;		// 0x06:
+	__le64 lsn;		// 0x08: Log file sequence number
+};
+
+static_assert(sizeof(struct NTFS_RECORD_HEADER) == 0x10);
+
+static inline int is_baad(const struct NTFS_RECORD_HEADER *hdr)
+{
+	return hdr->sign == NTFS_BAAD_SIGNATURE;
+}
+
+/* Possible bits in struct MFT_REC.flags */
+enum RECORD_FLAG {
+	RECORD_FLAG_IN_USE	= cpu_to_le16(0x0001),
+	RECORD_FLAG_DIR		= cpu_to_le16(0x0002),
+	RECORD_FLAG_SYSTEM	= cpu_to_le16(0x0004),
+	RECORD_FLAG_UNKNOWN	= cpu_to_le16(0x0008),
+};
+
+/* MFT Record structure */
+struct MFT_REC {
+	struct NTFS_RECORD_HEADER rhdr; // 'FILE'
+
+	__le16 seq;		// 0x10: Sequence number for this record
+	__le16 hard_links;	// 0x12: The number of hard links to record
+	__le16 attr_off;	// 0x14: Offset to attributes
+	__le16 flags;		// 0x16: See RECORD_FLAG
+	__le32 used;		// 0x18: The size of used part
+	__le32 total;		// 0x1C: Total record size
+
+	struct MFT_REF parent_ref; // 0x20: Parent MFT record
+	__le16 next_attr_id;	// 0x28: The next attribute Id
+
+	__le16 res;		// 0x2A: High part of mft record?
+	__le32 mft_record;	// 0x2C: Current mft record number
+	__le16 fixups[1];	// 0x30:
+};
+
+#define MFTRECORD_FIXUP_OFFSET_1 offsetof(struct MFT_REC, res)
+#define MFTRECORD_FIXUP_OFFSET_3 offsetof(struct MFT_REC, fixups)
+
+static_assert(MFTRECORD_FIXUP_OFFSET_1 == 0x2A);
+static_assert(MFTRECORD_FIXUP_OFFSET_3 == 0x30);
+
+static inline bool is_rec_base(const struct MFT_REC *rec)
+{
+	const struct MFT_REF *r = &rec->parent_ref;
+
+	return !r->low && !r->high && !r->seq;
+}
+
+static inline bool is_mft_rec5(const struct MFT_REC *rec)
+{
+	return le16_to_cpu(rec->rhdr.fix_off) >=
+	       offsetof(struct MFT_REC, fixups);
+}
+
+static inline bool is_rec_inuse(const struct MFT_REC *rec)
+{
+	return rec->flags & RECORD_FLAG_IN_USE;
+}
+
+static inline bool clear_rec_inuse(struct MFT_REC *rec)
+{
+	return rec->flags &= ~RECORD_FLAG_IN_USE;
+}
+
+/* Possible values of ATTR_RESIDENT.flags */
+#define RESIDENT_FLAG_INDEXED 0x01
+
+struct ATTR_RESIDENT {
+	__le32 data_size;	// 0x10: The size of data
+	__le16 data_off;	// 0x14: Offset to data
+	u8 flags;		// 0x16: resident flags ( 1 - indexed )
+	u8 res;			// 0x17:
+}; // sizeof() = 0x18
+
+struct ATTR_NONRESIDENT {
+	__le64 svcn;		// 0x10: Starting VCN of this segment
+	__le64 evcn;		// 0x18: End VCN of this segment
+	__le16 run_off;		// 0x20: Offset to packed runs
+	//  Unit of Compression size for this stream, expressed
+	//  as a log of the cluster size.
+	//
+	//	0 means file is not compressed
+	//	1, 2, 3, and 4 are potentially legal values if the
+	//	    stream is compressed, however the implementation
+	//	    may only choose to use 4, or possibly 3.  Note
+	//	    that 4 means cluster size time 16.	If convenient
+	//	    the implementation may wish to accept a
+	//	    reasonable range of legal values here (1-5?),
+	//	    even if the implementation only generates
+	//	    a smaller set of values itself.
+	u8 c_unit;		// 0x22
+	u8 res1[5];		// 0x23:
+	__le64 alloc_size;	// 0x28: The allocated size of attribute in bytes
+				// (multiple of cluster size)
+	__le64 data_size;	// 0x30: The size of attribute  in bytes <= alloc_size
+	__le64 valid_size;	// 0x38: The size of valid part in bytes <= data_size
+	__le64 total_size;	// 0x40: The sum of the allocated clusters for a file
+				// (present only for the first segment (0 == vcn)
+				// of compressed attribute)
+
+}; // sizeof()=0x40 or 0x48 (if compressed)
+
+/* Possible values of ATTRIB.flags: */
+#define ATTR_FLAG_COMPRESSED	  cpu_to_le16(0x0001)
+#define ATTR_FLAG_COMPRESSED_MASK cpu_to_le16(0x00FF)
+#define ATTR_FLAG_ENCRYPTED	  cpu_to_le16(0x4000)
+#define ATTR_FLAG_SPARSED	  cpu_to_le16(0x8000)
+
+struct ATTRIB {
+	enum ATTR_TYPE type;	// 0x00: The type of this attribute
+	__le32 size;		// 0x04: The size of this attribute
+	u8 non_res;		// 0x08: Is this attribute non-resident ?
+	u8 name_len;		// 0x09: This attribute name length
+	__le16 name_off;	// 0x0A: Offset to the attribute name
+	__le16 flags;		// 0x0C: See ATTR_FLAG_XXX
+	__le16 id;		// 0x0E: unique id (per record)
+
+	union {
+		struct ATTR_RESIDENT res;     // 0x10
+		struct ATTR_NONRESIDENT nres; // 0x10
+	};
+};
+
+/* Define attribute sizes */
+#define SIZEOF_RESIDENT			0x18
+#define SIZEOF_NONRESIDENT_EX		0x48
+#define SIZEOF_NONRESIDENT		0x40
+
+#define SIZEOF_RESIDENT_LE		cpu_to_le16(0x18)
+#define SIZEOF_NONRESIDENT_EX_LE	cpu_to_le16(0x48)
+#define SIZEOF_NONRESIDENT_LE		cpu_to_le16(0x40)
+
+static inline u64 attr_ondisk_size(const struct ATTRIB *attr)
+{
+	return attr->non_res ? ((attr->flags &
+				 (ATTR_FLAG_COMPRESSED | ATTR_FLAG_SPARSED)) ?
+					le64_to_cpu(attr->nres.total_size) :
+					le64_to_cpu(attr->nres.alloc_size)) :
+			       QuadAlign(le32_to_cpu(attr->res.data_size));
+}
+
+static inline u64 attr_size(const struct ATTRIB *attr)
+{
+	return attr->non_res ? le64_to_cpu(attr->nres.data_size) :
+			       le32_to_cpu(attr->res.data_size);
+}
+
+static inline bool is_attr_encrypted(const struct ATTRIB *attr)
+{
+	return attr->flags & ATTR_FLAG_ENCRYPTED;
+}
+
+static inline bool is_attr_sparsed(const struct ATTRIB *attr)
+{
+	return attr->flags & ATTR_FLAG_SPARSED;
+}
+
+static inline bool is_attr_compressed(const struct ATTRIB *attr)
+{
+	return attr->flags & ATTR_FLAG_COMPRESSED;
+}
+
+static inline bool is_attr_ext(const struct ATTRIB *attr)
+{
+	return attr->flags & (ATTR_FLAG_SPARSED | ATTR_FLAG_COMPRESSED);
+}
+
+static inline bool is_attr_indexed(const struct ATTRIB *attr)
+{
+	return !attr->non_res && (attr->res.flags & RESIDENT_FLAG_INDEXED);
+}
+
+static const inline __le16 *attr_name(const struct ATTRIB *attr)
+{
+	return Add2Ptr(attr, le16_to_cpu(attr->name_off));
+}
+
+static inline u64 attr_svcn(const struct ATTRIB *attr)
+{
+	return attr->non_res ? le64_to_cpu(attr->nres.svcn) : 0;
+}
+
+/* the size of resident attribute by its resident size */
+#define BYTES_PER_RESIDENT(b) (0x18 + (b))
+
+static_assert(sizeof(struct ATTRIB) == 0x48);
+static_assert(sizeof(((struct ATTRIB *)NULL)->res) == 0x08);
+static_assert(sizeof(((struct ATTRIB *)NULL)->nres) == 0x38);
+
+static inline void *resident_data_ex(const struct ATTRIB *attr, u32 datasize)
+{
+	u32 asize, rsize;
+	u16 off;
+
+	if (attr->non_res)
+		return NULL;
+
+	asize = le32_to_cpu(attr->size);
+	off = le16_to_cpu(attr->res.data_off);
+
+	if (asize < datasize + off)
+		return NULL;
+
+	rsize = le32_to_cpu(attr->res.data_size);
+	if (rsize < datasize)
+		return NULL;
+
+	return Add2Ptr(attr, off);
+}
+
+static inline void *resident_data(const struct ATTRIB *attr)
+{
+	return Add2Ptr(attr, le16_to_cpu(attr->res.data_off));
+}
+
+static inline void *attr_run(const struct ATTRIB *attr)
+{
+	return Add2Ptr(attr, le16_to_cpu(attr->nres.run_off));
+}
+
+/* Standard information attribute (0x10) */
+struct ATTR_STD_INFO {
+	__le64 cr_time;		// 0x00: File creation file
+	__le64 m_time;		// 0x08: File modification time
+	__le64 c_time;		// 0x10: Last time any attribute was modified
+	__le64 a_time;		// 0x18: File last access time
+	enum FILE_ATTRIBUTE fa; // 0x20: Standard DOS attributes & more
+	__le32 max_ver_num;	// 0x24: Maximum Number of Versions
+	__le32 ver_num;		// 0x28: Version Number
+	__le32 class_id;	// 0x2C: Class Id from bidirectional Class Id index
+};
+
+static_assert(sizeof(struct ATTR_STD_INFO) == 0x30);
+
+#define SECURITY_ID_INVALID 0x00000000
+#define SECURITY_ID_FIRST 0x00000100
+
+struct ATTR_STD_INFO5 {
+	__le64 cr_time;		// 0x00: File creation file
+	__le64 m_time;		// 0x08: File modification time
+	__le64 c_time;		// 0x10: Last time any attribute was modified
+	__le64 a_time;		// 0x18: File last access time
+	enum FILE_ATTRIBUTE fa; // 0x20: Standard DOS attributes & more
+	__le32 max_ver_num;	// 0x24: Maximum Number of Versions
+	__le32 ver_num;		// 0x28: Version Number
+	__le32 class_id;	// 0x2C: Class Id from bidirectional Class Id index
+
+	__le32 owner_id;	// 0x30: Owner Id of the user owning the file.
+	__le32 security_id;	// 0x34: The Security Id is a key in the $SII Index and $SDS
+	__le64 quota_charge;	// 0x38:
+	__le64 usn;		// 0x40: Last Update Sequence Number of the file. This is a direct
+				// index into the file $UsnJrnl. If zero, the USN Journal is
+				// disabled.
+};
+
+static_assert(sizeof(struct ATTR_STD_INFO5) == 0x48);
+
+/* attribute list entry structure (0x20) */
+struct ATTR_LIST_ENTRY {
+	enum ATTR_TYPE type;	// 0x00: The type of attribute
+	__le16 size;		// 0x04: The size of this record
+	u8 name_len;		// 0x06: The length of attribute name
+	u8 name_off;		// 0x07: The offset to attribute name
+	__le64 vcn;		// 0x08: Starting VCN of this attribute
+	struct MFT_REF ref;	// 0x10: MFT record number with attribute
+	__le16 id;		// 0x18: struct ATTRIB ID
+	__le16 name[3];		// 0x1A: Just to align. To get real name can use bNameOffset
+
+}; // sizeof(0x20)
+
+static_assert(sizeof(struct ATTR_LIST_ENTRY) == 0x20);
+
+static inline u32 le_size(u8 name_len)
+{
+	return QuadAlign(offsetof(struct ATTR_LIST_ENTRY, name) +
+			 name_len * sizeof(short));
+}
+
+/* returns 0 if 'attr' has the same type and name */
+static inline int le_cmp(const struct ATTR_LIST_ENTRY *le,
+			 const struct ATTRIB *attr)
+{
+	return le->type != attr->type || le->name_len != attr->name_len ||
+	       (!le->name_len &&
+		memcmp(Add2Ptr(le, le->name_off),
+		       Add2Ptr(attr, le16_to_cpu(attr->name_off)),
+		       le->name_len * sizeof(short)));
+}
+
+static const inline __le16 *le_name(const struct ATTR_LIST_ENTRY *le)
+{
+	return Add2Ptr(le, le->name_off);
+}
+
+/* File name types (the field type in struct ATTR_FILE_NAME ) */
+#define FILE_NAME_POSIX   0
+#define FILE_NAME_UNICODE 1
+#define FILE_NAME_DOS	  2
+#define FILE_NAME_UNICODE_AND_DOS (FILE_NAME_DOS | FILE_NAME_UNICODE)
+
+/* Filename attribute structure (0x30) */
+struct NTFS_DUP_INFO {
+	__le64 cr_time;		// 0x00: File creation file
+	__le64 m_time;		// 0x08: File modification time
+	__le64 c_time;		// 0x10: Last time any attribute was modified
+	__le64 a_time;		// 0x18: File last access time
+	__le64 alloc_size;	// 0x20: Data attribute allocated size, multiple of cluster size
+	__le64 data_size;	// 0x28: Data attribute size <= Dataalloc_size
+	enum FILE_ATTRIBUTE fa;	// 0x30: Standard DOS attributes & more
+	__le16 ea_size;		// 0x34: Packed EAs
+	__le16 reparse;		// 0x36: Used by Reparse
+
+}; // 0x38
+
+struct ATTR_FILE_NAME {
+	struct MFT_REF home;	// 0x00: MFT record for directory
+	struct NTFS_DUP_INFO dup;// 0x08
+	u8 name_len;		// 0x40: File name length in words
+	u8 type;		// 0x41: File name type
+	__le16 name[1];		// 0x42: File name
+};
+
+static_assert(sizeof(((struct ATTR_FILE_NAME *)NULL)->dup) == 0x38);
+static_assert(offsetof(struct ATTR_FILE_NAME, name) == 0x42);
+#define SIZEOF_ATTRIBUTE_FILENAME     0x44
+#define SIZEOF_ATTRIBUTE_FILENAME_MAX (0x42 + 255 * 2)
+
+static inline struct ATTRIB *attr_from_name(struct ATTR_FILE_NAME *fname)
+{
+	return (struct ATTRIB *)((char *)fname - SIZEOF_RESIDENT);
+}
+
+static inline u16 fname_full_size(const struct ATTR_FILE_NAME *fname)
+{
+	return offsetof(struct ATTR_FILE_NAME, name) +
+	       fname->name_len * sizeof(short);
+}
+
+static inline u8 paired_name(u8 type)
+{
+	if (type == FILE_NAME_UNICODE)
+		return FILE_NAME_DOS;
+	if (type == FILE_NAME_DOS)
+		return FILE_NAME_UNICODE;
+	return FILE_NAME_POSIX;
+}
+
+/* Index entry defines ( the field flags in NtfsDirEntry ) */
+#define NTFS_IE_HAS_SUBNODES	cpu_to_le16(1)
+#define NTFS_IE_LAST		cpu_to_le16(2)
+
+/* Directory entry structure */
+struct NTFS_DE {
+	union {
+		struct MFT_REF ref; // 0x00: MFT record number with this file
+		struct {
+			__le16 data_off;  // 0x00:
+			__le16 data_size; // 0x02:
+			__le32 res;	  // 0x04: must be 0
+		} view;
+	};
+	__le16 size;		// 0x08: The size of this entry
+	__le16 key_size;	// 0x0A: The size of File name length in bytes + 0x42
+	__le16 flags;		// 0x0C: Entry flags: NTFS_IE_XXX
+	__le16 res;		// 0x0E:
+
+	// Here any indexed attribute can be placed
+	// One of them is:
+	// struct ATTR_FILE_NAME AttrFileName;
+	//
+
+	// The last 8 bytes of this structure contains
+	// the VBN of subnode
+	// !!! Note !!!
+	// This field is presented only if (flags & NTFS_IE_HAS_SUBNODES)
+	// __le64 vbn;
+};
+
+static_assert(sizeof(struct NTFS_DE) == 0x10);
+
+static inline void de_set_vbn_le(struct NTFS_DE *e, __le64 vcn)
+{
+	__le64 *v = Add2Ptr(e, le16_to_cpu(e->size) - sizeof(__le64));
+
+	*v = vcn;
+}
+
+static inline void de_set_vbn(struct NTFS_DE *e, CLST vcn)
+{
+	__le64 *v = Add2Ptr(e, le16_to_cpu(e->size) - sizeof(__le64));
+
+	*v = cpu_to_le64(vcn);
+}
+
+static inline __le64 de_get_vbn_le(const struct NTFS_DE *e)
+{
+	return *(__le64 *)Add2Ptr(e, le16_to_cpu(e->size) - sizeof(__le64));
+}
+
+static inline CLST de_get_vbn(const struct NTFS_DE *e)
+{
+	__le64 *v = Add2Ptr(e, le16_to_cpu(e->size) - sizeof(__le64));
+
+	return le64_to_cpu(*v);
+}
+
+static inline struct NTFS_DE *de_get_next(const struct NTFS_DE *e)
+{
+	return Add2Ptr(e, le16_to_cpu(e->size));
+}
+
+static inline struct ATTR_FILE_NAME *de_get_fname(const struct NTFS_DE *e)
+{
+	return le16_to_cpu(e->key_size) >= SIZEOF_ATTRIBUTE_FILENAME ?
+		       Add2Ptr(e, sizeof(struct NTFS_DE)) :
+		       NULL;
+}
+
+static inline bool de_is_last(const struct NTFS_DE *e)
+{
+	return e->flags & NTFS_IE_LAST;
+}
+
+static inline bool de_has_vcn(const struct NTFS_DE *e)
+{
+	return e->flags & NTFS_IE_HAS_SUBNODES;
+}
+
+static inline bool de_has_vcn_ex(const struct NTFS_DE *e)
+{
+	return (e->flags & NTFS_IE_HAS_SUBNODES) &&
+	       (u64)(-1) != *((u64 *)Add2Ptr(e, le16_to_cpu(e->size) -
+							sizeof(__le64)));
+}
+
+#define MAX_BYTES_PER_NAME_ENTRY					       \
+	QuadAlign(sizeof(struct NTFS_DE) +				       \
+		  offsetof(struct ATTR_FILE_NAME, name) +		       \
+		  NTFS_NAME_LEN * sizeof(short))
+
+struct INDEX_HDR {
+	__le32 de_off;	// 0x00: The offset from the start of this structure
+			// to the first NTFS_DE
+	__le32 used;	// 0x04: The size of this structure plus all
+			// entries (quad-word aligned)
+	__le32 total;	// 0x08: The allocated size of for this structure plus all entries
+	u8 flags;	// 0x0C: 0x00 = Small directory, 0x01 = Large directory
+	u8 res[3];
+
+	//
+	// de_off + used <= total
+	//
+};
+
+static_assert(sizeof(struct INDEX_HDR) == 0x10);
+
+static inline struct NTFS_DE *hdr_first_de(const struct INDEX_HDR *hdr)
+{
+	u32 de_off = le32_to_cpu(hdr->de_off);
+	u32 used = le32_to_cpu(hdr->used);
+	struct NTFS_DE *e = Add2Ptr(hdr, de_off);
+	u16 esize;
+
+	if (de_off >= used || de_off >= le32_to_cpu(hdr->total))
+		return NULL;
+
+	esize = le16_to_cpu(e->size);
+	if (esize < sizeof(struct NTFS_DE) || de_off + esize > used)
+		return NULL;
+
+	return e;
+}
+
+static inline struct NTFS_DE *hdr_next_de(const struct INDEX_HDR *hdr,
+					  const struct NTFS_DE *e)
+{
+	size_t off = PtrOffset(hdr, e);
+	u32 used = le32_to_cpu(hdr->used);
+	u16 esize;
+
+	if (off >= used)
+		return NULL;
+
+	esize = le16_to_cpu(e->size);
+
+	if (esize < sizeof(struct NTFS_DE) ||
+	    off + esize + sizeof(struct NTFS_DE) > used)
+		return NULL;
+
+	return Add2Ptr(e, esize);
+}
+
+static inline bool hdr_has_subnode(const struct INDEX_HDR *hdr)
+{
+	return hdr->flags & 1;
+}
+
+struct INDEX_BUFFER {
+	struct NTFS_RECORD_HEADER rhdr; // 'INDX'
+	__le64 vbn; // 0x10: vcn if index >= cluster or vsn id index < cluster
+	struct INDEX_HDR ihdr; // 0x18:
+};
+
+static_assert(sizeof(struct INDEX_BUFFER) == 0x28);
+
+static inline bool ib_is_empty(const struct INDEX_BUFFER *ib)
+{
+	const struct NTFS_DE *first = hdr_first_de(&ib->ihdr);
+
+	return !first || de_is_last(first);
+}
+
+static inline bool ib_is_leaf(const struct INDEX_BUFFER *ib)
+{
+	return !(ib->ihdr.flags & 1);
+}
+
+/* Index root structure ( 0x90 ) */
+enum COLLATION_RULE {
+	NTFS_COLLATION_TYPE_BINARY	= cpu_to_le32(0),
+	// $I30
+	NTFS_COLLATION_TYPE_FILENAME	= cpu_to_le32(0x01),
+	// $SII of $Secure and $Q of Quota
+	NTFS_COLLATION_TYPE_UINT	= cpu_to_le32(0x10),
+	// $O of Quota
+	NTFS_COLLATION_TYPE_SID		= cpu_to_le32(0x11),
+	// $SDH of $Secure
+	NTFS_COLLATION_TYPE_SECURITY_HASH = cpu_to_le32(0x12),
+	// $O of ObjId and "$R" for Reparse
+	NTFS_COLLATION_TYPE_UINTS	= cpu_to_le32(0x13)
+};
+
+static_assert(sizeof(enum COLLATION_RULE) == 4);
+
+//
+struct INDEX_ROOT {
+	enum ATTR_TYPE type;	// 0x00: The type of attribute to index on
+	enum COLLATION_RULE rule; // 0x04: The rule
+	__le32 index_block_size;// 0x08: The size of index record
+	u8 index_block_clst;	// 0x0C: The number of clusters per index
+	u8 res[3];
+	struct INDEX_HDR ihdr;	// 0x10:
+};
+
+static_assert(sizeof(struct INDEX_ROOT) == 0x20);
+static_assert(offsetof(struct INDEX_ROOT, ihdr) == 0x10);
+
+#define VOLUME_FLAG_DIRTY	    cpu_to_le16(0x0001)
+#define VOLUME_FLAG_RESIZE_LOG_FILE cpu_to_le16(0x0002)
+
+struct VOLUME_INFO {
+	__le64 res1;	// 0x00
+	u8 major_ver;	// 0x08: NTFS major version number (before .)
+	u8 minor_ver;	// 0x09: NTFS minor version number (after .)
+	__le16 flags;	// 0x0A: Volume flags, see VOLUME_FLAG_XXX
+
+}; // sizeof=0xC
+
+#define SIZEOF_ATTRIBUTE_VOLUME_INFO 0xc
+
+#define NTFS_LABEL_MAX_LENGTH		(0x100 / sizeof(short))
+#define NTFS_ATTR_INDEXABLE		cpu_to_le32(0x00000002)
+#define NTFS_ATTR_DUPALLOWED		cpu_to_le32(0x00000004)
+#define NTFS_ATTR_MUST_BE_INDEXED	cpu_to_le32(0x00000010)
+#define NTFS_ATTR_MUST_BE_NAMED		cpu_to_le32(0x00000020)
+#define NTFS_ATTR_MUST_BE_RESIDENT	cpu_to_le32(0x00000040)
+#define NTFS_ATTR_LOG_ALWAYS		cpu_to_le32(0x00000080)
+
+/* $AttrDef file entry */
+struct ATTR_DEF_ENTRY {
+	__le16 name[0x40];	// 0x00: Attr name
+	enum ATTR_TYPE type;	// 0x80: struct ATTRIB type
+	__le32 res;		// 0x84:
+	enum COLLATION_RULE rule; // 0x88:
+	__le32 flags;		// 0x8C: NTFS_ATTR_XXX (see above)
+	__le64 min_sz;		// 0x90: Minimum attribute data size
+	__le64 max_sz;		// 0x98: Maximum attribute data size
+};
+
+static_assert(sizeof(struct ATTR_DEF_ENTRY) == 0xa0);
+
+/* Object ID (0x40) */
+struct OBJECT_ID {
+	struct GUID ObjId;	// 0x00: Unique Id assigned to file
+	struct GUID BirthVolumeId;// 0x10: Birth Volume Id is the Object Id of the Volume on
+				// which the Object Id was allocated. It never changes
+	struct GUID BirthObjectId; // 0x20: Birth Object Id is the first Object Id that was
+				// ever assigned to this MFT Record. I.e. If the Object Id
+				// is changed for some reason, this field will reflect the
+				// original value of the Object Id.
+	struct GUID DomainId;	// 0x30: Domain Id is currently unused but it is intended to be
+				// used in a network environment where the local machine is
+				// part of a Windows 2000 Domain. This may be used in a Windows
+				// 2000 Advanced Server managed domain.
+};
+
+static_assert(sizeof(struct OBJECT_ID) == 0x40);
+
+/* O Directory entry structure ( rule = 0x13 ) */
+struct NTFS_DE_O {
+	struct NTFS_DE de;
+	struct GUID ObjId;	// 0x10: Unique Id assigned to file
+	struct MFT_REF ref;	// 0x20: MFT record number with this file
+	struct GUID BirthVolumeId; // 0x28: Birth Volume Id is the Object Id of the Volume on
+				// which the Object Id was allocated. It never changes
+	struct GUID BirthObjectId; // 0x38: Birth Object Id is the first Object Id that was
+				// ever assigned to this MFT Record. I.e. If the Object Id
+				// is changed for some reason, this field will reflect the
+				// original value of the Object Id.
+				// This field is valid if data_size == 0x48
+	struct GUID BirthDomainId; // 0x48: Domain Id is currently unused but it is intended
+				// to be used in a network environment where the local
+				// machine is part of a Windows 2000 Domain. This may be
+				// used in a Windows 2000 Advanced Server managed domain.
+
+	// The last 8 bytes of this structure contains
+	// the VCN of subnode
+	// !!! Note !!!
+	// This field is presented only if (flags & 0x1)
+	// __le64	      SubnodesVCN;
+};
+
+static_assert(sizeof(struct NTFS_DE_O) == 0x58);
+
+#define NTFS_OBJECT_ENTRY_DATA_SIZE1					       \
+	0x38 // struct NTFS_DE_O.BirthDomainId is not used
+#define NTFS_OBJECT_ENTRY_DATA_SIZE2					       \
+	0x48 // struct NTFS_DE_O.BirthDomainId is used
+
+/* Q Directory entry structure ( rule = 0x11 ) */
+struct NTFS_DE_Q {
+	struct NTFS_DE de;
+	__le32 owner_id;	// 0x10: Unique Id assigned to file
+	__le32 Version;		// 0x14: 0x02
+	__le32 flags2;		// 0x18: Quota flags, see above
+	__le64 BytesUsed;	// 0x1C:
+	__le64 ChangeTime;	// 0x24:
+	__le64 WarningLimit;	// 0x28:
+	__le64 HardLimit;	// 0x34:
+	__le64 ExceededTime;	// 0x3C:
+
+	// SID is placed here
+
+	// The last 8 bytes of this structure contains
+	// the VCN of subnode
+	// !!! Note !!!
+	// This field is presented only if (flags & 0x1)
+	// __le64	      SubnodesVCN;
+
+}; // sizeof() = 0x44
+
+#define SIZEOF_NTFS_DE_Q 0x44
+
+#define SecurityDescriptorsBlockSize 0x40000 // 256K
+#define SecurityDescriptorMaxSize    0x20000 // 128K
+#define Log2OfSecurityDescriptorsBlockSize 18
+
+struct SECURITY_KEY {
+	__le32 hash; //  Hash value for descriptor
+	__le32 sec_id; //  Security Id (guaranteed unique)
+};
+
+/* Security descriptors (the content of $Secure::SDS data stream) */
+struct SECURITY_HDR {
+	struct SECURITY_KEY key;	// 0x00: Security Key
+	__le64 off;			// 0x08: Offset of this entry in the file
+	__le32 size;			// 0x10: Size of this entry, 8 byte aligned
+	//
+	// Security descriptor itself is placed here
+	// Total size is 16 byte aligned
+	//
+
+} __packed;
+
+#define SIZEOF_SECURITY_HDR 0x14
+
+/* SII Directory entry structure */
+struct NTFS_DE_SII {
+	struct NTFS_DE de;
+	__le32 sec_id;			// 0x10: Key: sizeof(security_id) = wKeySize
+	struct SECURITY_HDR sec_hdr;	// 0x14:
+
+} __packed;
+
+#define SIZEOF_SII_DIRENTRY 0x28
+
+/* SDH Directory entry structure */
+struct NTFS_DE_SDH {
+	struct NTFS_DE de;
+	struct SECURITY_KEY key;	// 0x10: Key
+	struct SECURITY_HDR sec_hdr;	// 0x18: Data
+	__le16 magic[2];		// 0x2C: 0x00490049 "I I"
+};
+
+#define SIZEOF_SDH_DIRENTRY 0x30
+
+struct REPARSE_KEY {
+	__le32 ReparseTag;		// 0x00: Reparse Tag
+	struct MFT_REF ref;		// 0x04: MFT record number with this file
+
+}; // sizeof() = 0x0C
+
+static_assert(offsetof(struct REPARSE_KEY, ref) == 0x04);
+#define SIZEOF_REPARSE_KEY 0x0C
+
+/* Reparse Directory entry structure */
+struct NTFS_DE_R {
+	struct NTFS_DE de;
+	struct REPARSE_KEY key;		// 0x10: Reparse Key
+
+	// The last 8 bytes of this structure contains
+	// the VCN of subnode
+	// !!! Note !!!
+	// This field is presented only if (flags & 0x1)
+	// __le64	      SubnodesVCN;
+
+}; // sizeof() = 0x1C
+
+#define SIZEOF_R_DIRENTRY 0x1C
+
+/* CompressReparseBuffer.WofVersion */
+#define WOF_CURRENT_VERSION		cpu_to_le32(1)
+/* CompressReparseBuffer.WofProvider */
+#define WOF_PROVIDER_WIM		cpu_to_le32(1)
+/* CompressReparseBuffer.WofProvider */
+#define WOF_PROVIDER_SYSTEM		cpu_to_le32(2)
+/* CompressReparseBuffer.ProviderVer */
+#define WOF_PROVIDER_CURRENT_VERSION	cpu_to_le32(1)
+
+#define WOF_COMPRESSION_XPRESS4K	0 // 4k
+#define WOF_COMPRESSION_LZX		1 // 32k
+#define WOF_COMPRESSION_XPRESS8K	2 // 8k
+#define WOF_COMPRESSION_XPRESS16K	3 // 16k
+
+/*
+ * ATTR_REPARSE (0xC0)
+ *
+ * The reparse struct GUID structure is used by all 3rd party layered drivers to
+ * store data in a reparse point. For non-Microsoft tags, The struct GUID field
+ * cannot be GUID_NULL.
+ * The constraints on reparse tags are defined below.
+ * Microsoft tags can also be used with this format of the reparse point buffer.
+ */
+struct REPARSE_POINT {
+	__le32 ReparseTag;	// 0x00:
+	__le16 ReparseDataLength;// 0x04:
+	__le16 Reserved;
+
+	struct GUID Guid;	// 0x08:
+
+	//
+	// Here GenericReparseBuffer is placed
+	//
+};
+
+static_assert(sizeof(struct REPARSE_POINT) == 0x18);
+
+//
+// Maximum allowed size of the reparse data.
+//
+#define MAXIMUM_REPARSE_DATA_BUFFER_SIZE	(16 * 1024)
+
+//
+// The value of the following constant needs to satisfy the following
+// conditions:
+//  (1) Be at least as large as the largest of the reserved tags.
+//  (2) Be strictly smaller than all the tags in use.
+//
+#define IO_REPARSE_TAG_RESERVED_RANGE		1
+
+//
+// The reparse tags are a ULONG. The 32 bits are laid out as follows:
+//
+//   3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
+//   1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
+//  +-+-+-+-+-----------------------+-------------------------------+
+//  |M|R|N|R|	  Reserved bits     |	    Reparse Tag Value	    |
+//  +-+-+-+-+-----------------------+-------------------------------+
+//
+// M is the Microsoft bit. When set to 1, it denotes a tag owned by Microsoft.
+//   All ISVs must use a tag with a 0 in this position.
+//   Note: If a Microsoft tag is used by non-Microsoft software, the
+//   behavior is not defined.
+//
+// R is reserved.  Must be zero for non-Microsoft tags.
+//
+// N is name surrogate. When set to 1, the file represents another named
+//   entity in the system.
+//
+// The M and N bits are OR-able.
+// The following macros check for the M and N bit values:
+//
+
+//
+// Macro to determine whether a reparse point tag corresponds to a tag
+// owned by Microsoft.
+//
+#define IsReparseTagMicrosoft(_tag)	(((_tag)&IO_REPARSE_TAG_MICROSOFT))
+
+//
+// Macro to determine whether a reparse point tag is a name surrogate
+//
+#define IsReparseTagNameSurrogate(_tag)	(((_tag)&IO_REPARSE_TAG_NAME_SURROGATE))
+
+//
+// The following constant represents the bits that are valid to use in
+// reparse tags.
+//
+#define IO_REPARSE_TAG_VALID_VALUES	0xF000FFFF
+
+//
+// Macro to determine whether a reparse tag is a valid tag.
+//
+#define IsReparseTagValid(_tag)						       \
+	(!((_tag) & ~IO_REPARSE_TAG_VALID_VALUES) &&			       \
+	 ((_tag) > IO_REPARSE_TAG_RESERVED_RANGE))
+
+//
+// Microsoft tags for reparse points.
+//
+
+enum IO_REPARSE_TAG {
+	IO_REPARSE_TAG_SYMBOLIC_LINK	= cpu_to_le32(0),
+	IO_REPARSE_TAG_NAME_SURROGATE	= cpu_to_le32(0x20000000),
+	IO_REPARSE_TAG_MICROSOFT	= cpu_to_le32(0x80000000),
+	IO_REPARSE_TAG_MOUNT_POINT	= cpu_to_le32(0xA0000003),
+	IO_REPARSE_TAG_SYMLINK		= cpu_to_le32(0xA000000C),
+	IO_REPARSE_TAG_HSM		= cpu_to_le32(0xC0000004),
+	IO_REPARSE_TAG_SIS		= cpu_to_le32(0x80000007),
+	IO_REPARSE_TAG_DEDUP		= cpu_to_le32(0x80000013),
+	IO_REPARSE_TAG_COMPRESS		= cpu_to_le32(0x80000017),
+
+	//
+	// The reparse tag 0x80000008 is reserved for Microsoft internal use
+	// (may be published in the future)
+	//
+
+	//
+	// Microsoft reparse tag reserved for DFS
+	//
+	IO_REPARSE_TAG_DFS		= cpu_to_le32(0x8000000A),
+
+	//
+	// Microsoft reparse tag reserved for the file system filter manager
+	//
+	IO_REPARSE_TAG_FILTER_MANAGER	= cpu_to_le32(0x8000000B),
+
+	//
+	// Non-Microsoft tags for reparse points
+	//
+
+	//
+	// Tag allocated to CONGRUENT, May 2000. Used by IFSTEST
+	//
+	IO_REPARSE_TAG_IFSTEST_CONGRUENT = cpu_to_le32(0x00000009),
+
+	//
+	// Tag allocated to ARKIVIO
+	//
+	IO_REPARSE_TAG_ARKIVIO		= cpu_to_le32(0x0000000C),
+
+	//
+	//  Tag allocated to SOLUTIONSOFT
+	//
+	IO_REPARSE_TAG_SOLUTIONSOFT	= cpu_to_le32(0x2000000D),
+
+	//
+	//  Tag allocated to COMMVAULT
+	//
+	IO_REPARSE_TAG_COMMVAULT	= cpu_to_le32(0x0000000E),
+
+	// OneDrive??
+	IO_REPARSE_TAG_CLOUD		= cpu_to_le32(0x9000001A),
+	IO_REPARSE_TAG_CLOUD_1		= cpu_to_le32(0x9000101A),
+	IO_REPARSE_TAG_CLOUD_2		= cpu_to_le32(0x9000201A),
+	IO_REPARSE_TAG_CLOUD_3		= cpu_to_le32(0x9000301A),
+	IO_REPARSE_TAG_CLOUD_4		= cpu_to_le32(0x9000401A),
+	IO_REPARSE_TAG_CLOUD_5		= cpu_to_le32(0x9000501A),
+	IO_REPARSE_TAG_CLOUD_6		= cpu_to_le32(0x9000601A),
+	IO_REPARSE_TAG_CLOUD_7		= cpu_to_le32(0x9000701A),
+	IO_REPARSE_TAG_CLOUD_8		= cpu_to_le32(0x9000801A),
+	IO_REPARSE_TAG_CLOUD_9		= cpu_to_le32(0x9000901A),
+	IO_REPARSE_TAG_CLOUD_A		= cpu_to_le32(0x9000A01A),
+	IO_REPARSE_TAG_CLOUD_B		= cpu_to_le32(0x9000B01A),
+	IO_REPARSE_TAG_CLOUD_C		= cpu_to_le32(0x9000C01A),
+	IO_REPARSE_TAG_CLOUD_D		= cpu_to_le32(0x9000D01A),
+	IO_REPARSE_TAG_CLOUD_E		= cpu_to_le32(0x9000E01A),
+	IO_REPARSE_TAG_CLOUD_F		= cpu_to_le32(0x9000F01A),
+
+};
+
+#define SYMLINK_FLAG_RELATIVE		1
+
+/* Microsoft reparse buffer. (see DDK for details) */
+struct REPARSE_DATA_BUFFER {
+	__le32 ReparseTag;		// 0x00:
+	__le16 ReparseDataLength;	// 0x04:
+	__le16 Reserved;
+
+	union {
+		// If ReparseTag == 0xA0000003 (IO_REPARSE_TAG_MOUNT_POINT)
+		struct {
+			__le16 SubstituteNameOffset; // 0x08
+			__le16 SubstituteNameLength; // 0x0A
+			__le16 PrintNameOffset;      // 0x0C
+			__le16 PrintNameLength;      // 0x0E
+			__le16 PathBuffer[1];	     // 0x10
+		} MountPointReparseBuffer;
+
+		// If ReparseTag == 0xA000000C (IO_REPARSE_TAG_SYMLINK)
+		// https://msdn.microsoft.com/en-us/library/cc232006.aspx
+		struct {
+			__le16 SubstituteNameOffset; // 0x08
+			__le16 SubstituteNameLength; // 0x0A
+			__le16 PrintNameOffset;      // 0x0C
+			__le16 PrintNameLength;      // 0x0E
+			// 0-absolute path 1- relative path, SYMLINK_FLAG_RELATIVE
+			__le32 Flags;		     // 0x10
+			__le16 PathBuffer[1];	     // 0x14
+		} SymbolicLinkReparseBuffer;
+
+		// If ReparseTag == 0x80000017U
+		struct {
+			__le32 WofVersion;  // 0x08 == 1
+			/* 1 - WIM backing provider ("WIMBoot"),
+			 * 2 - System compressed file provider
+			 */
+			__le32 WofProvider; // 0x0C
+			__le32 ProviderVer; // 0x10: == 1 WOF_FILE_PROVIDER_CURRENT_VERSION == 1
+			__le32 CompressionFormat; // 0x14: 0, 1, 2, 3. See WOF_COMPRESSION_XXX
+		} CompressReparseBuffer;
+
+		struct {
+			u8 DataBuffer[1];   // 0x08
+		} GenericReparseBuffer;
+	};
+};
+
+/* ATTR_EA_INFO (0xD0) */
+
+#define FILE_NEED_EA 0x80 // See ntifs.h
+/* FILE_NEED_EA, indicates that the file to which the EA belongs cannot be
+ * interpreted without understanding the associated extended attributes.
+ */
+struct EA_INFO {
+	__le16 size_pack;	// 0x00: Size of buffer to hold in packed form
+	__le16 count;		// 0x02: Count of EA's with FILE_NEED_EA bit set
+	__le32 size;		// 0x04: Size of buffer to hold in unpacked form
+};
+
+static_assert(sizeof(struct EA_INFO) == 8);
+
+/* ATTR_EA (0xE0) */
+struct EA_FULL {
+	__le32 size;		// 0x00: (not in packed)
+	u8 flags;		// 0x04
+	u8 name_len;		// 0x05
+	__le16 elength;		// 0x06
+	u8 name[1];		// 0x08
+};
+
+static_assert(offsetof(struct EA_FULL, name) == 8);
+
+#define MAX_EA_DATA_SIZE (256 * 1024)
+
+#define ACL_REVISION 2
+
+#define SE_SELF_RELATIVE 0x8000
+
+struct SECURITY_DESCRIPTOR_RELATIVE {
+	u8 Revision;
+	u8 Sbz1;
+	__le16 Control;
+	__le32 Owner;
+	__le32 Group;
+	__le32 Sacl;
+	__le32 Dacl;
+};
+static_assert(sizeof(struct SECURITY_DESCRIPTOR_RELATIVE) == 0x14);
+
+struct ACE_HEADER {
+	u8 AceType;
+	u8 AceFlags;
+	__le16 AceSize;
+};
+static_assert(sizeof(struct ACE_HEADER) == 4);
+
+struct ACL {
+	u8 AclRevision;
+	u8 Sbz1;
+	__le16 AclSize;
+	__le16 AceCount;
+	__le16 Sbz2;
+};
+static_assert(sizeof(struct ACL) == 8);
+
+struct SID {
+	u8 Revision;
+	u8 SubAuthorityCount;
+	u8 IdentifierAuthority[6];
+	__le32 SubAuthority[1];
+};
+static_assert(offsetof(struct SID, SubAuthority) == 8);
+
+// clang-format on
diff --git a/fs/ntfs3/ntfs_fs.h b/fs/ntfs3/ntfs_fs.h
new file mode 100644
index 000000000000..ce43b17a8f08
--- /dev/null
+++ b/fs/ntfs3/ntfs_fs.h
@@ -0,0 +1,988 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+// clang-format off
+#define MINUS_ONE_T			((size_t)(-1))
+/* Biggest MFT / smallest cluster */
+#define MAXIMUM_BYTES_PER_MFT		4096
+#define NTFS_BLOCKS_PER_MFT_RECORD	(MAXIMUM_BYTES_PER_MFT / 512)
+
+#define MAXIMUM_BYTES_PER_INDEX		4096
+#define NTFS_BLOCKS_PER_INODE		(MAXIMUM_BYTES_PER_INDEX / 512)
+
+/* ntfs specific error code when fixup failed*/
+#define E_NTFS_FIXUP			555
+
+/* sbi->flags */
+#define NTFS_FLAGS_NODISCARD		0x00000001
+#define NTFS_FLAGS_NEED_REPLAY		0x04000000
+
+/* ni->ni_flags */
+#define NI_FLAG_DIR			0x00000001
+#define NI_FLAG_RESIDENT		0x00000002
+#define NI_FLAG_UPDATE_PARENT		0x00000004
+/* Data attribute is external compressed (lzx/xpress) */
+#define NI_FLAG_COMPRESSED_MASK		0x00000f00
+/* Data attribute is deduplicated */
+#define NI_FLAG_DEDUPLICATED		0x00001000
+#define NI_FLAG_EA			0x00002000
+// clang-format on
+
+struct ntfs_inode;
+struct ntfs_sb_info;
+struct lznt;
+
+struct ntfs_mount_options {
+	struct nls_table *nls[2];
+
+	kuid_t fs_uid;
+	kgid_t fs_gid;
+	u16 fs_fmask_inv;
+	u16 fs_dmask_inv;
+
+	unsigned uid : 1, /* uid was set */
+		gid : 1, /* gid was set */
+		fmask : 1, /* fmask was set */
+		dmask : 1, /*dmask was set*/
+		sys_immutable : 1, /* set = system files are immutable */
+		discard : 1, /* issue discard requests on deletions */
+		sparse : 1, /*create sparse files*/
+		showmeta : 1, /*show meta files*/
+		nohidden : 1, /*do not show hidden files*/
+		force : 1, /*rw mount dirty volume*/
+		no_acs_rules : 1, /*exclude acs rules*/
+		prealloc : 1 /*preallocate space when file is growing*/
+		;
+};
+
+struct ntfs_run;
+
+/* TODO: use rb tree instead of array */
+struct runs_tree {
+	struct ntfs_run *runs_;
+	size_t count; // Currently used size a ntfs_run storage.
+	size_t allocated; // Currently allocated ntfs_run storage size.
+};
+
+struct ntfs_buffers {
+	/* Biggest MFT / smallest cluster = 4096 / 512 = 8 */
+	/* Biggest index / smallest cluster = 4096 / 512 = 8 */
+	struct buffer_head *bh[PAGE_SIZE >> SECTOR_SHIFT];
+	u32 bytes;
+	u32 nbufs;
+	u32 off;
+};
+
+enum ALLOCATE_OPT {
+	ALLOCATE_DEF = 0, // Allocate all clusters
+	ALLOCATE_MFT = 1, // Allocate for MFT
+};
+
+enum bitmap_mutex_classes {
+	BITMAP_MUTEX_CLUSTERS = 0,
+	BITMAP_MUTEX_MFT = 1,
+};
+
+struct wnd_bitmap {
+	struct super_block *sb;
+	struct rw_semaphore rw_lock;
+
+	struct runs_tree run;
+	size_t nbits;
+
+	u16 free_holder[8]; // holder for free_bits
+
+	size_t total_zeroes; // total number of free bits
+	u16 *free_bits; // free bits in each window
+	size_t nwnd;
+	u32 bits_last; // bits in last window
+
+	struct rb_root start_tree; // extents, sorted by 'start'
+	struct rb_root count_tree; // extents, sorted by 'count + start'
+	size_t count; // extents count
+
+	/*
+	 * -1 Tree is activated but not updated (too many fragments)
+	 * 0 - Tree is not activated
+	 * 1 - Tree is activated and updated
+	 */
+	int uptodated;
+	size_t extent_min; // Minimal extent used while building
+	size_t extent_max; // Upper estimate of biggest free block
+
+	/* Zone [bit, end) */
+	size_t zone_bit;
+	size_t zone_end;
+
+	bool set_tail; // not necessary in driver
+	bool inited;
+};
+
+typedef int (*NTFS_CMP_FUNC)(const void *key1, size_t len1, const void *key2,
+			     size_t len2, const void *param);
+
+enum index_mutex_classed {
+	INDEX_MUTEX_I30 = 0,
+	INDEX_MUTEX_SII = 1,
+	INDEX_MUTEX_SDH = 2,
+	INDEX_MUTEX_SO = 3,
+	INDEX_MUTEX_SQ = 4,
+	INDEX_MUTEX_SR = 5,
+	INDEX_MUTEX_TOTAL
+};
+
+/* This struct works with indexes */
+struct ntfs_index {
+	struct runs_tree bitmap_run;
+	struct runs_tree alloc_run;
+
+	/*TODO: remove 'cmp'*/
+	NTFS_CMP_FUNC cmp;
+
+	u8 index_bits; // log2(root->index_block_size)
+	u8 idx2vbn_bits; // log2(root->index_block_clst)
+	u8 vbn2vbo_bits; // index_block_size < cluster? 9 : cluster_bits
+	u8 changed; // set when tree is changed
+	u8 type; // index_mutex_classed
+};
+
+/* Set when $LogFile is replaying */
+#define NTFS_FLAGS_LOG_REPLAYING 0x00000008
+
+/* Set when we changed first MFT's which copy must be updated in $MftMirr */
+#define NTFS_FLAGS_MFTMIRR 0x00001000
+
+/* Minimum mft zone */
+#define NTFS_MIN_MFT_ZONE 100
+
+/* ntfs file system in-core superblock data */
+struct ntfs_sb_info {
+	struct super_block *sb;
+
+	u32 discard_granularity;
+	u64 discard_granularity_mask_inv; // ~(discard_granularity_mask_inv-1)
+
+	u32 cluster_size; // bytes per cluster
+	u32 cluster_mask; // == cluster_size - 1
+	u64 cluster_mask_inv; // ~(cluster_size - 1)
+	u32 block_mask; // sb->s_blocksize - 1
+	u32 blocks_per_cluster; // cluster_size / sb->s_blocksize
+
+	u32 record_size;
+	u32 sector_size;
+	u32 index_size;
+
+	u8 sector_bits;
+	u8 cluster_bits;
+	u8 record_bits;
+
+	u64 maxbytes; // Maximum size for normal files
+	u64 maxbytes_sparse; // Maximum size for sparse file
+
+	u32 flags; // See NTFS_FLAGS_XXX
+
+	CLST bad_clusters; // The count of marked bad clusters
+
+	u16 max_bytes_per_attr; // maximum attribute size in record
+	u16 attr_size_tr; // attribute size threshold (320 bytes)
+
+	/* Records in $Extend */
+	CLST objid_no;
+	CLST quota_no;
+	CLST reparse_no;
+	CLST usn_jrnl_no;
+
+	struct ATTR_DEF_ENTRY *def_table; // attribute definition table
+	u32 def_entries;
+
+	struct MFT_REC *new_rec;
+
+	u16 *upcase;
+
+	struct {
+		u64 lbo, lbo2;
+		struct ntfs_inode *ni;
+		struct wnd_bitmap bitmap; // $MFT::Bitmap
+		ulong reserved_bitmap;
+		size_t next_free; // The next record to allocate from
+		size_t used;
+		u32 recs_mirr; // Number of records MFTMirr
+		u8 next_reserved;
+		u8 reserved_bitmap_inited;
+	} mft;
+
+	struct {
+		struct wnd_bitmap bitmap; // $Bitmap::Data
+		CLST next_free_lcn;
+	} used;
+
+	struct {
+		u64 size; // in bytes
+		u64 blocks; // in blocks
+		u64 ser_num;
+		struct ntfs_inode *ni;
+		__le16 flags; // see VOLUME_FLAG_XXX
+		u8 major_ver;
+		u8 minor_ver;
+		char label[65];
+		bool real_dirty; /* real fs state*/
+	} volume;
+
+	struct {
+		struct ntfs_index index_sii;
+		struct ntfs_index index_sdh;
+		struct ntfs_inode *ni;
+		u32 next_id;
+		u64 next_off;
+
+		__le32 def_security_id;
+	} security;
+
+	struct {
+		struct ntfs_index index_r;
+		struct ntfs_inode *ni;
+		u64 max_size; // 16K
+	} reparse;
+
+	struct {
+		struct ntfs_index index_o;
+		struct ntfs_inode *ni;
+	} objid;
+
+	struct {
+		/*protect 'frame_unc' and 'ctx'*/
+		spinlock_t lock;
+		struct lznt *ctx;
+	} compress;
+
+	struct ntfs_mount_options options;
+	struct ratelimit_state msg_ratelimit;
+};
+
+struct mft_inode {
+	struct rb_node node;
+	struct ntfs_sb_info *sbi;
+
+	CLST rno;
+	struct MFT_REC *mrec;
+	struct ntfs_buffers nb;
+
+	bool dirty;
+};
+
+/* ntfs file system inode data memory */
+struct ntfs_inode {
+	struct mft_inode mi; // base record
+
+	loff_t i_valid; /* valid size */
+	struct timespec64 i_crtime;
+
+	struct mutex ni_lock;
+
+	/* file attributes from std */
+	enum FILE_ATTRIBUTE std_fa;
+	__le32 std_security_id;
+
+	// subrecords tree
+	struct rb_root mi_tree;
+
+	union {
+		struct ntfs_index dir;
+		struct {
+			struct rw_semaphore run_lock;
+			struct runs_tree run;
+		} file;
+	};
+
+	struct {
+		struct runs_tree run;
+		struct ATTR_LIST_ENTRY *le; // 1K aligned memory
+		size_t size;
+		bool dirty;
+	} attr_list;
+
+	size_t ni_flags; // NI_FLAG_XXX
+
+	struct inode vfs_inode;
+};
+
+struct indx_node {
+	struct ntfs_buffers nb;
+	struct INDEX_BUFFER *index;
+};
+
+struct ntfs_fnd {
+	int level;
+	struct indx_node *nodes[20];
+	struct NTFS_DE *de[20];
+	struct NTFS_DE *root_de;
+};
+
+enum REPARSE_SIGN {
+	REPARSE_NONE = 0,
+	REPARSE_COMPRESSED = 1,
+	REPARSE_DEDUPLICATED = 2,
+	REPARSE_LINK = 3
+};
+
+/* functions from attrib.c*/
+int attr_load_runs(struct ATTRIB *attr, struct ntfs_inode *ni,
+		   struct runs_tree *run, const CLST *vcn);
+int attr_allocate_clusters(struct ntfs_sb_info *sbi, struct runs_tree *run,
+			   CLST vcn, CLST lcn, CLST len, CLST *pre_alloc,
+			   enum ALLOCATE_OPT opt, CLST *alen, const size_t fr,
+			   CLST *new_lcn);
+int attr_set_size(struct ntfs_inode *ni, enum ATTR_TYPE type,
+		  const __le16 *name, u8 name_len, struct runs_tree *run,
+		  u64 new_size, const u64 *new_valid, bool keep_prealloc,
+		  struct ATTRIB **ret);
+int attr_data_get_block(struct ntfs_inode *ni, CLST vcn, CLST clen, CLST *lcn,
+			CLST *len, bool *new);
+int attr_load_runs_vcn(struct ntfs_inode *ni, enum ATTR_TYPE type,
+		       const __le16 *name, u8 name_len, struct runs_tree *run,
+		       CLST vcn);
+int attr_is_frame_compressed(struct ntfs_inode *ni, struct ATTRIB *attr,
+			     CLST frame, CLST *clst_data);
+int attr_allocate_frame(struct ntfs_inode *ni, CLST frame, size_t compr_size,
+			u64 new_valid);
+
+/* functions from attrlist.c*/
+void al_destroy(struct ntfs_inode *ni);
+bool al_verify(struct ntfs_inode *ni);
+int ntfs_load_attr_list(struct ntfs_inode *ni, struct ATTRIB *attr);
+struct ATTR_LIST_ENTRY *al_enumerate(struct ntfs_inode *ni,
+				     struct ATTR_LIST_ENTRY *le);
+struct ATTR_LIST_ENTRY *al_find_le(struct ntfs_inode *ni,
+				   struct ATTR_LIST_ENTRY *le,
+				   const struct ATTRIB *attr);
+struct ATTR_LIST_ENTRY *al_find_ex(struct ntfs_inode *ni,
+				   struct ATTR_LIST_ENTRY *le,
+				   enum ATTR_TYPE type, const __le16 *name,
+				   u8 name_len, const CLST *vcn);
+int al_add_le(struct ntfs_inode *ni, enum ATTR_TYPE type, const __le16 *name,
+	      u8 name_len, CLST svcn, __le16 id, const struct MFT_REF *ref,
+	      struct ATTR_LIST_ENTRY **new_le);
+bool al_remove_le(struct ntfs_inode *ni, struct ATTR_LIST_ENTRY *le);
+bool al_delete_le(struct ntfs_inode *ni, enum ATTR_TYPE type, CLST vcn,
+		  const __le16 *name, size_t name_len,
+		  const struct MFT_REF *ref);
+int al_update(struct ntfs_inode *ni);
+static inline size_t al_aligned(size_t size)
+{
+	return (size + 1023) & ~(size_t)1023;
+}
+
+/* globals from bitfunc.c */
+bool are_bits_clear(const ulong *map, size_t bit, size_t nbits);
+bool are_bits_set(const ulong *map, size_t bit, size_t nbits);
+size_t get_set_bits_ex(const ulong *map, size_t bit, size_t nbits);
+
+/* globals from dir.c */
+int ntfs_utf16_to_nls(struct ntfs_sb_info *sbi, const struct le_str *uni,
+		      u8 *buf, int buf_len);
+int ntfs_nls_to_utf16(struct ntfs_sb_info *sbi, const u8 *name, u32 name_len,
+		      struct cpu_str *uni, u32 max_ulen,
+		      enum utf16_endian endian);
+struct inode *dir_search_u(struct inode *dir, const struct cpu_str *uni,
+			   struct ntfs_fnd *fnd);
+bool dir_is_empty(struct inode *dir);
+extern const struct file_operations ntfs_dir_operations;
+
+/* globals from file.c*/
+int ntfs_getattr(const struct path *path, struct kstat *stat, u32 request_mask,
+		 u32 flags);
+void ntfs_sparse_cluster(struct inode *inode, struct page *page0, loff_t vbo,
+			 u32 bytes);
+int ntfs_file_fsync(struct file *filp, loff_t start, loff_t end, int datasync);
+void ntfs_truncate_blocks(struct inode *inode, loff_t offset);
+int ntfs_setattr(struct dentry *dentry, struct iattr *attr);
+int ntfs_file_open(struct inode *inode, struct file *file);
+int ntfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
+		__u64 start, __u64 len);
+extern const struct inode_operations ntfs_special_inode_operations;
+extern const struct inode_operations ntfs_file_inode_operations;
+extern const struct file_operations ntfs_file_operations;
+
+/* globals from frecord.c */
+void ni_remove_mi(struct ntfs_inode *ni, struct mft_inode *mi);
+struct ATTR_STD_INFO *ni_std(struct ntfs_inode *ni);
+struct ATTR_STD_INFO5 *ni_std5(struct ntfs_inode *ni);
+void ni_clear(struct ntfs_inode *ni);
+int ni_load_mi_ex(struct ntfs_inode *ni, CLST rno, struct mft_inode **mi);
+int ni_load_mi(struct ntfs_inode *ni, struct ATTR_LIST_ENTRY *le,
+	       struct mft_inode **mi);
+struct ATTRIB *ni_find_attr(struct ntfs_inode *ni, struct ATTRIB *attr,
+			    struct ATTR_LIST_ENTRY **entry_o,
+			    enum ATTR_TYPE type, const __le16 *name,
+			    u8 name_len, const CLST *vcn,
+			    struct mft_inode **mi);
+struct ATTRIB *ni_enum_attr_ex(struct ntfs_inode *ni, struct ATTRIB *attr,
+			       struct ATTR_LIST_ENTRY **le);
+struct ATTRIB *ni_load_attr(struct ntfs_inode *ni, enum ATTR_TYPE type,
+			    const __le16 *name, u8 name_len, CLST vcn,
+			    struct mft_inode **pmi);
+int ni_load_all_mi(struct ntfs_inode *ni);
+bool ni_add_subrecord(struct ntfs_inode *ni, CLST rno, struct mft_inode **mi);
+int ni_remove_attr(struct ntfs_inode *ni, enum ATTR_TYPE type,
+		   const __le16 *name, size_t name_len, bool base_only,
+		   const __le16 *id);
+int ni_create_attr_list(struct ntfs_inode *ni);
+int ni_expand_list(struct ntfs_inode *ni);
+int ni_insert_nonresident(struct ntfs_inode *ni, enum ATTR_TYPE type,
+			  const __le16 *name, u8 name_len,
+			  const struct runs_tree *run, CLST svcn, CLST len,
+			  __le16 flags, struct ATTRIB **new_attr,
+			  struct mft_inode **mi);
+int ni_insert_resident(struct ntfs_inode *ni, u32 data_size,
+		       enum ATTR_TYPE type, const __le16 *name, u8 name_len,
+		       struct ATTRIB **new_attr, struct mft_inode **mi);
+int ni_remove_attr_le(struct ntfs_inode *ni, struct ATTRIB *attr,
+		      struct ATTR_LIST_ENTRY *le);
+int ni_delete_all(struct ntfs_inode *ni);
+struct ATTR_FILE_NAME *ni_fname_name(struct ntfs_inode *ni,
+				     const struct cpu_str *uni,
+				     const struct MFT_REF *home,
+				     struct ATTR_LIST_ENTRY **entry);
+struct ATTR_FILE_NAME *ni_fname_type(struct ntfs_inode *ni, u8 name_type,
+				     struct ATTR_LIST_ENTRY **entry);
+u16 ni_fnames_count(struct ntfs_inode *ni);
+int ni_new_attr_flags(struct ntfs_inode *ni, enum FILE_ATTRIBUTE new_fa);
+enum REPARSE_SIGN ni_parse_reparse(struct ntfs_inode *ni, struct ATTRIB *attr,
+				   void *buffer);
+int ni_write_inode(struct inode *inode, int sync, const char *hint);
+#define _ni_write_inode(i, w) ni_write_inode(i, w, __func__)
+int ni_fiemap(struct ntfs_inode *ni, struct fiemap_extent_info *fieinfo,
+	      __u64 vbo, __u64 len);
+int ni_readpage_cmpr(struct ntfs_inode *ni, struct page *page);
+int ni_read_frame(struct ntfs_inode *ni, u64 frame_vbo, struct page **pages,
+		  u32 pages_per_frame);
+int ni_write_frame(struct ntfs_inode *ni, struct page **pages,
+		   u32 pages_per_frame);
+
+/* globals from fslog.c */
+int log_replay(struct ntfs_inode *ni);
+
+/* globals from fsntfs.c */
+bool ntfs_fix_pre_write(struct NTFS_RECORD_HEADER *rhdr, size_t bytes);
+int ntfs_fix_post_read(struct NTFS_RECORD_HEADER *rhdr, size_t bytes,
+		       bool simple);
+int ntfs_extend_init(struct ntfs_sb_info *sbi);
+int ntfs_loadlog_and_replay(struct ntfs_inode *ni, struct ntfs_sb_info *sbi);
+const struct ATTR_DEF_ENTRY *ntfs_query_def(struct ntfs_sb_info *sbi,
+					    enum ATTR_TYPE Type);
+int ntfs_look_for_free_space(struct ntfs_sb_info *sbi, CLST lcn, CLST len,
+			     CLST *new_lcn, CLST *new_len,
+			     enum ALLOCATE_OPT opt);
+int ntfs_look_free_mft(struct ntfs_sb_info *sbi, CLST *rno, bool mft,
+		       struct ntfs_inode *ni, struct mft_inode **mi);
+void ntfs_mark_rec_free(struct ntfs_sb_info *sbi, CLST nRecord);
+int ntfs_clear_mft_tail(struct ntfs_sb_info *sbi, size_t from, size_t to);
+int ntfs_refresh_zone(struct ntfs_sb_info *sbi);
+int ntfs_update_mftmirr(struct ntfs_sb_info *sbi, int wait);
+enum NTFS_DIRTY_FLAGS {
+	NTFS_DIRTY_CLEAR = 0,
+	NTFS_DIRTY_DIRTY = 1,
+	NTFS_DIRTY_ERROR = 2,
+};
+int ntfs_set_state(struct ntfs_sb_info *sbi, enum NTFS_DIRTY_FLAGS dirty);
+int ntfs_sb_read(struct super_block *sb, u64 lbo, size_t bytes, void *buffer);
+int ntfs_sb_write(struct super_block *sb, u64 lbo, size_t bytes,
+		  const void *buffer, int wait);
+int ntfs_sb_write_run(struct ntfs_sb_info *sbi, const struct runs_tree *run,
+		      u64 vbo, const void *buf, size_t bytes);
+struct buffer_head *ntfs_bread_run(struct ntfs_sb_info *sbi,
+				   const struct runs_tree *run, u64 vbo);
+int ntfs_read_run_nb(struct ntfs_sb_info *sbi, const struct runs_tree *run,
+		     u64 vbo, void *buf, u32 bytes, struct ntfs_buffers *nb);
+int ntfs_read_bh(struct ntfs_sb_info *sbi, const struct runs_tree *run, u64 vbo,
+		 struct NTFS_RECORD_HEADER *rhdr, u32 bytes,
+		 struct ntfs_buffers *nb);
+int ntfs_get_bh(struct ntfs_sb_info *sbi, const struct runs_tree *run, u64 vbo,
+		u32 bytes, struct ntfs_buffers *nb);
+int ntfs_write_bh(struct ntfs_sb_info *sbi, struct NTFS_RECORD_HEADER *rhdr,
+		  struct ntfs_buffers *nb, int sync);
+int ntfs_bio_pages(struct ntfs_sb_info *sbi, const struct runs_tree *run,
+		   struct page **pages, u32 nr_pages, u64 vbo, u32 bytes,
+		   u32 op);
+int ntfs_bio_fill_1(struct ntfs_sb_info *sbi, const struct runs_tree *run);
+int ntfs_vbo_to_lbo(struct ntfs_sb_info *sbi, const struct runs_tree *run,
+		    u64 vbo, u64 *lbo, u64 *bytes);
+struct ntfs_inode *ntfs_new_inode(struct ntfs_sb_info *sbi, CLST nRec,
+				  bool dir);
+extern const u8 s_default_security[0x50];
+bool is_sd_valid(const struct SECURITY_DESCRIPTOR_RELATIVE *sd, u32 len);
+int ntfs_security_init(struct ntfs_sb_info *sbi);
+int ntfs_get_security_by_id(struct ntfs_sb_info *sbi, __le32 security_id,
+			    struct SECURITY_DESCRIPTOR_RELATIVE **sd,
+			    size_t *size);
+int ntfs_insert_security(struct ntfs_sb_info *sbi,
+			 const struct SECURITY_DESCRIPTOR_RELATIVE *sd,
+			 u32 size, __le32 *security_id, bool *inserted);
+int ntfs_reparse_init(struct ntfs_sb_info *sbi);
+int ntfs_objid_init(struct ntfs_sb_info *sbi);
+int ntfs_objid_remove(struct ntfs_sb_info *sbi, struct GUID *guid);
+int ntfs_insert_reparse(struct ntfs_sb_info *sbi, __le32 rtag,
+			const struct MFT_REF *ref);
+int ntfs_remove_reparse(struct ntfs_sb_info *sbi, __le32 rtag,
+			const struct MFT_REF *ref);
+void mark_as_free_ex(struct ntfs_sb_info *sbi, CLST lcn, CLST len, bool trim);
+int run_deallocate(struct ntfs_sb_info *sbi, struct runs_tree *run, bool trim);
+
+/* globals from index.c */
+int indx_used_bit(struct ntfs_index *indx, struct ntfs_inode *ni, size_t *bit);
+void fnd_clear(struct ntfs_fnd *fnd);
+struct ntfs_fnd *fnd_get(struct ntfs_index *indx);
+void fnd_put(struct ntfs_fnd *fnd);
+void indx_clear(struct ntfs_index *idx);
+int indx_init(struct ntfs_index *indx, struct ntfs_sb_info *sbi,
+	      const struct ATTRIB *attr, enum index_mutex_classed type);
+struct INDEX_ROOT *indx_get_root(struct ntfs_index *indx, struct ntfs_inode *ni,
+				 struct ATTRIB **attr, struct mft_inode **mi);
+int indx_read(struct ntfs_index *idx, struct ntfs_inode *ni, CLST vbn,
+	      struct indx_node **node);
+int indx_find(struct ntfs_index *indx, struct ntfs_inode *dir,
+	      const struct INDEX_ROOT *root, const void *Key, size_t KeyLen,
+	      const void *param, int *diff, struct NTFS_DE **entry,
+	      struct ntfs_fnd *fnd);
+int indx_find_sort(struct ntfs_index *indx, struct ntfs_inode *ni,
+		   const struct INDEX_ROOT *root, struct NTFS_DE **entry,
+		   struct ntfs_fnd *fnd);
+int indx_find_raw(struct ntfs_index *indx, struct ntfs_inode *ni,
+		  const struct INDEX_ROOT *root, struct NTFS_DE **entry,
+		  size_t *off, struct ntfs_fnd *fnd);
+int indx_insert_entry(struct ntfs_index *indx, struct ntfs_inode *ni,
+		      const struct NTFS_DE *new_de, const void *param,
+		      struct ntfs_fnd *fnd);
+int indx_delete_entry(struct ntfs_index *indx, struct ntfs_inode *ni,
+		      const void *key, u32 key_len, const void *param);
+int indx_update_dup(struct ntfs_inode *ni, struct ntfs_sb_info *sbi,
+		    const struct ATTR_FILE_NAME *fname,
+		    const struct NTFS_DUP_INFO *dup, int sync);
+
+/* globals from inode.c */
+struct inode *ntfs_iget5(struct super_block *sb, const struct MFT_REF *ref,
+			 const struct cpu_str *name);
+int ntfs_set_size(struct inode *inode, u64 new_size);
+int reset_log_file(struct inode *inode);
+int ntfs_get_block(struct inode *inode, sector_t vbn,
+		   struct buffer_head *bh_result, int create);
+int ntfs_write_inode(struct inode *inode, struct writeback_control *wbc);
+int ntfs_sync_inode(struct inode *inode);
+int ntfs_flush_inodes(struct super_block *sb, struct inode *i1,
+		      struct inode *i2);
+int inode_write_data(struct inode *inode, const void *data, size_t bytes);
+int ntfs_create_inode(struct inode *dir, struct dentry *dentry,
+		      const struct cpu_str *uni, umode_t mode, dev_t dev,
+		      const char *symname, u32 size, int excl,
+		      struct ntfs_fnd *fnd, struct inode **new_inode);
+int ntfs_link_inode(struct inode *inode, struct dentry *dentry);
+int ntfs_unlink_inode(struct inode *dir, const struct dentry *dentry);
+void ntfs_evict_inode(struct inode *inode);
+int ntfs_readpage(struct file *file, struct page *page);
+extern const struct inode_operations ntfs_link_inode_operations;
+extern const struct address_space_operations ntfs_aops;
+extern const struct address_space_operations ntfs_aops_cmpr;
+
+/* globals from name_i.c*/
+int fill_name_de(struct ntfs_sb_info *sbi, void *buf, const struct qstr *name,
+		 const struct cpu_str *uni);
+struct dentry *ntfs_get_parent(struct dentry *child);
+
+extern const struct inode_operations ntfs_dir_inode_operations;
+
+/* globals from record.c */
+int mi_get(struct ntfs_sb_info *sbi, CLST rno, struct mft_inode **mi);
+void mi_put(struct mft_inode *mi);
+int mi_init(struct mft_inode *mi, struct ntfs_sb_info *sbi, CLST rno);
+int mi_read(struct mft_inode *mi, bool is_mft);
+struct ATTRIB *mi_enum_attr(struct mft_inode *mi, struct ATTRIB *attr);
+// TODO: id?
+struct ATTRIB *mi_find_attr(struct mft_inode *mi, struct ATTRIB *attr,
+			    enum ATTR_TYPE type, const __le16 *name,
+			    size_t name_len, const __le16 *id);
+static inline struct ATTRIB *rec_find_attr_le(struct mft_inode *rec,
+					      struct ATTR_LIST_ENTRY *le)
+{
+	return mi_find_attr(rec, NULL, le->type, le_name(le), le->name_len,
+			    &le->id);
+}
+int mi_write(struct mft_inode *mi, int wait);
+int mi_format_new(struct mft_inode *mi, struct ntfs_sb_info *sbi, CLST rno,
+		  __le16 flags, bool is_mft);
+void mi_mark_free(struct mft_inode *mi);
+struct ATTRIB *mi_insert_attr(struct mft_inode *mi, enum ATTR_TYPE type,
+			      const __le16 *name, u8 name_len, u32 asize,
+			      u16 name_off);
+
+bool mi_remove_attr(struct mft_inode *mi, struct ATTRIB *attr);
+bool mi_resize_attr(struct mft_inode *mi, struct ATTRIB *attr, int bytes);
+int mi_pack_runs(struct mft_inode *mi, struct ATTRIB *attr,
+		 struct runs_tree *run, CLST len);
+static inline bool mi_is_ref(const struct mft_inode *mi,
+			     const struct MFT_REF *ref)
+{
+	if (le32_to_cpu(ref->low) != mi->rno)
+		return false;
+	if (ref->seq != mi->mrec->seq)
+		return false;
+
+#ifdef NTFS3_64BIT_CLUSTER
+	return le16_to_cpu(ref->high) == (mi->rno >> 32);
+#else
+	return !ref->high;
+#endif
+}
+
+/* globals from run.c */
+bool run_lookup_entry(const struct runs_tree *run, CLST vcn, CLST *lcn,
+		      CLST *len, size_t *index);
+void run_truncate(struct runs_tree *run, CLST vcn);
+void run_truncate_head(struct runs_tree *run, CLST vcn);
+void run_truncate_around(struct runs_tree *run, CLST vcn);
+bool run_lookup(const struct runs_tree *run, CLST vcn, size_t *Index);
+bool run_add_entry(struct runs_tree *run, CLST vcn, CLST lcn, CLST len);
+bool run_get_entry(const struct runs_tree *run, size_t index, CLST *vcn,
+		   CLST *lcn, CLST *len);
+bool run_is_mapped_full(const struct runs_tree *run, CLST svcn, CLST evcn);
+
+int run_pack(const struct runs_tree *run, CLST svcn, CLST len, u8 *run_buf,
+	     u32 run_buf_size, CLST *packed_vcns);
+int run_unpack(struct runs_tree *run, struct ntfs_sb_info *sbi, CLST ino,
+	       CLST svcn, CLST evcn, CLST vcn, const u8 *run_buf,
+	       u32 run_buf_size);
+
+#ifdef NTFS3_CHECK_FREE_CLST
+int run_unpack_ex(struct runs_tree *run, struct ntfs_sb_info *sbi, CLST ino,
+		  CLST svcn, CLST evcn, CLST vcn, const u8 *run_buf,
+		  u32 run_buf_size);
+#else
+#define run_unpack_ex run_unpack
+#endif
+int run_get_highest_vcn(CLST vcn, const u8 *run_buf, u64 *highest_vcn);
+
+/* globals from super.c */
+void *ntfs_set_shared(void *ptr, u32 bytes);
+void *ntfs_put_shared(void *ptr);
+void ntfs_unmap_meta(struct super_block *sb, CLST lcn, CLST len);
+int ntfs_discard(struct ntfs_sb_info *sbi, CLST Lcn, CLST Len);
+
+/* globals from ubitmap.c*/
+void wnd_close(struct wnd_bitmap *wnd);
+static inline size_t wnd_zeroes(const struct wnd_bitmap *wnd)
+{
+	return wnd->total_zeroes;
+}
+void wnd_trace(struct wnd_bitmap *wnd);
+void wnd_trace_tree(struct wnd_bitmap *wnd, u32 nExtents, const char *Hint);
+int wnd_init(struct wnd_bitmap *wnd, struct super_block *sb, size_t nBits);
+int wnd_set_free(struct wnd_bitmap *wnd, size_t FirstBit, size_t Bits);
+int wnd_set_used(struct wnd_bitmap *wnd, size_t FirstBit, size_t Bits);
+bool wnd_is_free(struct wnd_bitmap *wnd, size_t FirstBit, size_t Bits);
+bool wnd_is_used(struct wnd_bitmap *wnd, size_t FirstBit, size_t Bits);
+
+/* Possible values for 'flags' 'wnd_find' */
+#define BITMAP_FIND_MARK_AS_USED 0x01
+#define BITMAP_FIND_FULL 0x02
+size_t wnd_find(struct wnd_bitmap *wnd, size_t to_alloc, size_t hint,
+		size_t flags, size_t *allocated);
+int wnd_extend(struct wnd_bitmap *wnd, size_t new_bits);
+void wnd_zone_set(struct wnd_bitmap *wnd, size_t Lcn, size_t Len);
+int ntfs_trim_fs(struct ntfs_sb_info *sbi, struct fstrim_range *range);
+
+/* globals from upcase.c */
+int ntfs_cmp_names(const __le16 *s1, size_t l1, const __le16 *s2, size_t l2,
+		   const u16 *upcase);
+int ntfs_cmp_names_cpu(const struct cpu_str *uni1, const struct le_str *uni2,
+		       const u16 *upcase);
+
+/* globals from xattr.c */
+struct posix_acl *ntfs_get_acl(struct inode *inode, int type);
+int ntfs_set_acl(struct inode *inode, struct posix_acl *acl, int type);
+int ntfs_acl_chmod(struct inode *inode);
+int ntfs_permission(struct inode *inode, int mask);
+ssize_t ntfs_listxattr(struct dentry *dentry, char *buffer, size_t size);
+int ntfs_init_acl(struct inode *inode, struct inode *dir);
+extern const struct xattr_handler *ntfs_xattr_handlers[];
+
+/* globals from lznt.c */
+struct lznt *get_compression_ctx(bool std);
+size_t compress_lznt(const void *uncompressed, size_t uncompressed_size,
+		     void *compressed, size_t compressed_size,
+		     struct lznt *ctx);
+ssize_t decompress_lznt(const void *compressed, size_t compressed_size,
+			void *uncompressed, size_t uncompressed_size);
+
+static inline bool is_ntfs3(struct ntfs_sb_info *sbi)
+{
+	return sbi->volume.major_ver >= 3;
+}
+
+/*(sb->s_flags & SB_ACTIVE)*/
+static inline bool is_mounted(struct ntfs_sb_info *sbi)
+{
+	return !!sbi->sb->s_root;
+}
+
+static inline bool ntfs_is_meta_file(struct ntfs_sb_info *sbi, CLST rno)
+{
+	return rno < MFT_REC_FREE || rno == sbi->objid_no ||
+	       rno == sbi->quota_no || rno == sbi->reparse_no ||
+	       rno == sbi->usn_jrnl_no;
+}
+
+static inline void ntfs_unmap_page(struct page *page)
+{
+	kunmap(page);
+	put_page(page);
+}
+
+static inline struct page *ntfs_map_page(struct address_space *mapping,
+					 unsigned long index)
+{
+	struct page *page = read_mapping_page(mapping, index, NULL);
+
+	if (!IS_ERR(page)) {
+		kmap(page);
+		if (!PageError(page))
+			return page;
+		ntfs_unmap_page(page);
+		return ERR_PTR(-EIO);
+	}
+	return page;
+}
+
+static inline size_t wnd_zone_bit(const struct wnd_bitmap *wnd)
+{
+	return wnd->zone_bit;
+}
+
+static inline size_t wnd_zone_len(const struct wnd_bitmap *wnd)
+{
+	return wnd->zone_end - wnd->zone_bit;
+}
+
+static inline void run_init(struct runs_tree *run)
+{
+	run->runs_ = NULL;
+	run->count = 0;
+	run->allocated = 0;
+}
+
+static inline struct runs_tree *run_alloc(void)
+{
+	return ntfs_alloc(sizeof(struct runs_tree), 1);
+}
+
+static inline void run_close(struct runs_tree *run)
+{
+	ntfs_free(run->runs_);
+	memset(run, 0, sizeof(*run));
+}
+
+static inline void run_free(struct runs_tree *run)
+{
+	if (run) {
+		ntfs_free(run->runs_);
+		ntfs_free(run);
+	}
+}
+
+static inline bool run_is_empty(struct runs_tree *run)
+{
+	return !run->count;
+}
+
+/* NTFS uses quad aligned bitmaps */
+static inline size_t bitmap_size(size_t bits)
+{
+	return QuadAlign((bits + 7) >> 3);
+}
+
+#define _100ns2seconds 10000000
+#define SecondsToStartOf1970 0x00000002B6109100
+
+#define NTFS_TIME_GRAN 100
+
+/*
+ * kernel2nt
+ *
+ * converts in-memory kernel timestamp into nt time
+ */
+static inline __le64 kernel2nt(const struct timespec64 *ts)
+{
+	// 10^7 units of 100 nanoseconds one second
+	return cpu_to_le64(_100ns2seconds *
+				   (ts->tv_sec + SecondsToStartOf1970) +
+			   ts->tv_nsec / NTFS_TIME_GRAN);
+}
+
+/*
+ * nt2kernel
+ *
+ * converts on-disk nt time into kernel timestamp
+ */
+static inline void nt2kernel(const __le64 tm, struct timespec64 *ts)
+{
+	u64 t = le64_to_cpu(tm) - _100ns2seconds * SecondsToStartOf1970;
+
+	// WARNING: do_div changes its first argument(!)
+	ts->tv_nsec = do_div(t, _100ns2seconds) * 100;
+	ts->tv_sec = t;
+}
+
+static inline struct ntfs_sb_info *ntfs_sb(struct super_block *sb)
+{
+	return sb->s_fs_info;
+}
+
+/* Align up on cluster boundary */
+static inline u64 ntfs_up_cluster(const struct ntfs_sb_info *sbi, u64 size)
+{
+	return (size + sbi->cluster_mask) & ~((u64)sbi->cluster_mask);
+}
+
+/* Align up on cluster boundary */
+static inline u64 ntfs_up_block(const struct super_block *sb, u64 size)
+{
+	return (size + sb->s_blocksize - 1) & ~(u64)(sb->s_blocksize - 1);
+}
+
+static inline CLST bytes_to_cluster(const struct ntfs_sb_info *sbi, u64 size)
+{
+	return (size + sbi->cluster_mask) >> sbi->cluster_bits;
+}
+
+static inline u64 bytes_to_block(const struct super_block *sb, u64 size)
+{
+	return (size + sb->s_blocksize - 1) >> sb->s_blocksize_bits;
+}
+
+/* calculates ((bytes + frame_size - 1)/frame_size)*frame_size; */
+static inline u64 ntfs_up_frame(const struct ntfs_sb_info *sbi, u64 bytes,
+				u8 c_unit)
+{
+	u32 bytes_per_frame = 1u << (c_unit + sbi->cluster_bits);
+
+	return (bytes + bytes_per_frame - 1) & ~(u64)(bytes_per_frame - 1);
+}
+
+static inline struct buffer_head *ntfs_bread(struct super_block *sb,
+					     sector_t block)
+{
+	struct buffer_head *bh;
+
+	bh = sb_bread(sb, block);
+	if (bh)
+		return bh;
+
+	ntfs_err(sb, "failed to read volume at offset 0x%llx",
+		 (u64)block << sb->s_blocksize_bits);
+	return NULL;
+}
+
+static inline bool is_power_of2(size_t v)
+{
+	return v && !(v & (v - 1));
+}
+
+static inline struct ntfs_inode *ntfs_i(struct inode *inode)
+{
+	return container_of(inode, struct ntfs_inode, vfs_inode);
+}
+
+static inline bool is_compressed(const struct ntfs_inode *ni)
+{
+	return (ni->std_fa & FILE_ATTRIBUTE_COMPRESSED) ||
+	       (ni->ni_flags & NI_FLAG_COMPRESSED_MASK);
+}
+
+static inline bool is_dedup(const struct ntfs_inode *ni)
+{
+	return ni->ni_flags & NI_FLAG_DEDUPLICATED;
+}
+
+static inline bool is_encrypted(const struct ntfs_inode *ni)
+{
+	return ni->std_fa & FILE_ATTRIBUTE_ENCRYPTED;
+}
+
+static inline bool is_sparsed(const struct ntfs_inode *ni)
+{
+	return ni->std_fa & FILE_ATTRIBUTE_SPARSE_FILE;
+}
+
+static inline void le16_sub_cpu(__le16 *var, u16 val)
+{
+	*var = cpu_to_le16(le16_to_cpu(*var) - val);
+}
+
+static inline void le32_sub_cpu(__le32 *var, u32 val)
+{
+	*var = cpu_to_le32(le32_to_cpu(*var) - val);
+}
+
+static inline void nb_put(struct ntfs_buffers *nb)
+{
+	u32 i, nbufs = nb->nbufs;
+
+	if (!nbufs)
+		return;
+
+	for (i = 0; i < nbufs; i++)
+		put_bh(nb->bh[i]);
+	nb->nbufs = 0;
+}
+
+static inline void put_indx_node(struct indx_node *in)
+{
+	if (!in)
+		return;
+
+	ntfs_free(in->index);
+	nb_put(&in->nb);
+	ntfs_free(in);
+}
+
+static inline void mi_clear(struct mft_inode *mi)
+{
+	nb_put(&mi->nb);
+	ntfs_free(mi->mrec);
+	mi->mrec = NULL;
+}
+
+static inline void ni_lock(struct ntfs_inode *ni)
+{
+	mutex_lock(&ni->ni_lock);
+}
+
+static inline void ni_unlock(struct ntfs_inode *ni)
+{
+	mutex_unlock(&ni->ni_lock);
+}
+
+static inline int ni_trylock(struct ntfs_inode *ni)
+{
+	return mutex_trylock(&ni->ni_lock);
+}
+
+static inline int ni_has_resident_data(struct ntfs_inode *ni)
+{
+	return ni->ni_flags & NI_FLAG_RESIDENT;
+}
+
+static inline int attr_load_runs_attr(struct ntfs_inode *ni,
+				      struct ATTRIB *attr,
+				      struct runs_tree *run, CLST vcn)
+{
+	return attr_load_runs_vcn(ni, attr->type, attr_name(attr),
+				  attr->name_len, run, vcn);
+}
+
+static inline void le64_sub_cpu(__le64 *var, u64 val)
+{
+	*var = cpu_to_le64(le64_to_cpu(*var) - val);
+}
diff --git a/fs/ntfs3/upcase.c b/fs/ntfs3/upcase.c
new file mode 100644
index 000000000000..87b65f9ccc16
--- /dev/null
+++ b/fs/ntfs3/upcase.c
@@ -0,0 +1,77 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/module.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+static inline u16 upcase_unicode_char(const u16 *upcase, u16 chr)
+{
+	if (chr < 'a')
+		return chr;
+
+	if (chr <= 'z')
+		return chr - ('a' - 'A');
+
+	return upcase[chr];
+}
+
+int ntfs_cmp_names(const __le16 *s1, size_t l1, const __le16 *s2, size_t l2,
+		   const u16 *upcase)
+{
+	int diff;
+	size_t len = l1 < l2 ? l1 : l2;
+
+	if (upcase) {
+		while (len--) {
+			diff = upcase_unicode_char(upcase, le16_to_cpu(*s1++)) -
+			       upcase_unicode_char(upcase, le16_to_cpu(*s2++));
+			if (diff)
+				return diff;
+		}
+	} else {
+		while (len--) {
+			diff = le16_to_cpu(*s1++) - le16_to_cpu(*s2++);
+			if (diff)
+				return diff;
+		}
+	}
+
+	return (int)(l1 - l2);
+}
+
+int ntfs_cmp_names_cpu(const struct cpu_str *uni1, const struct le_str *uni2,
+		       const u16 *upcase)
+{
+	const u16 *s1 = uni1->name;
+	const __le16 *s2 = uni2->name;
+	size_t l1 = uni1->len;
+	size_t l2 = uni2->len;
+	size_t len = l1 < l2 ? l1 : l2;
+	int diff;
+
+	if (upcase) {
+		while (len--) {
+			diff = upcase_unicode_char(upcase, *s1++) -
+			       upcase_unicode_char(upcase, le16_to_cpu(*s2++));
+			if (diff)
+				return diff;
+		}
+	} else {
+		while (len--) {
+			diff = *s1++ - le16_to_cpu(*s2++);
+			if (diff)
+				return diff;
+		}
+	}
+
+	return l1 - l2;
+}
-- 
2.25.4


^ permalink raw reply related	[flat|nested] 21+ messages in thread

* [PATCH v11 02/10] fs/ntfs3: Add initialization of super block
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 01/10] fs/ntfs3: Add headers and misc files Konstantin Komarov
@ 2020-10-30 15:02 ` Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 03/10] fs/ntfs3: Add bitmap Konstantin Komarov
                   ` (10 subsequent siblings)
  12 siblings, 0 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This adds initialization of super block

Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
---
 fs/ntfs3/fsntfs.c | 2565 +++++++++++++++++++++++++++++++++++++++++++
 fs/ntfs3/index.c  | 2665 +++++++++++++++++++++++++++++++++++++++++++++
 fs/ntfs3/inode.c  | 2030 ++++++++++++++++++++++++++++++++++
 fs/ntfs3/super.c  | 1479 +++++++++++++++++++++++++
 4 files changed, 8739 insertions(+)
 create mode 100644 fs/ntfs3/fsntfs.c
 create mode 100644 fs/ntfs3/index.c
 create mode 100644 fs/ntfs3/inode.c
 create mode 100644 fs/ntfs3/super.c

diff --git a/fs/ntfs3/fsntfs.c b/fs/ntfs3/fsntfs.c
new file mode 100644
index 000000000000..a75b48e0c003
--- /dev/null
+++ b/fs/ntfs3/fsntfs.c
@@ -0,0 +1,2565 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+const struct cpu_str NAME_MFT = {
+	4,
+	0,
+	{ '$', 'M', 'F', 'T' },
+};
+const struct cpu_str NAME_MIRROR = {
+	8,
+	0,
+	{ '$', 'M', 'F', 'T', 'M', 'i', 'r', 'r' },
+};
+const struct cpu_str NAME_LOGFILE = {
+	8,
+	0,
+	{ '$', 'L', 'o', 'g', 'F', 'i', 'l', 'e' },
+};
+const struct cpu_str NAME_VOLUME = {
+	7,
+	0,
+	{ '$', 'V', 'o', 'l', 'u', 'm', 'e' },
+};
+const struct cpu_str NAME_ATTRDEF = {
+	8,
+	0,
+	{ '$', 'A', 't', 't', 'r', 'D', 'e', 'f' },
+};
+const struct cpu_str NAME_ROOT = {
+	1,
+	0,
+	{ '.' },
+};
+const struct cpu_str NAME_BITMAP = {
+	7,
+	0,
+	{ '$', 'B', 'i', 't', 'm', 'a', 'p' },
+};
+const struct cpu_str NAME_BOOT = {
+	5,
+	0,
+	{ '$', 'B', 'o', 'o', 't' },
+};
+const struct cpu_str NAME_BADCLUS = {
+	8,
+	0,
+	{ '$', 'B', 'a', 'd', 'C', 'l', 'u', 's' },
+};
+const struct cpu_str NAME_QUOTA = {
+	6,
+	0,
+	{ '$', 'Q', 'u', 'o', 't', 'a' },
+};
+const struct cpu_str NAME_SECURE = {
+	7,
+	0,
+	{ '$', 'S', 'e', 'c', 'u', 'r', 'e' },
+};
+const struct cpu_str NAME_UPCASE = {
+	7,
+	0,
+	{ '$', 'U', 'p', 'C', 'a', 's', 'e' },
+};
+const struct cpu_str NAME_EXTEND = {
+	7,
+	0,
+	{ '$', 'E', 'x', 't', 'e', 'n', 'd' },
+};
+const struct cpu_str NAME_OBJID = {
+	6,
+	0,
+	{ '$', 'O', 'b', 'j', 'I', 'd' },
+};
+const struct cpu_str NAME_REPARSE = {
+	8,
+	0,
+	{ '$', 'R', 'e', 'p', 'a', 'r', 's', 'e' },
+};
+const struct cpu_str NAME_USNJRNL = {
+	8,
+	0,
+	{ '$', 'U', 's', 'n', 'J', 'r', 'n', 'l' },
+};
+const __le16 BAD_NAME[4] = {
+	cpu_to_le16('$'),
+	cpu_to_le16('B'),
+	cpu_to_le16('a'),
+	cpu_to_le16('d'),
+};
+const __le16 I30_NAME[4] = {
+	cpu_to_le16('$'),
+	cpu_to_le16('I'),
+	cpu_to_le16('3'),
+	cpu_to_le16('0'),
+};
+const __le16 SII_NAME[4] = {
+	cpu_to_le16('$'),
+	cpu_to_le16('S'),
+	cpu_to_le16('I'),
+	cpu_to_le16('I'),
+};
+const __le16 SDH_NAME[4] = {
+	cpu_to_le16('$'),
+	cpu_to_le16('S'),
+	cpu_to_le16('D'),
+	cpu_to_le16('H'),
+};
+const __le16 SDS_NAME[4] = {
+	cpu_to_le16('$'),
+	cpu_to_le16('S'),
+	cpu_to_le16('D'),
+	cpu_to_le16('S'),
+};
+const __le16 SO_NAME[2] = {
+	cpu_to_le16('$'),
+	cpu_to_le16('O'),
+};
+const __le16 SQ_NAME[2] = {
+	cpu_to_le16('$'),
+	cpu_to_le16('Q'),
+};
+const __le16 SR_NAME[2] = {
+	cpu_to_le16('$'),
+	cpu_to_le16('R'),
+};
+//const __le16 EFS_NAME[4] = { '$', 'E', 'F', 'S' };
+
+//const __le16 WOF_NAME[17] = { 'W', 'o', 'f', 'C', 'o', 'm', 'p', 'r', 'e',
+//			      'S', 'S', 'e', 'd', 'D', 'a', 't', 'a' };
+//const __le16 J_NAME[2] = { '$', 'J' };
+//const __le16 MAX_NAME[4] = { '$', 'M', 'a', 'x' };
+
+/*
+ * ntfs_fix_pre_write
+ *
+ * inserts fixups into 'rhdr' before writing to disk
+ */
+bool ntfs_fix_pre_write(struct NTFS_RECORD_HEADER *rhdr, size_t bytes)
+{
+	u16 *fixup, *ptr;
+	u16 sample;
+	u16 fo = le16_to_cpu(rhdr->fix_off);
+	u16 fn = le16_to_cpu(rhdr->fix_num);
+
+	if ((fo & 1) || fo + fn * sizeof(short) > SECTOR_SIZE || !fn-- ||
+	    fn * SECTOR_SIZE > bytes) {
+		return false;
+	}
+
+	/* Get fixup pointer */
+	fixup = Add2Ptr(rhdr, fo);
+
+	if (*fixup >= 0x7FFF)
+		*fixup = 1;
+	else
+		*fixup += 1;
+
+	sample = *fixup;
+
+	ptr = Add2Ptr(rhdr, SECTOR_SIZE - sizeof(short));
+
+	while (fn--) {
+		*++fixup = *ptr;
+		*ptr = sample;
+		ptr += SECTOR_SIZE / sizeof(short);
+	}
+	return true;
+}
+
+/*
+ * ntfs_fix_post_read
+ *
+ * remove fixups after reading from disk
+ * Returns < 0 if error, 0 if ok, 1 if need to update fixups
+ */
+int ntfs_fix_post_read(struct NTFS_RECORD_HEADER *rhdr, size_t bytes,
+		       bool simple)
+{
+	int ret;
+	u16 *fixup, *ptr;
+	u16 sample, fo, fn;
+
+	fo = le16_to_cpu(rhdr->fix_off);
+	fn = simple ? ((bytes >> SECTOR_SHIFT) + 1) :
+		      le16_to_cpu(rhdr->fix_num);
+
+	/* Check errors */
+	if ((fo & 1) || fo + fn * sizeof(short) > SECTOR_SIZE || !fn-- ||
+	    fn * SECTOR_SIZE > bytes) {
+		return -EINVAL; /* native chkntfs returns ok! */
+	}
+
+	/* Get fixup pointer */
+	fixup = Add2Ptr(rhdr, fo);
+	sample = *fixup;
+	ptr = Add2Ptr(rhdr, SECTOR_SIZE - sizeof(short));
+	ret = 0;
+
+	while (fn--) {
+		/* Test current word */
+		if (*ptr != sample) {
+			/* Fixup does not match! Is it serious error? */
+			ret = -E_NTFS_FIXUP;
+		}
+
+		/* Replace fixup */
+		*ptr = *++fixup;
+		ptr += SECTOR_SIZE / sizeof(short);
+	}
+
+	return ret;
+}
+
+/*
+ * ntfs_extend_init
+ *
+ * loads $Extend file
+ */
+int ntfs_extend_init(struct ntfs_sb_info *sbi)
+{
+	int err;
+	struct super_block *sb = sbi->sb;
+	struct inode *inode, *inode2;
+	struct MFT_REF ref;
+
+	if (sbi->volume.major_ver < 3) {
+		ntfs_notice(sb, "Skip $Extend 'cause NTFS version");
+		return 0;
+	}
+
+	ref.low = cpu_to_le32(MFT_REC_EXTEND);
+	ref.high = 0;
+	ref.seq = cpu_to_le16(MFT_REC_EXTEND);
+	inode = ntfs_iget5(sb, &ref, &NAME_EXTEND);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load $Extend.");
+		inode = NULL;
+		goto out;
+	}
+
+	/* if ntfs_iget5 reads from disk it never returns bad inode */
+	if (!S_ISDIR(inode->i_mode)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* Try to find $ObjId */
+	inode2 = dir_search_u(inode, &NAME_OBJID, NULL);
+	if (inode2 && !IS_ERR(inode2)) {
+		if (is_bad_inode(inode2)) {
+			iput(inode2);
+		} else {
+			sbi->objid.ni = ntfs_i(inode2);
+			sbi->objid_no = inode2->i_ino;
+		}
+	}
+
+	/* Try to find $Quota */
+	inode2 = dir_search_u(inode, &NAME_QUOTA, NULL);
+	if (inode2 && !IS_ERR(inode2)) {
+		sbi->quota_no = inode2->i_ino;
+		iput(inode2);
+	}
+
+	/* Try to find $Reparse */
+	inode2 = dir_search_u(inode, &NAME_REPARSE, NULL);
+	if (inode2 && !IS_ERR(inode2)) {
+		sbi->reparse.ni = ntfs_i(inode2);
+		sbi->reparse_no = inode2->i_ino;
+	}
+
+	/* Try to find $UsnJrnl */
+	inode2 = dir_search_u(inode, &NAME_USNJRNL, NULL);
+	if (inode2 && !IS_ERR(inode2)) {
+		sbi->usn_jrnl_no = inode2->i_ino;
+		iput(inode2);
+	}
+
+	err = 0;
+out:
+	iput(inode);
+	return err;
+}
+
+int ntfs_loadlog_and_replay(struct ntfs_inode *ni, struct ntfs_sb_info *sbi)
+{
+	int err = 0;
+	struct super_block *sb = sbi->sb;
+	struct inode *inode = &ni->vfs_inode;
+	struct MFT_REF ref;
+
+	/* Check for 4GB */
+	if (inode->i_size >= 0x100000000ull) {
+		ntfs_err(sb, "$LogFile is too big");
+		err = -EINVAL;
+		goto out;
+	}
+
+	sbi->flags |= NTFS_FLAGS_LOG_REPLAYING;
+
+	ref.low = cpu_to_le32(MFT_REC_MFT);
+	ref.high = 0;
+	ref.seq = cpu_to_le16(1);
+
+	inode = ntfs_iget5(sb, &ref, NULL);
+
+	if (IS_ERR(inode))
+		inode = NULL;
+
+	if (!inode) {
+		/* Try to use mft copy */
+		u64 t64 = sbi->mft.lbo;
+
+		sbi->mft.lbo = sbi->mft.lbo2;
+		inode = ntfs_iget5(sb, &ref, NULL);
+		sbi->mft.lbo = t64;
+		if (IS_ERR(inode))
+			inode = NULL;
+	}
+
+	if (!inode) {
+		err = -EINVAL;
+		ntfs_err(sb, "Failed to load $MFT.");
+		goto out;
+	}
+
+	sbi->mft.ni = ntfs_i(inode);
+
+	err = ni_load_all_mi(sbi->mft.ni);
+	if (!err)
+		err = log_replay(ni);
+
+	iput(inode);
+	sbi->mft.ni = NULL;
+
+	sync_blockdev(sb->s_bdev);
+	invalidate_bdev(sb->s_bdev);
+
+	/* reinit MFT */
+	if (sbi->flags & NTFS_FLAGS_NEED_REPLAY) {
+		err = 0;
+		goto out;
+	}
+
+	if (sb_rdonly(sb))
+		goto out;
+
+	err = ntfs_bio_fill_1(sbi, &ni->file.run);
+
+out:
+	sbi->flags &= ~NTFS_FLAGS_LOG_REPLAYING;
+
+	return err;
+}
+
+/*
+ * ntfs_query_def
+ *
+ * returns current ATTR_DEF_ENTRY for given attribute type
+ */
+const struct ATTR_DEF_ENTRY *ntfs_query_def(struct ntfs_sb_info *sbi,
+					    enum ATTR_TYPE type)
+{
+	int type_in = le32_to_cpu(type);
+	size_t min_idx = 0;
+	size_t max_idx = sbi->def_entries - 1;
+
+	while (min_idx <= max_idx) {
+		size_t i = min_idx + ((max_idx - min_idx) >> 1);
+		const struct ATTR_DEF_ENTRY *entry = sbi->def_table + i;
+		int diff = le32_to_cpu(entry->type) - type_in;
+
+		if (!diff)
+			return entry;
+		if (diff < 0)
+			min_idx = i + 1;
+		else if (i)
+			max_idx = i - 1;
+		else
+			return NULL;
+	}
+	return NULL;
+}
+
+/*
+ * ntfs_look_for_free_space
+ *
+ * looks for a free space in bitmap
+ */
+int ntfs_look_for_free_space(struct ntfs_sb_info *sbi, CLST lcn, CLST len,
+			     CLST *new_lcn, CLST *new_len,
+			     enum ALLOCATE_OPT opt)
+{
+	int err;
+	struct super_block *sb = sbi->sb;
+	size_t a_lcn, zlen, zeroes, zlcn, zlen2, ztrim, new_zlen;
+	struct wnd_bitmap *wnd = &sbi->used.bitmap;
+
+	down_write_nested(&wnd->rw_lock, BITMAP_MUTEX_CLUSTERS);
+	if (opt & ALLOCATE_MFT) {
+		CLST alen;
+
+		zlen = wnd_zone_len(wnd);
+
+		if (!zlen) {
+			err = ntfs_refresh_zone(sbi);
+			if (err)
+				goto out;
+
+			zlen = wnd_zone_len(wnd);
+
+			if (!zlen) {
+				ntfs_err(sbi->sb,
+					 "no free space to extend mft");
+				err = -ENOSPC;
+				goto out;
+			}
+		}
+
+		lcn = wnd_zone_bit(wnd);
+		alen = zlen > len ? len : zlen;
+
+		wnd_zone_set(wnd, lcn + alen, zlen - alen);
+
+		err = wnd_set_used(wnd, lcn, alen);
+		if (err)
+			goto out;
+
+		*new_lcn = lcn;
+		*new_len = alen;
+		goto ok;
+	}
+
+	/*
+	 * 'Cause cluster 0 is always used this value means that we should use
+	 * cached value of 'next_free_lcn' to improve performance
+	 */
+	if (!lcn)
+		lcn = sbi->used.next_free_lcn;
+
+	if (lcn >= wnd->nbits)
+		lcn = 0;
+
+	*new_len = wnd_find(wnd, len, lcn, BITMAP_FIND_MARK_AS_USED, &a_lcn);
+	if (*new_len) {
+		*new_lcn = a_lcn;
+		goto ok;
+	}
+
+	/* Try to use clusters from MftZone */
+	zlen = wnd_zone_len(wnd);
+	zeroes = wnd_zeroes(wnd);
+
+	/* Check too big request */
+	if (len > zeroes + zlen)
+		goto no_space;
+
+	if (zlen <= NTFS_MIN_MFT_ZONE)
+		goto no_space;
+
+	/* How many clusters to cat from zone */
+	zlcn = wnd_zone_bit(wnd);
+	zlen2 = zlen >> 1;
+	ztrim = len > zlen ? zlen : (len > zlen2 ? len : zlen2);
+	new_zlen = zlen - ztrim;
+
+	if (new_zlen < NTFS_MIN_MFT_ZONE) {
+		new_zlen = NTFS_MIN_MFT_ZONE;
+		if (new_zlen > zlen)
+			new_zlen = zlen;
+	}
+
+	wnd_zone_set(wnd, zlcn, new_zlen);
+
+	/* allocate continues clusters */
+	*new_len =
+		wnd_find(wnd, len, 0,
+			 BITMAP_FIND_MARK_AS_USED | BITMAP_FIND_FULL, &a_lcn);
+	if (*new_len) {
+		*new_lcn = a_lcn;
+		goto ok;
+	}
+
+no_space:
+	up_write(&wnd->rw_lock);
+
+	return -ENOSPC;
+
+ok:
+	err = 0;
+
+	ntfs_unmap_meta(sb, *new_lcn, *new_len);
+
+	if (opt & ALLOCATE_MFT)
+		goto out;
+
+	/* Set hint for next requests */
+	sbi->used.next_free_lcn = *new_lcn + *new_len;
+
+out:
+	up_write(&wnd->rw_lock);
+	return err;
+}
+
+/*
+ * ntfs_extend_mft
+ *
+ * allocates additional MFT records
+ * sbi->mft.bitmap is locked for write
+ *
+ * NOTE: recursive:
+ *	ntfs_look_free_mft ->
+ *	ntfs_extend_mft ->
+ *	attr_set_size ->
+ *	ni_insert_nonresident ->
+ *	ni_insert_attr ->
+ *	ni_ins_attr_ext ->
+ *	ntfs_look_free_mft ->
+ *	ntfs_extend_mft
+ * To avoid recursive always allocate space for two new mft records
+ * see attrib.c: "at least two mft to avoid recursive loop"
+ */
+static int ntfs_extend_mft(struct ntfs_sb_info *sbi)
+{
+	int err;
+	struct ntfs_inode *ni = sbi->mft.ni;
+	size_t new_mft_total;
+	u64 new_mft_bytes, new_bitmap_bytes;
+	struct ATTRIB *attr;
+	struct wnd_bitmap *wnd = &sbi->mft.bitmap;
+
+	new_mft_total = (wnd->nbits + MFT_INCREASE_CHUNK + 127) & (CLST)~127;
+	new_mft_bytes = (u64)new_mft_total << sbi->record_bits;
+
+	/* Step 1: Resize $MFT::DATA */
+	down_write(&ni->file.run_lock);
+	err = attr_set_size(ni, ATTR_DATA, NULL, 0, &ni->file.run,
+			    new_mft_bytes, NULL, false, &attr);
+
+	if (err) {
+		up_write(&ni->file.run_lock);
+		goto out;
+	}
+
+	attr->nres.valid_size = attr->nres.data_size;
+	new_mft_total = le64_to_cpu(attr->nres.alloc_size) >> sbi->record_bits;
+	ni->mi.dirty = true;
+
+	/* Step 2: Resize $MFT::BITMAP */
+	new_bitmap_bytes = bitmap_size(new_mft_total);
+
+	err = attr_set_size(ni, ATTR_BITMAP, NULL, 0, &sbi->mft.bitmap.run,
+			    new_bitmap_bytes, &new_bitmap_bytes, true, NULL);
+
+	/* Refresh Mft Zone if necessary */
+	down_write_nested(&sbi->used.bitmap.rw_lock, BITMAP_MUTEX_CLUSTERS);
+
+	ntfs_refresh_zone(sbi);
+
+	up_write(&sbi->used.bitmap.rw_lock);
+	up_write(&ni->file.run_lock);
+
+	if (err)
+		goto out;
+
+	err = wnd_extend(wnd, new_mft_total);
+
+	if (err)
+		goto out;
+
+	ntfs_clear_mft_tail(sbi, sbi->mft.used, new_mft_total);
+
+	err = _ni_write_inode(&ni->vfs_inode, 0);
+out:
+	return err;
+}
+
+/*
+ * ntfs_look_free_mft
+ *
+ * looks for a free MFT record
+ */
+int ntfs_look_free_mft(struct ntfs_sb_info *sbi, CLST *rno, bool mft,
+		       struct ntfs_inode *ni, struct mft_inode **mi)
+{
+	int err = 0;
+	size_t zbit, zlen, from, to, fr;
+	size_t mft_total;
+	struct MFT_REF ref;
+	struct super_block *sb = sbi->sb;
+	struct wnd_bitmap *wnd = &sbi->mft.bitmap;
+	u32 ir;
+
+	static_assert(sizeof(sbi->mft.reserved_bitmap) * 8 >=
+		      MFT_REC_FREE - MFT_REC_RESERVED);
+
+	if (!mft)
+		down_write_nested(&wnd->rw_lock, BITMAP_MUTEX_MFT);
+
+	zlen = wnd_zone_len(wnd);
+
+	/* Always reserve space for MFT */
+	if (zlen) {
+		if (mft) {
+			zbit = wnd_zone_bit(wnd);
+			*rno = zbit;
+			wnd_zone_set(wnd, zbit + 1, zlen - 1);
+		}
+		goto found;
+	}
+
+	/* No MFT zone. find the nearest to '0' free MFT */
+	if (!wnd_find(wnd, 1, MFT_REC_FREE, 0, &zbit)) {
+		/* Resize MFT */
+		mft_total = wnd->nbits;
+
+		err = ntfs_extend_mft(sbi);
+		if (!err) {
+			zbit = mft_total;
+			goto reserve_mft;
+		}
+
+		if (!mft || MFT_REC_FREE == sbi->mft.next_reserved)
+			goto out;
+
+		err = 0;
+
+		/*
+		 * Look for free record reserved area [11-16) ==
+		 * [MFT_REC_RESERVED, MFT_REC_FREE ) MFT bitmap always
+		 * marks it as used
+		 */
+		if (!sbi->mft.reserved_bitmap) {
+			/* Once per session create internal bitmap for 5 bits */
+			sbi->mft.reserved_bitmap = 0xFF;
+
+			ref.high = 0;
+			for (ir = MFT_REC_RESERVED; ir < MFT_REC_FREE; ir++) {
+				struct inode *i;
+				struct ntfs_inode *ni;
+				struct MFT_REC *mrec;
+
+				ref.low = cpu_to_le32(ir);
+				ref.seq = cpu_to_le16(ir);
+
+				i = ntfs_iget5(sb, &ref, NULL);
+				if (IS_ERR(i)) {
+next:
+					ntfs_notice(
+						sb,
+						"Invalid reserved record %x",
+						ref.low);
+					continue;
+				}
+				if (is_bad_inode(i)) {
+					iput(i);
+					goto next;
+				}
+
+				ni = ntfs_i(i);
+
+				mrec = ni->mi.mrec;
+
+				if (!is_rec_base(mrec))
+					goto next;
+
+				if (mrec->hard_links)
+					goto next;
+
+				if (!ni_std(ni))
+					goto next;
+
+				if (ni_find_attr(ni, NULL, NULL, ATTR_NAME,
+						 NULL, 0, NULL, NULL))
+					goto next;
+
+				__clear_bit(ir - MFT_REC_RESERVED,
+					    &sbi->mft.reserved_bitmap);
+			}
+		}
+
+		/* Scan 5 bits for zero. Bit 0 == MFT_REC_RESERVED */
+		zbit = find_next_zero_bit(&sbi->mft.reserved_bitmap,
+					  MFT_REC_FREE, MFT_REC_RESERVED);
+		if (zbit >= MFT_REC_FREE) {
+			sbi->mft.next_reserved = MFT_REC_FREE;
+			goto out;
+		}
+
+		zlen = 1;
+		sbi->mft.next_reserved = zbit;
+	} else {
+reserve_mft:
+		zlen = zbit == MFT_REC_FREE ? (MFT_REC_USER - MFT_REC_FREE) : 4;
+		if (zbit + zlen > wnd->nbits)
+			zlen = wnd->nbits - zbit;
+
+		while (zlen > 1 && !wnd_is_free(wnd, zbit, zlen))
+			zlen -= 1;
+
+		/* [zbit, zbit + zlen) will be used for Mft itself */
+		from = sbi->mft.used;
+		if (from < zbit)
+			from = zbit;
+		to = zbit + zlen;
+		if (from < to) {
+			ntfs_clear_mft_tail(sbi, from, to);
+			sbi->mft.used = to;
+		}
+	}
+
+	if (mft) {
+		*rno = zbit;
+		zbit += 1;
+		zlen -= 1;
+	}
+
+	wnd_zone_set(wnd, zbit, zlen);
+
+found:
+	if (!mft) {
+		/* The request to get record for general purpose */
+		if (sbi->mft.next_free < MFT_REC_USER)
+			sbi->mft.next_free = MFT_REC_USER;
+
+		for (;;) {
+			if (sbi->mft.next_free >= sbi->mft.bitmap.nbits) {
+			} else if (!wnd_find(wnd, 1, MFT_REC_USER, 0, &fr)) {
+				sbi->mft.next_free = sbi->mft.bitmap.nbits;
+			} else {
+				*rno = fr;
+				sbi->mft.next_free = *rno + 1;
+				break;
+			}
+
+			err = ntfs_extend_mft(sbi);
+			if (err)
+				goto out;
+		}
+	}
+
+	if (ni && !ni_add_subrecord(ni, *rno, mi)) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	/* We have found a record that are not reserved for next MFT */
+	if (*rno >= MFT_REC_FREE)
+		wnd_set_used(wnd, *rno, 1);
+	else if (*rno >= MFT_REC_RESERVED && sbi->mft.reserved_bitmap_inited)
+		__set_bit(*rno - MFT_REC_RESERVED, &sbi->mft.reserved_bitmap);
+
+out:
+	if (!mft)
+		up_write(&wnd->rw_lock);
+
+	return err;
+}
+
+/*
+ * ntfs_mark_rec_free
+ *
+ * marks record as free
+ */
+void ntfs_mark_rec_free(struct ntfs_sb_info *sbi, CLST rno)
+{
+	struct wnd_bitmap *wnd = &sbi->mft.bitmap;
+
+	down_write_nested(&wnd->rw_lock, BITMAP_MUTEX_MFT);
+	if (rno >= wnd->nbits)
+		goto out;
+
+	if (rno >= MFT_REC_FREE) {
+		if (!wnd_is_used(wnd, rno, 1))
+			ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
+		else
+			wnd_set_free(wnd, rno, 1);
+	} else if (rno >= MFT_REC_RESERVED && sbi->mft.reserved_bitmap_inited) {
+		__clear_bit(rno - MFT_REC_RESERVED, &sbi->mft.reserved_bitmap);
+	}
+
+	if (rno < wnd_zone_bit(wnd))
+		wnd_zone_set(wnd, rno, 1);
+	else if (rno < sbi->mft.next_free && rno >= MFT_REC_USER)
+		sbi->mft.next_free = rno;
+
+out:
+	up_write(&wnd->rw_lock);
+}
+
+/*
+ * ntfs_clear_mft_tail
+ *
+ * formats empty records [from, to)
+ * sbi->mft.bitmap is locked for write
+ */
+int ntfs_clear_mft_tail(struct ntfs_sb_info *sbi, size_t from, size_t to)
+{
+	int err;
+	u32 rs;
+	u64 vbo;
+	struct runs_tree *run;
+	struct ntfs_inode *ni;
+
+	if (from >= to)
+		return 0;
+
+	rs = sbi->record_size;
+	ni = sbi->mft.ni;
+	run = &ni->file.run;
+
+	down_read(&ni->file.run_lock);
+	vbo = (u64)from * rs;
+	for (; from < to; from++, vbo += rs) {
+		struct ntfs_buffers nb;
+
+		err = ntfs_get_bh(sbi, run, vbo, rs, &nb);
+		if (err)
+			goto out;
+
+		err = ntfs_write_bh(sbi, &sbi->new_rec->rhdr, &nb, 0);
+		nb_put(&nb);
+		if (err)
+			goto out;
+	}
+
+out:
+	sbi->mft.used = from;
+	up_read(&ni->file.run_lock);
+	return err;
+}
+
+/*
+ * ntfs_refresh_zone
+ *
+ * refreshes Mft zone
+ * sbi->used.bitmap is locked for rw
+ * sbi->mft.bitmap is locked for write
+ * sbi->mft.ni->file.run_lock for write
+ */
+int ntfs_refresh_zone(struct ntfs_sb_info *sbi)
+{
+	CLST zone_limit, zone_max, lcn, vcn, len;
+	size_t lcn_s, zlen;
+	struct wnd_bitmap *wnd = &sbi->used.bitmap;
+	struct ntfs_inode *ni = sbi->mft.ni;
+
+	/* Do not change anything unless we have non empty Mft zone */
+	if (wnd_zone_len(wnd))
+		return 0;
+
+	/*
+	 * Compute the mft zone at two steps
+	 * It would be nice if we are able to allocate
+	 * 1/8 of total clusters for MFT but not more then 512 MB
+	 */
+	zone_limit = (512 * 1024 * 1024) >> sbi->cluster_bits;
+	zone_max = wnd->nbits >> 3;
+	if (zone_max > zone_limit)
+		zone_max = zone_limit;
+
+	vcn = bytes_to_cluster(sbi,
+			       (u64)sbi->mft.bitmap.nbits << sbi->record_bits);
+
+	if (!run_lookup_entry(&ni->file.run, vcn - 1, &lcn, &len, NULL))
+		lcn = SPARSE_LCN;
+
+	/* We should always find Last Lcn for MFT */
+	if (lcn == SPARSE_LCN)
+		return -EINVAL;
+
+	lcn_s = lcn + 1;
+
+	/* Try to allocate clusters after last MFT run */
+	zlen = wnd_find(wnd, zone_max, lcn_s, 0, &lcn_s);
+	if (!zlen) {
+		ntfs_notice(sbi->sb, "MftZone: unavailable");
+		return 0;
+	}
+
+	/* Truncate too large zone */
+	wnd_zone_set(wnd, lcn_s, zlen);
+
+	return 0;
+}
+
+/*
+ * ntfs_update_mftmirr
+ *
+ * updates $MFTMirr data
+ */
+int ntfs_update_mftmirr(struct ntfs_sb_info *sbi, int wait)
+{
+	int err;
+	struct super_block *sb = sbi->sb;
+	u32 blocksize = sb->s_blocksize;
+	sector_t block1, block2;
+	u32 bytes;
+
+	if (!(sbi->flags & NTFS_FLAGS_MFTMIRR))
+		return 0;
+
+	err = 0;
+	bytes = sbi->mft.recs_mirr << sbi->record_bits;
+	block1 = sbi->mft.lbo >> sb->s_blocksize_bits;
+	block2 = sbi->mft.lbo2 >> sb->s_blocksize_bits;
+
+	for (; bytes >= blocksize; bytes -= blocksize) {
+		struct buffer_head *bh1, *bh2;
+
+		bh1 = sb_bread(sb, block1++);
+		if (!bh1) {
+			err = -EIO;
+			goto out;
+		}
+
+		bh2 = sb_getblk(sb, block2++);
+		if (!bh2) {
+			put_bh(bh1);
+			err = -EIO;
+			goto out;
+		}
+
+		if (buffer_locked(bh2))
+			__wait_on_buffer(bh2);
+
+		lock_buffer(bh2);
+		memcpy(bh2->b_data, bh1->b_data, blocksize);
+		set_buffer_uptodate(bh2);
+		mark_buffer_dirty(bh2);
+		unlock_buffer(bh2);
+
+		put_bh(bh1);
+		bh1 = NULL;
+
+		if (wait)
+			err = sync_dirty_buffer(bh2);
+
+		put_bh(bh2);
+		if (err)
+			goto out;
+	}
+
+	sbi->flags &= ~NTFS_FLAGS_MFTMIRR;
+
+out:
+	return err;
+}
+
+/*
+ * ntfs_set_state
+ *
+ * mount: ntfs_set_state(NTFS_DIRTY_DIRTY)
+ * umount: ntfs_set_state(NTFS_DIRTY_CLEAR)
+ * ntfs error: ntfs_set_state(NTFS_DIRTY_ERROR)
+ */
+int ntfs_set_state(struct ntfs_sb_info *sbi, enum NTFS_DIRTY_FLAGS dirty)
+{
+	int err;
+	struct ATTRIB *attr;
+	struct VOLUME_INFO *info;
+	struct mft_inode *mi;
+	struct ntfs_inode *ni;
+
+	/*
+	 * do not change state if fs was real_dirty
+	 * do not change state if fs already dirty(clear)
+	 * do not change any thing if mounted read only
+	 */
+	if (sbi->volume.real_dirty || sb_rdonly(sbi->sb))
+		return 0;
+
+	/* Check cached value */
+	if ((dirty == NTFS_DIRTY_CLEAR ? 0 : VOLUME_FLAG_DIRTY) ==
+	    (sbi->volume.flags & VOLUME_FLAG_DIRTY))
+		return 0;
+
+	ni = sbi->volume.ni;
+	if (!ni)
+		return -EINVAL;
+
+	inode_lock(&ni->vfs_inode);
+
+	attr = ni_find_attr(ni, NULL, NULL, ATTR_VOL_INFO, NULL, 0, NULL, &mi);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	info = resident_data_ex(attr, SIZEOF_ATTRIBUTE_VOLUME_INFO);
+	if (!info) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	switch (dirty) {
+	case NTFS_DIRTY_ERROR:
+		ntfs_notice(sbi->sb, "Mark volume as dirty due to NTFS errors");
+		sbi->volume.real_dirty = true;
+		fallthrough;
+	case NTFS_DIRTY_DIRTY:
+		info->flags |= VOLUME_FLAG_DIRTY;
+		break;
+	case NTFS_DIRTY_CLEAR:
+		info->flags &= ~VOLUME_FLAG_DIRTY;
+		break;
+	}
+	/* cache current volume flags*/
+	sbi->volume.flags = info->flags;
+	mi->dirty = true;
+	err = 0;
+
+out:
+	inode_unlock(&ni->vfs_inode);
+	if (err)
+		return err;
+
+	mark_inode_dirty(&ni->vfs_inode);
+	/*verify(!ntfs_update_mftmirr()); */
+	err = sync_inode_metadata(&ni->vfs_inode, 1);
+
+	return err;
+}
+
+/*
+ * security_hash
+ *
+ * calculates a hash of security descriptor
+ */
+static inline __le32 security_hash(const void *sd, size_t bytes)
+{
+	u32 hash = 0;
+	const __le32 *ptr = sd;
+
+	bytes >>= 2;
+	while (bytes--)
+		hash = ((hash >> 0x1D) | (hash << 3)) + le32_to_cpu(*ptr++);
+	return cpu_to_le32(hash);
+}
+
+int ntfs_sb_read(struct super_block *sb, u64 lbo, size_t bytes, void *buffer)
+{
+	struct block_device *bdev = sb->s_bdev;
+	u32 blocksize = sb->s_blocksize;
+	u64 block = lbo >> sb->s_blocksize_bits;
+	u32 off = lbo & (blocksize - 1);
+	u32 op = blocksize - off;
+
+	for (; bytes; block += 1, off = 0, op = blocksize) {
+		struct buffer_head *bh = __bread(bdev, block, blocksize);
+
+		if (!bh)
+			return -EIO;
+
+		if (op > bytes)
+			op = bytes;
+
+		memcpy(buffer, bh->b_data + off, op);
+
+		put_bh(bh);
+
+		bytes -= op;
+		buffer = Add2Ptr(buffer, op);
+	}
+
+	return 0;
+}
+
+int ntfs_sb_write(struct super_block *sb, u64 lbo, size_t bytes,
+		  const void *buf, int wait)
+{
+	u32 blocksize = sb->s_blocksize;
+	struct block_device *bdev = sb->s_bdev;
+	sector_t block = lbo >> sb->s_blocksize_bits;
+	u32 off = lbo & (blocksize - 1);
+	u32 op = blocksize - off;
+	struct buffer_head *bh;
+
+	if (!wait && (sb->s_flags & SB_SYNCHRONOUS))
+		wait = 1;
+
+	for (; bytes; block += 1, off = 0, op = blocksize) {
+		if (op > bytes)
+			op = bytes;
+
+		if (op < blocksize) {
+			bh = __bread(bdev, block, blocksize);
+			if (!bh) {
+				ntfs_err(sb, "failed to read block %llx",
+					 (u64)block);
+				return -EIO;
+			}
+		} else {
+			bh = __getblk(bdev, block, blocksize);
+			if (!bh)
+				return -ENOMEM;
+		}
+
+		if (buffer_locked(bh))
+			__wait_on_buffer(bh);
+
+		lock_buffer(bh);
+		if (buf) {
+			memcpy(bh->b_data + off, buf, op);
+			buf = Add2Ptr(buf, op);
+		} else {
+			memset(bh->b_data + off, -1, op);
+		}
+
+		set_buffer_uptodate(bh);
+		mark_buffer_dirty(bh);
+		unlock_buffer(bh);
+
+		if (wait) {
+			int err = sync_dirty_buffer(bh);
+
+			if (err) {
+				ntfs_err(
+					sb,
+					"failed to sync buffer at block %llx, error %d",
+					(u64)block, err);
+				put_bh(bh);
+				return err;
+			}
+		}
+
+		put_bh(bh);
+
+		bytes -= op;
+	}
+	return 0;
+}
+
+int ntfs_sb_write_run(struct ntfs_sb_info *sbi, const struct runs_tree *run,
+		      u64 vbo, const void *buf, size_t bytes)
+{
+	struct super_block *sb = sbi->sb;
+	u8 cluster_bits = sbi->cluster_bits;
+	u32 off = vbo & sbi->cluster_mask;
+	CLST lcn, clen;
+	u64 lbo, len;
+	size_t idx;
+
+	if (!run_lookup_entry(run, vbo >> cluster_bits, &lcn, &clen, &idx))
+		return -ENOENT;
+
+	if (lcn == SPARSE_LCN)
+		return -EINVAL;
+
+	lbo = ((u64)lcn << cluster_bits) + off;
+	len = ((u64)clen << cluster_bits) - off;
+
+	for (;;) {
+		u32 op = len < bytes ? len : bytes;
+		int err = ntfs_sb_write(sb, lbo, op, buf, 0);
+
+		if (err)
+			return err;
+
+		bytes -= op;
+		if (!bytes)
+			break;
+
+		if (!run_get_entry(run, ++idx, NULL, &lcn, &clen))
+			return -ENOENT;
+
+		if (lcn == SPARSE_LCN)
+			return -EINVAL;
+
+		if (buf)
+			buf = Add2Ptr(buf, op);
+
+		lbo = ((u64)lcn << cluster_bits) + off;
+		len = ((u64)clen << cluster_bits) - off;
+	}
+
+	return 0;
+}
+
+struct buffer_head *ntfs_bread_run(struct ntfs_sb_info *sbi,
+				   const struct runs_tree *run, u64 vbo)
+{
+	struct super_block *sb = sbi->sb;
+	u8 cluster_bits = sbi->cluster_bits;
+	CLST lcn;
+	u64 lbo;
+
+	if (!run_lookup_entry(run, vbo >> cluster_bits, &lcn, NULL, NULL))
+		return ERR_PTR(-ENOENT);
+
+	lbo = ((u64)lcn << cluster_bits) + (vbo & sbi->cluster_mask);
+
+	return ntfs_bread(sb, lbo >> sb->s_blocksize_bits);
+}
+
+int ntfs_read_run_nb(struct ntfs_sb_info *sbi, const struct runs_tree *run,
+		     u64 vbo, void *buf, u32 bytes, struct ntfs_buffers *nb)
+{
+	int err;
+	struct super_block *sb = sbi->sb;
+	u32 blocksize = sb->s_blocksize;
+	u8 cluster_bits = sbi->cluster_bits;
+	u32 off = vbo & sbi->cluster_mask;
+	u32 nbh = 0;
+	CLST lcn, clen;
+	u64 lbo, len;
+	size_t idx;
+	struct buffer_head *bh;
+
+	if (!run) {
+		/* first reading of $Volume + $MFTMirr + $LogFile goes here*/
+		if (vbo > MFT_REC_VOL * sbi->record_size) {
+			err = -ENOENT;
+			goto out;
+		}
+
+		/* use absolute boot's 'MFTCluster' to read record */
+		lbo = vbo + sbi->mft.lbo;
+		len = sbi->record_size;
+	} else if (!run_lookup_entry(run, vbo >> cluster_bits, &lcn, &clen,
+				     &idx)) {
+		err = -ENOENT;
+		goto out;
+	} else {
+		if (lcn == SPARSE_LCN) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		lbo = ((u64)lcn << cluster_bits) + off;
+		len = ((u64)clen << cluster_bits) - off;
+	}
+
+	off = lbo & (blocksize - 1);
+	if (nb) {
+		nb->off = off;
+		nb->bytes = bytes;
+	}
+
+	for (;;) {
+		u32 len32 = len >= bytes ? bytes : len;
+		sector_t block = lbo >> sb->s_blocksize_bits;
+
+		do {
+			u32 op = blocksize - off;
+
+			if (op > len32)
+				op = len32;
+
+			bh = ntfs_bread(sb, block);
+			if (!bh) {
+				err = -EIO;
+				goto out;
+			}
+
+			if (buf) {
+				memcpy(buf, bh->b_data + off, op);
+				buf = Add2Ptr(buf, op);
+			}
+
+			if (!nb) {
+				put_bh(bh);
+			} else if (nbh >= ARRAY_SIZE(nb->bh)) {
+				err = -EINVAL;
+				goto out;
+			} else {
+				nb->bh[nbh++] = bh;
+				nb->nbufs = nbh;
+			}
+
+			bytes -= op;
+			if (!bytes)
+				return 0;
+			len32 -= op;
+			block += 1;
+			off = 0;
+
+		} while (len32);
+
+		if (!run_get_entry(run, ++idx, NULL, &lcn, &clen)) {
+			err = -ENOENT;
+			goto out;
+		}
+
+		if (lcn == SPARSE_LCN) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		lbo = ((u64)lcn << cluster_bits);
+		len = ((u64)clen << cluster_bits);
+	}
+
+out:
+	if (!nbh)
+		return err;
+
+	while (nbh) {
+		put_bh(nb->bh[--nbh]);
+		nb->bh[nbh] = NULL;
+	}
+
+	nb->nbufs = 0;
+	return err;
+}
+
+/* Returns < 0 if error, 0 if ok, '-E_NTFS_FIXUP' if need to update fixups */
+int ntfs_read_bh(struct ntfs_sb_info *sbi, const struct runs_tree *run, u64 vbo,
+		 struct NTFS_RECORD_HEADER *rhdr, u32 bytes,
+		 struct ntfs_buffers *nb)
+{
+	int err = ntfs_read_run_nb(sbi, run, vbo, rhdr, bytes, nb);
+
+	if (err)
+		return err;
+	return ntfs_fix_post_read(rhdr, nb->bytes, true);
+}
+
+int ntfs_get_bh(struct ntfs_sb_info *sbi, const struct runs_tree *run, u64 vbo,
+		u32 bytes, struct ntfs_buffers *nb)
+{
+	int err = 0;
+	struct super_block *sb = sbi->sb;
+	u32 blocksize = sb->s_blocksize;
+	u8 cluster_bits = sbi->cluster_bits;
+	u32 off;
+	u32 nbh = 0;
+	CLST lcn, clen;
+	u64 lbo, len;
+	size_t idx;
+
+	nb->bytes = bytes;
+
+	if (!run_lookup_entry(run, vbo >> cluster_bits, &lcn, &clen, &idx)) {
+		err = -ENOENT;
+		goto out;
+	}
+
+	off = vbo & sbi->cluster_mask;
+	lbo = ((u64)lcn << cluster_bits) + off;
+	len = ((u64)clen << cluster_bits) - off;
+
+	nb->off = off = lbo & (blocksize - 1);
+
+	for (;;) {
+		u32 len32 = len < bytes ? len : bytes;
+		sector_t block = lbo >> sb->s_blocksize_bits;
+
+		do {
+			u32 op;
+			struct buffer_head *bh;
+
+			if (nbh >= ARRAY_SIZE(nb->bh)) {
+				err = -EINVAL;
+				goto out;
+			}
+
+			op = blocksize - off;
+			if (op > len32)
+				op = len32;
+
+			if (op == blocksize) {
+				bh = sb_getblk(sb, block);
+				if (!bh) {
+					err = -ENOMEM;
+					goto out;
+				}
+				if (buffer_locked(bh))
+					__wait_on_buffer(bh);
+				set_buffer_uptodate(bh);
+			} else {
+				bh = ntfs_bread(sb, block);
+				if (!bh) {
+					err = -EIO;
+					goto out;
+				}
+			}
+
+			nb->bh[nbh++] = bh;
+			bytes -= op;
+			if (!bytes) {
+				nb->nbufs = nbh;
+				return 0;
+			}
+
+			block += 1;
+			len32 -= op;
+			off = 0;
+		} while (len32);
+
+		if (!run_get_entry(run, ++idx, NULL, &lcn, &clen)) {
+			err = -ENOENT;
+			goto out;
+		}
+
+		lbo = ((u64)lcn << cluster_bits);
+		len = ((u64)clen << cluster_bits);
+	}
+
+out:
+	while (nbh) {
+		put_bh(nb->bh[--nbh]);
+		nb->bh[nbh] = NULL;
+	}
+
+	nb->nbufs = 0;
+
+	return err;
+}
+
+int ntfs_write_bh(struct ntfs_sb_info *sbi, struct NTFS_RECORD_HEADER *rhdr,
+		  struct ntfs_buffers *nb, int sync)
+{
+	int err = 0;
+	struct super_block *sb = sbi->sb;
+	u32 block_size = sb->s_blocksize;
+	u32 bytes = nb->bytes;
+	u32 off = nb->off;
+	u16 fo = le16_to_cpu(rhdr->fix_off);
+	u16 fn = le16_to_cpu(rhdr->fix_num);
+	u32 idx;
+	__le16 *fixup;
+	__le16 sample;
+
+	if ((fo & 1) || fo + fn * sizeof(short) > SECTOR_SIZE || !fn-- ||
+	    fn * SECTOR_SIZE > bytes) {
+		return -EINVAL;
+	}
+
+	for (idx = 0; bytes && idx < nb->nbufs; idx += 1, off = 0) {
+		u32 op = block_size - off;
+		char *bh_data;
+		struct buffer_head *bh = nb->bh[idx];
+		__le16 *ptr, *end_data;
+
+		if (op > bytes)
+			op = bytes;
+
+		if (buffer_locked(bh))
+			__wait_on_buffer(bh);
+
+		lock_buffer(nb->bh[idx]);
+
+		bh_data = bh->b_data + off;
+		end_data = Add2Ptr(bh_data, op);
+		memcpy(bh_data, rhdr, op);
+
+		if (!idx) {
+			u16 t16;
+
+			fixup = Add2Ptr(bh_data, fo);
+			sample = *fixup;
+			t16 = le16_to_cpu(sample);
+			if (t16 >= 0x7FFF) {
+				sample = *fixup = cpu_to_le16(1);
+			} else {
+				sample = cpu_to_le16(t16 + 1);
+				*fixup = sample;
+			}
+
+			*(__le16 *)Add2Ptr(rhdr, fo) = sample;
+		}
+
+		ptr = Add2Ptr(bh_data, SECTOR_SIZE - sizeof(short));
+
+		do {
+			*++fixup = *ptr;
+			*ptr = sample;
+			ptr += SECTOR_SIZE / sizeof(short);
+		} while (ptr < end_data);
+
+		set_buffer_uptodate(bh);
+		mark_buffer_dirty(bh);
+		unlock_buffer(bh);
+
+		if (sync) {
+			int err2 = sync_dirty_buffer(bh);
+
+			if (!err && err2)
+				err = err2;
+		}
+
+		bytes -= op;
+		rhdr = Add2Ptr(rhdr, op);
+	}
+
+	return err;
+}
+
+static inline struct bio *ntfs_alloc_bio(u32 nr_vecs)
+{
+	struct bio *bio = bio_alloc(GFP_NOFS | __GFP_HIGH, nr_vecs);
+
+	if (!bio && (current->flags & PF_MEMALLOC)) {
+		while (!bio && (nr_vecs /= 2))
+			bio = bio_alloc(GFP_NOFS | __GFP_HIGH, nr_vecs);
+	}
+	return bio;
+}
+
+/* read/write pages from/to disk*/
+int ntfs_bio_pages(struct ntfs_sb_info *sbi, const struct runs_tree *run,
+		   struct page **pages, u32 nr_pages, u64 vbo, u32 bytes,
+		   u32 op)
+{
+	int err = 0;
+	struct bio *new, *bio = NULL;
+	struct super_block *sb = sbi->sb;
+	struct block_device *bdev = sb->s_bdev;
+	struct page *page;
+	u8 cluster_bits = sbi->cluster_bits;
+	CLST lcn, clen, vcn = vbo >> cluster_bits;
+	u32 add, off, page_idx;
+	u64 lbo, len;
+	size_t run_idx;
+	struct blk_plug plug;
+
+	if (!bytes)
+		return 0;
+
+	blk_start_plug(&plug);
+
+	if (!run_lookup_entry(run, vcn, &lcn, &clen, &run_idx)) {
+		err = -ENOENT;
+		goto out;
+	}
+	off = vbo & sbi->cluster_mask;
+	page_idx = 0;
+	page = pages[0];
+
+	for (;;) {
+		lbo = ((u64)lcn << cluster_bits) + off;
+		len = ((u64)clen << cluster_bits) - off;
+new_bio:
+		new = ntfs_alloc_bio(nr_pages - page_idx);
+		if (!new) {
+			err = -ENOMEM;
+			goto out;
+		}
+		if (bio) {
+			bio_chain(bio, new);
+			submit_bio(bio);
+		}
+		bio = new;
+		bio_set_dev(bio, bdev);
+		bio->bi_iter.bi_sector = lbo >> 9;
+		bio->bi_opf = op;
+
+		while (len) {
+			off = vbo & (PAGE_SIZE - 1);
+			add = off + len > PAGE_SIZE ? (PAGE_SIZE - off) : len;
+
+			if (bio_add_page(bio, page, add, off) < add)
+				goto new_bio;
+
+			if (bytes <= add)
+				goto out;
+			bytes -= add;
+			vbo += add;
+
+			if (add + off == PAGE_SIZE) {
+				page_idx += 1;
+				if (page_idx >= nr_pages) {
+					WARN_ON(1);
+					err = -EINVAL;
+					goto out;
+				}
+				page = pages[page_idx];
+			}
+
+			if (len <= add)
+				break;
+			len -= add;
+			lbo += add;
+		}
+
+		if (!run_get_entry(run, ++run_idx, &vcn, &lcn, &clen)) {
+			err = -ENOENT;
+			goto out;
+		}
+		off = 0;
+	}
+out:
+	if (bio) {
+		if (!err)
+			err = submit_bio_wait(bio);
+		bio_put(bio);
+	}
+	blk_finish_plug(&plug);
+
+	return err;
+}
+
+/*
+ * Helper for ntfs_loadlog_and_replay
+ * fill on-disk logfile range by (-1)
+ * this means empty logfile
+ */
+int ntfs_bio_fill_1(struct ntfs_sb_info *sbi, const struct runs_tree *run)
+{
+	int err = 0;
+	struct super_block *sb = sbi->sb;
+	struct block_device *bdev = sb->s_bdev;
+	u8 cluster_bits = sbi->cluster_bits;
+	struct bio *new, *bio = NULL;
+	CLST lcn, clen;
+	u64 lbo, len;
+	size_t run_idx;
+	struct page *fill;
+	void *kaddr;
+	struct blk_plug plug;
+
+	fill = alloc_page(GFP_KERNEL);
+	if (!fill)
+		return -ENOMEM;
+
+	kaddr = kmap_atomic(fill);
+	memset(kaddr, -1, PAGE_SIZE);
+	kunmap_atomic(kaddr);
+	flush_dcache_page(fill);
+	lock_page(fill);
+
+	if (!run_lookup_entry(run, 0, &lcn, &clen, &run_idx)) {
+		err = -ENOENT;
+		goto out;
+	}
+
+	/*
+	 * TODO: try blkdev_issue_write_same
+	 */
+	blk_start_plug(&plug);
+	do {
+		lbo = (u64)lcn << cluster_bits;
+		len = (u64)clen << cluster_bits;
+new_bio:
+		new = ntfs_alloc_bio(BIO_MAX_PAGES);
+		if (!new) {
+			err = -ENOMEM;
+			break;
+		}
+		if (bio) {
+			bio_chain(bio, new);
+			submit_bio(bio);
+		}
+		bio = new;
+		bio_set_dev(bio, bdev);
+		bio->bi_opf = REQ_OP_WRITE;
+		bio->bi_iter.bi_sector = lbo >> 9;
+
+		for (;;) {
+			u32 add = len > PAGE_SIZE ? PAGE_SIZE : len;
+
+			if (bio_add_page(bio, fill, add, 0) < add)
+				goto new_bio;
+
+			lbo += add;
+			if (len <= add)
+				break;
+			len -= add;
+		}
+	} while (run_get_entry(run, ++run_idx, NULL, &lcn, &clen));
+
+	if (bio) {
+		if (!err)
+			err = submit_bio_wait(bio);
+		bio_put(bio);
+	}
+	blk_finish_plug(&plug);
+out:
+	unlock_page(fill);
+	put_page(fill);
+
+	return err;
+}
+
+int ntfs_vbo_to_lbo(struct ntfs_sb_info *sbi, const struct runs_tree *run,
+		    u64 vbo, u64 *lbo, u64 *bytes)
+{
+	u32 off;
+	CLST lcn, len;
+	u8 cluster_bits = sbi->cluster_bits;
+
+	if (!run_lookup_entry(run, vbo >> cluster_bits, &lcn, &len, NULL))
+		return -ENOENT;
+
+	off = vbo & sbi->cluster_mask;
+	*lbo = lcn == SPARSE_LCN ? -1 : (((u64)lcn << cluster_bits) + off);
+	*bytes = ((u64)len << cluster_bits) - off;
+
+	return 0;
+}
+
+struct ntfs_inode *ntfs_new_inode(struct ntfs_sb_info *sbi, CLST rno, bool dir)
+{
+	int err = 0;
+	struct super_block *sb = sbi->sb;
+	struct inode *inode = new_inode(sb);
+	struct ntfs_inode *ni;
+
+	if (!inode)
+		return ERR_PTR(-ENOMEM);
+
+	ni = ntfs_i(inode);
+
+	err = mi_format_new(&ni->mi, sbi, rno, dir ? RECORD_FLAG_DIR : 0,
+			    false);
+	if (err)
+		goto out;
+
+	inode->i_ino = rno;
+	if (insert_inode_locked(inode) < 0) {
+		err = -EIO;
+		goto out;
+	}
+
+out:
+	if (err) {
+		iput(inode);
+		ni = ERR_PTR(err);
+	}
+	return ni;
+}
+
+/*
+ * O:BAG:BAD:(A;OICI;FA;;;WD)
+ * owner S-1-5-32-544 (Administrators)
+ * group S-1-5-32-544 (Administrators)
+ * ACE: allow S-1-1-0 (Everyone) with FILE_ALL_ACCESS
+ */
+const u8 s_default_security[] __aligned(8) = {
+	0x01, 0x00, 0x04, 0x80, 0x30, 0x00, 0x00, 0x00, 0x40, 0x00, 0x00, 0x00,
+	0x00, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x02, 0x00, 0x1C, 0x00,
+	0x01, 0x00, 0x00, 0x00, 0x00, 0x03, 0x14, 0x00, 0xFF, 0x01, 0x1F, 0x00,
+	0x01, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00,
+	0x01, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x20, 0x00, 0x00, 0x00,
+	0x20, 0x02, 0x00, 0x00, 0x01, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05,
+	0x20, 0x00, 0x00, 0x00, 0x20, 0x02, 0x00, 0x00,
+};
+
+static_assert(sizeof(s_default_security) == 0x50);
+
+static inline u32 sid_length(const struct SID *sid)
+{
+	return offsetof(struct SID, SubAuthority[0]) +
+	       (sid->SubAuthorityCount * sizeof(u32));
+}
+
+/*
+ * Thanks Mark Harmstone for idea
+ */
+static bool is_acl_valid(const struct ACL *acl, u32 len)
+{
+	const struct ACE_HEADER *ace;
+	u32 i;
+	u16 ace_count, ace_size;
+
+	if (acl->AclRevision != ACL_REVISION)
+		return false;
+
+	if (acl->Sbz1)
+		return false;
+
+	if (le16_to_cpu(acl->AclSize) > len)
+		return false;
+
+	if (acl->Sbz2)
+		return false;
+
+	len -= sizeof(struct ACL);
+
+	ace = (struct ACE_HEADER *)&acl[1];
+	ace_count = le16_to_cpu(acl->AceCount);
+
+	for (i = 0; i < ace_count; i++) {
+		if (len < sizeof(struct ACE_HEADER))
+			return false;
+
+		ace_size = le16_to_cpu(ace->AceSize);
+		if (len < ace_size)
+			return false;
+
+		len -= ace_size;
+
+		ace = Add2Ptr(ace, ace_size);
+	}
+
+	return true;
+}
+
+bool is_sd_valid(const struct SECURITY_DESCRIPTOR_RELATIVE *sd, u32 len)
+{
+	u32 sd_owner, sd_group, sd_sacl, sd_dacl;
+
+	if (len < sizeof(struct SECURITY_DESCRIPTOR_RELATIVE))
+		return false;
+
+	if (sd->Revision != 1)
+		return false;
+
+	if (sd->Sbz1)
+		return false;
+
+	if (!(sd->Control & cpu_to_le16(SE_SELF_RELATIVE)))
+		return false;
+
+	sd_owner = le32_to_cpu(sd->Owner);
+	if (sd_owner) {
+		const struct SID *owner = Add2Ptr(sd, sd_owner);
+
+		if (sd_owner + offsetof(struct SID, SubAuthority) > len)
+			return false;
+
+		if (owner->Revision != 1)
+			return false;
+
+		if (sd_owner + sid_length(owner) > len)
+			return false;
+	}
+
+	sd_group = le32_to_cpu(sd->Group);
+	if (sd_group) {
+		const struct SID *group = Add2Ptr(sd, sd_group);
+
+		if (sd_group + offsetof(struct SID, SubAuthority) > len)
+			return false;
+
+		if (group->Revision != 1)
+			return false;
+
+		if (sd_group + sid_length(group) > len)
+			return false;
+	}
+
+	sd_sacl = le32_to_cpu(sd->Sacl);
+	if (sd_sacl) {
+		const struct ACL *sacl = Add2Ptr(sd, sd_sacl);
+
+		if (sd_sacl + sizeof(struct ACL) > len)
+			return false;
+
+		if (!is_acl_valid(sacl, len - sd_sacl))
+			return false;
+	}
+
+	sd_dacl = le32_to_cpu(sd->Dacl);
+	if (sd_dacl) {
+		const struct ACL *dacl = Add2Ptr(sd, sd_dacl);
+
+		if (sd_dacl + sizeof(struct ACL) > len)
+			return false;
+
+		if (!is_acl_valid(dacl, len - sd_dacl))
+			return false;
+	}
+
+	return true;
+}
+
+/*
+ * ntfs_security_init
+ *
+ * loads and parse $Secure
+ */
+int ntfs_security_init(struct ntfs_sb_info *sbi)
+{
+	int err;
+	struct super_block *sb = sbi->sb;
+	struct inode *inode;
+	struct ntfs_inode *ni;
+	struct MFT_REF ref;
+	struct ATTRIB *attr;
+	struct ATTR_LIST_ENTRY *le;
+	u64 sds_size;
+	size_t cnt, off;
+	struct NTFS_DE *ne;
+	struct NTFS_DE_SII *sii_e;
+	struct ntfs_fnd *fnd_sii = NULL;
+	const struct INDEX_ROOT *root_sii;
+	const struct INDEX_ROOT *root_sdh;
+	struct ntfs_index *indx_sdh = &sbi->security.index_sdh;
+	struct ntfs_index *indx_sii = &sbi->security.index_sii;
+
+	ref.low = cpu_to_le32(MFT_REC_SECURE);
+	ref.high = 0;
+	ref.seq = cpu_to_le16(MFT_REC_SECURE);
+
+	inode = ntfs_iget5(sb, &ref, &NAME_SECURE);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load $Secure.");
+		inode = NULL;
+		goto out;
+	}
+
+	ni = ntfs_i(inode);
+
+	le = NULL;
+
+	attr = ni_find_attr(ni, NULL, &le, ATTR_ROOT, SDH_NAME,
+			    ARRAY_SIZE(SDH_NAME), NULL, NULL);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	root_sdh = resident_data(attr);
+	if (root_sdh->type != ATTR_ZERO ||
+	    root_sdh->rule != NTFS_COLLATION_TYPE_SECURITY_HASH) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	err = indx_init(indx_sdh, sbi, attr, INDEX_MUTEX_SDH);
+	if (err)
+		goto out;
+
+	attr = ni_find_attr(ni, attr, &le, ATTR_ROOT, SII_NAME,
+			    ARRAY_SIZE(SII_NAME), NULL, NULL);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	root_sii = resident_data(attr);
+	if (root_sii->type != ATTR_ZERO ||
+	    root_sii->rule != NTFS_COLLATION_TYPE_UINT) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	err = indx_init(indx_sii, sbi, attr, INDEX_MUTEX_SII);
+	if (err)
+		goto out;
+
+	fnd_sii = fnd_get(indx_sii);
+	if (!fnd_sii) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	sds_size = inode->i_size;
+
+	/* Find the last valid Id */
+	sbi->security.next_id = SECURITY_ID_FIRST;
+	/* Always write new security at the end of bucket */
+	sbi->security.next_off =
+		Quad2Align(sds_size - SecurityDescriptorsBlockSize);
+
+	cnt = 0;
+	off = 0;
+	ne = NULL;
+
+	for (;;) {
+		u32 next_id;
+
+		err = indx_find_raw(indx_sii, ni, root_sii, &ne, &off, fnd_sii);
+		if (err || !ne)
+			break;
+
+		sii_e = (struct NTFS_DE_SII *)ne;
+		if (le16_to_cpu(ne->view.data_size) < SIZEOF_SECURITY_HDR)
+			continue;
+
+		next_id = le32_to_cpu(sii_e->sec_id) + 1;
+		if (next_id >= sbi->security.next_id)
+			sbi->security.next_id = next_id;
+
+		cnt += 1;
+	}
+
+	sbi->security.ni = ni;
+	inode = NULL;
+out:
+	iput(inode);
+	fnd_put(fnd_sii);
+
+	return err;
+}
+
+/*
+ * ntfs_get_security_by_id
+ *
+ * reads security descriptor by id
+ */
+int ntfs_get_security_by_id(struct ntfs_sb_info *sbi, __le32 security_id,
+			    struct SECURITY_DESCRIPTOR_RELATIVE **sd,
+			    size_t *size)
+{
+	int err;
+	int diff;
+	struct ntfs_inode *ni = sbi->security.ni;
+	struct ntfs_index *indx = &sbi->security.index_sii;
+	void *p = NULL;
+	struct NTFS_DE_SII *sii_e;
+	struct ntfs_fnd *fnd_sii;
+	struct SECURITY_HDR d_security;
+	const struct INDEX_ROOT *root_sii;
+	u32 t32;
+
+	*sd = NULL;
+
+	inode_lock_shared(&ni->vfs_inode);
+
+	fnd_sii = fnd_get(indx);
+	if (!fnd_sii) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	root_sii = indx_get_root(indx, ni, NULL, NULL);
+	if (!root_sii) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* Try to find this SECURITY descriptor in SII indexes */
+	err = indx_find(indx, ni, root_sii, &security_id, sizeof(security_id),
+			NULL, &diff, (struct NTFS_DE **)&sii_e, fnd_sii);
+	if (err)
+		goto out;
+
+	if (diff)
+		goto out;
+
+	t32 = le32_to_cpu(sii_e->sec_hdr.size);
+	if (t32 < SIZEOF_SECURITY_HDR) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (t32 > SIZEOF_SECURITY_HDR + 0x10000) {
+		/*
+		 * looks like too big security. 0x10000 - is arbitrary big number
+		 */
+		err = -EFBIG;
+		goto out;
+	}
+
+	*size = t32 - SIZEOF_SECURITY_HDR;
+
+	p = ntfs_alloc(*size, 0);
+	if (!p) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	err = ntfs_read_run_nb(sbi, &ni->file.run,
+			       le64_to_cpu(sii_e->sec_hdr.off), &d_security,
+			       sizeof(d_security), NULL);
+	if (err)
+		goto out;
+
+	if (memcmp(&d_security, &sii_e->sec_hdr, SIZEOF_SECURITY_HDR)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	err = ntfs_read_run_nb(sbi, &ni->file.run,
+			       le64_to_cpu(sii_e->sec_hdr.off) +
+				       SIZEOF_SECURITY_HDR,
+			       p, *size, NULL);
+	if (err)
+		goto out;
+
+	*sd = p;
+	p = NULL;
+
+out:
+	ntfs_free(p);
+	fnd_put(fnd_sii);
+	inode_unlock_shared(&ni->vfs_inode);
+
+	return err;
+}
+
+/*
+ * ntfs_insert_security
+ *
+ * inserts security descriptor into $Secure::SDS
+ *
+ * SECURITY Descriptor Stream data is organized into chunks of 256K bytes
+ * and it contains a mirror copy of each security descriptor.  When writing
+ * to a security descriptor at location X, another copy will be written at
+ * location (X+256K).
+ * When writing a security descriptor that will cross the 256K boundary,
+ * the pointer will be advanced by 256K to skip
+ * over the mirror portion.
+ */
+int ntfs_insert_security(struct ntfs_sb_info *sbi,
+			 const struct SECURITY_DESCRIPTOR_RELATIVE *sd,
+			 u32 size_sd, __le32 *security_id, bool *inserted)
+{
+	int err, diff;
+	struct ntfs_inode *ni = sbi->security.ni;
+	struct ntfs_index *indx_sdh = &sbi->security.index_sdh;
+	struct ntfs_index *indx_sii = &sbi->security.index_sii;
+	struct NTFS_DE_SDH *e;
+	struct NTFS_DE_SDH sdh_e;
+	struct NTFS_DE_SII sii_e;
+	struct SECURITY_HDR *d_security;
+	u32 new_sec_size = size_sd + SIZEOF_SECURITY_HDR;
+	u32 aligned_sec_size = Quad2Align(new_sec_size);
+	struct SECURITY_KEY hash_key;
+	struct ntfs_fnd *fnd_sdh = NULL;
+	const struct INDEX_ROOT *root_sdh;
+	const struct INDEX_ROOT *root_sii;
+	u64 mirr_off, new_sds_size;
+	u32 next, left;
+
+	static_assert((1 << Log2OfSecurityDescriptorsBlockSize) ==
+		      SecurityDescriptorsBlockSize);
+
+	hash_key.hash = security_hash(sd, size_sd);
+	hash_key.sec_id = SECURITY_ID_INVALID;
+
+	if (inserted)
+		*inserted = false;
+	*security_id = SECURITY_ID_INVALID;
+
+	/* Allocate a temporal buffer*/
+	d_security = ntfs_alloc(aligned_sec_size, 1);
+	if (!d_security)
+		return -ENOMEM;
+
+	inode_lock(&ni->vfs_inode);
+
+	fnd_sdh = fnd_get(indx_sdh);
+	if (!fnd_sdh) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	root_sdh = indx_get_root(indx_sdh, ni, NULL, NULL);
+	if (!root_sdh) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	root_sii = indx_get_root(indx_sii, ni, NULL, NULL);
+	if (!root_sii) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/*
+	 * Check if such security already exists
+	 * use "SDH" and hash -> to get the offset in "SDS"
+	 */
+	err = indx_find(indx_sdh, ni, root_sdh, &hash_key, sizeof(hash_key),
+			&d_security->key.sec_id, &diff, (struct NTFS_DE **)&e,
+			fnd_sdh);
+	if (err)
+		goto out;
+
+	while (e) {
+		if (le32_to_cpu(e->sec_hdr.size) == new_sec_size) {
+			err = ntfs_read_run_nb(sbi, &ni->file.run,
+					       le64_to_cpu(e->sec_hdr.off),
+					       d_security, new_sec_size, NULL);
+			if (err)
+				goto out;
+
+			if (le32_to_cpu(d_security->size) == new_sec_size &&
+			    d_security->key.hash == hash_key.hash &&
+			    !memcmp(d_security + 1, sd, size_sd)) {
+				*security_id = d_security->key.sec_id;
+				/*such security already exists*/
+				err = 0;
+				goto out;
+			}
+		}
+
+		err = indx_find_sort(indx_sdh, ni, root_sdh,
+				     (struct NTFS_DE **)&e, fnd_sdh);
+		if (err)
+			goto out;
+
+		if (!e || e->key.hash != hash_key.hash)
+			break;
+	}
+
+	/* Zero unused space */
+	next = sbi->security.next_off & (SecurityDescriptorsBlockSize - 1);
+	left = SecurityDescriptorsBlockSize - next;
+
+	/* Zero gap until SecurityDescriptorsBlockSize */
+	if (left < new_sec_size) {
+		/* zero "left" bytes from sbi->security.next_off */
+		sbi->security.next_off += SecurityDescriptorsBlockSize + left;
+	}
+
+	/* Zero tail of previous security */
+	//used = ni->vfs_inode.i_size & (SecurityDescriptorsBlockSize - 1);
+
+	/*
+	 * Example:
+	 * 0x40438 == ni->vfs_inode.i_size
+	 * 0x00440 == sbi->security.next_off
+	 * need to zero [0x438-0x440)
+	 * if (next > used) {
+	 *  u32 tozero = next - used;
+	 *  zero "tozero" bytes from sbi->security.next_off - tozero
+	 */
+
+	/* format new security descriptor */
+	d_security->key.hash = hash_key.hash;
+	d_security->key.sec_id = cpu_to_le32(sbi->security.next_id);
+	d_security->off = cpu_to_le64(sbi->security.next_off);
+	d_security->size = cpu_to_le32(new_sec_size);
+	memcpy(d_security + 1, sd, size_sd);
+
+	/* Write main SDS bucket */
+	err = ntfs_sb_write_run(sbi, &ni->file.run, sbi->security.next_off,
+				d_security, aligned_sec_size);
+
+	if (err)
+		goto out;
+
+	mirr_off = sbi->security.next_off + SecurityDescriptorsBlockSize;
+	new_sds_size = mirr_off + aligned_sec_size;
+
+	if (new_sds_size > ni->vfs_inode.i_size) {
+		err = attr_set_size(ni, ATTR_DATA, SDS_NAME,
+				    ARRAY_SIZE(SDS_NAME), &ni->file.run,
+				    new_sds_size, &new_sds_size, false, NULL);
+		if (err)
+			goto out;
+	}
+
+	/* Write copy SDS bucket */
+	err = ntfs_sb_write_run(sbi, &ni->file.run, mirr_off, d_security,
+				aligned_sec_size);
+	if (err)
+		goto out;
+
+	/* Fill SII entry */
+	sii_e.de.view.data_off =
+		cpu_to_le16(offsetof(struct NTFS_DE_SII, sec_hdr));
+	sii_e.de.view.data_size = cpu_to_le16(SIZEOF_SECURITY_HDR);
+	sii_e.de.view.res = 0;
+	sii_e.de.size = cpu_to_le16(SIZEOF_SII_DIRENTRY);
+	sii_e.de.key_size = cpu_to_le16(sizeof(d_security->key.sec_id));
+	sii_e.de.flags = 0;
+	sii_e.de.res = 0;
+	sii_e.sec_id = d_security->key.sec_id;
+	memcpy(&sii_e.sec_hdr, d_security, SIZEOF_SECURITY_HDR);
+
+	err = indx_insert_entry(indx_sii, ni, &sii_e.de, NULL, NULL);
+	if (err)
+		goto out;
+
+	/* Fill SDH entry */
+	sdh_e.de.view.data_off =
+		cpu_to_le16(offsetof(struct NTFS_DE_SDH, sec_hdr));
+	sdh_e.de.view.data_size = cpu_to_le16(SIZEOF_SECURITY_HDR);
+	sdh_e.de.view.res = 0;
+	sdh_e.de.size = cpu_to_le16(SIZEOF_SDH_DIRENTRY);
+	sdh_e.de.key_size = cpu_to_le16(sizeof(sdh_e.key));
+	sdh_e.de.flags = 0;
+	sdh_e.de.res = 0;
+	sdh_e.key.hash = d_security->key.hash;
+	sdh_e.key.sec_id = d_security->key.sec_id;
+	memcpy(&sdh_e.sec_hdr, d_security, SIZEOF_SECURITY_HDR);
+	sdh_e.magic[0] = cpu_to_le16('I');
+	sdh_e.magic[1] = cpu_to_le16('I');
+
+	fnd_clear(fnd_sdh);
+	err = indx_insert_entry(indx_sdh, ni, &sdh_e.de, (void *)(size_t)1,
+				fnd_sdh);
+	if (err)
+		goto out;
+
+	*security_id = d_security->key.sec_id;
+	if (inserted)
+		*inserted = true;
+
+	/* Update Id and offset for next descriptor */
+	sbi->security.next_id += 1;
+	sbi->security.next_off += aligned_sec_size;
+
+out:
+	fnd_put(fnd_sdh);
+	mark_inode_dirty(&ni->vfs_inode);
+	inode_unlock(&ni->vfs_inode);
+	ntfs_free(d_security);
+
+	return err;
+}
+
+/*
+ * ntfs_reparse_init
+ *
+ * loads and parse $Extend/$Reparse
+ */
+int ntfs_reparse_init(struct ntfs_sb_info *sbi)
+{
+	int err;
+	struct ntfs_inode *ni = sbi->reparse.ni;
+	struct ntfs_index *indx = &sbi->reparse.index_r;
+	struct ATTRIB *attr;
+	struct ATTR_LIST_ENTRY *le;
+	const struct INDEX_ROOT *root_r;
+
+	if (!ni)
+		return 0;
+
+	le = NULL;
+	attr = ni_find_attr(ni, NULL, &le, ATTR_ROOT, SR_NAME,
+			    ARRAY_SIZE(SR_NAME), NULL, NULL);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	root_r = resident_data(attr);
+	if (root_r->type != ATTR_ZERO ||
+	    root_r->rule != NTFS_COLLATION_TYPE_UINTS) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	err = indx_init(indx, sbi, attr, INDEX_MUTEX_SR);
+	if (err)
+		goto out;
+
+out:
+	return err;
+}
+
+/*
+ * ntfs_objid_init
+ *
+ * loads and parse $Extend/$ObjId
+ */
+int ntfs_objid_init(struct ntfs_sb_info *sbi)
+{
+	int err;
+	struct ntfs_inode *ni = sbi->objid.ni;
+	struct ntfs_index *indx = &sbi->objid.index_o;
+	struct ATTRIB *attr;
+	struct ATTR_LIST_ENTRY *le;
+	const struct INDEX_ROOT *root;
+
+	if (!ni)
+		return 0;
+
+	le = NULL;
+	attr = ni_find_attr(ni, NULL, &le, ATTR_ROOT, SO_NAME,
+			    ARRAY_SIZE(SO_NAME), NULL, NULL);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	root = resident_data(attr);
+	if (root->type != ATTR_ZERO ||
+	    root->rule != NTFS_COLLATION_TYPE_UINTS) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	err = indx_init(indx, sbi, attr, INDEX_MUTEX_SO);
+	if (err)
+		goto out;
+
+out:
+	return err;
+}
+
+int ntfs_objid_remove(struct ntfs_sb_info *sbi, struct GUID *guid)
+{
+	int err;
+	struct ntfs_inode *ni = sbi->objid.ni;
+	struct ntfs_index *indx = &sbi->objid.index_o;
+
+	if (!ni)
+		return -EINVAL;
+
+	inode_lock(&ni->vfs_inode);
+
+	err = indx_delete_entry(indx, ni, guid, sizeof(*guid), NULL);
+
+	mark_inode_dirty(&ni->vfs_inode);
+	inode_unlock(&ni->vfs_inode);
+
+	return err;
+}
+
+int ntfs_insert_reparse(struct ntfs_sb_info *sbi, __le32 rtag,
+			const struct MFT_REF *ref)
+{
+	int err;
+	struct ntfs_inode *ni = sbi->reparse.ni;
+	struct ntfs_index *indx = &sbi->reparse.index_r;
+	struct NTFS_DE_R re;
+
+	if (!ni)
+		return -EINVAL;
+
+	memset(&re, 0, sizeof(re));
+
+	re.de.view.data_off = cpu_to_le16(SIZEOF_R_DIRENTRY);
+	re.de.size = cpu_to_le16(QuadAlign(SIZEOF_R_DIRENTRY));
+	re.de.key_size = cpu_to_le16(sizeof(re.key));
+
+	re.key.ReparseTag = rtag;
+	memcpy(&re.key.ref, ref, sizeof(*ref));
+
+	inode_lock(&ni->vfs_inode);
+
+	err = indx_insert_entry(indx, ni, &re.de, NULL, NULL);
+
+	mark_inode_dirty(&ni->vfs_inode);
+	inode_unlock(&ni->vfs_inode);
+
+	return err;
+}
+
+int ntfs_remove_reparse(struct ntfs_sb_info *sbi, __le32 rtag,
+			const struct MFT_REF *ref)
+{
+	int err, diff;
+	struct ntfs_inode *ni = sbi->reparse.ni;
+	struct ntfs_index *indx = &sbi->reparse.index_r;
+	struct ntfs_fnd *fnd = NULL;
+	struct REPARSE_KEY rkey;
+	struct NTFS_DE_R *re;
+	struct INDEX_ROOT *root_r;
+
+	if (!ni)
+		return -EINVAL;
+
+	rkey.ReparseTag = rtag;
+	rkey.ref = *ref;
+
+	inode_lock(&ni->vfs_inode);
+
+	if (rtag) {
+		err = indx_delete_entry(indx, ni, &rkey, sizeof(rkey), NULL);
+		goto out1;
+	}
+
+	fnd = fnd_get(indx);
+	if (!fnd) {
+		err = -ENOMEM;
+		goto out1;
+	}
+
+	root_r = indx_get_root(indx, ni, NULL, NULL);
+	if (!root_r) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* 1 - forces to ignore rkey.ReparseTag when comparing keys */
+	err = indx_find(indx, ni, root_r, &rkey, sizeof(rkey), (void *)1, &diff,
+			(struct NTFS_DE **)&re, fnd);
+	if (err)
+		goto out;
+
+	if (memcmp(&re->key.ref, ref, sizeof(*ref))) {
+		/* Impossible. Looks like volume corrupt?*/
+		goto out;
+	}
+
+	memcpy(&rkey, &re->key, sizeof(rkey));
+
+	fnd_put(fnd);
+	fnd = NULL;
+
+	err = indx_delete_entry(indx, ni, &rkey, sizeof(rkey), NULL);
+	if (err)
+		goto out;
+
+out:
+	fnd_put(fnd);
+
+out1:
+	mark_inode_dirty(&ni->vfs_inode);
+	inode_unlock(&ni->vfs_inode);
+
+	return err;
+}
+
+static inline void ntfs_unmap_and_discard(struct ntfs_sb_info *sbi, CLST lcn,
+					  CLST len)
+{
+	ntfs_unmap_meta(sbi->sb, lcn, len);
+	ntfs_discard(sbi, lcn, len);
+}
+
+void mark_as_free_ex(struct ntfs_sb_info *sbi, CLST lcn, CLST len, bool trim)
+{
+	CLST end, i;
+	struct wnd_bitmap *wnd = &sbi->used.bitmap;
+
+	down_write_nested(&wnd->rw_lock, BITMAP_MUTEX_CLUSTERS);
+	if (!wnd_is_used(wnd, lcn, len)) {
+		ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
+
+		end = lcn + len;
+		len = 0;
+		for (i = lcn; i < end; i++) {
+			if (wnd_is_used(wnd, i, 1)) {
+				if (!len)
+					lcn = i;
+				len += 1;
+				continue;
+			}
+
+			if (!len)
+				continue;
+
+			if (trim)
+				ntfs_unmap_and_discard(sbi, lcn, len);
+
+			wnd_set_free(wnd, lcn, len);
+			len = 0;
+		}
+
+		if (!len)
+			goto out;
+	}
+
+	if (trim)
+		ntfs_unmap_and_discard(sbi, lcn, len);
+	wnd_set_free(wnd, lcn, len);
+
+out:
+	up_write(&wnd->rw_lock);
+}
+
+/*
+ * run_deallocate
+ *
+ * deallocate clusters
+ */
+int run_deallocate(struct ntfs_sb_info *sbi, struct runs_tree *run, bool trim)
+{
+	CLST lcn, len;
+	size_t idx = 0;
+
+	while (run_get_entry(run, idx++, NULL, &lcn, &len)) {
+		if (lcn == SPARSE_LCN)
+			continue;
+
+		mark_as_free_ex(sbi, lcn, len, trim);
+	}
+
+	return 0;
+}
diff --git a/fs/ntfs3/index.c b/fs/ntfs3/index.c
new file mode 100644
index 000000000000..b7caeb3da500
--- /dev/null
+++ b/fs/ntfs3/index.c
@@ -0,0 +1,2665 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+static const struct INDEX_NAMES {
+	const __le16 *name;
+	u8 name_len;
+} s_index_names[INDEX_MUTEX_TOTAL] = {
+	{ I30_NAME, ARRAY_SIZE(I30_NAME) }, { SII_NAME, ARRAY_SIZE(SII_NAME) },
+	{ SDH_NAME, ARRAY_SIZE(SDH_NAME) }, { SO_NAME, ARRAY_SIZE(SO_NAME) },
+	{ SQ_NAME, ARRAY_SIZE(SQ_NAME) },   { SR_NAME, ARRAY_SIZE(SR_NAME) },
+};
+
+/*
+ * compare two names in index
+ * if l1 != 0
+ *   both names are little endian on-disk ATTR_FILE_NAME structs
+ * else
+ *   f1 - cpu_str, f2 - ATTR_FILE_NAME
+ */
+static int cmp_fnames(const struct ATTR_FILE_NAME *f1, size_t l1,
+		      const struct ATTR_FILE_NAME *f2, size_t l2,
+		      const struct ntfs_sb_info *sbi)
+{
+	int diff;
+	u16 fsize2;
+
+	if (l2 <= offsetof(struct ATTR_FILE_NAME, name))
+		return -1;
+
+	fsize2 = fname_full_size(f2);
+	if (l2 < fsize2)
+		return -1;
+
+	if (!l1) {
+		const struct cpu_str *s1 = (struct cpu_str *)f1;
+		const struct le_str *s2 = (struct le_str *)&f2->name_len;
+
+		diff = ntfs_cmp_names_cpu(s1, s2, sbi->upcase);
+
+		if (diff)
+			return diff;
+
+		/*
+		 * If names are equal (case insensitive)
+		 * try to compare it case sensitive
+		 */
+		if (/*sbi->options.nocase || */ f2->type == FILE_NAME_DOS)
+			return 0;
+
+		return ntfs_cmp_names_cpu(s1, s2, NULL);
+	}
+
+	diff = ntfs_cmp_names(f1->name, f1->name_len, f2->name, f2->name_len,
+			      sbi->upcase);
+
+	if (diff)
+		return diff;
+
+	/*
+	 * If names are equal (case insensitive)
+	 * try to compare it case sensitive
+	 */
+	if (/*sbi->options.nocase || */ f2->type == FILE_NAME_DOS)
+		return 0;
+
+	return ntfs_cmp_names(f1->name, f1->name_len, f2->name, f2->name_len,
+			      NULL);
+}
+
+/* $SII of $Secure and $Q of Quota */
+static int cmp_uint(const u32 *k1, size_t l1, const u32 *k2, size_t l2,
+		    const void *p)
+{
+	if (l2 < sizeof(u32))
+		return -1;
+
+	if (*k1 < *k2)
+		return -1;
+	if (*k1 > *k2)
+		return 1;
+	return 0;
+}
+
+/* $SDH of $Secure */
+static int cmp_sdh(const struct SECURITY_KEY *k1, size_t l1,
+		   const struct SECURITY_KEY *k2, size_t l2, const void *p)
+{
+	u32 t1, t2;
+
+	if (l2 < sizeof(struct SECURITY_KEY))
+		return -1;
+
+	t1 = le32_to_cpu(k1->hash);
+	t2 = le32_to_cpu(k2->hash);
+
+	/* First value is a hash value itself */
+	if (t1 < t2)
+		return -1;
+	if (t1 > t2)
+		return 1;
+
+	/* Second value is security Id */
+	if (p) {
+		t1 = le32_to_cpu(k1->sec_id);
+		t2 = le32_to_cpu(k2->sec_id);
+		if (t1 < t2)
+			return -1;
+		if (t1 > t2)
+			return 1;
+	}
+
+	return 0;
+}
+
+/* $O of ObjId and "$R" for Reparse */
+static int cmp_uints(const __le32 *k1, size_t l1, const __le32 *k2, size_t l2,
+		     const void *p)
+{
+	size_t count;
+
+	if ((size_t)p == 1) {
+		/*
+		 * ni_delete_all -> ntfs_remove_reparse -> delete all with this reference
+		 * k1, k2 - pointers to REPARSE_KEY
+		 */
+
+		k1 += 1; // skip REPARSE_KEY.ReparseTag
+		k2 += 1; // skip REPARSE_KEY.ReparseTag
+		if (l2 <= sizeof(int))
+			return -1;
+		l2 -= sizeof(int);
+		if (l1 <= sizeof(int))
+			return 1;
+		l1 -= sizeof(int);
+	}
+
+	if (l2 < sizeof(int))
+		return -1;
+
+	for (count = min(l1, l2) >> 2; count > 0; --count, ++k1, ++k2) {
+		u32 t1 = le32_to_cpu(*k1);
+		u32 t2 = le32_to_cpu(*k2);
+
+		if (t1 > t2)
+			return 1;
+		if (t1 < t2)
+			return -1;
+	}
+
+	if (l1 > l2)
+		return 1;
+	if (l1 < l2)
+		return -1;
+
+	return 0;
+}
+
+static inline NTFS_CMP_FUNC get_cmp_func(const struct INDEX_ROOT *root)
+{
+	switch (root->type) {
+	case ATTR_NAME:
+		if (root->rule == NTFS_COLLATION_TYPE_FILENAME)
+			return (NTFS_CMP_FUNC)&cmp_fnames;
+		break;
+	case ATTR_ZERO:
+		switch (root->rule) {
+		case NTFS_COLLATION_TYPE_UINT:
+			return (NTFS_CMP_FUNC)&cmp_uint;
+		case NTFS_COLLATION_TYPE_SECURITY_HASH:
+			return (NTFS_CMP_FUNC)&cmp_sdh;
+		case NTFS_COLLATION_TYPE_UINTS:
+			return (NTFS_CMP_FUNC)&cmp_uints;
+		default:
+			break;
+		}
+	default:
+		break;
+	}
+
+	return NULL;
+}
+
+struct bmp_buf {
+	struct ATTRIB *b;
+	struct mft_inode *mi;
+	struct buffer_head *bh;
+	ulong *buf;
+	size_t bit;
+	u32 nbits;
+	u64 new_valid;
+};
+
+static int bmp_buf_get(struct ntfs_index *indx, struct ntfs_inode *ni,
+		       size_t bit, struct bmp_buf *bbuf)
+{
+	struct ATTRIB *b;
+	size_t data_size, valid_size, vbo, off = bit >> 3;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	CLST vcn = off >> sbi->cluster_bits;
+	struct ATTR_LIST_ENTRY *le = NULL;
+	struct buffer_head *bh;
+	struct super_block *sb;
+	u32 blocksize;
+	const struct INDEX_NAMES *in = &s_index_names[indx->type];
+
+	bbuf->bh = NULL;
+
+	b = ni_find_attr(ni, NULL, &le, ATTR_BITMAP, in->name, in->name_len,
+			 &vcn, &bbuf->mi);
+	bbuf->b = b;
+	if (!b)
+		return -EINVAL;
+
+	if (!b->non_res) {
+		data_size = le32_to_cpu(b->res.data_size);
+
+		if (off >= data_size)
+			return -EINVAL;
+
+		bbuf->buf = (ulong *)resident_data(b);
+		bbuf->bit = 0;
+		bbuf->nbits = data_size * 8;
+
+		return 0;
+	}
+
+	data_size = le64_to_cpu(b->nres.data_size);
+	if (off >= data_size) {
+		WARN_ON(1);
+		return -EINVAL;
+	}
+
+	valid_size = le64_to_cpu(b->nres.valid_size);
+
+	bh = ntfs_bread_run(sbi, &indx->bitmap_run, off);
+	if (!bh)
+		return -EIO;
+
+	if (IS_ERR(bh))
+		return PTR_ERR(bh);
+
+	bbuf->bh = bh;
+
+	if (buffer_locked(bh))
+		__wait_on_buffer(bh);
+
+	lock_buffer(bh);
+
+	sb = sbi->sb;
+	blocksize = sb->s_blocksize;
+
+	vbo = off & ~(size_t)sbi->block_mask;
+
+	bbuf->new_valid = vbo + blocksize;
+	if (bbuf->new_valid <= valid_size)
+		bbuf->new_valid = 0;
+	else if (bbuf->new_valid > data_size)
+		bbuf->new_valid = data_size;
+
+	if (vbo >= valid_size) {
+		memset(bh->b_data, 0, blocksize);
+	} else if (vbo + blocksize > valid_size) {
+		u32 voff = valid_size & sbi->block_mask;
+
+		memset(bh->b_data + voff, 0, blocksize - voff);
+	}
+
+	bbuf->buf = (ulong *)bh->b_data;
+	bbuf->bit = 8 * (off & ~(size_t)sbi->block_mask);
+	bbuf->nbits = 8 * blocksize;
+
+	return 0;
+}
+
+static void bmp_buf_put(struct bmp_buf *bbuf, bool dirty)
+{
+	struct buffer_head *bh = bbuf->bh;
+	struct ATTRIB *b = bbuf->b;
+
+	if (!bh) {
+		if (b && !b->non_res && dirty)
+			bbuf->mi->dirty = true;
+		return;
+	}
+
+	if (!dirty)
+		goto out;
+
+	if (bbuf->new_valid) {
+		b->nres.valid_size = cpu_to_le64(bbuf->new_valid);
+		bbuf->mi->dirty = true;
+	}
+
+	set_buffer_uptodate(bh);
+	mark_buffer_dirty(bh);
+
+out:
+	unlock_buffer(bh);
+	put_bh(bh);
+}
+
+/*
+ * indx_mark_used
+ *
+ * marks the bit 'bit' as used
+ */
+static int indx_mark_used(struct ntfs_index *indx, struct ntfs_inode *ni,
+			  size_t bit)
+{
+	int err;
+	struct bmp_buf bbuf;
+
+	err = bmp_buf_get(indx, ni, bit, &bbuf);
+	if (err)
+		return err;
+
+	__set_bit(bit - bbuf.bit, bbuf.buf);
+
+	bmp_buf_put(&bbuf, true);
+
+	return 0;
+}
+
+/*
+ * indx_mark_free
+ *
+ * the bit 'bit' as free
+ */
+static int indx_mark_free(struct ntfs_index *indx, struct ntfs_inode *ni,
+			  size_t bit)
+{
+	int err;
+	struct bmp_buf bbuf;
+
+	err = bmp_buf_get(indx, ni, bit, &bbuf);
+	if (err)
+		return err;
+
+	__clear_bit(bit - bbuf.bit, bbuf.buf);
+
+	bmp_buf_put(&bbuf, true);
+
+	return 0;
+}
+
+static int scan_nres_bitmap(struct ntfs_sb_info *sbi, struct ATTRIB *bitmap,
+			    struct runs_tree *run, size_t from,
+			    bool (*fn)(const ulong *buf, u32 bit, u32 bits,
+				       size_t *ret),
+			    size_t *ret)
+{
+	struct super_block *sb = sbi->sb;
+	u32 nbits = sb->s_blocksize * 8;
+	u32 blocksize = sb->s_blocksize;
+	u64 valid_size = le64_to_cpu(bitmap->nres.valid_size);
+	u64 data_size = le64_to_cpu(bitmap->nres.data_size);
+	sector_t eblock = bytes_to_block(sb, data_size);
+	size_t vbo = from >> 3;
+	sector_t blk = (vbo & sbi->cluster_mask) >> sb->s_blocksize_bits;
+	sector_t vblock = vbo >> sb->s_blocksize_bits;
+	sector_t blen, block;
+	CLST lcn, len;
+	size_t idx;
+	struct buffer_head *bh;
+
+	*ret = MINUS_ONE_T;
+
+	if (vblock >= eblock)
+		return 0;
+
+	from &= nbits - 1;
+
+	if (!run_lookup_entry(run, vbo >> sbi->cluster_bits, &lcn, &len,
+			      &idx)) {
+		return -ENOENT;
+	}
+
+	blen = (sector_t)len * sbi->blocks_per_cluster;
+	block = (sector_t)lcn * sbi->blocks_per_cluster;
+
+next_run:
+	for (; blk < blen; blk++, from = 0) {
+		bool ok;
+
+		bh = ntfs_bread(sb, block + blk);
+		if (!bh)
+			return -EIO;
+
+		vbo = (u64)vblock << sb->s_blocksize_bits;
+		if (vbo >= valid_size) {
+			memset(bh->b_data, 0, blocksize);
+		} else if (vbo + blocksize > valid_size) {
+			u32 voff = valid_size & sbi->block_mask;
+
+			memset(bh->b_data + voff, 0, blocksize - voff);
+		}
+
+		if (vbo + blocksize > data_size)
+			nbits = 8 * (data_size - vbo);
+
+		ok = nbits > from ?
+			     (*fn)((ulong *)bh->b_data, from, nbits, ret) :
+			     false;
+		put_bh(bh);
+
+		if (ok) {
+			*ret += 8 * vbo;
+			return 0;
+		}
+
+		if (++vblock >= eblock) {
+			*ret = MINUS_ONE_T;
+			return 0;
+		}
+	}
+
+	if (!run_get_entry(run, ++idx, NULL, &lcn, &len))
+		return -ENOENT;
+
+	blk = 0;
+	blen = (sector_t)len * sbi->blocks_per_cluster;
+	block = (sector_t)lcn * sbi->blocks_per_cluster;
+	goto next_run;
+}
+
+static bool scan_for_free(const ulong *buf, u32 bit, u32 bits, size_t *ret)
+{
+	size_t pos = find_next_zero_bit(buf, bits, bit);
+
+	if (pos >= bits)
+		return false;
+	*ret = pos;
+	return true;
+}
+
+/*
+ * indx_find_free
+ *
+ * looks for free bit
+ * returns -1 if no free bits
+ */
+static int indx_find_free(struct ntfs_index *indx, struct ntfs_inode *ni,
+			  size_t *bit, struct ATTRIB **bitmap)
+{
+	struct ATTRIB *b;
+	struct ATTR_LIST_ENTRY *le = NULL;
+	const struct INDEX_NAMES *in = &s_index_names[indx->type];
+
+	b = ni_find_attr(ni, NULL, &le, ATTR_BITMAP, in->name, in->name_len,
+			 NULL, NULL);
+
+	if (!b)
+		return -ENOENT;
+
+	*bitmap = b;
+	*bit = MINUS_ONE_T;
+
+	if (!b->non_res) {
+		u32 nbits = 8 * le32_to_cpu(b->res.data_size);
+		size_t pos = find_next_zero_bit(resident_data(b), nbits, 0);
+
+		if (pos < nbits)
+			*bit = pos;
+	} else {
+		int err = scan_nres_bitmap(ni->mi.sbi, b, &indx->bitmap_run, 0,
+					   &scan_for_free, bit);
+
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static bool scan_for_used(const ulong *buf, u32 bit, u32 bits, size_t *ret)
+{
+	size_t pos = find_next_bit(buf, bits, bit);
+
+	if (pos >= bits)
+		return false;
+	*ret = pos;
+	return true;
+}
+
+/*
+ * indx_used_bit
+ *
+ * looks for used bit
+ * returns MINUS_ONE_T if no used bits
+ */
+int indx_used_bit(struct ntfs_index *indx, struct ntfs_inode *ni, size_t *bit)
+{
+	struct ATTRIB *b;
+	struct ATTR_LIST_ENTRY *le = NULL;
+	size_t from = *bit;
+	const struct INDEX_NAMES *in = &s_index_names[indx->type];
+
+	b = ni_find_attr(ni, NULL, &le, ATTR_BITMAP, in->name, in->name_len,
+			 NULL, NULL);
+
+	if (!b)
+		return -ENOENT;
+
+	*bit = MINUS_ONE_T;
+
+	if (!b->non_res) {
+		u32 nbits = le32_to_cpu(b->res.data_size) * 8;
+		size_t pos = find_next_bit(resident_data(b), nbits, from);
+
+		if (pos < nbits)
+			*bit = pos;
+	} else {
+		int err = scan_nres_bitmap(ni->mi.sbi, b, &indx->bitmap_run,
+					   from, &scan_for_used, bit);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+/*
+ * hdr_find_split
+ *
+ * finds a point at which the index allocation buffer would like to
+ * be split.
+ * NOTE: This function should never return 'END' entry NULL returns on error
+ */
+static const inline struct NTFS_DE *hdr_find_split(const struct INDEX_HDR *hdr)
+{
+	size_t o;
+	const struct NTFS_DE *e = hdr_first_de(hdr);
+	u32 used_2 = le32_to_cpu(hdr->used) >> 1;
+	u16 esize = le16_to_cpu(e->size);
+
+	if (!e || de_is_last(e))
+		return NULL;
+
+	for (o = le32_to_cpu(hdr->de_off) + esize; o < used_2; o += esize) {
+		const struct NTFS_DE *p = e;
+
+		e = Add2Ptr(hdr, o);
+
+		/* We must not return END entry */
+		if (de_is_last(e))
+			return p;
+
+		esize = le16_to_cpu(e->size);
+	}
+
+	return e;
+}
+
+/*
+ * hdr_insert_head
+ *
+ * inserts some entries at the beginning of the buffer.
+ * It is used to insert entries into a newly-created buffer.
+ */
+static const inline struct NTFS_DE *
+hdr_insert_head(struct INDEX_HDR *hdr, const void *ins, u32 ins_bytes)
+{
+	u32 to_move;
+	struct NTFS_DE *e = hdr_first_de(hdr);
+	u32 used = le32_to_cpu(hdr->used);
+
+	if (!e)
+		return NULL;
+
+	/* Now we just make room for the inserted entries and jam it in. */
+	to_move = used - le32_to_cpu(hdr->de_off);
+	memmove(Add2Ptr(e, ins_bytes), e, to_move);
+	memcpy(e, ins, ins_bytes);
+	hdr->used = cpu_to_le32(used + ins_bytes);
+
+	return e;
+}
+
+void fnd_clear(struct ntfs_fnd *fnd)
+{
+	int i;
+
+	for (i = 0; i < fnd->level; i++) {
+		struct indx_node *n = fnd->nodes[i];
+
+		if (!n)
+			continue;
+
+		put_indx_node(n);
+		fnd->nodes[i] = NULL;
+	}
+	fnd->level = 0;
+	fnd->root_de = NULL;
+}
+
+static int fnd_push(struct ntfs_fnd *fnd, struct indx_node *n,
+		    struct NTFS_DE *e)
+{
+	int i;
+
+	i = fnd->level;
+	if (i < 0 || i >= ARRAY_SIZE(fnd->nodes))
+		return -EINVAL;
+	fnd->nodes[i] = n;
+	fnd->de[i] = e;
+	fnd->level += 1;
+	return 0;
+}
+
+static struct indx_node *fnd_pop(struct ntfs_fnd *fnd)
+{
+	struct indx_node *n;
+	int i = fnd->level;
+
+	i -= 1;
+	n = fnd->nodes[i];
+	fnd->nodes[i] = NULL;
+	fnd->level = i;
+
+	return n;
+}
+
+static bool fnd_is_empty(struct ntfs_fnd *fnd)
+{
+	if (!fnd->level)
+		return !fnd->root_de;
+
+	return !fnd->de[fnd->level - 1];
+}
+
+struct ntfs_fnd *fnd_get(struct ntfs_index *indx)
+{
+	struct ntfs_fnd *fnd = ntfs_alloc(sizeof(struct ntfs_fnd), 1);
+
+	if (!fnd)
+		return NULL;
+
+	return fnd;
+}
+
+void fnd_put(struct ntfs_fnd *fnd)
+{
+	if (!fnd)
+		return;
+	fnd_clear(fnd);
+	ntfs_free(fnd);
+}
+
+/*
+ * hdr_find_e
+ *
+ * locates an entry the index buffer.
+ * If no matching entry is found, it returns the first entry which is greater
+ * than the desired entry If the search key is greater than all the entries the
+ * buffer, it returns the 'end' entry. This function does a binary search of the
+ * current index buffer, for the first entry that is <= to the search value
+ * Returns NULL if error
+ */
+static struct NTFS_DE *hdr_find_e(const struct ntfs_index *indx,
+				  const struct INDEX_HDR *hdr, const void *key,
+				  size_t key_len, const void *ctx, int *diff)
+{
+	struct NTFS_DE *e;
+	NTFS_CMP_FUNC cmp = indx->cmp;
+	u32 e_size, e_key_len;
+	u32 end = le32_to_cpu(hdr->used);
+	u32 off = le32_to_cpu(hdr->de_off);
+
+#ifdef NTFS3_INDEX_BINARY_SEARCH
+	int max_idx = 0, fnd, min_idx;
+	int nslots = 64;
+	u16 *offs;
+
+	if (end > 0x10000)
+		goto next;
+
+	offs = ntfs_alloc(sizeof(u16) * nslots, 0);
+	if (!offs)
+		goto next;
+
+	/* use binary search algorithm */
+next1:
+	if (off + sizeof(struct NTFS_DE) > end) {
+		e = NULL;
+		goto out1;
+	}
+	e = Add2Ptr(hdr, off);
+	e_size = le16_to_cpu(e->size);
+
+	if (e_size < sizeof(struct NTFS_DE) || off + e_size > end) {
+		e = NULL;
+		goto out1;
+	}
+
+	if (max_idx >= nslots) {
+		u16 *ptr;
+		int new_slots = QuadAlign(2 * nslots);
+
+		ptr = ntfs_alloc(sizeof(u16) * new_slots, 0);
+		if (ptr)
+			memcpy(ptr, offs, sizeof(u16) * max_idx);
+		ntfs_free(offs);
+		offs = ptr;
+		nslots = new_slots;
+		if (!ptr)
+			goto next;
+	}
+
+	/* Store entry table */
+	offs[max_idx] = off;
+
+	if (!de_is_last(e)) {
+		off += e_size;
+		max_idx += 1;
+		goto next1;
+	}
+
+	/*
+	 * Table of pointers is created
+	 * Use binary search to find entry that is <= to the search value
+	 */
+	fnd = -1;
+	min_idx = 0;
+
+	while (min_idx <= max_idx) {
+		int mid_idx = min_idx + ((max_idx - min_idx) >> 1);
+		int diff2;
+
+		e = Add2Ptr(hdr, offs[mid_idx]);
+
+		e_key_len = le16_to_cpu(e->key_size);
+
+		diff2 = (*cmp)(key, key_len, e + 1, e_key_len, ctx);
+
+		if (!diff2) {
+			*diff = 0;
+			goto out1;
+		}
+
+		if (diff2 < 0) {
+			max_idx = mid_idx - 1;
+			fnd = mid_idx;
+			if (!fnd)
+				break;
+		} else {
+			min_idx = mid_idx + 1;
+		}
+	}
+
+	if (fnd == -1) {
+		e = NULL;
+		goto out1;
+	}
+
+	*diff = -1;
+	e = Add2Ptr(hdr, offs[fnd]);
+
+out1:
+	ntfs_free(offs);
+
+	return e;
+#endif
+
+next:
+	/*
+	 * Entries index are sorted
+	 * Enumerate all entries until we find entry that is <= to the search value
+	 */
+	if (off + sizeof(struct NTFS_DE) > end)
+		return NULL;
+
+	e = Add2Ptr(hdr, off);
+	e_size = le16_to_cpu(e->size);
+
+	if (e_size < sizeof(struct NTFS_DE) || off + e_size > end)
+		return NULL;
+
+	off += e_size;
+
+	e_key_len = le16_to_cpu(e->key_size);
+
+	*diff = (*cmp)(key, key_len, e + 1, e_key_len, ctx);
+	if (!*diff)
+		return e;
+
+	if (*diff <= 0)
+		return e;
+
+	if (de_is_last(e)) {
+		*diff = 1;
+		return e;
+	}
+	goto next;
+}
+
+/*
+ * hdr_insert_de
+ *
+ * inserts an index entry into the buffer.
+ * 'before' should be a pointer previously returned from hdr_find_e
+ */
+static struct NTFS_DE *hdr_insert_de(const struct ntfs_index *indx,
+				     struct INDEX_HDR *hdr,
+				     const struct NTFS_DE *de,
+				     struct NTFS_DE *before, const void *ctx)
+{
+	int diff;
+	size_t off = PtrOffset(hdr, before);
+	u32 used = le32_to_cpu(hdr->used);
+	u32 total = le32_to_cpu(hdr->total);
+	u16 de_size = le16_to_cpu(de->size);
+
+	/* First, check to see if there's enough room */
+	if (used + de_size > total)
+		return NULL;
+
+	/* We know there's enough space, so we know we'll succeed. */
+	if (before) {
+		/* Check that before is inside Index */
+		if (off >= used || off < le32_to_cpu(hdr->de_off) ||
+		    off + le16_to_cpu(before->size) > total) {
+			return NULL;
+		}
+		goto ok;
+	}
+	/* No insert point is applied. Get it manually */
+	before = hdr_find_e(indx, hdr, de + 1, le16_to_cpu(de->key_size), ctx,
+			    &diff);
+	if (!before)
+		return NULL;
+	off = PtrOffset(hdr, before);
+
+ok:
+	/* Now we just make room for the entry and jam it in. */
+	memmove(Add2Ptr(before, de_size), before, used - off);
+
+	hdr->used = cpu_to_le32(used + de_size);
+	memcpy(before, de, de_size);
+
+	return before;
+}
+
+/*
+ * hdr_delete_de
+ *
+ * removes an entry from the index buffer
+ */
+static inline struct NTFS_DE *hdr_delete_de(struct INDEX_HDR *hdr,
+					    struct NTFS_DE *re)
+{
+	u32 used = le32_to_cpu(hdr->used);
+	u16 esize = le16_to_cpu(re->size);
+	u32 off = PtrOffset(hdr, re);
+	int bytes = used - (off + esize);
+
+	if (off >= used || esize < sizeof(struct NTFS_DE) ||
+	    bytes < sizeof(struct NTFS_DE))
+		return NULL;
+
+	hdr->used = cpu_to_le32(used - esize);
+	memmove(re, Add2Ptr(re, esize), bytes);
+
+	return re;
+}
+
+void indx_clear(struct ntfs_index *indx)
+{
+	run_close(&indx->alloc_run);
+	run_close(&indx->bitmap_run);
+}
+
+int indx_init(struct ntfs_index *indx, struct ntfs_sb_info *sbi,
+	      const struct ATTRIB *attr, enum index_mutex_classed type)
+{
+	u32 t32;
+	const struct INDEX_ROOT *root = resident_data(attr);
+
+	/* Check root fields */
+	if (!root->index_block_clst)
+		return -EINVAL;
+
+	indx->type = type;
+	indx->idx2vbn_bits = __ffs(root->index_block_clst);
+
+	t32 = le32_to_cpu(root->index_block_size);
+	indx->index_bits = blksize_bits(t32);
+
+	/* Check index record size */
+	if (t32 < sbi->cluster_size) {
+		/* index record is smaller than a cluster, use 512 blocks */
+		if (t32 != root->index_block_clst * SECTOR_SIZE)
+			return -EINVAL;
+
+		/* Check alignment to a cluster */
+		if ((sbi->cluster_size >> SECTOR_SHIFT) &
+		    (root->index_block_clst - 1)) {
+			return -EINVAL;
+		}
+
+		indx->vbn2vbo_bits = SECTOR_SHIFT;
+	} else {
+		/* index record must be a multiple of cluster size */
+		if (t32 != root->index_block_clst << sbi->cluster_bits)
+			return -EINVAL;
+
+		indx->vbn2vbo_bits = sbi->cluster_bits;
+	}
+
+	indx->cmp = get_cmp_func(root);
+
+	return indx->cmp ? 0 : -EINVAL;
+}
+
+static struct indx_node *indx_new(struct ntfs_index *indx,
+				  struct ntfs_inode *ni, CLST vbn,
+				  const __le64 *sub_vbn)
+{
+	int err;
+	struct NTFS_DE *e;
+	struct indx_node *r;
+	struct INDEX_HDR *hdr;
+	struct INDEX_BUFFER *index;
+	u64 vbo = (u64)vbn << indx->vbn2vbo_bits;
+	u32 bytes = 1u << indx->index_bits;
+	u16 fn;
+	u32 eo;
+
+	r = ntfs_alloc(sizeof(struct indx_node), 1);
+	if (!r)
+		return ERR_PTR(-ENOMEM);
+
+	index = ntfs_alloc(bytes, 1);
+	if (!index) {
+		ntfs_free(r);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	err = ntfs_get_bh(ni->mi.sbi, &indx->alloc_run, vbo, bytes, &r->nb);
+
+	if (err) {
+		ntfs_free(index);
+		ntfs_free(r);
+		return ERR_PTR(err);
+	}
+
+	/* Create header */
+	index->rhdr.sign = NTFS_INDX_SIGNATURE;
+	index->rhdr.fix_off = cpu_to_le16(sizeof(struct INDEX_BUFFER)); // 0x28
+	fn = (bytes >> SECTOR_SHIFT) + 1; // 9
+	index->rhdr.fix_num = cpu_to_le16(fn);
+	index->vbn = cpu_to_le64(vbn);
+	hdr = &index->ihdr;
+	eo = QuadAlign(sizeof(struct INDEX_BUFFER) + fn * sizeof(short));
+	hdr->de_off = cpu_to_le32(eo);
+
+	e = Add2Ptr(hdr, eo);
+
+	if (sub_vbn) {
+		e->flags = NTFS_IE_LAST | NTFS_IE_HAS_SUBNODES;
+		e->size = cpu_to_le16(sizeof(struct NTFS_DE) + sizeof(u64));
+		hdr->used =
+			cpu_to_le32(eo + sizeof(struct NTFS_DE) + sizeof(u64));
+		de_set_vbn_le(e, *sub_vbn);
+		hdr->flags = 1;
+	} else {
+		e->size = cpu_to_le16(sizeof(struct NTFS_DE));
+		hdr->used = cpu_to_le32(eo + sizeof(struct NTFS_DE));
+		e->flags = NTFS_IE_LAST;
+	}
+
+	hdr->total = cpu_to_le32(bytes - offsetof(struct INDEX_BUFFER, ihdr));
+
+	r->index = index;
+	return r;
+}
+
+struct INDEX_ROOT *indx_get_root(struct ntfs_index *indx, struct ntfs_inode *ni,
+				 struct ATTRIB **attr, struct mft_inode **mi)
+{
+	struct ATTR_LIST_ENTRY *le = NULL;
+	struct ATTRIB *a;
+	const struct INDEX_NAMES *in = &s_index_names[indx->type];
+
+	a = ni_find_attr(ni, NULL, &le, ATTR_ROOT, in->name, in->name_len, NULL,
+			 mi);
+	if (!a)
+		return NULL;
+
+	if (attr)
+		*attr = a;
+
+	return resident_data_ex(a, sizeof(struct INDEX_ROOT));
+}
+
+static int indx_write(struct ntfs_index *indx, struct ntfs_inode *ni,
+		      struct indx_node *node, int sync)
+{
+	struct INDEX_BUFFER *ib = node->index;
+
+	return ntfs_write_bh(ni->mi.sbi, &ib->rhdr, &node->nb, sync);
+}
+
+int indx_read(struct ntfs_index *indx, struct ntfs_inode *ni, CLST vbn,
+	      struct indx_node **node)
+{
+	int err;
+	struct INDEX_BUFFER *ib;
+	u64 vbo = (u64)vbn << indx->vbn2vbo_bits;
+	u32 bytes = 1u << indx->index_bits;
+	struct indx_node *in = *node;
+	const struct INDEX_NAMES *name;
+
+	if (!in) {
+		in = ntfs_alloc(sizeof(struct indx_node), 1);
+		if (!in)
+			return -ENOMEM;
+	} else {
+		nb_put(&in->nb);
+	}
+
+	ib = in->index;
+	if (!ib) {
+		ib = ntfs_alloc(bytes, 0);
+		if (!ib) {
+			err = -ENOMEM;
+			goto out;
+		}
+	}
+
+	err = ntfs_read_bh(ni->mi.sbi, &indx->alloc_run, vbo, &ib->rhdr, bytes,
+			   &in->nb);
+
+	if (!err)
+		goto ok;
+
+	if (err == -E_NTFS_FIXUP)
+		goto ok;
+
+	if (err != -ENOENT)
+		goto out;
+
+	name = &s_index_names[indx->type];
+	err = attr_load_runs_vcn(ni, ATTR_ALLOC, name->name, name->name_len,
+				 &indx->alloc_run,
+				 vbo >> ni->mi.sbi->cluster_bits);
+	if (err)
+		goto out;
+
+	err = ntfs_read_bh(ni->mi.sbi, &indx->alloc_run, vbo, &ib->rhdr, bytes,
+			   &in->nb);
+	if (err == -E_NTFS_FIXUP)
+		goto ok;
+
+	if (err)
+		goto out;
+
+ok:
+	if (err == -E_NTFS_FIXUP) {
+		ntfs_write_bh(ni->mi.sbi, &ib->rhdr, &in->nb, 0);
+		err = 0;
+	}
+
+	in->index = ib;
+	*node = in;
+
+out:
+	if (ib != in->index)
+		ntfs_free(ib);
+
+	if (*node != in) {
+		nb_put(&in->nb);
+		ntfs_free(in);
+	}
+
+	return err;
+}
+
+/*
+ * indx_find
+ *
+ * scans NTFS directory for given entry
+ */
+int indx_find(struct ntfs_index *indx, struct ntfs_inode *ni,
+	      const struct INDEX_ROOT *root, const void *key, size_t key_len,
+	      const void *ctx, int *diff, struct NTFS_DE **entry,
+	      struct ntfs_fnd *fnd)
+{
+	int err;
+	struct NTFS_DE *e;
+	const struct INDEX_HDR *hdr;
+	struct indx_node *node;
+
+	if (!root)
+		root = indx_get_root(&ni->dir, ni, NULL, NULL);
+
+	if (!root) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	hdr = &root->ihdr;
+
+	/* Check cache */
+	e = fnd->level ? fnd->de[fnd->level - 1] : fnd->root_de;
+	if (e && !de_is_last(e) &&
+	    !(*indx->cmp)(key, key_len, e + 1, le16_to_cpu(e->key_size), ctx)) {
+		*entry = e;
+		*diff = 0;
+		return 0;
+	}
+
+	/* Soft finder reset */
+	fnd_clear(fnd);
+
+	/* Lookup entry that is <= to the search value */
+	e = hdr_find_e(indx, hdr, key, key_len, ctx, diff);
+	if (!e)
+		return -EINVAL;
+
+	if (fnd)
+		fnd->root_de = e;
+
+	err = 0;
+
+	for (;;) {
+		node = NULL;
+		if (*diff >= 0 || !de_has_vcn_ex(e)) {
+			*entry = e;
+			goto out;
+		}
+
+		/* Read next level. */
+		err = indx_read(indx, ni, de_get_vbn(e), &node);
+		if (err)
+			goto out;
+
+		/* Lookup entry that is <= to the search value */
+		e = hdr_find_e(indx, &node->index->ihdr, key, key_len, ctx,
+			       diff);
+		if (!e) {
+			err = -EINVAL;
+			put_indx_node(node);
+			goto out;
+		}
+
+		fnd_push(fnd, node, e);
+	}
+
+out:
+	return err;
+}
+
+int indx_find_sort(struct ntfs_index *indx, struct ntfs_inode *ni,
+		   const struct INDEX_ROOT *root, struct NTFS_DE **entry,
+		   struct ntfs_fnd *fnd)
+{
+	int err;
+	struct indx_node *n = NULL;
+	struct NTFS_DE *e;
+	size_t iter = 0;
+	int level = fnd->level;
+
+	if (!*entry) {
+		/* Start find */
+		e = hdr_first_de(&root->ihdr);
+		if (!e)
+			return 0;
+		fnd_clear(fnd);
+		fnd->root_de = e;
+	} else if (!level) {
+		if (de_is_last(fnd->root_de)) {
+			*entry = NULL;
+			return 0;
+		}
+
+		e = hdr_next_de(&root->ihdr, fnd->root_de);
+		if (!e)
+			return -EINVAL;
+		fnd->root_de = e;
+	} else {
+		n = fnd->nodes[level - 1];
+		e = fnd->de[level - 1];
+
+		if (de_is_last(e))
+			goto pop_level;
+
+		e = hdr_next_de(&n->index->ihdr, e);
+		if (!e)
+			return -EINVAL;
+
+		fnd->de[level - 1] = e;
+	}
+
+	/* Just to avoid tree cycle */
+next_iter:
+	if (iter++ >= 1000)
+		return -EINVAL;
+
+	while (de_has_vcn_ex(e)) {
+		if (le16_to_cpu(e->size) <
+		    sizeof(struct NTFS_DE) + sizeof(u64)) {
+			if (n) {
+				fnd_pop(fnd);
+				ntfs_free(n);
+			}
+			return -EINVAL;
+		}
+
+		/* Read next level */
+		err = indx_read(indx, ni, de_get_vbn(e), &n);
+		if (err)
+			return err;
+
+		/* Try next level */
+		e = hdr_first_de(&n->index->ihdr);
+		if (!e) {
+			ntfs_free(n);
+			return -EINVAL;
+		}
+
+		fnd_push(fnd, n, e);
+	}
+
+	if (le16_to_cpu(e->size) > sizeof(struct NTFS_DE)) {
+		*entry = e;
+		return 0;
+	}
+
+pop_level:
+	for (;;) {
+		if (!de_is_last(e))
+			goto next_iter;
+
+		/* Pop one level */
+		if (n) {
+			fnd_pop(fnd);
+			ntfs_free(n);
+		}
+
+		level = fnd->level;
+
+		if (level) {
+			n = fnd->nodes[level - 1];
+			e = fnd->de[level - 1];
+		} else if (fnd->root_de) {
+			n = NULL;
+			e = fnd->root_de;
+			fnd->root_de = NULL;
+		} else {
+			*entry = NULL;
+			return 0;
+		}
+
+		if (le16_to_cpu(e->size) > sizeof(struct NTFS_DE)) {
+			*entry = e;
+			if (!fnd->root_de)
+				fnd->root_de = e;
+			return 0;
+		}
+	}
+}
+
+int indx_find_raw(struct ntfs_index *indx, struct ntfs_inode *ni,
+		  const struct INDEX_ROOT *root, struct NTFS_DE **entry,
+		  size_t *off, struct ntfs_fnd *fnd)
+{
+	int err;
+	struct indx_node *n = NULL;
+	struct NTFS_DE *e = NULL;
+	struct NTFS_DE *e2;
+	size_t bit;
+	CLST next_used_vbn;
+	CLST next_vbn;
+	u32 record_size = ni->mi.sbi->record_size;
+
+	/* Use non sorted algorithm */
+	if (!*entry) {
+		/* This is the first call */
+		e = hdr_first_de(&root->ihdr);
+		if (!e)
+			return 0;
+		fnd_clear(fnd);
+		fnd->root_de = e;
+
+		/* The first call with setup of initial element */
+		if (*off >= record_size) {
+			next_vbn = (((*off - record_size) >> indx->index_bits))
+				   << indx->idx2vbn_bits;
+			/* jump inside cycle 'for'*/
+			goto next;
+		}
+
+		/* Start enumeration from root */
+		*off = 0;
+	} else if (!fnd->root_de)
+		return -EINVAL;
+
+	for (;;) {
+		/* Check if current entry can be used */
+		if (e && le16_to_cpu(e->size) > sizeof(struct NTFS_DE))
+			goto ok;
+
+		if (!fnd->level) {
+			/* Continue to enumerate root */
+			if (!de_is_last(fnd->root_de)) {
+				e = hdr_next_de(&root->ihdr, fnd->root_de);
+				if (!e)
+					return -EINVAL;
+				fnd->root_de = e;
+				continue;
+			}
+
+			/* Start to enumerate indexes from 0 */
+			next_vbn = 0;
+		} else {
+			/* Continue to enumerate indexes */
+			e2 = fnd->de[fnd->level - 1];
+
+			n = fnd->nodes[fnd->level - 1];
+
+			if (!de_is_last(e2)) {
+				e = hdr_next_de(&n->index->ihdr, e2);
+				if (!e)
+					return -EINVAL;
+				fnd->de[fnd->level - 1] = e;
+				continue;
+			}
+
+			/* Continue with next index */
+			next_vbn = le64_to_cpu(n->index->vbn) +
+				   root->index_block_clst;
+		}
+
+next:
+		/* Release current index */
+		if (n) {
+			fnd_pop(fnd);
+			put_indx_node(n);
+			n = NULL;
+		}
+
+		/* Skip all free indexes */
+		bit = next_vbn >> indx->idx2vbn_bits;
+		err = indx_used_bit(indx, ni, &bit);
+		if (err == -ENOENT || bit == MINUS_ONE_T) {
+			/* No used indexes */
+			*entry = NULL;
+			return 0;
+		}
+
+		next_used_vbn = bit << indx->idx2vbn_bits;
+
+		/* Read buffer into memory */
+		err = indx_read(indx, ni, next_used_vbn, &n);
+		if (err)
+			return err;
+
+		e = hdr_first_de(&n->index->ihdr);
+		fnd_push(fnd, n, e);
+		if (!e)
+			return -EINVAL;
+	}
+
+ok:
+	/* return offset to restore enumerator if necessary */
+	if (!n) {
+		/* 'e' points in root */
+		*off = PtrOffset(&root->ihdr, e);
+	} else {
+		/* 'e' points in index */
+		*off = (le64_to_cpu(n->index->vbn) << indx->vbn2vbo_bits) +
+		       record_size + PtrOffset(&n->index->ihdr, e);
+	}
+
+	*entry = e;
+	return 0;
+}
+
+/*
+ * indx_create_allocate
+ *
+ * create "Allocation + Bitmap" attributes
+ */
+static int indx_create_allocate(struct ntfs_index *indx, struct ntfs_inode *ni,
+				CLST *vbn)
+{
+	int err = -ENOMEM;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	struct ATTRIB *bitmap;
+	struct ATTRIB *alloc;
+	u32 alloc_size = ntfs_up_cluster(sbi, 1u << indx->index_bits);
+	CLST len = alloc_size >> sbi->cluster_bits;
+	const struct INDEX_NAMES *in = &s_index_names[indx->type];
+	CLST alen;
+	struct runs_tree run;
+
+	run_init(&run);
+
+	err = attr_allocate_clusters(sbi, &run, 0, 0, len, NULL, 0, &alen, 0,
+				     NULL);
+	if (err)
+		goto out;
+
+	err = ni_insert_nonresident(ni, ATTR_ALLOC, in->name, in->name_len,
+				    &run, 0, len, 0, &alloc, NULL);
+	if (err)
+		goto out1;
+
+	err = ni_insert_resident(ni, QuadAlign(1), ATTR_BITMAP, in->name,
+				 in->name_len, &bitmap, NULL);
+	if (err)
+		goto out2;
+
+	memcpy(&indx->alloc_run, &run, sizeof(run));
+
+	*vbn = 0;
+
+	if (in->name == I30_NAME)
+		ni->vfs_inode.i_size = alloc_size;
+
+	return 0;
+
+out2:
+	mi_remove_attr(&ni->mi, alloc);
+
+out1:
+	run_deallocate(sbi, &run, false);
+
+out:
+	return err;
+}
+
+/*
+ * indx_add_allocate
+ *
+ * add clusters to index
+ */
+static int indx_add_allocate(struct ntfs_index *indx, struct ntfs_inode *ni,
+			     CLST *vbn)
+{
+	int err;
+	size_t bit;
+	u64 data_size, alloc_size;
+	u64 bpb, vbpb;
+	struct ATTRIB *bmp, *alloc;
+	struct mft_inode *mi;
+	const struct INDEX_NAMES *in = &s_index_names[indx->type];
+
+	err = indx_find_free(indx, ni, &bit, &bmp);
+	if (err)
+		goto out1;
+
+	if (bit != MINUS_ONE_T) {
+		bmp = NULL;
+	} else {
+		if (bmp->non_res) {
+			bpb = le64_to_cpu(bmp->nres.data_size);
+			vbpb = le64_to_cpu(bmp->nres.valid_size);
+		} else {
+			bpb = vbpb = le32_to_cpu(bmp->res.data_size);
+		}
+
+		/* Increase bitmap */
+		err = attr_set_size(ni, ATTR_BITMAP, in->name, in->name_len,
+				    &indx->bitmap_run, QuadAlign(bpb + 8), NULL,
+				    true, NULL);
+		if (err)
+			goto out1;
+
+		bit = bpb << 3;
+	}
+
+	alloc = ni_find_attr(ni, NULL, NULL, ATTR_ALLOC, in->name, in->name_len,
+			     NULL, &mi);
+	if (!alloc) {
+		if (bmp)
+			goto out2;
+		goto out1;
+	}
+
+	data_size = (u64)(bit + 1) << indx->index_bits;
+	alloc_size = ntfs_up_cluster(ni->mi.sbi, data_size);
+
+	if (alloc_size > le64_to_cpu(alloc->nres.alloc_size)) {
+		/* Increase allocation */
+		err = attr_set_size(ni, ATTR_ALLOC, in->name, in->name_len,
+				    &indx->alloc_run, alloc_size, &alloc_size,
+				    true, NULL);
+		if (err) {
+			if (bmp)
+				goto out2;
+			goto out1;
+		}
+
+		if (in->name == I30_NAME)
+			ni->vfs_inode.i_size = alloc_size;
+	} else if (data_size > le64_to_cpu(alloc->nres.data_size)) {
+		alloc->nres.data_size = alloc->nres.valid_size =
+			cpu_to_le64(data_size);
+		mi->dirty = true;
+	}
+
+	*vbn = bit << indx->idx2vbn_bits;
+
+	return 0;
+
+out2:
+	/* Ops (no space?) */
+	attr_set_size(ni, ATTR_BITMAP, in->name, in->name_len,
+		      &indx->bitmap_run, bpb, &vbpb, false, NULL);
+
+out1:
+	return err;
+}
+
+/*
+ * indx_insert_into_root
+ *
+ * attempts to insert an entry into the index root
+ * If necessary, it will twiddle the index b-tree.
+ */
+static int indx_insert_into_root(struct ntfs_index *indx, struct ntfs_inode *ni,
+				 const struct NTFS_DE *new_de,
+				 struct NTFS_DE *root_de, const void *ctx,
+				 struct ntfs_fnd *fnd)
+{
+	int err = 0;
+	struct NTFS_DE *e, *e0, *re;
+	struct mft_inode *mi;
+	struct ATTRIB *attr;
+	struct MFT_REC *rec;
+	struct INDEX_HDR *hdr;
+	struct indx_node *n;
+	CLST new_vbn;
+	__le64 *sub_vbn, t_vbn;
+	u16 new_de_size;
+	u32 hdr_used, hdr_total, asize, tail, used, aoff, to_move;
+	u32 root_size, new_root_size;
+	struct ntfs_sb_info *sbi;
+	char *next;
+	int ds_root;
+	struct INDEX_ROOT *root, *a_root = NULL;
+
+	/* Get the record this root placed in */
+	root = indx_get_root(indx, ni, &attr, &mi);
+	if (!root)
+		goto out;
+
+	/*
+	 * Try easy case:
+	 * hdr_insert_de will succeed if there's room the root for the new entry.
+	 */
+	hdr = &root->ihdr;
+	sbi = ni->mi.sbi;
+	rec = mi->mrec;
+	aoff = PtrOffset(rec, attr);
+	used = le32_to_cpu(rec->used);
+	new_de_size = le16_to_cpu(new_de->size);
+	hdr_used = le32_to_cpu(hdr->used);
+	hdr_total = le32_to_cpu(hdr->total);
+	asize = le32_to_cpu(attr->size);
+	next = Add2Ptr(attr, asize);
+	tail = used - aoff - asize;
+	root_size = le32_to_cpu(attr->res.data_size);
+
+	ds_root = new_de_size + hdr_used - hdr_total;
+
+	if (used + ds_root < sbi->max_bytes_per_attr) {
+		/* make a room for new elements */
+		memmove(next + ds_root, next, used - aoff - asize);
+		hdr->total = cpu_to_le32(hdr_total + ds_root);
+		e = hdr_insert_de(indx, hdr, new_de, root_de, ctx);
+		WARN_ON(!e);
+		fnd_clear(fnd);
+		fnd->root_de = e;
+		attr->size = cpu_to_le32(asize + ds_root);
+		attr->res.data_size = cpu_to_le32(root_size + ds_root);
+		rec->used = cpu_to_le32(used + ds_root);
+		mi->dirty = true;
+
+		return 0;
+	}
+
+	/* Make a copy of root attribute to restore if error */
+	a_root = ntfs_memdup(attr, asize);
+	if (!a_root) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	/* copy all the non-end entries from the index root to the new buffer.*/
+	to_move = 0;
+	e0 = hdr_first_de(hdr);
+
+	/* Calculate the size to copy */
+	for (e = e0;; e = hdr_next_de(hdr, e)) {
+		if (!e) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		if (de_is_last(e))
+			break;
+		to_move += le16_to_cpu(e->size);
+	}
+
+	n = NULL;
+	if (!to_move) {
+		re = NULL;
+	} else {
+		re = ntfs_memdup(e0, to_move);
+		if (!re) {
+			err = -ENOMEM;
+			goto out;
+		}
+	}
+
+	sub_vbn = NULL;
+	if (de_has_vcn(e)) {
+		t_vbn = de_get_vbn_le(e);
+		sub_vbn = &t_vbn;
+	}
+
+	new_root_size = sizeof(struct INDEX_ROOT) + sizeof(struct NTFS_DE) +
+			sizeof(u64);
+	ds_root = new_root_size - root_size;
+
+	if (ds_root > 0 && used + ds_root > sbi->max_bytes_per_attr) {
+		/* make root external */
+		err = -EOPNOTSUPP;
+		goto out;
+	}
+
+	if (ds_root) {
+		memmove(next + ds_root, next, tail);
+		used += ds_root;
+		asize += ds_root;
+		rec->used = cpu_to_le32(used);
+		attr->size = cpu_to_le32(asize);
+		attr->res.data_size = cpu_to_le32(new_root_size);
+		mi->dirty = true;
+	}
+
+	/* Fill first entry (vcn will be set later) */
+	e = (struct NTFS_DE *)(root + 1);
+	memset(e, 0, sizeof(struct NTFS_DE));
+	e->size = cpu_to_le16(sizeof(struct NTFS_DE) + sizeof(u64));
+	e->flags = NTFS_IE_HAS_SUBNODES | NTFS_IE_LAST;
+
+	hdr->flags = 1;
+	hdr->used = hdr->total =
+		cpu_to_le32(new_root_size - offsetof(struct INDEX_ROOT, ihdr));
+
+	fnd->root_de = hdr_first_de(hdr);
+	mi->dirty = true;
+
+	/* Create alloc and bitmap attributes (if not) */
+	if (run_is_empty(&indx->alloc_run)) {
+		err = indx_create_allocate(indx, ni, &new_vbn);
+		if (err) {
+			/* restore root after 'indx_create_allocate' */
+			memmove(next - ds_root, next, tail);
+			used -= ds_root;
+			rec->used = cpu_to_le32(used);
+			memcpy(attr, a_root, asize);
+			goto out1;
+		}
+	} else {
+		err = indx_add_allocate(indx, ni, &new_vbn);
+		if (err)
+			goto out1;
+	}
+
+	/* layout of record may be changed, so rescan root */
+	root = indx_get_root(indx, ni, &attr, &mi);
+	if (!root) {
+		err = -EINVAL;
+		goto out1;
+	}
+
+	e = (struct NTFS_DE *)(root + 1);
+	*(__le64 *)(e + 1) = cpu_to_le64(new_vbn);
+	mi->dirty = true;
+
+	/* now we can create/format the new buffer and copy the entries into */
+	n = indx_new(indx, ni, new_vbn, sub_vbn);
+	if (IS_ERR(n)) {
+		err = PTR_ERR(n);
+		goto out1;
+	}
+
+	hdr = &n->index->ihdr;
+	hdr_used = le32_to_cpu(hdr->used);
+	hdr_total = le32_to_cpu(hdr->total);
+
+	/* Copy root entries into new buffer */
+	hdr_insert_head(hdr, re, to_move);
+
+	/* Update bitmap attribute */
+	indx_mark_used(indx, ni, new_vbn >> indx->idx2vbn_bits);
+
+	/* Check if we can insert new entry new index buffer */
+	if (hdr_used + new_de_size > hdr_total) {
+		/*
+		 * This occurs if mft record is the same or bigger than index
+		 * buffer. Move all root new index and have no space to add
+		 * new entry classic case when mft record is 1K and index
+		 * buffer 4K the problem should not occurs
+		 */
+		ntfs_free(re);
+		indx_write(indx, ni, n, 0);
+
+		put_indx_node(n);
+		fnd_clear(fnd);
+		err = indx_insert_entry(indx, ni, new_de, ctx, fnd);
+		goto out;
+	}
+
+	/*
+	 * Now root is a parent for new index buffer
+	 * Insert NewEntry a new buffer
+	 */
+	e = hdr_insert_de(indx, hdr, new_de, NULL, ctx);
+	if (!e) {
+		err = -EINVAL;
+		goto out1;
+	}
+	fnd_push(fnd, n, e);
+
+	/* Just write updates index into disk */
+	indx_write(indx, ni, n, 0);
+
+	n = NULL;
+
+out1:
+	ntfs_free(re);
+	if (n)
+		put_indx_node(n);
+
+out:
+	ntfs_free(a_root);
+	return err;
+}
+
+/*
+ * indx_insert_into_buffer
+ *
+ * attempts to insert an entry into an Index Allocation Buffer.
+ * If necessary, it will split the buffer.
+ */
+static int
+indx_insert_into_buffer(struct ntfs_index *indx, struct ntfs_inode *ni,
+			struct INDEX_ROOT *root, const struct NTFS_DE *new_de,
+			const void *ctx, int level, struct ntfs_fnd *fnd)
+{
+	int err;
+	const struct NTFS_DE *sp;
+	struct NTFS_DE *e, *de_t, *up_e = NULL;
+	struct indx_node *n2 = NULL;
+	struct indx_node *n1 = fnd->nodes[level];
+	struct INDEX_HDR *hdr1 = &n1->index->ihdr;
+	struct INDEX_HDR *hdr2;
+	u32 to_copy, used;
+	CLST new_vbn;
+	__le64 t_vbn, *sub_vbn;
+	u16 sp_size;
+
+	/* Try the most easy case */
+	e = fnd->level - 1 == level ? fnd->de[level] : NULL;
+	e = hdr_insert_de(indx, hdr1, new_de, e, ctx);
+	fnd->de[level] = e;
+	if (e) {
+		/* Just write updated index into disk */
+		indx_write(indx, ni, n1, 0);
+		return 0;
+	}
+
+	/*
+	 * No space to insert into buffer. Split it.
+	 * To split we:
+	 *  - Save split point ('cause index buffers will be changed)
+	 * - Allocate NewBuffer and copy all entries <= sp into new buffer
+	 * - Remove all entries (sp including) from TargetBuffer
+	 * - Insert NewEntry into left or right buffer (depending on sp <=>
+	 *     NewEntry)
+	 * - Insert sp into parent buffer (or root)
+	 * - Make sp a parent for new buffer
+	 */
+	sp = hdr_find_split(hdr1);
+	if (!sp)
+		return -EINVAL;
+
+	sp_size = le16_to_cpu(sp->size);
+	up_e = ntfs_alloc(sp_size + sizeof(u64), 0);
+	if (!up_e)
+		return -ENOMEM;
+	memcpy(up_e, sp, sp_size);
+
+	if (!hdr1->flags) {
+		up_e->flags |= NTFS_IE_HAS_SUBNODES;
+		up_e->size = cpu_to_le16(sp_size + sizeof(u64));
+		sub_vbn = NULL;
+	} else {
+		t_vbn = de_get_vbn_le(up_e);
+		sub_vbn = &t_vbn;
+	}
+
+	/* Allocate on disk a new index allocation buffer. */
+	err = indx_add_allocate(indx, ni, &new_vbn);
+	if (err)
+		goto out;
+
+	/* Allocate and format memory a new index buffer */
+	n2 = indx_new(indx, ni, new_vbn, sub_vbn);
+	if (IS_ERR(n2)) {
+		err = PTR_ERR(n2);
+		goto out;
+	}
+
+	hdr2 = &n2->index->ihdr;
+
+	/* Make sp a parent for new buffer */
+	de_set_vbn(up_e, new_vbn);
+
+	/* copy all the entries <= sp into the new buffer. */
+	de_t = hdr_first_de(hdr1);
+	to_copy = PtrOffset(de_t, sp);
+	hdr_insert_head(hdr2, de_t, to_copy);
+
+	/* remove all entries (sp including) from hdr1 */
+	used = le32_to_cpu(hdr1->used) - to_copy - sp_size;
+	memmove(de_t, Add2Ptr(sp, sp_size), used - le32_to_cpu(hdr1->de_off));
+	hdr1->used = cpu_to_le32(used);
+
+	/* Insert new entry into left or right buffer (depending on sp <=> new_de) */
+	hdr_insert_de(indx,
+		      (*indx->cmp)(new_de + 1, le16_to_cpu(new_de->key_size),
+				   up_e + 1, le16_to_cpu(up_e->key_size),
+				   ctx) < 0 ?
+			      hdr2 :
+			      hdr1,
+		      new_de, NULL, ctx);
+
+	indx_mark_used(indx, ni, new_vbn >> indx->idx2vbn_bits);
+
+	indx_write(indx, ni, n1, 0);
+	indx_write(indx, ni, n2, 0);
+
+	put_indx_node(n2);
+
+	/*
+	 * we've finished splitting everybody, so we are ready to
+	 * insert the promoted entry into the parent.
+	 */
+	if (!level) {
+		/* Insert in root */
+		err = indx_insert_into_root(indx, ni, up_e, NULL, ctx, fnd);
+		if (err)
+			goto out;
+	} else {
+		/*
+		 * The target buffer's parent is another index buffer
+		 * TODO: Remove recursion
+		 */
+		err = indx_insert_into_buffer(indx, ni, root, up_e, ctx,
+					      level - 1, fnd);
+		if (err)
+			goto out;
+	}
+
+out:
+	ntfs_free(up_e);
+
+	return err;
+}
+
+/*
+ * indx_insert_entry
+ *
+ * inserts new entry into index
+ */
+int indx_insert_entry(struct ntfs_index *indx, struct ntfs_inode *ni,
+		      const struct NTFS_DE *new_de, const void *ctx,
+		      struct ntfs_fnd *fnd)
+{
+	int err;
+	int diff;
+	struct NTFS_DE *e;
+	struct ntfs_fnd *fnd_a = NULL;
+	struct INDEX_ROOT *root;
+
+	if (!fnd) {
+		fnd_a = fnd_get(indx);
+		if (!fnd_a) {
+			err = -ENOMEM;
+			goto out1;
+		}
+		fnd = fnd_a;
+	}
+
+	root = indx_get_root(indx, ni, NULL, NULL);
+	if (!root) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (fnd_is_empty(fnd)) {
+		/* Find the spot the tree where we want to insert the new entry. */
+		err = indx_find(indx, ni, root, new_de + 1,
+				le16_to_cpu(new_de->key_size), ctx, &diff, &e,
+				fnd);
+		if (err)
+			goto out;
+
+		if (!diff) {
+			err = -EEXIST;
+			goto out;
+		}
+	}
+
+	if (!fnd->level) {
+		/* The root is also a leaf, so we'll insert the new entry into it. */
+		err = indx_insert_into_root(indx, ni, new_de, fnd->root_de, ctx,
+					    fnd);
+		if (err)
+			goto out;
+	} else {
+		/* found a leaf buffer, so we'll insert the new entry into it.*/
+		err = indx_insert_into_buffer(indx, ni, root, new_de, ctx,
+					      fnd->level - 1, fnd);
+		if (err)
+			goto out;
+	}
+
+out:
+	indx->changed = true;
+	fnd_put(fnd_a);
+out1:
+	return err;
+}
+
+/*
+ * indx_find_buffer
+ *
+ * locates a buffer the tree.
+ */
+static struct indx_node *indx_find_buffer(struct ntfs_index *indx,
+					  struct ntfs_inode *ni,
+					  const struct INDEX_ROOT *root,
+					  __le64 vbn, struct indx_node *n)
+{
+	int err;
+	const struct NTFS_DE *e;
+	struct indx_node *r;
+	const struct INDEX_HDR *hdr = n ? &n->index->ihdr : &root->ihdr;
+
+	/* Step 1: Scan one level */
+	for (e = hdr_first_de(hdr);; e = hdr_next_de(hdr, e)) {
+		if (!e)
+			return ERR_PTR(-EINVAL);
+
+		if (de_has_vcn(e) && vbn == de_get_vbn_le(e))
+			return n;
+
+		if (de_is_last(e))
+			break;
+	}
+
+	/* Step2: Do recursion */
+	e = Add2Ptr(hdr, le32_to_cpu(hdr->de_off));
+	for (;;) {
+		if (de_has_vcn_ex(e)) {
+			err = indx_read(indx, ni, de_get_vbn(e), &n);
+			if (err)
+				return ERR_PTR(err);
+
+			r = indx_find_buffer(indx, ni, root, vbn, n);
+			if (r)
+				return r;
+		}
+
+		if (de_is_last(e))
+			break;
+
+		e = Add2Ptr(e, le16_to_cpu(e->size));
+	}
+
+	return NULL;
+}
+
+/*
+ * indx_shrink
+ *
+ * deallocates unused tail indexes
+ */
+static int indx_shrink(struct ntfs_index *indx, struct ntfs_inode *ni,
+		       size_t bit)
+{
+	int err = 0;
+	u64 bpb, new_alloc;
+	size_t nbits;
+	struct ATTRIB *b;
+	struct ATTR_LIST_ENTRY *le = NULL;
+	const struct INDEX_NAMES *in = &s_index_names[indx->type];
+
+	b = ni_find_attr(ni, NULL, &le, ATTR_BITMAP, in->name, in->name_len,
+			 NULL, NULL);
+
+	if (!b)
+		return -ENOENT;
+
+	if (!b->non_res) {
+		unsigned long pos;
+		const unsigned long *bm = resident_data(b);
+
+		nbits = le32_to_cpu(b->res.data_size) * 8;
+
+		if (bit >= nbits)
+			return 0;
+
+		pos = find_next_bit(bm, nbits, bit);
+		if (pos < nbits)
+			return 0;
+	} else {
+		size_t used = MINUS_ONE_T;
+
+		nbits = le64_to_cpu(b->nres.data_size) * 8;
+
+		if (bit >= nbits)
+			return 0;
+
+		err = scan_nres_bitmap(ni->mi.sbi, b, &indx->bitmap_run, bit,
+				       &scan_for_used, &used);
+		if (err)
+			return err;
+
+		if (used != MINUS_ONE_T)
+			return 0;
+	}
+
+	new_alloc = (u64)bit << indx->index_bits;
+
+	err = attr_set_size(ni, ATTR_ALLOC, in->name, in->name_len,
+			    &indx->alloc_run, new_alloc, &new_alloc, false,
+			    NULL);
+	if (err)
+		return err;
+
+	if (in->name == I30_NAME)
+		ni->vfs_inode.i_size = new_alloc;
+
+	bpb = bitmap_size(bit);
+	if (bpb * 8 == nbits)
+		return 0;
+
+	err = attr_set_size(ni, ATTR_BITMAP, in->name, in->name_len,
+			    &indx->bitmap_run, bpb, &bpb, false, NULL);
+
+	return err;
+}
+
+static int indx_free_children(struct ntfs_index *indx, struct ntfs_inode *ni,
+			      const struct NTFS_DE *e, bool trim)
+{
+	int err;
+	struct indx_node *n;
+	struct INDEX_HDR *hdr;
+	CLST vbn = de_get_vbn(e);
+	size_t i;
+
+	err = indx_read(indx, ni, vbn, &n);
+	if (err)
+		return err;
+
+	hdr = &n->index->ihdr;
+	/* First, recurse into the children, if any.*/
+	if (hdr_has_subnode(hdr)) {
+		for (e = hdr_first_de(hdr); e; e = hdr_next_de(hdr, e)) {
+			indx_free_children(indx, ni, e, false);
+			if (de_is_last(e))
+				break;
+		}
+	}
+
+	put_indx_node(n);
+
+	i = vbn >> indx->idx2vbn_bits;
+	/* We've gotten rid of the children; add this buffer to the free list. */
+	indx_mark_free(indx, ni, i);
+
+	if (!trim)
+		return 0;
+
+	/*
+	 * If there are no used indexes after current free index
+	 * then we can truncate allocation and bitmap
+	 * Use bitmap to estimate the case
+	 */
+	indx_shrink(indx, ni, i + 1);
+	return 0;
+}
+
+/*
+ * indx_get_entry_to_replace
+ *
+ * finds a replacement entry for a deleted entry
+ * always returns a node entry:
+ * NTFS_IE_HAS_SUBNODES is set the flags and the size includes the sub_vcn
+ */
+static int indx_get_entry_to_replace(struct ntfs_index *indx,
+				     struct ntfs_inode *ni,
+				     const struct NTFS_DE *de_next,
+				     struct NTFS_DE **de_to_replace,
+				     struct ntfs_fnd *fnd)
+{
+	int err;
+	int level = -1;
+	CLST vbn;
+	struct NTFS_DE *e, *te, *re;
+	struct indx_node *n;
+	struct INDEX_BUFFER *ib;
+
+	*de_to_replace = NULL;
+
+	/* Find first leaf entry down from de_next */
+	vbn = de_get_vbn(de_next);
+	for (;;) {
+		n = NULL;
+		err = indx_read(indx, ni, vbn, &n);
+		if (err)
+			goto out;
+
+		e = hdr_first_de(&n->index->ihdr);
+		fnd_push(fnd, n, e);
+
+		if (!de_is_last(e)) {
+			/*
+			 * This buffer is non-empty, so its first entry could be used as the
+			 * replacement entry.
+			 */
+			level = fnd->level - 1;
+		}
+
+		if (!de_has_vcn(e))
+			break;
+
+		/* This buffer is a node. Continue to go down */
+		vbn = de_get_vbn(e);
+	}
+
+	if (level == -1)
+		goto out;
+
+	n = fnd->nodes[level];
+	te = hdr_first_de(&n->index->ihdr);
+	/* Copy the candidate entry into the replacement entry buffer. */
+	re = ntfs_alloc(le16_to_cpu(te->size) + sizeof(u64), 0);
+	if (!re) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	*de_to_replace = re;
+	memcpy(re, te, le16_to_cpu(te->size));
+
+	if (!de_has_vcn(re)) {
+		/*
+		 * The replacement entry we found doesn't have a sub_vcn. increase its size
+		 * to hold one.
+		 */
+		le16_add_cpu(&re->size, sizeof(u64));
+		re->flags |= NTFS_IE_HAS_SUBNODES;
+	} else {
+		/*
+		 * The replacement entry we found was a node entry, which means that all
+		 * its child buffers are empty. Return them to the free pool.
+		 */
+		indx_free_children(indx, ni, te, true);
+	}
+
+	/*
+	 * Expunge the replacement entry from its former location,
+	 * and then write that buffer.
+	 */
+	ib = n->index;
+	e = hdr_delete_de(&ib->ihdr, te);
+
+	fnd->de[level] = e;
+	indx_write(indx, ni, n, 0);
+
+	/* Check to see if this action created an empty leaf. */
+	if (ib_is_leaf(ib) && ib_is_empty(ib))
+		return 0;
+
+out:
+	fnd_clear(fnd);
+	return err;
+}
+
+/*
+ * indx_delete_entry
+ *
+ * deletes an entry from the index.
+ */
+int indx_delete_entry(struct ntfs_index *indx, struct ntfs_inode *ni,
+		      const void *key, u32 key_len, const void *ctx)
+{
+	int err, diff;
+	struct INDEX_ROOT *root;
+	struct INDEX_HDR *hdr;
+	struct ntfs_fnd *fnd, *fnd2;
+	struct INDEX_BUFFER *ib;
+	struct NTFS_DE *e, *re, *next, *prev, *me;
+	struct indx_node *n, *n2d = NULL;
+	__le64 sub_vbn;
+	int level, level2;
+	struct ATTRIB *attr;
+	struct mft_inode *mi;
+	u32 e_size, root_size, new_root_size;
+	size_t trim_bit;
+	const struct INDEX_NAMES *in;
+
+	fnd = fnd_get(indx);
+	if (!fnd) {
+		err = -ENOMEM;
+		goto out2;
+	}
+
+	fnd2 = fnd_get(NULL);
+	if (!fnd2) {
+		err = -ENOMEM;
+		goto out1;
+	}
+
+	root = indx_get_root(indx, ni, &attr, &mi);
+	if (!root) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* Locate the entry to remove. */
+	err = indx_find(indx, ni, root, key, key_len, ctx, &diff, &e, fnd);
+	if (err)
+		goto out;
+
+	if (!e || diff) {
+		err = -ENOENT;
+		goto out;
+	}
+
+	level = fnd->level;
+
+	if (level) {
+		n = fnd->nodes[level - 1];
+		e = fnd->de[level - 1];
+		ib = n->index;
+		hdr = &ib->ihdr;
+	} else {
+		hdr = &root->ihdr;
+		e = fnd->root_de;
+		n = NULL;
+	}
+
+	e_size = le16_to_cpu(e->size);
+
+	if (!de_has_vcn_ex(e)) {
+		/* The entry to delete is a leaf, so we can just rip it out */
+		hdr_delete_de(hdr, e);
+
+		if (!level) {
+			hdr->total = hdr->used;
+
+			/* Shrink resident root attribute */
+			mi_resize_attr(mi, attr, 0 - e_size);
+			goto out;
+		}
+
+		indx_write(indx, ni, n, 0);
+
+		/*
+		 * Check to see if removing that entry made
+		 * the leaf empty.
+		 */
+		if (ib_is_leaf(ib) && ib_is_empty(ib)) {
+			fnd_pop(fnd);
+			fnd_push(fnd2, n, e);
+		}
+	} else {
+		/*
+		 * The entry we wish to delete is a node buffer, so we
+		 * have to find a replacement for it.
+		 */
+		next = de_get_next(e);
+
+		err = indx_get_entry_to_replace(indx, ni, next, &re, fnd2);
+		if (err)
+			goto out;
+
+		if (re) {
+			de_set_vbn_le(re, de_get_vbn_le(e));
+			hdr_delete_de(hdr, e);
+
+			err = level ? indx_insert_into_buffer(indx, ni, root,
+							      re, ctx,
+							      fnd->level - 1,
+							      fnd) :
+				      indx_insert_into_root(indx, ni, re, e,
+							    ctx, fnd);
+			ntfs_free(re);
+
+			if (err)
+				goto out;
+		} else {
+			/*
+			 * There is no replacement for the current entry.
+			 * This means that the subtree rooted at its node is empty,
+			 * and can be deleted, which turn means that the node can
+			 * just inherit the deleted entry sub_vcn
+			 */
+			indx_free_children(indx, ni, next, true);
+
+			de_set_vbn_le(next, de_get_vbn_le(e));
+			hdr_delete_de(hdr, e);
+			if (level) {
+				indx_write(indx, ni, n, 0);
+			} else {
+				hdr->total = hdr->used;
+
+				/* Shrink resident root attribute */
+				mi_resize_attr(mi, attr, 0 - e_size);
+			}
+		}
+	}
+
+	/* Delete a branch of tree */
+	if (!fnd2 || !fnd2->level)
+		goto out;
+
+	/* Reinit root 'cause it can be changed */
+	root = indx_get_root(indx, ni, &attr, &mi);
+	if (!root) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	n2d = NULL;
+	sub_vbn = fnd2->nodes[0]->index->vbn;
+	level2 = 0;
+	level = fnd->level;
+
+	hdr = level ? &fnd->nodes[level - 1]->index->ihdr : &root->ihdr;
+
+	/* Scan current level */
+	for (e = hdr_first_de(hdr);; e = hdr_next_de(hdr, e)) {
+		if (!e) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		if (de_has_vcn(e) && sub_vbn == de_get_vbn_le(e))
+			break;
+
+		if (de_is_last(e)) {
+			e = NULL;
+			break;
+		}
+	}
+
+	if (!e) {
+		/* Do slow search from root */
+		struct indx_node *in;
+
+		fnd_clear(fnd);
+
+		in = indx_find_buffer(indx, ni, root, sub_vbn, NULL);
+		if (IS_ERR(in)) {
+			err = PTR_ERR(in);
+			goto out;
+		}
+
+		if (in)
+			fnd_push(fnd, in, NULL);
+	}
+
+	/* Merge fnd2 -> fnd */
+	for (level = 0; level < fnd2->level; level++) {
+		fnd_push(fnd, fnd2->nodes[level], fnd2->de[level]);
+		fnd2->nodes[level] = NULL;
+	}
+	fnd2->level = 0;
+
+	hdr = NULL;
+	for (level = fnd->level; level; level--) {
+		struct indx_node *in = fnd->nodes[level - 1];
+
+		ib = in->index;
+		if (ib_is_empty(ib)) {
+			sub_vbn = ib->vbn;
+		} else {
+			hdr = &ib->ihdr;
+			n2d = in;
+			level2 = level;
+			break;
+		}
+	}
+
+	if (!hdr)
+		hdr = &root->ihdr;
+
+	e = hdr_first_de(hdr);
+	if (!e) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (hdr != &root->ihdr || !de_is_last(e)) {
+		prev = NULL;
+		while (!de_is_last(e)) {
+			if (de_has_vcn(e) && sub_vbn == de_get_vbn_le(e))
+				break;
+			prev = e;
+			e = hdr_next_de(hdr, e);
+			if (!e) {
+				err = -EINVAL;
+				goto out;
+			}
+		}
+
+		if (sub_vbn != de_get_vbn_le(e)) {
+			/*
+			 * Didn't find the parent entry, although this buffer is the parent trail.
+			 * Something is corrupt.
+			 */
+			err = -EINVAL;
+			goto out;
+		}
+
+		if (de_is_last(e)) {
+			/*
+			 * Since we can't remove the end entry, we'll remove its
+			 * predecessor instead. This means we have to transfer the
+			 * predecessor's sub_vcn to the end entry.
+			 * Note: that this index block is not empty, so the
+			 * predecessor must exist
+			 */
+			if (!prev) {
+				err = -EINVAL;
+				goto out;
+			}
+
+			if (de_has_vcn(prev)) {
+				de_set_vbn_le(e, de_get_vbn_le(prev));
+			} else if (de_has_vcn(e)) {
+				le16_sub_cpu(&e->size, sizeof(u64));
+				e->flags &= ~NTFS_IE_HAS_SUBNODES;
+				le32_sub_cpu(&hdr->used, sizeof(u64));
+			}
+			e = prev;
+		}
+
+		/*
+		 * Copy the current entry into a temporary buffer (stripping off its
+		 * down-pointer, if any) and delete it from the current buffer or root,
+		 * as appropriate.
+		 */
+		e_size = le16_to_cpu(e->size);
+		me = ntfs_memdup(e, e_size);
+		if (!me) {
+			err = -ENOMEM;
+			goto out;
+		}
+
+		if (de_has_vcn(me)) {
+			me->flags &= ~NTFS_IE_HAS_SUBNODES;
+			le16_sub_cpu(&me->size, sizeof(u64));
+		}
+
+		hdr_delete_de(hdr, e);
+
+		if (hdr == &root->ihdr) {
+			level = 0;
+			hdr->total = hdr->used;
+
+			/* Shrink resident root attribute */
+			mi_resize_attr(mi, attr, 0 - e_size);
+		} else {
+			indx_write(indx, ni, n2d, 0);
+			level = level2;
+		}
+
+		/* Mark unused buffers as free */
+		trim_bit = -1;
+		for (; level < fnd->level; level++) {
+			ib = fnd->nodes[level]->index;
+			if (ib_is_empty(ib)) {
+				size_t k = le64_to_cpu(ib->vbn) >>
+					   indx->idx2vbn_bits;
+
+				indx_mark_free(indx, ni, k);
+				if (k < trim_bit)
+					trim_bit = k;
+			}
+		}
+
+		fnd_clear(fnd);
+		/*fnd->root_de = NULL;*/
+
+		/*
+		 * Re-insert the entry into the tree.
+		 * Find the spot the tree where we want to insert the new entry.
+		 */
+		err = indx_insert_entry(indx, ni, me, ctx, fnd);
+		ntfs_free(me);
+		if (err)
+			goto out;
+
+		if (trim_bit != -1)
+			indx_shrink(indx, ni, trim_bit);
+	} else {
+		/*
+		 * This tree needs to be collapsed down to an empty root.
+		 * Recreate the index root as an empty leaf and free all the bits the
+		 * index allocation bitmap.
+		 */
+		fnd_clear(fnd);
+		fnd_clear(fnd2);
+
+		in = &s_index_names[indx->type];
+
+		err = attr_set_size(ni, ATTR_ALLOC, in->name, in->name_len,
+				    &indx->alloc_run, 0, NULL, false, NULL);
+		err = ni_remove_attr(ni, ATTR_ALLOC, in->name, in->name_len,
+				     false, NULL);
+		run_close(&indx->alloc_run);
+
+		err = attr_set_size(ni, ATTR_BITMAP, in->name, in->name_len,
+				    &indx->bitmap_run, 0, NULL, false, NULL);
+		err = ni_remove_attr(ni, ATTR_BITMAP, in->name, in->name_len,
+				     false, NULL);
+		run_close(&indx->bitmap_run);
+
+		root = indx_get_root(indx, ni, &attr, &mi);
+		if (!root) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		root_size = le32_to_cpu(attr->res.data_size);
+		new_root_size =
+			sizeof(struct INDEX_ROOT) + sizeof(struct NTFS_DE);
+
+		if (new_root_size != root_size &&
+		    !mi_resize_attr(mi, attr, new_root_size - root_size)) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		/* Fill first entry */
+		e = (struct NTFS_DE *)(root + 1);
+		e->ref.low = 0;
+		e->ref.high = 0;
+		e->ref.seq = 0;
+		e->size = cpu_to_le16(sizeof(struct NTFS_DE));
+		e->flags = NTFS_IE_LAST; // 0x02
+		e->key_size = 0;
+		e->res = 0;
+
+		hdr = &root->ihdr;
+		hdr->flags = 0;
+		hdr->used = hdr->total = cpu_to_le32(
+			new_root_size - offsetof(struct INDEX_ROOT, ihdr));
+		mi->dirty = true;
+
+		if (in->name == I30_NAME)
+			ni->vfs_inode.i_size = 0;
+	}
+
+out:
+	fnd_put(fnd2);
+out1:
+	indx->changed = true;
+	fnd_put(fnd);
+
+out2:
+	return err;
+}
+
+int indx_update_dup(struct ntfs_inode *ni, struct ntfs_sb_info *sbi,
+		    const struct ATTR_FILE_NAME *fname,
+		    const struct NTFS_DUP_INFO *dup, int sync)
+{
+	int err, diff;
+	struct NTFS_DE *e = NULL;
+	struct ATTR_FILE_NAME *e_fname;
+	struct ntfs_fnd *fnd;
+	struct INDEX_ROOT *root;
+	struct mft_inode *mi;
+	struct ntfs_index *indx = &ni->dir;
+
+	fnd = fnd_get(indx);
+	if (!fnd) {
+		err = -ENOMEM;
+		goto out1;
+	}
+
+	root = indx_get_root(indx, ni, NULL, &mi);
+	if (!root) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* Find entries tree and on disk */
+	err = indx_find(indx, ni, root, fname, fname_full_size(fname), sbi,
+			&diff, &e, fnd);
+	if (err)
+		goto out;
+
+	if (!e) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (diff) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	e_fname = (struct ATTR_FILE_NAME *)(e + 1);
+
+	if (!memcmp(&e_fname->dup, dup, sizeof(*dup))) {
+		/* nothing to update in index! Try to avoid this call */
+		goto out;
+	}
+
+	memcpy(&e_fname->dup, dup, sizeof(*dup));
+
+	if (fnd->level) {
+		err = indx_write(indx, ni, fnd->nodes[fnd->level - 1], sync);
+	} else if (sync) {
+		mi->dirty = true;
+		err = mi_write(mi, 1);
+	} else {
+		mi->dirty = true;
+		mark_inode_dirty(&ni->vfs_inode);
+	}
+
+out:
+	fnd_put(fnd);
+
+out1:
+	return err;
+}
diff --git a/fs/ntfs3/inode.c b/fs/ntfs3/inode.c
new file mode 100644
index 000000000000..79160d00120d
--- /dev/null
+++ b/fs/ntfs3/inode.c
@@ -0,0 +1,2030 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/iversion.h>
+#include <linux/mpage.h>
+#include <linux/namei.h>
+#include <linux/nls.h>
+#include <linux/uio.h>
+#include <linux/version.h>
+#include <linux/writeback.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+/*
+ * ntfs_read_mft
+ *
+ * reads record and parses MFT
+ */
+static struct inode *ntfs_read_mft(struct inode *inode,
+				   const struct cpu_str *name,
+				   const struct MFT_REF *ref)
+{
+	int err = 0;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	struct super_block *sb = inode->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	mode_t mode = 0;
+	struct ATTR_STD_INFO5 *std5 = NULL;
+	struct ATTR_LIST_ENTRY *le;
+	struct ATTRIB *attr;
+	bool is_match = false;
+	bool is_root = false;
+	bool is_dir;
+	unsigned long ino = inode->i_ino;
+	u32 rp_fa = 0, asize, t32;
+	u16 roff, rsize, names = 0;
+	const struct ATTR_FILE_NAME *fname = NULL;
+	const struct INDEX_ROOT *root;
+	struct REPARSE_DATA_BUFFER rp; // 0x18 bytes
+	u64 t64;
+	struct MFT_REC *rec;
+	struct runs_tree *run;
+
+	inode->i_op = NULL;
+
+	err = mi_init(&ni->mi, sbi, ino);
+	if (err)
+		goto out;
+
+	if (!sbi->mft.ni && ino == MFT_REC_MFT && !sb->s_root) {
+		t64 = sbi->mft.lbo >> sbi->cluster_bits;
+		t32 = bytes_to_cluster(sbi, MFT_REC_VOL * sbi->record_size);
+		sbi->mft.ni = ni;
+		init_rwsem(&ni->file.run_lock);
+
+		if (!run_add_entry(&ni->file.run, 0, t64, t32)) {
+			err = -ENOMEM;
+			goto out;
+		}
+	}
+
+	err = mi_read(&ni->mi, ino == MFT_REC_MFT);
+
+	if (err)
+		goto out;
+
+	rec = ni->mi.mrec;
+
+	if (sbi->flags & NTFS_FLAGS_LOG_REPLAYING) {
+		;
+	} else if (ref->seq != rec->seq) {
+		err = -EINVAL;
+		ntfs_err(sb, "MFT: r=%lx, expect seq=%x instead of %x!", ino,
+			 le16_to_cpu(ref->seq), le16_to_cpu(rec->seq));
+		goto out;
+	} else if (!is_rec_inuse(rec)) {
+		err = -EINVAL;
+		ntfs_err(sb, "Inode r=%x is not in use!", (u32)ino);
+		goto out;
+	}
+
+	if (le32_to_cpu(rec->total) != sbi->record_size) {
+		// bad inode?
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (!is_rec_base(rec))
+		goto Ok;
+
+	/* record should contain $I30 root */
+	is_dir = rec->flags & RECORD_FLAG_DIR;
+
+	inode->i_generation = le16_to_cpu(rec->seq);
+
+	/* Enumerate all struct Attributes MFT */
+	le = NULL;
+	attr = NULL;
+
+	/*
+	 * to reduce tab pressure use goto instead of
+	 * while( (attr = ni_enum_attr_ex(ni, attr, &le) ))
+	 */
+next_attr:
+	run = NULL;
+	err = -EINVAL;
+	attr = ni_enum_attr_ex(ni, attr, &le);
+	if (!attr)
+		goto end_enum;
+
+	if (le && le->vcn) {
+		/* This is non primary attribute segment. Ignore if not MFT */
+		if (ino != MFT_REC_MFT || attr->type != ATTR_DATA)
+			goto next_attr;
+
+		run = &ni->file.run;
+		asize = le32_to_cpu(attr->size);
+		goto attr_unpack_run;
+	}
+
+	roff = attr->non_res ? 0 : le16_to_cpu(attr->res.data_off);
+	rsize = attr->non_res ? 0 : le32_to_cpu(attr->res.data_size);
+	asize = le32_to_cpu(attr->size);
+
+	switch (attr->type) {
+	case ATTR_STD:
+		if (attr->non_res ||
+		    asize < sizeof(struct ATTR_STD_INFO) + roff ||
+		    rsize < sizeof(struct ATTR_STD_INFO))
+			goto out;
+
+		if (std5)
+			goto next_attr;
+
+		std5 = Add2Ptr(attr, roff);
+
+#ifdef STATX_BTIME
+		nt2kernel(std5->cr_time, &ni->i_crtime);
+#endif
+		nt2kernel(std5->a_time, &inode->i_atime);
+		nt2kernel(std5->c_time, &inode->i_ctime);
+		nt2kernel(std5->m_time, &inode->i_mtime);
+
+		ni->std_fa = std5->fa;
+
+		if (asize >= sizeof(struct ATTR_STD_INFO5) + roff &&
+		    rsize >= sizeof(struct ATTR_STD_INFO5))
+			ni->std_security_id = std5->security_id;
+		goto next_attr;
+
+	case ATTR_LIST:
+		if (attr->name_len || le || ino == MFT_REC_LOG)
+			goto out;
+
+		err = ntfs_load_attr_list(ni, attr);
+		if (err)
+			goto out;
+
+		le = NULL;
+		attr = NULL;
+		goto next_attr;
+
+	case ATTR_NAME:
+		if (attr->non_res || asize < SIZEOF_ATTRIBUTE_FILENAME + roff ||
+		    rsize < SIZEOF_ATTRIBUTE_FILENAME)
+			goto out;
+
+		fname = Add2Ptr(attr, roff);
+		if (fname->type == FILE_NAME_DOS)
+			goto next_attr;
+
+		names += 1;
+		if (name && name->len == fname->name_len &&
+		    !ntfs_cmp_names_cpu(name, (struct le_str *)&fname->name_len,
+					NULL))
+			is_match = true;
+
+		goto next_attr;
+
+	case ATTR_DATA:
+		if (is_dir) {
+			/* ignore data attribute in dir record */
+			goto next_attr;
+		}
+
+		if (ino == MFT_REC_BADCLUST && !attr->non_res)
+			goto next_attr;
+
+		if (attr->name_len &&
+		    ((ino != MFT_REC_BADCLUST || !attr->non_res ||
+		      attr->name_len != ARRAY_SIZE(BAD_NAME) ||
+		      memcmp(attr_name(attr), BAD_NAME, sizeof(BAD_NAME))) &&
+		     (ino != MFT_REC_SECURE || !attr->non_res ||
+		      attr->name_len != ARRAY_SIZE(SDS_NAME) ||
+		      memcmp(attr_name(attr), SDS_NAME, sizeof(SDS_NAME))))) {
+			/* file contains stream attribute. ignore it */
+			goto next_attr;
+		}
+
+		if (is_attr_sparsed(attr))
+			ni->std_fa |= FILE_ATTRIBUTE_SPARSE_FILE;
+		else
+			ni->std_fa &= ~FILE_ATTRIBUTE_SPARSE_FILE;
+
+		if (is_attr_compressed(attr))
+			ni->std_fa |= FILE_ATTRIBUTE_COMPRESSED;
+		else
+			ni->std_fa &= ~FILE_ATTRIBUTE_COMPRESSED;
+
+		if (is_attr_encrypted(attr))
+			ni->std_fa |= FILE_ATTRIBUTE_ENCRYPTED;
+		else
+			ni->std_fa &= ~FILE_ATTRIBUTE_ENCRYPTED;
+
+		if (!attr->non_res) {
+			ni->i_valid = inode->i_size = rsize;
+			inode_set_bytes(inode, rsize);
+			t32 = asize;
+		} else {
+			t32 = le16_to_cpu(attr->nres.run_off);
+		}
+
+		mode = S_IFREG | (0777 & sbi->options.fs_fmask_inv);
+
+		if (!attr->non_res) {
+			ni->ni_flags |= NI_FLAG_RESIDENT;
+			goto next_attr;
+		}
+
+		inode_set_bytes(inode, attr_ondisk_size(attr));
+
+		ni->i_valid = le64_to_cpu(attr->nres.valid_size);
+		inode->i_size = le64_to_cpu(attr->nres.data_size);
+		if (!attr->nres.alloc_size)
+			goto next_attr;
+
+		run = ino == MFT_REC_BITMAP ? &sbi->used.bitmap.run :
+					      &ni->file.run;
+		break;
+
+	case ATTR_ROOT:
+		if (attr->non_res)
+			goto out;
+
+		root = Add2Ptr(attr, roff);
+		is_root = true;
+
+		if (attr->name_len != ARRAY_SIZE(I30_NAME) ||
+		    memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME)))
+			goto next_attr;
+
+		if (root->type != ATTR_NAME ||
+		    root->rule != NTFS_COLLATION_TYPE_FILENAME)
+			goto out;
+
+		if (!is_dir)
+			goto next_attr;
+
+		ni->ni_flags |= NI_FLAG_DIR;
+
+		err = indx_init(&ni->dir, sbi, attr, INDEX_MUTEX_I30);
+		if (err)
+			goto out;
+
+		mode = sb->s_root ?
+			       (S_IFDIR | (0777 & sbi->options.fs_dmask_inv)) :
+			       (S_IFDIR | 0777);
+		goto next_attr;
+
+	case ATTR_ALLOC:
+		if (!is_root || attr->name_len != ARRAY_SIZE(I30_NAME) ||
+		    memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME)))
+			goto next_attr;
+
+		inode->i_size = le64_to_cpu(attr->nres.data_size);
+		ni->i_valid = le64_to_cpu(attr->nres.valid_size);
+		inode_set_bytes(inode, le64_to_cpu(attr->nres.alloc_size));
+
+		run = &ni->dir.alloc_run;
+		break;
+
+	case ATTR_BITMAP:
+		if (ino == MFT_REC_MFT) {
+			if (!attr->non_res)
+				goto out;
+#ifndef NTFS3_64BIT_CLUSTER
+			/* 0x20000000 = 2^32 / 8 */
+			if (le64_to_cpu(attr->nres.alloc_size) >= 0x20000000)
+				goto out;
+#endif
+			run = &sbi->mft.bitmap.run;
+			break;
+		} else if (is_dir && attr->name_len == ARRAY_SIZE(I30_NAME) &&
+			   !memcmp(attr_name(attr), I30_NAME,
+				   sizeof(I30_NAME)) &&
+			   attr->non_res) {
+			run = &ni->dir.bitmap_run;
+			break;
+		}
+		goto next_attr;
+
+	case ATTR_REPARSE:
+		if (attr->name_len)
+			goto next_attr;
+
+		rp_fa = ni_parse_reparse(ni, attr, &rp);
+		switch (rp_fa) {
+		case REPARSE_LINK:
+			if (!attr->non_res) {
+				inode->i_size = rsize;
+				inode_set_bytes(inode, rsize);
+				t32 = asize;
+			} else {
+				inode->i_size =
+					le64_to_cpu(attr->nres.data_size);
+				t32 = le16_to_cpu(attr->nres.run_off);
+			}
+
+			/* Looks like normal symlink */
+			ni->i_valid = inode->i_size;
+
+			/* Clear directory bit */
+			if (ni->ni_flags & NI_FLAG_DIR) {
+				indx_clear(&ni->dir);
+				memset(&ni->dir, 0, sizeof(ni->dir));
+				ni->ni_flags &= ~NI_FLAG_DIR;
+			} else {
+				run_close(&ni->file.run);
+			}
+			mode = S_IFLNK | 0777;
+			is_dir = false;
+			if (attr->non_res) {
+				run = &ni->file.run;
+				goto attr_unpack_run; // double break
+			}
+			break;
+
+		case REPARSE_COMPRESSED:
+			break;
+
+		case REPARSE_DEDUPLICATED:
+			break;
+		}
+		goto next_attr;
+
+	case ATTR_EA_INFO:
+		if (!attr->name_len &&
+		    resident_data_ex(attr, sizeof(struct EA_INFO)))
+			ni->ni_flags |= NI_FLAG_EA;
+		goto next_attr;
+
+	default:
+		goto next_attr;
+	}
+
+attr_unpack_run:
+	roff = le16_to_cpu(attr->nres.run_off);
+
+	t64 = le64_to_cpu(attr->nres.svcn);
+	err = run_unpack_ex(run, sbi, ino, t64, le64_to_cpu(attr->nres.evcn),
+			    t64, Add2Ptr(attr, roff), asize - roff);
+	if (err < 0)
+		goto out;
+	err = 0;
+	goto next_attr;
+
+end_enum:
+
+	if (!std5)
+		goto out;
+
+	if (!is_match && name) {
+		/* reuse rec as buffer for ascii name */
+		err = -ENOENT;
+		goto out;
+	}
+
+	if (std5->fa & FILE_ATTRIBUTE_READONLY)
+		mode &= ~0222;
+
+	/* Setup 'uid' and 'gid' */
+	inode->i_uid = sbi->options.fs_uid;
+	inode->i_gid = sbi->options.fs_gid;
+
+	if (!names) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (S_ISDIR(mode)) {
+		ni->std_fa |= FILE_ATTRIBUTE_DIRECTORY;
+
+		/*
+		 * dot and dot-dot should be included in count but was not
+		 * included in enumeration.
+		 * Usually a hard links to directories are disabled
+		 */
+		set_nlink(inode, 1);
+		inode->i_op = &ntfs_dir_inode_operations;
+		inode->i_fop = &ntfs_dir_operations;
+		ni->i_valid = 0;
+	} else if (S_ISLNK(mode)) {
+		ni->std_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
+		inode->i_op = &ntfs_link_inode_operations;
+		inode->i_fop = NULL;
+		inode_nohighmem(inode); // ??
+		set_nlink(inode, names);
+	} else if (S_ISREG(mode)) {
+		ni->std_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
+
+		set_nlink(inode, names);
+
+		inode->i_op = &ntfs_file_inode_operations;
+		inode->i_fop = &ntfs_file_operations;
+		inode->i_mapping->a_ops =
+			is_compressed(ni) ? &ntfs_aops_cmpr : &ntfs_aops;
+
+		if (ino != MFT_REC_MFT)
+			init_rwsem(&ni->file.run_lock);
+	} else if (fname && fname->home.low == cpu_to_le32(MFT_REC_EXTEND) &&
+		   fname->home.seq == cpu_to_le16(MFT_REC_EXTEND)) {
+		/* Records in $Extend are not a files or general directories */
+	} else {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if ((sbi->options.sys_immutable &&
+	     (std5->fa & FILE_ATTRIBUTE_SYSTEM)) &&
+	    !S_ISFIFO(mode) && !S_ISSOCK(mode) && !S_ISLNK(mode)) {
+		inode->i_flags |= S_IMMUTABLE;
+	} else {
+		inode->i_flags &= ~S_IMMUTABLE;
+	}
+
+	inode->i_mode = mode;
+	if (!(ni->ni_flags & NI_FLAG_EA)) {
+		/* if no xattr then no security (stored in xattr) */
+		inode->i_flags |= S_NOSEC;
+	}
+
+Ok:
+	if (ino == MFT_REC_MFT && !sb->s_root)
+		sbi->mft.ni = NULL;
+
+	unlock_new_inode(inode);
+
+	return inode;
+
+out:
+	if (ino == MFT_REC_MFT && !sb->s_root)
+		sbi->mft.ni = NULL;
+
+	iget_failed(inode);
+	return ERR_PTR(err);
+}
+
+/* returns 1 if match */
+static int ntfs_test_inode(struct inode *inode, const struct MFT_REF *ref)
+{
+	return ino_get(ref) == inode->i_ino;
+}
+
+static int ntfs_set_inode(struct inode *inode, const struct MFT_REF *ref)
+{
+	inode->i_ino = ino_get(ref);
+
+	return 0;
+}
+
+struct inode *ntfs_iget5(struct super_block *sb, const struct MFT_REF *ref,
+			 const struct cpu_str *name)
+{
+	struct inode *inode;
+
+	inode = iget5_locked(sb, ino_get(ref),
+			     (int (*)(struct inode *, void *))ntfs_test_inode,
+			     (int (*)(struct inode *, void *))ntfs_set_inode,
+			     (void *)ref);
+	if (unlikely(!inode))
+		return ERR_PTR(-ENOMEM);
+
+	/* If this is a freshly allocated inode, need to read it now. */
+	if (inode->i_state & I_NEW)
+		inode = ntfs_read_mft(inode, name, ref);
+	else if (ref->seq != ntfs_i(inode)->mi.mrec->seq) {
+		/* inode overlaps? */
+		make_bad_inode(inode);
+	}
+
+	return inode;
+}
+
+enum get_block_ctx {
+	GET_BLOCK_GENERAL = 0,
+	GET_BLOCK_WRITE_BEGIN = 1,
+	GET_BLOCK_DIRECT_IO_R = 2,
+	GET_BLOCK_DIRECT_IO_W = 3,
+	GET_BLOCK_BMAP = 4,
+};
+
+static noinline int ntfs_get_block_vbo(struct inode *inode, u64 vbo,
+				       struct buffer_head *bh, int create,
+				       enum get_block_ctx ctx)
+{
+	struct super_block *sb = inode->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	u8 cluster_bits = sbi->cluster_bits;
+	u32 block_size = sb->s_blocksize;
+	u64 bytes, lbo, valid;
+	u32 off;
+	int err;
+	CLST vcn, lcn, len;
+	bool new;
+
+	/*clear previous state*/
+	clear_buffer_new(bh);
+	clear_buffer_uptodate(bh);
+
+	/* direct write uses 'create=0'*/
+	if (!create && vbo >= ni->i_valid) {
+		/* out of valid */
+		return 0;
+	}
+
+	if (vbo >= inode->i_size) {
+		/* out of size */
+		return 0;
+	}
+
+	vcn = vbo >> cluster_bits;
+	off = vbo & sbi->cluster_mask;
+	new = false;
+
+	err = attr_data_get_block(ni, vcn, 1, &lcn, &len, create ? &new : NULL);
+	if (err)
+		goto out;
+
+	if (!len)
+		return 0;
+
+	bytes = ((u64)len << cluster_bits) - off;
+
+	if (lcn == SPARSE_LCN) {
+		if (!create) {
+			if (bh->b_size > bytes)
+				bh->b_size = bytes;
+
+			return 0;
+		}
+		WARN_ON(1);
+	}
+
+	WARN_ON(lcn == RESIDENT_LCN);
+
+	if (new) {
+		u32 allocated = len * sbi->cluster_size;
+
+		set_buffer_new(bh);
+		if (allocated > block_size)
+			ntfs_sparse_cluster(inode, bh->b_page,
+					    (loff_t)vcn << sbi->cluster_bits,
+					    allocated);
+	}
+
+	lbo = ((u64)lcn << cluster_bits) + off;
+
+	set_buffer_mapped(bh);
+	bh->b_bdev = sb->s_bdev;
+	bh->b_blocknr = lbo >> sb->s_blocksize_bits;
+
+	valid = ni->i_valid;
+
+	if (ctx == GET_BLOCK_DIRECT_IO_W) {
+		/*ntfs_direct_IO will update ni->i_valid */
+		if (vbo >= valid)
+			set_buffer_new(bh);
+	} else if (create && ctx == GET_BLOCK_WRITE_BEGIN &&
+		   vbo + bh->b_size > valid) {
+		u32 voff = valid > vbo ? (valid - vbo) : 0;
+		struct page *page = bh->b_page;
+
+		off = bh_offset(bh);
+		zero_user_segment(page, off + voff, off + bh->b_size);
+		set_buffer_uptodate(bh);
+		ni->i_valid = vbo + bh->b_size;
+
+		/* ntfs_write_end will update ni->i_valid*/
+	} else if (create) {
+		/*normal write*/
+		if (vbo >= valid) {
+			set_buffer_new(bh);
+			if (bytes > bh->b_size)
+				bytes = bh->b_size;
+			ni->i_valid = vbo + bytes;
+			mark_inode_dirty(inode);
+		}
+	} else if (vbo >= valid) {
+		/* read out of valid data*/
+		/* should never be here 'cause already checked */
+		clear_buffer_mapped(bh);
+	} else if (vbo + bytes <= valid) {
+		/* normal read */
+	} else if (vbo + block_size <= valid) {
+		/* normal short read */
+		bytes = block_size;
+	} else {
+		/*
+		 * read across valid size: vbo < valid && valid < vbo + block_size
+		 */
+		u32 voff = valid - vbo;
+		struct page *page = bh->b_page;
+
+		bh->b_size = bytes = block_size;
+		ll_rw_block(REQ_OP_READ, 0, 1, &bh);
+		wait_on_buffer(bh);
+		/* Uhhuh. Read error. Complain and punt. */
+		if (!buffer_uptodate(bh)) {
+			err = -EIO;
+			goto out;
+		}
+
+		off = bh_offset(bh);
+		zero_user_segment(page, off + voff, off + block_size);
+	}
+
+	if (bh->b_size > bytes)
+		bh->b_size = bytes;
+
+#ifndef __LP64__
+	if (ctx == GET_BLOCK_DIRECT_IO_W || ctx == GET_BLOCK_DIRECT_IO_R) {
+		static_assert(sizeof(size_t) < sizeof(loff_t));
+		if (bytes > 0x40000000u)
+			bh->b_size = 0x40000000u;
+	}
+#endif
+
+	return 0;
+
+out:
+	return err;
+}
+
+int ntfs_get_block(struct inode *inode, sector_t vbn,
+		   struct buffer_head *bh_result, int create)
+{
+	return ntfs_get_block_vbo(inode, (u64)vbn << inode->i_blkbits,
+				  bh_result, create, GET_BLOCK_GENERAL);
+}
+
+static int ntfs_get_block_bmap(struct inode *inode, sector_t vsn,
+			       struct buffer_head *bh_result, int create)
+{
+	return ntfs_get_block_vbo(inode,
+				  (u64)vsn << inode->i_sb->s_blocksize_bits,
+				  bh_result, create, GET_BLOCK_BMAP);
+}
+
+static sector_t ntfs_bmap(struct address_space *mapping, sector_t block)
+{
+	return generic_block_bmap(mapping, block, ntfs_get_block_bmap);
+}
+
+int ntfs_readpage(struct file *file, struct page *page)
+{
+	int err;
+	struct address_space *mapping = page->mapping;
+	struct inode *inode = mapping->host;
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	if (ni_has_resident_data(ni)) {
+		ni_lock(ni);
+
+		if (!ni_has_resident_data(ni)) {
+			ni_unlock(ni);
+		} else {
+			u64 vbo = page->index << PAGE_SHIFT;
+			struct ATTRIB *attr;
+			const char *data;
+			u32 data_size;
+
+			attr = ni_find_attr(ni, NULL, NULL, ATTR_DATA, NULL, 0,
+					    NULL, NULL);
+			if (!attr) {
+				err = -EINVAL;
+				ni_unlock(ni);
+				unlock_page(page);
+				goto out;
+			}
+
+			WARN_ON(attr->non_res);
+
+			data = resident_data(attr);
+			data_size = le32_to_cpu(attr->res.data_size);
+
+			if (vbo < data_size) {
+				void *kaddr = kmap_atomic(page);
+
+				err = data_size - vbo;
+				if (err > PAGE_SIZE)
+					err = PAGE_SIZE;
+
+				memcpy(kaddr, data + vbo, err);
+				flush_dcache_page(page);
+				kunmap_atomic(kaddr);
+				zero_user_segment(page, err, PAGE_SIZE);
+				SetPageUptodate(page);
+			} else if (!PageUptodate(page)) {
+				zero_user_segment(page, 0, PAGE_SIZE);
+				SetPageUptodate(page);
+			}
+
+			ni_unlock(ni);
+			unlock_page(page);
+			return 0;
+		}
+	}
+
+	if (is_compressed(ni)) {
+		ni_lock(ni);
+		err = ni_readpage_cmpr(ni, page);
+		ni_unlock(ni);
+		return err;
+	}
+
+	/* normal + sparse files */
+	err = mpage_readpage(page, ntfs_get_block);
+out:
+	return err;
+}
+
+static void ntfs_readahead(struct readahead_control *rac)
+{
+	struct address_space *mapping = rac->mapping;
+	struct inode *inode = mapping->host;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	u64 valid;
+	loff_t pos;
+
+	if (ni_has_resident_data(ni)) {
+		/* no readahead for resident */
+		return;
+	}
+
+	if (is_compressed(ni)) {
+		/* no readahead for compressed */
+		return;
+	}
+
+	valid = ni->i_valid;
+	pos = readahead_pos(rac);
+
+	if (pos <= valid && valid < pos + readahead_length(rac)) {
+		/* range cross 'valid'. read it page by page */
+		return;
+	}
+
+	mpage_readahead(rac, ntfs_get_block);
+}
+
+static int ntfs_get_block_direct_IO_R(struct inode *inode, sector_t iblock,
+				      struct buffer_head *bh_result, int create)
+{
+	return ntfs_get_block_vbo(inode, (u64)iblock << inode->i_blkbits,
+				  bh_result, create, GET_BLOCK_DIRECT_IO_R);
+}
+
+static int ntfs_get_block_direct_IO_W(struct inode *inode, sector_t iblock,
+				      struct buffer_head *bh_result, int create)
+{
+	return ntfs_get_block_vbo(inode, (u64)iblock << inode->i_blkbits,
+				  bh_result, create, GET_BLOCK_DIRECT_IO_W);
+}
+
+static void ntfs_write_failed(struct address_space *mapping, loff_t to)
+{
+	struct inode *inode = mapping->host;
+
+	if (to > inode->i_size) {
+		truncate_pagecache(inode, inode->i_size);
+		ntfs_truncate_blocks(inode, inode->i_size);
+	}
+}
+
+static ssize_t ntfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
+{
+	struct file *file = iocb->ki_filp;
+	struct address_space *mapping = file->f_mapping;
+	struct inode *inode = mapping->host;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	size_t count = iov_iter_count(iter);
+	loff_t vbo = iocb->ki_pos;
+	loff_t end = vbo + count;
+	int wr = iov_iter_rw(iter) & WRITE;
+	const struct iovec *iov = iter->iov;
+	unsigned long nr_segs = iter->nr_segs;
+	loff_t valid;
+	ssize_t ret;
+
+	ret = blockdev_direct_IO(iocb, inode, iter,
+				 wr ? ntfs_get_block_direct_IO_W :
+				      ntfs_get_block_direct_IO_R);
+	valid = ni->i_valid;
+	if (wr) {
+		if (ret < 0)
+			ntfs_write_failed(mapping, end);
+		if (ret <= 0)
+			goto out;
+
+		vbo += ret;
+		if (vbo > valid && !S_ISBLK(inode->i_mode)) {
+			ni->i_valid = vbo;
+			mark_inode_dirty(inode);
+		}
+	} else if (vbo < valid && valid < end) {
+		/* fix page */
+		unsigned long uaddr = ~0ul;
+		struct page *page;
+		long i, npages;
+		size_t dvbo = valid - vbo;
+		size_t off = 0;
+
+		/*Find user address*/
+		for (i = 0; i < nr_segs; i++) {
+			if (off <= dvbo && dvbo < off + iov[i].iov_len) {
+				uaddr = (unsigned long)iov[i].iov_base + dvbo -
+					off;
+				break;
+			}
+			off += iov[i].iov_len;
+		}
+
+		if (uaddr == ~0ul)
+			goto fix_error;
+
+		npages = get_user_pages_unlocked(uaddr, 1, &page, FOLL_WRITE);
+
+		if (npages <= 0)
+			goto fix_error;
+
+		zero_user_segment(page, valid & (PAGE_SIZE - 1), PAGE_SIZE);
+		put_page(page);
+	}
+
+out:
+	return ret;
+fix_error:
+	ntfs_inode_warn(inode, "file garbage at 0x%llx", valid);
+	goto out;
+}
+
+int ntfs_set_size(struct inode *inode, u64 new_size)
+{
+	struct super_block *sb = inode->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	int err;
+
+	/* Check for maximum file size */
+	if (is_sparsed(ni) || is_compressed(ni)) {
+		if (new_size > sbi->maxbytes_sparse) {
+			err = -EFBIG;
+			goto out;
+		}
+	} else if (new_size > sbi->maxbytes) {
+		err = -EFBIG;
+		goto out;
+	}
+
+	ni_lock(ni);
+	down_write(&ni->file.run_lock);
+
+	err = attr_set_size(ni, ATTR_DATA, NULL, 0, &ni->file.run, new_size,
+			    &ni->i_valid, true, NULL);
+
+	up_write(&ni->file.run_lock);
+	ni_unlock(ni);
+
+	mark_inode_dirty(inode);
+
+out:
+	return err;
+}
+
+static int ntfs_writepage(struct page *page, struct writeback_control *wbc)
+{
+	return block_write_full_page(page, ntfs_get_block, wbc);
+}
+
+static int ntfs_writepages(struct address_space *mapping,
+			   struct writeback_control *wbc)
+{
+	return mpage_writepages(mapping, wbc, ntfs_get_block);
+}
+
+static int ntfs_get_block_write_begin(struct inode *inode, sector_t vbn,
+				      struct buffer_head *bh_result, int create)
+{
+	return ntfs_get_block_vbo(inode, (u64)vbn << inode->i_blkbits,
+				  bh_result, create, GET_BLOCK_WRITE_BEGIN);
+}
+
+static int ntfs_write_begin(struct file *file, struct address_space *mapping,
+			    loff_t pos, u32 len, u32 flags, struct page **pagep,
+			    void **fsdata)
+{
+	int err;
+
+	*pagep = NULL;
+
+	err = block_write_begin(mapping, pos, len, flags, pagep,
+				ntfs_get_block_write_begin);
+	if (err < 0)
+		ntfs_write_failed(mapping, pos + len);
+
+	return err;
+}
+
+/* address_space_operations::write_end */
+static int ntfs_write_end(struct file *file, struct address_space *mapping,
+			  loff_t pos, u32 len, u32 copied, struct page *page,
+			  void *fsdata)
+
+{
+	struct inode *inode = mapping->host;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	u64 valid = ni->i_valid;
+	int err;
+
+	err = generic_write_end(file, mapping, pos, len, copied, page, fsdata);
+
+	if (err < len)
+		ntfs_write_failed(mapping, pos + len);
+	if (err >= 0) {
+		bool dirty = false;
+
+		if (!(ni->std_fa & FILE_ATTRIBUTE_ARCHIVE)) {
+			inode->i_ctime = inode->i_mtime = current_time(inode);
+			ni->std_fa |= FILE_ATTRIBUTE_ARCHIVE;
+			dirty = true;
+		}
+
+		if (valid != ni->i_valid) {
+			/* ni->i_valid is changed in ntfs_get_block_vbo */
+			dirty = true;
+		}
+
+		if (dirty)
+			mark_inode_dirty(inode);
+	}
+
+	return err;
+}
+
+int reset_log_file(struct inode *inode)
+{
+	int err;
+	loff_t pos = 0;
+	u32 log_size = inode->i_size;
+	struct address_space *mapping = inode->i_mapping;
+
+	for (;;) {
+		u32 len;
+		void *kaddr;
+		struct page *page;
+
+		len = pos + PAGE_SIZE > log_size ? (log_size - pos) : PAGE_SIZE;
+
+		err = block_write_begin(mapping, pos, len, 0, &page,
+					ntfs_get_block_write_begin);
+		if (err)
+			goto out;
+
+		kaddr = kmap_atomic(page);
+		memset(kaddr, -1, len);
+		kunmap_atomic(kaddr);
+		flush_dcache_page(page);
+
+		err = block_write_end(NULL, mapping, pos, len, len, page, NULL);
+		if (err < 0)
+			goto out;
+		pos += len;
+
+		if (pos >= log_size)
+			break;
+		balance_dirty_pages_ratelimited(mapping);
+	}
+out:
+	mark_inode_dirty_sync(inode);
+
+	return err;
+}
+
+int ntfs_write_inode(struct inode *inode, struct writeback_control *wbc)
+{
+	return _ni_write_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
+}
+
+int ntfs_sync_inode(struct inode *inode)
+{
+	return _ni_write_inode(inode, 1);
+}
+
+/*
+ * helper function for ntfs_flush_inodes.  This writes both the inode
+ * and the file data blocks, waiting for in flight data blocks before
+ * the start of the call.  It does not wait for any io started
+ * during the call
+ */
+static int writeback_inode(struct inode *inode)
+{
+	int ret = sync_inode_metadata(inode, 0);
+
+	if (!ret)
+		ret = filemap_fdatawrite(inode->i_mapping);
+	return ret;
+}
+
+/*
+ * write data and metadata corresponding to i1 and i2.  The io is
+ * started but we do not wait for any of it to finish.
+ *
+ * filemap_flush is used for the block device, so if there is a dirty
+ * page for a block already in flight, we will not wait and start the
+ * io over again
+ */
+int ntfs_flush_inodes(struct super_block *sb, struct inode *i1,
+		      struct inode *i2)
+{
+	int ret = 0;
+
+	if (i1)
+		ret = writeback_inode(i1);
+	if (!ret && i2)
+		ret = writeback_inode(i2);
+	if (!ret)
+		ret = filemap_flush(sb->s_bdev->bd_inode->i_mapping);
+	return ret;
+}
+
+int inode_write_data(struct inode *inode, const void *data, size_t bytes)
+{
+	pgoff_t idx;
+
+	/* Write non resident data */
+	for (idx = 0; bytes; idx++) {
+		size_t op = bytes > PAGE_SIZE ? PAGE_SIZE : bytes;
+		struct page *page = ntfs_map_page(inode->i_mapping, idx);
+
+		if (IS_ERR(page))
+			return PTR_ERR(page);
+
+		lock_page(page);
+		WARN_ON(!PageUptodate(page));
+		ClearPageUptodate(page);
+
+		memcpy(page_address(page), data, op);
+
+		flush_dcache_page(page);
+		SetPageUptodate(page);
+		unlock_page(page);
+
+		ntfs_unmap_page(page);
+
+		bytes -= op;
+		data = Add2Ptr(data, PAGE_SIZE);
+	}
+	return 0;
+}
+
+/*
+ * number of bytes to for REPARSE_DATA_BUFFER(IO_REPARSE_TAG_SYMLINK)
+ * for unicode string of 'uni_len' length
+ */
+static inline u32 ntfs_reparse_bytes(u32 uni_len)
+{
+	/* header + unicode string + decorated unicode string */
+	return sizeof(short) * (2 * uni_len + 4) +
+	       offsetof(struct REPARSE_DATA_BUFFER,
+			SymbolicLinkReparseBuffer.PathBuffer);
+}
+
+static struct REPARSE_DATA_BUFFER *
+ntfs_create_reparse_buffer(struct ntfs_sb_info *sbi, const char *symname,
+			   u32 size, u16 *nsize)
+{
+	int i, err;
+	struct REPARSE_DATA_BUFFER *rp;
+	__le16 *rp_name;
+	typeof(rp->SymbolicLinkReparseBuffer) *rs;
+
+	rp = ntfs_alloc(ntfs_reparse_bytes(2 * size + 2), 1);
+	if (!rp)
+		return ERR_PTR(-ENOMEM);
+
+	rs = &rp->SymbolicLinkReparseBuffer;
+	rp_name = rs->PathBuffer;
+
+	/* Convert link name to utf16 */
+	err = ntfs_nls_to_utf16(sbi, symname, size,
+				(struct cpu_str *)(rp_name - 1), 2 * size,
+				UTF16_LITTLE_ENDIAN);
+	if (err < 0)
+		goto out;
+
+	/* err = the length of unicode name of symlink */
+	*nsize = ntfs_reparse_bytes(err);
+
+	if (*nsize > sbi->reparse.max_size) {
+		err = -EFBIG;
+		goto out;
+	}
+
+	/* translate linux '/' into windows '\' */
+	for (i = 0; i < err; i++) {
+		if (rp_name[i] == cpu_to_le16('/'))
+			rp_name[i] = cpu_to_le16('\\');
+	}
+
+	rp->ReparseTag = IO_REPARSE_TAG_SYMLINK;
+	rp->ReparseDataLength =
+		cpu_to_le16(*nsize - offsetof(struct REPARSE_DATA_BUFFER,
+					      SymbolicLinkReparseBuffer));
+
+	/* PrintName + SubstituteName */
+	rs->SubstituteNameOffset = cpu_to_le16(sizeof(short) * err);
+	rs->SubstituteNameLength = cpu_to_le16(sizeof(short) * err + 8);
+	rs->PrintNameLength = rs->SubstituteNameOffset;
+
+	/*
+	 * TODO: use relative path if possible to allow windows to parse this path
+	 * 0-absolute path 1- relative path (SYMLINK_FLAG_RELATIVE)
+	 */
+	rs->Flags = 0;
+
+	memmove(rp_name + err + 4, rp_name, sizeof(short) * err);
+
+	/* decorate SubstituteName */
+	rp_name += err;
+	rp_name[0] = cpu_to_le16('\\');
+	rp_name[1] = cpu_to_le16('?');
+	rp_name[2] = cpu_to_le16('?');
+	rp_name[3] = cpu_to_le16('\\');
+
+	return rp;
+out:
+	ntfs_free(rp);
+	return ERR_PTR(err);
+}
+
+int ntfs_create_inode(struct inode *dir, struct dentry *dentry,
+		      const struct cpu_str *uni, umode_t mode, dev_t dev,
+		      const char *symname, u32 size, int excl,
+		      struct ntfs_fnd *fnd, struct inode **new_inode)
+{
+	int err;
+	struct super_block *sb = dir->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	const struct qstr *name = &dentry->d_name;
+	CLST ino = 0;
+	struct ntfs_inode *dir_ni = ntfs_i(dir);
+	struct ntfs_inode *ni = NULL;
+	struct inode *inode = NULL;
+	struct ATTRIB *attr;
+	struct ATTR_STD_INFO5 *std5;
+	struct ATTR_FILE_NAME *fname;
+	struct MFT_REC *rec;
+	u32 asize, dsize, sd_size;
+	enum FILE_ATTRIBUTE fa;
+	__le32 security_id = SECURITY_ID_INVALID;
+	CLST vcn;
+	const void *sd;
+	u16 t16, nsize = 0, aid = 0;
+	struct INDEX_ROOT *root, *dir_root;
+	struct NTFS_DE *e, *new_de = NULL;
+	struct REPARSE_DATA_BUFFER *rp = NULL;
+	bool is_dir = S_ISDIR(mode);
+	bool is_link = S_ISLNK(mode);
+	bool rp_inserted = false;
+	bool is_sp = S_ISCHR(mode) || S_ISBLK(mode) || S_ISFIFO(mode) ||
+		     S_ISSOCK(mode);
+
+	if (is_sp)
+		return -EOPNOTSUPP;
+
+	dir_root = indx_get_root(&dir_ni->dir, dir_ni, NULL, NULL);
+	if (!dir_root)
+		return -EINVAL;
+
+	if (is_dir) {
+		/* use parent's directory attributes */
+		fa = dir_ni->std_fa | FILE_ATTRIBUTE_DIRECTORY |
+		     FILE_ATTRIBUTE_ARCHIVE;
+	} else if (is_link) {
+		/* It is good idea that link should be the same type (file/dir) as target */
+		fa = FILE_ATTRIBUTE_REPARSE_POINT;
+
+		/*
+		 * linux: there are dir/file/symlink and so on
+		 * NTFS: symlinks are "dir + reparse" or "file + reparse"
+		 * It is good idea to create:
+		 * dir + reparse if 'symname' points to directory
+		 * or
+		 * file + reparse if 'symname' points to file
+		 * Unfortunately kern_path hangs if symname contains 'dir'
+		 */
+
+		/*
+		 *	struct path path;
+		 *
+		 *	if (!kern_path(symname, LOOKUP_FOLLOW, &path)){
+		 *		struct inode *target = d_inode(path.dentry);
+		 *
+		 *		if (S_ISDIR(target->i_mode))
+		 *			fa |= FILE_ATTRIBUTE_DIRECTORY;
+		 *		// if ( target->i_sb == sb ){
+		 *		//	use relative path?
+		 *		// }
+		 *		path_put(&path);
+		 *	}
+		 */
+	} else if (sbi->options.sparse) {
+		/* sparsed regular file, cause option 'sparse' */
+		fa = FILE_ATTRIBUTE_SPARSE_FILE | FILE_ATTRIBUTE_ARCHIVE;
+	} else if (dir_ni->std_fa & FILE_ATTRIBUTE_COMPRESSED) {
+		/* compressed regular file, if parent is compressed */
+		fa = FILE_ATTRIBUTE_COMPRESSED | FILE_ATTRIBUTE_ARCHIVE;
+	} else {
+		/* regular file, default attributes */
+		fa = FILE_ATTRIBUTE_ARCHIVE;
+	}
+
+	if (!(mode & 0222))
+		fa |= FILE_ATTRIBUTE_READONLY;
+
+	/* allocate PATH_MAX bytes */
+	new_de = __getname();
+	if (!new_de) {
+		err = -ENOMEM;
+		goto out1;
+	}
+
+	/*mark rw ntfs as dirty. it will be cleared at umount*/
+	ntfs_set_state(sbi, NTFS_DIRTY_DIRTY);
+
+	/* Step 1: allocate and fill new mft record */
+	err = ntfs_look_free_mft(sbi, &ino, false, NULL, NULL);
+	if (err)
+		goto out2;
+
+	ni = ntfs_new_inode(sbi, ino, fa & FILE_ATTRIBUTE_DIRECTORY);
+	if (IS_ERR(ni)) {
+		err = PTR_ERR(ni);
+		ni = NULL;
+		goto out3;
+	}
+	inode = &ni->vfs_inode;
+
+	inode->i_atime = inode->i_mtime = inode->i_ctime = ni->i_crtime =
+		current_time(inode);
+
+	rec = ni->mi.mrec;
+	rec->hard_links = cpu_to_le16(1);
+	attr = Add2Ptr(rec, le16_to_cpu(rec->attr_off));
+
+	/* Get default security id */
+	sd = s_default_security;
+	sd_size = sizeof(s_default_security);
+
+	if (is_ntfs3(sbi)) {
+		security_id = dir_ni->std_security_id;
+		if (le32_to_cpu(security_id) < SECURITY_ID_FIRST) {
+			security_id = sbi->security.def_security_id;
+
+			if (security_id == SECURITY_ID_INVALID &&
+			    !ntfs_insert_security(sbi, sd, sd_size,
+						  &security_id, NULL))
+				sbi->security.def_security_id = security_id;
+		}
+	}
+
+	/* Insert standard info */
+	std5 = Add2Ptr(attr, SIZEOF_RESIDENT);
+
+	if (security_id == SECURITY_ID_INVALID) {
+		dsize = sizeof(struct ATTR_STD_INFO);
+	} else {
+		dsize = sizeof(struct ATTR_STD_INFO5);
+		std5->security_id = security_id;
+		ni->std_security_id = security_id;
+	}
+	asize = SIZEOF_RESIDENT + dsize;
+
+	attr->type = ATTR_STD;
+	attr->size = cpu_to_le32(asize);
+	attr->id = cpu_to_le16(aid++);
+	attr->res.data_off = SIZEOF_RESIDENT_LE;
+	attr->res.data_size = cpu_to_le32(dsize);
+
+	std5->cr_time = std5->m_time = std5->c_time = std5->a_time =
+		kernel2nt(&inode->i_atime);
+
+	ni->std_fa = fa;
+	std5->fa = fa;
+
+	attr = Add2Ptr(attr, asize);
+
+	/* Insert file name */
+	err = fill_name_de(sbi, new_de, name, uni);
+	if (err)
+		goto out4;
+
+	fname = (struct ATTR_FILE_NAME *)(new_de + 1);
+
+	new_de->ref.low = cpu_to_le32(ino);
+#ifdef NTFS3_64BIT_CLUSTER
+	new_de->ref.high = cpu_to_le16(ino >> 32);
+	fname->home.high = cpu_to_le16(dir->i_ino >> 32);
+#endif
+	new_de->ref.seq = rec->seq;
+
+	fname->home.low = cpu_to_le32(dir->i_ino & 0xffffffff);
+	fname->home.seq = dir_ni->mi.mrec->seq;
+
+	fname->dup.cr_time = fname->dup.m_time = fname->dup.c_time =
+		fname->dup.a_time = std5->cr_time;
+	fname->dup.alloc_size = fname->dup.data_size = 0;
+	fname->dup.fa = std5->fa;
+	fname->dup.ea_size = fname->dup.reparse = 0;
+
+	dsize = le16_to_cpu(new_de->key_size);
+	asize = QuadAlign(SIZEOF_RESIDENT + dsize);
+
+	attr->type = ATTR_NAME;
+	attr->size = cpu_to_le32(asize);
+	attr->res.data_off = SIZEOF_RESIDENT_LE;
+	attr->res.flags = RESIDENT_FLAG_INDEXED;
+	attr->id = cpu_to_le16(aid++);
+	attr->res.data_size = cpu_to_le32(dsize);
+	memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), fname, dsize);
+
+	attr = Add2Ptr(attr, asize);
+
+	if (security_id == SECURITY_ID_INVALID) {
+		/* Insert security attribute */
+		asize = SIZEOF_RESIDENT + QuadAlign(sd_size);
+
+		attr->type = ATTR_SECURE;
+		attr->size = cpu_to_le32(asize);
+		attr->id = cpu_to_le16(aid++);
+		attr->res.data_off = SIZEOF_RESIDENT_LE;
+		attr->res.data_size = cpu_to_le32(sd_size);
+		memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), sd, sd_size);
+
+		attr = Add2Ptr(attr, asize);
+	}
+
+	if (fa & FILE_ATTRIBUTE_DIRECTORY) {
+		/*
+		 * regular directory or symlink to directory
+		 * Create root attribute
+		 */
+		dsize = sizeof(struct INDEX_ROOT) + sizeof(struct NTFS_DE);
+		asize = sizeof(I30_NAME) + SIZEOF_RESIDENT + dsize;
+
+		attr->type = ATTR_ROOT;
+		attr->size = cpu_to_le32(asize);
+		attr->id = cpu_to_le16(aid++);
+
+		attr->name_len = ARRAY_SIZE(I30_NAME);
+		attr->name_off = SIZEOF_RESIDENT_LE;
+		attr->res.data_off =
+			cpu_to_le16(sizeof(I30_NAME) + SIZEOF_RESIDENT);
+		attr->res.data_size = cpu_to_le32(dsize);
+		memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), I30_NAME,
+		       sizeof(I30_NAME));
+
+		root = Add2Ptr(attr, sizeof(I30_NAME) + SIZEOF_RESIDENT);
+		memcpy(root, dir_root, offsetof(struct INDEX_ROOT, ihdr));
+		root->ihdr.de_off =
+			cpu_to_le32(sizeof(struct INDEX_HDR)); // 0x10
+		root->ihdr.used = cpu_to_le32(sizeof(struct INDEX_HDR) +
+					      sizeof(struct NTFS_DE));
+		root->ihdr.total = root->ihdr.used;
+
+		e = Add2Ptr(root, sizeof(struct INDEX_ROOT));
+		e->size = cpu_to_le16(sizeof(struct NTFS_DE));
+		e->flags = NTFS_IE_LAST;
+	} else if (is_link) {
+		/*
+		 * symlink to file
+		 * Create empty resident data attribute
+		 */
+		asize = SIZEOF_RESIDENT;
+
+		/* insert empty ATTR_DATA */
+		attr->type = ATTR_DATA;
+		attr->size = cpu_to_le32(SIZEOF_RESIDENT);
+		attr->id = cpu_to_le16(aid++);
+		attr->name_off = SIZEOF_RESIDENT_LE;
+		attr->res.data_off = SIZEOF_RESIDENT_LE;
+	} else {
+		/*
+		 * regular file
+		 * Create empty nonresident data attribute
+		 */
+		attr->type = ATTR_DATA;
+		attr->id = cpu_to_le16(aid++);
+		/* Create non resident data attribute */
+		attr->non_res = 1;
+		attr->nres.evcn = cpu_to_le64(-1ll);
+		if (fa & FILE_ATTRIBUTE_SPARSE_FILE) {
+			attr->size = cpu_to_le32(SIZEOF_NONRESIDENT_EX + 8);
+			attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
+			attr->flags = ATTR_FLAG_SPARSED;
+			asize = SIZEOF_NONRESIDENT_EX + 8;
+		} else if (fa & FILE_ATTRIBUTE_COMPRESSED) {
+			attr->size = cpu_to_le32(SIZEOF_NONRESIDENT_EX + 8);
+			attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
+			attr->flags = ATTR_FLAG_COMPRESSED;
+			attr->nres.c_unit = COMPRESSION_UNIT;
+			asize = SIZEOF_NONRESIDENT_EX + 8;
+		} else {
+			attr->size = cpu_to_le32(SIZEOF_NONRESIDENT + 8);
+			attr->name_off = SIZEOF_NONRESIDENT_LE;
+			asize = SIZEOF_NONRESIDENT + 8;
+		}
+		attr->nres.run_off = attr->name_off;
+	}
+
+	if (is_dir) {
+		ni->ni_flags |= NI_FLAG_DIR;
+		err = indx_init(&ni->dir, sbi, attr, INDEX_MUTEX_I30);
+		if (err)
+			goto out4;
+	} else if (is_link) {
+		rp = ntfs_create_reparse_buffer(sbi, symname, size, &nsize);
+
+		if (IS_ERR(rp)) {
+			err = PTR_ERR(rp);
+			rp = NULL;
+			goto out4;
+		}
+
+		/*
+		 * Insert ATTR_REPARSE
+		 */
+		attr = Add2Ptr(attr, asize);
+		attr->type = ATTR_REPARSE;
+		attr->id = cpu_to_le16(aid++);
+
+		/* resident or non resident? */
+		asize = QuadAlign(SIZEOF_RESIDENT + nsize);
+		t16 = PtrOffset(rec, attr);
+
+		if (asize + t16 + 8 > sbi->record_size) {
+			CLST alen;
+			CLST clst = bytes_to_cluster(sbi, nsize);
+
+			/* bytes per runs */
+			t16 = sbi->record_size - t16 - SIZEOF_NONRESIDENT;
+
+			attr->non_res = 1;
+			attr->nres.evcn = cpu_to_le64(clst - 1);
+			attr->name_off = SIZEOF_NONRESIDENT_LE;
+			attr->nres.run_off = attr->name_off;
+			attr->nres.data_size = cpu_to_le64(nsize);
+			attr->nres.valid_size = attr->nres.data_size;
+			attr->nres.alloc_size =
+				cpu_to_le64(ntfs_up_cluster(sbi, nsize));
+
+			err = attr_allocate_clusters(sbi, &ni->file.run, 0, 0,
+						     clst, NULL, 0, &alen, 0,
+						     NULL);
+			if (err)
+				goto out5;
+
+			err = run_pack(&ni->file.run, 0, clst,
+				       Add2Ptr(attr, SIZEOF_NONRESIDENT), t16,
+				       &vcn);
+			if (err < 0)
+				goto out5;
+
+			if (vcn != clst) {
+				err = -EINVAL;
+				goto out5;
+			}
+
+			asize = SIZEOF_NONRESIDENT + QuadAlign(err);
+			inode->i_size = nsize;
+		} else {
+			attr->res.data_off = SIZEOF_RESIDENT_LE;
+			attr->res.data_size = cpu_to_le32(nsize);
+			memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), rp, nsize);
+			inode->i_size = nsize;
+			nsize = 0;
+		}
+
+		attr->size = cpu_to_le32(asize);
+
+		err = ntfs_insert_reparse(sbi, IO_REPARSE_TAG_SYMLINK,
+					  &new_de->ref);
+		if (err)
+			goto out5;
+
+		rp_inserted = true;
+	}
+
+	attr = Add2Ptr(attr, asize);
+	attr->type = ATTR_END;
+
+	rec->used = cpu_to_le32(PtrOffset(rec, attr) + 8);
+	rec->next_attr_id = cpu_to_le16(aid);
+
+	/* Step 2: Add new name in index */
+	err = indx_insert_entry(&dir_ni->dir, dir_ni, new_de, sbi, fnd);
+	if (err)
+		goto out6;
+
+	/* Update current directory record */
+	mark_inode_dirty(dir);
+
+	/* Fill vfs inode fields */
+	inode->i_uid = sbi->options.uid ? sbi->options.fs_uid : current_fsuid();
+	inode->i_gid =
+		sbi->options.gid ?
+			sbi->options.fs_gid :
+			(dir->i_mode & S_ISGID) ? dir->i_gid : current_fsgid();
+	inode->i_generation = le16_to_cpu(rec->seq);
+
+	dir->i_mtime = dir->i_ctime = inode->i_atime;
+
+	if (is_dir) {
+		if (dir->i_mode & S_ISGID)
+			mode |= S_ISGID;
+		inode->i_op = &ntfs_dir_inode_operations;
+		inode->i_fop = &ntfs_dir_operations;
+	} else if (is_link) {
+		inode->i_op = &ntfs_link_inode_operations;
+		inode->i_fop = NULL;
+		inode->i_mapping->a_ops = &ntfs_aops;
+	} else {
+		inode->i_op = &ntfs_file_inode_operations;
+		inode->i_fop = &ntfs_file_operations;
+		inode->i_mapping->a_ops =
+			is_compressed(ni) ? &ntfs_aops_cmpr : &ntfs_aops;
+		init_rwsem(&ni->file.run_lock);
+	}
+
+	inode->i_mode = mode;
+
+	if (!is_link && (sb->s_flags & SB_POSIXACL)) {
+		err = ntfs_init_acl(inode, dir);
+		if (err)
+			goto out6;
+	} else {
+		inode->i_flags |= S_NOSEC;
+	}
+
+	/* Write non resident data */
+	if (nsize) {
+		err = ntfs_sb_write_run(sbi, &ni->file.run, 0, rp, nsize);
+		if (err)
+			goto out7;
+	}
+
+	/* call 'd_instantiate' after inode->i_op is set but before finish_open */
+	d_instantiate(dentry, inode);
+
+	mark_inode_dirty(inode);
+	mark_inode_dirty(dir);
+
+	/* normal exit */
+	goto out2;
+
+out7:
+
+	/* undo 'indx_insert_entry' */
+	indx_delete_entry(&dir_ni->dir, dir_ni, new_de + 1,
+			  le16_to_cpu(new_de->key_size), sbi);
+out6:
+	if (rp_inserted)
+		ntfs_remove_reparse(sbi, IO_REPARSE_TAG_SYMLINK, &new_de->ref);
+
+out5:
+	if (is_dir || run_is_empty(&ni->file.run))
+		goto out4;
+
+	run_deallocate(sbi, &ni->file.run, false);
+
+out4:
+	clear_rec_inuse(rec);
+	clear_nlink(inode);
+	ni->mi.dirty = false;
+	discard_new_inode(inode);
+out3:
+	ntfs_mark_rec_free(sbi, ino);
+
+out2:
+	__putname(new_de);
+	ntfs_free(rp);
+
+out1:
+	if (err)
+		return err;
+
+	unlock_new_inode(inode);
+
+	*new_inode = inode;
+	return 0;
+}
+
+int ntfs_link_inode(struct inode *inode, struct dentry *dentry)
+{
+	int err;
+	struct inode *dir = d_inode(dentry->d_parent);
+	struct ntfs_inode *dir_ni = ntfs_i(dir);
+	struct ntfs_inode *ni = ntfs_i(inode);
+	struct super_block *sb = inode->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	const struct qstr *name = &dentry->d_name;
+	struct NTFS_DE *new_de = NULL;
+	struct ATTR_FILE_NAME *fname;
+	struct ATTRIB *attr;
+	u16 key_size;
+	struct INDEX_ROOT *dir_root;
+
+	dir_root = indx_get_root(&dir_ni->dir, dir_ni, NULL, NULL);
+	if (!dir_root)
+		return -EINVAL;
+
+	/* allocate PATH_MAX bytes */
+	new_de = __getname();
+	if (!new_de)
+		return -ENOMEM;
+
+	/*mark rw ntfs as dirty. it will be cleared at umount*/
+	ntfs_set_state(ni->mi.sbi, NTFS_DIRTY_DIRTY);
+
+	// Insert file name
+	err = fill_name_de(sbi, new_de, name, NULL);
+	if (err)
+		goto out;
+
+	key_size = le16_to_cpu(new_de->key_size);
+	fname = (struct ATTR_FILE_NAME *)(new_de + 1);
+
+	err = ni_insert_resident(ni, key_size, ATTR_NAME, NULL, 0, &attr, NULL);
+	if (err)
+		goto out;
+
+	new_de->ref.low = cpu_to_le32(inode->i_ino);
+#ifdef NTFS3_64BIT_CLUSTER
+	new_de->ref.high = cpu_to_le16(inode->i_ino >> 32);
+	fname->home.high = cpu_to_le16(dir->i_ino >> 32);
+#endif
+	new_de->ref.seq = ni->mi.mrec->seq;
+
+	fname->home.low = cpu_to_le32(dir->i_ino & 0xffffffff);
+	fname->home.seq = dir_ni->mi.mrec->seq;
+
+	fname->dup.cr_time = fname->dup.m_time = fname->dup.c_time =
+		fname->dup.a_time = kernel2nt(&inode->i_ctime);
+	fname->dup.alloc_size = fname->dup.data_size = 0;
+	fname->dup.fa = ni->std_fa;
+	fname->dup.ea_size = fname->dup.reparse = 0;
+
+	memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), fname, key_size);
+
+	err = indx_insert_entry(&dir_ni->dir, dir_ni, new_de, sbi, NULL);
+	if (err)
+		goto out;
+
+	le16_add_cpu(&ni->mi.mrec->hard_links, 1);
+	ni->mi.dirty = true;
+
+out:
+	__putname(new_de);
+	return err;
+}
+
+/*
+ * ntfs_unlink_inode
+ *
+ * inode_operations::unlink
+ * inode_operations::rmdir
+ */
+int ntfs_unlink_inode(struct inode *dir, const struct dentry *dentry)
+{
+	int err;
+	struct super_block *sb = dir->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct inode *inode = d_inode(dentry);
+	struct ntfs_inode *ni = ntfs_i(inode);
+	const struct qstr *name = &dentry->d_name;
+	struct ntfs_inode *dir_ni = ntfs_i(dir);
+	struct ntfs_index *indx = &dir_ni->dir;
+	struct cpu_str *uni = NULL;
+	struct ATTR_FILE_NAME *fname;
+	u8 name_type;
+	struct ATTR_LIST_ENTRY *le;
+	struct MFT_REF ref;
+	bool is_dir = S_ISDIR(inode->i_mode);
+	struct INDEX_ROOT *dir_root;
+
+	dir_root = indx_get_root(indx, dir_ni, NULL, NULL);
+	if (!dir_root)
+		return -EINVAL;
+
+	ni_lock(ni);
+
+	if (is_dir && !dir_is_empty(inode)) {
+		err = -ENOTEMPTY;
+		goto out1;
+	}
+
+	if (ntfs_is_meta_file(sbi, inode->i_ino)) {
+		err = -EINVAL;
+		goto out1;
+	}
+
+	/* allocate PATH_MAX bytes */
+	uni = __getname();
+	if (!uni) {
+		err = -ENOMEM;
+		goto out1;
+	}
+
+	/* Convert input string to unicode */
+	err = ntfs_nls_to_utf16(sbi, name->name, name->len, uni, NTFS_NAME_LEN,
+				UTF16_HOST_ENDIAN);
+	if (err < 0)
+		goto out4;
+
+	le = NULL;
+
+	/*mark rw ntfs as dirty. it will be cleared at umount*/
+	ntfs_set_state(sbi, NTFS_DIRTY_DIRTY);
+
+	/* find name in record */
+#ifdef NTFS3_64BIT_CLUSTER
+	ref.low = cpu_to_le32(dir->i_ino & 0xffffffff);
+	ref.high = cpu_to_le16(dir->i_ino >> 32);
+#else
+	ref.low = cpu_to_le32(dir->i_ino & 0xffffffff);
+	ref.high = 0;
+#endif
+	ref.seq = dir_ni->mi.mrec->seq;
+
+	fname = ni_fname_name(ni, uni, &ref, &le);
+	if (!fname) {
+		err = -ENOENT;
+		goto out3;
+	}
+
+	name_type = paired_name(fname->type);
+
+	err = indx_delete_entry(indx, dir_ni, fname, fname_full_size(fname),
+				sbi);
+	if (err)
+		goto out4;
+
+	/* Then remove name from mft */
+	ni_remove_attr_le(ni, attr_from_name(fname), le);
+
+	le16_add_cpu(&ni->mi.mrec->hard_links, -1);
+	ni->mi.dirty = true;
+
+	if (name_type != FILE_NAME_POSIX) {
+		/* Now we should delete name by type */
+		fname = ni_fname_type(ni, name_type, &le);
+		if (fname) {
+			err = indx_delete_entry(indx, dir_ni, fname,
+						fname_full_size(fname), sbi);
+			if (err)
+				goto out4;
+
+			ni_remove_attr_le(ni, attr_from_name(fname), le);
+
+			le16_add_cpu(&ni->mi.mrec->hard_links, -1);
+		}
+	}
+
+out4:
+	switch (err) {
+	case 0:
+		drop_nlink(inode);
+	case -ENOTEMPTY:
+	case -ENOSPC:
+	case -EROFS:
+		break;
+	default:
+		make_bad_inode(inode);
+	}
+
+	dir->i_mtime = dir->i_ctime = current_time(dir);
+	mark_inode_dirty(dir);
+	inode->i_ctime = dir->i_ctime;
+	if (inode->i_nlink)
+		mark_inode_dirty(inode);
+
+out3:
+	__putname(uni);
+out1:
+	ni_unlock(ni);
+	return err;
+}
+
+void ntfs_evict_inode(struct inode *inode)
+{
+	truncate_inode_pages_final(&inode->i_data);
+
+	if (inode->i_nlink)
+		_ni_write_inode(inode, inode_needs_sync(inode));
+
+	invalidate_inode_buffers(inode);
+	clear_inode(inode);
+
+	ni_clear(ntfs_i(inode));
+}
+
+static noinline int ntfs_readlink_hlp(struct inode *inode, char *buffer,
+				      int buflen)
+{
+	int i, err = 0;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	struct super_block *sb = inode->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	u64 i_size = inode->i_size;
+	u16 nlen = 0;
+	void *to_free = NULL;
+	struct REPARSE_DATA_BUFFER *rp;
+	struct le_str *uni;
+	struct ATTRIB *attr;
+
+	/* Reparse data present. Try to parse it */
+	static_assert(!offsetof(struct REPARSE_DATA_BUFFER, ReparseTag));
+	static_assert(sizeof(u32) == sizeof(rp->ReparseTag));
+
+	*buffer = 0;
+
+	/* Read into temporal buffer */
+	if (i_size > sbi->reparse.max_size || i_size <= sizeof(u32)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	attr = ni_find_attr(ni, NULL, NULL, ATTR_REPARSE, NULL, 0, NULL, NULL);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (!attr->non_res) {
+		rp = resident_data_ex(attr, i_size);
+		if (!rp) {
+			err = -EINVAL;
+			goto out;
+		}
+	} else {
+		rp = ntfs_alloc(i_size, 0);
+		if (!rp) {
+			err = -ENOMEM;
+			goto out;
+		}
+		to_free = rp;
+		err = ntfs_read_run_nb(sbi, &ni->file.run, 0, rp, i_size, NULL);
+		if (err)
+			goto out;
+	}
+
+	err = -EINVAL;
+
+	/* Microsoft Tag */
+	switch (rp->ReparseTag) {
+	case IO_REPARSE_TAG_MOUNT_POINT:
+		/* Mount points and junctions */
+		/* Can we use 'Rp->MountPointReparseBuffer.PrintNameLength'? */
+		if (i_size <= offsetof(struct REPARSE_DATA_BUFFER,
+				       MountPointReparseBuffer.PathBuffer))
+			goto out;
+		uni = Add2Ptr(rp,
+			      offsetof(struct REPARSE_DATA_BUFFER,
+				       MountPointReparseBuffer.PathBuffer) +
+				      le16_to_cpu(rp->MountPointReparseBuffer
+							  .PrintNameOffset) -
+				      2);
+		nlen = le16_to_cpu(rp->MountPointReparseBuffer.PrintNameLength);
+		break;
+
+	case IO_REPARSE_TAG_SYMLINK:
+		/* FolderSymbolicLink */
+		/* Can we use 'Rp->SymbolicLinkReparseBuffer.PrintNameLength'? */
+		if (i_size <= offsetof(struct REPARSE_DATA_BUFFER,
+				       SymbolicLinkReparseBuffer.PathBuffer))
+			goto out;
+		uni = Add2Ptr(rp,
+			      offsetof(struct REPARSE_DATA_BUFFER,
+				       SymbolicLinkReparseBuffer.PathBuffer) +
+				      le16_to_cpu(rp->SymbolicLinkReparseBuffer
+							  .PrintNameOffset) -
+				      2);
+		nlen = le16_to_cpu(
+			rp->SymbolicLinkReparseBuffer.PrintNameLength);
+		break;
+
+	case IO_REPARSE_TAG_CLOUD:
+	case IO_REPARSE_TAG_CLOUD_1:
+	case IO_REPARSE_TAG_CLOUD_2:
+	case IO_REPARSE_TAG_CLOUD_3:
+	case IO_REPARSE_TAG_CLOUD_4:
+	case IO_REPARSE_TAG_CLOUD_5:
+	case IO_REPARSE_TAG_CLOUD_6:
+	case IO_REPARSE_TAG_CLOUD_7:
+	case IO_REPARSE_TAG_CLOUD_8:
+	case IO_REPARSE_TAG_CLOUD_9:
+	case IO_REPARSE_TAG_CLOUD_A:
+	case IO_REPARSE_TAG_CLOUD_B:
+	case IO_REPARSE_TAG_CLOUD_C:
+	case IO_REPARSE_TAG_CLOUD_D:
+	case IO_REPARSE_TAG_CLOUD_E:
+	case IO_REPARSE_TAG_CLOUD_F:
+		err = sizeof("OneDrive") - 1;
+		if (err > buflen)
+			err = buflen;
+		memcpy(buffer, "OneDrive", err);
+		goto out;
+
+	default:
+		if (IsReparseTagMicrosoft(rp->ReparseTag))
+			goto out;
+		if (!IsReparseTagNameSurrogate(rp->ReparseTag) ||
+		    i_size <= sizeof(struct REPARSE_POINT)) {
+			goto out;
+		}
+
+		/* Users tag */
+		uni = Add2Ptr(rp, sizeof(struct REPARSE_POINT) - 2);
+		nlen = le16_to_cpu(rp->ReparseDataLength) -
+		       sizeof(struct REPARSE_POINT);
+	}
+
+	/* Convert nlen from bytes to UNICODE chars */
+	nlen >>= 1;
+
+	/* Check that name is available */
+	if (!nlen || &uni->name[nlen] > (__le16 *)Add2Ptr(rp, i_size))
+		goto out;
+
+	/* If name is already zero terminated then truncate it now */
+	if (!uni->name[nlen - 1])
+		nlen -= 1;
+	uni->len = nlen;
+
+	err = ntfs_utf16_to_nls(sbi, uni, buffer, buflen);
+
+	if (err < 0)
+		goto out;
+
+	/* translate windows '\' into linux '/' */
+	for (i = 0; i < err; i++) {
+		if (buffer[i] == '\\')
+			buffer[i] = '/';
+	}
+
+	/* Always set last zero */
+	buffer[err] = 0;
+out:
+	ntfs_free(to_free);
+	return err;
+}
+
+static const char *ntfs_get_link(struct dentry *de, struct inode *inode,
+				 struct delayed_call *done)
+{
+	int err;
+	char *ret;
+
+	if (!de)
+		return ERR_PTR(-ECHILD);
+
+	ret = kmalloc(PAGE_SIZE, GFP_NOFS);
+	if (!ret)
+		return ERR_PTR(-ENOMEM);
+
+	err = ntfs_readlink_hlp(inode, ret, PAGE_SIZE);
+	if (err < 0) {
+		kfree(ret);
+		return ERR_PTR(err);
+	}
+
+	set_delayed_call(done, kfree_link, ret);
+
+	return ret;
+}
+
+const struct inode_operations ntfs_link_inode_operations = {
+	.get_link = ntfs_get_link,
+	.setattr = ntfs_setattr,
+	.listxattr = ntfs_listxattr,
+	.permission = ntfs_permission,
+	.get_acl = ntfs_get_acl,
+	.set_acl = ntfs_set_acl,
+};
+
+const struct address_space_operations ntfs_aops = {
+	.readpage = ntfs_readpage,
+	.readahead = ntfs_readahead,
+	.writepage = ntfs_writepage,
+	.writepages = ntfs_writepages,
+	.write_begin = ntfs_write_begin,
+	.write_end = ntfs_write_end,
+	.direct_IO = ntfs_direct_IO,
+	.bmap = ntfs_bmap,
+};
+
+const struct address_space_operations ntfs_aops_cmpr = {
+	.readpage = ntfs_readpage,
+	.readahead = ntfs_readahead,
+};
diff --git a/fs/ntfs3/super.c b/fs/ntfs3/super.c
new file mode 100644
index 000000000000..433fc4721c4e
--- /dev/null
+++ b/fs/ntfs3/super.c
@@ -0,0 +1,1479 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ *
+ *                 terminology
+ *
+ * vcn - virtual cluster number - offset inside the file in clusters
+ * vbo - virtual byte offset    - offset inside the file in bytes
+ * lcn - logical cluster number - 0 based cluster in clusters heap
+ * lbo - logical byte offset    - absolute position inside volume
+ *
+ */
+
+#include <linux/backing-dev.h>
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/exportfs.h>
+#include <linux/fs.h>
+#include <linux/iversion.h>
+#include <linux/module.h>
+#include <linux/nls.h>
+#include <linux/parser.h>
+#include <linux/seq_file.h>
+#include <linux/statfs.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+#ifdef CONFIG_PRINTK
+/*
+ * Trace warnings/notices/errors
+ * Thanks Joe Perches <joe@perches.com> for implementation
+ */
+void ntfs_printk(const struct super_block *sb, const char *fmt, ...)
+{
+	struct va_format vaf;
+	va_list args;
+	int level;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+
+	/*should we use different ratelimits for warnings/notices/errors? */
+	if (!___ratelimit(&sbi->msg_ratelimit, "ntfs3"))
+		return;
+
+	va_start(args, fmt);
+
+	level = printk_get_level(fmt);
+	vaf.fmt = printk_skip_level(fmt);
+	vaf.va = &args;
+	printk("%c%cntfs3: %s: %pV\n", KERN_SOH_ASCII, level, sb->s_id, &vaf);
+
+	va_end(args);
+}
+
+/* print warnings/notices/errors about inode using name or inode number */
+void ntfs_inode_printk(struct inode *inode, const char *fmt, ...)
+{
+	struct super_block *sb = inode->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct dentry *dentry;
+	char *name;
+	va_list args;
+	struct va_format vaf;
+	int level;
+	/*not necessary to print full inode name. just for information */
+	const u32 name_len = 63;
+
+	if (!___ratelimit(&sbi->msg_ratelimit, "ntfs3"))
+		return;
+
+	name = ntfs_alloc(name_len + 1, 0);
+	if (!name)
+		return;
+
+	dentry = d_find_alias(inode);
+	if (dentry) {
+		/*we can get the name of inode*/
+		spin_lock(&dentry->d_lock);
+		snprintf(name, name_len, "%s", dentry->d_name.name);
+		spin_unlock(&dentry->d_lock);
+		dput(dentry);
+		name[name_len] = 0; /* to be sure*/
+	} else {
+		/*use inode number as inode info*/
+		snprintf(name, name_len, "ino=%lx", inode->i_ino);
+	}
+
+	va_start(args, fmt);
+
+	level = printk_get_level(fmt);
+	vaf.fmt = printk_skip_level(fmt);
+	vaf.va = &args;
+
+	printk("%c%cntfs3: %s: %s %pV\n", KERN_SOH_ASCII, level, sb->s_id, name,
+	       &vaf);
+
+	va_end(args);
+
+	ntfs_free(name);
+}
+#endif
+
+/*
+ * Shared memory struct.
+ *
+ * on-disk ntfs's upcase table is created by ntfs formater
+ * 'upcase' table is 128K bytes of memory
+ * we should read it into memory when mounting
+ * Several ntfs volumes likely use the same 'upcase' table
+ * It is good idea to share in-memory 'upcase' table between different volumes
+ * Unfortunately winxp/vista/win7 use different upcase tables
+ */
+static DEFINE_SPINLOCK(s_shared_lock);
+
+static struct {
+	void *ptr;
+	u32 len;
+	int cnt;
+} s_shared[8];
+
+/*
+ * ntfs_set_shared
+ *
+ * Returns 'ptr' if pointer was saved in shared memory
+ * Returns NULL if pointer was not shared
+ */
+void *ntfs_set_shared(void *ptr, u32 bytes)
+{
+	void *ret = NULL;
+	int i, j = -1;
+
+	spin_lock(&s_shared_lock);
+	for (i = 0; i < ARRAY_SIZE(s_shared); i++) {
+		if (!s_shared[i].cnt) {
+			j = i;
+		} else if (bytes == s_shared[i].len &&
+			   !memcmp(s_shared[i].ptr, ptr, bytes)) {
+			s_shared[i].cnt += 1;
+			ret = s_shared[i].ptr;
+			break;
+		}
+	}
+
+	if (!ret && j != -1) {
+		s_shared[j].ptr = ptr;
+		s_shared[j].len = bytes;
+		s_shared[j].cnt = 1;
+		ret = ptr;
+	}
+	spin_unlock(&s_shared_lock);
+
+	return ret;
+}
+
+/*
+ * ntfs_put_shared
+ *
+ * Returns 'ptr' if pointer is not shared anymore
+ * Returns NULL if pointer is still shared
+ */
+void *ntfs_put_shared(void *ptr)
+{
+	void *ret = ptr;
+	int i;
+
+	spin_lock(&s_shared_lock);
+	for (i = 0; i < ARRAY_SIZE(s_shared); i++) {
+		if (s_shared[i].cnt && s_shared[i].ptr == ptr) {
+			if (--s_shared[i].cnt)
+				ret = NULL;
+			break;
+		}
+	}
+	spin_unlock(&s_shared_lock);
+
+	return ret;
+}
+
+static void clear_mount_options(struct ntfs_mount_options *options)
+{
+	static_assert(ARRAY_SIZE(options->nls) == 2);
+
+	unload_nls(options->nls[0]);
+	unload_nls(options->nls[1]);
+}
+
+enum Opt {
+	Opt_uid,
+	Opt_gid,
+	Opt_umask,
+	Opt_dmask,
+	Opt_fmask,
+	Opt_immutable,
+	Opt_discard,
+	Opt_force,
+	Opt_sparse,
+	Opt_nohidden,
+	Opt_showmeta,
+	Opt_acl,
+	Opt_noatime,
+	Opt_nls,
+	Opt_nls_alt,
+	Opt_prealloc,
+	Opt_no_acs_rules,
+	Opt_err,
+};
+
+static const match_table_t ntfs_tokens = {
+	{ Opt_uid, "uid=%u" },
+	{ Opt_gid, "gid=%u" },
+	{ Opt_umask, "umask=%o" },
+	{ Opt_dmask, "dmask=%o" },
+	{ Opt_fmask, "fmask=%o" },
+	{ Opt_immutable, "sys_immutable" },
+	{ Opt_discard, "discard" },
+	{ Opt_force, "force" },
+	{ Opt_sparse, "sparse" },
+	{ Opt_nohidden, "nohidden" },
+	{ Opt_acl, "acl" },
+	{ Opt_noatime, "noatime" },
+	{ Opt_showmeta, "showmeta" },
+	{ Opt_nls, "nls=%s" },
+	{ Opt_nls_alt, "nls_alt=%s" },
+	{ Opt_prealloc, "prealloc" },
+	{ Opt_no_acs_rules, "no_acs_rules" },
+	{ Opt_err, NULL },
+};
+
+static noinline int ntfs_parse_options(struct super_block *sb, char *options,
+				       int silent,
+				       struct ntfs_mount_options *opts)
+{
+	char *p;
+	substring_t args[MAX_OPT_ARGS];
+	int i, option;
+	char nls_name[2][30];
+
+	opts->fs_uid = current_uid();
+	opts->fs_gid = current_gid();
+	opts->fs_fmask_inv = opts->fs_dmask_inv = ~current_umask();
+	nls_name[0][0] = 0;
+	nls_name[1][0] = 0;
+
+	if (!options)
+		goto out;
+
+	while ((p = strsep(&options, ","))) {
+		int token;
+
+		if (!*p)
+			continue;
+
+		token = match_token(p, ntfs_tokens, args);
+		switch (token) {
+		case Opt_immutable:
+			opts->sys_immutable = 1;
+			break;
+		case Opt_uid:
+			if (match_int(&args[0], &option))
+				return -EINVAL;
+			opts->fs_uid = make_kuid(current_user_ns(), option);
+			if (!uid_valid(opts->fs_uid))
+				return -EINVAL;
+			opts->uid = 1;
+			break;
+		case Opt_gid:
+			if (match_int(&args[0], &option))
+				return -EINVAL;
+			opts->fs_gid = make_kgid(current_user_ns(), option);
+			if (!gid_valid(opts->fs_gid))
+				return -EINVAL;
+			opts->gid = 1;
+			break;
+		case Opt_umask:
+			if (match_octal(&args[0], &option))
+				return -EINVAL;
+			opts->fs_fmask_inv = opts->fs_dmask_inv = ~option;
+			opts->fmask = opts->dmask = 1;
+			break;
+		case Opt_dmask:
+			if (match_octal(&args[0], &option))
+				return -EINVAL;
+			opts->fs_dmask_inv = ~option;
+			opts->dmask = 1;
+			break;
+		case Opt_fmask:
+			if (match_octal(&args[0], &option))
+				return -EINVAL;
+			opts->fs_fmask_inv = ~option;
+			opts->fmask = 1;
+			break;
+		case Opt_discard:
+			opts->discard = 1;
+			break;
+		case Opt_force:
+			opts->force = 1;
+			break;
+		case Opt_sparse:
+			opts->sparse = 1;
+			break;
+		case Opt_nohidden:
+			opts->nohidden = 1;
+			break;
+		case Opt_acl:
+			sb->s_flags |= SB_POSIXACL;
+			break;
+		case Opt_noatime:
+			sb->s_flags |= SB_NOATIME;
+			break;
+		case Opt_showmeta:
+			opts->showmeta = 1;
+			break;
+		case Opt_nls:
+		case Opt_nls_alt:
+			match_strlcpy(nls_name[token == Opt_nls_alt], &args[0],
+				      sizeof(nls_name[0]));
+			break;
+		case Opt_prealloc:
+			opts->prealloc = 1;
+			break;
+		case Opt_no_acs_rules:
+			opts->no_acs_rules = 1;
+			break;
+		default:
+			if (!silent)
+				ntfs_err(
+					sb,
+					"Unrecognized mount option \"%s\" or missing value",
+					p);
+			//return -EINVAL;
+		}
+	}
+
+out:
+
+	for (i = 0; i < ARRAY_SIZE(opts->nls); i++) {
+		struct nls_table *nls;
+
+		if (nls_name[i][0]) {
+			nls = load_nls(nls_name[i]);
+			if (!nls) {
+				ntfs_err(sb, "failed to load \"%s\"",
+					 nls_name[i]);
+				return -EINVAL;
+			}
+		} else {
+			nls = load_nls_default();
+			if (!nls) {
+				ntfs_err(sb, "failed to load default nls");
+				return -EINVAL;
+			}
+		}
+
+		opts->nls[i] = nls;
+	}
+
+	/* unload alternative nls if it equals primary one */
+	if (!strcmp(opts->nls[0]->charset, opts->nls[1]->charset)) {
+		unload_nls(opts->nls[1]);
+		opts->nls[1] = NULL;
+	}
+
+	if (!strcmp(opts->nls[0]->charset, "utf8")) {
+		/*use utf16s_to_utf8s/utf8s_to_utf16s instead of nls*/
+		unload_nls(opts->nls[0]);
+		opts->nls[0] = NULL;
+	}
+
+	return 0;
+}
+
+static int ntfs_remount(struct super_block *sb, int *flags, char *data)
+{
+	int err, ro_rw;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_mount_options old_opts;
+	char *orig_data = kstrdup(data, GFP_KERNEL);
+
+	if (data && !orig_data)
+		return -ENOMEM;
+
+	/* Store  original options */
+	memcpy(&old_opts, &sbi->options, sizeof(old_opts));
+	clear_mount_options(&sbi->options);
+	memset(&sbi->options, 0, sizeof(sbi->options));
+
+	err = ntfs_parse_options(sb, data, 0, &sbi->options);
+	if (err)
+		goto restore_opts;
+
+	ro_rw = 0;
+	if (sb_rdonly(sb) && !(*flags & SB_RDONLY)) {
+		/* ro -> rw */
+		ro_rw = 1;
+		if (sbi->flags & NTFS_FLAGS_NEED_REPLAY) {
+			ntfs_warn(
+				sb,
+				"Couldn't remount rw because journal is not replayed. Please umount/remount instead\n");
+			err = -EINVAL;
+			goto restore_opts;
+		}
+	}
+
+	sync_filesystem(sb);
+
+	if (ro_rw && (sbi->volume.flags & VOLUME_FLAG_DIRTY) &&
+	    !sbi->options.force) {
+		ntfs_warn(sb, "volume is dirty and \"force\" flag is not set!");
+		err = -EINVAL;
+		goto restore_opts;
+	}
+
+	clear_mount_options(&old_opts);
+
+	*flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME) |
+		 SB_NODIRATIME | SB_NOATIME;
+	ntfs_info(sb, "re-mounted. Opts: %s", orig_data);
+	err = 0;
+	goto out;
+
+restore_opts:
+	clear_mount_options(&sbi->options);
+	memcpy(&sbi->options, &old_opts, sizeof(old_opts));
+
+out:
+	kfree(orig_data);
+	return err;
+}
+
+static struct kmem_cache *ntfs_inode_cachep;
+
+static struct inode *ntfs_alloc_inode(struct super_block *sb)
+{
+	struct ntfs_inode *ni = kmem_cache_alloc(ntfs_inode_cachep, GFP_NOFS);
+
+	if (!ni)
+		return NULL;
+
+	memset(ni, 0, offsetof(struct ntfs_inode, vfs_inode));
+
+	mutex_init(&ni->ni_lock);
+
+	return &ni->vfs_inode;
+}
+
+static void ntfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	mutex_destroy(&ni->ni_lock);
+
+	kmem_cache_free(ntfs_inode_cachep, ni);
+}
+
+static void ntfs_destroy_inode(struct inode *inode)
+{
+	call_rcu(&inode->i_rcu, ntfs_i_callback);
+}
+
+static void init_once(void *foo)
+{
+	struct ntfs_inode *ni = foo;
+
+	inode_init_once(&ni->vfs_inode);
+}
+
+/* noinline to reduce binary size*/
+static noinline void put_ntfs(struct ntfs_sb_info *sbi)
+{
+	ntfs_free(sbi->new_rec);
+	ntfs_free(ntfs_put_shared(sbi->upcase));
+	ntfs_free(sbi->def_table);
+
+	wnd_close(&sbi->mft.bitmap);
+	wnd_close(&sbi->used.bitmap);
+
+	if (sbi->mft.ni)
+		iput(&sbi->mft.ni->vfs_inode);
+
+	if (sbi->security.ni)
+		iput(&sbi->security.ni->vfs_inode);
+
+	if (sbi->reparse.ni)
+		iput(&sbi->reparse.ni->vfs_inode);
+
+	if (sbi->objid.ni)
+		iput(&sbi->objid.ni->vfs_inode);
+
+	if (sbi->volume.ni)
+		iput(&sbi->volume.ni->vfs_inode);
+
+	ntfs_update_mftmirr(sbi, 0);
+
+	indx_clear(&sbi->security.index_sii);
+	indx_clear(&sbi->security.index_sdh);
+	indx_clear(&sbi->reparse.index_r);
+	indx_clear(&sbi->objid.index_o);
+	ntfs_free(sbi->compress.ctx);
+
+	clear_mount_options(&sbi->options);
+
+	ntfs_free(sbi);
+}
+
+static void ntfs_put_super(struct super_block *sb)
+{
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+
+	/*mark rw ntfs as clear, if possible*/
+	ntfs_set_state(sbi, NTFS_DIRTY_CLEAR);
+
+	put_ntfs(sbi);
+
+	sync_blockdev(sb->s_bdev);
+}
+
+static int ntfs_statfs(struct dentry *dentry, struct kstatfs *buf)
+{
+	struct super_block *sb = dentry->d_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct wnd_bitmap *wnd = &sbi->used.bitmap;
+
+	buf->f_type = sb->s_magic;
+	buf->f_bsize = sbi->cluster_size;
+	buf->f_blocks = wnd->nbits;
+
+	buf->f_bfree = buf->f_bavail = wnd_zeroes(wnd);
+	buf->f_fsid.val[0] = (u32)sbi->volume.ser_num;
+	buf->f_fsid.val[1] = (u32)(sbi->volume.ser_num >> 32);
+	buf->f_namelen = NTFS_NAME_LEN;
+
+	return 0;
+}
+
+static int ntfs_show_options(struct seq_file *m, struct dentry *root)
+{
+	struct super_block *sb = root->d_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_mount_options *opts = &sbi->options;
+
+	if (opts->uid)
+		seq_printf(m, ",uid=%u",
+			   from_kuid_munged(&init_user_ns, opts->fs_uid));
+	if (opts->gid)
+		seq_printf(m, ",gid=%u",
+			   from_kgid_munged(&init_user_ns, opts->fs_gid));
+	if (opts->fmask)
+		seq_printf(m, ",fmask=%04o", ~opts->fs_fmask_inv);
+	if (opts->dmask)
+		seq_printf(m, ",dmask=%04o", ~opts->fs_dmask_inv);
+	if (opts->nls[0])
+		seq_printf(m, ",nls=%s", opts->nls[0]->charset);
+	else
+		seq_puts(m, ",nls=utf8");
+	if (opts->nls[1])
+		seq_printf(m, ",nls_alt=%s", opts->nls[1]->charset);
+	if (opts->sys_immutable)
+		seq_puts(m, ",sys_immutable");
+	if (opts->discard)
+		seq_puts(m, ",discard");
+	if (opts->sparse)
+		seq_puts(m, ",sparse");
+	if (opts->showmeta)
+		seq_puts(m, ",showmeta");
+	if (opts->nohidden)
+		seq_puts(m, ",nohidden");
+	if (opts->force)
+		seq_puts(m, ",force");
+	if (opts->no_acs_rules)
+		seq_puts(m, ",no_acs_rules");
+	if (opts->prealloc)
+		seq_puts(m, ",prealloc");
+	if (sb->s_flags & SB_POSIXACL)
+		seq_puts(m, ",acl");
+	if (sb->s_flags & SB_NOATIME)
+		seq_puts(m, ",noatime");
+
+	return 0;
+}
+
+/*super_operations::sync_fs*/
+static int ntfs_sync_fs(struct super_block *sb, int wait)
+{
+	int err = 0, err2;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_inode *ni;
+	struct inode *inode;
+
+	ni = sbi->security.ni;
+	if (ni) {
+		inode = &ni->vfs_inode;
+		err2 = _ni_write_inode(inode, wait);
+		if (err2 && !err)
+			err = err2;
+	}
+
+	ni = sbi->objid.ni;
+	if (ni) {
+		inode = &ni->vfs_inode;
+		err2 = _ni_write_inode(inode, wait);
+		if (err2 && !err)
+			err = err2;
+	}
+
+	ni = sbi->reparse.ni;
+	if (ni) {
+		inode = &ni->vfs_inode;
+		err2 = _ni_write_inode(inode, wait);
+		if (err2 && !err)
+			err = err2;
+	}
+
+	if (!err)
+		ntfs_set_state(sbi, NTFS_DIRTY_CLEAR);
+
+	ntfs_update_mftmirr(sbi, wait);
+
+	return err;
+}
+
+static const struct super_operations ntfs_sops = {
+	.alloc_inode = ntfs_alloc_inode,
+	.destroy_inode = ntfs_destroy_inode,
+	.evict_inode = ntfs_evict_inode,
+	.put_super = ntfs_put_super,
+	.statfs = ntfs_statfs,
+	.show_options = ntfs_show_options,
+	.sync_fs = ntfs_sync_fs,
+	.remount_fs = ntfs_remount,
+	.write_inode = ntfs_write_inode,
+};
+
+static struct inode *ntfs_export_get_inode(struct super_block *sb, u64 ino,
+					   u32 generation)
+{
+	struct MFT_REF ref;
+	struct inode *inode;
+
+	ref.low = cpu_to_le32(ino);
+#ifdef NTFS3_64BIT_CLUSTER
+	ref.high = cpu_to_le16(ino >> 32);
+#else
+	ref.high = 0;
+#endif
+	ref.seq = cpu_to_le16(generation);
+
+	inode = ntfs_iget5(sb, &ref, NULL);
+	if (!IS_ERR(inode) && is_bad_inode(inode)) {
+		iput(inode);
+		inode = ERR_PTR(-ESTALE);
+	}
+
+	return inode;
+}
+
+static struct dentry *ntfs_fh_to_dentry(struct super_block *sb, struct fid *fid,
+					int fh_len, int fh_type)
+{
+	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
+				    ntfs_export_get_inode);
+}
+
+static struct dentry *ntfs_fh_to_parent(struct super_block *sb, struct fid *fid,
+					int fh_len, int fh_type)
+{
+	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
+				    ntfs_export_get_inode);
+}
+
+/* TODO: == ntfs_sync_inode */
+static int ntfs_nfs_commit_metadata(struct inode *inode)
+{
+	return _ni_write_inode(inode, 1);
+}
+
+static const struct export_operations ntfs_export_ops = {
+	.fh_to_dentry = ntfs_fh_to_dentry,
+	.fh_to_parent = ntfs_fh_to_parent,
+	.get_parent = ntfs_get_parent,
+	.commit_metadata = ntfs_nfs_commit_metadata,
+};
+
+/* Returns Gb,Mb to print with "%u.%02u Gb" */
+static u32 format_size_gb(const u64 bytes, u32 *mb)
+{
+	/* Do simple right 30 bit shift of 64 bit value */
+	u64 kbytes = bytes >> 10;
+	u32 kbytes32 = (u32)kbytes;
+
+	*mb = (100 * (kbytes32 & 0xfffff) + 0x7ffff) >> 20;
+	if (*mb >= 100)
+		*mb = 99;
+
+	return (kbytes32 >> 20) | (((u32)(kbytes >> 32)) << 12);
+}
+
+static u32 true_sectors_per_clst(const struct NTFS_BOOT *boot)
+{
+	return boot->sectors_per_clusters <= 0x80 ?
+		       boot->sectors_per_clusters :
+		       (1u << (0 - boot->sectors_per_clusters));
+}
+
+/* inits internal info from on-disk boot sector*/
+static int ntfs_init_from_boot(struct super_block *sb, u32 sector_size,
+			       u64 dev_size)
+{
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	int err;
+	u32 mb, gb, boot_sector_size, sct_per_clst, record_size;
+	u64 sectors, clusters, fs_size, mlcn, mlcn2;
+	struct NTFS_BOOT *boot;
+	struct buffer_head *bh;
+	struct MFT_REC *rec;
+	u16 fn, ao;
+
+	sbi->volume.blocks = dev_size >> PAGE_SHIFT;
+
+	bh = ntfs_bread(sb, 0);
+	if (!bh)
+		return -EIO;
+
+	err = -EINVAL;
+	boot = (struct NTFS_BOOT *)bh->b_data;
+
+	if (memcmp(boot->system_id, "NTFS    ", sizeof("NTFS    ") - 1))
+		goto out;
+
+	/* 0x55AA is not mandaroty. Thanks Maxim Suhanov*/
+	/*if (0x55 != boot->boot_magic[0] || 0xAA != boot->boot_magic[1])
+	 *	goto out;
+	 */
+
+	boot_sector_size = (u32)boot->bytes_per_sector[1] << 8;
+	if (boot->bytes_per_sector[0] || boot_sector_size < SECTOR_SIZE ||
+	    !is_power_of2(boot_sector_size)) {
+		goto out;
+	}
+
+	sct_per_clst = true_sectors_per_clst(boot);
+	if (!is_power_of2(sct_per_clst))
+		goto out;
+
+	mlcn = le64_to_cpu(boot->mft_clst);
+	mlcn2 = le64_to_cpu(boot->mft2_clst);
+	sectors = le64_to_cpu(boot->sectors_per_volume);
+
+	if (mlcn * sct_per_clst >= sectors)
+		goto out;
+
+	if (mlcn2 * sct_per_clst >= sectors)
+		goto out;
+
+	/* Check MFT record size */
+	if ((boot->record_size < 0 &&
+	     SECTOR_SIZE > (2U << (-boot->record_size))) ||
+	    (boot->record_size >= 0 && !is_power_of2(boot->record_size))) {
+		goto out;
+	}
+
+	/* Check index record size */
+	if ((boot->index_size < 0 &&
+	     SECTOR_SIZE > (2U << (-boot->index_size))) ||
+	    (boot->index_size >= 0 && !is_power_of2(boot->index_size))) {
+		goto out;
+	}
+
+	sbi->sector_size = boot_sector_size;
+	sbi->sector_bits = blksize_bits(boot_sector_size);
+	fs_size = (sectors + 1) << sbi->sector_bits;
+
+	gb = format_size_gb(fs_size, &mb);
+
+	/*
+	 * - Volume formatted and mounted with the same sector size
+	 * - Volume formatted 4K and mounted as 512
+	 * - Volume formatted 512 and mounted as 4K
+	 */
+	if (sbi->sector_size != sector_size) {
+		ntfs_warn(sb,
+			  "Different NTFS' sector size and media sector size");
+		dev_size += sector_size - 1;
+	}
+
+	sbi->cluster_size = boot_sector_size * sct_per_clst;
+	sbi->cluster_bits = blksize_bits(sbi->cluster_size);
+
+	sbi->mft.lbo = mlcn << sbi->cluster_bits;
+	sbi->mft.lbo2 = mlcn2 << sbi->cluster_bits;
+
+	if (sbi->cluster_size < sbi->sector_size)
+		goto out;
+
+	sbi->cluster_mask = sbi->cluster_size - 1;
+	sbi->cluster_mask_inv = ~(u64)sbi->cluster_mask;
+	sbi->record_size = record_size = boot->record_size < 0 ?
+						 1 << (-boot->record_size) :
+						 (u32)boot->record_size
+							 << sbi->cluster_bits;
+
+	if (record_size > MAXIMUM_BYTES_PER_MFT)
+		goto out;
+
+	sbi->record_bits = blksize_bits(record_size);
+	sbi->attr_size_tr = (5 * record_size >> 4); // ~320 bytes
+
+	sbi->max_bytes_per_attr =
+		record_size - QuadAlign(MFTRECORD_FIXUP_OFFSET_1) -
+		QuadAlign(((record_size >> SECTOR_SHIFT) * sizeof(short))) -
+		QuadAlign(sizeof(enum ATTR_TYPE));
+
+	sbi->index_size = boot->index_size < 0 ?
+				  1u << (-boot->index_size) :
+				  (u32)boot->index_size << sbi->cluster_bits;
+
+	sbi->volume.ser_num = le64_to_cpu(boot->serial_num);
+	sbi->volume.size = sectors << sbi->sector_bits;
+
+	/* warning if RAW volume */
+	if (dev_size < fs_size) {
+		u32 mb0, gb0;
+
+		gb0 = format_size_gb(dev_size, &mb0);
+		ntfs_warn(
+			sb,
+			"RAW NTFS volume: Filesystem size %u.%02u Gb > volume size %u.%02u Gb. Mount in read-only",
+			gb, mb, gb0, mb0);
+		sb->s_flags |= SB_RDONLY;
+	}
+
+	clusters = sbi->volume.size >> sbi->cluster_bits;
+#ifdef NTFS3_64BIT_CLUSTER
+#if BITS_PER_LONG < 64
+#error "NTFS3_64BIT_CLUSTER incompatible in 32 bit OS"
+#endif
+#else
+	/* 32 bits per cluster */
+	if (clusters >> 32) {
+		ntfs_notice(
+			sb,
+			"NTFS %u.%02u Gb is too big to use 32 bits per cluster",
+			gb, mb);
+		goto out;
+	}
+#endif
+
+	sbi->used.bitmap.nbits = clusters;
+
+	rec = ntfs_alloc(record_size, 1);
+	if (!rec) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	sbi->new_rec = rec;
+	rec->rhdr.sign = NTFS_FILE_SIGNATURE;
+	rec->rhdr.fix_off = cpu_to_le16(MFTRECORD_FIXUP_OFFSET_1);
+	fn = (sbi->record_size >> SECTOR_SHIFT) + 1;
+	rec->rhdr.fix_num = cpu_to_le16(fn);
+	ao = QuadAlign(MFTRECORD_FIXUP_OFFSET_1 + sizeof(short) * fn);
+	rec->attr_off = cpu_to_le16(ao);
+	rec->used = cpu_to_le32(ao + QuadAlign(sizeof(enum ATTR_TYPE)));
+	rec->total = cpu_to_le32(sbi->record_size);
+	((struct ATTRIB *)Add2Ptr(rec, ao))->type = ATTR_END;
+
+	if (sbi->cluster_size < PAGE_SIZE)
+		sb_set_blocksize(sb, sbi->cluster_size);
+
+	sbi->block_mask = sb->s_blocksize - 1;
+	sbi->blocks_per_cluster = sbi->cluster_size >> sb->s_blocksize_bits;
+	sbi->volume.blocks = sbi->volume.size >> sb->s_blocksize_bits;
+
+	/* Maximum size for normal files */
+	sbi->maxbytes = (clusters << sbi->cluster_bits) - 1;
+
+#ifdef NTFS3_64BIT_CLUSTER
+	if (clusters >= (1ull << (64 - sbi->cluster_bits)))
+		sbi->maxbytes = -1;
+	sbi->maxbytes_sparse = -1;
+#else
+	/* Maximum size for sparse file */
+	sbi->maxbytes_sparse = (1ull << (sbi->cluster_bits + 32)) - 1;
+#endif
+
+	err = 0;
+
+out:
+	brelse(bh);
+
+	return err;
+}
+
+/* try to mount*/
+static int ntfs_fill_super(struct super_block *sb, void *data, int silent)
+{
+	int err;
+	struct ntfs_sb_info *sbi;
+	struct block_device *bdev = sb->s_bdev;
+	struct inode *bd_inode = bdev->bd_inode;
+	struct request_queue *rq = bdev_get_queue(bdev);
+	struct inode *inode = NULL;
+	struct ntfs_inode *ni;
+	size_t i, tt;
+	CLST vcn, lcn, len;
+	struct ATTRIB *attr;
+	const struct VOLUME_INFO *info;
+	u32 idx, done, bytes;
+	struct ATTR_DEF_ENTRY *t;
+	u16 *upcase = NULL;
+	u16 *shared;
+	bool is_ro;
+	struct MFT_REF ref;
+
+	ref.high = 0;
+
+	sbi = ntfs_alloc(sizeof(struct ntfs_sb_info), true);
+	if (!sbi)
+		return -ENOMEM;
+
+	sb->s_fs_info = sbi;
+	sbi->sb = sb;
+	sb->s_flags |= SB_NODIRATIME;
+	sb->s_magic = 0x7366746e; // "ntfs"
+	sb->s_op = &ntfs_sops;
+	sb->s_export_op = &ntfs_export_ops;
+	sb->s_time_gran = NTFS_TIME_GRAN; // 100 nsec
+	sb->s_xattr = ntfs_xattr_handlers;
+	sb->s_maxbytes = MAX_LFS_FILESIZE;
+
+	ratelimit_state_init(&sbi->msg_ratelimit, DEFAULT_RATELIMIT_INTERVAL,
+			     DEFAULT_RATELIMIT_BURST);
+
+	err = ntfs_parse_options(sb, data, silent, &sbi->options);
+	if (err)
+		goto out;
+
+	if (!rq || !blk_queue_discard(rq) || !rq->limits.discard_granularity) {
+		;
+	} else {
+		sbi->discard_granularity = rq->limits.discard_granularity;
+		sbi->discard_granularity_mask_inv =
+			~(u64)(sbi->discard_granularity - 1);
+	}
+
+	sb_set_blocksize(sb, PAGE_SIZE);
+
+	/* parse boot */
+	err = ntfs_init_from_boot(sb, rq ? queue_logical_block_size(rq) : 512,
+				  bd_inode->i_size);
+	if (err)
+		goto out;
+
+	spin_lock_init(&sbi->compress.lock);
+	if (sbi->cluster_size <= NTFS_LZNT_MAX_CLUSTER) {
+		sbi->compress.ctx = get_compression_ctx(true);
+		if (!sbi->compress.ctx) {
+			err = -ENOMEM;
+			goto out;
+		}
+	}
+
+	/*
+	 * Load $Volume. This should be done before $LogFile
+	 * 'cause 'sbi->volume.ni' is used 'ntfs_set_state'
+	 */
+	ref.low = cpu_to_le32(MFT_REC_VOL);
+	ref.seq = cpu_to_le16(MFT_REC_VOL);
+	inode = ntfs_iget5(sb, &ref, &NAME_VOLUME);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load $Volume.");
+		inode = NULL;
+		goto out;
+	}
+
+	ni = ntfs_i(inode);
+
+	/* Load and save label (not necessary) */
+	attr = ni_find_attr(ni, NULL, NULL, ATTR_LABEL, NULL, 0, NULL, NULL);
+
+	if (!attr) {
+		;
+	} else if (!attr->non_res && !is_attr_ext(attr)) {
+		/* $AttrDef allows labels to be up to 128 symbols */
+		err = utf16s_to_utf8s(resident_data(attr),
+				      le32_to_cpu(attr->res.data_size) >> 1,
+				      UTF16_LITTLE_ENDIAN, sbi->volume.label,
+				      sizeof(sbi->volume.label));
+		if (err < 0)
+			sbi->volume.label[0] = 0;
+	} else {
+		/* should we break mounting here? */
+		//err = -EINVAL;
+		//goto out;
+	}
+
+	attr = ni_find_attr(ni, attr, NULL, ATTR_VOL_INFO, NULL, 0, NULL, NULL);
+	if (!attr || is_attr_ext(attr)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	info = resident_data_ex(attr, SIZEOF_ATTRIBUTE_VOLUME_INFO);
+	if (!info) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	sbi->volume.major_ver = info->major_ver;
+	sbi->volume.minor_ver = info->minor_ver;
+	sbi->volume.flags = info->flags;
+
+	sbi->volume.ni = ni;
+	inode = NULL;
+
+	/* Load $MFTMirr to estimate recs_mirr */
+	ref.low = cpu_to_le32(MFT_REC_MIRR);
+	ref.seq = cpu_to_le16(MFT_REC_MIRR);
+	inode = ntfs_iget5(sb, &ref, &NAME_MIRROR);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load $MFTMirr.");
+		inode = NULL;
+		goto out;
+	}
+
+	sbi->mft.recs_mirr =
+		ntfs_up_cluster(sbi, inode->i_size) >> sbi->record_bits;
+
+	iput(inode);
+
+	/* Load $LogFile to replay */
+	ref.low = cpu_to_le32(MFT_REC_LOG);
+	ref.seq = cpu_to_le16(MFT_REC_LOG);
+	inode = ntfs_iget5(sb, &ref, &NAME_LOGFILE);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load $LogFile.");
+		inode = NULL;
+		goto out;
+	}
+
+	ni = ntfs_i(inode);
+
+	err = ntfs_loadlog_and_replay(ni, sbi);
+	if (err)
+		goto out;
+
+	iput(inode);
+	inode = NULL;
+
+	is_ro = sb_rdonly(sbi->sb);
+
+	if (sbi->flags & NTFS_FLAGS_NEED_REPLAY) {
+		if (!is_ro) {
+			ntfs_warn(sb,
+				  "failed to replay log file. Can't mount rw!");
+			err = -EINVAL;
+			goto out;
+		}
+	} else if (sbi->volume.flags & VOLUME_FLAG_DIRTY) {
+		if (!is_ro && !sbi->options.force) {
+			ntfs_warn(
+				sb,
+				"volume is dirty and \"force\" flag is not set!");
+			err = -EINVAL;
+			goto out;
+		}
+	}
+
+	/* Load $MFT */
+	ref.low = cpu_to_le32(MFT_REC_MFT);
+	ref.seq = cpu_to_le16(1);
+
+	inode = ntfs_iget5(sb, &ref, &NAME_MFT);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load $MFT.");
+		inode = NULL;
+		goto out;
+	}
+
+	ni = ntfs_i(inode);
+
+	sbi->mft.used = ni->i_valid >> sbi->record_bits;
+	tt = inode->i_size >> sbi->record_bits;
+	sbi->mft.next_free = MFT_REC_USER;
+
+	err = wnd_init(&sbi->mft.bitmap, sb, tt);
+	if (err)
+		goto out;
+
+	err = ni_load_all_mi(ni);
+	if (err)
+		goto out;
+
+	sbi->mft.ni = ni;
+
+	/* Load $BadClus */
+	ref.low = cpu_to_le32(MFT_REC_BADCLUST);
+	ref.seq = cpu_to_le16(MFT_REC_BADCLUST);
+	inode = ntfs_iget5(sb, &ref, &NAME_BADCLUS);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load $BadClus.");
+		inode = NULL;
+		goto out;
+	}
+
+	ni = ntfs_i(inode);
+
+	for (i = 0; run_get_entry(&ni->file.run, i, &vcn, &lcn, &len); i++) {
+		if (lcn == SPARSE_LCN)
+			continue;
+
+		if (!sbi->bad_clusters)
+			ntfs_notice(sb, "Volume contains bad blocks");
+
+		sbi->bad_clusters += len;
+	}
+
+	iput(inode);
+
+	/* Load $Bitmap */
+	ref.low = cpu_to_le32(MFT_REC_BITMAP);
+	ref.seq = cpu_to_le16(MFT_REC_BITMAP);
+	inode = ntfs_iget5(sb, &ref, &NAME_BITMAP);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load $Bitmap.");
+		inode = NULL;
+		goto out;
+	}
+
+	ni = ntfs_i(inode);
+
+#ifndef NTFS3_64BIT_CLUSTER
+	if (inode->i_size >> 32) {
+		err = -EINVAL;
+		goto out;
+	}
+#endif
+
+	/* Check bitmap boundary */
+	tt = sbi->used.bitmap.nbits;
+	if (inode->i_size < bitmap_size(tt)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* Not necessary */
+	sbi->used.bitmap.set_tail = true;
+	err = wnd_init(&sbi->used.bitmap, sbi->sb, tt);
+	if (err)
+		goto out;
+
+	iput(inode);
+
+	/* Compute the mft zone */
+	err = ntfs_refresh_zone(sbi);
+	if (err)
+		goto out;
+
+	/* Load $AttrDef */
+	ref.low = cpu_to_le32(MFT_REC_ATTR);
+	ref.seq = cpu_to_le16(MFT_REC_ATTR);
+	inode = ntfs_iget5(sbi->sb, &ref, &NAME_ATTRDEF);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load $AttrDef -> %d", err);
+		inode = NULL;
+		goto out;
+	}
+
+	if (inode->i_size < sizeof(struct ATTR_DEF_ENTRY)) {
+		err = -EINVAL;
+		goto out;
+	}
+	bytes = inode->i_size;
+	sbi->def_table = t = ntfs_alloc(bytes, 0);
+	if (!t) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	for (done = idx = 0; done < bytes; done += PAGE_SIZE, idx++) {
+		unsigned long tail = bytes - done;
+		struct page *page = ntfs_map_page(inode->i_mapping, idx);
+
+		if (IS_ERR(page)) {
+			err = PTR_ERR(page);
+			goto out;
+		}
+		memcpy(Add2Ptr(t, done), page_address(page),
+		       min(PAGE_SIZE, tail));
+		ntfs_unmap_page(page);
+
+		if (!idx && ATTR_STD != t->type) {
+			err = -EINVAL;
+			goto out;
+		}
+	}
+
+	t += 1;
+	sbi->def_entries = 1;
+	done = sizeof(struct ATTR_DEF_ENTRY);
+	sbi->reparse.max_size = MAXIMUM_REPARSE_DATA_BUFFER_SIZE;
+
+	while (done + sizeof(struct ATTR_DEF_ENTRY) <= bytes) {
+		u32 t32 = le32_to_cpu(t->type);
+
+		if ((t32 & 0xF) || le32_to_cpu(t[-1].type) >= t32)
+			break;
+
+		if (t->type == ATTR_REPARSE)
+			sbi->reparse.max_size = le64_to_cpu(t->max_sz);
+
+		done += sizeof(struct ATTR_DEF_ENTRY);
+		t += 1;
+		sbi->def_entries += 1;
+	}
+	iput(inode);
+
+	/* Load $UpCase */
+	ref.low = cpu_to_le32(MFT_REC_UPCASE);
+	ref.seq = cpu_to_le16(MFT_REC_UPCASE);
+	inode = ntfs_iget5(sb, &ref, &NAME_UPCASE);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load $LogFile.");
+		inode = NULL;
+		goto out;
+	}
+
+	ni = ntfs_i(inode);
+
+	if (inode->i_size != 0x10000 * sizeof(short)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	sbi->upcase = upcase = ntfs_alloc(0x10000 * sizeof(short), 0);
+	if (!upcase) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	for (idx = 0; idx < (0x10000 * sizeof(short) >> PAGE_SHIFT); idx++) {
+		const u16 *src;
+		u16 *dst = Add2Ptr(upcase, idx << PAGE_SHIFT);
+		struct page *page = ntfs_map_page(inode->i_mapping, idx);
+
+		if (IS_ERR(page)) {
+			err = PTR_ERR(page);
+			goto out;
+		}
+
+		src = page_address(page);
+
+#ifdef __BIG_ENDIAN
+		for (i = 0; i < PAGE_SIZE / sizeof(u16); i++)
+			*dst++ = le16_to_cpu(*src++);
+#else
+		memcpy(dst, src, PAGE_SIZE);
+#endif
+		ntfs_unmap_page(page);
+	}
+
+	shared = ntfs_set_shared(upcase, 0x10000 * sizeof(short));
+	if (shared && upcase != shared) {
+		sbi->upcase = shared;
+		ntfs_free(upcase);
+	}
+
+	iput(inode);
+	inode = NULL;
+
+	if (is_ntfs3(sbi)) {
+		/* Load $Secure */
+		err = ntfs_security_init(sbi);
+		if (err)
+			goto out;
+
+		/* Load $Extend */
+		err = ntfs_extend_init(sbi);
+		if (err)
+			goto load_root;
+
+		/* Load $Extend\$Reparse */
+		err = ntfs_reparse_init(sbi);
+		if (err)
+			goto load_root;
+
+		/* Load $Extend\$ObjId */
+		err = ntfs_objid_init(sbi);
+		if (err)
+			goto load_root;
+	}
+
+load_root:
+
+	/* Load root */
+	ref.low = cpu_to_le32(MFT_REC_ROOT);
+	ref.seq = cpu_to_le16(MFT_REC_ROOT);
+	inode = ntfs_iget5(sb, &ref, &NAME_ROOT);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
+		ntfs_err(sb, "Failed to load root.");
+		inode = NULL;
+		goto out;
+	}
+
+	ni = ntfs_i(inode);
+
+	sb->s_root = d_make_root(inode);
+
+	if (!sb->s_root) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	return 0;
+
+out:
+	iput(inode);
+
+	if (sb->s_root) {
+		d_drop(sb->s_root);
+		sb->s_root = NULL;
+	}
+
+	put_ntfs(sbi);
+
+	sb->s_fs_info = NULL;
+	return err;
+}
+
+void ntfs_unmap_meta(struct super_block *sb, CLST lcn, CLST len)
+{
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct block_device *bdev = sb->s_bdev;
+	sector_t devblock = (u64)lcn * sbi->blocks_per_cluster;
+	unsigned long blocks = (u64)len * sbi->blocks_per_cluster;
+	unsigned long cnt = 0;
+	unsigned long limit = global_zone_page_state(NR_FREE_PAGES)
+			      << (PAGE_SHIFT - sb->s_blocksize_bits);
+
+	if (limit >= 0x2000)
+		limit -= 0x1000;
+	else if (limit < 32)
+		limit = 32;
+	else
+		limit >>= 1;
+
+	while (blocks--) {
+		clean_bdev_aliases(bdev, devblock++, 1);
+		if (cnt++ >= limit) {
+			sync_blockdev(bdev);
+			cnt = 0;
+		}
+	}
+}
+
+/*
+ * ntfs_discard
+ *
+ * issue a discard request (trim for SSD)
+ */
+int ntfs_discard(struct ntfs_sb_info *sbi, CLST lcn, CLST len)
+{
+	int err;
+	u64 lbo, bytes, start, end;
+	struct super_block *sb;
+
+	if (sbi->used.next_free_lcn == lcn + len)
+		sbi->used.next_free_lcn = lcn;
+
+	if (sbi->flags & NTFS_FLAGS_NODISCARD)
+		return -EOPNOTSUPP;
+
+	if (!sbi->options.discard)
+		return -EOPNOTSUPP;
+
+	lbo = (u64)lcn << sbi->cluster_bits;
+	bytes = (u64)len << sbi->cluster_bits;
+
+	/* Align up 'start' on discard_granularity */
+	start = (lbo + sbi->discard_granularity - 1) &
+		sbi->discard_granularity_mask_inv;
+	/* Align down 'end' on discard_granularity */
+	end = (lbo + bytes) & sbi->discard_granularity_mask_inv;
+
+	sb = sbi->sb;
+	if (start >= end)
+		return 0;
+
+	err = blkdev_issue_discard(sb->s_bdev, start >> 9, (end - start) >> 9,
+				   GFP_NOFS, 0);
+
+	if (err == -EOPNOTSUPP)
+		sbi->flags |= NTFS_FLAGS_NODISCARD;
+
+	return err;
+}
+
+static struct dentry *ntfs_mount(struct file_system_type *fs_type, int flags,
+				 const char *dev_name, void *data)
+{
+	return mount_bdev(fs_type, flags, dev_name, data, ntfs_fill_super);
+}
+
+static struct file_system_type ntfs_fs_type = {
+	.owner = THIS_MODULE,
+	.name = "ntfs3",
+	.mount = ntfs_mount,
+	.kill_sb = kill_block_super,
+	.fs_flags = FS_REQUIRES_DEV,
+};
+
+static int __init init_ntfs_fs(void)
+{
+	int err;
+
+#ifdef NTFS3_INDEX_BINARY_SEARCH
+	pr_notice("ntfs3: +index binary search");
+#endif
+
+#ifdef NTFS3_CHECK_FREE_CLST
+	pr_notice("ntfs3: +check free clusters");
+#endif
+
+#if NTFS_LINK_MAX < 0xffff
+	pr_notice("ntfs3: max link count %u", NTFS_LINK_MAX);
+#endif
+
+#ifdef NTFS3_64BIT_CLUSTER
+	pr_notice("ntfs3: 64 bits per cluster");
+#else
+	pr_notice("ntfs3: 32 bits per cluster");
+#endif
+
+	ntfs_inode_cachep = kmem_cache_create(
+		"ntfs_inode_cache", sizeof(struct ntfs_inode), 0,
+		(SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD | SLAB_ACCOUNT),
+		init_once);
+	if (!ntfs_inode_cachep) {
+		err = -ENOMEM;
+		goto failed;
+	}
+
+	err = register_filesystem(&ntfs_fs_type);
+	if (!err)
+		return 0;
+
+failed:
+	return err;
+}
+
+static void __exit exit_ntfs_fs(void)
+{
+	if (ntfs_inode_cachep) {
+		rcu_barrier();
+		kmem_cache_destroy(ntfs_inode_cachep);
+	}
+
+	unregister_filesystem(&ntfs_fs_type);
+}
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("ntfs3 filesystem");
+MODULE_AUTHOR("Konstantin Komarov");
+MODULE_ALIAS_FS("ntfs3");
+
+module_init(init_ntfs_fs);
+module_exit(exit_ntfs_fs);
-- 
2.25.4


^ permalink raw reply related	[flat|nested] 21+ messages in thread

* [PATCH v11 03/10] fs/ntfs3: Add bitmap
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 01/10] fs/ntfs3: Add headers and misc files Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 02/10] fs/ntfs3: Add initialization of super block Konstantin Komarov
@ 2020-10-30 15:02 ` Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 04/10] fs/ntfs3: Add file operations and implementation Konstantin Komarov
                   ` (9 subsequent siblings)
  12 siblings, 0 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This adds bitmap

Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
---
 fs/ntfs3/bitfunc.c |  135 ++++
 fs/ntfs3/bitmap.c  | 1504 ++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 1639 insertions(+)
 create mode 100644 fs/ntfs3/bitfunc.c
 create mode 100644 fs/ntfs3/bitmap.c

diff --git a/fs/ntfs3/bitfunc.c b/fs/ntfs3/bitfunc.c
new file mode 100644
index 000000000000..2d43d718d645
--- /dev/null
+++ b/fs/ntfs3/bitfunc.c
@@ -0,0 +1,135 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+#define BITS_IN_SIZE_T (sizeof(size_t) * 8)
+
+/*
+ * fill_mask[i] - first i bits are '1' , i = 0,1,2,3,4,5,6,7,8
+ * fill_mask[i] = 0xFF >> (8-i)
+ */
+static const u8 fill_mask[] = { 0x00, 0x01, 0x03, 0x07, 0x0F,
+				0x1F, 0x3F, 0x7F, 0xFF };
+
+/*
+ * zero_mask[i] - first i bits are '0' , i = 0,1,2,3,4,5,6,7,8
+ * zero_mask[i] = 0xFF << i
+ */
+static const u8 zero_mask[] = { 0xFF, 0xFE, 0xFC, 0xF8, 0xF0,
+				0xE0, 0xC0, 0x80, 0x00 };
+
+/*
+ * are_bits_clear
+ *
+ * Returns true if all bits [bit, bit+nbits) are zeros "0"
+ */
+bool are_bits_clear(const ulong *lmap, size_t bit, size_t nbits)
+{
+	size_t pos = bit & 7;
+	const u8 *map = (u8 *)lmap + (bit >> 3);
+
+	if (pos) {
+		if (8 - pos >= nbits)
+			return !nbits || !(*map & fill_mask[pos + nbits] &
+					   zero_mask[pos]);
+
+		if (*map++ & zero_mask[pos])
+			return false;
+		nbits -= 8 - pos;
+	}
+
+	pos = ((size_t)map) & (sizeof(size_t) - 1);
+	if (pos) {
+		pos = sizeof(size_t) - pos;
+		if (nbits >= pos * 8) {
+			for (nbits -= pos * 8; pos; pos--, map++) {
+				if (*map)
+					return false;
+			}
+		}
+	}
+
+	for (pos = nbits / BITS_IN_SIZE_T; pos; pos--, map += sizeof(size_t)) {
+		if (*((size_t *)map))
+			return false;
+	}
+
+	for (pos = (nbits % BITS_IN_SIZE_T) >> 3; pos; pos--, map++) {
+		if (*map)
+			return false;
+	}
+
+	pos = nbits & 7;
+	if (pos && (*map & fill_mask[pos]))
+		return false;
+
+	// All bits are zero
+	return true;
+}
+
+/*
+ * are_bits_set
+ *
+ * Returns true if all bits [bit, bit+nbits) are ones "1"
+ */
+bool are_bits_set(const ulong *lmap, size_t bit, size_t nbits)
+{
+	u8 mask;
+	size_t pos = bit & 7;
+	const u8 *map = (u8 *)lmap + (bit >> 3);
+
+	if (pos) {
+		if (8 - pos >= nbits) {
+			mask = fill_mask[pos + nbits] & zero_mask[pos];
+			return !nbits || (*map & mask) == mask;
+		}
+
+		mask = zero_mask[pos];
+		if ((*map++ & mask) != mask)
+			return false;
+		nbits -= 8 - pos;
+	}
+
+	pos = ((size_t)map) & (sizeof(size_t) - 1);
+	if (pos) {
+		pos = sizeof(size_t) - pos;
+		if (nbits >= pos * 8) {
+			for (nbits -= pos * 8; pos; pos--, map++) {
+				if (*map != 0xFF)
+					return false;
+			}
+		}
+	}
+
+	for (pos = nbits / BITS_IN_SIZE_T; pos; pos--, map += sizeof(size_t)) {
+		if (*((size_t *)map) != MINUS_ONE_T)
+			return false;
+	}
+
+	for (pos = (nbits % BITS_IN_SIZE_T) >> 3; pos; pos--, map++) {
+		if (*map != 0xFF)
+			return false;
+	}
+
+	pos = nbits & 7;
+	if (pos) {
+		u8 mask = fill_mask[pos];
+
+		if ((*map & mask) != mask)
+			return false;
+	}
+
+	// All bits are ones
+	return true;
+}
diff --git a/fs/ntfs3/bitmap.c b/fs/ntfs3/bitmap.c
new file mode 100644
index 000000000000..db7c4567c59a
--- /dev/null
+++ b/fs/ntfs3/bitmap.c
@@ -0,0 +1,1504 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+struct rb_node_key {
+	struct rb_node node;
+	size_t key;
+};
+
+/*
+ * Tree is sorted by start (key)
+ */
+struct e_node {
+	struct rb_node_key start; /* Tree sorted by start */
+	struct rb_node_key count; /* Tree sorted by len*/
+};
+
+static int wnd_rescan(struct wnd_bitmap *wnd);
+static struct buffer_head *wnd_map(struct wnd_bitmap *wnd, size_t iw);
+static bool wnd_is_free_hlp(struct wnd_bitmap *wnd, size_t bit, size_t bits);
+
+static inline u32 wnd_bits(const struct wnd_bitmap *wnd, size_t i)
+{
+	return i + 1 == wnd->nwnd ? wnd->bits_last : wnd->sb->s_blocksize * 8;
+}
+
+/*
+ * b_pos + b_len - biggest fragment
+ * Scan range [wpos wbits) window 'buf'
+ * Returns -1 if not found
+ */
+static size_t wnd_scan(const ulong *buf, size_t wbit, u32 wpos, u32 wend,
+		       size_t to_alloc, size_t *prev_tail, size_t *b_pos,
+		       size_t *b_len)
+{
+	while (wpos < wend) {
+		size_t free_len;
+		u32 free_bits, end;
+		u32 used = find_next_zero_bit(buf, wend, wpos);
+
+		if (used >= wend) {
+			if (*b_len < *prev_tail) {
+				*b_pos = wbit - *prev_tail;
+				*b_len = *prev_tail;
+			}
+
+			*prev_tail = 0;
+			return -1;
+		}
+
+		if (used > wpos) {
+			wpos = used;
+			if (*b_len < *prev_tail) {
+				*b_pos = wbit - *prev_tail;
+				*b_len = *prev_tail;
+			}
+
+			*prev_tail = 0;
+		}
+
+		/*
+		 * Now we have a fragment [wpos, wend) staring with 0
+		 */
+		end = wpos + to_alloc - *prev_tail;
+		free_bits = find_next_bit(buf, min(end, wend), wpos);
+
+		free_len = *prev_tail + free_bits - wpos;
+
+		if (*b_len < free_len) {
+			*b_pos = wbit + wpos - *prev_tail;
+			*b_len = free_len;
+		}
+
+		if (free_len >= to_alloc)
+			return wbit + wpos - *prev_tail;
+
+		if (free_bits >= wend) {
+			*prev_tail += free_bits - wpos;
+			return -1;
+		}
+
+		wpos = free_bits + 1;
+
+		*prev_tail = 0;
+	}
+
+	return -1;
+}
+
+/*
+ * wnd_close
+ *
+ *
+ */
+void wnd_close(struct wnd_bitmap *wnd)
+{
+	struct rb_node *node, *next;
+
+	if (wnd->free_bits != wnd->free_holder)
+		ntfs_free(wnd->free_bits);
+	run_close(&wnd->run);
+
+	node = rb_first(&wnd->start_tree);
+
+	while (node) {
+		next = rb_next(node);
+		rb_erase(node, &wnd->start_tree);
+		ntfs_free(rb_entry(node, struct e_node, start.node));
+		node = next;
+	}
+}
+
+static struct rb_node *rb_lookup(struct rb_root *root, size_t v)
+{
+	struct rb_node **p = &root->rb_node;
+	struct rb_node *r = NULL;
+
+	while (*p) {
+		struct rb_node_key *k;
+
+		k = rb_entry(*p, struct rb_node_key, node);
+		if (v < k->key) {
+			p = &(*p)->rb_left;
+		} else if (v > k->key) {
+			r = &k->node;
+			p = &(*p)->rb_right;
+		} else {
+			return &k->node;
+		}
+	}
+
+	return r;
+}
+
+/*
+ * rb_insert_count
+ *
+ * Helper function to insert special kind of 'count' tree
+ */
+static inline bool rb_insert_count(struct rb_root *root, struct e_node *e)
+{
+	struct rb_node **p = &root->rb_node;
+	struct rb_node *parent = NULL;
+	size_t e_ckey = e->count.key;
+	size_t e_skey = e->start.key;
+
+	while (*p) {
+		struct e_node *k =
+			rb_entry(parent = *p, struct e_node, count.node);
+
+		if (e_ckey > k->count.key) {
+			p = &(*p)->rb_left;
+		} else if (e_ckey < k->count.key) {
+			p = &(*p)->rb_right;
+		} else if (e_skey < k->start.key) {
+			p = &(*p)->rb_left;
+		} else if (e_skey > k->start.key) {
+			p = &(*p)->rb_right;
+		} else {
+			WARN_ON(1);
+			return false;
+		}
+	}
+
+	rb_link_node(&e->count.node, parent, p);
+	rb_insert_color(&e->count.node, root);
+	return true;
+}
+
+/*
+ * inline bool rb_insert_start
+ *
+ * Helper function to insert special kind of 'count' tree
+ */
+static inline bool rb_insert_start(struct rb_root *root, struct e_node *e)
+{
+	struct rb_node **p = &root->rb_node;
+	struct rb_node *parent = NULL;
+	size_t e_skey = e->start.key;
+
+	while (*p) {
+		struct e_node *k;
+
+		parent = *p;
+
+		k = rb_entry(parent, struct e_node, start.node);
+		if (e_skey < k->start.key) {
+			p = &(*p)->rb_left;
+		} else if (e_skey > k->start.key) {
+			p = &(*p)->rb_right;
+		} else {
+			WARN_ON(1);
+			return false;
+		}
+	}
+
+	rb_link_node(&e->start.node, parent, p);
+	rb_insert_color(&e->start.node, root);
+	return true;
+}
+
+#define NTFS_MAX_WND_EXTENTS (32u * 1024u)
+
+/*
+ * wnd_add_free_ext
+ *
+ * adds a new extent of free space
+ * build = 1 when building tree
+ */
+static void wnd_add_free_ext(struct wnd_bitmap *wnd, size_t bit, size_t len,
+			     bool build)
+{
+	struct e_node *e, *e0 = NULL;
+	size_t ib, end_in = bit + len;
+	struct rb_node *n;
+
+	if (build) {
+		/* Use extent_min to filter too short extents */
+		if (wnd->count >= NTFS_MAX_WND_EXTENTS &&
+		    len <= wnd->extent_min) {
+			wnd->uptodated = -1;
+			return;
+		}
+	} else {
+		/* Try to find extent before 'bit' */
+		n = rb_lookup(&wnd->start_tree, bit);
+
+		if (!n) {
+			n = rb_first(&wnd->start_tree);
+		} else {
+			e = rb_entry(n, struct e_node, start.node);
+			n = rb_next(n);
+			if (e->start.key + e->count.key == bit) {
+				/* Remove left */
+				bit = e->start.key;
+				len += e->count.key;
+				rb_erase(&e->start.node, &wnd->start_tree);
+				rb_erase(&e->count.node, &wnd->count_tree);
+				wnd->count -= 1;
+				e0 = e;
+			}
+		}
+
+		while (n) {
+			size_t next_end;
+
+			e = rb_entry(n, struct e_node, start.node);
+			next_end = e->start.key + e->count.key;
+			if (e->start.key > end_in)
+				break;
+
+			/* Remove right */
+			n = rb_next(n);
+			len += next_end - end_in;
+			end_in = next_end;
+			rb_erase(&e->start.node, &wnd->start_tree);
+			rb_erase(&e->count.node, &wnd->count_tree);
+			wnd->count -= 1;
+
+			if (!e0)
+				e0 = e;
+			else
+				ntfs_free(e);
+		}
+
+		if (wnd->uptodated != 1) {
+			/* Check bits before 'bit' */
+			ib = wnd->zone_bit == wnd->zone_end ||
+					     bit < wnd->zone_end ?
+				     0 :
+				     wnd->zone_end;
+
+			while (bit > ib && wnd_is_free_hlp(wnd, bit - 1, 1)) {
+				bit -= 1;
+				len += 1;
+			}
+
+			/* Check bits after 'end_in' */
+			ib = wnd->zone_bit == wnd->zone_end ||
+					     end_in > wnd->zone_bit ?
+				     wnd->nbits :
+				     wnd->zone_bit;
+
+			while (end_in < ib && wnd_is_free_hlp(wnd, end_in, 1)) {
+				end_in += 1;
+				len += 1;
+			}
+		}
+	}
+	/* Insert new fragment */
+	if (wnd->count >= NTFS_MAX_WND_EXTENTS) {
+		if (e0)
+			ntfs_free(e0);
+
+		wnd->uptodated = -1;
+
+		/* Compare with smallest fragment */
+		n = rb_last(&wnd->count_tree);
+		e = rb_entry(n, struct e_node, count.node);
+		if (len <= e->count.key)
+			goto out; /* Do not insert small fragments */
+
+		if (build) {
+			struct e_node *e2;
+
+			n = rb_prev(n);
+			e2 = rb_entry(n, struct e_node, count.node);
+			/* smallest fragment will be 'e2->count.key' */
+			wnd->extent_min = e2->count.key;
+		}
+
+		/* Replace smallest fragment by new one */
+		rb_erase(&e->start.node, &wnd->start_tree);
+		rb_erase(&e->count.node, &wnd->count_tree);
+		wnd->count -= 1;
+	} else {
+		e = e0 ? e0 : ntfs_alloc(sizeof(struct e_node), 0);
+		if (!e) {
+			wnd->uptodated = -1;
+			goto out;
+		}
+
+		if (build && len <= wnd->extent_min)
+			wnd->extent_min = len;
+	}
+	e->start.key = bit;
+	e->count.key = len;
+	if (len > wnd->extent_max)
+		wnd->extent_max = len;
+
+	rb_insert_start(&wnd->start_tree, e);
+	rb_insert_count(&wnd->count_tree, e);
+	wnd->count += 1;
+
+out:;
+}
+
+/*
+ * wnd_remove_free_ext
+ *
+ * removes a run from the cached free space
+ */
+static void wnd_remove_free_ext(struct wnd_bitmap *wnd, size_t bit, size_t len)
+{
+	struct rb_node *n, *n3;
+	struct e_node *e, *e3;
+	size_t end_in = bit + len;
+	size_t end3, end, new_key, new_len, max_new_len;
+
+	/* Try to find extent before 'bit' */
+	n = rb_lookup(&wnd->start_tree, bit);
+
+	if (!n)
+		return;
+
+	e = rb_entry(n, struct e_node, start.node);
+	end = e->start.key + e->count.key;
+
+	new_key = new_len = 0;
+	len = e->count.key;
+
+	/* Range [bit,end_in) must be inside 'e' or outside 'e' and 'n' */
+	if (e->start.key > bit)
+		;
+	else if (end_in <= end) {
+		/* Range [bit,end_in) inside 'e' */
+		new_key = end_in;
+		new_len = end - end_in;
+		len = bit - e->start.key;
+	} else if (bit > end) {
+		bool bmax = false;
+
+		n3 = rb_next(n);
+
+		while (n3) {
+			e3 = rb_entry(n3, struct e_node, start.node);
+			if (e3->start.key >= end_in)
+				break;
+
+			if (e3->count.key == wnd->extent_max)
+				bmax = true;
+
+			end3 = e3->start.key + e3->count.key;
+			if (end3 > end_in) {
+				e3->start.key = end_in;
+				rb_erase(&e3->count.node, &wnd->count_tree);
+				e3->count.key = end3 - end_in;
+				rb_insert_count(&wnd->count_tree, e3);
+				break;
+			}
+
+			n3 = rb_next(n3);
+			rb_erase(&e3->start.node, &wnd->start_tree);
+			rb_erase(&e3->count.node, &wnd->count_tree);
+			wnd->count -= 1;
+			ntfs_free(e3);
+		}
+		if (!bmax)
+			return;
+		n3 = rb_first(&wnd->count_tree);
+		wnd->extent_max =
+			n3 ? rb_entry(n3, struct e_node, count.node)->count.key :
+			     0;
+		return;
+	}
+
+	if (e->count.key != wnd->extent_max) {
+		;
+	} else if (rb_prev(&e->count.node)) {
+		;
+	} else {
+		n3 = rb_next(&e->count.node);
+		max_new_len = len > new_len ? len : new_len;
+		if (!n3) {
+			wnd->extent_max = max_new_len;
+		} else {
+			e3 = rb_entry(n3, struct e_node, count.node);
+			wnd->extent_max = max(e3->count.key, max_new_len);
+		}
+	}
+
+	if (!len) {
+		if (new_len) {
+			e->start.key = new_key;
+			rb_erase(&e->count.node, &wnd->count_tree);
+			e->count.key = new_len;
+			rb_insert_count(&wnd->count_tree, e);
+		} else {
+			rb_erase(&e->start.node, &wnd->start_tree);
+			rb_erase(&e->count.node, &wnd->count_tree);
+			wnd->count -= 1;
+			ntfs_free(e);
+		}
+		goto out;
+	}
+	rb_erase(&e->count.node, &wnd->count_tree);
+	e->count.key = len;
+	rb_insert_count(&wnd->count_tree, e);
+
+	if (!new_len)
+		goto out;
+
+	if (wnd->count >= NTFS_MAX_WND_EXTENTS) {
+		wnd->uptodated = -1;
+
+		/* Get minimal extent */
+		e = rb_entry(rb_last(&wnd->count_tree), struct e_node,
+			     count.node);
+		if (e->count.key > new_len)
+			goto out;
+
+		/* Replace minimum */
+		rb_erase(&e->start.node, &wnd->start_tree);
+		rb_erase(&e->count.node, &wnd->count_tree);
+		wnd->count -= 1;
+	} else {
+		e = ntfs_alloc(sizeof(struct e_node), 0);
+		if (!e)
+			wnd->uptodated = -1;
+	}
+
+	if (e) {
+		e->start.key = new_key;
+		e->count.key = new_len;
+		rb_insert_start(&wnd->start_tree, e);
+		rb_insert_count(&wnd->count_tree, e);
+		wnd->count += 1;
+	}
+
+out:
+	if (!wnd->count && 1 != wnd->uptodated)
+		wnd_rescan(wnd);
+}
+
+/*
+ * wnd_rescan
+ *
+ * Scan all bitmap. used while initialization.
+ */
+static int wnd_rescan(struct wnd_bitmap *wnd)
+{
+	int err = 0;
+	size_t prev_tail = 0;
+	struct super_block *sb = wnd->sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	u64 lbo, len = 0;
+	u32 blocksize = sb->s_blocksize;
+	u8 cluster_bits = sbi->cluster_bits;
+	u32 wbits = 8 * sb->s_blocksize;
+	u32 used, frb;
+	const ulong *buf;
+	size_t wpos, wbit, iw, vbo;
+	struct buffer_head *bh = NULL;
+	CLST lcn, clen;
+
+	wnd->uptodated = 0;
+	wnd->extent_max = 0;
+	wnd->extent_min = MINUS_ONE_T;
+	wnd->total_zeroes = 0;
+
+	vbo = 0;
+
+	for (iw = 0; iw < wnd->nwnd; iw++) {
+		if (iw + 1 == wnd->nwnd)
+			wbits = wnd->bits_last;
+
+		if (wnd->inited) {
+			if (!wnd->free_bits[iw]) {
+				/* all ones */
+				if (prev_tail) {
+					wnd_add_free_ext(wnd,
+							 vbo * 8 - prev_tail,
+							 prev_tail, true);
+					prev_tail = 0;
+				}
+				goto next_wnd;
+			}
+			if (wbits == wnd->free_bits[iw]) {
+				/* all zeroes */
+				prev_tail += wbits;
+				wnd->total_zeroes += wbits;
+				goto next_wnd;
+			}
+		}
+
+		if (!len) {
+			if (!run_lookup_entry(&wnd->run, vbo >> cluster_bits,
+					      &lcn, &clen, NULL)) {
+				err = -ENOENT;
+				goto out;
+			}
+
+			lbo = (u64)lcn << cluster_bits;
+			len = (u64)clen << cluster_bits;
+		}
+
+		bh = ntfs_bread(sb, lbo >> sb->s_blocksize_bits);
+		if (!bh) {
+			err = -EIO;
+			goto out;
+		}
+
+		buf = (ulong *)bh->b_data;
+
+		used = __bitmap_weight(buf, wbits);
+		if (used < wbits) {
+			frb = wbits - used;
+			wnd->free_bits[iw] = frb;
+			wnd->total_zeroes += frb;
+		}
+
+		wpos = 0;
+		wbit = vbo * 8;
+
+		if (wbit + wbits > wnd->nbits)
+			wbits = wnd->nbits - wbit;
+
+		do {
+			used = find_next_zero_bit(buf, wbits, wpos);
+
+			if (used > wpos && prev_tail) {
+				wnd_add_free_ext(wnd, wbit + wpos - prev_tail,
+						 prev_tail, true);
+				prev_tail = 0;
+			}
+
+			wpos = used;
+
+			if (wpos >= wbits) {
+				/* No free blocks */
+				prev_tail = 0;
+				break;
+			}
+
+			frb = find_next_bit(buf, wbits, wpos);
+			if (frb >= wbits) {
+				/* keep last free block */
+				prev_tail += frb - wpos;
+				break;
+			}
+
+			wnd_add_free_ext(wnd, wbit + wpos - prev_tail,
+					 frb + prev_tail - wpos, true);
+
+			/* Skip free block and first '1' */
+			wpos = frb + 1;
+			/* Reset previous tail */
+			prev_tail = 0;
+		} while (wpos < wbits);
+
+next_wnd:
+
+		if (bh)
+			put_bh(bh);
+		bh = NULL;
+
+		vbo += blocksize;
+		if (len) {
+			len -= blocksize;
+			lbo += blocksize;
+		}
+	}
+
+	/* Add last block */
+	if (prev_tail)
+		wnd_add_free_ext(wnd, wnd->nbits - prev_tail, prev_tail, true);
+
+	/*
+	 * Before init cycle wnd->uptodated was 0
+	 * If any errors or limits occurs while initialization then
+	 * wnd->uptodated will be -1
+	 * If 'uptodated' is still 0 then Tree is really updated
+	 */
+	if (!wnd->uptodated)
+		wnd->uptodated = 1;
+
+	if (wnd->zone_bit != wnd->zone_end) {
+		size_t zlen = wnd->zone_end - wnd->zone_bit;
+
+		wnd->zone_end = wnd->zone_bit;
+		wnd_zone_set(wnd, wnd->zone_bit, zlen);
+	}
+
+out:
+	return err;
+}
+
+/*
+ * wnd_init
+ */
+int wnd_init(struct wnd_bitmap *wnd, struct super_block *sb, size_t nbits)
+{
+	int err;
+	u32 blocksize = sb->s_blocksize;
+	u32 wbits = blocksize * 8;
+
+	init_rwsem(&wnd->rw_lock);
+
+	wnd->sb = sb;
+	wnd->nbits = nbits;
+	wnd->total_zeroes = nbits;
+	wnd->extent_max = MINUS_ONE_T;
+	wnd->zone_bit = wnd->zone_end = 0;
+	wnd->nwnd = bytes_to_block(sb, bitmap_size(nbits));
+	wnd->bits_last = nbits & (wbits - 1);
+	if (!wnd->bits_last)
+		wnd->bits_last = wbits;
+
+	if (wnd->nwnd <= ARRAY_SIZE(wnd->free_holder)) {
+		wnd->free_bits = wnd->free_holder;
+	} else {
+		wnd->free_bits = ntfs_alloc(wnd->nwnd * sizeof(u16), 1);
+		if (!wnd->free_bits)
+			return -ENOMEM;
+	}
+
+	err = wnd_rescan(wnd);
+	if (err)
+		return err;
+
+	wnd->inited = true;
+
+	return 0;
+}
+
+/*
+ * wnd_map
+ *
+ * call sb_bread for requested window
+ */
+static struct buffer_head *wnd_map(struct wnd_bitmap *wnd, size_t iw)
+{
+	size_t vbo;
+	CLST lcn, clen;
+	struct super_block *sb = wnd->sb;
+	struct ntfs_sb_info *sbi;
+	struct buffer_head *bh;
+	u64 lbo;
+
+	sbi = sb->s_fs_info;
+	vbo = (u64)iw << sb->s_blocksize_bits;
+
+	if (!run_lookup_entry(&wnd->run, vbo >> sbi->cluster_bits, &lcn, &clen,
+			      NULL)) {
+		return ERR_PTR(-ENOENT);
+	}
+
+	lbo = ((u64)lcn << sbi->cluster_bits) + (vbo & sbi->cluster_mask);
+
+	bh = ntfs_bread(wnd->sb, lbo >> sb->s_blocksize_bits);
+	if (!bh)
+		return ERR_PTR(-EIO);
+
+	return bh;
+}
+
+/*
+ * wnd_set_free
+ *
+ * Marks the bits range from bit to bit + bits as free
+ */
+int wnd_set_free(struct wnd_bitmap *wnd, size_t bit, size_t bits)
+{
+	int err = 0;
+	struct super_block *sb = wnd->sb;
+	size_t bits0 = bits;
+	u32 wbits = 8 * sb->s_blocksize;
+	size_t iw = bit >> (sb->s_blocksize_bits + 3);
+	u32 wbit = bit & (wbits - 1);
+	struct buffer_head *bh;
+
+	while (iw < wnd->nwnd && bits) {
+		u32 tail, op;
+		ulong *buf;
+
+		if (iw + 1 == wnd->nwnd)
+			wbits = wnd->bits_last;
+
+		tail = wbits - wbit;
+		op = tail < bits ? tail : bits;
+
+		bh = wnd_map(wnd, iw);
+		if (IS_ERR(bh)) {
+			err = PTR_ERR(bh);
+			break;
+		}
+
+		buf = (ulong *)bh->b_data;
+
+		lock_buffer(bh);
+
+		__bitmap_clear(buf, wbit, op);
+
+		wnd->free_bits[iw] += op;
+
+		set_buffer_uptodate(bh);
+		mark_buffer_dirty(bh);
+		unlock_buffer(bh);
+		put_bh(bh);
+
+		wnd->total_zeroes += op;
+		bits -= op;
+		wbit = 0;
+		iw += 1;
+	}
+
+	wnd_add_free_ext(wnd, bit, bits0, false);
+
+	return err;
+}
+
+/*
+ * wnd_set_used
+ *
+ * Marks the bits range from bit to bit + bits as used
+ */
+int wnd_set_used(struct wnd_bitmap *wnd, size_t bit, size_t bits)
+{
+	int err = 0;
+	struct super_block *sb = wnd->sb;
+	size_t bits0 = bits;
+	size_t iw = bit >> (sb->s_blocksize_bits + 3);
+	u32 wbits = 8 * sb->s_blocksize;
+	u32 wbit = bit & (wbits - 1);
+	struct buffer_head *bh;
+
+	while (iw < wnd->nwnd && bits) {
+		u32 tail, op;
+		ulong *buf;
+
+		if (unlikely(iw + 1 == wnd->nwnd))
+			wbits = wnd->bits_last;
+
+		tail = wbits - wbit;
+		op = tail < bits ? tail : bits;
+
+		bh = wnd_map(wnd, iw);
+		if (IS_ERR(bh)) {
+			err = PTR_ERR(bh);
+			break;
+		}
+		buf = (ulong *)bh->b_data;
+
+		lock_buffer(bh);
+
+		__bitmap_set(buf, wbit, op);
+		wnd->free_bits[iw] -= op;
+
+		set_buffer_uptodate(bh);
+		mark_buffer_dirty(bh);
+		unlock_buffer(bh);
+		put_bh(bh);
+
+		wnd->total_zeroes -= op;
+		bits -= op;
+		wbit = 0;
+		iw += 1;
+	}
+
+	if (!RB_EMPTY_ROOT(&wnd->start_tree))
+		wnd_remove_free_ext(wnd, bit, bits0);
+
+	return err;
+}
+
+/*
+ * wnd_is_free_hlp
+ *
+ * Returns true if all clusters [bit, bit+bits) are free (bitmap only)
+ */
+static bool wnd_is_free_hlp(struct wnd_bitmap *wnd, size_t bit, size_t bits)
+{
+	struct super_block *sb = wnd->sb;
+	size_t iw = bit >> (sb->s_blocksize_bits + 3);
+	u32 wbits = 8 * sb->s_blocksize;
+	u32 wbit = bit & (wbits - 1);
+
+	while (iw < wnd->nwnd && bits) {
+		u32 tail, op;
+
+		if (unlikely(iw + 1 == wnd->nwnd))
+			wbits = wnd->bits_last;
+
+		tail = wbits - wbit;
+		op = tail < bits ? tail : bits;
+
+		if (wbits != wnd->free_bits[iw]) {
+			bool ret;
+			struct buffer_head *bh = wnd_map(wnd, iw);
+
+			if (IS_ERR(bh))
+				return false;
+
+			ret = are_bits_clear((ulong *)bh->b_data, wbit, op);
+
+			put_bh(bh);
+			if (!ret)
+				return false;
+		}
+
+		bits -= op;
+		wbit = 0;
+		iw += 1;
+	}
+
+	return true;
+}
+
+/*
+ * wnd_is_free
+ *
+ * Returns true if all clusters [bit, bit+bits) are free
+ */
+bool wnd_is_free(struct wnd_bitmap *wnd, size_t bit, size_t bits)
+{
+	bool ret;
+	struct rb_node *n;
+	size_t end;
+	struct e_node *e;
+
+	if (RB_EMPTY_ROOT(&wnd->start_tree))
+		goto use_wnd;
+
+	n = rb_lookup(&wnd->start_tree, bit);
+	if (!n)
+		goto use_wnd;
+
+	e = rb_entry(n, struct e_node, start.node);
+
+	end = e->start.key + e->count.key;
+
+	if (bit < end && bit + bits <= end)
+		return true;
+
+use_wnd:
+	ret = wnd_is_free_hlp(wnd, bit, bits);
+
+	return ret;
+}
+
+/*
+ * wnd_is_used
+ *
+ * Returns true if all clusters [bit, bit+bits) are used
+ */
+bool wnd_is_used(struct wnd_bitmap *wnd, size_t bit, size_t bits)
+{
+	bool ret = false;
+	struct super_block *sb = wnd->sb;
+	size_t iw = bit >> (sb->s_blocksize_bits + 3);
+	u32 wbits = 8 * sb->s_blocksize;
+	u32 wbit = bit & (wbits - 1);
+	size_t end;
+	struct rb_node *n;
+	struct e_node *e;
+
+	if (RB_EMPTY_ROOT(&wnd->start_tree))
+		goto use_wnd;
+
+	end = bit + bits;
+	n = rb_lookup(&wnd->start_tree, end - 1);
+	if (!n)
+		goto use_wnd;
+
+	e = rb_entry(n, struct e_node, start.node);
+	if (e->start.key + e->count.key > bit)
+		return false;
+
+use_wnd:
+	while (iw < wnd->nwnd && bits) {
+		u32 tail, op;
+
+		if (unlikely(iw + 1 == wnd->nwnd))
+			wbits = wnd->bits_last;
+
+		tail = wbits - wbit;
+		op = tail < bits ? tail : bits;
+
+		if (wnd->free_bits[iw]) {
+			bool ret;
+			struct buffer_head *bh = wnd_map(wnd, iw);
+
+			if (IS_ERR(bh))
+				goto out;
+
+			ret = are_bits_set((ulong *)bh->b_data, wbit, op);
+			put_bh(bh);
+			if (!ret)
+				goto out;
+		}
+
+		bits -= op;
+		wbit = 0;
+		iw += 1;
+	}
+	ret = true;
+
+out:
+	return ret;
+}
+
+/*
+ * wnd_find
+ * - flags - BITMAP_FIND_XXX flags
+ *
+ * looks for free space
+ * Returns 0 if not found
+ */
+size_t wnd_find(struct wnd_bitmap *wnd, size_t to_alloc, size_t hint,
+		size_t flags, size_t *allocated)
+{
+	struct super_block *sb;
+	u32 wbits, wpos, wzbit, wzend;
+	size_t fnd, max_alloc, b_len, b_pos;
+	size_t iw, prev_tail, nwnd, wbit, ebit, zbit, zend;
+	size_t to_alloc0 = to_alloc;
+	const ulong *buf;
+	const struct e_node *e;
+	const struct rb_node *pr, *cr;
+	u8 log2_bits;
+	bool fbits_valid;
+	struct buffer_head *bh;
+
+	/* fast checking for available free space */
+	if (flags & BITMAP_FIND_FULL) {
+		size_t zeroes = wnd_zeroes(wnd);
+
+		zeroes -= wnd->zone_end - wnd->zone_bit;
+		if (zeroes < to_alloc0)
+			goto no_space;
+
+		if (to_alloc0 > wnd->extent_max)
+			goto no_space;
+	} else {
+		if (to_alloc > wnd->extent_max)
+			to_alloc = wnd->extent_max;
+	}
+
+	if (wnd->zone_bit <= hint && hint < wnd->zone_end)
+		hint = wnd->zone_end;
+
+	max_alloc = wnd->nbits;
+	b_len = b_pos = 0;
+
+	if (hint >= max_alloc)
+		hint = 0;
+
+	if (RB_EMPTY_ROOT(&wnd->start_tree)) {
+		if (wnd->uptodated == 1) {
+			/* extents tree is updated -> no free space */
+			goto no_space;
+		}
+		goto scan_bitmap;
+	}
+
+	e = NULL;
+	if (!hint)
+		goto allocate_biggest;
+
+	/* Use hint: enumerate extents by start >= hint */
+	pr = NULL;
+	cr = wnd->start_tree.rb_node;
+
+	for (;;) {
+		e = rb_entry(cr, struct e_node, start.node);
+
+		if (e->start.key == hint)
+			break;
+
+		if (e->start.key < hint) {
+			pr = cr;
+			cr = cr->rb_right;
+			if (!cr)
+				break;
+			continue;
+		}
+
+		cr = cr->rb_left;
+		if (!cr) {
+			e = pr ? rb_entry(pr, struct e_node, start.node) : NULL;
+			break;
+		}
+	}
+
+	if (!e)
+		goto allocate_biggest;
+
+	if (e->start.key + e->count.key > hint) {
+		/* We have found extension with 'hint' inside */
+		size_t len = e->start.key + e->count.key - hint;
+
+		if (len >= to_alloc && hint + to_alloc <= max_alloc) {
+			fnd = hint;
+			goto found;
+		}
+
+		if (!(flags & BITMAP_FIND_FULL)) {
+			if (len > to_alloc)
+				len = to_alloc;
+
+			if (hint + len <= max_alloc) {
+				fnd = hint;
+				to_alloc = len;
+				goto found;
+			}
+		}
+	}
+
+allocate_biggest:
+	/* Allocate from biggest free extent */
+	e = rb_entry(rb_first(&wnd->count_tree), struct e_node, count.node);
+	if (e->count.key != wnd->extent_max)
+		wnd->extent_max = e->count.key;
+
+	if (e->count.key < max_alloc) {
+		if (e->count.key >= to_alloc) {
+			;
+		} else if (flags & BITMAP_FIND_FULL) {
+			if (e->count.key < to_alloc0) {
+				/* Biggest free block is less then requested */
+				goto no_space;
+			}
+			to_alloc = e->count.key;
+		} else if (-1 != wnd->uptodated) {
+			to_alloc = e->count.key;
+		} else {
+			/* Check if we can use more bits */
+			size_t op, max_check;
+			struct rb_root start_tree;
+
+			memcpy(&start_tree, &wnd->start_tree,
+			       sizeof(struct rb_root));
+			memset(&wnd->start_tree, 0, sizeof(struct rb_root));
+
+			max_check = e->start.key + to_alloc;
+			if (max_check > max_alloc)
+				max_check = max_alloc;
+			for (op = e->start.key + e->count.key; op < max_check;
+			     op++) {
+				if (!wnd_is_free(wnd, op, 1))
+					break;
+			}
+			memcpy(&wnd->start_tree, &start_tree,
+			       sizeof(struct rb_root));
+			to_alloc = op - e->start.key;
+		}
+
+		/* Prepare to return */
+		fnd = e->start.key;
+		if (e->start.key + to_alloc > max_alloc)
+			to_alloc = max_alloc - e->start.key;
+		goto found;
+	}
+
+	if (wnd->uptodated == 1) {
+		/* extents tree is updated -> no free space */
+		goto no_space;
+	}
+
+	b_len = e->count.key;
+	b_pos = e->start.key;
+
+scan_bitmap:
+	sb = wnd->sb;
+	log2_bits = sb->s_blocksize_bits + 3;
+
+	/* At most two ranges [hint, max_alloc) + [0, hint) */
+Again:
+
+	/* TODO: optimize request for case nbits > wbits */
+	iw = hint >> log2_bits;
+	wbits = sb->s_blocksize * 8;
+	wpos = hint & (wbits - 1);
+	prev_tail = 0;
+	fbits_valid = true;
+
+	if (max_alloc == wnd->nbits) {
+		nwnd = wnd->nwnd;
+	} else {
+		size_t t = max_alloc + wbits - 1;
+
+		nwnd = likely(t > max_alloc) ? (t >> log2_bits) : wnd->nwnd;
+	}
+
+	/* Enumerate all windows */
+	for (; iw < nwnd; iw++) {
+		wbit = iw << log2_bits;
+
+		if (!wnd->free_bits[iw]) {
+			if (prev_tail > b_len) {
+				b_pos = wbit - prev_tail;
+				b_len = prev_tail;
+			}
+
+			/* Skip full used window */
+			prev_tail = 0;
+			wpos = 0;
+			continue;
+		}
+
+		if (unlikely(iw + 1 == nwnd)) {
+			if (max_alloc == wnd->nbits) {
+				wbits = wnd->bits_last;
+			} else {
+				size_t t = max_alloc & (wbits - 1);
+
+				if (t) {
+					wbits = t;
+					fbits_valid = false;
+				}
+			}
+		}
+
+		if (wnd->zone_end > wnd->zone_bit) {
+			ebit = wbit + wbits;
+			zbit = max(wnd->zone_bit, wbit);
+			zend = min(wnd->zone_end, ebit);
+
+			/* Here we have a window [wbit, ebit) and zone [zbit, zend) */
+			if (zend <= zbit) {
+				/* Zone does not overlap window */
+			} else {
+				wzbit = zbit - wbit;
+				wzend = zend - wbit;
+
+				/* Zone overlaps window */
+				if (wnd->free_bits[iw] == wzend - wzbit) {
+					prev_tail = 0;
+					wpos = 0;
+					continue;
+				}
+
+				/* Scan two ranges window: [wbit, zbit) and [zend, ebit) */
+				bh = wnd_map(wnd, iw);
+
+				if (IS_ERR(bh)) {
+					/* TODO: error */
+					prev_tail = 0;
+					wpos = 0;
+					continue;
+				}
+
+				buf = (ulong *)bh->b_data;
+
+				/* Scan range [wbit, zbit) */
+				if (wpos < wzbit) {
+					/* Scan range [wpos, zbit) */
+					fnd = wnd_scan(buf, wbit, wpos, wzbit,
+						       to_alloc, &prev_tail,
+						       &b_pos, &b_len);
+					if (fnd != MINUS_ONE_T) {
+						put_bh(bh);
+						goto found;
+					}
+				}
+
+				prev_tail = 0;
+
+				/* Scan range [zend, ebit) */
+				if (wzend < wbits) {
+					fnd = wnd_scan(buf, wbit,
+						       max(wzend, wpos), wbits,
+						       to_alloc, &prev_tail,
+						       &b_pos, &b_len);
+					if (fnd != MINUS_ONE_T) {
+						put_bh(bh);
+						goto found;
+					}
+				}
+
+				wpos = 0;
+				put_bh(bh);
+				continue;
+			}
+		}
+
+		/* Current window does not overlap zone */
+		if (!wpos && fbits_valid && wnd->free_bits[iw] == wbits) {
+			/* window is empty */
+			if (prev_tail + wbits >= to_alloc) {
+				fnd = wbit + wpos - prev_tail;
+				goto found;
+			}
+
+			/* Increase 'prev_tail' and process next window */
+			prev_tail += wbits;
+			wpos = 0;
+			continue;
+		}
+
+		/* read window */
+		bh = wnd_map(wnd, iw);
+		if (IS_ERR(bh)) {
+			// TODO: error
+			prev_tail = 0;
+			wpos = 0;
+			continue;
+		}
+
+		buf = (ulong *)bh->b_data;
+
+		/* Scan range [wpos, eBits) */
+		fnd = wnd_scan(buf, wbit, wpos, wbits, to_alloc, &prev_tail,
+			       &b_pos, &b_len);
+		put_bh(bh);
+		if (fnd != MINUS_ONE_T)
+			goto found;
+	}
+
+	if (b_len < prev_tail) {
+		/* The last fragment */
+		b_len = prev_tail;
+		b_pos = max_alloc - prev_tail;
+	}
+
+	if (hint) {
+		/*
+		 * We have scanned range [hint max_alloc)
+		 * Prepare to scan range [0 hint + to_alloc)
+		 */
+		size_t nextmax = hint + to_alloc;
+
+		if (likely(nextmax >= hint) && nextmax < max_alloc)
+			max_alloc = nextmax;
+		hint = 0;
+		goto Again;
+	}
+
+	if (!b_len)
+		goto no_space;
+
+	wnd->extent_max = b_len;
+
+	if (flags & BITMAP_FIND_FULL)
+		goto no_space;
+
+	fnd = b_pos;
+	to_alloc = b_len;
+
+found:
+	if (flags & BITMAP_FIND_MARK_AS_USED) {
+		/* TODO optimize remove extent (pass 'e'?) */
+		if (wnd_set_used(wnd, fnd, to_alloc))
+			goto no_space;
+	} else if (wnd->extent_max != MINUS_ONE_T &&
+		   to_alloc > wnd->extent_max) {
+		wnd->extent_max = to_alloc;
+	}
+
+	*allocated = fnd;
+	return to_alloc;
+
+no_space:
+	return 0;
+}
+
+/*
+ * wnd_extend
+ *
+ * Extend bitmap ($MFT bitmap)
+ */
+int wnd_extend(struct wnd_bitmap *wnd, size_t new_bits)
+{
+	int err;
+	struct super_block *sb = wnd->sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	u32 blocksize = sb->s_blocksize;
+	u32 wbits = blocksize * 8;
+	u32 b0, new_last;
+	size_t bits, iw, new_wnd;
+	size_t old_bits = wnd->nbits;
+	u16 *new_free;
+
+	if (new_bits <= old_bits)
+		return -EINVAL;
+
+	/* align to 8 byte boundary */
+	new_wnd = bytes_to_block(sb, bitmap_size(new_bits));
+	new_last = new_bits & (wbits - 1);
+	if (!new_last)
+		new_last = wbits;
+
+	if (new_wnd != wnd->nwnd) {
+		if (new_wnd <= ARRAY_SIZE(wnd->free_holder)) {
+			new_free = wnd->free_holder;
+		} else {
+			new_free = ntfs_alloc(new_wnd * sizeof(u16), 0);
+			if (!new_free)
+				return -ENOMEM;
+		}
+
+		if (new_free != wnd->free_bits)
+			memcpy(new_free, wnd->free_bits,
+			       wnd->nwnd * sizeof(short));
+		memset(new_free + wnd->nwnd, 0,
+		       (new_wnd - wnd->nwnd) * sizeof(short));
+		if (wnd->free_bits != wnd->free_holder)
+			ntfs_free(wnd->free_bits);
+
+		wnd->free_bits = new_free;
+	}
+
+	/* Zero bits [old_bits,new_bits) */
+	bits = new_bits - old_bits;
+	b0 = old_bits & (wbits - 1);
+
+	for (iw = old_bits >> (sb->s_blocksize_bits + 3); bits; iw += 1) {
+		u32 op;
+		size_t frb;
+		u64 vbo, lbo, bytes;
+		struct buffer_head *bh;
+		ulong *buf;
+
+		if (iw + 1 == new_wnd)
+			wbits = new_last;
+
+		op = b0 + bits > wbits ? wbits - b0 : bits;
+		vbo = (u64)iw * blocksize;
+
+		err = ntfs_vbo_to_lbo(sbi, &wnd->run, vbo, &lbo, &bytes);
+		if (err)
+			break;
+
+		bh = ntfs_bread(sb, lbo >> sb->s_blocksize_bits);
+		if (!bh)
+			return -EIO;
+
+		lock_buffer(bh);
+		buf = (ulong *)bh->b_data;
+
+		__bitmap_clear(buf, b0, blocksize * 8 - b0);
+		frb = wbits - __bitmap_weight(buf, wbits);
+		wnd->total_zeroes += frb - wnd->free_bits[iw];
+		wnd->free_bits[iw] = frb;
+
+		set_buffer_uptodate(bh);
+		mark_buffer_dirty(bh);
+		unlock_buffer(bh);
+		/*err = sync_dirty_buffer(bh);*/
+
+		b0 = 0;
+		bits -= op;
+	}
+
+	wnd->nbits = new_bits;
+	wnd->nwnd = new_wnd;
+	wnd->bits_last = new_last;
+
+	wnd_add_free_ext(wnd, old_bits, new_bits - old_bits, false);
+
+	return 0;
+}
+
+/*
+ * wnd_zone_set
+ */
+void wnd_zone_set(struct wnd_bitmap *wnd, size_t lcn, size_t len)
+{
+	size_t zlen;
+
+	zlen = wnd->zone_end - wnd->zone_bit;
+	if (zlen)
+		wnd_add_free_ext(wnd, wnd->zone_bit, zlen, false);
+
+	if (!RB_EMPTY_ROOT(&wnd->start_tree) && len)
+		wnd_remove_free_ext(wnd, lcn, len);
+
+	wnd->zone_bit = lcn;
+	wnd->zone_end = lcn + len;
+}
+
+int ntfs_trim_fs(struct ntfs_sb_info *sbi, struct fstrim_range *range)
+{
+	int err = 0;
+	struct super_block *sb = sbi->sb;
+	struct wnd_bitmap *wnd = &sbi->used.bitmap;
+	u32 wbits = 8 * sb->s_blocksize;
+	CLST len = 0, lcn = 0, done = 0;
+	CLST minlen = bytes_to_cluster(sbi, range->minlen);
+	CLST lcn_from = bytes_to_cluster(sbi, range->start);
+	size_t iw = lcn_from >> (sb->s_blocksize_bits + 3);
+	u32 wbit = lcn_from & (wbits - 1);
+	const ulong *buf;
+	CLST lcn_to;
+
+	if (!minlen)
+		minlen = 1;
+
+	if (range->len == (u64)-1)
+		lcn_to = wnd->nbits;
+	else
+		lcn_to = bytes_to_cluster(sbi, range->start + range->len);
+
+	down_read_nested(&wnd->rw_lock, BITMAP_MUTEX_CLUSTERS);
+
+	for (; iw < wnd->nbits; iw++, wbit = 0) {
+		CLST lcn_wnd = iw * wbits;
+		struct buffer_head *bh;
+
+		if (lcn_wnd > lcn_to)
+			break;
+
+		if (!wnd->free_bits[iw])
+			continue;
+
+		if (iw + 1 == wnd->nwnd)
+			wbits = wnd->bits_last;
+
+		if (lcn_wnd + wbits > lcn_to)
+			wbits = lcn_to - lcn_wnd;
+
+		bh = wnd_map(wnd, iw);
+		if (IS_ERR(bh)) {
+			err = PTR_ERR(bh);
+			break;
+		}
+
+		buf = (ulong *)bh->b_data;
+
+		for (; wbit < wbits; wbit++) {
+			if (!test_bit(wbit, buf)) {
+				if (!len)
+					lcn = lcn_wnd + wbit;
+				len += 1;
+				continue;
+			}
+			if (len >= minlen) {
+				err = ntfs_discard(sbi, lcn, len);
+				if (err)
+					goto out;
+				done += len;
+			}
+			len = 0;
+		}
+		put_bh(bh);
+	}
+
+	/* Process the last fragment */
+	if (len >= minlen) {
+		err = ntfs_discard(sbi, lcn, len);
+		if (err)
+			goto out;
+		done += len;
+	}
+
+out:
+	range->len = (u64)done << sbi->cluster_bits;
+
+	up_read(&wnd->rw_lock);
+
+	return err;
+}
-- 
2.25.4


^ permalink raw reply related	[flat|nested] 21+ messages in thread

* [PATCH v11 04/10] fs/ntfs3: Add file operations and implementation
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
                   ` (2 preceding siblings ...)
  2020-10-30 15:02 ` [PATCH v11 03/10] fs/ntfs3: Add bitmap Konstantin Komarov
@ 2020-10-30 15:02 ` Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 05/10] fs/ntfs3: Add attrib operations Konstantin Komarov
                   ` (8 subsequent siblings)
  12 siblings, 0 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This adds file operations and implementation

Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
---
 fs/ntfs3/dir.c     |  610 ++++++++++
 fs/ntfs3/file.c    | 1153 +++++++++++++++++++
 fs/ntfs3/frecord.c | 2696 ++++++++++++++++++++++++++++++++++++++++++++
 fs/ntfs3/namei.c   |  576 ++++++++++
 fs/ntfs3/record.c  |  613 ++++++++++
 fs/ntfs3/run.c     | 1190 +++++++++++++++++++
 6 files changed, 6838 insertions(+)
 create mode 100644 fs/ntfs3/dir.c
 create mode 100644 fs/ntfs3/file.c
 create mode 100644 fs/ntfs3/frecord.c
 create mode 100644 fs/ntfs3/namei.c
 create mode 100644 fs/ntfs3/record.c
 create mode 100644 fs/ntfs3/run.c

diff --git a/fs/ntfs3/dir.c b/fs/ntfs3/dir.c
new file mode 100644
index 000000000000..f8abe459c39d
--- /dev/null
+++ b/fs/ntfs3/dir.c
@@ -0,0 +1,610 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ *  directory handling functions for ntfs-based filesystems
+ *
+ */
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/iversion.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+/*
+ * Convert little endian utf16 to nls string
+ */
+int ntfs_utf16_to_nls(struct ntfs_sb_info *sbi, const struct le_str *uni,
+		      u8 *buf, int buf_len)
+{
+	int ret, uni_len;
+	const __le16 *ip;
+	u8 *op;
+	struct nls_table *nls = sbi->options.nls[0];
+
+	static_assert(sizeof(wchar_t) == sizeof(__le16));
+
+	if (!nls) {
+		/* utf16 -> utf8 */
+		ret = utf16s_to_utf8s((wchar_t *)uni->name, uni->len,
+				      UTF16_LITTLE_ENDIAN, buf, buf_len);
+		buf[ret] = '\0';
+		return ret;
+	}
+
+	ip = uni->name;
+	op = buf;
+	uni_len = uni->len;
+
+	while (uni_len--) {
+		u16 ec;
+		int charlen;
+
+		if (buf_len < NLS_MAX_CHARSET_SIZE) {
+			ntfs_warn(sbi->sb,
+				  "filename was truncated while converting.");
+			break;
+		}
+
+		ec = le16_to_cpu(*ip++);
+		charlen = nls->uni2char(ec, op, buf_len);
+
+		if (charlen > 0) {
+			op += charlen;
+			buf_len -= charlen;
+		} else {
+			*op++ = ':';
+			op = hex_byte_pack(op, ec >> 8);
+			op = hex_byte_pack(op, ec);
+			buf_len -= 5;
+		}
+	}
+
+	*op = '\0';
+	return op - buf;
+}
+
+// clang-format off
+#define PLANE_SIZE	0x00010000
+
+#define SURROGATE_PAIR	0x0000d800
+#define SURROGATE_LOW	0x00000400
+#define SURROGATE_BITS	0x000003ff
+// clang-format off
+
+static inline void put_utf16(wchar_t *s, unsigned int c, enum utf16_endian endian)
+{
+	switch (endian) {
+	default:
+		*s = (wchar_t)c;
+		break;
+	case UTF16_LITTLE_ENDIAN:
+		*s = __cpu_to_le16(c);
+		break;
+	case UTF16_BIG_ENDIAN:
+		*s = __cpu_to_be16(c);
+		break;
+	}
+}
+
+/*
+ * modified version of 'utf8s_to_utf16s' allows to
+ * detect -ENAMETOOLONG without writing out of expected maximum
+ */
+static int _utf8s_to_utf16s(const u8 *s, int inlen, enum utf16_endian endian,
+			    wchar_t *pwcs, int maxout)
+{
+	u16 *op;
+	int size;
+	unicode_t u;
+
+	op = pwcs;
+	while (inlen > 0 && *s) {
+		if (*s & 0x80) {
+			size = utf8_to_utf32(s, inlen, &u);
+			if (size < 0)
+				return -EINVAL;
+			s += size;
+			inlen -= size;
+
+			if (u >= PLANE_SIZE) {
+				if (maxout < 2)
+					return -ENAMETOOLONG;
+
+				u -= PLANE_SIZE;
+				put_utf16(op++,
+					  SURROGATE_PAIR |
+						  ((u >> 10) & SURROGATE_BITS),
+					  endian);
+				put_utf16(op++,
+					  SURROGATE_PAIR | SURROGATE_LOW |
+						  (u & SURROGATE_BITS),
+					  endian);
+				maxout -= 2;
+			} else {
+				if (maxout < 1)
+					return -ENAMETOOLONG;
+
+				put_utf16(op++, u, endian);
+				maxout--;
+			}
+		} else {
+			if (maxout < 1)
+				return -ENAMETOOLONG;
+
+			put_utf16(op++, *s++, endian);
+			inlen--;
+			maxout--;
+		}
+	}
+	return op - pwcs;
+}
+
+/* helper for ntfs_nls_to_utf16 */
+static int _ntfs_nls_to_utf16(struct nls_table *nls, const u8 *name,
+			      u32 name_len, wchar_t *uname, u32 max_ulen,
+			      enum utf16_endian endian)
+{
+	int ret, slen;
+	const u8 *end;
+
+	if (!nls) {
+		/* utf8 -> utf16 */
+		return _utf8s_to_utf16s(name, name_len, endian, uname,
+					max_ulen);
+	}
+
+	ret = 0;
+
+	for (end = name + name_len; name < end; ret++, name += slen) {
+		if (ret >= max_ulen)
+			return -ENAMETOOLONG;
+
+		slen = nls->char2uni(name, end - name, uname + ret);
+		if (!slen)
+			return -EINVAL;
+		if (slen < 0)
+			return slen;
+	}
+
+#ifdef __BIG_ENDIAN
+	if (endian == UTF16_LITTLE_ENDIAN) {
+		int i = ret;
+
+		while (i--) {
+			__cpu_to_le16s(uname);
+			uname++;
+		}
+	}
+#else
+	if (endian == UTF16_BIG_ENDIAN) {
+		int i = ret;
+
+		while (i--) {
+			__cpu_to_be16s(uname);
+			uname++;
+		}
+	}
+#endif
+
+	return ret;
+}
+
+/*
+ * Convert input string to utf16
+ *
+ * name, name_len - input name
+ * uni, max_ulen - destination memory
+ * endian - endian of target utf16 string
+ *
+ * This function is called:
+ * - to create ntfs name
+ * - to create symlink
+ *
+ * returns utf16 string length or error (if negative)
+ */
+int ntfs_nls_to_utf16(struct ntfs_sb_info *sbi, const u8 *name, u32 name_len,
+		      struct cpu_str *uni, u32 max_ulen,
+		      enum utf16_endian endian)
+{
+	int ret;
+	struct nls_table *nls = sbi->options.nls[0];
+
+	static_assert(sizeof(wchar_t) == sizeof(u16));
+
+	/* use primary nls (may be NULL) */
+	ret = _ntfs_nls_to_utf16(nls, name, name_len, uni->name, max_ulen,
+				 endian);
+
+	if (ret == -ENAMETOOLONG)
+		return -ENAMETOOLONG;
+
+	if (ret > 0)
+		return uni->len = ret;
+
+	if (!sbi->options.nls[1]) {
+		ntfs_warn(sbi->sb, "%s failed to convert \"%.*s\" into utf16",
+			  nls ? nls->charset : "utf8", name_len, name);
+		return ret;
+	}
+
+	/* use alternative nls */
+	ret = _ntfs_nls_to_utf16(sbi->options.nls[1], name, name_len, uni->name,
+				 max_ulen, endian);
+
+	if (ret == -ENAMETOOLONG)
+		return -ENAMETOOLONG;
+
+	if (ret > 0) {
+		/*
+		 * primary nls failed but alternative nls ok
+		 * should we print something?
+		 */
+		return uni->len = ret;
+	}
+
+	ntfs_warn(sbi->sb, "%s and %s failed to convert \"%.*s\" into utf16",
+		  nls ? nls->charset : "utf8", sbi->options.nls[1]->charset,
+		  name_len, name);
+	return ret;
+}
+
+/* helper function */
+struct inode *dir_search_u(struct inode *dir, const struct cpu_str *uni,
+			   struct ntfs_fnd *fnd)
+{
+	int err = 0;
+	struct super_block *sb = dir->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct NTFS_DE *e;
+	int diff;
+	struct inode *inode = NULL;
+	struct ntfs_fnd *fnd_a = NULL;
+
+	if (!fnd) {
+		fnd_a = fnd_get(&ni->dir);
+		if (!fnd_a) {
+			err = -ENOMEM;
+			goto out;
+		}
+		fnd = fnd_a;
+	}
+
+	err = indx_find(&ni->dir, ni, NULL, uni, 0, sbi, &diff, &e, fnd);
+
+	if (err)
+		goto out;
+
+	if (diff) {
+		err = -ENOENT;
+		goto out;
+	}
+
+	inode = ntfs_iget5(sb, &e->ref, uni);
+	if (!IS_ERR(inode) && is_bad_inode(inode)) {
+		iput(inode);
+		err = -EINVAL;
+	}
+out:
+	fnd_put(fnd_a);
+
+	return err == -ENOENT ? NULL : err ? ERR_PTR(err) : inode;
+}
+
+static inline int ntfs_filldir(struct ntfs_sb_info *sbi, struct ntfs_inode *ni,
+			       const struct NTFS_DE *e, u8 *name,
+			       struct dir_context *ctx)
+{
+	const struct ATTR_FILE_NAME *fname;
+	unsigned long ino;
+	int name_len;
+	u32 dt_type;
+
+	fname = Add2Ptr(e, sizeof(struct NTFS_DE));
+
+	if (fname->type == FILE_NAME_DOS)
+		return 0;
+
+	if (!mi_is_ref(&ni->mi, &fname->home))
+		return 0;
+
+	ino = ino_get(&e->ref);
+
+	if (ino == MFT_REC_ROOT)
+		return 0;
+
+	/* Skip meta files ( unless option to show metafiles is set ) */
+	if (!sbi->options.showmeta && ntfs_is_meta_file(sbi, ino))
+		return 0;
+
+	if (sbi->options.nohidden && (fname->dup.fa & FILE_ATTRIBUTE_HIDDEN))
+		return 0;
+
+	name_len = ntfs_utf16_to_nls(sbi, (struct le_str *)&fname->name_len,
+				     name, PATH_MAX);
+	if (name_len <= 0) {
+		ntfs_warn(sbi->sb, "failed to convert name for inode %lx.",
+			  ino);
+		return 0;
+	}
+
+	dt_type = (fname->dup.fa & FILE_ATTRIBUTE_DIRECTORY) ? DT_DIR : DT_REG;
+
+	return !dir_emit(ctx, (s8 *)name, name_len, ino, dt_type);
+}
+
+/*
+ * ntfs_read_hdr
+ *
+ * helper function 'ntfs_readdir'
+ */
+static int ntfs_read_hdr(struct ntfs_sb_info *sbi, struct ntfs_inode *ni,
+			 const struct INDEX_HDR *hdr, u64 vbo, u64 pos,
+			 u8 *name, struct dir_context *ctx)
+{
+	int err;
+	const struct NTFS_DE *e;
+	u32 e_size;
+	u32 end = le32_to_cpu(hdr->used);
+	u32 off = le32_to_cpu(hdr->de_off);
+
+	for (;; off += e_size) {
+		if (off + sizeof(struct NTFS_DE) > end)
+			return -1;
+
+		e = Add2Ptr(hdr, off);
+		e_size = le16_to_cpu(e->size);
+		if (e_size < sizeof(struct NTFS_DE) || off + e_size > end)
+			return -1;
+
+		if (de_is_last(e))
+			return 0;
+
+		/* Skip already enumerated*/
+		if (vbo + off < pos)
+			continue;
+
+		if (le16_to_cpu(e->key_size) < SIZEOF_ATTRIBUTE_FILENAME)
+			return -1;
+
+		ctx->pos = vbo + off;
+
+		/* Submit the name to the filldir callback. */
+		err = ntfs_filldir(sbi, ni, e, name, ctx);
+		if (err)
+			return err;
+	}
+}
+
+/*
+ * file_operations::iterate_shared
+ *
+ * Use non sorted enumeration.
+ * We have an example of broken volume where sorted enumeration
+ * counts each name twice
+ */
+static int ntfs_readdir(struct file *file, struct dir_context *ctx)
+{
+	const struct INDEX_ROOT *root;
+	u64 vbo;
+	size_t bit;
+	loff_t eod;
+	int err = 0;
+	struct inode *dir = file_inode(file);
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct super_block *sb = dir->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	loff_t i_size = dir->i_size;
+	u32 pos = ctx->pos;
+	u8 *name = NULL;
+	struct indx_node *node = NULL;
+	u8 index_bits = ni->dir.index_bits;
+
+	/* name is a buffer of PATH_MAX length */
+	static_assert(NTFS_NAME_LEN * 4 < PATH_MAX);
+
+	if (ni->dir.changed) {
+		ni->dir.changed = false;
+		pos = 0;
+	}
+
+	eod = i_size + sbi->record_size;
+
+	if (pos >= eod)
+		return 0;
+
+	if (!dir_emit_dots(file, ctx))
+		return 0;
+
+	/* allocate PATH_MAX bytes */
+	name = __getname();
+	if (!name)
+		return -ENOMEM;
+
+	ni_lock(ni);
+
+	root = indx_get_root(&ni->dir, ni, NULL, NULL);
+	if (!root) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (pos >= sbi->record_size) {
+		bit = (pos - sbi->record_size) >> index_bits;
+	} else {
+		err = ntfs_read_hdr(sbi, ni, &root->ihdr, 0, pos, name, ctx);
+		if (err)
+			goto out;
+		bit = 0;
+	}
+
+	if (!i_size) {
+		ctx->pos = eod;
+		goto out;
+	}
+
+	for (;;) {
+		vbo = (u64)bit << index_bits;
+		if (vbo >= i_size) {
+			ctx->pos = eod;
+			goto out;
+		}
+
+		err = indx_used_bit(&ni->dir, ni, &bit);
+		if (err)
+			goto out;
+
+		if (bit == MINUS_ONE_T) {
+			ctx->pos = eod;
+			goto out;
+		}
+
+		vbo = (u64)bit << index_bits;
+		if (vbo >= i_size) {
+			ntfs_inode_err(dir, "Looks like your dir is corrupt");
+			err = -EINVAL;
+			goto out;
+		}
+
+		err = indx_read(&ni->dir, ni, bit << ni->dir.idx2vbn_bits,
+				&node);
+		if (err)
+			goto out;
+
+		err = ntfs_read_hdr(sbi, ni, &node->index->ihdr,
+				    vbo + sbi->record_size, pos, name, ctx);
+		if (err)
+			goto out;
+
+		bit += 1;
+	}
+
+out:
+
+	__putname(name);
+	put_indx_node(node);
+
+	if (err == -ENOENT) {
+		err = 0;
+		ctx->pos = pos;
+	}
+
+	ni_unlock(ni);
+
+	return err;
+}
+
+static int ntfs_dir_count(struct inode *dir, bool *is_empty, size_t *dirs,
+			  size_t *files)
+{
+	int err = 0;
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct NTFS_DE *e = NULL;
+	struct INDEX_ROOT *root;
+	struct INDEX_HDR *hdr;
+	const struct ATTR_FILE_NAME *fname;
+	u32 e_size, off, end;
+	u64 vbo = 0;
+	size_t drs = 0, fles = 0, bit = 0;
+	loff_t i_size = ni->vfs_inode.i_size;
+	struct indx_node *node = NULL;
+	u8 index_bits = ni->dir.index_bits;
+
+	if (is_empty)
+		*is_empty = true;
+
+	root = indx_get_root(&ni->dir, ni, NULL, NULL);
+	if (!root)
+		return -EINVAL;
+
+	hdr = &root->ihdr;
+
+	for (;;) {
+		end = le32_to_cpu(hdr->used);
+		off = le32_to_cpu(hdr->de_off);
+
+		for (; off + sizeof(struct NTFS_DE) <= end; off += e_size) {
+			e = Add2Ptr(hdr, off);
+			e_size = le16_to_cpu(e->size);
+			if (e_size < sizeof(struct NTFS_DE) ||
+			    off + e_size > end)
+				break;
+
+			if (de_is_last(e))
+				break;
+
+			fname = de_get_fname(e);
+			if (!fname)
+				continue;
+
+			if (fname->type == FILE_NAME_DOS)
+				continue;
+
+			if (is_empty) {
+				*is_empty = false;
+				if (!dirs && !files)
+					goto out;
+			}
+
+			if (fname->dup.fa & FILE_ATTRIBUTE_DIRECTORY)
+				drs += 1;
+			else
+				fles += 1;
+		}
+
+		if (vbo >= i_size)
+			goto out;
+
+		err = indx_used_bit(&ni->dir, ni, &bit);
+		if (err)
+			goto out;
+
+		if (bit == MINUS_ONE_T)
+			goto out;
+
+		vbo = (u64)bit << index_bits;
+		if (vbo >= i_size)
+			goto out;
+
+		err = indx_read(&ni->dir, ni, bit << ni->dir.idx2vbn_bits,
+				&node);
+		if (err)
+			goto out;
+
+		hdr = &node->index->ihdr;
+		bit += 1;
+		vbo = (u64)bit << ni->dir.idx2vbn_bits;
+	}
+
+out:
+	put_indx_node(node);
+	if (dirs)
+		*dirs = drs;
+	if (files)
+		*files = fles;
+
+	return err;
+}
+
+bool dir_is_empty(struct inode *dir)
+{
+	bool is_empty = false;
+
+	ntfs_dir_count(dir, &is_empty, NULL, NULL);
+
+	return is_empty;
+}
+
+const struct file_operations ntfs_dir_operations = {
+	.llseek = generic_file_llseek,
+	.read = generic_read_dir,
+	.iterate = ntfs_readdir,
+	.fsync = ntfs_file_fsync,
+	.open = ntfs_file_open,
+};
diff --git a/fs/ntfs3/file.c b/fs/ntfs3/file.c
new file mode 100644
index 000000000000..38177db431b4
--- /dev/null
+++ b/fs/ntfs3/file.c
@@ -0,0 +1,1153 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ *  regular file handling primitives for ntfs-based filesystems
+ */
+#include <linux/backing-dev.h>
+#include <linux/buffer_head.h>
+#include <linux/compat.h>
+#include <linux/falloc.h>
+#include <linux/fiemap.h>
+#include <linux/msdos_fs.h> /* FAT_IOCTL_XXX */
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+static int ntfs_ioctl_fitrim(struct ntfs_sb_info *sbi, unsigned long arg)
+{
+	struct fstrim_range __user *user_range;
+	struct fstrim_range range;
+	struct request_queue *q = bdev_get_queue(sbi->sb->s_bdev);
+	int err;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	if (!blk_queue_discard(q))
+		return -EOPNOTSUPP;
+
+	user_range = (struct fstrim_range __user *)arg;
+	if (copy_from_user(&range, user_range, sizeof(range)))
+		return -EFAULT;
+
+	range.minlen = max_t(u32, range.minlen, q->limits.discard_granularity);
+
+	err = ntfs_trim_fs(sbi, &range);
+	if (err < 0)
+		return err;
+
+	if (copy_to_user(user_range, &range, sizeof(range)))
+		return -EFAULT;
+
+	return 0;
+}
+
+static long ntfs_ioctl(struct file *filp, u32 cmd, unsigned long arg)
+{
+	struct inode *inode = file_inode(filp);
+	struct ntfs_sb_info *sbi = inode->i_sb->s_fs_info;
+	u32 __user *user_attr = (u32 __user *)arg;
+
+	switch (cmd) {
+	case FAT_IOCTL_GET_ATTRIBUTES:
+		return put_user(le32_to_cpu(ntfs_i(inode)->std_fa), user_attr);
+
+	case FAT_IOCTL_GET_VOLUME_ID:
+		return put_user(sbi->volume.ser_num, user_attr);
+
+	case FITRIM:
+		return ntfs_ioctl_fitrim(sbi, arg);
+	}
+	return -ENOTTY; /* Inappropriate ioctl for device */
+}
+
+#ifdef CONFIG_COMPAT
+static long ntfs_compat_ioctl(struct file *filp, u32 cmd, unsigned long arg)
+
+{
+	return ntfs_ioctl(filp, cmd, (unsigned long)compat_ptr(arg));
+}
+#endif
+
+/*
+ * inode_operations::getattr
+ */
+int ntfs_getattr(const struct path *path, struct kstat *stat, u32 request_mask,
+		 u32 flags)
+{
+	struct inode *inode = d_inode(path->dentry);
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	if (is_compressed(ni))
+		stat->attributes |= STATX_ATTR_COMPRESSED;
+
+	if (is_encrypted(ni))
+		stat->attributes |= STATX_ATTR_ENCRYPTED;
+
+	stat->attributes_mask |= STATX_ATTR_COMPRESSED | STATX_ATTR_ENCRYPTED;
+
+	generic_fillattr(inode, stat);
+
+	stat->result_mask |= STATX_BTIME;
+	stat->btime = ni->i_crtime;
+
+	return 0;
+}
+
+static int ntfs_extend_initialized_size(struct file *file,
+					struct ntfs_inode *ni,
+					const loff_t valid,
+					const loff_t new_valid)
+{
+	struct inode *inode = &ni->vfs_inode;
+	struct address_space *mapping = inode->i_mapping;
+	struct ntfs_sb_info *sbi = inode->i_sb->s_fs_info;
+	loff_t pos = valid;
+	int err;
+
+	WARN_ON(is_compressed(ni));
+	WARN_ON(valid >= new_valid);
+
+	for (;;) {
+		u32 zerofrom, len;
+		struct page *page;
+		void *fsdata;
+		u8 bits;
+		CLST vcn, lcn, clen;
+
+		if (is_sparsed(ni)) {
+			bits = sbi->cluster_bits;
+			vcn = pos >> bits;
+
+			err = attr_data_get_block(ni, vcn, 0, &lcn, &clen,
+						  NULL);
+			if (err)
+				goto out;
+
+			if (lcn == SPARSE_LCN) {
+				loff_t vbo = (loff_t)vcn << bits;
+				loff_t to = vbo + ((loff_t)clen << bits);
+
+				if (to <= new_valid) {
+					ni->i_valid = to;
+					pos = to;
+					goto next;
+				}
+
+				if (vbo < pos) {
+					pos = vbo;
+				} else {
+					to = (new_valid >> bits) << bits;
+					if (pos < to) {
+						ni->i_valid = to;
+						pos = to;
+						goto next;
+					}
+				}
+			}
+		}
+
+		zerofrom = pos & (PAGE_SIZE - 1);
+		len = PAGE_SIZE - zerofrom;
+
+		if (pos + len > new_valid)
+			len = new_valid - pos;
+
+		err = pagecache_write_begin(file, mapping, pos, len, 0, &page,
+					    &fsdata);
+		if (err)
+			goto out;
+
+		zero_user_segment(page, zerofrom, PAGE_SIZE);
+
+		/* this function in any case puts page*/
+		err = pagecache_write_end(file, mapping, pos, len, len, page,
+					  fsdata);
+		if (err < 0)
+			goto out;
+		pos += len;
+
+next:
+		if (pos >= new_valid)
+			break;
+
+		balance_dirty_pages_ratelimited(mapping);
+		cond_resched();
+	}
+
+	mark_inode_dirty(inode);
+
+	return 0;
+
+out:
+	ni->i_valid = valid;
+	ntfs_inode_warn(inode, "failed to extend initialized size to %llx.",
+			new_valid);
+	return err;
+}
+
+/*
+ * ntfs_sparse_cluster
+ *
+ * Helper function to zero a new allocated clusters
+ */
+void ntfs_sparse_cluster(struct inode *inode, struct page *page0, loff_t vbo,
+			 u32 bytes)
+{
+	struct address_space *mapping = inode->i_mapping;
+	struct ntfs_sb_info *sbi = inode->i_sb->s_fs_info;
+	u32 blocksize = 1 << inode->i_blkbits;
+	pgoff_t idx0 = page0 ? page0->index : -1;
+	loff_t vbo_clst = vbo & sbi->cluster_mask_inv;
+	loff_t end = ntfs_up_cluster(sbi, vbo + bytes);
+	pgoff_t idx = vbo_clst >> PAGE_SHIFT;
+	u32 from = vbo_clst & (PAGE_SIZE - 1);
+	pgoff_t idx_end = (end + PAGE_SIZE - 1) >> PAGE_SHIFT;
+	loff_t page_off;
+	u32 to;
+	bool partial;
+	struct page *page;
+
+	for (; idx < idx_end; idx += 1, from = 0) {
+		page = idx == idx0 ? page0 : grab_cache_page(mapping, idx);
+
+		if (!page)
+			continue;
+
+		page_off = (loff_t)idx << PAGE_SHIFT;
+		to = (page_off + PAGE_SIZE) > end ? (end - page_off) :
+						    PAGE_SIZE;
+		partial = false;
+
+		if ((from || PAGE_SIZE != to) &&
+		    likely(!page_has_buffers(page))) {
+			create_empty_buffers(page, blocksize, 0);
+			if (!page_has_buffers(page)) {
+				ntfs_inode_err(
+					inode,
+					"failed to allocate page buffers.");
+				/*err = -ENOMEM;*/
+				goto unlock_page;
+			}
+		}
+
+		if (page_has_buffers(page)) {
+			struct buffer_head *head, *bh;
+			u32 bh_off = 0;
+
+			bh = head = page_buffers(page);
+			do {
+				u32 bh_next = bh_off + blocksize;
+
+				if (from <= bh_off && bh_next <= to) {
+					set_buffer_uptodate(bh);
+					mark_buffer_dirty(bh);
+				} else if (!buffer_uptodate(bh)) {
+					partial = true;
+				}
+				bh_off = bh_next;
+			} while (head != (bh = bh->b_this_page));
+		}
+
+		zero_user_segment(page, from, to);
+
+		if (!partial) {
+			if (!PageUptodate(page))
+				SetPageUptodate(page);
+			set_page_dirty(page);
+		}
+
+unlock_page:
+		if (idx != idx0) {
+			unlock_page(page);
+			put_page(page);
+		}
+	}
+	mark_inode_dirty(inode);
+}
+
+struct vma_cookie {
+	loff_t to;
+};
+
+/*
+ * vm_operations_struct::close
+ */
+static void ntfs_filemap_close(struct vm_area_struct *vma)
+{
+	struct inode *inode = file_inode(vma->vm_file);
+	struct ntfs_inode *ni = ntfs_i(inode);
+	loff_t to;
+
+	if (sizeof(void *) >= sizeof(loff_t))
+		to = (loff_t)vma->vm_private_data;
+	else
+		to = ((struct vma_cookie *)vma->vm_private_data)->to;
+
+	// Update valid size
+	ni->i_valid = max_t(loff_t, ni->i_valid,
+			    min_t(loff_t, i_size_read(inode), to));
+
+	if (sizeof(void *) < sizeof(loff_t)) {
+		ntfs_free(vma->vm_private_data);
+		vma->vm_private_data = NULL; // is it necessary?
+	}
+}
+
+/*
+ * vm_operations_struct::fault
+ */
+static vm_fault_t ntfs_filemap_fault(struct vm_fault *vmf)
+{
+	struct vm_area_struct *vma = vmf->vma;
+	vm_fault_t ret;
+	loff_t to, *vmc_to;
+
+	ret = filemap_fault(vmf);
+
+	if (!(ret & VM_FAULT_LOCKED))
+		return ret;
+
+	/* Update maximum mapped range */
+	to = (loff_t)(vmf->pgoff + 1) << PAGE_SHIFT;
+
+	if (sizeof(void *) >= sizeof(loff_t))
+		vmc_to = (loff_t *)&vma->vm_private_data;
+	else
+		vmc_to = &((struct vma_cookie *)vma->vm_private_data)->to;
+
+	if (*vmc_to < to)
+		*vmc_to = to;
+
+	return ret;
+}
+
+static const struct vm_operations_struct vma_cookie = {
+	.close = ntfs_filemap_close,
+	.fault = ntfs_filemap_fault,
+	.map_pages = filemap_map_pages,
+	.page_mkwrite = filemap_page_mkwrite,
+};
+
+/*
+ * file_operations::mmap
+ */
+static int ntfs_file_mmap(struct file *file, struct vm_area_struct *vma)
+{
+	struct address_space *mapping = file->f_mapping;
+	struct inode *inode = mapping->host;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	u64 to, from = ((u64)vma->vm_pgoff << PAGE_SHIFT);
+	bool rw = vma->vm_flags & VM_WRITE;
+	int err;
+
+	if (is_encrypted(ni)) {
+		err = -EOPNOTSUPP;
+		goto out;
+	}
+
+	if (!rw) {
+		err = generic_file_mmap(file, vma);
+		goto out;
+	}
+
+	if (is_compressed(ni)) {
+		err = -EOPNOTSUPP;
+		goto out;
+	}
+
+	// map for write
+	inode_lock(inode);
+
+	to = from + vma->vm_end - vma->vm_start;
+
+	if (to > inode->i_size)
+		to = inode->i_size;
+
+	if (is_sparsed(ni)) {
+		/* allocate clusters for rw map */
+		struct ntfs_sb_info *sbi = inode->i_sb->s_fs_info;
+		CLST vcn, lcn, len;
+		CLST end = bytes_to_cluster(sbi, to);
+		bool new;
+
+		for (vcn = from >> sbi->cluster_bits; vcn < end; vcn += len) {
+			err = attr_data_get_block(ni, vcn, 1, &lcn, &len, &new);
+			if (err) {
+				inode_unlock(inode);
+				goto out;
+			}
+			if (!new)
+				continue;
+			ntfs_sparse_cluster(inode, NULL,
+					    (u64)vcn << sbi->cluster_bits,
+					    sbi->cluster_size);
+		}
+	}
+
+	err = ni->i_valid < to ?
+		      ntfs_extend_initialized_size(file, ni, ni->i_valid, to) :
+		      0;
+
+	inode_unlock(inode);
+	if (err)
+		goto out;
+
+	if (vma->vm_private_data) {
+		ntfs_inode_warn(
+			inode,
+			"mmap: failed to keep private vma_data, possible garbidge data");
+		err = generic_file_mmap(file, vma);
+		goto out;
+	}
+
+	if (!mapping->a_ops->readpage) {
+		err = -ENOEXEC;
+		goto out;
+	}
+
+	/*
+	 * Allocate and init small struct to keep track the mapping operations
+	 * It is useful when mmap(size) + truncate(size/2) + unmap(). see
+	 * xfstests/generic/039
+	 *
+	 * TODO: if sizeof(void*) == 8 bytes we can store the only field 'to'
+	 * of vma_cookie in vma->vm_private_data
+	 */
+	if (sizeof(void *) >= sizeof(loff_t))
+		vma->vm_private_data = (void *)to;
+	else {
+		struct vma_cookie *vmc =
+			ntfs_alloc(sizeof(struct vma_cookie), 0);
+
+		if (unlikely(!vmc)) {
+			err = -ENOMEM;
+			goto out;
+		}
+		vmc->to = to;
+		vma->vm_private_data = vmc;
+	}
+
+	file_accessed(file);
+	vma->vm_ops = &vma_cookie;
+
+out:
+	return err;
+}
+
+/*
+ * file_operations::fsync
+ */
+int ntfs_file_fsync(struct file *filp, loff_t start, loff_t end, int datasync)
+{
+	return generic_file_fsync(filp, start, end, datasync);
+}
+
+static int ntfs_extend_ex(struct inode *inode, loff_t pos, size_t count,
+			  struct file *file)
+{
+	struct ntfs_inode *ni = ntfs_i(inode);
+	struct address_space *mapping = inode->i_mapping;
+	loff_t end = pos + count;
+	int err;
+	bool extend_init = file && pos > ni->i_valid;
+
+	if (end <= inode->i_size && !extend_init)
+		return 0;
+
+	/*mark rw ntfs as dirty. it will be cleared at umount*/
+	ntfs_set_state(ni->mi.sbi, NTFS_DIRTY_DIRTY);
+
+	if (end > inode->i_size) {
+		err = ntfs_set_size(inode, end);
+		if (err)
+			goto out;
+		inode->i_size = end;
+	}
+
+	if (extend_init && !is_compressed(ni)) {
+		err = ntfs_extend_initialized_size(file, ni, ni->i_valid, pos);
+		if (err)
+			goto out;
+	} else {
+		err = 0;
+	}
+
+	inode->i_ctime = inode->i_mtime = current_time(inode);
+	mark_inode_dirty(inode);
+
+	if (IS_SYNC(inode)) {
+		int err2;
+
+		err = filemap_fdatawrite_range(mapping, pos, end - 1);
+		err2 = sync_mapping_buffers(mapping);
+		if (!err)
+			err = err2;
+		err2 = write_inode_now(inode, 1);
+		if (!err)
+			err = err2;
+		if (!err)
+			err = filemap_fdatawait_range(mapping, pos, end - 1);
+	}
+
+out:
+	return err;
+}
+
+/*
+ * Preallocate space for a file. This implements ntfs's fallocate file
+ * operation, which gets called from sys_fallocate system call. User
+ * space requests 'len' bytes at 'vbo'. If FALLOC_FL_KEEP_SIZE is set
+ * we just allocate clusters without zeroing them out. Otherwise we
+ * allocate and zero out clusters via an expanding truncate.
+ */
+static long ntfs_fallocate(struct file *file, int mode, loff_t vbo, loff_t len)
+{
+	struct inode *inode = file->f_mapping->host;
+	struct super_block *sb = inode->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	loff_t i_size;
+	loff_t end;
+	int err;
+
+	/* No support for dir */
+	if (!S_ISREG(inode->i_mode))
+		return -EOPNOTSUPP;
+
+	/* Return error if mode is not supported */
+	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
+		     FALLOC_FL_COLLAPSE_RANGE))
+		return -EOPNOTSUPP;
+
+	inode_lock(inode);
+	i_size = inode->i_size;
+
+	if (mode & FALLOC_FL_PUNCH_HOLE) {
+		if (!(mode & FALLOC_FL_KEEP_SIZE)) {
+			err = -EINVAL;
+			goto out;
+		}
+		/*TODO: add support*/
+		err = -EOPNOTSUPP;
+		goto out;
+	}
+
+	if (mode & FALLOC_FL_COLLAPSE_RANGE) {
+		if (mode & ~FALLOC_FL_COLLAPSE_RANGE) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		/*TODO: add support*/
+		err = -EOPNOTSUPP;
+		goto out;
+	}
+
+	end = vbo + len;
+
+	ntfs_set_state(sbi, NTFS_DIRTY_DIRTY);
+
+	/*
+	 * normal file: allocate clusters, do not change 'valid' size
+	 */
+	err = ntfs_set_size(inode, max(end, i_size));
+	if (err)
+		goto out;
+
+	if (is_sparsed(ni) || is_compressed(ni)) {
+		CLST vcn = vbo >> sbi->cluster_bits;
+		CLST cend = bytes_to_cluster(sbi, end);
+		CLST lcn, clen;
+		bool new;
+
+		/*
+		 * allocate but not zero new clusters
+		 * this breaks security (one can read unused on-disk areas)
+		 * zeroing these clusters may be too long
+		 * may be we should check here for root rights?
+		 */
+		for (; vcn < cend; vcn += clen) {
+			err = attr_data_get_block(ni, vcn, cend - vcn, &lcn,
+						  &clen, &new);
+			if (err)
+				goto out;
+		}
+	}
+
+	if (mode & FALLOC_FL_KEEP_SIZE) {
+		ni_lock(ni);
+		/*true - keep preallocated*/
+		err = attr_set_size(ni, ATTR_DATA, NULL, 0, &ni->file.run,
+				    i_size, &ni->i_valid, true, NULL);
+		ni_unlock(ni);
+		if (err)
+			goto out;
+	}
+
+	inode->i_ctime = inode->i_mtime = current_time(inode);
+	mark_inode_dirty(inode);
+
+out:
+	if (err == -EFBIG)
+		err = -ENOSPC;
+
+	inode_unlock(inode);
+	return err;
+}
+
+void ntfs_truncate_blocks(struct inode *inode, loff_t new_size)
+{
+	struct super_block *sb = inode->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	int err, dirty = 0;
+	u32 vcn;
+	u64 new_valid;
+
+	if (!S_ISREG(inode->i_mode))
+		return;
+
+	vcn = bytes_to_cluster(sbi, new_size);
+	new_valid = ntfs_up_block(sb, min(ni->i_valid, new_size));
+
+	ni_lock(ni);
+
+	truncate_setsize(inode, new_size);
+
+	if (new_valid < ni->i_valid)
+		ni->i_valid = new_valid;
+
+	ni_unlock(ni);
+
+	ni->std_fa |= FILE_ATTRIBUTE_ARCHIVE;
+	inode->i_ctime = inode->i_mtime = current_time(inode);
+	if (!IS_DIRSYNC(inode)) {
+		dirty = 1;
+	} else {
+		err = ntfs_sync_inode(inode);
+		if (err)
+			return;
+	}
+
+	inode->i_blocks = vcn;
+
+	if (dirty)
+		mark_inode_dirty(inode);
+
+	/*ntfs_flush_inodes(inode->i_sb, inode, NULL);*/
+}
+
+/*
+ * inode_operations::setattr
+ */
+int ntfs_setattr(struct dentry *dentry, struct iattr *attr)
+{
+	struct super_block *sb = dentry->d_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct inode *inode = d_inode(dentry);
+	struct ntfs_inode *ni = ntfs_i(inode);
+	u32 ia_valid = attr->ia_valid;
+	umode_t mode = inode->i_mode;
+	int err;
+
+	if (sbi->options.no_acs_rules) {
+		/* "no access rules" - force any changes of time etc. */
+		attr->ia_valid |= ATTR_FORCE;
+		/* and disable for editing some attributes */
+		attr->ia_valid &= ~(ATTR_UID | ATTR_GID | ATTR_MODE);
+		ia_valid = attr->ia_valid;
+	}
+
+	err = setattr_prepare(dentry, attr);
+	if (err)
+		goto out;
+
+	if (ia_valid & ATTR_SIZE) {
+		loff_t oldsize = inode->i_size;
+
+		inode_dio_wait(inode);
+
+		if (attr->ia_size < oldsize) {
+			if (is_compressed(ni)) {
+				if (ni->i_valid > attr->ia_size)
+					ni->i_valid = attr->ia_size;
+			} else {
+				err = block_truncate_page(inode->i_mapping,
+							  attr->ia_size,
+							  ntfs_get_block);
+				if (err)
+					goto out;
+			}
+			ntfs_truncate_blocks(inode, attr->ia_size);
+		} else if (attr->ia_size > oldsize) {
+			err = ntfs_extend_ex(inode, attr->ia_size, 0, NULL);
+			if (err)
+				goto out;
+		}
+
+		ni->ni_flags |= NI_FLAG_UPDATE_PARENT;
+	}
+
+	setattr_copy(inode, attr);
+
+	if (mode != inode->i_mode) {
+		err = ntfs_acl_chmod(inode);
+		if (err)
+			goto out;
+
+		/* linux 'w' -> windows 'ro' */
+		if (0222 & inode->i_mode)
+			ni->std_fa &= ~FILE_ATTRIBUTE_READONLY;
+		else
+			ni->std_fa |= FILE_ATTRIBUTE_READONLY;
+	}
+
+	mark_inode_dirty(inode);
+out:
+	return err;
+}
+
+static ssize_t ntfs_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
+{
+	ssize_t err;
+	size_t count = iov_iter_count(iter);
+	struct file *file = iocb->ki_filp;
+	struct inode *inode = file->f_mapping->host;
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	if (is_encrypted(ni)) {
+		ntfs_inode_warn(inode, "encrypted i/o not supported");
+		return -EOPNOTSUPP;
+	}
+
+	if (is_compressed(ni) && (iocb->ki_flags & IOCB_DIRECT)) {
+		ntfs_inode_warn(inode, "direct i/o + compressed not supported");
+		return -EOPNOTSUPP;
+	}
+
+	if (ni->ni_flags & NI_FLAG_COMPRESSED_MASK) {
+		ntfs_inode_warn(
+			inode,
+			"read external compressed file not supported (temporary)");
+		return -EOPNOTSUPP;
+	}
+
+	if (is_dedup(ni)) {
+		ntfs_inode_warn(inode, "read deduplicated not supported");
+		return -EOPNOTSUPP;
+	}
+
+	err = count ? generic_file_read_iter(iocb, iter) : 0;
+
+	return err;
+}
+
+/* returns array of locked pages */
+static int ntfs_get_frame_pages(struct address_space *mapping, pgoff_t index,
+				struct page **pages, u32 pages_per_frame,
+				bool *frame_uptodate)
+{
+	gfp_t gfp_mask = mapping_gfp_mask(mapping);
+	u32 npages;
+
+	*frame_uptodate = true;
+
+	for (npages = 0; npages < pages_per_frame; npages++, index++) {
+		struct page *page;
+
+		page = find_or_create_page(mapping, index, gfp_mask);
+		if (!page) {
+			while (npages--) {
+				page = pages[npages];
+				unlock_page(page);
+				put_page(page);
+			}
+
+			return -ENOMEM;
+		}
+
+		if (!PageUptodate(page))
+			*frame_uptodate = false;
+
+		pages[npages] = page;
+	}
+
+	return 0;
+}
+
+/*helper for ntfs_file_write_iter (compressed files)*/
+static noinline ssize_t ntfs_compress_write(struct kiocb *iocb,
+					    struct iov_iter *from)
+{
+	int err;
+	struct file *file = iocb->ki_filp;
+	size_t count = iov_iter_count(from);
+	loff_t pos = iocb->ki_pos;
+	struct inode *inode = file_inode(file);
+	loff_t i_size = inode->i_size;
+	struct address_space *mapping = inode->i_mapping;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	u64 valid = ni->i_valid;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	struct page *page, **pages = NULL;
+	size_t written = 0;
+	u8 frame_bits = NTFS_LZNT_CUNIT + sbi->cluster_bits;
+	u32 frame_size = 1u << frame_bits;
+	u32 pages_per_frame = frame_size >> PAGE_SHIFT;
+	u32 ip, off;
+	CLST frame;
+	u64 frame_vbo;
+	pgoff_t index;
+	bool frame_uptodate;
+
+	if (frame_size < PAGE_SIZE) {
+		/*
+		 * frame_size == 8K if cluster 512
+		 * frame_size == 64K if cluster 4096
+		 */
+		ntfs_inode_warn(inode, "page size is bigger than frame size");
+		return -EOPNOTSUPP;
+	}
+
+	pages = ntfs_alloc(pages_per_frame * sizeof(struct page *), 0);
+	if (!pages)
+		return -ENOMEM;
+
+	current->backing_dev_info = inode_to_bdi(inode);
+	err = file_remove_privs(file);
+	if (err)
+		goto out;
+
+	err = file_update_time(file);
+	if (err)
+		goto out;
+
+	/* zero range [valid : pos) */
+	while (valid < pos) {
+		CLST lcn, clen;
+
+		frame = valid >> frame_bits;
+		frame_vbo = valid & ~(frame_size - 1);
+		off = valid & (frame_size - 1);
+
+		err = attr_data_get_block(ni, frame << NTFS_LZNT_CUNIT, 0, &lcn,
+					  &clen, NULL);
+		if (err)
+			goto out;
+
+		if (lcn == SPARSE_LCN) {
+			ni->i_valid = valid =
+				frame_vbo + ((u64)clen << sbi->cluster_bits);
+			continue;
+		}
+
+		/* Load full frame */
+		err = ntfs_get_frame_pages(mapping, frame_vbo >> PAGE_SHIFT,
+					   pages, pages_per_frame,
+					   &frame_uptodate);
+		if (err)
+			goto out;
+
+		if (!frame_uptodate && off) {
+			err = ni_read_frame(ni, frame_vbo, pages,
+					    pages_per_frame);
+			if (err) {
+				for (ip = 0; ip < pages_per_frame; ip++) {
+					page = pages[ip];
+					unlock_page(page);
+					put_page(page);
+				}
+				goto out;
+			}
+		}
+
+		ip = off >> PAGE_SHIFT;
+		off = offset_in_page(valid);
+		for (; ip < pages_per_frame; ip++, off = 0) {
+			page = pages[ip];
+			zero_user_segment(page, off, PAGE_SIZE);
+			flush_dcache_page(page);
+			SetPageUptodate(page);
+		}
+
+		ni_lock(ni);
+		err = ni_write_frame(ni, pages, pages_per_frame);
+		ni_unlock(ni);
+
+		for (ip = 0; ip < pages_per_frame; ip++) {
+			page = pages[ip];
+			SetPageUptodate(page);
+			unlock_page(page);
+			put_page(page);
+		}
+
+		if (err)
+			goto out;
+
+		ni->i_valid = valid = frame_vbo + frame_size;
+	}
+
+	/* copy user data [pos : pos + count) */
+	while (count) {
+		size_t copied, bytes;
+
+		off = pos & (frame_size - 1);
+		bytes = frame_size - off;
+		if (bytes > count)
+			bytes = count;
+
+		frame = pos >> frame_bits;
+		frame_vbo = pos & ~(frame_size - 1);
+		index = frame_vbo >> PAGE_SHIFT;
+
+		if (unlikely(iov_iter_fault_in_readable(from, bytes))) {
+			err = -EFAULT;
+			goto out;
+		}
+
+		/* Load full frame */
+		err = ntfs_get_frame_pages(mapping, index, pages,
+					   pages_per_frame, &frame_uptodate);
+		if (err)
+			goto out;
+
+		if (!frame_uptodate) {
+			loff_t to = pos + bytes;
+
+			if (off || (to < i_size && (to & (frame_size - 1)))) {
+				err = ni_read_frame(ni, frame_vbo, pages,
+						    pages_per_frame);
+				if (err) {
+					for (ip = 0; ip < pages_per_frame;
+					     ip++) {
+						page = pages[ip];
+						unlock_page(page);
+						put_page(page);
+					}
+					goto out;
+				}
+			}
+		}
+
+		WARN_ON(!bytes);
+		copied = 0;
+		ip = off >> PAGE_SHIFT;
+		off = offset_in_page(pos);
+
+		/* copy user data to pages */
+		for (;;) {
+			size_t cp, tail = PAGE_SIZE - off;
+
+			page = pages[ip];
+			cp = iov_iter_copy_from_user_atomic(page, from, off,
+							    min(tail, bytes));
+			flush_dcache_page(page);
+			iov_iter_advance(from, cp);
+			copied += cp;
+			bytes -= cp;
+			if (!bytes || !cp)
+				break;
+
+			if (cp < tail) {
+				off += cp;
+			} else {
+				ip++;
+				off = 0;
+			}
+		}
+
+		ni_lock(ni);
+		err = ni_write_frame(ni, pages, pages_per_frame);
+		ni_unlock(ni);
+
+		for (ip = 0; ip < pages_per_frame; ip++) {
+			page = pages[ip];
+			ClearPageDirty(page);
+			SetPageUptodate(page);
+			unlock_page(page);
+			put_page(page);
+		}
+
+		if (err)
+			goto out;
+
+		/*
+		 * We can loop for a long time in here. Be nice and allow
+		 * us to schedule out to avoid softlocking if preempt
+		 * is disabled.
+		 */
+		cond_resched();
+
+		pos += copied;
+		written += copied;
+
+		count = iov_iter_count(from);
+	}
+
+out:
+	ntfs_free(pages);
+
+	current->backing_dev_info = NULL;
+
+	if (err < 0)
+		return err;
+
+	iocb->ki_pos += written;
+	if (iocb->ki_pos > ni->i_valid)
+		ni->i_valid = iocb->ki_pos;
+
+	return written;
+}
+
+/*
+ * file_operations::write_iter
+ */
+static ssize_t ntfs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
+{
+	struct file *file = iocb->ki_filp;
+	struct address_space *mapping = file->f_mapping;
+	struct inode *inode = mapping->host;
+	ssize_t ret;
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	if (is_encrypted(ni)) {
+		ntfs_inode_warn(inode, "encrypted i/o not supported");
+		return -EOPNOTSUPP;
+	}
+
+	if (is_compressed(ni) && (iocb->ki_flags & IOCB_DIRECT)) {
+		ntfs_inode_warn(inode, "direct i/o + compressed not supported");
+		return -EOPNOTSUPP;
+	}
+
+	if (ni->ni_flags & NI_FLAG_COMPRESSED_MASK) {
+		ntfs_inode_warn(
+			inode,
+			"write into external compressed file not supported (temporary)");
+		return -EOPNOTSUPP;
+	}
+
+	if (is_dedup(ni)) {
+		ntfs_inode_warn(inode, "write into deduplicated not supported");
+		return -EOPNOTSUPP;
+	}
+
+	if (!inode_trylock(inode)) {
+		if (iocb->ki_flags & IOCB_NOWAIT)
+			return -EAGAIN;
+		inode_lock(inode);
+	}
+
+	ret = generic_write_checks(iocb, from);
+	if (ret <= 0)
+		goto out;
+
+	ret = ntfs_extend_ex(inode, iocb->ki_pos, ret, file);
+	if (ret)
+		goto out;
+
+	ret = is_compressed(ni) ? ntfs_compress_write(iocb, from) :
+				  __generic_file_write_iter(iocb, from);
+
+out:
+	inode_unlock(inode);
+
+	if (ret > 0)
+		ret = generic_write_sync(iocb, ret);
+
+	return ret;
+}
+
+/*
+ * file_operations::open
+ */
+int ntfs_file_open(struct inode *inode, struct file *file)
+{
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	if (unlikely((is_compressed(ni) || is_encrypted(ni)) &&
+		     (file->f_flags & O_DIRECT))) {
+		return -EOPNOTSUPP;
+	}
+
+	return generic_file_open(inode, file);
+}
+
+/*
+ * file_operations::release
+ */
+static int ntfs_file_release(struct inode *inode, struct file *file)
+{
+	struct ntfs_inode *ni = ntfs_i(inode);
+	struct ntfs_sb_info *sbi;
+	int err;
+
+	/* if we are the last writer on the inode, drop the block reservation */
+	if (!(file->f_mode & FMODE_WRITE) ||
+	    atomic_read(&inode->i_writecount) != 1)
+		return 0;
+
+	sbi = ni->mi.sbi;
+
+	ni_lock(ni);
+	down_write(&ni->file.run_lock);
+
+	err = attr_set_size(ni, ATTR_DATA, NULL, 0, &ni->file.run,
+			    inode->i_size, &ni->i_valid, false, NULL);
+
+	up_write(&ni->file.run_lock);
+	ni_unlock(ni);
+
+	return err;
+}
+
+/* file_operations::fiemap */
+int ntfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
+		__u64 start, __u64 len)
+{
+	int err;
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR)
+		return -EOPNOTSUPP;
+
+	ni_lock(ni);
+
+	err = ni_fiemap(ni, fieinfo, start, len);
+
+	ni_unlock(ni);
+
+	return err;
+}
+
+const struct inode_operations ntfs_file_inode_operations = {
+	.getattr = ntfs_getattr,
+	.setattr = ntfs_setattr,
+	.listxattr = ntfs_listxattr,
+	.permission = ntfs_permission,
+	.get_acl = ntfs_get_acl,
+	.set_acl = ntfs_set_acl,
+	.fiemap = ntfs_fiemap,
+};
+
+const struct file_operations ntfs_file_operations = {
+	.llseek = generic_file_llseek,
+	.read_iter = ntfs_file_read_iter,
+	.write_iter = ntfs_file_write_iter,
+	.unlocked_ioctl = ntfs_ioctl,
+#ifdef CONFIG_COMPAT
+	.compat_ioctl = ntfs_compat_ioctl,
+#endif
+	.splice_read = generic_file_splice_read,
+	.mmap = ntfs_file_mmap,
+	.open = ntfs_file_open,
+	.fsync = ntfs_file_fsync,
+	.splice_write = iter_file_splice_write,
+	.fallocate = ntfs_fallocate,
+	.release = ntfs_file_release,
+};
diff --git a/fs/ntfs3/frecord.c b/fs/ntfs3/frecord.c
new file mode 100644
index 000000000000..f8f762ccdbcc
--- /dev/null
+++ b/fs/ntfs3/frecord.c
@@ -0,0 +1,2696 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fiemap.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+static inline void get_mi_ref(const struct mft_inode *mi, struct MFT_REF *ref)
+{
+#ifdef NTFS3_64BIT_CLUSTER
+	ref->low = cpu_to_le32(mi->rno);
+	ref->high = cpu_to_le16(mi->rno >> 32);
+#else
+	ref->low = cpu_to_le32(mi->rno);
+	ref->high = 0;
+#endif
+	ref->seq = mi->mrec->seq;
+}
+
+static struct mft_inode *ni_ins_mi(struct ntfs_inode *ni, struct rb_root *tree,
+				   CLST ino, struct rb_node *ins)
+{
+	struct rb_node **p = &tree->rb_node;
+	struct rb_node *pr = NULL;
+
+	while (*p) {
+		struct mft_inode *mi;
+
+		pr = *p;
+		mi = rb_entry(pr, struct mft_inode, node);
+		if (mi->rno > ino)
+			p = &pr->rb_left;
+		else if (mi->rno < ino)
+			p = &pr->rb_right;
+		else
+			return mi;
+	}
+
+	if (!ins)
+		return NULL;
+
+	rb_link_node(ins, pr, p);
+	rb_insert_color(ins, tree);
+	return rb_entry(ins, struct mft_inode, node);
+}
+
+/*
+ * ni_find_mi
+ *
+ * finds mft_inode by record number
+ */
+static struct mft_inode *ni_find_mi(struct ntfs_inode *ni, CLST rno)
+{
+	return ni_ins_mi(ni, &ni->mi_tree, rno, NULL);
+}
+
+/*
+ * ni_add_mi
+ *
+ * adds new mft_inode into ntfs_inode
+ */
+static void ni_add_mi(struct ntfs_inode *ni, struct mft_inode *mi)
+{
+	ni_ins_mi(ni, &ni->mi_tree, mi->rno, &mi->node);
+}
+
+/*
+ * ni_remove_mi
+ *
+ * removes mft_inode from ntfs_inode
+ */
+void ni_remove_mi(struct ntfs_inode *ni, struct mft_inode *mi)
+{
+	rb_erase(&mi->node, &ni->mi_tree);
+}
+
+/*
+ * ni_std
+ *
+ * returns pointer into std_info from primary record
+ */
+struct ATTR_STD_INFO *ni_std(struct ntfs_inode *ni)
+{
+	const struct ATTRIB *attr;
+
+	attr = mi_find_attr(&ni->mi, NULL, ATTR_STD, NULL, 0, NULL);
+	return attr ? resident_data_ex(attr, sizeof(struct ATTR_STD_INFO)) :
+		      NULL;
+}
+
+/*
+ * ni_std5
+ *
+ * returns pointer into std_info from primary record
+ */
+struct ATTR_STD_INFO5 *ni_std5(struct ntfs_inode *ni)
+{
+	const struct ATTRIB *attr;
+
+	attr = mi_find_attr(&ni->mi, NULL, ATTR_STD, NULL, 0, NULL);
+
+	return attr ? resident_data_ex(attr, sizeof(struct ATTR_STD_INFO5)) :
+		      NULL;
+}
+
+/*
+ * ni_clear
+ *
+ * clears resources allocated by ntfs_inode
+ */
+void ni_clear(struct ntfs_inode *ni)
+{
+	struct rb_node *node;
+
+	if (!ni->vfs_inode.i_nlink && is_rec_inuse(ni->mi.mrec))
+		ni_delete_all(ni);
+
+	al_destroy(ni);
+
+	for (node = rb_first(&ni->mi_tree); node;) {
+		struct rb_node *next = rb_next(node);
+		struct mft_inode *mi = rb_entry(node, struct mft_inode, node);
+
+		rb_erase(node, &ni->mi_tree);
+		mi_put(mi);
+		node = next;
+	}
+
+	/* bad inode always has mode == S_IFREG */
+	if (ni->ni_flags & NI_FLAG_DIR)
+		indx_clear(&ni->dir);
+	else
+		run_close(&ni->file.run);
+
+	mi_clear(&ni->mi);
+}
+
+/*
+ * ni_load_mi_ex
+ *
+ * finds mft_inode by record number.
+ */
+int ni_load_mi_ex(struct ntfs_inode *ni, CLST rno, struct mft_inode **mi)
+{
+	int err;
+	struct mft_inode *r;
+
+	r = ni_find_mi(ni, rno);
+	if (r)
+		goto out;
+
+	err = mi_get(ni->mi.sbi, rno, &r);
+	if (err)
+		return err;
+
+	ni_add_mi(ni, r);
+
+out:
+	if (mi)
+		*mi = r;
+	return 0;
+}
+
+/*
+ * ni_load_mi
+ *
+ * load mft_inode corresponded list_entry
+ */
+int ni_load_mi(struct ntfs_inode *ni, struct ATTR_LIST_ENTRY *le,
+	       struct mft_inode **mi)
+{
+	CLST rno;
+
+	if (!le) {
+		*mi = &ni->mi;
+		return 0;
+	}
+
+	rno = ino_get(&le->ref);
+	if (rno == ni->mi.rno) {
+		*mi = &ni->mi;
+		return 0;
+	}
+	return ni_load_mi_ex(ni, rno, mi);
+}
+
+/*
+ * ni_find_attr
+ *
+ * returns attribute and record this attribute belongs to
+ */
+struct ATTRIB *ni_find_attr(struct ntfs_inode *ni, struct ATTRIB *attr,
+			    struct ATTR_LIST_ENTRY **le_o, enum ATTR_TYPE type,
+			    const __le16 *name, u8 name_len, const CLST *vcn,
+			    struct mft_inode **mi)
+{
+	struct ATTR_LIST_ENTRY *le;
+	struct mft_inode *m;
+
+	if (!ni->attr_list.size ||
+	    (!name_len && (type == ATTR_LIST || type == ATTR_STD))) {
+		if (le_o)
+			*le_o = NULL;
+		if (mi)
+			*mi = &ni->mi;
+
+		/* Look for required attribute in primary record */
+		return mi_find_attr(&ni->mi, attr, type, name, name_len, NULL);
+	}
+
+	/* first look for list entry of required type */
+	le = al_find_ex(ni, le_o ? *le_o : NULL, type, name, name_len, vcn);
+	if (!le)
+		return NULL;
+
+	if (le_o)
+		*le_o = le;
+
+	/* Load record that contains this attribute */
+	if (ni_load_mi(ni, le, &m))
+		return NULL;
+
+	/* Look for required attribute */
+	attr = mi_find_attr(m, NULL, type, name, name_len, &le->id);
+
+	if (!attr)
+		goto out;
+
+	if (!attr->non_res) {
+		if (vcn && *vcn)
+			goto out;
+	} else if (!vcn) {
+		if (attr->nres.svcn)
+			goto out;
+	} else if (le64_to_cpu(attr->nres.svcn) > *vcn ||
+		   *vcn > le64_to_cpu(attr->nres.evcn)) {
+		goto out;
+	}
+
+	if (mi)
+		*mi = m;
+	return attr;
+
+out:
+	ntfs_set_state(ni->mi.sbi, NTFS_DIRTY_ERROR);
+	return NULL;
+}
+
+/*
+ * ni_enum_attr_ex
+ *
+ * enumerates attributes in ntfs_inode
+ */
+struct ATTRIB *ni_enum_attr_ex(struct ntfs_inode *ni, struct ATTRIB *attr,
+			       struct ATTR_LIST_ENTRY **le)
+{
+	struct mft_inode *mi;
+	struct ATTR_LIST_ENTRY *le2;
+
+	/* Do we have an attribute list? */
+	if (!ni->attr_list.size) {
+		*le = NULL;
+		/* Enum attributes in primary record */
+		return mi_enum_attr(&ni->mi, attr);
+	}
+
+	/* get next list entry */
+	le2 = *le = al_enumerate(ni, attr ? *le : NULL);
+	if (!le2)
+		return NULL;
+
+	/* Load record that contains the required attribute */
+	if (ni_load_mi(ni, le2, &mi))
+		return NULL;
+
+	/* Find attribute in loaded record */
+	attr = rec_find_attr_le(mi, le2);
+	return attr;
+}
+
+/*
+ * ni_load_attr
+ *
+ * loads attribute that contains given vcn
+ */
+struct ATTRIB *ni_load_attr(struct ntfs_inode *ni, enum ATTR_TYPE type,
+			    const __le16 *name, u8 name_len, CLST vcn,
+			    struct mft_inode **pmi)
+{
+	struct ATTR_LIST_ENTRY *le;
+	struct ATTRIB *attr;
+	struct mft_inode *mi;
+	struct ATTR_LIST_ENTRY *next;
+
+	if (!ni->attr_list.size) {
+		if (pmi)
+			*pmi = &ni->mi;
+		return mi_find_attr(&ni->mi, NULL, type, name, name_len, NULL);
+	}
+
+	le = al_find_ex(ni, NULL, type, name, name_len, NULL);
+	if (!le)
+		return NULL;
+
+	/*
+	 * Unfortunately ATTR_LIST_ENTRY contains only start vcn
+	 * So to find the ATTRIB segment that contains 'vcn' we should
+	 * enumerate some entries
+	 */
+	if (vcn) {
+		for (;; le = next) {
+			next = al_find_ex(ni, le, type, name, name_len, NULL);
+			if (!next || le64_to_cpu(next->vcn) > vcn)
+				break;
+		}
+	}
+
+	if (ni_load_mi(ni, le, &mi))
+		return NULL;
+
+	if (pmi)
+		*pmi = mi;
+
+	attr = mi_find_attr(mi, NULL, type, name, name_len, &le->id);
+	if (!attr)
+		return NULL;
+
+	if (!attr->non_res)
+		return attr;
+
+	if (le64_to_cpu(attr->nres.svcn) <= vcn &&
+	    vcn <= le64_to_cpu(attr->nres.evcn))
+		return attr;
+
+	return NULL;
+}
+
+/*
+ * ni_load_all_mi
+ *
+ * loads all subrecords
+ */
+int ni_load_all_mi(struct ntfs_inode *ni)
+{
+	int err;
+	struct ATTR_LIST_ENTRY *le;
+
+	if (!ni->attr_list.size)
+		return 0;
+
+	le = NULL;
+
+	while ((le = al_enumerate(ni, le))) {
+		CLST rno = ino_get(&le->ref);
+
+		if (rno == ni->mi.rno)
+			continue;
+
+		err = ni_load_mi_ex(ni, rno, NULL);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+/*
+ * ni_add_subrecord
+ *
+ * allocate + format + attach a new subrecord
+ */
+bool ni_add_subrecord(struct ntfs_inode *ni, CLST rno, struct mft_inode **mi)
+{
+	struct mft_inode *m;
+
+	m = ntfs_alloc(sizeof(struct mft_inode), 1);
+	if (!m)
+		return false;
+
+	if (mi_format_new(m, ni->mi.sbi, rno, 0, ni->mi.rno == MFT_REC_MFT)) {
+		mi_put(m);
+		return false;
+	}
+
+	get_mi_ref(&ni->mi, &m->mrec->parent_ref);
+
+	ni_add_mi(ni, m);
+	*mi = m;
+	return true;
+}
+
+/*
+ * ni_remove_attr
+ *
+ * removes all attributes for the given type/name/id
+ */
+int ni_remove_attr(struct ntfs_inode *ni, enum ATTR_TYPE type,
+		   const __le16 *name, size_t name_len, bool base_only,
+		   const __le16 *id)
+{
+	int err;
+	struct ATTRIB *attr;
+	struct ATTR_LIST_ENTRY *le;
+	struct mft_inode *mi;
+	u32 type_in;
+	int diff;
+
+	if (base_only || type == ATTR_LIST || !ni->attr_list.size) {
+		attr = mi_find_attr(&ni->mi, NULL, type, name, name_len, id);
+		if (!attr)
+			return -ENOENT;
+
+		mi_remove_attr(&ni->mi, attr);
+		return 0;
+	}
+
+	type_in = le32_to_cpu(type);
+	le = NULL;
+
+	for (;;) {
+		le = al_enumerate(ni, le);
+		if (!le)
+			return 0;
+
+next_le2:
+		diff = le32_to_cpu(le->type) - type_in;
+		if (diff < 0)
+			continue;
+
+		if (diff > 0)
+			return 0;
+
+		if (le->name_len != name_len)
+			continue;
+
+		if (name_len &&
+		    memcmp(le_name(le), name, name_len * sizeof(short)))
+			continue;
+
+		if (id && le->id != *id)
+			continue;
+		err = ni_load_mi(ni, le, &mi);
+		if (err)
+			return err;
+
+		al_remove_le(ni, le);
+
+		attr = mi_find_attr(mi, NULL, type, name, name_len, id);
+		if (!attr)
+			return -ENOENT;
+
+		mi_remove_attr(mi, attr);
+
+		if (PtrOffset(ni->attr_list.le, le) >= ni->attr_list.size)
+			return 0;
+		goto next_le2;
+	}
+}
+
+/*
+ * ni_ins_new_attr
+ *
+ * inserts the attribute into record
+ * Returns not full constructed attribute or NULL if not possible to create
+ */
+static struct ATTRIB *ni_ins_new_attr(struct ntfs_inode *ni,
+				      struct mft_inode *mi,
+				      struct ATTR_LIST_ENTRY *le,
+				      enum ATTR_TYPE type, const __le16 *name,
+				      u8 name_len, u32 asize, u16 name_off,
+				      CLST svcn)
+{
+	int err;
+	struct ATTRIB *attr;
+	bool le_added = false;
+	struct MFT_REF ref;
+
+	get_mi_ref(mi, &ref);
+
+	if (type != ATTR_LIST && !le && ni->attr_list.size) {
+		err = al_add_le(ni, type, name, name_len, svcn, cpu_to_le16(-1),
+				&ref, &le);
+		if (err) {
+			/* no memory or no space */
+			return NULL;
+		}
+		le_added = true;
+
+		/*
+		 * al_add_le -> attr_set_size (list) -> ni_expand_list
+		 * which moves some attributes out of primary record
+		 * this means that name may point into moved memory
+		 * reinit 'name' from le
+		 */
+		name = le->name;
+	}
+
+	attr = mi_insert_attr(mi, type, name, name_len, asize, name_off);
+	if (!attr) {
+		if (le_added)
+			al_remove_le(ni, le);
+		return NULL;
+	}
+
+	if (type == ATTR_LIST) {
+		/*attr list is not in list entry array*/
+		goto out;
+	}
+
+	if (!le)
+		goto out;
+
+	/* Update ATTRIB Id and record reference */
+	le->id = attr->id;
+	ni->attr_list.dirty = true;
+	le->ref = ref;
+
+out:
+
+	return attr;
+}
+
+/*
+ * random write access to sparsed or compressed file may result to
+ * not optimized packed runs.
+ * Here it is the place to optimize it
+ */
+static int ni_repack(struct ntfs_inode *ni)
+{
+	int err = 0;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	struct mft_inode *mi, *mi_p = NULL;
+	struct ATTRIB *attr = NULL, *attr_p;
+	struct ATTR_LIST_ENTRY *le = NULL, *le_p;
+	CLST alloc = 0;
+	u8 cluster_bits = sbi->cluster_bits;
+	CLST svcn, evcn = 0, svcn_p, evcn_p, next_svcn;
+	u32 roff, rs = sbi->record_size;
+	struct runs_tree run;
+
+	run_init(&run);
+
+	while ((attr = ni_enum_attr_ex(ni, attr, &le))) {
+		if (!attr->non_res)
+			continue;
+
+		if (ni_load_mi(ni, le, &mi)) {
+			err = -EINVAL;
+			break;
+		}
+
+		svcn = le64_to_cpu(attr->nres.svcn);
+		if (svcn != le64_to_cpu(le->vcn)) {
+			err = -EINVAL;
+			break;
+		}
+
+		if (!svcn) {
+			alloc = le32_to_cpu(attr->nres.alloc_size) >>
+				cluster_bits;
+			mi_p = NULL;
+		} else if (svcn != evcn + 1) {
+			err = -EINVAL;
+			break;
+		}
+
+		evcn = le64_to_cpu(attr->nres.evcn);
+
+		if (svcn > evcn + 1) {
+			err = -EINVAL;
+			break;
+		}
+
+		if (!mi_p) {
+			/* do not try if too little free space */
+			if (le32_to_cpu(mi->mrec->used) + 8 >= rs)
+				continue;
+
+			/* do not try if last attribute segment */
+			if (evcn + 1 == alloc)
+				continue;
+			run_close(&run);
+		}
+
+		roff = le16_to_cpu(attr->nres.run_off);
+		err = run_unpack(&run, sbi, ni->mi.rno, svcn, evcn, svcn,
+				 Add2Ptr(attr, roff),
+				 le32_to_cpu(attr->size) - roff);
+		if (err < 0)
+			break;
+
+		if (!mi_p) {
+			mi_p = mi;
+			attr_p = attr;
+			svcn_p = svcn;
+			evcn_p = evcn;
+			le_p = le;
+			continue;
+		}
+
+		/*
+		 * run contains data from two records: mi_p and mi
+		 * try to pack in one
+		 */
+		err = mi_pack_runs(mi_p, attr_p, &run, evcn + 1 - svcn_p);
+		if (err)
+			break;
+
+		next_svcn = le64_to_cpu(attr_p->nres.evcn) + 1;
+
+		if (next_svcn >= evcn + 1) {
+			/* we can remove this attribute segment */
+			al_remove_le(ni, le);
+			mi_remove_attr(mi, attr);
+			le = le_p;
+			continue;
+		}
+
+		attr->nres.svcn = le->vcn = cpu_to_le64(next_svcn);
+		mi->dirty = true;
+		ni->attr_list.dirty = true;
+
+		if (evcn + 1 == alloc) {
+			err = mi_pack_runs(mi, attr, &run,
+					   evcn + 1 - next_svcn);
+			if (err)
+				break;
+			mi_p = NULL;
+		} else {
+			mi_p = mi;
+			attr_p = attr;
+			svcn_p = next_svcn;
+			evcn_p = evcn;
+			le_p = le;
+			run_truncate_head(&run, next_svcn);
+		}
+	}
+
+	if (err) {
+		ntfs_inode_warn(&ni->vfs_inode, "there is a problem");
+		ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
+
+		/* Pack loaded but not packed runs */
+		if (mi_p)
+			mi_pack_runs(mi_p, attr_p, &run, evcn_p + 1 - svcn_p);
+	}
+
+	run_close(&run);
+	return err;
+}
+
+/*
+ * ni_try_remove_attr_list
+ *
+ * Can we remove attribute list?
+ * Check the case when primary record contains enough space for all attributes
+ */
+static int ni_try_remove_attr_list(struct ntfs_inode *ni)
+{
+	int err = 0;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	struct ATTRIB *attr, *attr_list, *attr_ins;
+	struct ATTR_LIST_ENTRY *le;
+	struct mft_inode *mi;
+	u32 asize, free;
+	struct MFT_REF ref;
+	__le16 id;
+
+	if (!ni->attr_list.dirty)
+		return 0;
+
+	err = ni_repack(ni);
+	if (err)
+		return err;
+
+	attr_list = mi_find_attr(&ni->mi, NULL, ATTR_LIST, NULL, 0, NULL);
+	if (!attr_list)
+		return 0;
+
+	asize = le32_to_cpu(attr_list->size);
+
+	/* free space in primary record without attribute list */
+	free = sbi->record_size - le32_to_cpu(ni->mi.mrec->used) + asize;
+	get_mi_ref(&ni->mi, &ref);
+
+	le = NULL;
+	while ((le = al_enumerate(ni, le))) {
+		if (!memcmp(&le->ref, &ref, sizeof(ref)))
+			continue;
+
+		if (le->vcn)
+			return 0;
+
+		mi = ni_find_mi(ni, ino_get(&le->ref));
+		if (!mi)
+			return 0;
+
+		attr = mi_find_attr(mi, NULL, le->type, le_name(le),
+				    le->name_len, &le->id);
+		if (!attr)
+			return 0;
+
+		asize = le32_to_cpu(attr->size);
+		if (asize > free)
+			return 0;
+
+		free -= asize;
+	}
+
+	/* Is seems that attribute list can be removed from primary record */
+	mi_remove_attr(&ni->mi, attr_list);
+
+	/*
+	 * Repeat the cycle above and move all attributes to primary record.
+	 * It should be success!
+	 */
+	le = NULL;
+	while ((le = al_enumerate(ni, le))) {
+		if (!memcmp(&le->ref, &ref, sizeof(ref)))
+			continue;
+
+		mi = ni_find_mi(ni, ino_get(&le->ref));
+
+		attr = mi_find_attr(mi, NULL, le->type, le_name(le),
+				    le->name_len, &le->id);
+		asize = le32_to_cpu(attr->size);
+
+		/* insert into primary record */
+		attr_ins = mi_insert_attr(&ni->mi, le->type, le_name(le),
+					  le->name_len, asize,
+					  le16_to_cpu(attr->name_off));
+		id = attr_ins->id;
+
+		/* copy all except id */
+		memcpy(attr_ins, attr, asize);
+		attr_ins->id = id;
+
+		/* remove from original record */
+		mi_remove_attr(mi, attr);
+	}
+
+	run_deallocate(sbi, &ni->attr_list.run, true);
+	run_close(&ni->attr_list.run);
+	ni->attr_list.size = 0;
+	ntfs_free(ni->attr_list.le);
+	ni->attr_list.le = NULL;
+	ni->attr_list.dirty = false;
+
+	return 0;
+}
+
+/*
+ * ni_create_attr_list
+ *
+ * generates an attribute list for this primary record
+ */
+int ni_create_attr_list(struct ntfs_inode *ni)
+{
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	int err;
+	u32 lsize;
+	struct ATTRIB *attr;
+	struct ATTRIB *arr_move[7];
+	struct ATTR_LIST_ENTRY *le, *le_b[7];
+	struct MFT_REC *rec;
+	bool is_mft;
+	CLST rno = 0;
+	struct mft_inode *mi;
+	u32 free_b, nb, to_free, rs;
+	u16 sz;
+
+	is_mft = ni->mi.rno == MFT_REC_MFT;
+	rec = ni->mi.mrec;
+	rs = sbi->record_size;
+
+	/*
+	 * Skip estimating exact memory requirement
+	 * Looks like one record_size is always enough
+	 */
+	le = ntfs_alloc(al_aligned(rs), 0);
+	if (!le) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	get_mi_ref(&ni->mi, &le->ref);
+	ni->attr_list.le = le;
+
+	attr = NULL;
+	nb = 0;
+	free_b = 0;
+	attr = NULL;
+
+	for (; (attr = mi_enum_attr(&ni->mi, attr)); le = Add2Ptr(le, sz)) {
+		sz = le_size(attr->name_len);
+		le->type = attr->type;
+		le->size = cpu_to_le16(sz);
+		le->name_len = attr->name_len;
+		le->name_off = offsetof(struct ATTR_LIST_ENTRY, name);
+		le->vcn = 0;
+		if (le != ni->attr_list.le)
+			le->ref = ni->attr_list.le->ref;
+		le->id = attr->id;
+
+		if (attr->name_len)
+			memcpy(le->name, attr_name(attr),
+			       sizeof(short) * attr->name_len);
+		else if (attr->type == ATTR_STD)
+			continue;
+		else if (attr->type == ATTR_LIST)
+			continue;
+		else if (is_mft && attr->type == ATTR_DATA)
+			continue;
+
+		if (!nb || nb < ARRAY_SIZE(arr_move)) {
+			le_b[nb] = le;
+			arr_move[nb++] = attr;
+			free_b += le32_to_cpu(attr->size);
+		}
+	}
+
+	lsize = PtrOffset(ni->attr_list.le, le);
+	ni->attr_list.size = lsize;
+
+	to_free = le32_to_cpu(rec->used) + lsize + SIZEOF_RESIDENT;
+	if (to_free <= rs) {
+		to_free = 0;
+	} else {
+		to_free -= rs;
+
+		if (to_free > free_b) {
+			err = -EINVAL;
+			goto out1;
+		}
+	}
+
+	/* Allocate child mft. */
+	err = ntfs_look_free_mft(sbi, &rno, is_mft, ni, &mi);
+	if (err)
+		goto out1;
+
+	/* Call 'mi_remove_attr' in reverse order to keep pointers 'arr_move' valid */
+	while (to_free > 0) {
+		struct ATTRIB *b = arr_move[--nb];
+		u32 asize = le32_to_cpu(b->size);
+		u16 name_off = le16_to_cpu(b->name_off);
+
+		attr = mi_insert_attr(mi, b->type, Add2Ptr(b, name_off),
+				      b->name_len, asize, name_off);
+		WARN_ON(!attr);
+
+		get_mi_ref(mi, &le_b[nb]->ref);
+		le_b[nb]->id = attr->id;
+
+		/* copy all except id */
+		memcpy(attr, b, asize);
+		attr->id = le_b[nb]->id;
+
+		WARN_ON(!mi_remove_attr(&ni->mi, b));
+
+		if (to_free <= asize)
+			break;
+		to_free -= asize;
+		WARN_ON(!nb);
+	}
+
+	attr = mi_insert_attr(&ni->mi, ATTR_LIST, NULL, 0,
+			      lsize + SIZEOF_RESIDENT, SIZEOF_RESIDENT);
+	WARN_ON(!attr);
+
+	attr->non_res = 0;
+	attr->flags = 0;
+	attr->res.data_size = cpu_to_le32(lsize);
+	attr->res.data_off = SIZEOF_RESIDENT_LE;
+	attr->res.flags = 0;
+	attr->res.res = 0;
+
+	memcpy(resident_data_ex(attr, lsize), ni->attr_list.le, lsize);
+
+	ni->attr_list.dirty = false;
+
+	mark_inode_dirty(&ni->vfs_inode);
+	goto out;
+
+out1:
+	ntfs_free(ni->attr_list.le);
+	ni->attr_list.le = NULL;
+	ni->attr_list.size = 0;
+
+out:
+	return err;
+}
+
+/*
+ * ni_ins_attr_ext
+ *
+ * This method adds an external attribute to the ntfs_inode.
+ */
+static int ni_ins_attr_ext(struct ntfs_inode *ni, struct ATTR_LIST_ENTRY *le,
+			   enum ATTR_TYPE type, const __le16 *name, u8 name_len,
+			   u32 asize, CLST svcn, u16 name_off, bool force_ext,
+			   struct ATTRIB **ins_attr, struct mft_inode **ins_mi)
+{
+	struct ATTRIB *attr;
+	struct mft_inode *mi;
+	CLST rno;
+	u64 vbo;
+	struct rb_node *node;
+	int err;
+	bool is_mft, is_mft_data;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+
+	is_mft = ni->mi.rno == MFT_REC_MFT;
+	is_mft_data = is_mft && type == ATTR_DATA && !name_len;
+
+	if (asize > sbi->max_bytes_per_attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/*
+	 * standard information and attr_list cannot be made external.
+	 * The Log File cannot have any external attributes
+	 */
+	if (type == ATTR_STD || type == ATTR_LIST ||
+	    ni->mi.rno == MFT_REC_LOG) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* Create attribute list if it is not already existed */
+	if (!ni->attr_list.size) {
+		err = ni_create_attr_list(ni);
+		if (err)
+			goto out;
+	}
+
+	vbo = is_mft_data ? ((u64)svcn << sbi->cluster_bits) : 0;
+
+	if (force_ext)
+		goto insert_ext;
+
+	/* Load all subrecords into memory. */
+	err = ni_load_all_mi(ni);
+	if (err)
+		goto out;
+
+	/* Check each of loaded subrecord */
+	for (node = rb_first(&ni->mi_tree); node; node = rb_next(node)) {
+		mi = rb_entry(node, struct mft_inode, node);
+
+		if (is_mft_data &&
+		    (mi_enum_attr(mi, NULL) ||
+		     vbo <= ((u64)mi->rno << sbi->record_bits))) {
+			/* We can't accept this record 'case MFT's bootstrapping */
+			continue;
+		}
+		if (is_mft &&
+		    mi_find_attr(mi, NULL, ATTR_DATA, NULL, 0, NULL)) {
+			/*
+			 * This child record already has a ATTR_DATA.
+			 * So it can't accept any other records.
+			 */
+			continue;
+		}
+
+		if ((type != ATTR_NAME || name_len) &&
+		    mi_find_attr(mi, NULL, type, name, name_len, NULL)) {
+			/* Only indexed attributes can share same record */
+			continue;
+		}
+
+		/* Try to insert attribute into this subrecord */
+		attr = ni_ins_new_attr(ni, mi, le, type, name, name_len, asize,
+				       name_off, svcn);
+		if (!attr)
+			continue;
+
+		if (ins_attr)
+			*ins_attr = attr;
+		return 0;
+	}
+
+insert_ext:
+	/* We have to allocate a new child subrecord*/
+	err = ntfs_look_free_mft(sbi, &rno, is_mft_data, ni, &mi);
+	if (err)
+		goto out;
+
+	if (is_mft_data && vbo <= ((u64)rno << sbi->record_bits)) {
+		err = -EINVAL;
+		goto out1;
+	}
+
+	attr = ni_ins_new_attr(ni, mi, le, type, name, name_len, asize,
+			       name_off, svcn);
+	if (!attr)
+		goto out2;
+
+	if (ins_attr)
+		*ins_attr = attr;
+	if (ins_mi)
+		*ins_mi = mi;
+
+	return 0;
+
+out2:
+	ni_remove_mi(ni, mi);
+	mi_put(mi);
+	err = -EINVAL;
+
+out1:
+	ntfs_mark_rec_free(sbi, rno);
+
+out:
+	return err;
+}
+
+/*
+ * ni_insert_attr
+ *
+ * inserts an attribute into the file.
+ *
+ * If the primary record has room, it will just insert the attribute.
+ * If not, it may make the attribute external.
+ * For $MFT::Data it may make room for the attribute by
+ * making other attributes external.
+ *
+ * NOTE:
+ * The ATTR_LIST and ATTR_STD cannot be made external.
+ * This function does not fill new attribute full
+ * It only fills 'size'/'type'/'id'/'name_len' fields
+ */
+static int ni_insert_attr(struct ntfs_inode *ni, enum ATTR_TYPE type,
+			  const __le16 *name, u8 name_len, u32 asize,
+			  u16 name_off, CLST svcn, struct ATTRIB **ins_attr,
+			  struct mft_inode **ins_mi)
+{
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	int err;
+	struct ATTRIB *attr, *eattr;
+	struct MFT_REC *rec;
+	bool is_mft;
+	struct ATTR_LIST_ENTRY *le;
+	u32 list_reserve, max_free, free, used, t32;
+	__le16 id;
+	u16 t16;
+
+	is_mft = ni->mi.rno == MFT_REC_MFT;
+	rec = ni->mi.mrec;
+
+	list_reserve = SIZEOF_NONRESIDENT + 3 * (1 + 2 * sizeof(u32));
+	used = le32_to_cpu(rec->used);
+	free = sbi->record_size - used;
+
+	if (is_mft && type != ATTR_LIST) {
+		/* Reserve space for the ATTRIB List. */
+		if (free < list_reserve)
+			free = 0;
+		else
+			free -= list_reserve;
+	}
+
+	if (asize <= free) {
+		attr = ni_ins_new_attr(ni, &ni->mi, NULL, type, name, name_len,
+				       asize, name_off, svcn);
+		if (attr) {
+			if (ins_attr)
+				*ins_attr = attr;
+			if (ins_mi)
+				*ins_mi = &ni->mi;
+			err = 0;
+			goto out;
+		}
+	}
+
+	if (!is_mft || type != ATTR_DATA || svcn) {
+		/* This ATTRIB will be external. */
+		err = ni_ins_attr_ext(ni, NULL, type, name, name_len, asize,
+				      svcn, name_off, false, ins_attr, ins_mi);
+		goto out;
+	}
+
+	/*
+	 * Here we have: "is_mft && type == ATTR_DATA && !svcn
+	 *
+	 * The first chunk of the $MFT::Data ATTRIB must be the base record.
+	 * Evict as many other attributes as possible.
+	 */
+	max_free = free;
+
+	/* Estimate the result of moving all possible attributes away.*/
+	attr = NULL;
+
+	while ((attr = mi_enum_attr(&ni->mi, attr))) {
+		if (attr->type == ATTR_STD)
+			continue;
+		if (attr->type == ATTR_LIST)
+			continue;
+		max_free += le32_to_cpu(attr->size);
+	}
+
+	if (max_free < asize + list_reserve) {
+		/* Impossible to insert this attribute into primary record */
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* Start real attribute moving */
+	attr = NULL;
+
+	for (;;) {
+		attr = mi_enum_attr(&ni->mi, attr);
+		if (!attr) {
+			/* We should never be here 'cause we have already check this case */
+			err = -EINVAL;
+			goto out;
+		}
+
+		/* Skip attributes that MUST be primary record */
+		if (attr->type == ATTR_STD || attr->type == ATTR_LIST)
+			continue;
+
+		le = NULL;
+		if (ni->attr_list.size) {
+			le = al_find_le(ni, NULL, attr);
+			if (!le) {
+				/* Really this is a serious bug */
+				err = -EINVAL;
+				goto out;
+			}
+		}
+
+		t32 = le32_to_cpu(attr->size);
+		t16 = le16_to_cpu(attr->name_off);
+		err = ni_ins_attr_ext(ni, le, attr->type, Add2Ptr(attr, t16),
+				      attr->name_len, t32, attr_svcn(attr), t16,
+				      false, &eattr, NULL);
+		if (err)
+			return err;
+
+		id = eattr->id;
+		memcpy(eattr, attr, t32);
+		eattr->id = id;
+
+		/* remove attrib from primary record */
+		mi_remove_attr(&ni->mi, attr);
+
+		/* attr now points to next attribute */
+		if (attr->type == ATTR_END)
+			goto out;
+	}
+	while (asize + list_reserve > sbi->record_size - le32_to_cpu(rec->used))
+		;
+
+	attr = ni_ins_new_attr(ni, &ni->mi, NULL, type, name, name_len, asize,
+			       name_off, svcn);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (ins_attr)
+		*ins_attr = attr;
+	if (ins_mi)
+		*ins_mi = &ni->mi;
+
+out:
+	return err;
+}
+
+/*
+ * ni_expand_mft_list
+ *
+ * This method splits ATTR_DATA of $MFT
+ */
+static int ni_expand_mft_list(struct ntfs_inode *ni)
+{
+	int err = 0;
+	struct runs_tree *run = &ni->file.run;
+	u32 asize, run_size, done = 0;
+	struct ATTRIB *attr;
+	struct rb_node *node;
+	CLST mft_min, mft_new, svcn, evcn, plen;
+	struct mft_inode *mi, *mi_min, *mi_new;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+
+	/* Find the nearest Mft */
+	mft_min = 0;
+	mft_new = 0;
+	mi_min = NULL;
+
+	for (node = rb_first(&ni->mi_tree); node; node = rb_next(node)) {
+		mi = rb_entry(node, struct mft_inode, node);
+
+		attr = mi_enum_attr(mi, NULL);
+
+		if (!attr) {
+			mft_min = mi->rno;
+			mi_min = mi;
+			break;
+		}
+	}
+
+	if (ntfs_look_free_mft(sbi, &mft_new, true, ni, &mi_new)) {
+		mft_new = 0;
+		// really this is not critical
+	} else if (mft_min > mft_new) {
+		mft_min = mft_new;
+		mi_min = mi_new;
+	} else {
+		ntfs_mark_rec_free(sbi, mft_new);
+		mft_new = 0;
+		ni_remove_mi(ni, mi_new);
+	}
+
+	attr = mi_find_attr(&ni->mi, NULL, ATTR_DATA, NULL, 0, NULL);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	asize = le32_to_cpu(attr->size);
+
+	evcn = le64_to_cpu(attr->nres.evcn);
+	svcn = bytes_to_cluster(sbi, (u64)(mft_min + 1) << sbi->record_bits);
+	if (evcn + 1 >= svcn) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/*
+	 * split primary attribute [0 evcn] in two parts [0 svcn) + [svcn evcn]
+	 *
+	 * Update first part of ATTR_DATA in 'primary MFT
+	 */
+	err = run_pack(run, 0, svcn, Add2Ptr(attr, SIZEOF_NONRESIDENT),
+		       asize - SIZEOF_NONRESIDENT, &plen);
+	if (err < 0)
+		goto out;
+
+	run_size = QuadAlign(err);
+	err = 0;
+
+	if (plen < svcn) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	attr->nres.evcn = cpu_to_le64(svcn - 1);
+	attr->size = cpu_to_le32(run_size + SIZEOF_NONRESIDENT);
+	/* 'done' - how many bytes of primary MFT becomes free */
+	done = asize - run_size - SIZEOF_NONRESIDENT;
+	le32_sub_cpu(&ni->mi.mrec->used, done);
+
+	/* Estimate the size of second part: run_buf=NULL */
+	err = run_pack(run, svcn, evcn + 1 - svcn, NULL, sbi->record_size,
+		       &plen);
+	if (err < 0)
+		goto out;
+
+	run_size = QuadAlign(err);
+	err = 0;
+
+	if (plen < evcn + 1 - svcn) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/*
+	 * This function may implicitly call expand attr_list
+	 * Insert second part of ATTR_DATA in 'mi_min'
+	 */
+	attr = ni_ins_new_attr(ni, mi_min, NULL, ATTR_DATA, NULL, 0,
+			       SIZEOF_NONRESIDENT + run_size,
+			       SIZEOF_NONRESIDENT, svcn);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	attr->non_res = 1;
+	attr->name_off = SIZEOF_NONRESIDENT_LE;
+	attr->flags = 0;
+
+	run_pack(run, svcn, evcn + 1 - svcn, Add2Ptr(attr, SIZEOF_NONRESIDENT),
+		 run_size, &plen);
+
+	attr->nres.svcn = cpu_to_le64(svcn);
+	attr->nres.evcn = cpu_to_le64(evcn);
+	attr->nres.run_off = cpu_to_le16(SIZEOF_NONRESIDENT);
+
+out:
+	if (mft_new) {
+		ntfs_mark_rec_free(sbi, mft_new);
+		ni_remove_mi(ni, mi_new);
+	}
+
+	return !err && !done ? -EOPNOTSUPP : err;
+}
+
+/*
+ * ni_expand_list
+ *
+ * This method moves all possible attributes out of primary record
+ */
+int ni_expand_list(struct ntfs_inode *ni)
+{
+	int err = 0;
+	u32 asize, done = 0;
+	struct ATTRIB *attr, *ins_attr;
+	struct ATTR_LIST_ENTRY *le;
+	bool is_mft = ni->mi.rno == MFT_REC_MFT;
+	struct MFT_REF ref;
+
+	get_mi_ref(&ni->mi, &ref);
+	le = NULL;
+
+	while ((le = al_enumerate(ni, le))) {
+		if (le->type == ATTR_STD)
+			continue;
+
+		if (memcmp(&ref, &le->ref, sizeof(struct MFT_REF)))
+			continue;
+
+		if (is_mft && le->type == ATTR_DATA)
+			continue;
+
+		/* Find attribute in primary record */
+		attr = rec_find_attr_le(&ni->mi, le);
+		if (!attr) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		asize = le32_to_cpu(attr->size);
+
+		/* Always insert into new record to avoid collisions (deep recursive) */
+		err = ni_ins_attr_ext(ni, le, attr->type, attr_name(attr),
+				      attr->name_len, asize, attr_svcn(attr),
+				      le16_to_cpu(attr->name_off), true,
+				      &ins_attr, NULL);
+
+		if (err)
+			goto out;
+
+		memcpy(ins_attr, attr, asize);
+		ins_attr->id = le->id;
+		mi_remove_attr(&ni->mi, attr);
+
+		done += asize;
+		goto out;
+	}
+
+	if (!is_mft) {
+		err = -EFBIG; /* attr list is too big(?) */
+		goto out;
+	}
+
+	/* split mft data as much as possible */
+	err = ni_expand_mft_list(ni);
+	if (err)
+		goto out;
+
+out:
+	return !err && !done ? -EOPNOTSUPP : err;
+}
+
+/*
+ * ni_insert_nonresident
+ *
+ * inserts new nonresident attribute
+ */
+int ni_insert_nonresident(struct ntfs_inode *ni, enum ATTR_TYPE type,
+			  const __le16 *name, u8 name_len,
+			  const struct runs_tree *run, CLST svcn, CLST len,
+			  __le16 flags, struct ATTRIB **new_attr,
+			  struct mft_inode **mi)
+{
+	int err;
+	CLST plen;
+	struct ATTRIB *attr;
+	bool is_ext =
+		(flags & (ATTR_FLAG_SPARSED | ATTR_FLAG_COMPRESSED)) && !svcn;
+	u32 name_size = QuadAlign(name_len * sizeof(short));
+	u32 name_off = is_ext ? SIZEOF_NONRESIDENT_EX : SIZEOF_NONRESIDENT;
+	u32 run_off = name_off + name_size;
+	u32 run_size, asize;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+
+	err = run_pack(run, svcn, len, NULL, sbi->max_bytes_per_attr - run_off,
+		       &plen);
+	if (err < 0)
+		goto out;
+
+	run_size = QuadAlign(err);
+
+	if (plen < len) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	asize = run_off + run_size;
+
+	if (asize > sbi->max_bytes_per_attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	err = ni_insert_attr(ni, type, name, name_len, asize, name_off, svcn,
+			     &attr, mi);
+
+	if (err)
+		goto out;
+
+	attr->non_res = 1;
+	attr->name_off = cpu_to_le16(name_off);
+	attr->flags = flags;
+
+	run_pack(run, svcn, len, Add2Ptr(attr, run_off), run_size, &plen);
+
+	attr->nres.svcn = cpu_to_le64(svcn);
+	attr->nres.evcn = cpu_to_le64((u64)svcn + len - 1);
+
+	err = 0;
+	if (new_attr)
+		*new_attr = attr;
+
+	*(__le64 *)&attr->nres.run_off = cpu_to_le64(run_off);
+
+	attr->nres.alloc_size =
+		svcn ? 0 : cpu_to_le64((u64)len << ni->mi.sbi->cluster_bits);
+	attr->nres.data_size = attr->nres.alloc_size;
+	attr->nres.valid_size = attr->nres.alloc_size;
+
+	if (is_ext) {
+		if (flags & ATTR_FLAG_COMPRESSED)
+			attr->nres.c_unit = COMPRESSION_UNIT;
+		attr->nres.total_size = attr->nres.alloc_size;
+	}
+
+out:
+	return err;
+}
+
+/*
+ * ni_insert_resident
+ *
+ * inserts new resident attribute
+ */
+int ni_insert_resident(struct ntfs_inode *ni, u32 data_size,
+		       enum ATTR_TYPE type, const __le16 *name, u8 name_len,
+		       struct ATTRIB **new_attr, struct mft_inode **mi)
+{
+	int err;
+	u32 name_size = QuadAlign(name_len * sizeof(short));
+	u32 asize = SIZEOF_RESIDENT + name_size + QuadAlign(data_size);
+	struct ATTRIB *attr;
+
+	err = ni_insert_attr(ni, type, name, name_len, asize, SIZEOF_RESIDENT,
+			     0, &attr, mi);
+	if (err)
+		return err;
+
+	attr->non_res = 0;
+	attr->flags = 0;
+
+	attr->res.data_size = cpu_to_le32(data_size);
+	attr->res.data_off = cpu_to_le16(SIZEOF_RESIDENT + name_size);
+	if (type == ATTR_NAME)
+		attr->res.flags = RESIDENT_FLAG_INDEXED;
+	attr->res.res = 0;
+
+	if (new_attr)
+		*new_attr = attr;
+
+	return 0;
+}
+
+/*
+ * ni_remove_attr_le
+ *
+ * removes attribute from record
+ */
+int ni_remove_attr_le(struct ntfs_inode *ni, struct ATTRIB *attr,
+		      struct ATTR_LIST_ENTRY *le)
+{
+	int err;
+	struct mft_inode *mi;
+
+	err = ni_load_mi(ni, le, &mi);
+	if (err)
+		return err;
+
+	mi_remove_attr(mi, attr);
+
+	if (le)
+		al_remove_le(ni, le);
+
+	return 0;
+}
+
+/*
+ * ni_delete_all
+ *
+ * removes all attributes and frees allocates space
+ * ntfs_evict_inode->ntfs_clear_inode->ni_delete_all (if no links)
+ */
+int ni_delete_all(struct ntfs_inode *ni)
+{
+	int err;
+	struct ATTR_LIST_ENTRY *le = NULL;
+	struct ATTRIB *attr = NULL;
+	struct rb_node *node;
+	u16 roff;
+	u32 asize;
+	CLST svcn, evcn;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	bool nt3 = is_ntfs3(sbi);
+	struct MFT_REF ref;
+
+	while ((attr = ni_enum_attr_ex(ni, attr, &le))) {
+		if (!nt3 || attr->name_len) {
+			;
+		} else if (attr->type == ATTR_REPARSE) {
+			get_mi_ref(&ni->mi, &ref);
+			ntfs_remove_reparse(sbi, 0, &ref);
+		} else if (attr->type == ATTR_ID && !attr->non_res &&
+			   le32_to_cpu(attr->res.data_size) >=
+				   sizeof(struct GUID)) {
+			ntfs_objid_remove(sbi, resident_data(attr));
+		}
+
+		if (!attr->non_res)
+			continue;
+
+		svcn = le64_to_cpu(attr->nres.svcn);
+		evcn = le64_to_cpu(attr->nres.evcn);
+
+		if (evcn + 1 <= svcn)
+			continue;
+
+		asize = le32_to_cpu(attr->size);
+		roff = le16_to_cpu(attr->nres.run_off);
+
+		/*run==1 means unpack and deallocate*/
+		run_unpack_ex((struct runs_tree *)(size_t)1, sbi, ni->mi.rno,
+			      svcn, evcn, svcn, Add2Ptr(attr, roff),
+			      asize - roff);
+	}
+
+	if (ni->attr_list.size) {
+		run_deallocate(ni->mi.sbi, &ni->attr_list.run, true);
+		al_destroy(ni);
+	}
+
+	/* Free all subrecords */
+	for (node = rb_first(&ni->mi_tree); node;) {
+		struct rb_node *next = rb_next(node);
+		struct mft_inode *mi = rb_entry(node, struct mft_inode, node);
+
+		clear_rec_inuse(mi->mrec);
+		mi->dirty = true;
+		mi_write(mi, 0);
+
+		ntfs_mark_rec_free(sbi, mi->rno);
+		ni_remove_mi(ni, mi);
+		mi_put(mi);
+		node = next;
+	}
+
+	// Free base record
+	clear_rec_inuse(ni->mi.mrec);
+	ni->mi.dirty = true;
+	err = mi_write(&ni->mi, 0);
+
+	ntfs_mark_rec_free(sbi, ni->mi.rno);
+
+	return err;
+}
+
+/*
+ * ni_fname_name
+ *
+ * returns file name attribute by its value
+ */
+struct ATTR_FILE_NAME *ni_fname_name(struct ntfs_inode *ni,
+				     const struct cpu_str *uni,
+				     const struct MFT_REF *home_dir,
+				     struct ATTR_LIST_ENTRY **le)
+{
+	struct ATTRIB *attr = NULL;
+	struct ATTR_FILE_NAME *fname;
+
+	*le = NULL;
+
+	/* Enumerate all names */
+next:
+	attr = ni_find_attr(ni, attr, le, ATTR_NAME, NULL, 0, NULL, NULL);
+	if (!attr)
+		return NULL;
+
+	fname = resident_data_ex(attr, SIZEOF_ATTRIBUTE_FILENAME);
+	if (!fname)
+		goto next;
+
+	if (home_dir && memcmp(home_dir, &fname->home, sizeof(*home_dir)))
+		goto next;
+
+	if (!uni)
+		goto next;
+
+	if (uni->len != fname->name_len)
+		goto next;
+
+	if (ntfs_cmp_names_cpu(uni, (struct le_str *)&fname->name_len, NULL))
+		goto next;
+
+	return fname;
+}
+
+/*
+ * ni_fname_type
+ *
+ * returns file name attribute with given type
+ */
+struct ATTR_FILE_NAME *ni_fname_type(struct ntfs_inode *ni, u8 name_type,
+				     struct ATTR_LIST_ENTRY **le)
+{
+	struct ATTRIB *attr = NULL;
+	struct ATTR_FILE_NAME *fname;
+
+	*le = NULL;
+
+	/* Enumerate all names */
+	for (;;) {
+		attr = ni_find_attr(ni, attr, le, ATTR_NAME, NULL, 0, NULL,
+				    NULL);
+		if (!attr)
+			return NULL;
+
+		fname = resident_data_ex(attr, SIZEOF_ATTRIBUTE_FILENAME);
+		if (fname && name_type == fname->type)
+			return fname;
+	}
+}
+
+/*
+ * Process compressed/sparsed in special way
+ * NOTE: you need to set ni->std_fa = new_fa
+ * after this function to keep internal structures in consistency
+ */
+int ni_new_attr_flags(struct ntfs_inode *ni, enum FILE_ATTRIBUTE new_fa)
+{
+	struct ATTRIB *attr;
+	struct mft_inode *mi;
+	__le16 new_aflags;
+	u32 new_asize;
+
+	attr = ni_find_attr(ni, NULL, NULL, ATTR_DATA, NULL, 0, NULL, &mi);
+	if (!attr)
+		return -EINVAL;
+
+	new_aflags = attr->flags;
+
+	if (new_fa & FILE_ATTRIBUTE_SPARSE_FILE)
+		new_aflags |= ATTR_FLAG_SPARSED;
+	else
+		new_aflags &= ~ATTR_FLAG_SPARSED;
+
+	if (new_fa & FILE_ATTRIBUTE_COMPRESSED)
+		new_aflags |= ATTR_FLAG_COMPRESSED;
+	else
+		new_aflags &= ~ATTR_FLAG_COMPRESSED;
+
+	if (new_aflags == attr->flags)
+		return 0;
+
+	if ((new_aflags & (ATTR_FLAG_COMPRESSED | ATTR_FLAG_SPARSED)) ==
+	    (ATTR_FLAG_COMPRESSED | ATTR_FLAG_SPARSED)) {
+		ntfs_inode_warn(&ni->vfs_inode,
+				"file can't be sparsed and compressed");
+		return -EOPNOTSUPP;
+	}
+
+	if (!attr->non_res)
+		goto out;
+
+	if (attr->nres.data_size) {
+		ntfs_inode_warn(
+			&ni->vfs_inode,
+			"one can change sparsed/compressed only for empty files");
+		return -EOPNOTSUPP;
+	}
+
+	/* resize nonresident empty attribute in-place only*/
+	new_asize = (new_aflags & (ATTR_FLAG_COMPRESSED | ATTR_FLAG_SPARSED)) ?
+			    (SIZEOF_NONRESIDENT_EX + 8) :
+			    (SIZEOF_NONRESIDENT + 8);
+
+	if (!mi_resize_attr(mi, attr, new_asize - le32_to_cpu(attr->size)))
+		return -EOPNOTSUPP;
+
+	if (new_aflags & ATTR_FLAG_SPARSED) {
+		attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
+		/* windows uses 16 clusters per frame but supports one cluster per frame too*/
+		attr->nres.c_unit = 0;
+		ni->vfs_inode.i_mapping->a_ops = &ntfs_aops;
+	} else if (new_aflags & ATTR_FLAG_COMPRESSED) {
+		attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
+		/* the only allowed: 16 clusters per frame */
+		attr->nres.c_unit = NTFS_LZNT_CUNIT;
+		ni->vfs_inode.i_mapping->a_ops = &ntfs_aops_cmpr;
+	} else {
+		attr->name_off = SIZEOF_NONRESIDENT_LE;
+		/* normal files */
+		attr->nres.c_unit = 0;
+		ni->vfs_inode.i_mapping->a_ops = &ntfs_aops;
+	}
+	attr->nres.run_off = attr->name_off;
+out:
+	attr->flags = new_aflags;
+	mi->dirty = true;
+
+	return 0;
+}
+
+/*
+ * ni_parse_reparse
+ *
+ * buffer is at least 24 bytes
+ */
+enum REPARSE_SIGN ni_parse_reparse(struct ntfs_inode *ni, struct ATTRIB *attr,
+				   void *buffer)
+{
+	const struct REPARSE_DATA_BUFFER *rp = NULL;
+	u32 c_format;
+	u16 len;
+	typeof(rp->CompressReparseBuffer) *cmpr;
+
+	static_assert(sizeof(struct REPARSE_DATA_BUFFER) <= 24);
+
+	/* Try to estimate reparse point */
+	if (!attr->non_res) {
+		rp = resident_data_ex(attr, sizeof(struct REPARSE_DATA_BUFFER));
+	} else if (le64_to_cpu(attr->nres.data_size) >=
+		   sizeof(struct REPARSE_DATA_BUFFER)) {
+		struct runs_tree run;
+
+		run_init(&run);
+
+		if (!attr_load_runs_vcn(ni, ATTR_REPARSE, NULL, 0, &run, 0) &&
+		    !ntfs_read_run_nb(ni->mi.sbi, &run, 0, buffer,
+				      sizeof(struct REPARSE_DATA_BUFFER),
+				      NULL)) {
+			rp = buffer;
+		}
+
+		run_close(&run);
+	}
+
+	if (!rp)
+		return REPARSE_NONE;
+
+	len = le16_to_cpu(rp->ReparseDataLength);
+	switch (rp->ReparseTag) {
+	case (IO_REPARSE_TAG_MICROSOFT | IO_REPARSE_TAG_SYMBOLIC_LINK):
+		break; /* Symbolic link */
+	case IO_REPARSE_TAG_MOUNT_POINT:
+		break; /* Mount points and junctions */
+	case IO_REPARSE_TAG_SYMLINK:
+		break;
+	case IO_REPARSE_TAG_COMPRESS:
+		cmpr = &rp->CompressReparseBuffer;
+		if (len < sizeof(*cmpr) ||
+		    cmpr->WofVersion != WOF_CURRENT_VERSION ||
+		    cmpr->WofProvider != WOF_PROVIDER_SYSTEM ||
+		    cmpr->ProviderVer != WOF_PROVIDER_CURRENT_VERSION) {
+			return REPARSE_NONE;
+		}
+		c_format = le32_to_cpu(cmpr->CompressionFormat);
+		if (c_format > 3)
+			return REPARSE_NONE;
+
+		ni->ni_flags |= (c_format + 1) << 8;
+		return REPARSE_COMPRESSED;
+
+	case IO_REPARSE_TAG_DEDUP:
+		ni->ni_flags |= NI_FLAG_DEDUPLICATED;
+		return REPARSE_DEDUPLICATED;
+
+	default:
+		if (rp->ReparseTag & IO_REPARSE_TAG_NAME_SURROGATE)
+			break;
+
+		return REPARSE_NONE;
+	}
+
+	/* Looks like normal symlink */
+	return REPARSE_LINK;
+}
+
+/*
+ * helper for file_fiemap
+ * assumed ni_lock
+ * TODO: less aggressive locks
+ */
+int ni_fiemap(struct ntfs_inode *ni, struct fiemap_extent_info *fieinfo,
+	      __u64 vbo, __u64 len)
+{
+	int err = 0;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	u8 cluster_bits = sbi->cluster_bits;
+	struct runs_tree *run;
+	struct rw_semaphore *run_lock;
+	struct ATTRIB *attr;
+	CLST vcn = vbo >> cluster_bits;
+	CLST lcn, clen;
+	u64 valid = ni->i_valid;
+	u64 lbo, bytes;
+	u64 end, alloc_size;
+	size_t idx = -1;
+	u32 flags;
+	bool ok;
+
+	if (S_ISDIR(ni->vfs_inode.i_mode)) {
+		run = &ni->dir.alloc_run;
+		attr = ni_find_attr(ni, NULL, NULL, ATTR_ALLOC, I30_NAME,
+				    ARRAY_SIZE(I30_NAME), NULL, NULL);
+		run_lock = NULL;
+	} else {
+		run = &ni->file.run;
+		attr = ni_find_attr(ni, NULL, NULL, ATTR_DATA, NULL, 0, NULL,
+				    NULL);
+		if (!attr) {
+			err = -EINVAL;
+			goto out;
+		}
+		if (is_attr_compressed(attr)) {
+			/*unfortunately cp -r incorrectly treats compressed clusters*/
+			err = -EOPNOTSUPP;
+			ntfs_inode_warn(
+				&ni->vfs_inode,
+				"fiemap is not supported for compressed file (cp -r)");
+			goto out;
+		}
+		run_lock = &ni->file.run_lock;
+	}
+
+	if (!attr || !attr->non_res) {
+		err = fiemap_fill_next_extent(
+			fieinfo, 0, 0,
+			attr ? le32_to_cpu(attr->res.data_size) : 0,
+			FIEMAP_EXTENT_DATA_INLINE | FIEMAP_EXTENT_LAST |
+				FIEMAP_EXTENT_MERGED);
+		goto out;
+	}
+
+	end = vbo + len;
+	alloc_size = le64_to_cpu(attr->nres.alloc_size);
+	if (end > alloc_size)
+		end = alloc_size;
+
+	if (run_lock)
+		down_read(run_lock);
+
+	while (vbo < end) {
+		if (idx == -1) {
+			ok = run_lookup_entry(run, vcn, &lcn, &clen, &idx);
+		} else {
+			CLST next_vcn = vcn;
+
+			ok = run_get_entry(run, ++idx, &vcn, &lcn, &clen);
+			if (ok && vcn != next_vcn) {
+				ok = false;
+				vcn = next_vcn;
+			}
+		}
+
+		if (!ok) {
+			if (run_lock) {
+				up_read(run_lock);
+				down_write(run_lock);
+			}
+
+			err = attr_load_runs_vcn(ni, attr->type,
+						 attr_name(attr),
+						 attr->name_len, run, vcn);
+
+			if (run_lock) {
+				up_write(run_lock);
+				down_read(run_lock);
+			}
+
+			if (err)
+				break;
+
+			ok = run_lookup_entry(run, vcn, &lcn, &clen, &idx);
+
+			if (!ok) {
+				err = -EINVAL;
+				break;
+			}
+		}
+
+		if (!clen) {
+			err = -EINVAL; // ?
+			break;
+		}
+
+		if (lcn == SPARSE_LCN) {
+			vcn += clen;
+			vbo = (u64)vcn << cluster_bits;
+			continue;
+		}
+
+		flags = FIEMAP_EXTENT_MERGED;
+		if (S_ISDIR(ni->vfs_inode.i_mode)) {
+			;
+		} else if (is_attr_compressed(attr)) {
+			CLST clst_data;
+
+			err = attr_is_frame_compressed(
+				ni, attr, vcn >> attr->nres.c_unit, &clst_data);
+			if (err)
+				break;
+			if (clst_data < NTFS_LZNT_CLUSTERS)
+				flags |= FIEMAP_EXTENT_ENCODED;
+		} else if (is_attr_encrypted(attr)) {
+			flags |= FIEMAP_EXTENT_DATA_ENCRYPTED;
+		}
+
+		vbo = (u64)vcn << cluster_bits;
+		bytes = (u64)clen << cluster_bits;
+		lbo = (u64)lcn << cluster_bits;
+
+		vcn += clen;
+
+		if (vbo + bytes >= end) {
+			bytes = end - vbo;
+			flags |= FIEMAP_EXTENT_LAST;
+		}
+
+		if (vbo + bytes <= valid) {
+			;
+		} else if (vbo >= valid) {
+			flags |= FIEMAP_EXTENT_UNWRITTEN;
+		} else {
+			/* vbo < valid && valid < vbo + bytes */
+			u64 dlen = valid - vbo;
+
+			err = fiemap_fill_next_extent(fieinfo, vbo, lbo, dlen,
+						      flags);
+			if (err < 0)
+				break;
+			if (err == 1) {
+				err = 0;
+				break;
+			}
+
+			vbo = valid;
+			bytes -= dlen;
+			if (!bytes)
+				continue;
+
+			lbo += dlen;
+			flags |= FIEMAP_EXTENT_UNWRITTEN;
+		}
+
+		err = fiemap_fill_next_extent(fieinfo, vbo, lbo, bytes, flags);
+		if (err < 0)
+			break;
+		if (err == 1) {
+			err = 0;
+			break;
+		}
+
+		vbo += bytes;
+	}
+
+	if (run_lock)
+		up_read(run_lock);
+
+out:
+	return err;
+}
+
+/*
+ * When decompressing, we typically obtain more than one page per reference.
+ * We inject the additional pages into the page cache.
+ */
+int ni_readpage_cmpr(struct ntfs_inode *ni, struct page *page)
+{
+	int err;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	struct address_space *mapping = page->mapping;
+	pgoff_t index = page->index;
+	u64 frame_vbo, vbo = (u64)index << PAGE_SHIFT;
+	struct page **pages = NULL; /*array of at most 16 pages. stack?*/
+	u8 frame_bits;
+	CLST frame;
+	u32 i, idx, frame_size, pages_per_frame;
+	gfp_t gfp_mask;
+	struct page *pg;
+
+	if (vbo >= ni->vfs_inode.i_size) {
+		SetPageUptodate(page);
+		err = 0;
+		goto out;
+	}
+
+	if (ni->ni_flags & NI_FLAG_COMPRESSED_MASK) {
+		u32 cmpr = ((ni->ni_flags & NI_FLAG_COMPRESSED_MASK) >> 8) - 1;
+
+		switch (cmpr) {
+		case WOF_COMPRESSION_XPRESS4K:
+			/* xpress 4k*/
+			frame_bits = 12;
+			break;
+		case WOF_COMPRESSION_LZX:
+			/* lzx 32k*/
+			frame_bits = 15;
+			break;
+		case WOF_COMPRESSION_XPRESS8K:
+			/* xpress 8k*/
+			frame_bits = 13;
+			break;
+		case WOF_COMPRESSION_XPRESS16K:
+			/* xpress 16k*/
+			frame_bits = 14;
+			break;
+		}
+		/* TODO: port lzx/xpress */
+		err = -EOPNOTSUPP;
+		goto out;
+	} else {
+		/* lznt compression*/
+		frame_bits = NTFS_LZNT_CUNIT + sbi->cluster_bits;
+	}
+	frame_size = 1u << frame_bits;
+	frame = vbo >> frame_bits;
+	frame_vbo = (u64)frame << frame_bits;
+	idx = (vbo - frame_vbo) >> PAGE_SHIFT;
+
+	pages_per_frame = frame_size >> PAGE_SHIFT;
+	pages = ntfs_alloc(pages_per_frame * sizeof(struct page *), 1);
+	if (!pages) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	pages[idx] = page;
+	index = frame_vbo >> PAGE_SHIFT;
+	gfp_mask = mapping_gfp_mask(mapping);
+
+	for (i = 0; i < pages_per_frame; i++, index++) {
+		if (i == idx)
+			continue;
+
+		pg = find_or_create_page(mapping, index, gfp_mask);
+		if (!pg) {
+			err = -ENOMEM;
+			goto out1;
+		}
+		pages[i] = pg;
+	}
+
+	err = ni_read_frame(ni, frame_vbo, pages, pages_per_frame);
+
+out1:
+	if (err)
+		SetPageError(page);
+
+	for (i = 0; i < pages_per_frame; i++) {
+		pg = pages[i];
+		if (i == idx)
+			continue;
+		unlock_page(pg);
+		put_page(pg);
+	}
+
+out:
+	/* At this point, err contains 0 or -EIO depending on the "critical" page */
+	ntfs_free(pages);
+	unlock_page(page);
+
+	return err;
+}
+
+/*
+ * ni_read_frame
+ *
+ * pages - array of locked pages
+ */
+int ni_read_frame(struct ntfs_inode *ni, u64 frame_vbo, struct page **pages,
+		  u32 pages_per_frame)
+{
+	int err;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	char *frame_ondisk = NULL;
+	char *frame_mem = NULL;
+	struct page **pages_disk = NULL;
+	u64 valid_size = ni->i_valid;
+	u32 frame_size;
+	size_t unc_size;
+	u32 ondisk_size, i, npages_disk;
+	struct page *pg;
+	struct ATTRIB *attr;
+	CLST frame, clst_data;
+
+	attr = ni_find_attr(ni, NULL, NULL, ATTR_DATA, NULL, 0, NULL, NULL);
+	if (!attr) {
+		err = -ENOENT;
+		goto out;
+	}
+
+	if (!attr->non_res) {
+		u32 data_size = le32_to_cpu(attr->res.data_size);
+		void *kaddr = kmap_atomic(pages[0]);
+
+		memset(kaddr, 0, PAGE_SIZE);
+		if (frame_vbo < data_size) {
+			err = data_size - frame_vbo;
+			if (err > PAGE_SIZE)
+				err = PAGE_SIZE;
+			memcpy(kaddr, resident_data(attr) + frame_vbo, err);
+		}
+		flush_dcache_page(pages[0]);
+		kunmap_atomic(kaddr);
+		err = 0;
+		goto out;
+	}
+
+	if (frame_vbo >= valid_size) {
+all_zero:
+		for (i = 0; i < pages_per_frame; i++)
+			zero_user_segment(pages[i], 0, PAGE_SIZE);
+		err = 0;
+		goto out;
+	}
+
+	if (ni->ni_flags & NI_FLAG_COMPRESSED_MASK) {
+		u32 cmpr = ((ni->ni_flags & NI_FLAG_COMPRESSED_MASK) >> 8) - 1;
+
+		switch (cmpr) {
+		case WOF_COMPRESSION_XPRESS4K:
+			/* xpress 4k*/
+			frame_size = 0x1000;
+			break;
+		case WOF_COMPRESSION_LZX:
+			/* lzx 32k*/
+			frame_size = 0x8000;
+			break;
+		case WOF_COMPRESSION_XPRESS8K:
+			/* xpress 8k*/
+			frame_size = 0x2000;
+			break;
+		case WOF_COMPRESSION_XPRESS16K:
+			/* xpress 16k*/
+			frame_size = 0x4000;
+			break;
+		}
+		/* TODO: port lzx/xpress */
+		err = -EOPNOTSUPP;
+		goto out;
+	} else if (is_attr_compressed(attr)) {
+		/* lznt compression*/
+		if (sbi->cluster_size > NTFS_LZNT_MAX_CLUSTER) {
+			err = -EOPNOTSUPP;
+			goto out;
+		}
+
+		if (attr->nres.c_unit != NTFS_LZNT_CUNIT) {
+			err = -EOPNOTSUPP;
+			goto out;
+		}
+
+		down_write(&ni->file.run_lock);
+		run_truncate_around(&ni->file.run,
+				    le64_to_cpu(attr->nres.svcn));
+		frame = frame_vbo >> (sbi->cluster_bits + NTFS_LZNT_CUNIT);
+		err = attr_is_frame_compressed(ni, attr, frame, &clst_data);
+		up_write(&ni->file.run_lock);
+		if (err)
+			goto out;
+
+		if (!clst_data)
+			goto all_zero;
+
+		frame_size = sbi->cluster_size << NTFS_LZNT_CUNIT;
+		ondisk_size = clst_data << sbi->cluster_bits;
+
+		if (clst_data >= NTFS_LZNT_CLUSTERS) {
+			/* frame is not compressed */
+			down_read(&ni->file.run_lock);
+			err = ntfs_bio_pages(sbi, &ni->file.run, pages,
+					     pages_per_frame, frame_vbo,
+					     ondisk_size, REQ_OP_READ);
+			up_read(&ni->file.run_lock);
+			goto out;
+		}
+	} else {
+		__builtin_unreachable();
+		err = -EINVAL;
+		goto out;
+	}
+
+	npages_disk = (ondisk_size + PAGE_SIZE - 1) >> PAGE_SHIFT;
+	pages_disk = ntfs_alloc(npages_disk * sizeof(struct page *), 1);
+	if (!pages_disk) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	for (i = 0; i < npages_disk; i++) {
+		pg = alloc_page(GFP_KERNEL);
+		if (!pg) {
+			err = -ENOMEM;
+			goto out1;
+		}
+		pages_disk[i] = pg;
+		lock_page(pg);
+		kmap(pg);
+	}
+
+	/* read 'ondisk_size' bytes from disk */
+	down_read(&ni->file.run_lock);
+	err = ntfs_bio_pages(sbi, &ni->file.run, pages_disk, npages_disk,
+			     frame_vbo, ondisk_size, REQ_OP_READ);
+	up_read(&ni->file.run_lock);
+	if (err)
+		goto out1;
+
+	/*
+	 * To simplify decompress algorithm do vmap for source and target pages
+	 */
+	frame_ondisk = vmap(pages_disk, npages_disk, VM_MAP, PAGE_KERNEL);
+	if (!frame_ondisk) {
+		err = -ENOMEM;
+		goto out1;
+	}
+
+	for (i = 0; i < pages_per_frame; i++)
+		kmap(pages[i]);
+
+	frame_mem = vmap(pages, pages_per_frame, VM_MAP, PAGE_KERNEL);
+	if (!frame_mem) {
+		err = -ENOMEM;
+		goto out2;
+	}
+
+	/* decompress: frame_ondisk -> frame_mem */
+	unc_size = decompress_lznt(frame_ondisk, ondisk_size, frame_mem,
+				   frame_size);
+
+	if ((ssize_t)unc_size < 0) {
+		err = unc_size;
+	} else if (!unc_size || unc_size > frame_size) {
+		err = -EINVAL;
+	} else {
+		if (valid_size < frame_vbo + frame_size) {
+			size_t ok = valid_size - frame_vbo;
+
+			memset(frame_mem + ok, 0, frame_size - ok);
+		}
+	}
+
+	vunmap(frame_mem);
+out2:
+	for (i = 0; i < pages_per_frame; i++) {
+		pg = pages[i];
+		kunmap(pg);
+	}
+	vunmap(frame_ondisk);
+out1:
+	for (i = 0; i < npages_disk; i++) {
+		pg = pages_disk[i];
+		if (pg) {
+			kunmap(pg);
+			unlock_page(pg);
+			put_page(pg);
+		}
+	}
+	ntfs_free(pages_disk);
+out:
+	for (i = 0; i < pages_per_frame; i++) {
+		pg = pages[i];
+		ClearPageError(pg);
+		SetPageUptodate(pg);
+	}
+
+	return err;
+}
+
+/*
+ * ni_write_frame
+ *
+ * pages - array of locked pages
+ */
+int ni_write_frame(struct ntfs_inode *ni, struct page **pages,
+		   u32 pages_per_frame)
+{
+	int err;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	u8 frame_bits = NTFS_LZNT_CUNIT + sbi->cluster_bits;
+	u32 frame_size = sbi->cluster_size << NTFS_LZNT_CUNIT;
+	u64 frame_vbo = (u64)pages[0]->index << PAGE_SHIFT;
+	CLST frame = frame_vbo >> frame_bits;
+	char *frame_ondisk = NULL;
+	struct page **pages_disk = NULL;
+	char *frame_mem;
+	struct ATTRIB *attr;
+	u32 i;
+	struct page *pg;
+	size_t compr_size, ondisk_size;
+
+	attr = ni_find_attr(ni, NULL, NULL, ATTR_DATA, NULL, 0, NULL, NULL);
+	if (!attr) {
+		err = -ENOENT;
+		goto out;
+	}
+
+	if (!attr->non_res) {
+		WARN_ON(1);
+		err = 0;
+		goto out;
+	}
+
+	if (!is_attr_compressed(attr)) {
+		WARN_ON(1);
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (sbi->cluster_size > NTFS_LZNT_MAX_CLUSTER) {
+		err = -EOPNOTSUPP;
+		goto out;
+	}
+
+	if (attr->nres.c_unit != NTFS_LZNT_CUNIT) {
+		err = -EOPNOTSUPP;
+		goto out;
+	}
+
+	pages_disk = ntfs_alloc(pages_per_frame * sizeof(struct page *), 1);
+	if (!pages_disk) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	for (i = 0; i < pages_per_frame; i++) {
+		pg = alloc_page(GFP_KERNEL);
+		if (!pg) {
+			err = -ENOMEM;
+			goto out1;
+		}
+		pages_disk[i] = pg;
+		lock_page(pg);
+		kmap(pg);
+	}
+
+	/*
+	 * To simplify compress algorithm do vmap for source and target pages
+	 */
+	frame_ondisk = vmap(pages_disk, pages_per_frame, VM_MAP, PAGE_KERNEL);
+	if (!frame_ondisk) {
+		err = -ENOMEM;
+		goto out1;
+	}
+
+	for (i = 0; i < pages_per_frame; i++)
+		kmap(pages[i]);
+
+	frame_mem = vmap(pages, pages_per_frame, VM_MAP, PAGE_KERNEL_RO);
+	if (!frame_mem) {
+		err = -ENOMEM;
+		goto out2;
+	}
+
+	spin_lock(&sbi->compress.lock);
+	/* compress: frame_mem -> frame_ondisk */
+	compr_size = compress_lznt(frame_mem, frame_size, frame_ondisk,
+				   frame_size, sbi->compress.ctx);
+	spin_unlock(&sbi->compress.lock);
+
+	if (compr_size + sbi->cluster_size > frame_size) {
+		/* frame is not compressed */
+		compr_size = frame_size;
+		ondisk_size = frame_size;
+	} else if (compr_size) {
+		/* frame is compressed */
+		ondisk_size = ntfs_up_cluster(sbi, compr_size);
+		memset(frame_ondisk + compr_size, 0, ondisk_size - compr_size);
+	} else {
+		/* frame is sparsed */
+		ondisk_size = 0;
+	}
+
+	down_write(&ni->file.run_lock);
+	run_truncate_around(&ni->file.run, le64_to_cpu(attr->nres.svcn));
+	err = attr_allocate_frame(ni, frame, compr_size, ni->i_valid);
+	up_write(&ni->file.run_lock);
+	if (err)
+		goto out2;
+
+	if (!ondisk_size)
+		goto out2;
+
+	down_read(&ni->file.run_lock);
+	err = ntfs_bio_pages(sbi, &ni->file.run,
+			     ondisk_size < frame_size ? pages_disk : pages,
+			     pages_per_frame, frame_vbo, ondisk_size,
+			     REQ_OP_WRITE);
+	up_read(&ni->file.run_lock);
+
+	vunmap(frame_mem);
+
+out2:
+	for (i = 0; i < pages_per_frame; i++)
+		kunmap(pages[i]);
+
+	vunmap(frame_ondisk);
+out1:
+	for (i = 0; i < pages_per_frame; i++) {
+		pg = pages_disk[i];
+		if (pg) {
+			kunmap(pg);
+			unlock_page(pg);
+			put_page(pg);
+		}
+	}
+	ntfs_free(pages_disk);
+out:
+	return err;
+}
+
+/*
+ * update duplicate info of ATTR_FILE_NAME in MFT and in parent directories
+ */
+static bool ni_update_parent(struct ntfs_inode *ni, struct NTFS_DUP_INFO *dup,
+			     int sync)
+{
+	struct ATTRIB *attr;
+	struct mft_inode *mi;
+	struct ATTR_LIST_ENTRY *le = NULL;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	struct super_block *sb = sbi->sb;
+	bool re_dirty = false;
+	bool active = sb->s_flags & SB_ACTIVE;
+	bool upd_parent = ni->ni_flags & NI_FLAG_UPDATE_PARENT;
+
+	if (ni->mi.mrec->flags & RECORD_FLAG_DIR) {
+		dup->fa |= FILE_ATTRIBUTE_DIRECTORY;
+		attr = NULL;
+		dup->alloc_size = 0;
+		dup->data_size = 0;
+	} else {
+		dup->fa &= ~FILE_ATTRIBUTE_DIRECTORY;
+
+		attr = ni_find_attr(ni, NULL, &le, ATTR_DATA, NULL, 0, NULL,
+				    &mi);
+		if (!attr) {
+			dup->alloc_size = dup->data_size = 0;
+		} else if (!attr->non_res) {
+			u32 data_size = le32_to_cpu(attr->res.data_size);
+
+			dup->alloc_size = cpu_to_le64(QuadAlign(data_size));
+			dup->data_size = cpu_to_le64(data_size);
+		} else {
+			u64 new_valid = ni->i_valid;
+			u64 data_size = le64_to_cpu(attr->nres.data_size);
+			__le64 valid_le;
+
+			dup->alloc_size = is_attr_ext(attr) ?
+						  attr->nres.total_size :
+						  attr->nres.alloc_size;
+			dup->data_size = attr->nres.data_size;
+
+			if (new_valid > data_size)
+				new_valid = data_size;
+
+			valid_le = cpu_to_le64(new_valid);
+			if (valid_le != attr->nres.valid_size) {
+				attr->nres.valid_size = valid_le;
+				mi->dirty = true;
+			}
+		}
+	}
+
+	/* TODO: fill reparse info */
+	dup->reparse = 0;
+	dup->ea_size = 0;
+
+	if (ni->ni_flags & NI_FLAG_EA) {
+		attr = ni_find_attr(ni, attr, &le, ATTR_EA_INFO, NULL, 0, NULL,
+				    NULL);
+		if (attr) {
+			const struct EA_INFO *info;
+
+			info = resident_data_ex(attr, sizeof(struct EA_INFO));
+			dup->ea_size = info->size_pack;
+		}
+	}
+
+	attr = NULL;
+	le = NULL;
+
+	while ((attr = ni_find_attr(ni, attr, &le, ATTR_NAME, NULL, 0, NULL,
+				    &mi))) {
+		struct inode *dir;
+		struct ATTR_FILE_NAME *fname;
+
+		fname = resident_data_ex(attr, SIZEOF_ATTRIBUTE_FILENAME);
+		if (!fname)
+			continue;
+
+		if (memcmp(&fname->dup, dup, sizeof(fname->dup))) {
+			memcpy(&fname->dup, dup, sizeof(fname->dup));
+			mi->dirty = true;
+		} else if (!upd_parent) {
+			continue;
+		}
+
+		if (!active)
+			continue; /*avoid __wait_on_freeing_inode(inode); */
+
+		/*ntfs_iget5 may sleep*/
+		dir = ntfs_iget5(sb, &fname->home, NULL);
+		if (IS_ERR(dir)) {
+			ntfs_inode_warn(
+				&ni->vfs_inode,
+				"failed to open parent directory r=%lx to update",
+				(long)ino_get(&fname->home));
+			continue;
+		}
+
+		if (!is_bad_inode(dir)) {
+			struct ntfs_inode *dir_ni = ntfs_i(dir);
+
+			if (!ni_trylock(dir_ni)) {
+				re_dirty = true;
+			} else {
+				indx_update_dup(dir_ni, sbi, fname, dup, sync);
+				ni_unlock(dir_ni);
+			}
+		}
+		iput(dir);
+	}
+
+	return re_dirty;
+}
+
+/*
+ * ni_write_inode
+ *
+ * write mft base record and all subrecords to disk
+ */
+int ni_write_inode(struct inode *inode, int sync, const char *hint)
+{
+	int err = 0, err2;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	struct super_block *sb = inode->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	bool re_dirty = false;
+	struct ATTR_STD_INFO *std;
+	struct rb_node *node, *next;
+	struct NTFS_DUP_INFO dup;
+
+	if (is_bad_inode(inode) || sb_rdonly(sb))
+		return 0;
+
+	if (!ni_trylock(ni)) {
+		/* 'ni' is under modification, skip for now */
+		mark_inode_dirty_sync(inode);
+		return 0;
+	}
+
+	if (is_rec_inuse(ni->mi.mrec) &&
+	    !(sbi->flags & NTFS_FLAGS_LOG_REPLAYING) && inode->i_nlink) {
+		bool modified = false;
+
+		/* update times in standard attribute */
+		std = ni_std(ni);
+		if (!std) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		/* Update the access times if they have changed. */
+		dup.m_time = kernel2nt(&inode->i_mtime);
+		if (std->m_time != dup.m_time) {
+			std->m_time = dup.m_time;
+			modified = true;
+		}
+
+		dup.c_time = kernel2nt(&inode->i_ctime);
+		if (std->c_time != dup.c_time) {
+			std->c_time = dup.c_time;
+			modified = true;
+		}
+
+		dup.a_time = kernel2nt(&inode->i_atime);
+		if (std->a_time != dup.a_time) {
+			std->a_time = dup.a_time;
+			modified = true;
+		}
+
+		dup.fa = ni->std_fa;
+		if (std->fa != dup.fa) {
+			std->fa = dup.fa;
+			modified = true;
+		}
+
+		if (modified)
+			ni->mi.dirty = true;
+
+		if (!ntfs_is_meta_file(sbi, inode->i_ino) &&
+		    (modified || (ni->ni_flags & NI_FLAG_UPDATE_PARENT))) {
+			dup.cr_time = std->cr_time;
+			/* Not critical if this function fail */
+			re_dirty = ni_update_parent(ni, &dup, sync);
+
+			if (re_dirty)
+				ni->ni_flags |= NI_FLAG_UPDATE_PARENT;
+			else
+				ni->ni_flags &= ~NI_FLAG_UPDATE_PARENT;
+		}
+
+		/* update attribute list */
+		if (ni->attr_list.size && ni->attr_list.dirty) {
+			if (inode->i_ino != MFT_REC_MFT || sync) {
+				err = ni_try_remove_attr_list(ni);
+				if (err)
+					goto out;
+			}
+
+			err = al_update(ni);
+			if (err)
+				goto out;
+		}
+	}
+
+	for (node = rb_first(&ni->mi_tree); node; node = next) {
+		struct mft_inode *mi = rb_entry(node, struct mft_inode, node);
+		bool is_empty;
+
+		next = rb_next(node);
+
+		if (!mi->dirty)
+			continue;
+
+		is_empty = !mi_enum_attr(mi, NULL);
+
+		if (is_empty)
+			clear_rec_inuse(mi->mrec);
+
+		err2 = mi_write(mi, sync);
+		if (!err && err2)
+			err = err2;
+
+		if (is_empty) {
+			ntfs_mark_rec_free(sbi, mi->rno);
+			rb_erase(node, &ni->mi_tree);
+			mi_put(mi);
+		}
+	}
+
+	if (ni->mi.dirty) {
+		err2 = mi_write(&ni->mi, sync);
+		if (!err && err2)
+			err = err2;
+	}
+out:
+	ni_unlock(ni);
+
+	if (err) {
+		ntfs_err(sb, "%s r=%lx failed, %d.", hint, inode->i_ino, err);
+		ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
+		return err;
+	}
+
+	if (re_dirty && (sb->s_flags & SB_ACTIVE))
+		mark_inode_dirty_sync(inode);
+
+	return 0;
+}
diff --git a/fs/ntfs3/namei.c b/fs/ntfs3/namei.c
new file mode 100644
index 000000000000..860e65f6d384
--- /dev/null
+++ b/fs/ntfs3/namei.c
@@ -0,0 +1,576 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/iversion.h>
+#include <linux/namei.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+/*
+ * fill_name_de
+ *
+ * formats NTFS_DE in 'buf'
+ */
+int fill_name_de(struct ntfs_sb_info *sbi, void *buf, const struct qstr *name,
+		 const struct cpu_str *uni)
+{
+	int err;
+	struct NTFS_DE *e = buf;
+	u16 data_size;
+	struct ATTR_FILE_NAME *fname = (struct ATTR_FILE_NAME *)(e + 1);
+
+#ifndef NTFS3_64BIT_CLUSTER
+	e->ref.high = fname->home.high = 0;
+#endif
+	if (uni) {
+#ifdef __BIG_ENDIAN
+		int ulen = uni->len;
+		__le16 *uname = fname->name;
+		const u16 *name_cpu = uni->name;
+
+		while (ulen--)
+			*uname++ = cpu_to_le16(*name_cpu++);
+#else
+		memcpy(fname->name, uni->name, uni->len * sizeof(u16));
+#endif
+		fname->name_len = uni->len;
+
+	} else {
+		/* Convert input string to unicode */
+		err = ntfs_nls_to_utf16(sbi, name->name, name->len,
+					(struct cpu_str *)&fname->name_len,
+					NTFS_NAME_LEN, UTF16_LITTLE_ENDIAN);
+		if (err < 0)
+			return err;
+	}
+
+	fname->type = FILE_NAME_POSIX;
+	data_size = fname_full_size(fname);
+
+	e->size = cpu_to_le16(QuadAlign(data_size) + sizeof(struct NTFS_DE));
+	e->key_size = cpu_to_le16(data_size);
+	e->flags = 0;
+	e->res = 0;
+
+	return 0;
+}
+
+/*
+ * ntfs_lookup
+ *
+ * inode_operations::lookup
+ */
+static struct dentry *ntfs_lookup(struct inode *dir, struct dentry *dentry,
+				  u32 flags)
+{
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct cpu_str *uni = __getname();
+	struct inode *inode;
+	int err;
+
+	if (!uni)
+		inode = ERR_PTR(-ENOMEM);
+	else {
+		err = ntfs_nls_to_utf16(ni->mi.sbi, dentry->d_name.name,
+					dentry->d_name.len, uni, NTFS_NAME_LEN,
+					UTF16_HOST_ENDIAN);
+		if (err < 0)
+			inode = ERR_PTR(err);
+		else {
+			ni_lock(ni);
+			inode = dir_search_u(dir, uni, NULL);
+			ni_unlock(ni);
+		}
+		__putname(uni);
+	}
+
+	return d_splice_alias(inode, dentry);
+}
+
+/*
+ * ntfs_create
+ *
+ * inode_operations::create
+ */
+static int ntfs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
+		       bool excl)
+{
+	int err;
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct inode *inode;
+
+	ni_lock(ni);
+
+	err = ntfs_create_inode(dir, dentry, NULL, S_IFREG | mode, 0, NULL, 0,
+				excl, NULL, &inode);
+
+	ni_unlock(ni);
+
+	return err;
+}
+
+/*
+ * ntfs_link
+ *
+ * inode_operations::link
+ */
+static int ntfs_link(struct dentry *ode, struct inode *dir, struct dentry *de)
+{
+	int err;
+	struct inode *inode = d_inode(ode);
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	if (S_ISDIR(inode->i_mode))
+		return -EPERM;
+
+	if (inode->i_nlink >= NTFS_LINK_MAX)
+		return -EMLINK;
+
+	ni_lock(ni);
+	if (inode != dir)
+		ni_lock(ntfs_i(dir));
+
+	dir->i_ctime = dir->i_mtime = inode->i_ctime = current_time(inode);
+	inc_nlink(inode);
+	ihold(inode);
+
+	err = ntfs_link_inode(inode, de);
+	if (!err) {
+		mark_inode_dirty(inode);
+		mark_inode_dirty(dir);
+		d_instantiate(de, inode);
+	} else {
+		drop_nlink(inode);
+		iput(inode);
+	}
+
+	if (inode != dir)
+		ni_unlock(ntfs_i(dir));
+	ni_unlock(ni);
+
+	return err;
+}
+
+/*
+ * ntfs_unlink
+ *
+ * inode_operations::unlink
+ */
+static int ntfs_unlink(struct inode *dir, struct dentry *dentry)
+{
+	struct ntfs_inode *ni = ntfs_i(dir);
+	int err;
+
+	ni_lock(ni);
+
+	err = ntfs_unlink_inode(dir, dentry);
+
+	ni_unlock(ni);
+
+	return err;
+}
+
+/*
+ * ntfs_symlink
+ *
+ * inode_operations::symlink
+ */
+static int ntfs_symlink(struct inode *dir, struct dentry *dentry,
+			const char *symname)
+{
+	int err;
+	u32 size = strlen(symname);
+	struct inode *inode;
+	struct ntfs_inode *ni = ntfs_i(dir);
+
+	ni_lock(ni);
+
+	err = ntfs_create_inode(dir, dentry, NULL, S_IFLNK | 0777, 0, symname,
+				size, 0, NULL, &inode);
+
+	ni_unlock(ni);
+
+	return err;
+}
+
+/*
+ * ntfs_mkdir
+ *
+ * inode_operations::mkdir
+ */
+static int ntfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
+{
+	int err;
+	struct inode *inode;
+	struct ntfs_inode *ni = ntfs_i(dir);
+
+	ni_lock(ni);
+
+	err = ntfs_create_inode(dir, dentry, NULL, S_IFDIR | mode, 0, NULL, -1,
+				0, NULL, &inode);
+
+	ni_unlock(ni);
+
+	return err;
+}
+
+/*
+ * ntfs_rmdir
+ *
+ * inode_operations::rm_dir
+ */
+static int ntfs_rmdir(struct inode *dir, struct dentry *dentry)
+{
+	struct ntfs_inode *ni = ntfs_i(dir);
+	int err;
+
+	ni_lock(ni);
+
+	err = ntfs_unlink_inode(dir, dentry);
+
+	ni_unlock(ni);
+
+	return err;
+}
+
+/*
+ * ntfs_rename
+ *
+ * inode_operations::rename
+ */
+static int ntfs_rename(struct inode *old_dir, struct dentry *old_dentry,
+		       struct inode *new_dir, struct dentry *new_dentry,
+		       u32 flags)
+{
+	int err;
+	struct super_block *sb = old_dir->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	struct ntfs_inode *old_diri = ntfs_i(old_dir);
+	struct ntfs_inode *new_diri = ntfs_i(new_dir);
+	struct ntfs_inode *ni;
+	struct ATTR_FILE_NAME *old_name, *new_name, *fname;
+	u8 name_type;
+	bool is_same;
+	struct inode *old_inode, *new_inode;
+	struct NTFS_DE *old_de, *new_de;
+	struct ATTRIB *attr;
+	struct ATTR_LIST_ENTRY *le;
+	u16 new_de_key_size;
+
+	static_assert(SIZEOF_ATTRIBUTE_FILENAME_MAX + SIZEOF_RESIDENT < 1024);
+	static_assert(SIZEOF_ATTRIBUTE_FILENAME_MAX + sizeof(struct NTFS_DE) <
+		      1024);
+	static_assert(PATH_MAX >= 4 * 1024);
+
+	if (flags & ~RENAME_NOREPLACE)
+		return -EINVAL;
+
+	old_inode = d_inode(old_dentry);
+	new_inode = d_inode(new_dentry);
+
+	ni = ntfs_i(old_inode);
+
+	ni_lock(ni);
+	ni_lock(old_diri);
+
+	is_same = old_dentry->d_name.len == new_dentry->d_name.len &&
+		  !memcmp(old_dentry->d_name.name, new_dentry->d_name.name,
+			  old_dentry->d_name.len);
+
+	if (is_same && old_dir == new_dir) {
+		/* Nothing to do */
+		err = 0;
+		goto out1;
+	}
+
+	if (ntfs_is_meta_file(sbi, old_inode->i_ino)) {
+		err = -EINVAL;
+		goto out1;
+	}
+
+	if (new_inode) {
+		/*target name exists. unlink it*/
+		dget(new_dentry);
+		if (old_diri != new_diri)
+			ni_lock(new_diri);
+		err = ntfs_unlink_inode(new_dir, new_dentry);
+		if (old_diri != new_diri)
+			ni_unlock(new_diri);
+
+		dput(new_dentry);
+		if (err)
+			goto out1;
+	}
+
+	/* allocate PATH_MAX bytes */
+	old_de = __getname();
+	if (!old_de) {
+		err = -ENOMEM;
+		goto out1;
+	}
+
+	err = fill_name_de(sbi, old_de, &old_dentry->d_name, NULL);
+	if (err < 0)
+		goto out2;
+
+	old_name = (struct ATTR_FILE_NAME *)(old_de + 1);
+
+	if (is_same) {
+		new_de = old_de;
+	} else {
+		new_de = Add2Ptr(old_de, 1024);
+		err = fill_name_de(sbi, new_de, &new_dentry->d_name, NULL);
+		if (err < 0)
+			goto out2;
+	}
+
+	old_name->home.low = cpu_to_le32(old_dir->i_ino);
+#ifdef NTFS3_64BIT_CLUSTER
+	old_name->home.high = cpu_to_le16(old_dir->i_ino >> 32);
+#endif
+	old_name->home.seq = ntfs_i(old_dir)->mi.mrec->seq;
+
+	/*get pointer to file_name in mft*/
+	fname = ni_fname_name(ni, (struct cpu_str *)&old_name->name_len,
+			      &old_name->home, &le);
+	if (!fname) {
+		err = -EINVAL;
+		goto out2;
+	}
+
+	/* Copy fname info from record into new fname */
+	new_name = (struct ATTR_FILE_NAME *)(new_de + 1);
+	memcpy(&new_name->dup, &fname->dup, sizeof(fname->dup));
+
+	name_type = paired_name(fname->type);
+
+	/* remove first name from directory */
+	err = indx_delete_entry(&old_diri->dir, old_diri, old_de + 1,
+				le16_to_cpu(old_de->key_size), sbi);
+	if (err)
+		goto out3;
+
+	/* remove first name from mft */
+	err = ni_remove_attr_le(ni, attr_from_name(fname), le);
+	if (err)
+		goto out4;
+
+	le16_add_cpu(&ni->mi.mrec->hard_links, -1);
+	ni->mi.dirty = true;
+
+	if (name_type != FILE_NAME_POSIX) {
+		/* get paired name */
+		fname = ni_fname_type(ni, name_type, &le);
+		if (fname) {
+			/* remove second name from directory */
+			err = indx_delete_entry(&old_diri->dir, old_diri, fname,
+						fname_full_size(fname), sbi);
+			if (err)
+				goto out5;
+
+			/* remove second name from mft */
+			err = ni_remove_attr_le(ni, attr_from_name(fname), le);
+			if (err)
+				goto out6;
+
+			le16_add_cpu(&ni->mi.mrec->hard_links, -1);
+			ni->mi.dirty = true;
+		}
+	}
+
+	/* Add new name */
+	new_de->ref.low = cpu_to_le32(old_inode->i_ino);
+#ifdef NTFS3_64BIT_CLUSTER
+	new_de->ref.high = cpu_to_le16(old_inode->i_ino >> 32);
+	new_name->home.high = cpu_to_le16(new_dir->i_ino >> 32);
+#endif
+	new_de->ref.seq = ni->mi.mrec->seq;
+
+	new_name->home.low = cpu_to_le32(new_dir->i_ino);
+	new_name->home.seq = ntfs_i(new_dir)->mi.mrec->seq;
+
+	new_de_key_size = le16_to_cpu(new_de->key_size);
+
+	/* insert new name in mft */
+	err = ni_insert_resident(ni, new_de_key_size, ATTR_NAME, NULL, 0, &attr,
+				 NULL);
+	if (err)
+		goto out7;
+
+	attr->res.flags = RESIDENT_FLAG_INDEXED;
+
+	memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), new_name, new_de_key_size);
+
+	le16_add_cpu(&ni->mi.mrec->hard_links, 1);
+	ni->mi.dirty = true;
+
+	/* insert new name in directory */
+	err = indx_insert_entry(&new_diri->dir, new_diri, new_de, sbi, NULL);
+	if (err)
+		goto out8;
+
+	if (IS_DIRSYNC(new_dir))
+		err = ntfs_sync_inode(old_inode);
+	else
+		mark_inode_dirty(old_inode);
+
+	old_dir->i_ctime = old_dir->i_mtime = current_time(old_dir);
+	if (IS_DIRSYNC(old_dir))
+		(void)ntfs_sync_inode(old_dir);
+	else
+		mark_inode_dirty(old_dir);
+
+	if (old_dir != new_dir) {
+		new_dir->i_mtime = new_dir->i_ctime = old_dir->i_ctime;
+		mark_inode_dirty(new_dir);
+	}
+
+	if (old_inode) {
+		old_inode->i_ctime = old_dir->i_ctime;
+		mark_inode_dirty(old_inode);
+	}
+
+	err = 0;
+	goto out2;
+
+out8:
+	mi_remove_attr(&ni->mi, attr);
+
+out7:
+out6:
+out5:
+out4:
+	/* Undo:
+	 *err = indx_delete_entry(&old_diri->dir, old_diri, old_de + 1,
+	 *			old_de->key_size, NULL);
+	 */
+
+out3:
+out2:
+	__putname(old_de);
+out1:
+	ni_unlock(old_diri);
+	ni_unlock(ni);
+
+	return err;
+}
+
+/*
+ * ntfs_atomic_open
+ *
+ * inode_operations::atomic_open
+ */
+static int ntfs_atomic_open(struct inode *dir, struct dentry *dentry,
+			    struct file *file, u32 flags, umode_t mode)
+{
+	int err;
+	bool excl = !!(flags & O_EXCL);
+	struct inode *inode;
+	struct ntfs_fnd *fnd = NULL;
+	struct ntfs_inode *ni = ntfs_i(dir);
+	struct dentry *d = NULL;
+	struct cpu_str *uni = __getname();
+
+	if (!uni)
+		return -ENOMEM;
+
+	err = ntfs_nls_to_utf16(ni->mi.sbi, dentry->d_name.name,
+				dentry->d_name.len, uni, NTFS_NAME_LEN,
+				UTF16_HOST_ENDIAN);
+	if (err < 0)
+		goto out;
+
+	ni_lock(ni);
+
+	if (d_in_lookup(dentry)) {
+		fnd = fnd_get(&ntfs_i(dir)->dir);
+		if (!fnd) {
+			err = -ENOMEM;
+			goto out1;
+		}
+
+		d = d_splice_alias(dir_search_u(dir, uni, fnd), dentry);
+		if (IS_ERR(d)) {
+			err = PTR_ERR(d);
+			d = NULL;
+			goto out2;
+		}
+
+		if (d)
+			dentry = d;
+	}
+
+	if (!(flags & O_CREAT) || d_really_is_positive(dentry)) {
+		err = finish_no_open(file, d);
+		goto out2;
+	}
+
+	file->f_mode |= FMODE_CREATED;
+
+	/*fnd contains tree's path to insert to*/
+	err = ntfs_create_inode(dir, dentry, uni, mode, 0, NULL, 0, excl, fnd,
+				&inode);
+	if (!err)
+		err = finish_open(file, dentry, ntfs_file_open);
+	dput(d);
+
+out2:
+	fnd_put(fnd);
+out1:
+	ni_unlock(ni);
+out:
+	__putname(uni);
+
+	return err;
+}
+
+struct dentry *ntfs_get_parent(struct dentry *child)
+{
+	struct inode *inode = d_inode(child);
+	struct ntfs_inode *ni = ntfs_i(inode);
+
+	struct ATTR_LIST_ENTRY *le = NULL;
+	struct ATTRIB *attr = NULL;
+	struct ATTR_FILE_NAME *fname;
+
+	while ((attr = ni_find_attr(ni, attr, &le, ATTR_NAME, NULL, 0, NULL,
+				    NULL))) {
+		fname = resident_data_ex(attr, SIZEOF_ATTRIBUTE_FILENAME);
+		if (!fname)
+			continue;
+
+		return d_obtain_alias(
+			ntfs_iget5(inode->i_sb, &fname->home, NULL));
+	}
+
+	return ERR_PTR(-ENOENT);
+}
+
+const struct inode_operations ntfs_dir_inode_operations = {
+	.lookup = ntfs_lookup,
+	.create = ntfs_create,
+	.link = ntfs_link,
+	.unlink = ntfs_unlink,
+	.symlink = ntfs_symlink,
+	.mkdir = ntfs_mkdir,
+	.rmdir = ntfs_rmdir,
+	.rename = ntfs_rename,
+	.permission = ntfs_permission,
+	.get_acl = ntfs_get_acl,
+	.set_acl = ntfs_set_acl,
+	.setattr = ntfs_setattr,
+	.getattr = ntfs_getattr,
+	.listxattr = ntfs_listxattr,
+	.atomic_open = ntfs_atomic_open,
+	.fiemap = ntfs_fiemap,
+};
diff --git a/fs/ntfs3/record.c b/fs/ntfs3/record.c
new file mode 100644
index 000000000000..f79756a955be
--- /dev/null
+++ b/fs/ntfs3/record.c
@@ -0,0 +1,613 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+static inline int compare_attr(const struct ATTRIB *left, enum ATTR_TYPE type,
+			       const __le16 *name, u8 name_len,
+			       const u16 *upcase)
+{
+	/* First, compare the type codes: */
+	int diff = le32_to_cpu(left->type) - le32_to_cpu(type);
+
+	if (diff)
+		return diff;
+
+	/*
+	 * They have the same type code, so we have to compare the names.
+	 * First compare case insensitive
+	 */
+	diff = ntfs_cmp_names(attr_name(left), left->name_len, name, name_len,
+			      upcase);
+	if (diff)
+		return diff;
+
+	/* Second compare case sensitive */
+	return ntfs_cmp_names(attr_name(left), left->name_len, name, name_len,
+			      NULL);
+}
+
+/*
+ * mi_new_attt_id
+ *
+ * returns unused attribute id that is less than mrec->next_attr_id
+ */
+static __le16 mi_new_attt_id(struct mft_inode *mi)
+{
+	u16 free_id, max_id, t16;
+	struct MFT_REC *rec = mi->mrec;
+	struct ATTRIB *attr;
+	__le16 id;
+
+	id = rec->next_attr_id;
+	free_id = le16_to_cpu(id);
+	if (free_id < 0x7FFF) {
+		rec->next_attr_id = cpu_to_le16(free_id + 1);
+		return id;
+	}
+
+	/* One record can store up to 1024/24 ~= 42 attributes */
+	free_id = 0;
+	max_id = 0;
+
+	attr = NULL;
+
+	for (;;) {
+		attr = mi_enum_attr(mi, attr);
+		if (!attr) {
+			rec->next_attr_id = cpu_to_le16(max_id + 1);
+			mi->dirty = true;
+			return cpu_to_le16(free_id);
+		}
+
+		t16 = le16_to_cpu(attr->id);
+		if (t16 == free_id) {
+			free_id += 1;
+			attr = NULL;
+		} else if (max_id < t16)
+			max_id = t16;
+	}
+}
+
+int mi_get(struct ntfs_sb_info *sbi, CLST rno, struct mft_inode **mi)
+{
+	int err;
+	struct mft_inode *m = ntfs_alloc(sizeof(struct mft_inode), 1);
+
+	if (!m)
+		return -ENOMEM;
+
+	err = mi_init(m, sbi, rno);
+	if (!err)
+		err = mi_read(m, false);
+
+	if (err) {
+		mi_put(m);
+		return err;
+	}
+
+	*mi = m;
+	return 0;
+}
+
+void mi_put(struct mft_inode *mi)
+{
+	mi_clear(mi);
+	ntfs_free(mi);
+}
+
+int mi_init(struct mft_inode *mi, struct ntfs_sb_info *sbi, CLST rno)
+{
+	mi->sbi = sbi;
+	mi->rno = rno;
+	mi->mrec = ntfs_alloc(sbi->record_size, 0);
+	if (!mi->mrec)
+		return -ENOMEM;
+
+	return 0;
+}
+
+/*
+ * mi_read
+ *
+ * reads MFT data
+ */
+int mi_read(struct mft_inode *mi, bool is_mft)
+{
+	int err;
+	struct MFT_REC *rec = mi->mrec;
+	struct ntfs_sb_info *sbi = mi->sbi;
+	u32 bpr = sbi->record_size;
+	u64 vbo = (u64)mi->rno << sbi->record_bits;
+	struct ntfs_inode *mft_ni = sbi->mft.ni;
+	struct runs_tree *run = mft_ni ? &mft_ni->file.run : NULL;
+	struct rw_semaphore *rw_lock = NULL;
+
+	if (is_mounted(sbi)) {
+		if (!is_mft) {
+			rw_lock = &mft_ni->file.run_lock;
+			down_read(rw_lock);
+		}
+	}
+
+	err = ntfs_read_bh(sbi, run, vbo, &rec->rhdr, bpr, &mi->nb);
+	if (rw_lock)
+		up_read(rw_lock);
+	if (!err)
+		goto ok;
+
+	if (err == -E_NTFS_FIXUP) {
+		mi->dirty = true;
+		goto ok;
+	}
+
+	if (err != -ENOENT)
+		goto out;
+
+	if (rw_lock) {
+		ni_lock(mft_ni);
+		down_write(rw_lock);
+	}
+	err = attr_load_runs_vcn(mft_ni, ATTR_DATA, NULL, 0, &mft_ni->file.run,
+				 vbo >> sbi->cluster_bits);
+	if (rw_lock) {
+		up_write(rw_lock);
+		ni_unlock(mft_ni);
+	}
+	if (err)
+		goto out;
+
+	if (rw_lock)
+		down_read(rw_lock);
+	err = ntfs_read_bh(sbi, run, vbo, &rec->rhdr, bpr, &mi->nb);
+	if (rw_lock)
+		up_read(rw_lock);
+
+	if (err == -E_NTFS_FIXUP) {
+		mi->dirty = true;
+		goto ok;
+	}
+	if (err)
+		goto out;
+
+ok:
+	/* check field 'total' only here */
+	if (le32_to_cpu(rec->total) != bpr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	return 0;
+
+out:
+	return err;
+}
+
+struct ATTRIB *mi_enum_attr(struct mft_inode *mi, struct ATTRIB *attr)
+{
+	const struct MFT_REC *rec = mi->mrec;
+	u32 used = le32_to_cpu(rec->used);
+	u32 t32, off, asize;
+	u16 t16;
+
+	if (!attr) {
+		u32 total = le32_to_cpu(rec->total);
+
+		off = le16_to_cpu(rec->attr_off);
+
+		if (used > total)
+			goto out;
+
+		if (off >= used || off < MFTRECORD_FIXUP_OFFSET_1 ||
+		    !IsDwordAligned(off)) {
+			goto out;
+		}
+
+		/* Skip non-resident records */
+		if (!is_rec_inuse(rec))
+			goto out;
+
+		attr = Add2Ptr(rec, off);
+	} else {
+		/* Check if input attr inside record */
+		off = PtrOffset(rec, attr);
+		if (off >= used)
+			goto out;
+
+		asize = le32_to_cpu(attr->size);
+		if (asize < SIZEOF_RESIDENT)
+			goto out;
+
+		attr = Add2Ptr(attr, asize);
+		off += asize;
+	}
+
+	asize = le32_to_cpu(attr->size);
+
+	/* Can we use the first field (attr->type) */
+	if (off + 8 > used) {
+		static_assert(QuadAlign(sizeof(enum ATTR_TYPE)) == 8);
+		goto out;
+	}
+
+	if (attr->type == ATTR_END) {
+		if (used != off + 8)
+			goto out;
+		return NULL;
+	}
+
+	t32 = le32_to_cpu(attr->type);
+	if ((t32 & 0xf) || (t32 > 0x100))
+		goto out;
+
+	/* Check boundary */
+	if (off + asize > used)
+		goto out;
+
+	/* Check size of attribute */
+	if (!attr->non_res) {
+		if (asize < SIZEOF_RESIDENT)
+			goto out;
+
+		t16 = le16_to_cpu(attr->res.data_off);
+
+		if (t16 > asize)
+			goto out;
+
+		t32 = le32_to_cpu(attr->res.data_size);
+		if (t16 + t32 > asize)
+			goto out;
+
+		return attr;
+	}
+
+	/* Check some nonresident fields */
+	if (attr->name_len &&
+	    le16_to_cpu(attr->name_off) + sizeof(short) * attr->name_len >
+		    le16_to_cpu(attr->nres.run_off)) {
+		goto out;
+	}
+
+	if (attr->nres.svcn || !is_attr_ext(attr)) {
+		if (asize + 8 < SIZEOF_NONRESIDENT)
+			goto out;
+
+		if (attr->nres.c_unit)
+			goto out;
+	} else if (asize + 8 < SIZEOF_NONRESIDENT_EX)
+		goto out;
+
+	return attr;
+
+out:
+	return NULL;
+}
+
+/*
+ * mi_find_attr
+ *
+ * finds the attribute by type and name and id
+ */
+struct ATTRIB *mi_find_attr(struct mft_inode *mi, struct ATTRIB *attr,
+			    enum ATTR_TYPE type, const __le16 *name,
+			    size_t name_len, const __le16 *id)
+{
+	u32 type_in = le32_to_cpu(type);
+	u32 atype;
+
+next_attr:
+	attr = mi_enum_attr(mi, attr);
+	if (!attr)
+		return NULL;
+
+	atype = le32_to_cpu(attr->type);
+	if (atype > type_in)
+		return NULL;
+
+	if (atype < type_in)
+		goto next_attr;
+
+	if (attr->name_len != name_len)
+		goto next_attr;
+
+	if (name_len && memcmp(attr_name(attr), name, name_len * sizeof(short)))
+		goto next_attr;
+
+	if (id && *id != attr->id)
+		goto next_attr;
+
+	return attr;
+}
+
+int mi_write(struct mft_inode *mi, int wait)
+{
+	struct MFT_REC *rec;
+	int err;
+	struct ntfs_sb_info *sbi;
+
+	if (!mi->dirty)
+		return 0;
+
+	sbi = mi->sbi;
+	rec = mi->mrec;
+
+	err = ntfs_write_bh(sbi, &rec->rhdr, &mi->nb, wait);
+	if (err)
+		return err;
+
+	if (mi->rno < sbi->mft.recs_mirr)
+		sbi->flags |= NTFS_FLAGS_MFTMIRR;
+
+	mi->dirty = false;
+
+	return 0;
+}
+
+int mi_format_new(struct mft_inode *mi, struct ntfs_sb_info *sbi, CLST rno,
+		  __le16 flags, bool is_mft)
+{
+	int err;
+	u16 seq = 1;
+	struct MFT_REC *rec;
+	u64 vbo = (u64)rno << sbi->record_bits;
+
+	err = mi_init(mi, sbi, rno);
+	if (err)
+		return err;
+
+	rec = mi->mrec;
+
+	if (rno == MFT_REC_MFT) {
+		;
+	} else if (rno < MFT_REC_FREE) {
+		seq = rno;
+	} else if (rno >= sbi->mft.used) {
+		;
+	} else if (mi_read(mi, is_mft)) {
+		;
+	} else if (rec->rhdr.sign == NTFS_FILE_SIGNATURE) {
+		/* Record is reused. Update its sequence number */
+		seq = le16_to_cpu(rec->seq) + 1;
+		if (!seq)
+			seq = 1;
+	}
+
+	memcpy(rec, sbi->new_rec, sbi->record_size);
+
+	rec->seq = cpu_to_le16(seq);
+	rec->flags = RECORD_FLAG_IN_USE | flags;
+
+	mi->dirty = true;
+
+	if (!mi->nb.nbufs) {
+		struct ntfs_inode *ni = sbi->mft.ni;
+		bool lock = false;
+
+		if (is_mounted(sbi) && !is_mft) {
+			down_read(&ni->file.run_lock);
+			lock = true;
+		}
+
+		err = ntfs_get_bh(sbi, &ni->file.run, vbo, sbi->record_size,
+				  &mi->nb);
+		if (lock)
+			up_read(&ni->file.run_lock);
+	}
+
+	return err;
+}
+
+/*
+ * mi_mark_free
+ *
+ * marks record as unused and marks it as free in bitmap
+ */
+void mi_mark_free(struct mft_inode *mi)
+{
+	CLST rno = mi->rno;
+	struct ntfs_sb_info *sbi = mi->sbi;
+
+	if (rno >= MFT_REC_RESERVED && rno < MFT_REC_FREE) {
+		ntfs_clear_mft_tail(sbi, rno, rno + 1);
+		mi->dirty = false;
+		return;
+	}
+
+	if (mi->mrec) {
+		clear_rec_inuse(mi->mrec);
+		mi->dirty = true;
+		mi_write(mi, 0);
+	}
+	ntfs_mark_rec_free(sbi, rno);
+}
+
+/*
+ * mi_insert_attr
+ *
+ * reserves space for new attribute
+ * returns not full constructed attribute or NULL if not possible to create
+ */
+struct ATTRIB *mi_insert_attr(struct mft_inode *mi, enum ATTR_TYPE type,
+			      const __le16 *name, u8 name_len, u32 asize,
+			      u16 name_off)
+{
+	size_t tail;
+	struct ATTRIB *attr;
+	__le16 id;
+	struct MFT_REC *rec = mi->mrec;
+	struct ntfs_sb_info *sbi = mi->sbi;
+	u32 used = le32_to_cpu(rec->used);
+	const u16 *upcase = sbi->upcase;
+	int diff;
+
+	/* Can we insert mi attribute? */
+	if (used + asize > mi->sbi->record_size)
+		return NULL;
+
+	/*
+	 * Scan through the list of attributes to find the point
+	 * at which we should insert it.
+	 */
+	attr = NULL;
+	while ((attr = mi_enum_attr(mi, attr))) {
+		diff = compare_attr(attr, type, name, name_len, upcase);
+		if (diff > 0)
+			break;
+		if (diff < 0)
+			continue;
+
+		if (!is_attr_indexed(attr))
+			return NULL;
+		break;
+	}
+
+	if (!attr) {
+		tail = 8; /* not used, just to suppress warning */
+		attr = Add2Ptr(rec, used - 8);
+	} else {
+		tail = used - PtrOffset(rec, attr);
+	}
+
+	id = mi_new_attt_id(mi);
+
+	memmove(Add2Ptr(attr, asize), attr, tail);
+	memset(attr, 0, asize);
+
+	attr->type = type;
+	attr->size = cpu_to_le32(asize);
+	attr->name_len = name_len;
+	attr->name_off = cpu_to_le16(name_off);
+	attr->id = id;
+
+	memmove(Add2Ptr(attr, name_off), name, name_len * sizeof(short));
+	rec->used = cpu_to_le32(used + asize);
+
+	mi->dirty = true;
+
+	return attr;
+}
+
+/*
+ * mi_remove_attr
+ *
+ * removes the attribute from record
+ * NOTE: The source attr will point to next attribute
+ */
+bool mi_remove_attr(struct mft_inode *mi, struct ATTRIB *attr)
+{
+	struct MFT_REC *rec = mi->mrec;
+	u32 aoff = PtrOffset(rec, attr);
+	u32 used = le32_to_cpu(rec->used);
+	u32 asize = le32_to_cpu(attr->size);
+
+	if (aoff + asize > used)
+		return false;
+
+	used -= asize;
+	memmove(attr, Add2Ptr(attr, asize), used - aoff);
+	rec->used = cpu_to_le32(used);
+	mi->dirty = true;
+
+	return true;
+}
+
+bool mi_resize_attr(struct mft_inode *mi, struct ATTRIB *attr, int bytes)
+{
+	struct MFT_REC *rec = mi->mrec;
+	u32 aoff = PtrOffset(rec, attr);
+	u32 total, used = le32_to_cpu(rec->used);
+	u32 nsize, asize = le32_to_cpu(attr->size);
+	u32 rsize = le32_to_cpu(attr->res.data_size);
+	int tail = (int)(used - aoff - asize);
+	int dsize;
+	char *next;
+
+	if (tail < 0 || aoff >= used)
+		return false;
+
+	if (!bytes)
+		return true;
+
+	total = le32_to_cpu(rec->total);
+	next = Add2Ptr(attr, asize);
+
+	if (bytes > 0) {
+		dsize = QuadAlign(bytes);
+		if (used + dsize > total)
+			return false;
+		nsize = asize + dsize;
+		// move tail
+		memmove(next + dsize, next, tail);
+		memset(next, 0, dsize);
+		used += dsize;
+		rsize += dsize;
+	} else {
+		dsize = QuadAlign(-bytes);
+		if (dsize > asize)
+			return false;
+		nsize = asize - dsize;
+		memmove(next - dsize, next, tail);
+		used -= dsize;
+		rsize -= dsize;
+	}
+
+	rec->used = cpu_to_le32(used);
+	attr->size = cpu_to_le32(nsize);
+	if (!attr->non_res)
+		attr->res.data_size = cpu_to_le32(rsize);
+	mi->dirty = true;
+
+	return true;
+}
+
+int mi_pack_runs(struct mft_inode *mi, struct ATTRIB *attr,
+		 struct runs_tree *run, CLST len)
+{
+	int err = 0;
+	struct ntfs_sb_info *sbi = mi->sbi;
+	u32 new_run_size;
+	CLST plen;
+	struct MFT_REC *rec = mi->mrec;
+	CLST svcn = le64_to_cpu(attr->nres.svcn);
+	u32 used = le32_to_cpu(rec->used);
+	u32 aoff = PtrOffset(rec, attr);
+	u32 asize = le32_to_cpu(attr->size);
+	char *next = Add2Ptr(attr, asize);
+	u16 run_off = le16_to_cpu(attr->nres.run_off);
+	u32 run_size = asize - run_off;
+	u32 tail = used - aoff - asize;
+	u32 dsize = sbi->record_size - used;
+
+	/* Make a maximum gap in current record */
+	memmove(next + dsize, next, tail);
+
+	/* Pack as much as possible */
+	err = run_pack(run, svcn, len, Add2Ptr(attr, run_off), run_size + dsize,
+		       &plen);
+	if (err < 0) {
+		memmove(next, next + dsize, tail);
+		return err;
+	}
+
+	new_run_size = QuadAlign(err);
+
+	memmove(next + new_run_size - run_size, next + dsize, tail);
+
+	attr->size = cpu_to_le32(asize + new_run_size - run_size);
+	attr->nres.evcn = cpu_to_le64(svcn + plen - 1);
+	rec->used = cpu_to_le32(used + new_run_size - run_size);
+	mi->dirty = true;
+
+	return 0;
+}
diff --git a/fs/ntfs3/run.c b/fs/ntfs3/run.c
new file mode 100644
index 000000000000..85321249e592
--- /dev/null
+++ b/fs/ntfs3/run.c
@@ -0,0 +1,1190 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+/* runs_tree is a continues memory. Try to avoid big size  */
+#define NTFS3_RUN_MAX_BYTES 0x10000
+
+struct ntfs_run {
+	CLST vcn; /* virtual cluster number */
+	CLST len; /* length in clusters */
+	CLST lcn; /* logical cluster number */
+};
+
+/*
+ * run_lookup
+ *
+ * Lookup the index of a MCB entry that is first <= vcn.
+ * case of success it will return non-zero value and set
+ * 'index' parameter to index of entry been found.
+ * case of entry missing from list 'index' will be set to
+ * point to insertion position for the entry question.
+ */
+bool run_lookup(const struct runs_tree *run, CLST vcn, size_t *index)
+{
+	size_t min_idx, max_idx, mid_idx;
+	struct ntfs_run *r;
+
+	if (!run->count) {
+		*index = 0;
+		return false;
+	}
+
+	min_idx = 0;
+	max_idx = run->count - 1;
+
+	/* Check boundary cases specially, 'cause they cover the often requests */
+	r = run->runs_;
+	if (vcn < r->vcn) {
+		*index = 0;
+		return false;
+	}
+
+	if (vcn < r->vcn + r->len) {
+		*index = 0;
+		return true;
+	}
+
+	r += max_idx;
+	if (vcn >= r->vcn + r->len) {
+		*index = run->count;
+		return false;
+	}
+
+	if (vcn >= r->vcn) {
+		*index = max_idx;
+		return true;
+	}
+
+	do {
+		mid_idx = min_idx + ((max_idx - min_idx) >> 1);
+		r = run->runs_ + mid_idx;
+
+		if (vcn < r->vcn) {
+			max_idx = mid_idx - 1;
+			if (!mid_idx)
+				break;
+		} else if (vcn >= r->vcn + r->len) {
+			min_idx = mid_idx + 1;
+		} else {
+			*index = mid_idx;
+			return true;
+		}
+	} while (min_idx <= max_idx);
+
+	*index = max_idx + 1;
+	return false;
+}
+
+/*
+ * run_consolidate
+ *
+ * consolidate runs starting from a given one.
+ */
+static void run_consolidate(struct runs_tree *run, size_t index)
+{
+	size_t i;
+	struct ntfs_run *r = run->runs_ + index;
+
+	while (index + 1 < run->count) {
+		/*
+		 * I should merge current run with next
+		 * if start of the next run lies inside one being tested.
+		 */
+		struct ntfs_run *n = r + 1;
+		CLST end = r->vcn + r->len;
+		CLST dl;
+
+		/* Stop if runs are not aligned one to another. */
+		if (n->vcn > end)
+			break;
+
+		dl = end - n->vcn;
+
+		/*
+		 * If range at index overlaps with next one
+		 * then I will either adjust it's start position
+		 * or (if completely matches) dust remove one from the list.
+		 */
+		if (dl > 0) {
+			if (n->len <= dl)
+				goto remove_next_range;
+
+			n->len -= dl;
+			n->vcn += dl;
+			if (n->lcn != SPARSE_LCN)
+				n->lcn += dl;
+			dl = 0;
+		}
+
+		/*
+		 * Stop if sparse mode does not match
+		 * both current and next runs.
+		 */
+		if ((n->lcn == SPARSE_LCN) != (r->lcn == SPARSE_LCN)) {
+			index += 1;
+			r = n;
+			continue;
+		}
+
+		/*
+		 * Check if volume block
+		 * of a next run lcn does not match
+		 * last volume block of the current run.
+		 */
+		if (n->lcn != SPARSE_LCN && n->lcn != r->lcn + r->len)
+			break;
+
+		/*
+		 * Next and current are siblings.
+		 * Eat/join.
+		 */
+		r->len += n->len - dl;
+
+remove_next_range:
+		i = run->count - (index + 1);
+		if (i > 1)
+			memmove(n, n + 1, sizeof(*n) * (i - 1));
+
+		run->count -= 1;
+	}
+}
+
+/* returns true if range [svcn - evcn] is mapped*/
+bool run_is_mapped_full(const struct runs_tree *run, CLST svcn, CLST evcn)
+{
+	size_t i;
+	const struct ntfs_run *r, *end;
+	CLST next_vcn;
+
+	if (!run_lookup(run, svcn, &i))
+		return false;
+
+	end = run->runs_ + run->count;
+	r = run->runs_ + i;
+
+	for (;;) {
+		next_vcn = r->vcn + r->len;
+		if (next_vcn > evcn)
+			return true;
+
+		if (++r >= end)
+			return false;
+
+		if (r->vcn != next_vcn)
+			return false;
+	}
+}
+
+bool run_lookup_entry(const struct runs_tree *run, CLST vcn, CLST *lcn,
+		      CLST *len, size_t *index)
+{
+	size_t idx;
+	CLST gap;
+	struct ntfs_run *r;
+
+	/* Fail immediately if nrun was not touched yet. */
+	if (!run->runs_)
+		return false;
+
+	if (!run_lookup(run, vcn, &idx))
+		return false;
+
+	r = run->runs_ + idx;
+
+	if (vcn >= r->vcn + r->len)
+		return false;
+
+	gap = vcn - r->vcn;
+	if (r->len <= gap)
+		return false;
+
+	*lcn = r->lcn == SPARSE_LCN ? SPARSE_LCN : (r->lcn + gap);
+
+	if (len)
+		*len = r->len - gap;
+	if (index)
+		*index = idx;
+
+	return true;
+}
+
+/*
+ * run_truncate_head
+ *
+ * decommit the range before vcn
+ */
+void run_truncate_head(struct runs_tree *run, CLST vcn)
+{
+	size_t index;
+	struct ntfs_run *r;
+
+	if (run_lookup(run, vcn, &index)) {
+		r = run->runs_ + index;
+
+		if (vcn > r->vcn) {
+			CLST dlen = vcn - r->vcn;
+
+			r->vcn = vcn;
+			r->len -= dlen;
+			if (r->lcn != SPARSE_LCN)
+				r->lcn += dlen;
+		}
+
+		if (!index)
+			return;
+	}
+	r = run->runs_;
+	memmove(r, r + index, sizeof(*r) * (run->count - index));
+
+	run->count -= index;
+
+	if (!run->count) {
+		ntfs_free(run->runs_);
+		run->runs_ = NULL;
+		run->allocated = 0;
+	}
+}
+
+/*
+ * run_truncate
+ *
+ * decommit the range after vcn
+ */
+void run_truncate(struct runs_tree *run, CLST vcn)
+{
+	size_t index;
+
+	/*
+	 * If I hit the range then
+	 * I have to truncate one.
+	 * If range to be truncated is becoming empty
+	 * then it will entirely be removed.
+	 */
+	if (run_lookup(run, vcn, &index)) {
+		struct ntfs_run *r = run->runs_ + index;
+
+		r->len = vcn - r->vcn;
+
+		if (r->len > 0)
+			index += 1;
+	}
+
+	/*
+	 * At this point 'index' is set to
+	 * position that should be thrown away (including index itself)
+	 * Simple one - just set the limit.
+	 */
+	run->count = index;
+
+	/* Do not reallocate array 'runs'. Only free if possible */
+	if (!index) {
+		ntfs_free(run->runs_);
+		run->runs_ = NULL;
+		run->allocated = 0;
+	}
+}
+
+/* trim head and tail if necessary*/
+void run_truncate_around(struct runs_tree *run, CLST vcn)
+{
+	run_truncate_head(run, vcn);
+
+	if (run->count >= NTFS3_RUN_MAX_BYTES / sizeof(struct ntfs_run) / 2)
+		run_truncate(run, (run->runs_ + (run->count >> 1))->vcn);
+}
+
+/*
+ * run_add_entry
+ *
+ * sets location to known state.
+ * run to be added may overlap with existing location.
+ * returns false if of memory
+ */
+bool run_add_entry(struct runs_tree *run, CLST vcn, CLST lcn, CLST len)
+{
+	size_t used, index;
+	struct ntfs_run *r;
+	bool inrange;
+	CLST tail_vcn = 0, tail_len = 0, tail_lcn = 0;
+	bool should_add_tail = false;
+
+	/*
+	 * Lookup the insertion point.
+	 *
+	 * Execute bsearch for the entry containing
+	 * start position question.
+	 */
+	inrange = run_lookup(run, vcn, &index);
+
+	/*
+	 * Shortcut here would be case of
+	 * range not been found but one been added
+	 * continues previous run.
+	 * this case I can directly make use of
+	 * existing range as my start point.
+	 */
+	if (!inrange && index > 0) {
+		struct ntfs_run *t = run->runs_ + index - 1;
+
+		if (t->vcn + t->len == vcn &&
+		    (t->lcn == SPARSE_LCN) == (lcn == SPARSE_LCN) &&
+		    (lcn == SPARSE_LCN || lcn == t->lcn + t->len)) {
+			inrange = true;
+			index -= 1;
+		}
+	}
+
+	/*
+	 * At this point 'index' either points to the range
+	 * containing start position or to the insertion position
+	 * for a new range.
+	 * So first let's check if range I'm probing is here already.
+	 */
+	if (!inrange) {
+requires_new_range:
+		/*
+		 * Range was not found.
+		 * Insert at position 'index'
+		 */
+		used = run->count * sizeof(struct ntfs_run);
+
+		/*
+		 * Check allocated space.
+		 * If one is not enough to get one more entry
+		 * then it will be reallocated
+		 */
+		if (run->allocated < used + sizeof(struct ntfs_run)) {
+			size_t bytes;
+			struct ntfs_run *new_ptr;
+
+			/* Use power of 2 for 'bytes'*/
+			if (!used) {
+				bytes = 64;
+			} else if (used <= 16 * PAGE_SIZE) {
+				if (is_power_of2(run->allocated))
+					bytes = run->allocated << 1;
+				else
+					bytes = (size_t)1
+						<< (2 + blksize_bits(used));
+			} else {
+				bytes = run->allocated + (16 * PAGE_SIZE);
+			}
+
+			//BUG_ON(bytes > NTFS3_RUN_MAX_BYTES);
+			WARN_ON(bytes > NTFS3_RUN_MAX_BYTES);
+
+			new_ptr = ntfs_alloc(bytes, 0);
+
+			if (!new_ptr)
+				return false;
+
+			r = new_ptr + index;
+			memcpy(new_ptr, run->runs_,
+			       index * sizeof(struct ntfs_run));
+			memcpy(r + 1, run->runs_ + index,
+			       sizeof(struct ntfs_run) * (run->count - index));
+
+			ntfs_free(run->runs_);
+			run->runs_ = new_ptr;
+			run->allocated = bytes;
+
+		} else {
+			size_t i = run->count - index;
+
+			r = run->runs_ + index;
+
+			/* memmove appears to be a bottle neck here... */
+			if (i > 0)
+				memmove(r + 1, r, sizeof(struct ntfs_run) * i);
+		}
+
+		r->vcn = vcn;
+		r->lcn = lcn;
+		r->len = len;
+		run->count += 1;
+	} else {
+		r = run->runs_ + index;
+
+		/*
+		 * If one of ranges was not allocated
+		 * then I have to split location I just matched.
+		 * and insert current one
+		 * a common case this requires tail to be reinserted
+		 * a recursive call.
+		 */
+		if (((lcn == SPARSE_LCN) != (r->lcn == SPARSE_LCN)) ||
+		    (lcn != SPARSE_LCN && lcn != r->lcn + (vcn - r->vcn))) {
+			CLST to_eat = vcn - r->vcn;
+			CLST Tovcn = to_eat + len;
+
+			should_add_tail = Tovcn < r->len;
+
+			if (should_add_tail) {
+				tail_lcn = r->lcn == SPARSE_LCN ?
+						   SPARSE_LCN :
+						   (r->lcn + Tovcn);
+				tail_vcn = r->vcn + Tovcn;
+				tail_len = r->len - Tovcn;
+			}
+
+			if (to_eat > 0) {
+				r->len = to_eat;
+				inrange = false;
+				index += 1;
+				goto requires_new_range;
+			}
+
+			/* lcn should match one I'm going to add. */
+			r->lcn = lcn;
+		}
+
+		/*
+		 * If existing range fits then I'm done.
+		 * Otherwise extend found one and fall back to range jocode.
+		 */
+		if (r->vcn + r->len < vcn + len)
+			r->len += len - ((r->vcn + r->len) - vcn);
+	}
+
+	/*
+	 * And normalize it starting from insertion point.
+	 * It's possible that no insertion needed case if
+	 * start point lies within the range of an entry
+	 * that 'index' points to.
+	 */
+	if (inrange && index > 0)
+		index -= 1;
+	run_consolidate(run, index);
+	run_consolidate(run, index + 1);
+
+	/*
+	 * a special case
+	 * I have to add extra range a tail.
+	 */
+	if (should_add_tail &&
+	    !run_add_entry(run, tail_vcn, tail_lcn, tail_len))
+		return false;
+
+	return true;
+}
+
+/*
+ * run_get_entry
+ *
+ * returns index-th mapped region
+ */
+bool run_get_entry(const struct runs_tree *run, size_t index, CLST *vcn,
+		   CLST *lcn, CLST *len)
+{
+	const struct ntfs_run *r;
+
+	if (index >= run->count)
+		return false;
+
+	r = run->runs_ + index;
+
+	if (!r->len)
+		return false;
+
+	if (vcn)
+		*vcn = r->vcn;
+	if (lcn)
+		*lcn = r->lcn;
+	if (len)
+		*len = r->len;
+	return true;
+}
+
+/*
+ * run_packed_size
+ *
+ * calculates the size of packed int64
+ */
+static inline int run_packed_size(const s64 *n)
+{
+#ifdef __BIG_ENDIAN
+	const u8 *p = (const u8 *)n + sizeof(*n) - 1;
+
+	if (*n >= 0) {
+		if (p[-7] || p[-6] || p[-5] || p[-4])
+			p -= 4;
+		if (p[-3] || p[-2])
+			p -= 2;
+		if (p[-1])
+			p -= 1;
+		if (p[0] & 0x80)
+			p -= 1;
+	} else {
+		if (p[-7] != 0xff || p[-6] != 0xff || p[-5] != 0xff ||
+		    p[-4] != 0xff)
+			p -= 4;
+		if (p[-3] != 0xff || p[-2] != 0xff)
+			p -= 2;
+		if (p[-1] != 0xff)
+			p -= 1;
+		if (!(p[0] & 0x80))
+			p -= 1;
+	}
+	return (const u8 *)n + sizeof(*n) - p;
+#else
+	const u8 *p = (const u8 *)n;
+
+	if (*n >= 0) {
+		if (p[7] || p[6] || p[5] || p[4])
+			p += 4;
+		if (p[3] || p[2])
+			p += 2;
+		if (p[1])
+			p += 1;
+		if (p[0] & 0x80)
+			p += 1;
+	} else {
+		if (p[7] != 0xff || p[6] != 0xff || p[5] != 0xff ||
+		    p[4] != 0xff)
+			p += 4;
+		if (p[3] != 0xff || p[2] != 0xff)
+			p += 2;
+		if (p[1] != 0xff)
+			p += 1;
+		if (!(p[0] & 0x80))
+			p += 1;
+	}
+
+	return 1 + p - (const u8 *)n;
+#endif
+}
+
+/*
+ * run_pack
+ *
+ * packs runs into buffer
+ * packed_vcns - how much runs we have packed
+ * packed_size - how much bytes we have used run_buf
+ */
+int run_pack(const struct runs_tree *run, CLST svcn, CLST len, u8 *run_buf,
+	     u32 run_buf_size, CLST *packed_vcns)
+{
+	CLST next_vcn, vcn, lcn;
+	CLST prev_lcn = 0;
+	CLST evcn1 = svcn + len;
+	int packed_size = 0;
+	size_t i;
+	bool ok;
+	s64 dlcn, len64;
+	int offset_size, size_size, t;
+	const u8 *p;
+
+	next_vcn = vcn = svcn;
+
+	*packed_vcns = 0;
+
+	if (!len)
+		goto out;
+
+	ok = run_lookup_entry(run, vcn, &lcn, &len, &i);
+
+	if (!ok)
+		goto error;
+
+	if (next_vcn != vcn)
+		goto error;
+
+	for (;;) {
+		/* offset of current fragment relatively to previous fragment */
+		dlcn = 0;
+		next_vcn = vcn + len;
+
+		if (next_vcn > evcn1)
+			len = evcn1 - vcn;
+
+		/*
+		 * mirror of len, but signed, because run_packed_size()
+		 * works with signed int only
+		 */
+		len64 = len;
+
+		/* how much bytes is packed len64 */
+		size_size = run_packed_size(&len64);
+
+		/* offset_size - how much bytes is packed dlcn */
+		if (lcn == SPARSE_LCN) {
+			offset_size = 0;
+		} else {
+			/* NOTE: lcn can be less than prev_lcn! */
+			dlcn = (s64)lcn - prev_lcn;
+			offset_size = run_packed_size(&dlcn);
+			prev_lcn = lcn;
+		}
+
+		t = run_buf_size - packed_size - 2 - offset_size;
+		if (t <= 0)
+			goto out;
+
+		/* can we store this entire run */
+		if (t < size_size)
+			goto out;
+
+		if (run_buf) {
+			p = (u8 *)&len64;
+
+			/* pack run header */
+			run_buf[0] = ((u8)(size_size | (offset_size << 4)));
+			run_buf += 1;
+
+			/* Pack the length of run */
+			switch (size_size) {
+#ifdef __BIG_ENDIAN
+			case 8:
+				run_buf[7] = p[0];
+				fallthrough;
+			case 7:
+				run_buf[6] = p[1];
+				fallthrough;
+			case 6:
+				run_buf[5] = p[2];
+				fallthrough;
+			case 5:
+				run_buf[4] = p[3];
+				fallthrough;
+			case 4:
+				run_buf[3] = p[4];
+				fallthrough;
+			case 3:
+				run_buf[2] = p[5];
+				fallthrough;
+			case 2:
+				run_buf[1] = p[6];
+				fallthrough;
+			case 1:
+				run_buf[0] = p[7];
+#else
+			case 8:
+				run_buf[7] = p[7];
+				fallthrough;
+			case 7:
+				run_buf[6] = p[6];
+				fallthrough;
+			case 6:
+				run_buf[5] = p[5];
+				fallthrough;
+			case 5:
+				run_buf[4] = p[4];
+				fallthrough;
+			case 4:
+				run_buf[3] = p[3];
+				fallthrough;
+			case 3:
+				run_buf[2] = p[2];
+				fallthrough;
+			case 2:
+				run_buf[1] = p[1];
+				fallthrough;
+			case 1:
+				run_buf[0] = p[0];
+#endif
+			}
+
+			run_buf += size_size;
+			p = (u8 *)&dlcn;
+
+			/* Pack the offset from previous lcn */
+			switch (offset_size) {
+#ifdef __BIG_ENDIAN
+			case 8:
+				run_buf[7] = p[0];
+				fallthrough;
+			case 7:
+				run_buf[6] = p[1];
+				fallthrough;
+			case 6:
+				run_buf[5] = p[2];
+				fallthrough;
+			case 5:
+				run_buf[4] = p[3];
+				fallthrough;
+			case 4:
+				run_buf[3] = p[4];
+				fallthrough;
+			case 3:
+				run_buf[2] = p[5];
+				fallthrough;
+			case 2:
+				run_buf[1] = p[6];
+				fallthrough;
+			case 1:
+				run_buf[0] = p[7];
+#else
+			case 8:
+				run_buf[7] = p[7];
+				fallthrough;
+			case 7:
+				run_buf[6] = p[6];
+				fallthrough;
+			case 6:
+				run_buf[5] = p[5];
+				fallthrough;
+			case 5:
+				run_buf[4] = p[4];
+				fallthrough;
+			case 4:
+				run_buf[3] = p[3];
+				fallthrough;
+			case 3:
+				run_buf[2] = p[2];
+				fallthrough;
+			case 2:
+				run_buf[1] = p[1];
+				fallthrough;
+			case 1:
+				run_buf[0] = p[0];
+#endif
+			}
+
+			run_buf += offset_size;
+		}
+
+		packed_size += 1 + offset_size + size_size;
+		*packed_vcns += len;
+
+		if (packed_size + 1 >= run_buf_size || next_vcn >= evcn1)
+			goto out;
+
+		ok = run_get_entry(run, ++i, &vcn, &lcn, &len);
+		if (!ok)
+			goto error;
+
+		if (next_vcn != vcn)
+			goto error;
+	}
+
+out:
+	/* Store last zero */
+	if (run_buf)
+		run_buf[0] = 0;
+
+	return packed_size + 1;
+
+error:
+	return -EOPNOTSUPP;
+}
+
+/*
+ * run_unpack
+ *
+ * unpacks packed runs from "run_buf"
+ * returns error, if negative, or real used bytes
+ */
+int run_unpack(struct runs_tree *run, struct ntfs_sb_info *sbi, CLST ino,
+	       CLST svcn, CLST evcn, CLST vcn, const u8 *run_buf,
+	       u32 run_buf_size)
+{
+	u64 prev_lcn, vcn64, lcn, next_vcn;
+	const u8 *run_last, *run_0;
+
+	/* Check for empty */
+	if (evcn + 1 == svcn)
+		return 0;
+
+	if (evcn < svcn)
+		return -EINVAL;
+
+	run_0 = run_buf;
+	run_last = run_buf + run_buf_size;
+	prev_lcn = 0;
+	vcn64 = svcn;
+
+	/* Read all runs the chain */
+	/* size_size - how much bytes is packed len */
+	while (run_buf < run_last) {
+		/* size_size - how much bytes is packed len */
+		u8 size_size = *run_buf & 0xF;
+		/* offset_size - how much bytes is packed dlcn */
+		u8 offset_size = *run_buf++ >> 4;
+		u64 len = 0;
+		u8 *p = (u8 *)&len;
+		s64 dlcn;
+
+		if (!size_size)
+			break;
+
+		/*
+		 * Unpack runs.
+		 * NOTE: runs are stored little endian order
+		 * "len" is unsigned value, "dlcn" is signed
+		 * Large positive number requires to store 5 bytes
+		 * e.g.: 05 FF 7E FF FF 00 00 00
+		 */
+
+		switch (size_size) {
+		default:
+error:
+			return -EINVAL;
+
+#ifdef __BIG_ENDIAN
+		case 8:
+			p[0] = run_buf[7];
+			fallthrough;
+		case 7:
+			p[1] = run_buf[6];
+			fallthrough;
+		case 6:
+			p[2] = run_buf[5];
+			fallthrough;
+		case 5:
+			p[3] = run_buf[4];
+			fallthrough;
+		case 4:
+			p[4] = run_buf[3];
+			fallthrough;
+		case 3:
+			p[5] = run_buf[2];
+			fallthrough;
+		case 2:
+			p[6] = run_buf[1];
+			fallthrough;
+		case 1:
+			p[7] = run_buf[0];
+#else
+		case 8:
+			p[7] = run_buf[7];
+			fallthrough;
+		case 7:
+			p[6] = run_buf[6];
+			fallthrough;
+		case 6:
+			p[5] = run_buf[5];
+			fallthrough;
+		case 5:
+			p[4] = run_buf[4];
+			fallthrough;
+		case 4:
+			p[3] = run_buf[3];
+			fallthrough;
+		case 3:
+			p[2] = run_buf[2];
+			fallthrough;
+		case 2:
+			p[1] = run_buf[1];
+			fallthrough;
+		case 1:
+			p[0] = run_buf[0];
+#endif
+		}
+
+		/* skip size_size */
+		run_buf += size_size;
+
+		if (!len)
+			goto error;
+
+		if (!offset_size) {
+			lcn = SPARSE_LCN;
+		} else {
+			/* Check sign */
+			dlcn = (run_buf[offset_size - 1] & 0x80) ? (s64)-1 : 0;
+
+			p = (u8 *)&dlcn;
+
+			switch (offset_size) {
+			default:
+				goto error;
+
+#ifdef __BIG_ENDIAN
+			case 8:
+				p[0] = run_buf[7];
+				fallthrough;
+			case 7:
+				p[1] = run_buf[6];
+				fallthrough;
+			case 6:
+				p[2] = run_buf[5];
+				fallthrough;
+			case 5:
+				p[3] = run_buf[4];
+				fallthrough;
+			case 4:
+				p[4] = run_buf[3];
+				fallthrough;
+			case 3:
+				p[5] = run_buf[2];
+				fallthrough;
+			case 2:
+				p[6] = run_buf[1];
+				fallthrough;
+			case 1:
+				p[7] = run_buf[0];
+#else
+			case 8:
+				p[7] = run_buf[7];
+				fallthrough;
+			case 7:
+				p[6] = run_buf[6];
+				fallthrough;
+			case 6:
+				p[5] = run_buf[5];
+				fallthrough;
+			case 5:
+				p[4] = run_buf[4];
+				fallthrough;
+			case 4:
+				p[3] = run_buf[3];
+				fallthrough;
+			case 3:
+				p[2] = run_buf[2];
+				fallthrough;
+			case 2:
+				p[1] = run_buf[1];
+				fallthrough;
+			case 1:
+				p[0] = run_buf[0];
+#endif
+			}
+
+			/* skip offset_size */
+			run_buf += offset_size;
+			lcn = prev_lcn + dlcn;
+			prev_lcn = lcn;
+		}
+
+		next_vcn = vcn64 + len;
+		/* check boundary */
+		if (next_vcn > evcn + 1)
+			goto error;
+
+#ifndef NTFS3_64BIT_CLUSTER
+		if ((vcn64 >> 32)
+		    /* 0xffffffffffffffff is a valid 'lcn' */
+		    || (lcn + 1) > 0x100000000ull || (len >> 32)) {
+			goto error;
+		}
+#endif
+
+		if (!run)
+			; /* called from check_attr(fslog.c) to check run */
+		else if ((size_t)run == 1) {
+			/* called from ni_delete_all to free clusters without storing in run */
+			if (lcn != SPARSE_LCN)
+				mark_as_free_ex(sbi, lcn, len, true);
+		} else if (vcn64 >= vcn) {
+			if (!run_add_entry(run, vcn64, lcn, len))
+				return -ENOMEM;
+		} else if (next_vcn > vcn) {
+			u64 dlen = vcn - vcn64;
+
+			if (!run_add_entry(run, vcn, lcn + dlen, len - dlen))
+				return -ENOMEM;
+		}
+
+		if (lcn != SPARSE_LCN && lcn + len > sbi->used.bitmap.nbits)
+			return -EINVAL;
+
+		vcn64 = next_vcn;
+	}
+
+	/* Check vcn consistency */
+	if (vcn64 == evcn + 1)
+		return run_buf - run_0;
+
+	return -EINVAL;
+}
+
+#ifdef NTFS3_CHECK_FREE_CLST
+/*
+ * run_unpack_ex
+ *
+ * unpacks packed runs from "run_buf"
+ * checks unpacked runs to be used in bitmap
+ * returns error, if negative, or real used bytes
+ */
+int run_unpack_ex(struct runs_tree *run, struct ntfs_sb_info *sbi, CLST ino,
+		  CLST svcn, CLST evcn, CLST vcn, const u8 *run_buf,
+		  u32 run_buf_size)
+{
+	int ret, err;
+	CLST next_vcn, lcn, len;
+	size_t index;
+	bool ok;
+	struct wnd_bitmap *wnd;
+
+	ret = run_unpack(run, sbi, ino, svcn, evcn, vcn, run_buf, run_buf_size);
+	if (ret < 0)
+		return ret;
+
+	if (!sbi->used.bitmap.sb || !run || (size_t)run == 1)
+		return ret;
+
+	if (ino == MFT_REC_BADCLUST)
+		return ret;
+
+	next_vcn = vcn = svcn;
+	wnd = &sbi->used.bitmap;
+
+	for (ok = run_lookup_entry(run, vcn, &lcn, &len, &index);
+	     next_vcn <= evcn;
+	     ok = run_get_entry(run, ++index, &vcn, &lcn, &len)) {
+		CLST real_free, i;
+
+		if (!ok || next_vcn != vcn)
+			return -EINVAL;
+
+		next_vcn = vcn + len;
+
+		if (lcn == SPARSE_LCN)
+			continue;
+
+		if (sbi->flags & NTFS_FLAGS_NEED_REPLAY)
+			continue;
+
+next:
+		down_read_nested(&wnd->rw_lock, BITMAP_MUTEX_CLUSTERS);
+		/* Check for free blocks */
+		ok = wnd_is_used(wnd, lcn, len);
+		up_read(&wnd->rw_lock);
+		if (ok)
+			continue;
+
+		ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
+
+		if (!down_write_trylock(&wnd->rw_lock))
+			continue;
+
+		/* Find first free */
+		real_free = len;
+		while (real_free && !wnd_is_free(wnd, lcn, 1)) {
+			lcn += 1;
+			real_free -= 1;
+		}
+
+		if (!real_free) {
+			up_write(&wnd->rw_lock);
+			continue;
+		}
+
+		/* Find total free */
+		i = 1;
+		while (i < real_free && wnd_is_free(wnd, lcn + i, 1))
+			i += 1;
+
+		real_free = i;
+
+		err = wnd_set_used(wnd, lcn, real_free);
+		up_write(&wnd->rw_lock);
+
+		if (err)
+			return err;
+
+		if (len != real_free) {
+			len -= real_free + 1;
+			lcn += real_free + 1;
+			goto next;
+		}
+	}
+
+	return ret;
+}
+#endif
+
+/*
+ * run_get_highest_vcn
+ *
+ * returns the highest vcn from a mapping pairs array
+ * it used while replaying log file
+ */
+int run_get_highest_vcn(CLST vcn, const u8 *run_buf, u64 *highest_vcn)
+{
+	const u8 *run = run_buf;
+	u64 vcn64 = vcn;
+	u8 size_size;
+
+	while ((size_size = *run & 0xF)) {
+		u8 offset_size = *run++ >> 4;
+		u64 len = 0;
+		u8 *p = (u8 *)&len;
+
+		switch (size_size) {
+		default:
+error:
+			return -EINVAL;
+
+#ifdef __BIG_ENDIAN
+		case 8:
+			p[0] = run[7];
+			fallthrough;
+		case 7:
+			p[1] = run[6];
+			fallthrough;
+		case 6:
+			p[2] = run[5];
+			fallthrough;
+		case 5:
+			p[3] = run[4];
+			fallthrough;
+		case 4:
+			p[4] = run[3];
+			fallthrough;
+		case 3:
+			p[5] = run[2];
+			fallthrough;
+		case 2:
+			p[6] = run[1];
+			fallthrough;
+		case 1:
+			p[7] = run[0];
+#else
+		case 8:
+			p[7] = run[7];
+			fallthrough;
+		case 7:
+			p[6] = run[6];
+			fallthrough;
+		case 6:
+			p[5] = run[5];
+			fallthrough;
+		case 5:
+			p[4] = run[4];
+			fallthrough;
+		case 4:
+			p[3] = run[3];
+			fallthrough;
+		case 3:
+			p[2] = run[2];
+			fallthrough;
+		case 2:
+			p[1] = run[1];
+			fallthrough;
+		case 1:
+			p[0] = run[0];
+#endif
+		}
+
+		/* skip size_size */
+		run += size_size;
+
+		if (!len)
+			goto error;
+
+		run += offset_size;
+
+#ifdef NTFS3_64BIT_CLUSTER
+		if ((vcn >> 32) || (len >> 32))
+			goto error;
+#endif
+		vcn64 += len;
+	}
+
+	*highest_vcn = vcn64 - 1;
+	return 0;
+}
-- 
2.25.4


^ permalink raw reply related	[flat|nested] 21+ messages in thread

* [PATCH v11 05/10] fs/ntfs3: Add attrib operations
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
                   ` (3 preceding siblings ...)
  2020-10-30 15:02 ` [PATCH v11 04/10] fs/ntfs3: Add file operations and implementation Konstantin Komarov
@ 2020-10-30 15:02 ` Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 06/10] fs/ntfs3: Add compression Konstantin Komarov
                   ` (7 subsequent siblings)
  12 siblings, 0 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This adds attrib operations

Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
---
 fs/ntfs3/attrib.c   | 1400 +++++++++++++++++++++++++++++++++++++++++++
 fs/ntfs3/attrlist.c |  463 ++++++++++++++
 fs/ntfs3/xattr.c    | 1069 +++++++++++++++++++++++++++++++++
 3 files changed, 2932 insertions(+)
 create mode 100644 fs/ntfs3/attrib.c
 create mode 100644 fs/ntfs3/attrlist.c
 create mode 100644 fs/ntfs3/xattr.c

diff --git a/fs/ntfs3/attrib.c b/fs/ntfs3/attrib.c
new file mode 100644
index 000000000000..e5997b48e7a4
--- /dev/null
+++ b/fs/ntfs3/attrib.c
@@ -0,0 +1,1400 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ * TODO: merge attr_set_size/attr_data_get_block/attr_allocate_frame?
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/hash.h>
+#include <linux/nls.h>
+#include <linux/ratelimit.h>
+#include <linux/slab.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+/*
+ * You can set external NTFS_MIN_LOG2_OF_CLUMP/NTFS_MAX_LOG2_OF_CLUMP to manage
+ * preallocate algorithm
+ */
+#ifndef NTFS_MIN_LOG2_OF_CLUMP
+#define NTFS_MIN_LOG2_OF_CLUMP 16
+#endif
+
+#ifndef NTFS_MAX_LOG2_OF_CLUMP
+#define NTFS_MAX_LOG2_OF_CLUMP 26
+#endif
+
+// 16M
+#define NTFS_CLUMP_MIN (1 << (NTFS_MIN_LOG2_OF_CLUMP + 8))
+// 16G
+#define NTFS_CLUMP_MAX (1ull << (NTFS_MAX_LOG2_OF_CLUMP + 8))
+
+/*
+ * get_pre_allocated
+ *
+ */
+static inline u64 get_pre_allocated(u64 size)
+{
+	u32 clump;
+	u8 align_shift;
+	u64 ret;
+
+	if (size <= NTFS_CLUMP_MIN) {
+		clump = 1 << NTFS_MIN_LOG2_OF_CLUMP;
+		align_shift = NTFS_MIN_LOG2_OF_CLUMP;
+	} else if (size >= NTFS_CLUMP_MAX) {
+		clump = 1 << NTFS_MAX_LOG2_OF_CLUMP;
+		align_shift = NTFS_MAX_LOG2_OF_CLUMP;
+	} else {
+		align_shift = NTFS_MIN_LOG2_OF_CLUMP - 1 +
+			      __ffs(size >> (8 + NTFS_MIN_LOG2_OF_CLUMP));
+		clump = 1u << align_shift;
+	}
+
+	ret = (((size + clump - 1) >> align_shift)) << align_shift;
+
+	return ret;
+}
+
+/*
+ * attr_must_be_resident
+ *
+ * returns true if attribute must be resident
+ */
+static inline bool attr_must_be_resident(struct ntfs_sb_info *sbi,
+					 enum ATTR_TYPE type)
+{
+	const struct ATTR_DEF_ENTRY *de;
+
+	switch (type) {
+	case ATTR_STD:
+	case ATTR_NAME:
+	case ATTR_ID:
+	case ATTR_LABEL:
+	case ATTR_VOL_INFO:
+	case ATTR_ROOT:
+	case ATTR_EA_INFO:
+		return true;
+	default:
+		de = ntfs_query_def(sbi, type);
+		if (de && (de->flags & NTFS_ATTR_MUST_BE_RESIDENT))
+			return true;
+		return false;
+	}
+}
+
+/*
+ * attr_load_runs
+ *
+ * load all runs stored in 'attr'
+ */
+int attr_load_runs(struct ATTRIB *attr, struct ntfs_inode *ni,
+		   struct runs_tree *run, const CLST *vcn)
+{
+	int err;
+	CLST svcn = le64_to_cpu(attr->nres.svcn);
+	CLST evcn = le64_to_cpu(attr->nres.evcn);
+	u32 asize;
+	u16 run_off;
+
+	if (svcn >= evcn + 1 || run_is_mapped_full(run, svcn, evcn))
+		return 0;
+
+	if (vcn && (evcn < *vcn || *vcn < svcn))
+		return -EINVAL;
+
+	asize = le32_to_cpu(attr->size);
+	run_off = le16_to_cpu(attr->nres.run_off);
+	err = run_unpack_ex(run, ni->mi.sbi, ni->mi.rno, svcn, evcn,
+			    vcn ? *vcn : svcn, Add2Ptr(attr, run_off),
+			    asize - run_off);
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+/*
+ * int run_deallocate_ex
+ *
+ * Deallocate clusters
+ */
+static int run_deallocate_ex(struct ntfs_sb_info *sbi, struct runs_tree *run,
+			     CLST vcn, CLST len, CLST *done, bool trim)
+{
+	int err = 0;
+	CLST vcn0 = vcn, lcn, clen, dn = 0;
+	size_t idx;
+
+	if (!len)
+		goto out;
+
+	if (!run_lookup_entry(run, vcn, &lcn, &clen, &idx)) {
+failed:
+		run_truncate(run, vcn0);
+		err = -EINVAL;
+		goto out;
+	}
+
+	for (;;) {
+		if (clen > len)
+			clen = len;
+
+		if (!clen) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		if (lcn != SPARSE_LCN) {
+			mark_as_free_ex(sbi, lcn, clen, trim);
+			dn += clen;
+		}
+
+		len -= clen;
+		if (!len)
+			break;
+
+		if (!run_get_entry(run, ++idx, &vcn, &lcn, &clen)) {
+			// save memory - don't load entire run
+			goto failed;
+		}
+	}
+
+out:
+	if (done)
+		*done = dn;
+
+	return err;
+}
+
+/*
+ * attr_allocate_clusters
+ *
+ * find free space, mark it as used and store in 'run'
+ */
+int attr_allocate_clusters(struct ntfs_sb_info *sbi, struct runs_tree *run,
+			   CLST vcn, CLST lcn, CLST len, CLST *pre_alloc,
+			   enum ALLOCATE_OPT opt, CLST *alen, const size_t fr,
+			   CLST *new_lcn)
+{
+	int err;
+	CLST flen, vcn0 = vcn, pre = pre_alloc ? *pre_alloc : 0;
+	struct wnd_bitmap *wnd = &sbi->used.bitmap;
+	size_t cnt = run->count;
+
+	for (;;) {
+		err = ntfs_look_for_free_space(sbi, lcn, len + pre, &lcn, &flen,
+					       opt);
+
+		if (err == -ENOSPC && pre) {
+			pre = 0;
+			if (*pre_alloc)
+				*pre_alloc = 0;
+			continue;
+		}
+
+		if (err)
+			goto out;
+
+		if (new_lcn && vcn == vcn0)
+			*new_lcn = lcn;
+
+		/* Add new fragment into run storage */
+		if (!run_add_entry(run, vcn, lcn, flen)) {
+			down_write_nested(&wnd->rw_lock, BITMAP_MUTEX_CLUSTERS);
+			wnd_set_free(wnd, lcn, flen);
+			up_write(&wnd->rw_lock);
+			err = -ENOMEM;
+			goto out;
+		}
+
+		vcn += flen;
+
+		if (flen >= len || opt == ALLOCATE_MFT ||
+		    (fr && run->count - cnt >= fr)) {
+			*alen = vcn - vcn0;
+			return 0;
+		}
+
+		len -= flen;
+	}
+
+out:
+	/* undo */
+	run_deallocate_ex(sbi, run, vcn0, vcn - vcn0, NULL, false);
+	run_truncate(run, vcn0);
+
+	return err;
+}
+
+/*
+ * attr_set_size_res
+ *
+ * helper for attr_set_size
+ */
+static int attr_set_size_res(struct ntfs_inode *ni, struct ATTRIB *attr,
+			     struct ATTR_LIST_ENTRY *le, struct mft_inode *mi,
+			     u64 new_size, struct runs_tree *run,
+			     struct ATTRIB **ins_attr)
+{
+	int err = 0;
+	struct ntfs_sb_info *sbi = mi->sbi;
+	struct MFT_REC *rec = mi->mrec;
+	u32 used = le32_to_cpu(rec->used);
+	u32 asize = le32_to_cpu(attr->size);
+	u32 aoff = PtrOffset(rec, attr);
+	u32 rsize = le32_to_cpu(attr->res.data_size);
+	u32 tail = used - aoff - asize;
+	char *next = Add2Ptr(attr, asize);
+	int dsize = QuadAlign(new_size) - QuadAlign(rsize);
+	CLST len, alen;
+	struct ATTRIB *attr_s = NULL;
+	bool is_ext;
+
+	if (dsize < 0) {
+		memmove(next + dsize, next, tail);
+	} else if (dsize > 0) {
+		if (used + dsize > sbi->max_bytes_per_attr)
+			goto resident2nonresident;
+
+		memmove(next + dsize, next, tail);
+		memset(next, 0, dsize);
+	}
+
+	rec->used = cpu_to_le32(used + dsize);
+	attr->size = cpu_to_le32(asize + dsize);
+	attr->res.data_size = cpu_to_le32(new_size);
+	mi->dirty = true;
+	*ins_attr = attr;
+
+	return 0;
+
+resident2nonresident:
+	len = bytes_to_cluster(sbi, rsize);
+
+	run_init(run);
+
+	is_ext = is_attr_ext(attr);
+
+	if (!len) {
+		alen = 0;
+	} else if (is_ext) {
+		if (!run_add_entry(run, 0, SPARSE_LCN, len)) {
+			err = -ENOMEM;
+			goto out;
+		}
+		alen = len;
+	} else {
+		err = attr_allocate_clusters(sbi, run, 0, 0, len, NULL,
+					     ALLOCATE_DEF, &alen, 0, NULL);
+		if (err)
+			goto out;
+
+		err = ntfs_sb_write_run(sbi, run, 0, resident_data(attr),
+					rsize);
+		if (err)
+			goto out;
+	}
+
+	attr_s = ntfs_memdup(attr, asize);
+	if (!attr_s) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	/*verify(mi_remove_attr(mi, attr));*/
+	used -= asize;
+	memmove(attr, Add2Ptr(attr, asize), used - aoff);
+	rec->used = cpu_to_le32(used);
+	mi->dirty = true;
+	if (le)
+		al_remove_le(ni, le);
+
+	err = ni_insert_nonresident(ni, attr_s->type, attr_name(attr_s),
+				    attr_s->name_len, run, 0, alen,
+				    attr_s->flags, &attr, NULL);
+	if (err)
+		goto out;
+
+	ntfs_free(attr_s);
+	attr->nres.data_size = cpu_to_le64(rsize);
+	attr->nres.valid_size = attr->nres.data_size;
+
+	*ins_attr = attr;
+
+	if (attr_s->type == ATTR_DATA && !attr_s->name_len &&
+	    run == &ni->file.run) {
+		ni->ni_flags &= ~NI_FLAG_RESIDENT;
+	}
+
+	/* Resident attribute becomes non resident */
+	return 0;
+
+out:
+	/* undo: do not trim new allocated clusters */
+	run_deallocate(sbi, run, false);
+	run_close(run);
+
+	if (attr_s) {
+		memmove(next, Add2Ptr(rec, aoff), used - aoff);
+		memcpy(Add2Ptr(rec, aoff), attr_s, asize);
+		rec->used = cpu_to_le32(used + asize);
+		mi->dirty = true;
+		ntfs_free(attr_s);
+		/*reinsert le*/
+	}
+
+	return err;
+}
+
+/*
+ * attr_set_size
+ *
+ * change the size of attribute
+ * Extend:
+ *   - sparse/compressed: no allocated clusters
+ *   - normal: append allocated and preallocated new clusters
+ * Shrink:
+ *   - no deallocate if keep_prealloc is set
+ */
+int attr_set_size(struct ntfs_inode *ni, enum ATTR_TYPE type,
+		  const __le16 *name, u8 name_len, struct runs_tree *run,
+		  u64 new_size, const u64 *new_valid, bool keep_prealloc,
+		  struct ATTRIB **ret)
+{
+	int err = 0;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	u8 cluster_bits = sbi->cluster_bits;
+	bool is_mft =
+		ni->mi.rno == MFT_REC_MFT && type == ATTR_DATA && !name_len;
+	u64 old_valid, old_size, old_alloc, new_alloc, new_alloc_tmp;
+	struct ATTRIB *attr, *attr_b;
+	struct ATTR_LIST_ENTRY *le, *le_b;
+	struct mft_inode *mi, *mi_b;
+	CLST alen, vcn, lcn, new_alen, old_alen, svcn, evcn;
+	CLST next_svcn, pre_alloc = -1, done = 0;
+	bool is_ext;
+	u32 align;
+	struct MFT_REC *rec;
+
+again:
+	le_b = NULL;
+	attr_b = ni_find_attr(ni, NULL, &le_b, type, name, name_len, NULL,
+			      &mi_b);
+	if (!attr_b) {
+		err = -ENOENT;
+		goto out;
+	}
+
+	if (!attr_b->non_res) {
+		err = attr_set_size_res(ni, attr_b, le_b, mi_b, new_size, run,
+					&attr_b);
+		if (err || !attr_b->non_res)
+			goto out;
+
+		/* layout of records may be changed, so do a full search */
+		goto again;
+	}
+
+	is_ext = is_attr_ext(attr_b);
+
+again_1:
+	if (is_ext) {
+		align = 1u << (attr_b->nres.c_unit + cluster_bits);
+		if (is_attr_sparsed(attr_b))
+			keep_prealloc = false;
+	} else {
+		align = sbi->cluster_size;
+	}
+
+	old_valid = le64_to_cpu(attr_b->nres.valid_size);
+	old_size = le64_to_cpu(attr_b->nres.data_size);
+	old_alloc = le64_to_cpu(attr_b->nres.alloc_size);
+	old_alen = old_alloc >> cluster_bits;
+
+	new_alloc = (new_size + align - 1) & ~(u64)(align - 1);
+	new_alen = new_alloc >> cluster_bits;
+
+	if (keep_prealloc && is_ext)
+		keep_prealloc = false;
+
+	if (keep_prealloc && new_size < old_size) {
+		attr_b->nres.data_size = cpu_to_le64(new_size);
+		mi_b->dirty = true;
+		goto ok;
+	}
+
+	vcn = old_alen - 1;
+
+	svcn = le64_to_cpu(attr_b->nres.svcn);
+	evcn = le64_to_cpu(attr_b->nres.evcn);
+
+	if (svcn <= vcn && vcn <= evcn) {
+		attr = attr_b;
+		le = le_b;
+		mi = mi_b;
+	} else if (!le_b) {
+		err = -EINVAL;
+		goto out;
+	} else {
+		le = le_b;
+		attr = ni_find_attr(ni, attr_b, &le, type, name, name_len, &vcn,
+				    &mi);
+		if (!attr) {
+			err = -EINVAL;
+			goto out;
+		}
+
+next_le_1:
+		svcn = le64_to_cpu(attr->nres.svcn);
+		evcn = le64_to_cpu(attr->nres.evcn);
+	}
+
+next_le:
+	rec = mi->mrec;
+
+	err = attr_load_runs(attr, ni, run, NULL);
+	if (err)
+		goto out;
+
+	if (new_size > old_size) {
+		CLST to_allocate;
+		size_t free;
+
+		if (new_alloc <= old_alloc) {
+			attr_b->nres.data_size = cpu_to_le64(new_size);
+			mi_b->dirty = true;
+			goto ok;
+		}
+
+		to_allocate = new_alen - old_alen;
+add_alloc_in_same_attr_seg:
+		lcn = 0;
+		if (is_mft) {
+			/* mft allocates clusters from mftzone */
+			pre_alloc = 0;
+		} else if (is_ext) {
+			/* no preallocate for sparse/compress */
+			pre_alloc = 0;
+		} else if (pre_alloc == -1) {
+			pre_alloc = 0;
+			if (type == ATTR_DATA && !name_len &&
+			    sbi->options.prealloc) {
+				CLST new_alen2 = bytes_to_cluster(
+					sbi, get_pre_allocated(new_size));
+				pre_alloc = new_alen2 - new_alen;
+			}
+
+			/* Get the last lcn to allocate from */
+			if (old_alen &&
+			    !run_lookup_entry(run, vcn, &lcn, NULL, NULL)) {
+				lcn = SPARSE_LCN;
+			}
+
+			if (lcn == SPARSE_LCN)
+				lcn = 0;
+			else if (lcn)
+				lcn += 1;
+
+			free = wnd_zeroes(&sbi->used.bitmap);
+			if (to_allocate > free) {
+				err = -ENOSPC;
+				goto out;
+			}
+
+			if (pre_alloc && to_allocate + pre_alloc > free)
+				pre_alloc = 0;
+		}
+
+		vcn = old_alen;
+
+		if (is_ext) {
+			if (!run_add_entry(run, vcn, SPARSE_LCN, to_allocate)) {
+				err = -ENOMEM;
+				goto out;
+			}
+			alen = to_allocate;
+		} else {
+			/* ~3 bytes per fragment */
+			err = attr_allocate_clusters(
+				sbi, run, vcn, lcn, to_allocate, &pre_alloc,
+				is_mft ? ALLOCATE_MFT : 0, &alen,
+				is_mft ? 0 :
+					 (sbi->record_size -
+					  le32_to_cpu(rec->used) + 8) /
+							 3 +
+						 1,
+				NULL);
+			if (err)
+				goto out;
+		}
+
+		done += alen;
+		vcn += alen;
+		if (to_allocate > alen)
+			to_allocate -= alen;
+		else
+			to_allocate = 0;
+
+pack_runs:
+		err = mi_pack_runs(mi, attr, run, vcn - svcn);
+		if (err)
+			goto out;
+
+		next_svcn = le64_to_cpu(attr->nres.evcn) + 1;
+		new_alloc_tmp = (u64)next_svcn << cluster_bits;
+		attr_b->nres.alloc_size = cpu_to_le64(new_alloc_tmp);
+		mi_b->dirty = true;
+
+		if (next_svcn >= vcn && !to_allocate) {
+			/* Normal way. update attribute and exit */
+			attr_b->nres.data_size = cpu_to_le64(new_size);
+			goto ok;
+		}
+
+		/* at least two mft to avoid recursive loop*/
+		if (is_mft && next_svcn == vcn &&
+		    ((u64)done << sbi->cluster_bits) >= 2 * sbi->record_size) {
+			new_size = new_alloc_tmp;
+			attr_b->nres.data_size = attr_b->nres.alloc_size;
+			goto ok;
+		}
+
+		if (le32_to_cpu(rec->used) < sbi->record_size) {
+			old_alen = next_svcn;
+			evcn = old_alen - 1;
+			goto add_alloc_in_same_attr_seg;
+		}
+
+		attr_b->nres.data_size = attr_b->nres.alloc_size;
+		if (new_alloc_tmp < old_valid)
+			attr_b->nres.valid_size = attr_b->nres.data_size;
+
+		if (type == ATTR_LIST) {
+			err = ni_expand_list(ni);
+			if (err)
+				goto out;
+			if (next_svcn < vcn)
+				goto pack_runs;
+
+			/* layout of records is changed */
+			goto again;
+		}
+
+		if (!ni->attr_list.size) {
+			err = ni_create_attr_list(ni);
+			if (err)
+				goto out;
+			/* layout of records is changed */
+		}
+
+		if (next_svcn >= vcn) {
+			/* this is mft data, repeat */
+			goto again;
+		}
+
+		/* insert new attribute segment */
+		err = ni_insert_nonresident(ni, type, name, name_len, run,
+					    next_svcn, vcn - next_svcn,
+					    attr_b->flags, &attr, &mi);
+		if (err)
+			goto out;
+
+		if (!is_mft)
+			run_truncate_head(run, evcn + 1);
+
+		svcn = le64_to_cpu(attr->nres.svcn);
+		evcn = le64_to_cpu(attr->nres.evcn);
+
+		le_b = NULL;
+		/* layout of records maybe changed */
+		/* find base attribute to update*/
+		attr_b = ni_find_attr(ni, NULL, &le_b, type, name, name_len,
+				      NULL, &mi_b);
+		if (!attr_b) {
+			err = -ENOENT;
+			goto out;
+		}
+
+		attr_b->nres.alloc_size = cpu_to_le64((u64)vcn << cluster_bits);
+		attr_b->nres.data_size = attr_b->nres.alloc_size;
+		attr_b->nres.valid_size = attr_b->nres.alloc_size;
+		mi_b->dirty = true;
+		goto again_1;
+	}
+
+	if (new_size != old_size ||
+	    (new_alloc != old_alloc && !keep_prealloc)) {
+		vcn = max(svcn, new_alen);
+		new_alloc_tmp = (u64)vcn << cluster_bits;
+
+		err = run_deallocate_ex(sbi, run, vcn, evcn - vcn + 1, &alen,
+					true);
+		if (err)
+			goto out;
+
+		run_truncate(run, vcn);
+
+		if (vcn > svcn) {
+			err = mi_pack_runs(mi, attr, run, vcn - svcn);
+			if (err)
+				goto out;
+		} else if (le && le->vcn) {
+			u16 le_sz = le16_to_cpu(le->size);
+
+			/*
+			 * NOTE: list entries for one attribute are always
+			 * the same size. We deal with last entry (vcn==0)
+			 * and it is not first in entries array
+			 * (list entry for std attribute always first)
+			 * So it is safe to step back
+			 */
+			mi_remove_attr(mi, attr);
+
+			if (!al_remove_le(ni, le)) {
+				err = -EINVAL;
+				goto out;
+			}
+
+			le = (struct ATTR_LIST_ENTRY *)((u8 *)le - le_sz);
+		} else {
+			attr->nres.evcn = cpu_to_le64((u64)vcn - 1);
+			mi->dirty = true;
+		}
+
+		attr_b->nres.alloc_size = cpu_to_le64(new_alloc_tmp);
+
+		if (vcn == new_alen) {
+			attr_b->nres.data_size = cpu_to_le64(new_size);
+			if (new_size < old_valid)
+				attr_b->nres.valid_size =
+					attr_b->nres.data_size;
+		} else {
+			if (new_alloc_tmp <=
+			    le64_to_cpu(attr_b->nres.data_size))
+				attr_b->nres.data_size =
+					attr_b->nres.alloc_size;
+			if (new_alloc_tmp <
+			    le64_to_cpu(attr_b->nres.valid_size))
+				attr_b->nres.valid_size =
+					attr_b->nres.alloc_size;
+		}
+
+		if (is_ext)
+			le64_sub_cpu(&attr_b->nres.total_size,
+				     ((u64)alen << cluster_bits));
+
+		mi_b->dirty = true;
+
+		if (new_alloc_tmp <= new_alloc)
+			goto ok;
+
+		old_size = new_alloc_tmp;
+		vcn = svcn - 1;
+
+		if (le == le_b) {
+			attr = attr_b;
+			mi = mi_b;
+			evcn = svcn - 1;
+			svcn = 0;
+			goto next_le;
+		}
+
+		if (le->type != type || le->name_len != name_len ||
+		    memcmp(le_name(le), name, name_len * sizeof(short))) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		err = ni_load_mi(ni, le, &mi);
+		if (err)
+			goto out;
+
+		attr = mi_find_attr(mi, NULL, type, name, name_len, &le->id);
+		if (!attr) {
+			err = -EINVAL;
+			goto out;
+		}
+		goto next_le_1;
+	}
+
+ok:
+	if (new_valid) {
+		__le64 valid = cpu_to_le64(min(*new_valid, new_size));
+
+		if (attr_b->nres.valid_size != valid) {
+			attr_b->nres.valid_size = valid;
+			mi_b->dirty = true;
+		}
+	}
+
+out:
+	if (!err && attr_b && ret)
+		*ret = attr_b;
+
+	/* update inode_set_bytes*/
+	if (!err && attr_b && attr_b->non_res &&
+	    ((type == ATTR_DATA && !name_len) ||
+	     (type == ATTR_ALLOC && name == I30_NAME))) {
+		bool dirty = false;
+
+		if (ni->vfs_inode.i_size != new_size) {
+			ni->vfs_inode.i_size = new_size;
+			dirty = true;
+		}
+
+		new_alloc = le64_to_cpu(attr_b->nres.alloc_size);
+		if (inode_get_bytes(&ni->vfs_inode) != new_alloc) {
+			inode_set_bytes(&ni->vfs_inode, new_alloc);
+			dirty = true;
+		}
+
+		if (dirty) {
+			ni->ni_flags |= NI_FLAG_UPDATE_PARENT;
+			mark_inode_dirty(&ni->vfs_inode);
+		}
+	}
+
+	return err;
+}
+
+int attr_data_get_block(struct ntfs_inode *ni, CLST vcn, CLST clen, CLST *lcn,
+			CLST *len, bool *new)
+{
+	int err = 0;
+	struct runs_tree *run = &ni->file.run;
+	struct ntfs_sb_info *sbi;
+	u8 cluster_bits;
+	struct ATTRIB *attr, *attr_b;
+	struct ATTR_LIST_ENTRY *le, *le_b;
+	struct mft_inode *mi, *mi_b;
+	CLST hint, svcn, to_alloc, evcn1, next_svcn, asize, end;
+	u64 new_size, total_size;
+	u32 clst_per_frame;
+	bool ok;
+
+	if (new)
+		*new = false;
+
+	down_read(&ni->file.run_lock);
+	ok = run_lookup_entry(run, vcn, lcn, len, NULL);
+	up_read(&ni->file.run_lock);
+
+	if (ok && (*lcn != SPARSE_LCN || !new)) {
+		/* normal way */
+		return 0;
+	}
+
+	if (!clen)
+		clen = 1;
+
+	if (ok && clen > *len)
+		clen = *len;
+
+	sbi = ni->mi.sbi;
+	cluster_bits = sbi->cluster_bits;
+	new_size = ((u64)vcn + clen) << cluster_bits;
+
+	ni_lock(ni);
+	down_write(&ni->file.run_lock);
+
+again:
+	le_b = NULL;
+	attr_b = ni_find_attr(ni, NULL, &le_b, ATTR_DATA, NULL, 0, NULL, &mi_b);
+	if (!attr_b) {
+		err = -ENOENT;
+		goto out;
+	}
+
+	if (!attr_b->non_res) {
+		if (!new) {
+			*lcn = RESIDENT_LCN;
+			goto out;
+		}
+
+		err = attr_set_size_res(ni, attr_b, le_b, mi_b, new_size, run,
+					&attr_b);
+		if (err)
+			goto out;
+
+		if (!attr_b->non_res) {
+			/* Resident attribute still resident */
+			*lcn = RESIDENT_LCN;
+			goto out;
+		}
+
+		/* Resident attribute becomes non resident */
+		goto again;
+	}
+
+	asize = le64_to_cpu(attr_b->nres.alloc_size) >> sbi->cluster_bits;
+	if (vcn >= asize) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	clst_per_frame = 1u << attr_b->nres.c_unit;
+	to_alloc = (clen + clst_per_frame - 1) & ~(clst_per_frame - 1);
+
+	if (vcn + to_alloc > asize)
+		to_alloc = asize - vcn;
+
+	svcn = le64_to_cpu(attr_b->nres.svcn);
+	evcn1 = le64_to_cpu(attr_b->nres.evcn) + 1;
+
+	attr = attr_b;
+	le = le_b;
+	mi = mi_b;
+
+	if (le_b && (vcn < svcn || evcn1 <= vcn)) {
+		attr = ni_find_attr(ni, attr_b, &le, ATTR_DATA, NULL, 0, &vcn,
+				    &mi);
+		if (!attr) {
+			err = -EINVAL;
+			goto out;
+		}
+		svcn = le64_to_cpu(attr->nres.svcn);
+		evcn1 = le64_to_cpu(attr->nres.evcn) + 1;
+	}
+
+	err = attr_load_runs(attr, ni, run, NULL);
+	if (err)
+		goto out;
+
+	if (!ok) {
+		ok = run_lookup_entry(run, vcn, lcn, len, NULL);
+		if (ok && (*lcn != SPARSE_LCN || !new)) {
+			/* normal way */
+			err = 0;
+			goto ok;
+		}
+
+		if (!ok && !new) {
+			*len = 0;
+			err = 0;
+			goto ok;
+		}
+
+		if (ok && clen > *len) {
+			clen = *len;
+			new_size = ((u64)vcn + clen) << cluster_bits;
+			to_alloc = (clen + clst_per_frame - 1) &
+				   ~(clst_per_frame - 1);
+		}
+	}
+
+	if (!is_attr_ext(attr_b)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* Get the last lcn to allocate from */
+	hint = 0;
+
+	if (vcn > evcn1) {
+		if (!run_add_entry(run, evcn1, SPARSE_LCN, vcn - evcn1)) {
+			err = -ENOMEM;
+			goto out;
+		}
+	} else if (vcn && !run_lookup_entry(run, vcn - 1, &hint, NULL, NULL)) {
+		hint = -1;
+	}
+
+	err = attr_allocate_clusters(
+		sbi, run, vcn, hint + 1, to_alloc, NULL, 0, len,
+		(sbi->record_size - le32_to_cpu(mi->mrec->used) + 8) / 3 + 1,
+		lcn);
+	if (err)
+		goto out;
+	*new = true;
+
+	end = vcn + *len;
+
+	total_size = le64_to_cpu(attr_b->nres.total_size) +
+		     ((u64)*len << cluster_bits);
+
+repack:
+	err = mi_pack_runs(mi, attr, run, max(end, evcn1) - svcn);
+	if (err)
+		goto out;
+
+	attr_b->nres.total_size = cpu_to_le64(total_size);
+	inode_set_bytes(&ni->vfs_inode, total_size);
+
+	mi_b->dirty = true;
+	mark_inode_dirty(&ni->vfs_inode);
+
+	/* stored [vcn : next_svcn) from [vcn : end) */
+	next_svcn = le64_to_cpu(attr->nres.evcn) + 1;
+
+	if (end <= evcn1) {
+		if (next_svcn == evcn1) {
+			/* Normal way. update attribute and exit */
+			goto ok;
+		}
+		/* add new segment [next_svcn : evcn1 - next_svcn )*/
+		if (!ni->attr_list.size) {
+			err = ni_create_attr_list(ni);
+			if (err)
+				goto out;
+			/* layout of records is changed */
+			le_b = NULL;
+			attr_b = ni_find_attr(ni, NULL, &le_b, ATTR_DATA, NULL,
+					      0, NULL, &mi_b);
+			if (!attr_b) {
+				err = -ENOENT;
+				goto out;
+			}
+
+			attr = attr_b;
+			le = le_b;
+			mi = mi_b;
+			goto repack;
+		}
+	}
+
+	svcn = evcn1;
+
+	/* Estimate next attribute */
+	attr = ni_find_attr(ni, attr, &le, ATTR_DATA, NULL, 0, &svcn, &mi);
+
+	if (attr) {
+		CLST alloc = bytes_to_cluster(
+			sbi, le64_to_cpu(attr_b->nres.alloc_size));
+		CLST evcn = le64_to_cpu(attr->nres.evcn);
+
+		if (end < next_svcn)
+			end = next_svcn;
+		while (end > evcn) {
+			/* remove segment [svcn : evcn)*/
+			mi_remove_attr(mi, attr);
+
+			if (!al_remove_le(ni, le)) {
+				err = -EINVAL;
+				goto out;
+			}
+
+			if (evcn + 1 >= alloc) {
+				/* last attribute segment */
+				evcn1 = evcn + 1;
+				goto ins_ext;
+			}
+
+			if (ni_load_mi(ni, le, &mi)) {
+				attr = NULL;
+				goto out;
+			}
+
+			attr = mi_find_attr(mi, NULL, ATTR_DATA, NULL, 0,
+					    &le->id);
+			if (!attr) {
+				err = -EINVAL;
+				goto out;
+			}
+			svcn = le64_to_cpu(attr->nres.svcn);
+			evcn = le64_to_cpu(attr->nres.evcn);
+		}
+
+		if (end < svcn)
+			end = svcn;
+
+		err = attr_load_runs(attr, ni, run, &end);
+		if (err)
+			goto out;
+
+		evcn1 = evcn + 1;
+		attr->nres.svcn = cpu_to_le64(next_svcn);
+		err = mi_pack_runs(mi, attr, run, evcn1 - next_svcn);
+		if (err)
+			goto out;
+
+		le->vcn = cpu_to_le64(next_svcn);
+		ni->attr_list.dirty = true;
+		mi->dirty = true;
+
+		next_svcn = le64_to_cpu(attr->nres.evcn) + 1;
+	}
+ins_ext:
+	if (evcn1 > next_svcn) {
+		err = ni_insert_nonresident(ni, ATTR_DATA, NULL, 0, run,
+					    next_svcn, evcn1 - next_svcn,
+					    attr_b->flags, &attr, &mi);
+		if (err)
+			goto out;
+	}
+ok:
+	run_truncate_around(run, vcn);
+out:
+	up_write(&ni->file.run_lock);
+	ni_unlock(ni);
+
+	return err;
+}
+
+/*
+ * attr_load_runs_vcn
+ *
+ * load runs with vcn
+ */
+int attr_load_runs_vcn(struct ntfs_inode *ni, enum ATTR_TYPE type,
+		       const __le16 *name, u8 name_len, struct runs_tree *run,
+		       CLST vcn)
+{
+	struct ATTRIB *attr;
+	int err;
+	CLST svcn, evcn;
+	u16 ro;
+
+	attr = ni_find_attr(ni, NULL, NULL, type, name, name_len, &vcn, NULL);
+	if (!attr)
+		return -ENOENT;
+
+	svcn = le64_to_cpu(attr->nres.svcn);
+	evcn = le64_to_cpu(attr->nres.evcn);
+
+	if (evcn < vcn || vcn < svcn)
+		return -EINVAL;
+
+	ro = le16_to_cpu(attr->nres.run_off);
+	err = run_unpack_ex(run, ni->mi.sbi, ni->mi.rno, svcn, evcn, svcn,
+			    Add2Ptr(attr, ro), le32_to_cpu(attr->size) - ro);
+	if (err < 0)
+		return err;
+	return 0;
+}
+
+/*
+ * attr_is_frame_compressed
+ *
+ * This function is used to detect compressed frame
+ */
+int attr_is_frame_compressed(struct ntfs_inode *ni, struct ATTRIB *attr,
+			     CLST frame, CLST *clst_data)
+{
+	int err;
+	u32 clst_frame;
+	CLST len, lcn, vcn, alen, slen, vcn1;
+	size_t idx;
+	struct runs_tree *run;
+
+	*clst_data = 0;
+
+	if (!is_attr_compressed(attr))
+		return 0;
+
+	if (!attr->non_res)
+		return 0;
+
+	clst_frame = 1u << attr->nres.c_unit;
+	vcn = frame * clst_frame;
+	run = &ni->file.run;
+
+	if (!run_lookup_entry(run, vcn, &lcn, &len, &idx)) {
+		err = attr_load_runs_vcn(ni, attr->type, attr_name(attr),
+					 attr->name_len, run, vcn);
+		if (err)
+			return err;
+
+		if (!run_lookup_entry(run, vcn, &lcn, &len, &idx))
+			return -ENOENT;
+	}
+
+	if (lcn == SPARSE_LCN) {
+		/* sparsed frame */
+		return 0;
+	}
+
+	if (len >= clst_frame) {
+		/*
+		 * The frame is not compressed 'cause
+		 * it does not contain any sparse clusters
+		 */
+		*clst_data = clst_frame;
+		return 0;
+	}
+
+	alen = bytes_to_cluster(ni->mi.sbi, le64_to_cpu(attr->nres.alloc_size));
+	slen = 0;
+	*clst_data = len;
+
+	/*
+	 * The frame is compressed if *clst_data + slen >= clst_frame
+	 * Check next fragments
+	 */
+	while ((vcn += len) < alen) {
+		vcn1 = vcn;
+
+		if (!run_get_entry(run, ++idx, &vcn, &lcn, &len) ||
+		    vcn1 != vcn) {
+			err = attr_load_runs_vcn(ni, attr->type,
+						 attr_name(attr),
+						 attr->name_len, run, vcn1);
+			if (err)
+				return err;
+			vcn = vcn1;
+
+			if (!run_lookup_entry(run, vcn, &lcn, &len, &idx))
+				return -ENOENT;
+		}
+
+		if (lcn == SPARSE_LCN) {
+			slen += len;
+		} else {
+			if (slen) {
+				/*
+				 * data_clusters + sparse_clusters =
+				 * not enough for frame
+				 */
+				return -EINVAL;
+			}
+			*clst_data += len;
+		}
+
+		if (*clst_data + slen >= clst_frame) {
+			if (!slen) {
+				/*
+				 * There is no sparsed clusters in this frame
+				 * So it is not compressed
+				 */
+				*clst_data = clst_frame;
+			} else {
+				/*frame is compressed*/
+			}
+			break;
+		}
+	}
+
+	return 0;
+}
+
+/*
+ * attr_allocate_frame
+ *
+ * allocate/free clusters for 'frame'
+ * assumed: down_write(&ni->file.run_lock);
+ */
+int attr_allocate_frame(struct ntfs_inode *ni, CLST frame, size_t compr_size,
+			u64 new_valid)
+{
+	int err = 0;
+	struct runs_tree *run = &ni->file.run;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	struct ATTRIB *attr, *attr_b;
+	struct ATTR_LIST_ENTRY *le, *le_b;
+	struct mft_inode *mi, *mi_b;
+	CLST svcn, evcn1, next_svcn, lcn, len;
+	CLST vcn, end, clst_data;
+	u64 total_size, valid_size, data_size;
+
+	le_b = NULL;
+	attr_b = ni_find_attr(ni, NULL, &le_b, ATTR_DATA, NULL, 0, NULL, &mi_b);
+	if (!attr_b)
+		return -ENOENT;
+
+	if (!is_attr_ext(attr_b))
+		return -EINVAL;
+
+	vcn = frame << NTFS_LZNT_CUNIT;
+	total_size = le64_to_cpu(attr_b->nres.total_size);
+
+	svcn = le64_to_cpu(attr_b->nres.svcn);
+	evcn1 = le64_to_cpu(attr_b->nres.evcn) + 1;
+	data_size = le64_to_cpu(attr_b->nres.data_size);
+
+	if (svcn <= vcn && vcn < evcn1) {
+		attr = attr_b;
+		le = le_b;
+		mi = mi_b;
+	} else if (!le_b) {
+		err = -EINVAL;
+		goto out;
+	} else {
+		le = le_b;
+		attr = ni_find_attr(ni, attr_b, &le, ATTR_DATA, NULL, 0, &vcn,
+				    &mi);
+		if (!attr) {
+			err = -EINVAL;
+			goto out;
+		}
+		svcn = le64_to_cpu(attr->nres.svcn);
+		evcn1 = le64_to_cpu(attr->nres.evcn) + 1;
+	}
+
+	err = attr_load_runs(attr, ni, run, NULL);
+	if (err)
+		goto out;
+
+	err = attr_is_frame_compressed(ni, attr_b, frame, &clst_data);
+	if (err)
+		goto out;
+
+	total_size -= (u64)clst_data << sbi->cluster_bits;
+
+	len = bytes_to_cluster(sbi, compr_size);
+
+	if (len == clst_data)
+		goto out;
+
+	if (len < clst_data) {
+		err = run_deallocate_ex(sbi, run, vcn + len, clst_data - len,
+					NULL, true);
+		if (err)
+			goto out;
+
+		if (!run_add_entry(run, vcn + len, SPARSE_LCN,
+				   clst_data - len)) {
+			err = -ENOMEM;
+			goto out;
+		}
+		end = vcn + clst_data;
+		/* run contains updated range [vcn + len : end) */
+	} else {
+		CLST alen, hint;
+		/* Get the last lcn to allocate from */
+		if (vcn + clst_data &&
+		    !run_lookup_entry(run, vcn + clst_data - 1, &hint, NULL,
+				      NULL)) {
+			hint = -1;
+		}
+
+		err = attr_allocate_clusters(sbi, run, vcn + clst_data,
+					     hint + 1, len - clst_data, NULL, 0,
+					     &alen, 0, &lcn);
+		if (err)
+			goto out;
+
+		end = vcn + len;
+		/* run contains updated range [vcn + clst_data : end) */
+	}
+
+	total_size += (u64)len << sbi->cluster_bits;
+
+repack:
+	err = mi_pack_runs(mi, attr, run, max(end, evcn1) - svcn);
+	if (err)
+		goto out;
+
+	attr_b->nres.total_size = cpu_to_le64(total_size);
+	inode_set_bytes(&ni->vfs_inode, total_size);
+
+	mi_b->dirty = true;
+	mark_inode_dirty(&ni->vfs_inode);
+
+	/* stored [vcn : next_svcn) from [vcn : end) */
+	next_svcn = le64_to_cpu(attr->nres.evcn) + 1;
+
+	if (end <= evcn1) {
+		if (next_svcn == evcn1) {
+			/* Normal way. update attribute and exit */
+			goto ok;
+		}
+		/* add new segment [next_svcn : evcn1 - next_svcn )*/
+		if (!ni->attr_list.size) {
+			err = ni_create_attr_list(ni);
+			if (err)
+				goto out;
+			/* layout of records is changed */
+			le_b = NULL;
+			attr_b = ni_find_attr(ni, NULL, &le_b, ATTR_DATA, NULL,
+					      0, NULL, &mi_b);
+			if (!attr_b) {
+				err = -ENOENT;
+				goto out;
+			}
+
+			attr = attr_b;
+			le = le_b;
+			mi = mi_b;
+			goto repack;
+		}
+	}
+
+	svcn = evcn1;
+
+	/* Estimate next attribute */
+	attr = ni_find_attr(ni, attr, &le, ATTR_DATA, NULL, 0, &svcn, &mi);
+
+	if (attr) {
+		CLST alloc = bytes_to_cluster(
+			sbi, le64_to_cpu(attr_b->nres.alloc_size));
+		CLST evcn = le64_to_cpu(attr->nres.evcn);
+
+		if (end < next_svcn)
+			end = next_svcn;
+		while (end > evcn) {
+			/* remove segment [svcn : evcn)*/
+			mi_remove_attr(mi, attr);
+
+			if (!al_remove_le(ni, le)) {
+				err = -EINVAL;
+				goto out;
+			}
+
+			if (evcn + 1 >= alloc) {
+				/* last attribute segment */
+				evcn1 = evcn + 1;
+				goto ins_ext;
+			}
+
+			if (ni_load_mi(ni, le, &mi)) {
+				attr = NULL;
+				goto out;
+			}
+
+			attr = mi_find_attr(mi, NULL, ATTR_DATA, NULL, 0,
+					    &le->id);
+			if (!attr) {
+				err = -EINVAL;
+				goto out;
+			}
+			svcn = le64_to_cpu(attr->nres.svcn);
+			evcn = le64_to_cpu(attr->nres.evcn);
+		}
+
+		if (end < svcn)
+			end = svcn;
+
+		err = attr_load_runs(attr, ni, run, &end);
+		if (err)
+			goto out;
+
+		evcn1 = evcn + 1;
+		attr->nres.svcn = cpu_to_le64(next_svcn);
+		err = mi_pack_runs(mi, attr, run, evcn1 - next_svcn);
+		if (err)
+			goto out;
+
+		le->vcn = cpu_to_le64(next_svcn);
+		ni->attr_list.dirty = true;
+		mi->dirty = true;
+
+		next_svcn = le64_to_cpu(attr->nres.evcn) + 1;
+	}
+ins_ext:
+	if (evcn1 > next_svcn) {
+		err = ni_insert_nonresident(ni, ATTR_DATA, NULL, 0, run,
+					    next_svcn, evcn1 - next_svcn,
+					    attr_b->flags, &attr, &mi);
+		if (err)
+			goto out;
+	}
+ok:
+	run_truncate_around(run, vcn);
+out:
+	if (new_valid > data_size)
+		new_valid = data_size;
+
+	valid_size = le64_to_cpu(attr_b->nres.valid_size);
+	if (new_valid != valid_size) {
+		attr_b->nres.valid_size = cpu_to_le64(valid_size);
+		mi_b->dirty = true;
+	}
+
+	return err;
+}
diff --git a/fs/ntfs3/attrlist.c b/fs/ntfs3/attrlist.c
new file mode 100644
index 000000000000..e94512259354
--- /dev/null
+++ b/fs/ntfs3/attrlist.c
@@ -0,0 +1,463 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+/* Returns true if le is valid */
+static inline bool al_is_valid_le(const struct ntfs_inode *ni,
+				  struct ATTR_LIST_ENTRY *le)
+{
+	if (!le || !ni->attr_list.le || !ni->attr_list.size)
+		return false;
+
+	return PtrOffset(ni->attr_list.le, le) + le16_to_cpu(le->size) <=
+	       ni->attr_list.size;
+}
+
+void al_destroy(struct ntfs_inode *ni)
+{
+	run_close(&ni->attr_list.run);
+	ntfs_free(ni->attr_list.le);
+	ni->attr_list.le = NULL;
+	ni->attr_list.size = 0;
+	ni->attr_list.dirty = false;
+}
+
+/*
+ * ntfs_load_attr_list
+ *
+ * This method makes sure that the ATTRIB list, if present,
+ * has been properly set up.
+ */
+int ntfs_load_attr_list(struct ntfs_inode *ni, struct ATTRIB *attr)
+{
+	int err;
+	size_t lsize;
+	void *le = NULL;
+
+	if (ni->attr_list.size)
+		return 0;
+
+	if (!attr->non_res) {
+		lsize = le32_to_cpu(attr->res.data_size);
+		le = ntfs_alloc(al_aligned(lsize), 0);
+		if (!le) {
+			err = -ENOMEM;
+			goto out;
+		}
+		memcpy(le, resident_data(attr), lsize);
+	} else if (attr->nres.svcn) {
+		err = -EINVAL;
+		goto out;
+	} else {
+		u16 run_off = le16_to_cpu(attr->nres.run_off);
+
+		lsize = le64_to_cpu(attr->nres.data_size);
+
+		run_init(&ni->attr_list.run);
+
+		err = run_unpack_ex(&ni->attr_list.run, ni->mi.sbi, ni->mi.rno,
+				    0, le64_to_cpu(attr->nres.evcn), 0,
+				    Add2Ptr(attr, run_off),
+				    le32_to_cpu(attr->size) - run_off);
+		if (err < 0)
+			goto out;
+
+		le = ntfs_alloc(al_aligned(lsize), 0);
+		if (!le) {
+			err = -ENOMEM;
+			goto out;
+		}
+
+		err = ntfs_read_run_nb(ni->mi.sbi, &ni->attr_list.run, 0, le,
+				       lsize, NULL);
+		if (err)
+			goto out;
+	}
+
+	ni->attr_list.size = lsize;
+	ni->attr_list.le = le;
+
+	return 0;
+
+out:
+	ni->attr_list.le = le;
+	al_destroy(ni);
+
+	return err;
+}
+
+/*
+ * al_enumerate
+ *
+ * Returns the next list le
+ * if le is NULL then returns the first le
+ */
+struct ATTR_LIST_ENTRY *al_enumerate(struct ntfs_inode *ni,
+				     struct ATTR_LIST_ENTRY *le)
+{
+	size_t off;
+	u16 sz;
+
+	if (!le) {
+		le = ni->attr_list.le;
+	} else {
+		sz = le16_to_cpu(le->size);
+		if (sz < sizeof(struct ATTR_LIST_ENTRY)) {
+			/* Impossible 'cause we should not return such le */
+			return NULL;
+		}
+		le = Add2Ptr(le, sz);
+	}
+
+	/* Check boundary */
+	off = PtrOffset(ni->attr_list.le, le);
+	if (off + sizeof(struct ATTR_LIST_ENTRY) > ni->attr_list.size) {
+		// The regular end of list
+		return NULL;
+	}
+
+	sz = le16_to_cpu(le->size);
+
+	/* Check le for errors */
+	if (sz < sizeof(struct ATTR_LIST_ENTRY) ||
+	    off + sz > ni->attr_list.size ||
+	    sz < le->name_off + le->name_len * sizeof(short)) {
+		return NULL;
+	}
+
+	return le;
+}
+
+/*
+ * al_find_le
+ *
+ * finds the first le in the list which matches type, name and vcn
+ * Returns NULL if not found
+ */
+struct ATTR_LIST_ENTRY *al_find_le(struct ntfs_inode *ni,
+				   struct ATTR_LIST_ENTRY *le,
+				   const struct ATTRIB *attr)
+{
+	CLST svcn = attr_svcn(attr);
+
+	return al_find_ex(ni, le, attr->type, attr_name(attr), attr->name_len,
+			  &svcn);
+}
+
+/*
+ * al_find_ex
+ *
+ * finds the first le in the list which matches type, name and vcn
+ * Returns NULL if not found
+ */
+struct ATTR_LIST_ENTRY *al_find_ex(struct ntfs_inode *ni,
+				   struct ATTR_LIST_ENTRY *le,
+				   enum ATTR_TYPE type, const __le16 *name,
+				   u8 name_len, const CLST *vcn)
+{
+	struct ATTR_LIST_ENTRY *ret = NULL;
+	u32 type_in = le32_to_cpu(type);
+
+	while ((le = al_enumerate(ni, le))) {
+		u64 le_vcn;
+		int diff;
+
+		/* List entries are sorted by type, name and vcn */
+		diff = le32_to_cpu(le->type) - type_in;
+		if (diff < 0)
+			continue;
+
+		if (diff > 0)
+			return ret;
+
+		if (le->name_len != name_len)
+			continue;
+
+		if (name_len &&
+		    memcmp(le_name(le), name, name_len * sizeof(short)))
+			continue;
+
+		if (!vcn)
+			return le;
+
+		le_vcn = le64_to_cpu(le->vcn);
+		if (*vcn == le_vcn)
+			return le;
+
+		if (*vcn < le_vcn)
+			return ret;
+
+		ret = le;
+	}
+
+	return ret;
+}
+
+/*
+ * al_find_le_to_insert
+ *
+ * finds the first list entry which matches type, name and vcn
+ * Returns NULL if not found
+ */
+static struct ATTR_LIST_ENTRY *
+al_find_le_to_insert(struct ntfs_inode *ni, enum ATTR_TYPE type,
+		     const __le16 *name, u8 name_len, const CLST *vcn)
+{
+	struct ATTR_LIST_ENTRY *le = NULL, *prev;
+	u32 type_in = le32_to_cpu(type);
+	int diff;
+
+	/* List entries are sorted by type, name, vcn */
+next:
+	le = al_enumerate(ni, prev = le);
+	if (!le)
+		goto out;
+	diff = le32_to_cpu(le->type) - type_in;
+	if (diff < 0)
+		goto next;
+	if (diff > 0)
+		goto out;
+
+	if (ntfs_cmp_names(name, name_len, le_name(le), le->name_len, NULL) > 0)
+		goto next;
+
+	if (!vcn || *vcn > le64_to_cpu(le->vcn))
+		goto next;
+
+out:
+	if (!le)
+		le = prev ? Add2Ptr(prev, le16_to_cpu(prev->size)) :
+			    ni->attr_list.le;
+
+	return le;
+}
+
+/*
+ * al_add_le
+ *
+ * adds an "attribute list entry" to the list.
+ */
+int al_add_le(struct ntfs_inode *ni, enum ATTR_TYPE type, const __le16 *name,
+	      u8 name_len, CLST svcn, __le16 id, const struct MFT_REF *ref,
+	      struct ATTR_LIST_ENTRY **new_le)
+{
+	int err;
+	struct ATTRIB *attr;
+	struct ATTR_LIST_ENTRY *le;
+	size_t off;
+	u16 sz;
+	size_t asize, new_asize;
+	u64 new_size;
+	typeof(ni->attr_list) *al = &ni->attr_list;
+
+	/*
+	 * Compute the size of the new le and the new length of the
+	 * list with al le added.
+	 */
+	sz = le_size(name_len);
+	new_size = al->size + sz;
+	asize = al_aligned(al->size);
+	new_asize = al_aligned(new_size);
+
+	/* Scan forward to the point at which the new le should be inserted. */
+	le = al_find_le_to_insert(ni, type, name, name_len, &svcn);
+	off = PtrOffset(al->le, le);
+
+	if (new_size > asize) {
+		void *ptr = ntfs_alloc(new_asize, 0);
+
+		if (!ptr)
+			return -ENOMEM;
+
+		memcpy(ptr, al->le, off);
+		memcpy(Add2Ptr(ptr, off + sz), le, al->size - off);
+		le = Add2Ptr(ptr, off);
+		ntfs_free(al->le);
+		al->le = ptr;
+	} else {
+		memmove(Add2Ptr(le, sz), le, al->size - off);
+	}
+
+	al->size = new_size;
+
+	le->type = type;
+	le->size = cpu_to_le16(sz);
+	le->name_len = name_len;
+	le->name_off = offsetof(struct ATTR_LIST_ENTRY, name);
+	le->vcn = cpu_to_le64(svcn);
+	le->ref = *ref;
+	le->id = id;
+	memcpy(le->name, name, sizeof(short) * name_len);
+
+	al->dirty = true;
+
+	err = attr_set_size(ni, ATTR_LIST, NULL, 0, &al->run, new_size,
+			    &new_size, true, &attr);
+	if (err)
+		return err;
+
+	if (attr && attr->non_res) {
+		err = ntfs_sb_write_run(ni->mi.sbi, &al->run, 0, al->le,
+					al->size);
+		if (err)
+			return err;
+	}
+
+	al->dirty = false;
+	*new_le = le;
+
+	return 0;
+}
+
+/*
+ * al_remove_le
+ *
+ * removes 'le' from attribute list
+ */
+bool al_remove_le(struct ntfs_inode *ni, struct ATTR_LIST_ENTRY *le)
+{
+	u16 size;
+	size_t off;
+	typeof(ni->attr_list) *al = &ni->attr_list;
+
+	if (!al_is_valid_le(ni, le))
+		return false;
+
+	/* Save on stack the size of le */
+	size = le16_to_cpu(le->size);
+	off = PtrOffset(al->le, le);
+
+	memmove(le, Add2Ptr(le, size), al->size - (off + size));
+
+	al->size -= size;
+	al->dirty = true;
+
+	return true;
+}
+
+/*
+ * al_delete_le
+ *
+ * deletes from the list the first le which matches its parameters.
+ */
+bool al_delete_le(struct ntfs_inode *ni, enum ATTR_TYPE type, CLST vcn,
+		  const __le16 *name, size_t name_len,
+		  const struct MFT_REF *ref)
+{
+	u16 size;
+	struct ATTR_LIST_ENTRY *le;
+	size_t off;
+	typeof(ni->attr_list) *al = &ni->attr_list;
+
+	/* Scan forward to the first le that matches the input */
+	le = al_find_ex(ni, NULL, type, name, name_len, &vcn);
+	if (!le)
+		return false;
+
+	off = PtrOffset(al->le, le);
+
+	if (!ref)
+		goto del;
+
+	/*
+	 * The caller specified a segment reference, so we have to
+	 * scan through the matching entries until we find that segment
+	 * reference or we run of matching entries.
+	 */
+next:
+	if (off + sizeof(struct ATTR_LIST_ENTRY) > al->size)
+		goto del;
+	if (le->type != type)
+		goto del;
+	if (le->name_len != name_len)
+		goto del;
+	if (name_len &&
+	    memcmp(name, Add2Ptr(le, le->name_off), name_len * sizeof(short)))
+		goto del;
+	if (le64_to_cpu(le->vcn) != vcn)
+		goto del;
+	if (!memcmp(ref, &le->ref, sizeof(*ref)))
+		goto del;
+
+	off += le16_to_cpu(le->size);
+	le = Add2Ptr(al->le, off);
+	goto next;
+
+del:
+	/*
+	 * If we've gone off the end of the list, or if the type, name,
+	 * and vcn don't match, then we don't have any matching records.
+	 */
+	if (off >= al->size)
+		return false;
+	if (le->type != type)
+		return false;
+	if (le->name_len != name_len)
+		return false;
+	if (name_len &&
+	    memcmp(name, Add2Ptr(le, le->name_off), name_len * sizeof(short)))
+		return false;
+	if (le64_to_cpu(le->vcn) != vcn)
+		return false;
+
+	/* Save on stack the size of le */
+	size = le16_to_cpu(le->size);
+	/* Delete the le. */
+	memmove(le, Add2Ptr(le, size), al->size - (off + size));
+
+	al->size -= size;
+	al->dirty = true;
+
+	return true;
+}
+
+/*
+ * al_update
+ */
+int al_update(struct ntfs_inode *ni)
+{
+	int err;
+	struct ATTRIB *attr;
+	typeof(ni->attr_list) *al = &ni->attr_list;
+
+	if (!al->dirty || !al->size)
+		return 0;
+
+	/*
+	 * attribute list increased on demand in al_add_le
+	 * attribute list decreased here
+	 */
+	err = attr_set_size(ni, ATTR_LIST, NULL, 0, &al->run, al->size, NULL,
+			    false, &attr);
+	if (err)
+		goto out;
+
+	if (!attr->non_res) {
+		memcpy(resident_data(attr), al->le, al->size);
+	} else {
+		err = ntfs_sb_write_run(ni->mi.sbi, &al->run, 0, al->le,
+					al->size);
+		if (err)
+			goto out;
+
+		attr->nres.valid_size = attr->nres.data_size;
+	}
+
+	ni->mi.dirty = true;
+	al->dirty = false;
+
+out:
+	return err;
+}
diff --git a/fs/ntfs3/xattr.c b/fs/ntfs3/xattr.c
new file mode 100644
index 000000000000..e21c06388677
--- /dev/null
+++ b/fs/ntfs3/xattr.c
@@ -0,0 +1,1069 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+#include <linux/posix_acl.h>
+#include <linux/posix_acl_xattr.h>
+#include <linux/xattr.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+// clang-format off
+#define SYSTEM_DOS_ATTRIB    "system.dos_attrib"
+#define SYSTEM_NTFS_ATTRIB   "system.ntfs_attrib"
+#define SYSTEM_NTFS_SECURITY "system.ntfs_security"
+#define USER_DOSATTRIB       "user.DOSATTRIB"
+// clang-format on
+
+static inline size_t unpacked_ea_size(const struct EA_FULL *ea)
+{
+	return !ea->size ? DwordAlign(offsetof(struct EA_FULL, name) + 1 +
+				      ea->name_len + le16_to_cpu(ea->elength)) :
+			   le32_to_cpu(ea->size);
+}
+
+static inline size_t packed_ea_size(const struct EA_FULL *ea)
+{
+	return offsetof(struct EA_FULL, name) + 1 -
+	       offsetof(struct EA_FULL, flags) + ea->name_len +
+	       le16_to_cpu(ea->elength);
+}
+
+/*
+ * find_ea
+ *
+ * assume there is at least one xattr in the list
+ */
+static inline bool find_ea(const struct EA_FULL *ea_all, u32 bytes,
+			   const char *name, u8 name_len, u32 *off)
+{
+	*off = 0;
+
+	if (!ea_all || !bytes)
+		return false;
+
+	for (;;) {
+		const struct EA_FULL *ea = Add2Ptr(ea_all, *off);
+		u32 next_off = *off + unpacked_ea_size(ea);
+
+		if (next_off > bytes)
+			return false;
+
+		if (ea->name_len == name_len &&
+		    !memcmp(ea->name, name, name_len))
+			return true;
+
+		*off = next_off;
+		if (next_off >= bytes)
+			return false;
+	}
+}
+
+/*
+ * ntfs_read_ea
+ *
+ * reads all extended attributes
+ * ea - new allocated memory
+ * info - pointer into resident data
+ */
+static int ntfs_read_ea(struct ntfs_inode *ni, struct EA_FULL **ea,
+			size_t add_bytes, const struct EA_INFO **info)
+{
+	int err;
+	struct ATTR_LIST_ENTRY *le = NULL;
+	struct ATTRIB *attr_info, *attr_ea;
+	void *ea_p;
+	u32 size;
+
+	static_assert(le32_to_cpu(ATTR_EA_INFO) < le32_to_cpu(ATTR_EA));
+
+	*ea = NULL;
+	*info = NULL;
+
+	attr_info =
+		ni_find_attr(ni, NULL, &le, ATTR_EA_INFO, NULL, 0, NULL, NULL);
+	attr_ea =
+		ni_find_attr(ni, attr_info, &le, ATTR_EA, NULL, 0, NULL, NULL);
+
+	if (!attr_ea || !attr_info)
+		return 0;
+
+	*info = resident_data_ex(attr_info, sizeof(struct EA_INFO));
+	if (!*info)
+		return -EINVAL;
+
+	/* Check Ea limit */
+	size = le32_to_cpu((*info)->size);
+	if (size > MAX_EA_DATA_SIZE || size + add_bytes > MAX_EA_DATA_SIZE)
+		return -EINVAL;
+
+	/* Allocate memory for packed Ea */
+	ea_p = ntfs_alloc(size + add_bytes, 0);
+	if (!ea_p)
+		return -ENOMEM;
+
+	if (attr_ea->non_res) {
+		struct runs_tree run;
+
+		run_init(&run);
+
+		err = attr_load_runs(attr_ea, ni, &run, NULL);
+		if (!err)
+			err = ntfs_read_run_nb(ni->mi.sbi, &run, 0, ea_p, size,
+					       NULL);
+		run_close(&run);
+
+		if (err)
+			goto out;
+	} else {
+		void *p = resident_data_ex(attr_ea, size);
+
+		if (!p) {
+			err = -EINVAL;
+			goto out;
+		}
+		memcpy(ea_p, p, size);
+	}
+
+	memset(Add2Ptr(ea_p, size), 0, add_bytes);
+	*ea = ea_p;
+	return 0;
+
+out:
+	ntfs_free(ea_p);
+	*ea = NULL;
+	return err;
+}
+
+/*
+ * ntfs_listxattr_hlp
+ *
+ * copy a list of xattrs names into the buffer
+ * provided, or compute the buffer size required
+ */
+static int ntfs_listxattr_hlp(struct ntfs_inode *ni, char *buffer,
+			      size_t bytes_per_buffer, size_t *bytes)
+{
+	const struct EA_INFO *info;
+	struct EA_FULL *ea_all = NULL;
+	const struct EA_FULL *ea;
+	u32 off, size;
+	int err;
+
+	*bytes = 0;
+
+	err = ntfs_read_ea(ni, &ea_all, 0, &info);
+	if (err)
+		return err;
+
+	if (!info || !ea_all)
+		return 0;
+
+	size = le32_to_cpu(info->size);
+
+	/* Enumerate all xattrs */
+	for (off = 0; off < size; off += unpacked_ea_size(ea)) {
+		ea = Add2Ptr(ea_all, off);
+
+		if (buffer) {
+			if (*bytes + ea->name_len + 1 > bytes_per_buffer) {
+				err = -ERANGE;
+				goto out;
+			}
+
+			memcpy(buffer + *bytes, ea->name, ea->name_len);
+			buffer[*bytes + ea->name_len] = 0;
+		}
+
+		*bytes += ea->name_len + 1;
+	}
+
+out:
+	ntfs_free(ea_all);
+	return err;
+}
+
+/*
+ * ntfs_get_ea
+ *
+ * reads xattr
+ */
+static int ntfs_get_ea(struct ntfs_inode *ni, const char *name, size_t name_len,
+		       void *buffer, size_t bytes_per_buffer, u32 *len)
+{
+	const struct EA_INFO *info;
+	struct EA_FULL *ea_all = NULL;
+	const struct EA_FULL *ea;
+	u32 off;
+	int err;
+
+	*len = 0;
+
+	if (name_len > 255) {
+		err = -ENAMETOOLONG;
+		goto out;
+	}
+
+	err = ntfs_read_ea(ni, &ea_all, 0, &info);
+	if (err)
+		goto out;
+
+	if (!info)
+		goto out;
+
+	/* Enumerate all xattrs */
+	if (!find_ea(ea_all, le32_to_cpu(info->size), name, name_len, &off)) {
+		err = -ENODATA;
+		goto out;
+	}
+	ea = Add2Ptr(ea_all, off);
+
+	*len = le16_to_cpu(ea->elength);
+	if (!buffer) {
+		err = 0;
+		goto out;
+	}
+
+	if (*len > bytes_per_buffer) {
+		err = -ERANGE;
+		goto out;
+	}
+	memcpy(buffer, ea->name + ea->name_len + 1, *len);
+	err = 0;
+
+out:
+	ntfs_free(ea_all);
+
+	return err;
+}
+
+static noinline int ntfs_getxattr_hlp(struct inode *inode, const char *name,
+				      void *value, size_t size,
+				      size_t *required)
+{
+	struct ntfs_inode *ni = ntfs_i(inode);
+	int err;
+	u32 len;
+
+	if (!(ni->ni_flags & NI_FLAG_EA))
+		return -ENODATA;
+
+	if (!required)
+		ni_lock(ni);
+
+	err = ntfs_get_ea(ni, name, strlen(name), value, size, &len);
+	if (!err)
+		err = len;
+	else if (-ERANGE == err && required)
+		*required = len;
+
+	if (!required)
+		ni_unlock(ni);
+
+	return err;
+}
+
+static noinline int ntfs_set_ea(struct inode *inode, const char *name,
+				const void *value, size_t val_size, int flags,
+				int locked)
+{
+	struct ntfs_inode *ni = ntfs_i(inode);
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	int err;
+	struct EA_INFO ea_info;
+	const struct EA_INFO *info;
+	struct EA_FULL *new_ea;
+	struct EA_FULL *ea_all = NULL;
+	size_t name_len, add;
+	u32 off, size;
+	__le16 size_pack;
+	struct ATTRIB *attr;
+	struct ATTR_LIST_ENTRY *le;
+	struct mft_inode *mi;
+	struct runs_tree ea_run;
+	u64 new_sz;
+	void *p;
+
+	if (!locked)
+		ni_lock(ni);
+
+	run_init(&ea_run);
+	name_len = strlen(name);
+
+	if (name_len > 255) {
+		err = -ENAMETOOLONG;
+		goto out;
+	}
+
+	add = DwordAlign(offsetof(struct EA_FULL, name) + 1 + name_len +
+			 val_size);
+
+	err = ntfs_read_ea(ni, &ea_all, add, &info);
+	if (err)
+		goto out;
+
+	if (!info) {
+		memset(&ea_info, 0, sizeof(ea_info));
+		size = 0;
+		size_pack = 0;
+	} else {
+		memcpy(&ea_info, info, sizeof(ea_info));
+		size = le32_to_cpu(ea_info.size);
+		size_pack = ea_info.size_pack;
+	}
+
+	if (info && find_ea(ea_all, size, name, name_len, &off)) {
+		struct EA_FULL *ea;
+		size_t ea_sz;
+
+		if (flags & XATTR_CREATE) {
+			err = -EEXIST;
+			goto out;
+		}
+
+		/* Remove current xattr */
+		ea = Add2Ptr(ea_all, off);
+		if (ea->flags & FILE_NEED_EA)
+			le16_add_cpu(&ea_info.count, -1);
+
+		ea_sz = unpacked_ea_size(ea);
+
+		le16_add_cpu(&ea_info.size_pack, 0 - packed_ea_size(ea));
+
+		memmove(ea, Add2Ptr(ea, ea_sz), size - off - ea_sz);
+
+		size -= ea_sz;
+		memset(Add2Ptr(ea_all, size), 0, ea_sz);
+
+		ea_info.size = cpu_to_le32(size);
+
+		if ((flags & XATTR_REPLACE) && !val_size)
+			goto update_ea;
+	} else {
+		if (flags & XATTR_REPLACE) {
+			err = -ENODATA;
+			goto out;
+		}
+
+		if (!ea_all) {
+			ea_all = ntfs_alloc(add, 1);
+			if (!ea_all) {
+				err = -ENOMEM;
+				goto out;
+			}
+		}
+	}
+
+	/* append new xattr */
+	new_ea = Add2Ptr(ea_all, size);
+	new_ea->size = cpu_to_le32(add);
+	new_ea->flags = 0;
+	new_ea->name_len = name_len;
+	new_ea->elength = cpu_to_le16(val_size);
+	memcpy(new_ea->name, name, name_len);
+	new_ea->name[name_len] = 0;
+	memcpy(new_ea->name + name_len + 1, value, val_size);
+
+	le16_add_cpu(&ea_info.size_pack, packed_ea_size(new_ea));
+	size += add;
+	ea_info.size = cpu_to_le32(size);
+
+update_ea:
+
+	if (!info) {
+		/* Create xattr */
+		if (!size) {
+			err = 0;
+			goto out;
+		}
+
+		err = ni_insert_resident(ni, sizeof(struct EA_INFO),
+					 ATTR_EA_INFO, NULL, 0, NULL, NULL);
+		if (err)
+			goto out;
+
+		err = ni_insert_resident(ni, 0, ATTR_EA, NULL, 0, NULL, NULL);
+		if (err)
+			goto out;
+	}
+
+	new_sz = size;
+	err = attr_set_size(ni, ATTR_EA, NULL, 0, &ea_run, new_sz, &new_sz,
+			    false, NULL);
+	if (err)
+		goto out;
+
+	le = NULL;
+	attr = ni_find_attr(ni, NULL, &le, ATTR_EA_INFO, NULL, 0, NULL, &mi);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (!size) {
+		/* delete xattr, ATTR_EA_INFO */
+		err = ni_remove_attr_le(ni, attr, le);
+		if (err)
+			goto out;
+	} else {
+		p = resident_data_ex(attr, sizeof(struct EA_INFO));
+		if (!p) {
+			err = -EINVAL;
+			goto out;
+		}
+		memcpy(p, &ea_info, sizeof(struct EA_INFO));
+		mi->dirty = true;
+	}
+
+	le = NULL;
+	attr = ni_find_attr(ni, NULL, &le, ATTR_EA, NULL, 0, NULL, &mi);
+	if (!attr) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (!size) {
+		/* delete xattr, ATTR_EA */
+		err = ni_remove_attr_le(ni, attr, le);
+		if (err)
+			goto out;
+	} else if (attr->non_res) {
+		err = ntfs_sb_write_run(sbi, &ea_run, 0, ea_all, size);
+		if (err)
+			goto out;
+	} else {
+		p = resident_data_ex(attr, size);
+		if (!p) {
+			err = -EINVAL;
+			goto out;
+		}
+		memcpy(p, ea_all, size);
+		mi->dirty = true;
+	}
+
+	if (ea_info.size_pack != size_pack)
+		ni->ni_flags |= NI_FLAG_UPDATE_PARENT;
+	mark_inode_dirty(&ni->vfs_inode);
+
+	/* Check if we delete the last xattr */
+	if (val_size || flags != XATTR_REPLACE ||
+	    ntfs_listxattr_hlp(ni, NULL, 0, &val_size) || val_size) {
+		ni->ni_flags |= NI_FLAG_EA;
+	} else {
+		ni->ni_flags &= ~NI_FLAG_EA;
+	}
+
+out:
+	if (!locked)
+		ni_unlock(ni);
+
+	run_close(&ea_run);
+	ntfs_free(ea_all);
+
+	return err;
+}
+
+static inline void ntfs_posix_acl_release(struct posix_acl *acl)
+{
+	if (acl && refcount_dec_and_test(&acl->a_refcount))
+		kfree(acl);
+}
+
+static struct posix_acl *ntfs_get_acl_ex(struct inode *inode, int type,
+					 int locked)
+{
+	struct ntfs_inode *ni = ntfs_i(inode);
+	const char *name;
+	struct posix_acl *acl;
+	size_t req;
+	int err;
+	void *buf;
+
+	/* allocate PATH_MAX bytes */
+	buf = __getname();
+	if (!buf)
+		return ERR_PTR(-ENOMEM);
+
+	/* Possible values of 'type' was already checked above */
+	name = type == ACL_TYPE_ACCESS ? XATTR_NAME_POSIX_ACL_ACCESS :
+					 XATTR_NAME_POSIX_ACL_DEFAULT;
+
+	if (!locked)
+		ni_lock(ni);
+
+	err = ntfs_getxattr_hlp(inode, name, buf, PATH_MAX, &req);
+
+	if (!locked)
+		ni_unlock(ni);
+
+	/* Translate extended attribute to acl */
+	if (err > 0) {
+		acl = posix_acl_from_xattr(&init_user_ns, buf, err);
+		if (!IS_ERR(acl))
+			set_cached_acl(inode, type, acl);
+	} else {
+		acl = err == -ENODATA ? NULL : ERR_PTR(err);
+	}
+
+	__putname(buf);
+
+	return acl;
+}
+
+/*
+ * ntfs_get_acl
+ *
+ * inode_operations::get_acl
+ */
+struct posix_acl *ntfs_get_acl(struct inode *inode, int type)
+{
+	return ntfs_get_acl_ex(inode, type, 0);
+}
+
+static noinline int ntfs_set_acl_ex(struct inode *inode, struct posix_acl *acl,
+				    int type, int locked)
+{
+	const char *name;
+	size_t size;
+	void *value = NULL;
+	int err = 0;
+
+	if (S_ISLNK(inode->i_mode))
+		return -EOPNOTSUPP;
+
+	switch (type) {
+	case ACL_TYPE_ACCESS:
+		if (acl) {
+			umode_t mode = inode->i_mode;
+
+			err = posix_acl_equiv_mode(acl, &mode);
+			if (err < 0)
+				return err;
+
+			if (inode->i_mode != mode) {
+				inode->i_mode = mode;
+				mark_inode_dirty(inode);
+			}
+
+			if (!err) {
+				/*
+				 * acl can be exactly represented in the
+				 * traditional file mode permission bits
+				 */
+				acl = NULL;
+				goto out;
+			}
+		}
+		name = XATTR_NAME_POSIX_ACL_ACCESS;
+		break;
+
+	case ACL_TYPE_DEFAULT:
+		if (!S_ISDIR(inode->i_mode))
+			return acl ? -EACCES : 0;
+		name = XATTR_NAME_POSIX_ACL_DEFAULT;
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	if (!acl)
+		goto out;
+
+	size = posix_acl_xattr_size(acl->a_count);
+	value = ntfs_alloc(size, 0);
+	if (!value)
+		return -ENOMEM;
+
+	err = posix_acl_to_xattr(&init_user_ns, acl, value, size);
+	if (err)
+		goto out;
+
+	err = ntfs_set_ea(inode, name, value, size, 0, locked);
+	if (err)
+		goto out;
+
+	inode->i_flags &= ~S_NOSEC;
+
+out:
+	if (!err)
+		set_cached_acl(inode, type, acl);
+
+	kfree(value);
+
+	return err;
+}
+
+/*
+ * ntfs_set_acl
+ *
+ * inode_operations::set_acl
+ */
+int ntfs_set_acl(struct inode *inode, struct posix_acl *acl, int type)
+{
+	return ntfs_set_acl_ex(inode, acl, type, 0);
+}
+
+static int ntfs_xattr_get_acl(struct inode *inode, int type, void *buffer,
+			      size_t size)
+{
+	struct super_block *sb = inode->i_sb;
+	struct posix_acl *acl;
+	int err;
+
+	if (!(sb->s_flags & SB_POSIXACL))
+		return -EOPNOTSUPP;
+
+	acl = ntfs_get_acl(inode, type);
+	if (IS_ERR(acl))
+		return PTR_ERR(acl);
+
+	if (!acl)
+		return -ENODATA;
+
+	err = posix_acl_to_xattr(&init_user_ns, acl, buffer, size);
+	ntfs_posix_acl_release(acl);
+
+	return err;
+}
+
+static int ntfs_xattr_set_acl(struct inode *inode, int type, const void *value,
+			      size_t size)
+{
+	struct super_block *sb = inode->i_sb;
+	struct posix_acl *acl;
+	int err;
+
+	if (!(sb->s_flags & SB_POSIXACL))
+		return -EOPNOTSUPP;
+
+	if (!inode_owner_or_capable(inode))
+		return -EPERM;
+
+	if (!value)
+		return 0;
+
+	acl = posix_acl_from_xattr(&init_user_ns, value, size);
+	if (IS_ERR(acl))
+		return PTR_ERR(acl);
+
+	if (acl) {
+		err = posix_acl_valid(sb->s_user_ns, acl);
+		if (err)
+			goto release_and_out;
+	}
+
+	err = ntfs_set_acl(inode, acl, type);
+
+release_and_out:
+	ntfs_posix_acl_release(acl);
+	return err;
+}
+
+/*
+ * ntfs_acl_chmod
+ *
+ * helper for 'ntfs_setattr'
+ */
+int ntfs_acl_chmod(struct inode *inode)
+{
+	struct super_block *sb = inode->i_sb;
+	int err;
+
+	if (!(sb->s_flags & SB_POSIXACL))
+		return 0;
+
+	if (S_ISLNK(inode->i_mode))
+		return -EOPNOTSUPP;
+
+	err = posix_acl_chmod(inode, inode->i_mode);
+
+	return err;
+}
+
+/*
+ * ntfs_permission
+ *
+ * inode_operations::permission
+ */
+int ntfs_permission(struct inode *inode, int mask)
+{
+	struct super_block *sb = inode->i_sb;
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+	int err;
+
+	if (sbi->options.no_acs_rules) {
+		/* "no access rules" mode - allow all changes */
+		return 0;
+	}
+
+	err = generic_permission(inode, mask);
+
+	return err;
+}
+
+/*
+ * ntfs_listxattr
+ *
+ * inode_operations::listxattr
+ */
+ssize_t ntfs_listxattr(struct dentry *dentry, char *buffer, size_t size)
+{
+	struct inode *inode = d_inode(dentry);
+	struct ntfs_inode *ni = ntfs_i(inode);
+	ssize_t ret = -1;
+	int err;
+
+	if (!(ni->ni_flags & NI_FLAG_EA)) {
+		ret = 0;
+		goto out;
+	}
+
+	ni_lock(ni);
+
+	err = ntfs_listxattr_hlp(ni, buffer, size, (size_t *)&ret);
+
+	ni_unlock(ni);
+
+	if (err)
+		ret = err;
+out:
+
+	return ret;
+}
+
+static int ntfs_getxattr(const struct xattr_handler *handler, struct dentry *de,
+			 struct inode *inode, const char *name, void *buffer,
+			 size_t size)
+{
+	int err;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	size_t name_len = strlen(name);
+
+	/* Dispatch request */
+	if (name_len == sizeof(SYSTEM_DOS_ATTRIB) - 1 &&
+	    !memcmp(name, SYSTEM_DOS_ATTRIB, sizeof(SYSTEM_DOS_ATTRIB))) {
+		/* system.dos_attrib */
+		if (!buffer) {
+			err = sizeof(u8);
+		} else if (size < sizeof(u8)) {
+			err = -ENODATA;
+		} else {
+			err = sizeof(u8);
+			*(u8 *)buffer = le32_to_cpu(ni->std_fa);
+		}
+		goto out;
+	}
+
+	if (name_len == sizeof(SYSTEM_NTFS_ATTRIB) - 1 &&
+	    !memcmp(name, SYSTEM_NTFS_ATTRIB, sizeof(SYSTEM_NTFS_ATTRIB))) {
+		/* system.ntfs_attrib */
+		if (!buffer) {
+			err = sizeof(u32);
+		} else if (size < sizeof(u32)) {
+			err = -ENODATA;
+		} else {
+			err = sizeof(u32);
+			*(u32 *)buffer = le32_to_cpu(ni->std_fa);
+		}
+		goto out;
+	}
+
+	if (name_len == sizeof(USER_DOSATTRIB) - 1 &&
+	    !memcmp(name, USER_DOSATTRIB, sizeof(USER_DOSATTRIB))) {
+		/* user.DOSATTRIB */
+		if (!buffer) {
+			err = 5;
+		} else if (size < 5) {
+			err = -ENODATA;
+		} else {
+			err = sprintf((char *)buffer, "0x%x",
+				      le32_to_cpu(ni->std_fa) & 0xff) +
+			      1;
+		}
+		goto out;
+	}
+
+	if (name_len == sizeof(SYSTEM_NTFS_SECURITY) - 1 &&
+	    !memcmp(name, SYSTEM_NTFS_SECURITY, sizeof(SYSTEM_NTFS_SECURITY))) {
+		/* system.ntfs_security*/
+		struct SECURITY_DESCRIPTOR_RELATIVE *sd = NULL;
+		size_t sd_size = 0;
+
+		if (!is_ntfs3(ni->mi.sbi)) {
+			/* we should get nt4 security */
+			err = -EINVAL;
+			goto out;
+		} else if (le32_to_cpu(ni->std_security_id) <
+			   SECURITY_ID_FIRST) {
+			err = -ENOENT;
+			goto out;
+		}
+
+		err = ntfs_get_security_by_id(ni->mi.sbi, ni->std_security_id,
+					      &sd, &sd_size);
+		if (err)
+			goto out;
+
+		if (!is_sd_valid(sd, sd_size)) {
+			ntfs_inode_warn(
+				inode,
+				"looks like you get incorrect security descriptor id=%u",
+				ni->std_security_id);
+		}
+
+		if (!buffer) {
+			err = sd_size;
+		} else if (size < sd_size) {
+			err = -ENODATA;
+		} else {
+			err = sd_size;
+			memcpy(buffer, sd, sd_size);
+		}
+		ntfs_free(sd);
+		goto out;
+	}
+
+	if ((name_len == sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1 &&
+	     !memcmp(name, XATTR_NAME_POSIX_ACL_ACCESS,
+		     sizeof(XATTR_NAME_POSIX_ACL_ACCESS))) ||
+	    (name_len == sizeof(XATTR_NAME_POSIX_ACL_DEFAULT) - 1 &&
+	     !memcmp(name, XATTR_NAME_POSIX_ACL_DEFAULT,
+		     sizeof(XATTR_NAME_POSIX_ACL_DEFAULT)))) {
+		err = ntfs_xattr_get_acl(
+			inode,
+			name_len == sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1 ?
+				ACL_TYPE_ACCESS :
+				ACL_TYPE_DEFAULT,
+			buffer, size);
+	} else {
+		err = ntfs_getxattr_hlp(inode, name, buffer, size, NULL);
+	}
+
+out:
+	return err;
+}
+
+/*
+ * ntfs_setxattr
+ *
+ * inode_operations::setxattr
+ */
+static noinline int ntfs_setxattr(const struct xattr_handler *handler,
+				  struct dentry *de, struct inode *inode,
+				  const char *name, const void *value,
+				  size_t size, int flags)
+{
+	int err = -EINVAL;
+	struct ntfs_inode *ni = ntfs_i(inode);
+	size_t name_len = strlen(name);
+	enum FILE_ATTRIBUTE new_fa;
+
+	/* Dispatch request */
+	if (name_len == sizeof(SYSTEM_DOS_ATTRIB) - 1 &&
+	    !memcmp(name, SYSTEM_DOS_ATTRIB, sizeof(SYSTEM_DOS_ATTRIB))) {
+		if (sizeof(u8) != size)
+			goto out;
+		new_fa = cpu_to_le32(*(u8 *)value);
+		goto set_new_fa;
+	}
+
+	if (name_len == sizeof(SYSTEM_NTFS_ATTRIB) - 1 &&
+	    !memcmp(name, SYSTEM_NTFS_ATTRIB, sizeof(SYSTEM_NTFS_ATTRIB))) {
+		if (sizeof(u32) != size)
+			goto out;
+		new_fa = cpu_to_le32(*(u32 *)value);
+
+		if (S_ISREG(inode->i_mode)) {
+			/* Process compressed/sparsed in special way*/
+			ni_lock(ni);
+			err = ni_new_attr_flags(ni, new_fa);
+			ni_unlock(ni);
+			if (err)
+				goto out;
+		}
+		goto set_new_fa;
+	}
+
+	if (name_len == sizeof(USER_DOSATTRIB) - 1 &&
+	    !memcmp(name, USER_DOSATTRIB, sizeof(USER_DOSATTRIB))) {
+		u32 attrib;
+
+		if (size < 4 || ((char *)value)[size - 1])
+			goto out;
+
+		/*
+		 * The input value must be string in form 0x%x with last zero
+		 * This means that the 'size' must be 4, 5, ...
+		 *  E.g: 0x1 - 4 bytes, 0x20 - 5 bytes
+		 */
+		if (sscanf((char *)value, "0x%x", &attrib) != 1)
+			goto out;
+
+		new_fa = cpu_to_le32(attrib);
+set_new_fa:
+		/*
+		 * Thanks Mark Harmstone:
+		 * keep directory bit consistency
+		 */
+		if (S_ISDIR(inode->i_mode))
+			new_fa |= FILE_ATTRIBUTE_DIRECTORY;
+		else
+			new_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
+
+		if (ni->std_fa != new_fa) {
+			ni->std_fa = new_fa;
+			/* std attribute always in primary record */
+			ni->mi.dirty = true;
+			mark_inode_dirty(inode);
+		}
+		err = 0;
+
+		goto out;
+	}
+
+	if (name_len == sizeof(SYSTEM_NTFS_SECURITY) - 1 &&
+	    !memcmp(name, SYSTEM_NTFS_SECURITY, sizeof(SYSTEM_NTFS_SECURITY))) {
+		/* system.ntfs_security*/
+		__le32 security_id;
+		bool inserted;
+		struct ATTR_STD_INFO5 *std;
+
+		if (!is_ntfs3(ni->mi.sbi)) {
+			/*
+			 * we should replace ATTR_SECURE
+			 * Skip this way cause it is nt4 feature
+			 */
+			err = -EINVAL;
+			goto out;
+		}
+
+		if (!is_sd_valid(value, size)) {
+			err = -EINVAL;
+			ntfs_inode_warn(
+				inode,
+				"you try to set invalid security descriptor");
+			goto out;
+		}
+
+		err = ntfs_insert_security(ni->mi.sbi, value, size,
+					   &security_id, &inserted);
+		if (err)
+			goto out;
+
+		ni_lock(ni);
+		std = ni_std5(ni);
+		if (!std) {
+			err = -EINVAL;
+		} else if (std->security_id != security_id) {
+			std->security_id = ni->std_security_id = security_id;
+			/* std attribute always in primary record */
+			ni->mi.dirty = true;
+			mark_inode_dirty(&ni->vfs_inode);
+		}
+		ni_unlock(ni);
+		goto out;
+	}
+
+	if ((name_len == sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1 &&
+	     !memcmp(name, XATTR_NAME_POSIX_ACL_ACCESS,
+		     sizeof(XATTR_NAME_POSIX_ACL_ACCESS))) ||
+	    (name_len == sizeof(XATTR_NAME_POSIX_ACL_DEFAULT) - 1 &&
+	     !memcmp(name, XATTR_NAME_POSIX_ACL_DEFAULT,
+		     sizeof(XATTR_NAME_POSIX_ACL_DEFAULT)))) {
+		err = ntfs_xattr_set_acl(
+			inode,
+			name_len == sizeof(XATTR_NAME_POSIX_ACL_ACCESS) - 1 ?
+				ACL_TYPE_ACCESS :
+				ACL_TYPE_DEFAULT,
+			value, size);
+	} else {
+		err = ntfs_set_ea(inode, name, value, size, flags, 0);
+	}
+
+out:
+	return err;
+}
+
+/*
+ * Initialize the ACLs of a new inode. Called from ntfs_create_inode.
+ */
+int ntfs_init_acl(struct inode *inode, struct inode *dir)
+{
+	struct posix_acl *default_acl, *acl;
+	int err;
+
+	/*
+	 * TODO refactoring lock
+	 * ni_lock(dir) ... -> posix_acl_create(dir,...) -> ntfs_get_acl -> ni_lock(dir)
+	 */
+	inode->i_default_acl = NULL;
+
+	default_acl = ntfs_get_acl_ex(dir, ACL_TYPE_DEFAULT, 1);
+
+	if (!default_acl || default_acl == ERR_PTR(-EOPNOTSUPP)) {
+		inode->i_mode &= ~current_umask();
+		err = 0;
+		goto out;
+	}
+
+	if (IS_ERR(default_acl)) {
+		err = PTR_ERR(default_acl);
+		goto out;
+	}
+
+	acl = default_acl;
+	err = __posix_acl_create(&acl, GFP_NOFS, &inode->i_mode);
+	if (err < 0)
+		goto out1;
+	if (!err) {
+		posix_acl_release(acl);
+		acl = NULL;
+	}
+
+	if (!S_ISDIR(inode->i_mode)) {
+		posix_acl_release(default_acl);
+		default_acl = NULL;
+	}
+
+	if (default_acl)
+		err = ntfs_set_acl_ex(inode, default_acl, ACL_TYPE_DEFAULT, 1);
+
+	if (!acl)
+		inode->i_acl = NULL;
+	else if (!err)
+		err = ntfs_set_acl_ex(inode, acl, ACL_TYPE_ACCESS, 1);
+
+	posix_acl_release(acl);
+out1:
+	posix_acl_release(default_acl);
+
+out:
+	return err;
+}
+
+static bool ntfs_xattr_user_list(struct dentry *dentry)
+{
+	return 1;
+}
+
+static const struct xattr_handler ntfs_xattr_handler = {
+	.prefix = "",
+	.get = ntfs_getxattr,
+	.set = ntfs_setxattr,
+	.list = ntfs_xattr_user_list,
+};
+
+const struct xattr_handler *ntfs_xattr_handlers[] = {
+	&ntfs_xattr_handler,
+	NULL,
+};
-- 
2.25.4


^ permalink raw reply related	[flat|nested] 21+ messages in thread

* [PATCH v11 06/10] fs/ntfs3: Add compression
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
                   ` (4 preceding siblings ...)
  2020-10-30 15:02 ` [PATCH v11 05/10] fs/ntfs3: Add attrib operations Konstantin Komarov
@ 2020-10-30 15:02 ` Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 07/10] fs/ntfs3: Add NTFS journal Konstantin Komarov
                   ` (6 subsequent siblings)
  12 siblings, 0 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This adds compression

Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
---
 fs/ntfs3/lznt.c | 452 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 452 insertions(+)
 create mode 100644 fs/ntfs3/lznt.c

diff --git a/fs/ntfs3/lznt.c b/fs/ntfs3/lznt.c
new file mode 100644
index 000000000000..d7b794591d2e
--- /dev/null
+++ b/fs/ntfs3/lznt.c
@@ -0,0 +1,452 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/nls.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+// clang-format off
+/* src buffer is zero */
+#define LZNT_ERROR_ALL_ZEROS	1
+#define LZNT_CHUNK_SIZE		0x1000
+// clang-format on
+
+struct lznt_hash {
+	const u8 *p1;
+	const u8 *p2;
+};
+
+struct lznt {
+	const u8 *unc;
+	const u8 *unc_end;
+	const u8 *best_match;
+	size_t max_len;
+	bool std;
+
+	struct lznt_hash hash[LZNT_CHUNK_SIZE];
+};
+
+static inline size_t get_match_len(const u8 *ptr, const u8 *end, const u8 *prev,
+				   size_t max_len)
+{
+	size_t len = 0;
+
+	while (ptr + len < end && ptr[len] == prev[len] && ++len < max_len)
+		;
+	return len;
+}
+
+static size_t longest_match_std(const u8 *src, struct lznt *ctx)
+{
+	size_t hash_index;
+	size_t len1 = 0, len2 = 0;
+	const u8 **hash;
+
+	hash_index =
+		((40543U * ((((src[0] << 4) ^ src[1]) << 4) ^ src[2])) >> 4) &
+		(LZNT_CHUNK_SIZE - 1);
+
+	hash = &(ctx->hash[hash_index].p1);
+
+	if (hash[0] >= ctx->unc && hash[0] < src && hash[0][0] == src[0] &&
+	    hash[0][1] == src[1] && hash[0][2] == src[2]) {
+		len1 = 3;
+		if (ctx->max_len > 3)
+			len1 += get_match_len(src + 3, ctx->unc_end,
+					      hash[0] + 3, ctx->max_len - 3);
+	}
+
+	if (hash[1] >= ctx->unc && hash[1] < src && hash[1][0] == src[0] &&
+	    hash[1][1] == src[1] && hash[1][2] == src[2]) {
+		len2 = 3;
+		if (ctx->max_len > 3)
+			len2 += get_match_len(src + 3, ctx->unc_end,
+					      hash[1] + 3, ctx->max_len - 3);
+	}
+
+	/* Compare two matches and select the best one */
+	if (len1 < len2) {
+		ctx->best_match = hash[1];
+		len1 = len2;
+	} else {
+		ctx->best_match = hash[0];
+	}
+
+	hash[1] = hash[0];
+	hash[0] = src;
+	return len1;
+}
+
+static size_t longest_match_best(const u8 *src, struct lznt *ctx)
+{
+	size_t max_len;
+	const u8 *ptr;
+
+	if (ctx->unc >= src || !ctx->max_len)
+		return 0;
+
+	max_len = 0;
+	for (ptr = ctx->unc; ptr < src; ++ptr) {
+		size_t len =
+			get_match_len(src, ctx->unc_end, ptr, ctx->max_len);
+		if (len >= max_len) {
+			max_len = len;
+			ctx->best_match = ptr;
+		}
+	}
+
+	return max_len >= 3 ? max_len : 0;
+}
+
+static const size_t s_max_len[] = {
+	0x1002, 0x802, 0x402, 0x202, 0x102, 0x82, 0x42, 0x22, 0x12,
+};
+
+static const size_t s_max_off[] = {
+	0x10, 0x20, 0x40, 0x80, 0x100, 0x200, 0x400, 0x800, 0x1000,
+};
+
+static inline u16 make_pair(size_t offset, size_t len, size_t index)
+{
+	return ((offset - 1) << (12 - index)) |
+	       ((len - 3) & (((1 << (12 - index)) - 1)));
+}
+
+static inline size_t parse_pair(u16 pair, size_t *offset, size_t index)
+{
+	*offset = 1 + (pair >> (12 - index));
+	return 3 + (pair & ((1 << (12 - index)) - 1));
+}
+
+/*
+ * compress_chunk
+ *
+ * returns one of the three values:
+ * 0 - ok, 'cmpr' contains 'cmpr_chunk_size' bytes of compressed data
+ * 1 - input buffer is full zero
+ * -2 - the compressed buffer is too small to hold the compressed data
+ */
+static inline int compress_chunk(size_t (*match)(const u8 *, struct lznt *),
+				 const u8 *unc, const u8 *unc_end, u8 *cmpr,
+				 u8 *cmpr_end, size_t *cmpr_chunk_size,
+				 struct lznt *ctx)
+{
+	size_t cnt = 0;
+	size_t idx = 0;
+	const u8 *up = unc;
+	u8 *cp = cmpr + 3;
+	u8 *cp2 = cmpr + 2;
+	u8 not_zero = 0;
+	/* Control byte of 8-bit values: ( 0 - means byte as is, 1 - short pair ) */
+	u8 ohdr = 0;
+	u8 *last;
+	u16 t16;
+
+	if (unc + LZNT_CHUNK_SIZE < unc_end)
+		unc_end = unc + LZNT_CHUNK_SIZE;
+
+	last = min(cmpr + LZNT_CHUNK_SIZE + sizeof(short), cmpr_end);
+
+	ctx->unc = unc;
+	ctx->unc_end = unc_end;
+	ctx->max_len = s_max_len[0];
+
+	while (up < unc_end) {
+		size_t max_len;
+
+		while (unc + s_max_off[idx] < up)
+			ctx->max_len = s_max_len[++idx];
+
+		// Find match
+		max_len = up + 3 <= unc_end ? (*match)(up, ctx) : 0;
+
+		if (!max_len) {
+			if (cp >= last)
+				goto NotCompressed;
+			not_zero |= *cp++ = *up++;
+		} else if (cp + 1 >= last) {
+			goto NotCompressed;
+		} else {
+			t16 = make_pair(up - ctx->best_match, max_len, idx);
+			*cp++ = t16;
+			*cp++ = t16 >> 8;
+
+			ohdr |= 1 << cnt;
+			up += max_len;
+		}
+
+		cnt = (cnt + 1) & 7;
+		if (!cnt) {
+			*cp2 = ohdr;
+			ohdr = 0;
+			cp2 = cp;
+			cp += 1;
+		}
+	}
+
+	if (cp2 < last)
+		*cp2 = ohdr;
+	else
+		cp -= 1;
+
+	*cmpr_chunk_size = cp - cmpr;
+
+	t16 = (*cmpr_chunk_size - 3) | 0xB000;
+	cmpr[0] = t16;
+	cmpr[1] = t16 >> 8;
+
+	return not_zero ? 0 : LZNT_ERROR_ALL_ZEROS;
+
+NotCompressed:
+
+	if ((cmpr + LZNT_CHUNK_SIZE + sizeof(short)) > last)
+		return -2;
+
+	/*
+	 * Copy non cmpr data
+	 * 0x3FFF == ((LZNT_CHUNK_SIZE + 2 - 3) | 0x3000)
+	 */
+	cmpr[0] = 0xff;
+	cmpr[1] = 0x3f;
+
+	memcpy(cmpr + sizeof(short), unc, LZNT_CHUNK_SIZE);
+	*cmpr_chunk_size = LZNT_CHUNK_SIZE + sizeof(short);
+
+	return 0;
+}
+
+static inline ssize_t decompress_chunk(u8 *unc, u8 *unc_end, const u8 *cmpr,
+				       const u8 *cmpr_end)
+{
+	u8 *up = unc;
+	u8 ch = *cmpr++;
+	size_t bit = 0;
+	size_t index = 0;
+	u16 pair;
+	size_t offset, length;
+
+	/* Do decompression until pointers are inside range */
+	while (up < unc_end && cmpr < cmpr_end) {
+		/* Correct index */
+		while (unc + s_max_off[index] < up)
+			index += 1;
+
+		/* Check the current flag for zero */
+		if (!(ch & (1 << bit))) {
+			/* Just copy byte */
+			*up++ = *cmpr++;
+			goto next;
+		}
+
+		/* Check for boundary */
+		if (cmpr + 1 >= cmpr_end)
+			return -EINVAL;
+
+		/* Read a short from little endian stream */
+		pair = cmpr[1];
+		pair <<= 8;
+		pair |= cmpr[0];
+
+		cmpr += 2;
+
+		/* Translate packed information into offset and length */
+		length = parse_pair(pair, &offset, index);
+
+		/* Check offset for boundary */
+		if (unc + offset > up)
+			return -EINVAL;
+
+		/* Truncate the length if necessary */
+		if (up + length >= unc_end)
+			length = unc_end - up;
+
+		/* Now we copy bytes. This is the heart of LZ algorithm. */
+		for (; length > 0; length--, up++)
+			*up = *(up - offset);
+
+next:
+		/* Advance flag bit value */
+		bit = (bit + 1) & 7;
+
+		if (!bit) {
+			if (cmpr >= cmpr_end)
+				break;
+
+			ch = *cmpr++;
+		}
+	}
+
+	/* return the size of uncompressed data */
+	return up - unc;
+}
+
+/*
+ * std = true - standard compression
+ * std = false - best compression, requires a lot of cpu
+ */
+struct lznt *get_compression_ctx(bool std)
+{
+	struct lznt *r = ntfs_alloc(
+		std ? sizeof(struct lznt) : offsetof(struct lznt, hash), 1);
+
+	if (r)
+		r->std = std;
+	return r;
+}
+
+/*
+ * compress_lznt
+ *
+ * Compresses "unc" into "cmpr"
+ * +x - ok, 'cmpr' contains 'final_compressed_size' bytes of compressed data
+ * 0 - input buffer is full zero
+ */
+size_t compress_lznt(const void *unc, size_t unc_size, void *cmpr,
+		     size_t cmpr_size, struct lznt *ctx)
+{
+	int err;
+	size_t (*match)(const u8 *src, struct lznt *ctx);
+	u8 *p = cmpr;
+	u8 *end = p + cmpr_size;
+	const u8 *unc_chunk = unc;
+	const u8 *unc_end = unc_chunk + unc_size;
+	bool is_zero = true;
+
+	if (ctx->std) {
+		match = &longest_match_std;
+		memset(ctx->hash, 0, sizeof(ctx->hash));
+	} else {
+		match = &longest_match_best;
+	}
+
+	/* compression cycle */
+	for (; unc_chunk < unc_end; unc_chunk += LZNT_CHUNK_SIZE) {
+		cmpr_size = 0;
+		err = compress_chunk(match, unc_chunk, unc_end, p, end,
+				     &cmpr_size, ctx);
+		if (err < 0)
+			return unc_size;
+
+		if (is_zero && err != LZNT_ERROR_ALL_ZEROS)
+			is_zero = false;
+
+		p += cmpr_size;
+	}
+
+	if (p <= end - 2)
+		p[0] = p[1] = 0;
+
+	return is_zero ? 0 : PtrOffset(cmpr, p);
+}
+
+/*
+ * decompress_lznt
+ *
+ * decompresses "cmpr" into "unc"
+ */
+ssize_t decompress_lznt(const void *cmpr, size_t cmpr_size, void *unc,
+			size_t unc_size)
+{
+	const u8 *cmpr_chunk = cmpr;
+	const u8 *cmpr_end = cmpr_chunk + cmpr_size;
+	u8 *unc_chunk = unc;
+	u8 *unc_end = unc_chunk + unc_size;
+	u16 chunk_hdr;
+
+	if (cmpr_size < sizeof(short))
+		return -EINVAL;
+
+	/* read chunk header */
+	chunk_hdr = cmpr_chunk[1];
+	chunk_hdr <<= 8;
+	chunk_hdr |= cmpr_chunk[0];
+
+	/* loop through decompressing chunks */
+	for (;;) {
+		size_t chunk_size_saved;
+		size_t unc_use;
+		size_t cmpr_use = 3 + (chunk_hdr & (LZNT_CHUNK_SIZE - 1));
+
+		/* Check that the chunk actually fits the supplied buffer */
+		if (cmpr_chunk + cmpr_use > cmpr_end)
+			return -EINVAL;
+
+		/* First make sure the chunk contains compressed data */
+		if (chunk_hdr & 0x8000) {
+			/* Decompress a chunk and return if we get an error */
+			ssize_t err =
+				decompress_chunk(unc_chunk, unc_end,
+						 cmpr_chunk + sizeof(chunk_hdr),
+						 cmpr_chunk + cmpr_use);
+			if (err < 0)
+				return err;
+			unc_use = err;
+		} else {
+			/* This chunk does not contain compressed data */
+			unc_use = unc_chunk + LZNT_CHUNK_SIZE > unc_end ?
+					  unc_end - unc_chunk :
+					  LZNT_CHUNK_SIZE;
+
+			if (cmpr_chunk + sizeof(chunk_hdr) + unc_use >
+			    cmpr_end) {
+				return -EINVAL;
+			}
+
+			memcpy(unc_chunk, cmpr_chunk + sizeof(chunk_hdr),
+			       unc_use);
+		}
+
+		/* Advance pointers */
+		cmpr_chunk += cmpr_use;
+		unc_chunk += unc_use;
+
+		/* Check for the end of unc buffer */
+		if (unc_chunk >= unc_end)
+			break;
+
+		/* Proceed the next chunk */
+		if (cmpr_chunk > cmpr_end - 2)
+			break;
+
+		chunk_size_saved = LZNT_CHUNK_SIZE;
+
+		/* read chunk header */
+		chunk_hdr = cmpr_chunk[1];
+		chunk_hdr <<= 8;
+		chunk_hdr |= cmpr_chunk[0];
+
+		if (!chunk_hdr)
+			break;
+
+		/* Check the size of unc buffer */
+		if (unc_use < chunk_size_saved) {
+			size_t t1 = chunk_size_saved - unc_use;
+			u8 *t2 = unc_chunk + t1;
+
+			/* 'Zero' memory */
+			if (t2 >= unc_end)
+				break;
+
+			memset(unc_chunk, 0, t1);
+			unc_chunk = t2;
+		}
+	}
+
+	/* Check compression boundary */
+	if (cmpr_chunk > cmpr_end)
+		return -EINVAL;
+
+	/*
+	 * The unc size is just a difference between current
+	 * pointer and original one
+	 */
+	return PtrOffset(unc, unc_chunk);
+}
-- 
2.25.4


^ permalink raw reply related	[flat|nested] 21+ messages in thread

* [PATCH v11 07/10] fs/ntfs3: Add NTFS journal
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
                   ` (5 preceding siblings ...)
  2020-10-30 15:02 ` [PATCH v11 06/10] fs/ntfs3: Add compression Konstantin Komarov
@ 2020-10-30 15:02 ` Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 08/10] fs/ntfs3: Add Kconfig, Makefile and doc Konstantin Komarov
                   ` (5 subsequent siblings)
  12 siblings, 0 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This adds NTFS journal

Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
---
 fs/ntfs3/fslog.c | 5221 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 5221 insertions(+)
 create mode 100644 fs/ntfs3/fslog.c

diff --git a/fs/ntfs3/fslog.c b/fs/ntfs3/fslog.c
new file mode 100644
index 000000000000..7335bb6f8b24
--- /dev/null
+++ b/fs/ntfs3/fslog.c
@@ -0,0 +1,5221 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *
+ * Copyright (C) 2019-2020 Paragon Software GmbH, All rights reserved.
+ *
+ */
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/fs.h>
+#include <linux/hash.h>
+#include <linux/nls.h>
+#include <linux/random.h>
+#include <linux/ratelimit.h>
+#include <linux/slab.h>
+
+#include "debug.h"
+#include "ntfs.h"
+#include "ntfs_fs.h"
+
+/*
+ * LOG FILE structs
+ */
+
+// clang-format off
+
+#define MaxLogFileSize     0x100000000ull
+#define DefaultLogPageSize 4096
+#define MinLogRecordPages  0x30
+
+struct RESTART_HDR {
+	struct NTFS_RECORD_HEADER rhdr; // 'RSTR'
+	__le32 sys_page_size; // 0x10: Page size of the system which initialized the log
+	__le32 page_size;     // 0x14: Log page size used for this log file
+	__le16 ra_off;        // 0x18:
+	__le16 minor_ver;     // 0x1A:
+	__le16 major_ver;     // 0x1C:
+	__le16 fixups[1];
+};
+
+#define LFS_NO_CLIENT 0xffff
+#define LFS_NO_CLIENT_LE cpu_to_le16(0xffff)
+
+struct CLIENT_REC {
+	__le64 oldest_lsn;
+	__le64 restart_lsn; // 0x08:
+	__le16 prev_client; // 0x10:
+	__le16 next_client; // 0x12:
+	__le16 seq_num;     // 0x14:
+	u8 align[6];        // 0x16
+	__le32 name_bytes;  // 0x1C: in bytes
+	__le16 name[32];    // 0x20: name of client
+};
+
+static_assert(sizeof(struct CLIENT_REC) == 0x60);
+
+/* Two copies of these will exist at the beginning of the log file */
+struct RESTART_AREA {
+	__le64 current_lsn;    // 0x00: Current logical end of log file
+	__le16 log_clients;    // 0x08: Maximum number of clients
+	__le16 client_idx[2];  // 0x0A: free/use index into the client record arrays
+	__le16 flags;          // 0x0E: See RESTART_SINGLE_PAGE_IO
+	__le32 seq_num_bits;   // 0x10: the number of bits in sequence number.
+	__le16 ra_len;         // 0x14:
+	__le16 client_off;     // 0x16:
+	__le64 l_size;         // 0x18: Usable log file size.
+	__le32 last_lsn_data_len; // 0x20:
+	__le16 rec_hdr_len;    // 0x24: log page data offset
+	__le16 data_off;       // 0x26: log page data length
+	__le32 open_log_count; // 0x28:
+	__le32 align[5];       // 0x2C:
+	struct CLIENT_REC clients[1]; // 0x40:
+};
+
+struct LOG_REC_HDR {
+	__le16 redo_op;      // 0x00:  NTFS_LOG_OPERATION
+	__le16 undo_op;      // 0x02:  NTFS_LOG_OPERATION
+	__le16 redo_off;     // 0x04:  Offset to Redo record
+	__le16 redo_len;     // 0x06:  Redo length
+	__le16 undo_off;     // 0x08:  Offset to Undo record
+	__le16 undo_len;     // 0x0A:  Undo length
+	__le16 target_attr;  // 0x0C:
+	__le16 lcns_follow;  // 0x0E:
+	__le16 record_off;   // 0x10:
+	__le16 attr_off;     // 0x12:
+	__le16 cluster_off;  // 0x14:
+	__le16 reserved;     // 0x16:
+	__le64 target_vcn;   // 0x18:
+	__le64 page_lcns[1]; // 0x20:
+};
+
+static_assert(sizeof(struct LOG_REC_HDR) == 0x28);
+
+#define RESTART_ENTRY_ALLOCATED    0xFFFFFFFF
+#define RESTART_ENTRY_ALLOCATED_LE cpu_to_le32(0xFFFFFFFF)
+
+struct RESTART_TABLE {
+	__le16 size;       // 0x00:  In bytes
+	__le16 used;       // 0x02: entries
+	__le16 total;      // 0x04: entries
+	__le16 res[3];     // 0x06:
+	__le32 free_goal;  // 0x0C:
+	__le32 first_free; // 0x10
+	__le32 last_free;  // 0x14
+
+};
+
+static_assert(sizeof(struct RESTART_TABLE) == 0x18);
+
+struct ATTR_NAME_ENTRY {
+	__le16 off; // offset in the Open attribute Table
+	__le16 name_bytes;
+	__le16 name[1];
+};
+
+struct OPEN_ATTR_ENRTY {
+	__le32 next;            // 0x00: RESTART_ENTRY_ALLOCATED if allocated
+	__le32 bytes_per_index; // 0x04:
+	enum ATTR_TYPE type;    // 0x08:
+	u8 is_dirty_pages;      // 0x0C:
+	u8 is_attr_name;        // 0x0B: Faked field to manage 'ptr'
+	u8 name_len;            // 0x0C: Faked field to manage 'ptr'
+	u8 res;
+	struct MFT_REF ref; // 0x10: File Reference of file containing attribute
+	__le64 open_record_lsn; // 0x18:
+	void *ptr;              // 0x20:
+};
+
+/* 32 bit version of 'struct OPEN_ATTR_ENRTY' */
+struct OPEN_ATTR_ENRTY_32 {
+	__le32 next;            // 0x00: RESTART_ENTRY_ALLOCATED if allocated
+	__le32 ptr;             // 0x04:
+	struct MFT_REF ref;     // 0x08:
+	__le64 open_record_lsn; // 0x10:
+	u8 is_dirty_pages;      // 0x18:
+	u8 is_attr_name;        // 0x19
+	u8 res1[2];
+	enum ATTR_TYPE type;    // 0x1C:
+	u8 name_len;            // 0x20:  in wchar
+	u8 res2[3];
+	__le32 AttributeName;   // 0x24:
+	__le32 bytes_per_index; // 0x28:
+};
+
+#define SIZEOF_OPENATTRIBUTEENTRY0 0x2c
+// static_assert( 0x2C == sizeof(struct OPEN_ATTR_ENRTY_32) );
+static_assert(sizeof(struct OPEN_ATTR_ENRTY) < SIZEOF_OPENATTRIBUTEENTRY0);
+
+/*
+ * One entry exists in the Dirty Pages Table for each page which is dirty at the
+ * time the Restart Area is written
+ */
+struct DIR_PAGE_ENTRY {
+	__le32 next;         // 0x00:  RESTART_ENTRY_ALLOCATED if allocated
+	__le32 target_attr;  // 0x04:  Index into the Open attribute Table
+	__le32 transfer_len; // 0x08:
+	__le32 lcns_follow;  // 0x0C:
+	__le64 vcn;          // 0x10:  Vcn of dirty page
+	__le64 oldest_lsn;   // 0x18:
+	__le64 page_lcns[1]; // 0x20:
+};
+
+static_assert(sizeof(struct DIR_PAGE_ENTRY) == 0x28);
+
+/* 32 bit version of 'struct DIR_PAGE_ENTRY' */
+struct DIR_PAGE_ENTRY_32 {
+	__le32 next;         // 0x00:  RESTART_ENTRY_ALLOCATED if allocated
+	__le32 target_attr;  // 0x04:  Index into the Open attribute Table
+	__le32 transfer_len; // 0x08:
+	__le32 lcns_follow;  // 0x0C:
+	__le32 reserved;     // 0x10:
+	__le32 vcn_low;      // 0x14:  Vcn of dirty page
+	__le32 vcn_hi;       // 0x18:  Vcn of dirty page
+	__le32 oldest_lsn_low; // 0x1C:
+	__le32 oldest_lsn_hi; // 0x1C:
+	__le32 page_lcns_low; // 0x24:
+	__le32 page_lcns_hi; // 0x24:
+};
+
+static_assert(offsetof(struct DIR_PAGE_ENTRY_32, vcn_low) == 0x14);
+static_assert(sizeof(struct DIR_PAGE_ENTRY_32) == 0x2c);
+
+enum transact_state {
+	TransactionUninitialized = 0,
+	TransactionActive,
+	TransactionPrepared,
+	TransactionCommitted
+};
+
+struct TRANSACTION_ENTRY {
+	__le32 next;          // 0x00: RESTART_ENTRY_ALLOCATED if allocated
+	u8 transact_state;    // 0x04:
+	u8 reserved[3];       // 0x05:
+	__le64 first_lsn;     // 0x08:
+	__le64 prev_lsn;      // 0x10:
+	__le64 undo_next_lsn; // 0x18:
+	__le32 undo_records;  // 0x20: Number of undo log records pending abort
+	__le32 undo_len;      // 0x24: Total undo size
+};
+
+static_assert(sizeof(struct TRANSACTION_ENTRY) == 0x28);
+
+struct NTFS_RESTART {
+	__le32 major_ver;             // 0x00:
+	__le32 minor_ver;             // 0x04:
+	__le64 check_point_start;     // 0x08:
+	__le64 open_attr_table_lsn;   // 0x10:
+	__le64 attr_names_lsn;        // 0x18:
+	__le64 dirty_pages_table_lsn; // 0x20:
+	__le64 transact_table_lsn;    // 0x28:
+	__le32 open_attr_len;         // 0x30: In bytes
+	__le32 attr_names_len;        // 0x34: In bytes
+	__le32 dirty_pages_len;       // 0x38: In bytes
+	__le32 transact_table_len;    // 0x3C: In bytes
+};
+
+static_assert(sizeof(struct NTFS_RESTART) == 0x40);
+
+struct NEW_ATTRIBUTE_SIZES {
+	__le64 alloc_size;
+	__le64 valid_size;
+	__le64 data_size;
+	__le64 total_size;
+};
+
+struct BITMAP_RANGE {
+	__le32 bitmap_off;
+	__le32 bits;
+};
+
+struct LCN_RANGE {
+	__le64 lcn;
+	__le64 len;
+};
+
+/* The following type defines the different log record types */
+#define LfsClientRecord  cpu_to_le32(1)
+#define LfsClientRestart cpu_to_le32(2)
+
+/* This is used to uniquely identify a client for a particular log file */
+struct CLIENT_ID {
+	__le16 seq_num;
+	__le16 client_idx;
+};
+
+/* This is the header that begins every Log Record in the log file */
+struct LFS_RECORD_HDR {
+	__le64 this_lsn;    // 0x00:
+	__le64 client_prev_lsn;  // 0x08:
+	__le64 client_undo_next_lsn; // 0x10:
+	__le32 client_data_len;  // 0x18:
+	struct CLIENT_ID client; // 0x1C: Owner of this log record
+	__le32 record_type; // 0x20: LfsClientRecord or LfsClientRestart
+	__le32 transact_id; // 0x24:
+	__le16 flags;       // 0x28:	LOG_RECORD_MULTI_PAGE
+	u8 align[6];        // 0x2A:
+};
+
+#define LOG_RECORD_MULTI_PAGE cpu_to_le16(1)
+
+static_assert(sizeof(struct LFS_RECORD_HDR) == 0x30);
+
+struct LFS_RECORD {
+	__le16 next_record_off; // 0x00: Offset of the free space in the page
+	u8 align[6];         // 0x02:
+	__le64 last_end_lsn; // 0x08: lsn for the last log record which ends on the page
+};
+
+static_assert(sizeof(struct LFS_RECORD) == 0x10);
+
+struct RECORD_PAGE_HDR {
+	struct NTFS_RECORD_HEADER rhdr; // 'RCRD'
+	__le32 rflags;     // 0x10:  See LOG_PAGE_LOG_RECORD_END
+	__le16 page_count; // 0x14:
+	__le16 page_pos;   // 0x16:
+	struct LFS_RECORD record_hdr; // 0x18
+	__le16 fixups[10]; // 0x28
+	__le32 file_off;   // 0x3c: used when major version >= 2
+};
+
+// clang-format on
+
+// Page contains the end of a log record
+#define LOG_PAGE_LOG_RECORD_END cpu_to_le32(0x00000001)
+
+static inline bool is_log_record_end(const struct RECORD_PAGE_HDR *hdr)
+{
+	return hdr->rflags & LOG_PAGE_LOG_RECORD_END;
+}
+
+static_assert(offsetof(struct RECORD_PAGE_HDR, file_off) == 0x3c);
+
+/*
+ * END of NTFS LOG structures
+ */
+
+/* Define some tuning parameters to keep the restart tables a reasonable size */
+#define INITIAL_NUMBER_TRANSACTIONS 5
+
+enum NTFS_LOG_OPERATION {
+
+	Noop = 0x00,
+	CompensationLogRecord = 0x01,
+	InitializeFileRecordSegment = 0x02,
+	DeallocateFileRecordSegment = 0x03,
+	WriteEndOfFileRecordSegment = 0x04,
+	CreateAttribute = 0x05,
+	DeleteAttribute = 0x06,
+	UpdateResidentValue = 0x07,
+	UpdateNonresidentValue = 0x08,
+	UpdateMappingPairs = 0x09,
+	DeleteDirtyClusters = 0x0A,
+	SetNewAttributeSizes = 0x0B,
+	AddIndexEntryRoot = 0x0C,
+	DeleteIndexEntryRoot = 0x0D,
+	AddIndexEntryAllocation = 0x0E,
+	DeleteIndexEntryAllocation = 0x0F,
+	WriteEndOfIndexBuffer = 0x10,
+	SetIndexEntryVcnRoot = 0x11,
+	SetIndexEntryVcnAllocation = 0x12,
+	UpdateFileNameRoot = 0x13,
+	UpdateFileNameAllocation = 0x14,
+	SetBitsInNonresidentBitMap = 0x15,
+	ClearBitsInNonresidentBitMap = 0x16,
+	HotFix = 0x17,
+	EndTopLevelAction = 0x18,
+	PrepareTransaction = 0x19,
+	CommitTransaction = 0x1A,
+	ForgetTransaction = 0x1B,
+	OpenNonresidentAttribute = 0x1C,
+	OpenAttributeTableDump = 0x1D,
+	AttributeNamesDump = 0x1E,
+	DirtyPageTableDump = 0x1F,
+	TransactionTableDump = 0x20,
+	UpdateRecordDataRoot = 0x21,
+	UpdateRecordDataAllocation = 0x22,
+
+	UpdateRelativeDataInIndex =
+		0x23, // NtOfsRestartUpdateRelativeDataInIndex
+	UpdateRelativeDataInIndex2 = 0x24,
+	ZeroEndOfFileRecord = 0x25,
+};
+
+/*
+ * Array for log records which require a target attribute
+ * A true indicates that the corresponding restart operation requires a target attribute
+ */
+static const u8 AttributeRequired[] = {
+	0xFC, 0xFB, 0xFF, 0x10, 0x06,
+};
+
+static inline bool is_target_required(u16 op)
+{
+	bool ret = op <= UpdateRecordDataAllocation &&
+		   (AttributeRequired[op >> 3] >> (op & 7) & 1);
+	return ret;
+}
+
+static inline bool can_skip_action(enum NTFS_LOG_OPERATION op)
+{
+	switch (op) {
+	case Noop:
+	case DeleteDirtyClusters:
+	case HotFix:
+	case EndTopLevelAction:
+	case PrepareTransaction:
+	case CommitTransaction:
+	case ForgetTransaction:
+	case CompensationLogRecord:
+	case OpenNonresidentAttribute:
+	case OpenAttributeTableDump:
+	case AttributeNamesDump:
+	case DirtyPageTableDump:
+	case TransactionTableDump:
+		return true;
+	default:
+		return false;
+	}
+}
+
+enum { lcb_ctx_undo_next, lcb_ctx_prev, lcb_ctx_next };
+
+/* bytes per restart table */
+static inline u32 bytes_per_rt(const struct RESTART_TABLE *rt)
+{
+	return le16_to_cpu(rt->used) * le16_to_cpu(rt->size) +
+	       sizeof(struct RESTART_TABLE);
+}
+
+/* log record length */
+static inline u32 lrh_length(const struct LOG_REC_HDR *lr)
+{
+	u16 t16 = le16_to_cpu(lr->lcns_follow);
+
+	return t16 > 1 ? sizeof(struct LOG_REC_HDR) + (t16 - 1) * sizeof(u64) :
+			 sizeof(struct LOG_REC_HDR);
+}
+
+struct lcb {
+	struct LFS_RECORD_HDR *lrh; // Log record header of the current lsn
+	struct LOG_REC_HDR *log_rec;
+	u32 ctx_mode; // lcb_ctx_undo_next/lcb_ctx_prev/lcb_ctx_next
+	struct CLIENT_ID client;
+	bool alloc; // if true the we should deallocate 'log_rec'
+};
+
+static void lcb_put(struct lcb *lcb)
+{
+	if (lcb->alloc)
+		ntfs_free(lcb->log_rec);
+	ntfs_free(lcb->lrh);
+	ntfs_free(lcb);
+}
+
+/*
+ * oldest_client_lsn
+ *
+ * find the oldest lsn from active clients.
+ */
+static inline void oldest_client_lsn(const struct CLIENT_REC *ca,
+				     __le16 next_client, u64 *oldest_lsn)
+{
+	while (next_client != LFS_NO_CLIENT_LE) {
+		const struct CLIENT_REC *cr = ca + le16_to_cpu(next_client);
+		u64 lsn = le64_to_cpu(cr->oldest_lsn);
+
+		/* ignore this block if it's oldest lsn is 0 */
+		if (lsn && lsn < *oldest_lsn)
+			*oldest_lsn = lsn;
+
+		next_client = cr->next_client;
+	}
+}
+
+static inline bool is_rst_page_hdr_valid(u32 file_off,
+					 const struct RESTART_HDR *rhdr)
+{
+	u32 sys_page = le32_to_cpu(rhdr->sys_page_size);
+	u32 page_size = le32_to_cpu(rhdr->page_size);
+	u32 end_usa;
+	u16 ro;
+
+	if (sys_page < SECTOR_SIZE || page_size < SECTOR_SIZE ||
+	    sys_page & (sys_page - 1) || page_size & (page_size - 1)) {
+		return false;
+	}
+
+	/* Check that if the file offset isn't 0, it is the system page size */
+	if (file_off && file_off != sys_page)
+		return false;
+
+	/* Check support version 1.1+ */
+	if (le16_to_cpu(rhdr->major_ver) <= 1 && !rhdr->minor_ver)
+		return false;
+
+	if (le16_to_cpu(rhdr->major_ver) > 2)
+		return false;
+
+	ro = le16_to_cpu(rhdr->ra_off);
+	if (!IsQuadAligned(ro) || ro > sys_page)
+		return false;
+
+	end_usa = ((sys_page >> SECTOR_SHIFT) + 1) * sizeof(short);
+	end_usa += le16_to_cpu(rhdr->rhdr.fix_off);
+
+	if (ro < end_usa)
+		return false;
+
+	return true;
+}
+
+static inline bool is_rst_area_valid(const struct RESTART_HDR *rhdr)
+{
+	const struct RESTART_AREA *ra;
+	u16 cl, fl, ul;
+	u32 off, l_size, file_dat_bits, file_size_round;
+	u16 ro = le16_to_cpu(rhdr->ra_off);
+	u32 sys_page = le32_to_cpu(rhdr->sys_page_size);
+
+	if (ro + offsetof(struct RESTART_AREA, l_size) >
+	    SECTOR_SIZE - sizeof(short))
+		return false;
+
+	ra = Add2Ptr(rhdr, ro);
+	cl = le16_to_cpu(ra->log_clients);
+
+	if (cl > 1)
+		return false;
+
+	off = le16_to_cpu(ra->client_off);
+
+	if (!IsQuadAligned(off) || ro + off > SECTOR_SIZE - sizeof(short))
+		return false;
+
+	off += cl * sizeof(struct CLIENT_REC);
+
+	if (off > sys_page)
+		return false;
+
+	/*
+	 * Check the restart length field and whether the entire
+	 * restart area is contained that length
+	 */
+	if (le16_to_cpu(rhdr->ra_off) + le16_to_cpu(ra->ra_len) > sys_page ||
+	    off > le16_to_cpu(ra->ra_len)) {
+		return false;
+	}
+
+	/*
+	 * As a final check make sure that the use list and the free list
+	 * are either empty or point to a valid client
+	 */
+	fl = le16_to_cpu(ra->client_idx[0]);
+	ul = le16_to_cpu(ra->client_idx[1]);
+	if ((fl != LFS_NO_CLIENT && fl >= cl) ||
+	    (ul != LFS_NO_CLIENT && ul >= cl))
+		return false;
+
+	/* Make sure the sequence number bits match the log file size */
+	l_size = le64_to_cpu(ra->l_size);
+
+	file_dat_bits = sizeof(u64) * 8 - le32_to_cpu(ra->seq_num_bits);
+	file_size_round = 1u << (file_dat_bits + 3);
+	if (file_size_round != l_size &&
+	    (file_size_round < l_size || (file_size_round / 2) > l_size)) {
+		return false;
+	}
+
+	/* The log page data offset and record header length must be quad-aligned */
+	if (!IsQuadAligned(le16_to_cpu(ra->data_off)) ||
+	    !IsQuadAligned(le16_to_cpu(ra->rec_hdr_len)))
+		return false;
+
+	return true;
+}
+
+static inline bool is_client_area_valid(const struct RESTART_HDR *rhdr,
+					bool usa_error)
+{
+	u16 ro = le16_to_cpu(rhdr->ra_off);
+	const struct RESTART_AREA *ra = Add2Ptr(rhdr, ro);
+	u16 ra_len = le16_to_cpu(ra->ra_len);
+	const struct CLIENT_REC *ca;
+	u32 i;
+
+	if (usa_error && ra_len + ro > SECTOR_SIZE - sizeof(short))
+		return false;
+
+	/* Find the start of the client array */
+	ca = Add2Ptr(ra, le16_to_cpu(ra->client_off));
+
+	/*
+	 * Start with the free list
+	 * Check that all the clients are valid and that there isn't a cycle
+	 * Do the in-use list on the second pass
+	 */
+	for (i = 0; i < 2; i++) {
+		u16 client_idx = le16_to_cpu(ra->client_idx[i]);
+		bool first_client = true;
+		u16 clients = le16_to_cpu(ra->log_clients);
+
+		while (client_idx != LFS_NO_CLIENT) {
+			const struct CLIENT_REC *cr;
+
+			if (!clients ||
+			    client_idx >= le16_to_cpu(ra->log_clients))
+				return false;
+
+			clients -= 1;
+			cr = ca + client_idx;
+
+			client_idx = le16_to_cpu(cr->next_client);
+
+			if (first_client) {
+				first_client = false;
+				if (cr->prev_client != LFS_NO_CLIENT_LE)
+					return false;
+			}
+		}
+	}
+
+	return true;
+}
+
+/*
+ * remove_client
+ *
+ * remove a client record from a client record list an restart area
+ */
+static inline void remove_client(struct CLIENT_REC *ca,
+				 const struct CLIENT_REC *cr, __le16 *head)
+{
+	if (cr->prev_client == LFS_NO_CLIENT_LE)
+		*head = cr->next_client;
+	else
+		ca[le16_to_cpu(cr->prev_client)].next_client = cr->next_client;
+
+	if (cr->next_client != LFS_NO_CLIENT_LE)
+		ca[le16_to_cpu(cr->next_client)].prev_client = cr->prev_client;
+}
+
+/*
+ * add_client
+ *
+ * add a client record to the start of a list
+ */
+static inline void add_client(struct CLIENT_REC *ca, u16 index, __le16 *head)
+{
+	struct CLIENT_REC *cr = ca + index;
+
+	cr->prev_client = LFS_NO_CLIENT_LE;
+	cr->next_client = *head;
+
+	if (*head != LFS_NO_CLIENT_LE)
+		ca[le16_to_cpu(*head)].prev_client = cpu_to_le16(index);
+
+	*head = cpu_to_le16(index);
+}
+
+/*
+ * enum_rstbl
+ *
+ */
+static inline void *enum_rstbl(struct RESTART_TABLE *t, void *c)
+{
+	__le32 *e;
+	u32 bprt;
+	u16 rsize = t ? le16_to_cpu(t->size) : 0;
+
+	if (!c) {
+		if (!t || !t->total)
+			return NULL;
+		e = Add2Ptr(t, sizeof(struct RESTART_TABLE));
+	} else {
+		e = Add2Ptr(c, rsize);
+	}
+
+	/* Loop until we hit the first one allocated, or the end of the list */
+	for (bprt = bytes_per_rt(t); PtrOffset(t, e) < bprt;
+	     e = Add2Ptr(e, rsize)) {
+		if (*e == RESTART_ENTRY_ALLOCATED_LE)
+			return e;
+	}
+	return NULL;
+}
+
+/*
+ * find_dp
+ *
+ * searches for a 'vcn' in Dirty Page Table,
+ */
+static inline struct DIR_PAGE_ENTRY *find_dp(struct RESTART_TABLE *dptbl,
+					     u32 target_attr, u64 vcn)
+{
+	__le32 ta = cpu_to_le32(target_attr);
+	struct DIR_PAGE_ENTRY *dp = NULL;
+
+	while ((dp = enum_rstbl(dptbl, dp))) {
+		u64 dp_vcn = le64_to_cpu(dp->vcn);
+
+		if (dp->target_attr == ta && vcn >= dp_vcn &&
+		    vcn < dp_vcn + le32_to_cpu(dp->lcns_follow)) {
+			return dp;
+		}
+	}
+	return NULL;
+}
+
+static inline u32 norm_file_page(u32 page_size, u32 *l_size, bool use_default)
+{
+	if (use_default)
+		page_size = DefaultLogPageSize;
+
+	/* Round the file size down to a system page boundary */
+	*l_size &= ~(page_size - 1);
+
+	/* File should contain at least 2 restart pages and MinLogRecordPages pages */
+	if (*l_size < (MinLogRecordPages + 2) * page_size)
+		return 0;
+
+	return page_size;
+}
+
+static bool check_log_rec(const struct LOG_REC_HDR *lr, u32 bytes, u32 tr,
+			  u32 bytes_per_attr_entry)
+{
+	u16 t16;
+
+	if (bytes < sizeof(struct LOG_REC_HDR))
+		return false;
+	if (!tr)
+		return false;
+
+	if ((tr - sizeof(struct RESTART_TABLE)) %
+	    sizeof(struct TRANSACTION_ENTRY))
+		return false;
+
+	if (le16_to_cpu(lr->redo_off) & 7)
+		return false;
+
+	if (le16_to_cpu(lr->undo_off) & 7)
+		return false;
+
+	if (lr->target_attr)
+		goto check_lcns;
+
+	if (is_target_required(le16_to_cpu(lr->redo_op)))
+		return false;
+
+	if (is_target_required(le16_to_cpu(lr->undo_op)))
+		return false;
+
+check_lcns:
+	if (!lr->lcns_follow)
+		goto check_length;
+
+	t16 = le16_to_cpu(lr->target_attr);
+	if ((t16 - sizeof(struct RESTART_TABLE)) % bytes_per_attr_entry)
+		return false;
+
+check_length:
+	if (bytes < lrh_length(lr))
+		return false;
+
+	return true;
+}
+
+static bool check_rstbl(const struct RESTART_TABLE *rt, size_t bytes)
+{
+	u32 ts;
+	u32 i, off;
+	u16 rsize = le16_to_cpu(rt->size);
+	u16 ne = le16_to_cpu(rt->used);
+	u32 ff = le32_to_cpu(rt->first_free);
+	u32 lf = le32_to_cpu(rt->last_free);
+
+	ts = rsize * ne + sizeof(struct RESTART_TABLE);
+
+	if (!rsize || rsize > bytes ||
+	    rsize + sizeof(struct RESTART_TABLE) > bytes || bytes < ts ||
+	    le16_to_cpu(rt->total) > ne || ff > ts || lf > ts ||
+	    (ff && ff < sizeof(struct RESTART_TABLE)) ||
+	    (lf && lf < sizeof(struct RESTART_TABLE))) {
+		return false;
+	}
+
+	/* Verify each entry is either allocated or points
+	 * to a valid offset the table
+	 */
+	for (i = 0; i < ne; i++) {
+		off = le32_to_cpu(*(__le32 *)Add2Ptr(
+			rt, i * rsize + sizeof(struct RESTART_TABLE)));
+
+		if (off != RESTART_ENTRY_ALLOCATED && off &&
+		    (off < sizeof(struct RESTART_TABLE) ||
+		     ((off - sizeof(struct RESTART_TABLE)) % rsize))) {
+			return false;
+		}
+	}
+
+	/* Walk through the list headed by the first entry to make
+	 * sure none of the entries are currently being used
+	 */
+	for (off = ff; off;) {
+		if (off == RESTART_ENTRY_ALLOCATED)
+			return false;
+
+		off = le32_to_cpu(*(__le32 *)Add2Ptr(rt, off));
+	}
+
+	return true;
+}
+
+/*
+ * free_rsttbl_idx
+ *
+ * frees a previously allocated index a Restart Table.
+ */
+static inline void free_rsttbl_idx(struct RESTART_TABLE *rt, u32 off)
+{
+	__le32 *e;
+	u32 lf = le32_to_cpu(rt->last_free);
+	__le32 off_le = cpu_to_le32(off);
+
+	e = Add2Ptr(rt, off);
+
+	if (off < le32_to_cpu(rt->free_goal)) {
+		*e = rt->first_free;
+		rt->first_free = off_le;
+		if (!lf)
+			rt->last_free = off_le;
+	} else {
+		if (lf)
+			*(__le32 *)Add2Ptr(rt, lf) = off_le;
+		else
+			rt->first_free = off_le;
+
+		rt->last_free = off_le;
+		*e = 0;
+	}
+
+	le16_sub_cpu(&rt->total, 1);
+}
+
+static inline struct RESTART_TABLE *init_rsttbl(u16 esize, u16 used)
+{
+	__le32 *e, *last_free;
+	u32 off;
+	u32 bytes = esize * used + sizeof(struct RESTART_TABLE);
+	u32 lf = sizeof(struct RESTART_TABLE) + (used - 1) * esize;
+	struct RESTART_TABLE *t = ntfs_alloc(bytes, 1);
+
+	t->size = cpu_to_le16(esize);
+	t->used = cpu_to_le16(used);
+	t->free_goal = cpu_to_le32(~0u);
+	t->first_free = cpu_to_le32(sizeof(struct RESTART_TABLE));
+	t->last_free = cpu_to_le32(lf);
+
+	e = (__le32 *)(t + 1);
+	last_free = Add2Ptr(t, lf);
+
+	for (off = sizeof(struct RESTART_TABLE) + esize; e < last_free;
+	     e = Add2Ptr(e, esize), off += esize) {
+		*e = cpu_to_le32(off);
+	}
+	return t;
+}
+
+static inline struct RESTART_TABLE *extend_rsttbl(struct RESTART_TABLE *tbl,
+						  u32 add, u32 free_goal)
+{
+	u16 esize = le16_to_cpu(tbl->size);
+	__le32 osize = cpu_to_le32(bytes_per_rt(tbl));
+	u32 used = le16_to_cpu(tbl->used);
+	struct RESTART_TABLE *rt = init_rsttbl(esize, used + add);
+
+	memcpy(rt + 1, tbl + 1, esize * used);
+
+	rt->free_goal = free_goal == ~0u ?
+				cpu_to_le32(~0u) :
+				cpu_to_le32(sizeof(struct RESTART_TABLE) +
+					    free_goal * esize);
+
+	if (tbl->first_free) {
+		rt->first_free = tbl->first_free;
+		*(__le32 *)Add2Ptr(rt, le32_to_cpu(tbl->last_free)) = osize;
+	} else {
+		rt->first_free = osize;
+	}
+
+	rt->total = tbl->total;
+
+	ntfs_free(tbl);
+	return rt;
+}
+
+/*
+ * alloc_rsttbl_idx
+ *
+ * allocates an index from within a previously initialized Restart Table
+ */
+static inline void *alloc_rsttbl_idx(struct RESTART_TABLE **tbl)
+{
+	u32 off;
+	__le32 *e;
+	struct RESTART_TABLE *t = *tbl;
+
+	if (!t->first_free)
+		*tbl = t = extend_rsttbl(t, 16, ~0u);
+
+	off = le32_to_cpu(t->first_free);
+
+	/* Dequeue this entry and zero it. */
+	e = Add2Ptr(t, off);
+
+	t->first_free = *e;
+
+	memset(e, 0, le16_to_cpu(t->size));
+
+	*e = RESTART_ENTRY_ALLOCATED_LE;
+
+	/* If list is going empty, then we fix the last_free as well. */
+	if (!t->first_free)
+		t->last_free = 0;
+
+	le16_add_cpu(&t->total, 1);
+
+	return Add2Ptr(t, off);
+}
+
+/*
+ * alloc_rsttbl_from_idx
+ *
+ * allocates a specific index from within a previously initialized Restart Table
+ */
+static inline void *alloc_rsttbl_from_idx(struct RESTART_TABLE **tbl, u32 vbo)
+{
+	u32 off;
+	__le32 *e;
+	struct RESTART_TABLE *rt = *tbl;
+	u32 bytes = bytes_per_rt(rt);
+	u16 esize = le16_to_cpu(rt->size);
+
+	/* If the entry is not the table, we will have to extend the table */
+	if (vbo >= bytes) {
+		/*
+		 * extend the size by computing the number of entries between
+		 * the existing size and the desired index and adding
+		 * 1 to that
+		 */
+		u32 bytes2idx = vbo - bytes;
+
+		/* There should always be an integral number of entries being added */
+		/* Now extend the table */
+		*tbl = rt = extend_rsttbl(rt, bytes2idx / esize + 1, bytes);
+		if (!rt)
+			return NULL;
+	}
+
+	/* see if the entry is already allocated, and just return if it is. */
+	e = Add2Ptr(rt, vbo);
+
+	if (*e == RESTART_ENTRY_ALLOCATED_LE)
+		return e;
+
+	/*
+	 * Walk through the table, looking for the entry we're
+	 * interested and the previous entry
+	 */
+	off = le32_to_cpu(rt->first_free);
+	e = Add2Ptr(rt, off);
+
+	if (off == vbo) {
+		/* this is a match */
+		rt->first_free = *e;
+		goto skip_looking;
+	}
+
+	/*
+	 * need to walk through the list looking for the predecessor of our entry
+	 */
+	for (;;) {
+		/* Remember the entry just found */
+		u32 last_off = off;
+		__le32 *last_e = e;
+
+		/* should never run of entries. */
+
+		/* Lookup up the next entry the list */
+		off = le32_to_cpu(*last_e);
+		e = Add2Ptr(rt, off);
+
+		/* If this is our match we are done */
+		if (off == vbo) {
+			*last_e = *e;
+
+			/* If this was the last entry, we update that the table as well */
+			if (le32_to_cpu(rt->last_free) == off)
+				rt->last_free = cpu_to_le32(last_off);
+			break;
+		}
+	}
+
+skip_looking:
+	/* If the list is now empty, we fix the last_free as well */
+	if (!rt->first_free)
+		rt->last_free = 0;
+
+	/* Zero this entry */
+	memset(e, 0, esize);
+	*e = RESTART_ENTRY_ALLOCATED_LE;
+
+	le16_add_cpu(&rt->total, 1);
+
+	return e;
+}
+
+#define RESTART_SINGLE_PAGE_IO cpu_to_le16(0x0001)
+
+#define NTFSLOG_WRAPPED 0x00000001
+#define NTFSLOG_MULTIPLE_PAGE_IO 0x00000002
+#define NTFSLOG_NO_LAST_LSN 0x00000004
+#define NTFSLOG_REUSE_TAIL 0x00000010
+#define NTFSLOG_NO_OLDEST_LSN 0x00000020
+
+/*
+ * Helper struct to work with NTFS $LogFile
+ */
+struct ntfs_log {
+	struct ntfs_inode *ni;
+
+	u32 l_size;
+	u32 sys_page_size;
+	u32 sys_page_mask;
+	u32 page_size;
+	u32 page_mask; // page_size - 1
+	u8 page_bits;
+	struct RECORD_PAGE_HDR *one_page_buf;
+
+	struct RESTART_TABLE *open_attr_tbl;
+	u32 transaction_id;
+	u32 clst_per_page;
+
+	u32 first_page;
+	u32 next_page;
+	u32 ra_off;
+	u32 data_off;
+	u32 restart_size;
+	u32 data_size;
+	u16 record_header_len;
+	u64 seq_num;
+	u32 seq_num_bits;
+	u32 file_data_bits;
+	u32 seq_num_mask; /* (1 << file_data_bits) - 1 */
+
+	struct RESTART_AREA *ra; /* in-memory image of the next restart area */
+	u32 ra_size; /* the usable size of the restart area */
+
+	/*
+	 * If true, then the in-memory restart area is to be written
+	 * to the first position on the disk
+	 */
+	bool init_ra;
+	bool set_dirty; /* true if we need to set dirty flag */
+
+	u64 oldest_lsn;
+
+	u32 oldest_lsn_off;
+	u64 last_lsn;
+
+	u32 total_avail;
+	u32 total_avail_pages;
+	u32 total_undo_commit;
+	u32 max_current_avail;
+	u32 current_avail;
+	u32 reserved;
+
+	short major_ver;
+	short minor_ver;
+
+	u32 l_flags; /* See NTFSLOG_XXX */
+	u32 current_openlog_count; /* On-disk value for open_log_count */
+
+	struct CLIENT_ID client_id;
+	u32 client_undo_commit;
+};
+
+static inline u32 lsn_to_vbo(struct ntfs_log *log, const u64 lsn)
+{
+	u32 vbo = (lsn << log->seq_num_bits) >> (log->seq_num_bits - 3);
+
+	return vbo;
+}
+
+/* compute the offset in the log file of the next log page */
+static inline u32 next_page_off(struct ntfs_log *log, u32 off)
+{
+	off = (off & ~log->sys_page_mask) + log->page_size;
+	return off >= log->l_size ? log->first_page : off;
+}
+
+static inline u32 lsn_to_page_off(struct ntfs_log *log, u64 lsn)
+{
+	return (((u32)lsn) << 3) & log->page_mask;
+}
+
+static inline u64 vbo_to_lsn(struct ntfs_log *log, u32 off, u64 Seq)
+{
+	return (off >> 3) + (Seq << log->file_data_bits);
+}
+
+static inline bool is_lsn_in_file(struct ntfs_log *log, u64 lsn)
+{
+	return lsn >= log->oldest_lsn &&
+	       lsn <= le64_to_cpu(log->ra->current_lsn);
+}
+
+static inline u32 hdr_file_off(struct ntfs_log *log,
+			       struct RECORD_PAGE_HDR *hdr)
+{
+	if (log->major_ver < 2)
+		return le64_to_cpu(hdr->rhdr.lsn);
+
+	return le32_to_cpu(hdr->file_off);
+}
+
+static inline u64 base_lsn(struct ntfs_log *log,
+			   const struct RECORD_PAGE_HDR *hdr, u64 lsn)
+{
+	u64 h_lsn = le64_to_cpu(hdr->rhdr.lsn);
+	u64 ret = (((h_lsn >> log->file_data_bits) +
+		    (lsn < (lsn_to_vbo(log, h_lsn) & ~log->page_mask) ? 1 : 0))
+		   << log->file_data_bits) +
+		  ((((is_log_record_end(hdr) &&
+		      h_lsn <= le64_to_cpu(hdr->record_hdr.last_end_lsn)) ?
+			     le16_to_cpu(hdr->record_hdr.next_record_off) :
+			     log->page_size) +
+		    lsn) >>
+		   3);
+
+	return ret;
+}
+
+static inline bool verify_client_lsn(struct ntfs_log *log,
+				     const struct CLIENT_REC *client, u64 lsn)
+{
+	return lsn >= le64_to_cpu(client->oldest_lsn) &&
+	       lsn <= le64_to_cpu(log->ra->current_lsn) && lsn;
+}
+
+struct restart_info {
+	u64 last_lsn;
+	struct RESTART_HDR *r_page;
+	u32 vbo;
+	bool chkdsk_was_run;
+	bool valid_page;
+	bool initialized;
+	bool restart;
+};
+
+static int read_log_page(struct ntfs_log *log, u32 vbo,
+			 struct RECORD_PAGE_HDR **buffer, bool allow_errors,
+			 bool ignore_usa_error, bool *usa_error)
+{
+	int err = 0;
+	u32 page_idx = vbo >> log->page_bits;
+	u32 page_off = vbo & log->page_mask;
+	u32 bytes = log->page_size - page_off;
+	void *to_free = NULL;
+	u32 page_vbo = page_idx << log->page_bits;
+	struct RECORD_PAGE_HDR *page_buf;
+	struct ntfs_inode *ni = log->ni;
+	bool bBAAD;
+
+	if (vbo >= log->l_size)
+		return -EINVAL;
+
+	if (!*buffer) {
+		to_free = ntfs_alloc(bytes, 0);
+		if (!to_free)
+			return -ENOMEM;
+		*buffer = to_free;
+	}
+
+	page_buf = page_off ? log->one_page_buf : *buffer;
+
+	err = ntfs_read_run_nb(ni->mi.sbi, &ni->file.run, page_vbo, page_buf,
+			       log->page_size, NULL);
+	if (err)
+		goto out;
+
+	if (page_buf->rhdr.sign != NTFS_FFFF_SIGNATURE)
+		ntfs_fix_post_read(&page_buf->rhdr, PAGE_SIZE, false);
+
+	if (page_buf != *buffer)
+		memcpy(*buffer, Add2Ptr(page_buf, page_off), bytes);
+
+	bBAAD = page_buf->rhdr.sign == NTFS_BAAD_SIGNATURE;
+
+	/* Check that the update sequence array for this page is valid */
+	if (bBAAD) {
+		/* If we don't allow errors, raise an error status */
+		if (!ignore_usa_error) {
+			err = -EINVAL;
+			goto out;
+		}
+	}
+
+	if (usa_error)
+		*usa_error = bBAAD;
+
+out:
+	if (err && to_free) {
+		ntfs_free(to_free);
+		*buffer = NULL;
+	}
+
+	return err;
+}
+
+/*
+ * log_read_rst
+ *
+ * it walks through 512 blocks of the file looking for a valid restart page header
+ * It will stop the first time we find a valid page header
+ */
+static int log_read_rst(struct ntfs_log *log, u32 l_size, bool first,
+			struct restart_info *info)
+{
+	int err;
+	u32 skip, vbo;
+	struct RESTART_HDR *r_page = ntfs_alloc(DefaultLogPageSize, 0);
+
+	if (!r_page)
+		return -ENOMEM;
+
+	memset(info, 0, sizeof(struct restart_info));
+
+	/* Determine which restart area we are looking for */
+	if (first) {
+		vbo = 0;
+		skip = 512;
+	} else {
+		vbo = 512;
+		skip = 0;
+	}
+
+	/* loop continuously until we succeed */
+	for (; vbo < l_size; vbo = 2 * vbo + skip, skip = 0) {
+		bool usa_error;
+		u32 sys_page_size;
+		bool brst, bchk;
+		struct RESTART_AREA *ra;
+
+		/* Read a page header at the current offset */
+		err = read_log_page(log, vbo,
+				    (struct RECORD_PAGE_HDR **)&r_page, true,
+				    true, &usa_error);
+
+		if (err)
+			continue;
+
+		/* exit if the signature is a log record page */
+		if (r_page->rhdr.sign == NTFS_RCRD_SIGNATURE) {
+			info->initialized = true;
+			break;
+		}
+
+		brst = r_page->rhdr.sign == NTFS_RSTR_SIGNATURE;
+		bchk = r_page->rhdr.sign == NTFS_CHKD_SIGNATURE;
+
+		if (!bchk && !brst) {
+			if (r_page->rhdr.sign != NTFS_FFFF_SIGNATURE) {
+				/*
+				 * Remember if the signature does not
+				 * indicate uninitialized file
+				 */
+				info->initialized = true;
+			}
+			continue;
+		}
+
+		ra = NULL;
+		info->valid_page = false;
+		info->initialized = true;
+		info->vbo = vbo;
+
+		/* Let's check the restart area if this is a valid page */
+		if (!is_rst_page_hdr_valid(vbo, r_page))
+			goto check_result;
+		ra = Add2Ptr(r_page, le16_to_cpu(r_page->ra_off));
+
+		if (!is_rst_area_valid(r_page))
+			goto check_result;
+
+		/*
+		 * We have a valid restart page header and restart area.
+		 * If chkdsk was run or we have no clients then we have
+		 * no more checking to do
+		 */
+		if (bchk || ra->client_idx[1] == LFS_NO_CLIENT_LE) {
+			info->valid_page = true;
+			goto check_result;
+		}
+
+		/* Read the entire restart area */
+		sys_page_size = le32_to_cpu(r_page->sys_page_size);
+		if (DefaultLogPageSize != sys_page_size) {
+			ntfs_free(r_page);
+			r_page = ntfs_alloc(sys_page_size, 1);
+			if (!r_page)
+				return -ENOMEM;
+
+			err = read_log_page(log, vbo,
+					    (struct RECORD_PAGE_HDR **)&r_page,
+					    true, true, &usa_error);
+
+			if (err) {
+				ntfs_free(r_page);
+				return err;
+			}
+		}
+
+		if (is_client_area_valid(r_page, usa_error)) {
+			info->valid_page = true;
+			ra = Add2Ptr(r_page, le16_to_cpu(r_page->ra_off));
+		}
+
+check_result:
+		/* If chkdsk was run then update the caller's values and return */
+		if (r_page->rhdr.sign == NTFS_CHKD_SIGNATURE) {
+			info->chkdsk_was_run = true;
+			info->last_lsn = le64_to_cpu(r_page->rhdr.lsn);
+			info->restart = true;
+			info->r_page = r_page;
+			return 0;
+		}
+
+		/* If we have a valid page then copy the values we need from it */
+		if (info->valid_page) {
+			info->last_lsn = le64_to_cpu(ra->current_lsn);
+			info->restart = true;
+			info->r_page = r_page;
+			return 0;
+		}
+	}
+
+	ntfs_free(r_page);
+
+	return 0;
+}
+
+/*
+ * log_init_pg_hdr
+ *
+ * init "log' from restart page header
+ */
+static void log_init_pg_hdr(struct ntfs_log *log, u32 sys_page_size,
+			    u32 page_size, u16 major_ver, u16 minor_ver)
+{
+	log->sys_page_size = sys_page_size;
+	log->sys_page_mask = sys_page_size - 1;
+	log->page_size = page_size;
+	log->page_mask = page_size - 1;
+	log->page_bits = blksize_bits(page_size);
+
+	log->clst_per_page = log->page_size >> log->ni->mi.sbi->cluster_bits;
+	if (!log->clst_per_page)
+		log->clst_per_page = 1;
+
+	log->first_page = major_ver >= 2 ?
+				  0x22 * page_size :
+				  ((sys_page_size << 1) + (page_size << 1));
+	log->major_ver = major_ver;
+	log->minor_ver = minor_ver;
+}
+
+/*
+ * log_create
+ *
+ * init "log" in cases when we don't have a restart area to use
+ */
+static void log_create(struct ntfs_log *log, u32 l_size, const u64 last_lsn,
+		       u32 open_log_count, bool wrapped, bool use_multi_page)
+{
+	log->l_size = l_size;
+	/* All file offsets must be quadword aligned */
+	log->file_data_bits = blksize_bits(l_size) - 3;
+	log->seq_num_mask = (8 << log->file_data_bits) - 1;
+	log->seq_num_bits = sizeof(u64) * 8 - log->file_data_bits;
+	log->seq_num = (last_lsn >> log->file_data_bits) + 2;
+	log->next_page = log->first_page;
+	log->oldest_lsn = log->seq_num << log->file_data_bits;
+	log->oldest_lsn_off = 0;
+	log->last_lsn = log->oldest_lsn;
+
+	log->l_flags |= NTFSLOG_NO_LAST_LSN | NTFSLOG_NO_OLDEST_LSN;
+
+	/* Set the correct flags for the I/O and indicate if we have wrapped */
+	if (wrapped)
+		log->l_flags |= NTFSLOG_WRAPPED;
+
+	if (use_multi_page)
+		log->l_flags |= NTFSLOG_MULTIPLE_PAGE_IO;
+
+	/* Compute the log page values */
+	log->data_off = QuadAlign(
+		offsetof(struct RECORD_PAGE_HDR, fixups) +
+		sizeof(short) * ((log->page_size >> SECTOR_SHIFT) + 1));
+	log->data_size = log->page_size - log->data_off;
+	log->record_header_len = sizeof(struct LFS_RECORD_HDR);
+
+	/* Remember the different page sizes for reservation */
+	log->reserved = log->data_size - log->record_header_len;
+
+	/* Compute the restart page values. */
+	log->ra_off = QuadAlign(
+		offsetof(struct RESTART_HDR, fixups) +
+		sizeof(short) * ((log->sys_page_size >> SECTOR_SHIFT) + 1));
+	log->restart_size = log->sys_page_size - log->ra_off;
+	log->ra_size = offsetof(struct RESTART_AREA, clients) +
+		       sizeof(struct CLIENT_REC);
+	log->current_openlog_count = open_log_count;
+
+	/*
+	 * The total available log file space is the number of
+	 * log file pages times the space available on each page
+	 */
+	log->total_avail_pages = log->l_size - log->first_page;
+	log->total_avail = log->total_avail_pages >> log->page_bits;
+
+	/*
+	 * We assume that we can't use the end of the page less than
+	 * the file record size
+	 * Then we won't need to reserve more than the caller asks for
+	 */
+	log->max_current_avail = log->total_avail * log->reserved;
+	log->total_avail = log->total_avail * log->data_size;
+	log->current_avail = log->max_current_avail;
+}
+
+/*
+ * log_create_ra
+ *
+ * This routine is called to fill a restart area from the values stored in 'log'
+ */
+static struct RESTART_AREA *log_create_ra(struct ntfs_log *log)
+{
+	struct CLIENT_REC *cr;
+	struct RESTART_AREA *ra = ntfs_alloc(log->restart_size, 1);
+
+	if (!ra)
+		return NULL;
+
+	ra->current_lsn = cpu_to_le64(log->last_lsn);
+	ra->log_clients = cpu_to_le16(1);
+	ra->client_idx[1] = LFS_NO_CLIENT_LE;
+	if (log->l_flags & NTFSLOG_MULTIPLE_PAGE_IO)
+		ra->flags = RESTART_SINGLE_PAGE_IO;
+	ra->seq_num_bits = cpu_to_le32(log->seq_num_bits);
+	ra->ra_len = cpu_to_le16(log->ra_size);
+	ra->client_off = cpu_to_le16(offsetof(struct RESTART_AREA, clients));
+	ra->l_size = cpu_to_le64(log->l_size);
+	ra->rec_hdr_len = cpu_to_le16(log->record_header_len);
+	ra->data_off = cpu_to_le16(log->data_off);
+	ra->open_log_count = cpu_to_le32(log->current_openlog_count + 1);
+
+	cr = ra->clients;
+
+	cr->prev_client = LFS_NO_CLIENT_LE;
+	cr->next_client = LFS_NO_CLIENT_LE;
+
+	return ra;
+}
+
+static u32 final_log_off(struct ntfs_log *log, u64 lsn, u32 data_len)
+{
+	u32 base_vbo = lsn << 3;
+	u32 final_log_off = (base_vbo & log->seq_num_mask) & ~log->page_mask;
+	u32 page_off = base_vbo & log->page_mask;
+	u32 tail = log->page_size - page_off;
+
+	page_off -= 1;
+
+	/* Add the length of the header */
+	data_len += log->record_header_len;
+
+	/*
+	 * If this lsn is contained this log page we are done
+	 * Otherwise we need to walk through several log pages
+	 */
+	if (data_len > tail) {
+		data_len -= tail;
+		tail = log->data_size;
+		page_off = log->data_off - 1;
+
+		for (;;) {
+			final_log_off = next_page_off(log, final_log_off);
+
+			/* We are done if the remaining bytes fit on this page */
+			if (data_len <= tail)
+				break;
+			data_len -= tail;
+		}
+	}
+
+	/*
+	 * We add the remaining bytes to our starting position on this page
+	 * and then add that value to the file offset of this log page
+	 */
+	return final_log_off + data_len + page_off;
+}
+
+static int next_log_lsn(struct ntfs_log *log, const struct LFS_RECORD_HDR *rh,
+			u64 *lsn)
+{
+	int err;
+	u64 this_lsn = le64_to_cpu(rh->this_lsn);
+	u32 vbo = lsn_to_vbo(log, this_lsn);
+	u32 end =
+		final_log_off(log, this_lsn, le32_to_cpu(rh->client_data_len));
+	u32 hdr_off = end & ~log->sys_page_mask;
+	u64 seq = this_lsn >> log->file_data_bits;
+	struct RECORD_PAGE_HDR *page = NULL;
+
+	/* Remember if we wrapped */
+	if (end <= vbo)
+		seq += 1;
+
+	/* log page header for this page */
+	err = read_log_page(log, hdr_off, &page, false, false, NULL);
+	if (err)
+		return err;
+
+	/*
+	 * If the lsn we were given was not the last lsn on this page,
+	 * then the starting offset for the next lsn is on a quad word
+	 * boundary following the last file offset for the current lsn
+	 * Otherwise the file offset is the start of the data on the next page
+	 */
+	if (this_lsn == le64_to_cpu(page->rhdr.lsn)) {
+		/* If we wrapped, we need to increment the sequence number */
+		hdr_off = next_page_off(log, hdr_off);
+		if (hdr_off == log->first_page)
+			seq += 1;
+
+		vbo = hdr_off + log->data_off;
+	} else {
+		vbo = QuadAlign(end);
+	}
+
+	/* Compute the lsn based on the file offset and the sequence count */
+	*lsn = vbo_to_lsn(log, vbo, seq);
+
+	/*
+	 * If this lsn is within the legal range for the file, we return true
+	 * Otherwise false indicates that there are no more lsn's
+	 */
+	if (!is_lsn_in_file(log, *lsn))
+		*lsn = 0;
+
+	ntfs_free(page);
+
+	return 0;
+}
+
+/*
+ * current_log_avail
+ *
+ * calculate the number of bytes available for log records
+ */
+static u32 current_log_avail(struct ntfs_log *log)
+{
+	u32 oldest_off, next_free_off, free_bytes;
+
+	if (log->l_flags & NTFSLOG_NO_LAST_LSN) {
+		/* The entire file is available */
+		return log->max_current_avail;
+	}
+
+	/*
+	 * If there is a last lsn the restart area then we know that we will
+	 * have to compute the free range
+	 * If there is no oldest lsn then start at the first page of the file
+	 */
+	oldest_off = (log->l_flags & NTFSLOG_NO_OLDEST_LSN) ?
+			     log->first_page :
+			     (log->oldest_lsn_off & ~log->sys_page_mask);
+
+	/*
+	 * We will use the next log page offset to compute the next free page\
+	 * If we are going to reuse this page go to the next page
+	 * If we are at the first page then use the end of the file
+	 */
+	next_free_off = (log->l_flags & NTFSLOG_REUSE_TAIL) ?
+				log->next_page + log->page_size :
+				log->next_page == log->first_page ?
+				log->l_size :
+				log->next_page;
+
+	/* If the two offsets are the same then there is no available space */
+	if (oldest_off == next_free_off)
+		return 0;
+	/*
+	 * If the free offset follows the oldest offset then subtract
+	 * this range from the total available pages
+	 */
+	free_bytes =
+		oldest_off < next_free_off ?
+			log->total_avail_pages - (next_free_off - oldest_off) :
+			oldest_off - next_free_off;
+
+	free_bytes >>= log->page_bits;
+	return free_bytes * log->reserved;
+}
+
+static bool check_subseq_log_page(struct ntfs_log *log,
+				  const struct RECORD_PAGE_HDR *rp, u32 vbo,
+				  u64 seq)
+{
+	u64 lsn_seq;
+	const struct NTFS_RECORD_HEADER *rhdr = &rp->rhdr;
+	u64 lsn = le64_to_cpu(rhdr->lsn);
+
+	if (rhdr->sign == NTFS_FFFF_SIGNATURE || !rhdr->sign)
+		return false;
+
+	/*
+	 * If the last lsn on the page occurs was written after the page
+	 * that caused the original error then we have a fatal error
+	 */
+	lsn_seq = lsn >> log->file_data_bits;
+
+	/*
+	 * If the sequence number for the lsn the page is equal or greater
+	 * than lsn we expect, then this is a subsequent write
+	 */
+	return lsn_seq >= seq ||
+	       (lsn_seq == seq - 1 && log->first_page == vbo &&
+		vbo != (lsn_to_vbo(log, lsn) & ~log->page_mask));
+}
+
+/*
+ * last_log_lsn
+ *
+ * This routine walks through the log pages for a file, searching for the
+ * last log page written to the file
+ */
+static int last_log_lsn(struct ntfs_log *log)
+{
+	int err;
+	bool usa_error = false;
+	bool replace_page = false;
+	bool reuse_page = log->l_flags & NTFSLOG_REUSE_TAIL;
+	bool wrapped_file, wrapped;
+
+	u32 page_cnt = 1, page_pos = 1;
+	u32 page_off = 0, page_off1 = 0, saved_off = 0;
+	u32 final_off, second_off, final_off_prev = 0, second_off_prev = 0;
+	u32 first_file_off = 0, second_file_off = 0;
+	u32 part_io_count = 0;
+	u32 tails = 0;
+	u32 this_off, curpage_off, nextpage_off, remain_pages;
+
+	u64 expected_seq, seq_base = 0, lsn_base = 0;
+	u64 best_lsn, best_lsn1, best_lsn2;
+	u64 lsn_cur, lsn1, lsn2;
+	u64 last_ok_lsn = reuse_page ? log->last_lsn : 0;
+
+	u16 cur_pos, best_page_pos;
+
+	struct RECORD_PAGE_HDR *page = NULL;
+	struct RECORD_PAGE_HDR *tst_page = NULL;
+	struct RECORD_PAGE_HDR *first_tail = NULL;
+	struct RECORD_PAGE_HDR *second_tail = NULL;
+	struct RECORD_PAGE_HDR *tail_page = NULL;
+	struct RECORD_PAGE_HDR *second_tail_prev = NULL,
+			       *first_tail_prev = NULL;
+	struct RECORD_PAGE_HDR *page_bufs = NULL;
+	struct RECORD_PAGE_HDR *best_page;
+
+	if (log->major_ver >= 2) {
+		final_off = 0x02 * log->page_size;
+		second_off = 0x12 * log->page_size;
+
+		// 0x10 == 0x12 - 0x2
+		page_bufs = ntfs_alloc(log->page_size * 0x10, 0);
+		if (!page_bufs)
+			return -ENOMEM;
+	} else {
+		second_off = log->first_page - log->page_size;
+		final_off = second_off - log->page_size;
+	}
+
+next_tail:
+	/* Read second tail page (at pos 3/0x12000) */
+	if (read_log_page(log, second_off, &second_tail, true, true,
+			  &usa_error) ||
+	    usa_error || second_tail->rhdr.sign != NTFS_RCRD_SIGNATURE) {
+		ntfs_free(second_tail);
+		second_tail = NULL;
+		second_file_off = 0;
+		lsn2 = 0;
+	} else {
+		second_file_off = hdr_file_off(log, second_tail);
+		lsn2 = le64_to_cpu(second_tail->record_hdr.last_end_lsn);
+	}
+
+	/* Read first tail page (at pos 2/0x2000 ) */
+	if (read_log_page(log, final_off, &first_tail, true, true,
+			  &usa_error) ||
+	    usa_error || first_tail->rhdr.sign != NTFS_RCRD_SIGNATURE) {
+		ntfs_free(first_tail);
+		first_tail = NULL;
+		first_file_off = 0;
+		lsn1 = 0;
+	} else {
+		first_file_off = hdr_file_off(log, first_tail);
+		lsn1 = le64_to_cpu(first_tail->record_hdr.last_end_lsn);
+	}
+
+	if (log->major_ver < 2) {
+		int best_page;
+
+		first_tail_prev = first_tail;
+		final_off_prev = first_file_off;
+		second_tail_prev = second_tail;
+		second_off_prev = second_file_off;
+		tails = 1;
+
+		if (!first_tail && !second_tail)
+			goto tail_read;
+
+		if (first_tail && second_tail)
+			best_page = lsn1 < lsn2 ? 1 : 0;
+		else if (first_tail)
+			best_page = 0;
+		else
+			best_page = 1;
+
+		page_off = best_page ? second_file_off : first_file_off;
+		seq_base = (best_page ? lsn2 : lsn1) >> log->file_data_bits;
+		goto tail_read;
+	}
+
+	best_lsn1 = first_tail ? base_lsn(log, first_tail, first_file_off) : 0;
+	best_lsn2 =
+		second_tail ? base_lsn(log, second_tail, second_file_off) : 0;
+
+	if (first_tail && second_tail) {
+		if (best_lsn1 > best_lsn2) {
+			best_lsn = best_lsn1;
+			best_page = first_tail;
+			this_off = first_file_off;
+		} else {
+			best_lsn = best_lsn2;
+			best_page = second_tail;
+			this_off = second_file_off;
+		}
+	} else if (first_tail) {
+		best_lsn = best_lsn1;
+		best_page = first_tail;
+		this_off = first_file_off;
+	} else if (second_tail) {
+		best_lsn = best_lsn2;
+		best_page = second_tail;
+		this_off = second_file_off;
+	} else {
+		goto free_and_tail_read;
+	}
+
+	best_page_pos = le16_to_cpu(best_page->page_pos);
+
+	if (!tails) {
+		if (best_page_pos == page_pos) {
+			seq_base = best_lsn >> log->file_data_bits;
+			saved_off = page_off = le32_to_cpu(best_page->file_off);
+			lsn_base = best_lsn;
+
+			memmove(page_bufs, best_page, log->page_size);
+
+			page_cnt = le16_to_cpu(best_page->page_count);
+			if (page_cnt > 1)
+				page_pos += 1;
+
+			tails = 1;
+		}
+	} else if (seq_base == (best_lsn >> log->file_data_bits) &&
+		   saved_off + log->page_size == this_off &&
+		   lsn_base < best_lsn &&
+		   (page_pos != page_cnt || best_page_pos == page_pos ||
+		    best_page_pos == 1) &&
+		   (page_pos >= page_cnt || best_page_pos == page_pos)) {
+		u16 bppc = le16_to_cpu(best_page->page_count);
+
+		saved_off += log->page_size;
+		lsn_base = best_lsn;
+
+		memmove(Add2Ptr(page_bufs, tails * log->page_size), best_page,
+			log->page_size);
+
+		tails += 1;
+
+		if (best_page_pos != bppc) {
+			page_cnt = bppc;
+			page_pos = best_page_pos;
+
+			if (page_cnt > 1)
+				page_pos += 1;
+		} else {
+			page_pos = page_cnt = 1;
+		}
+	} else {
+free_and_tail_read:
+		ntfs_free(first_tail);
+		ntfs_free(second_tail);
+		goto tail_read;
+	}
+
+	ntfs_free(first_tail_prev);
+	first_tail_prev = first_tail;
+	final_off_prev = first_file_off;
+	first_tail = NULL;
+
+	ntfs_free(second_tail_prev);
+	second_tail_prev = second_tail;
+	second_off_prev = second_file_off;
+	second_tail = NULL;
+
+	final_off += log->page_size;
+	second_off += log->page_size;
+
+	if (tails < 0x10)
+		goto next_tail;
+tail_read:
+	first_tail = first_tail_prev;
+	final_off = final_off_prev;
+
+	second_tail = second_tail_prev;
+	second_off = second_off_prev;
+
+	page_cnt = page_pos = 1;
+
+	curpage_off = seq_base == log->seq_num ? min(log->next_page, page_off) :
+						 log->next_page;
+
+	wrapped_file =
+		curpage_off == log->first_page &&
+		!(log->l_flags & (NTFSLOG_NO_LAST_LSN | NTFSLOG_REUSE_TAIL));
+
+	expected_seq = wrapped_file ? (log->seq_num + 1) : log->seq_num;
+
+	nextpage_off = curpage_off;
+
+next_page:
+	tail_page = NULL;
+	/* Read the next log page, allowing errors */
+	err = read_log_page(log, curpage_off, &page, true, true, &usa_error);
+
+	/* Compute the next log page offset the file */
+	nextpage_off = next_page_off(log, curpage_off);
+	wrapped = nextpage_off == log->first_page;
+
+	if (tails > 1) {
+		struct RECORD_PAGE_HDR *cur_page =
+			Add2Ptr(page_bufs, curpage_off - page_off);
+
+		if (curpage_off == saved_off) {
+			tail_page = cur_page;
+			goto use_tail_page;
+		}
+
+		if (page_off > curpage_off || curpage_off >= saved_off)
+			goto use_tail_page;
+
+		if (page_off1)
+			goto use_cur_page;
+
+		if (!err && !usa_error &&
+		    page->rhdr.sign == NTFS_RCRD_SIGNATURE &&
+		    cur_page->rhdr.lsn == page->rhdr.lsn &&
+		    cur_page->record_hdr.next_record_off ==
+			    page->record_hdr.next_record_off &&
+		    ((page_pos == page_cnt &&
+		      le16_to_cpu(page->page_pos) == 1) ||
+		     (page_pos != page_cnt &&
+		      le16_to_cpu(page->page_pos) == page_pos + 1 &&
+		      le16_to_cpu(page->page_count) == page_cnt))) {
+			cur_page = NULL;
+			goto use_tail_page;
+		}
+
+		page_off1 = page_off;
+
+use_cur_page:
+
+		lsn_cur = le64_to_cpu(cur_page->rhdr.lsn);
+
+		if (last_ok_lsn !=
+			    le64_to_cpu(cur_page->record_hdr.last_end_lsn) &&
+		    ((lsn_cur >> log->file_data_bits) +
+		     ((curpage_off <
+		       (lsn_to_vbo(log, lsn_cur) & ~log->page_mask)) ?
+			      1 :
+			      0)) != expected_seq) {
+			goto check_tail;
+		}
+
+		if (!is_log_record_end(cur_page)) {
+			tail_page = NULL;
+			last_ok_lsn = lsn_cur;
+			goto next_page_1;
+		}
+
+		log->seq_num = expected_seq;
+		log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
+		log->last_lsn = le64_to_cpu(cur_page->record_hdr.last_end_lsn);
+		log->ra->current_lsn = cur_page->record_hdr.last_end_lsn;
+
+		if (log->record_header_len <=
+		    log->page_size -
+			    le16_to_cpu(cur_page->record_hdr.next_record_off)) {
+			log->l_flags |= NTFSLOG_REUSE_TAIL;
+			log->next_page = curpage_off;
+		} else {
+			log->l_flags &= ~NTFSLOG_REUSE_TAIL;
+			log->next_page = nextpage_off;
+		}
+
+		if (wrapped_file)
+			log->l_flags |= NTFSLOG_WRAPPED;
+
+		last_ok_lsn = le64_to_cpu(cur_page->record_hdr.last_end_lsn);
+		goto next_page_1;
+	}
+
+	/*
+	 * If we are at the expected first page of a transfer check to see
+	 * if either tail copy is at this offset
+	 * If this page is the last page of a transfer, check if we wrote
+	 * a subsequent tail copy
+	 */
+	if (page_cnt == page_pos || page_cnt == page_pos + 1) {
+		/*
+		 * Check if the offset matches either the first or second
+		 * tail copy. It is possible it will match both
+		 */
+		if (curpage_off == final_off)
+			tail_page = first_tail;
+
+		/*
+		 * If we already matched on the first page then
+		 * check the ending lsn's.
+		 */
+		if (curpage_off == second_off) {
+			if (!tail_page ||
+			    (second_tail &&
+			     le64_to_cpu(second_tail->record_hdr.last_end_lsn) >
+				     le64_to_cpu(first_tail->record_hdr
+							 .last_end_lsn))) {
+				tail_page = second_tail;
+			}
+		}
+	}
+
+use_tail_page:
+	if (tail_page) {
+		/* we have a candidate for a tail copy */
+		lsn_cur = le64_to_cpu(tail_page->record_hdr.last_end_lsn);
+
+		if (last_ok_lsn < lsn_cur) {
+			/*
+			 * If the sequence number is not expected,
+			 * then don't use the tail copy
+			 */
+			if (expected_seq != (lsn_cur >> log->file_data_bits))
+				tail_page = NULL;
+		} else if (last_ok_lsn > lsn_cur) {
+			/*
+			 * If the last lsn is greater than the one on
+			 * this page then forget this tail
+			 */
+			tail_page = NULL;
+		}
+	}
+
+	/* If we have an error on the current page, we will break of this loop */
+	if (err || usa_error)
+		goto check_tail;
+
+	/*
+	 * Done if the last lsn on this page doesn't match the previous known
+	 * last lsn or the sequence number is not expected
+	 */
+	lsn_cur = le64_to_cpu(page->rhdr.lsn);
+	if (last_ok_lsn != lsn_cur &&
+	    expected_seq != (lsn_cur >> log->file_data_bits)) {
+		goto check_tail;
+	}
+
+	/*
+	 * Check that the page position and page count values are correct
+	 * If this is the first page of a transfer the position must be 1
+	 * and the count will be unknown
+	 */
+	if (page_cnt == page_pos) {
+		if (page->page_pos != cpu_to_le16(1) &&
+		    (!reuse_page || page->page_pos != page->page_count)) {
+			/*
+			 * If the current page is the first page we are
+			 * looking at and we are reusing this page then
+			 * it can be either the first or last page of a
+			 * transfer. Otherwise it can only be the first.
+			 */
+			goto check_tail;
+		}
+	} else if (le16_to_cpu(page->page_count) != page_cnt ||
+		   le16_to_cpu(page->page_pos) != page_pos + 1) {
+		/*
+		 * The page position better be 1 more than the last page
+		 * position and the page count better match
+		 */
+		goto check_tail;
+	}
+
+	/*
+	 * We have a valid page the file and may have a valid page
+	 * the tail copy area
+	 * If the tail page was written after the page the file then
+	 * break of the loop
+	 */
+	if (tail_page &&
+	    le64_to_cpu(tail_page->record_hdr.last_end_lsn) > lsn_cur) {
+		/* Remember if we will replace the page */
+		replace_page = true;
+		goto check_tail;
+	}
+
+	tail_page = NULL;
+
+	if (is_log_record_end(page)) {
+		/*
+		 * Since we have read this page we know the sequence number
+		 * is the same as our expected value
+		 */
+		log->seq_num = expected_seq;
+		log->last_lsn = le64_to_cpu(page->record_hdr.last_end_lsn);
+		log->ra->current_lsn = page->record_hdr.last_end_lsn;
+		log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
+
+		/*
+		 * If there is room on this page for another header then
+		 * remember we want to reuse the page
+		 */
+		if (log->record_header_len <=
+		    log->page_size -
+			    le16_to_cpu(page->record_hdr.next_record_off)) {
+			log->l_flags |= NTFSLOG_REUSE_TAIL;
+			log->next_page = curpage_off;
+		} else {
+			log->l_flags &= ~NTFSLOG_REUSE_TAIL;
+			log->next_page = nextpage_off;
+		}
+
+		/* Remember if we wrapped the log file */
+		if (wrapped_file)
+			log->l_flags |= NTFSLOG_WRAPPED;
+	}
+
+	/*
+	 * Remember the last page count and position.
+	 * Also remember the last known lsn
+	 */
+	page_cnt = le16_to_cpu(page->page_count);
+	page_pos = le16_to_cpu(page->page_pos);
+	last_ok_lsn = le64_to_cpu(page->rhdr.lsn);
+
+next_page_1:
+
+	if (wrapped) {
+		expected_seq += 1;
+		wrapped_file = 1;
+	}
+
+	curpage_off = nextpage_off;
+	ntfs_free(page);
+	page = NULL;
+	reuse_page = 0;
+	goto next_page;
+
+check_tail:
+	if (tail_page) {
+		log->seq_num = expected_seq;
+		log->last_lsn = le64_to_cpu(tail_page->record_hdr.last_end_lsn);
+		log->ra->current_lsn = tail_page->record_hdr.last_end_lsn;
+		log->l_flags &= ~NTFSLOG_NO_LAST_LSN;
+
+		if (log->page_size -
+			    le16_to_cpu(
+				    tail_page->record_hdr.next_record_off) >=
+		    log->record_header_len) {
+			log->l_flags |= NTFSLOG_REUSE_TAIL;
+			log->next_page = curpage_off;
+		} else {
+			log->l_flags &= ~NTFSLOG_REUSE_TAIL;
+			log->next_page = nextpage_off;
+		}
+
+		if (wrapped)
+			log->l_flags |= NTFSLOG_WRAPPED;
+	}
+
+	/* Remember that the partial IO will start at the next page */
+	second_off = nextpage_off;
+
+	/*
+	 * If the next page is the first page of the file then update
+	 * the sequence number for log records which begon the next page
+	 */
+	if (wrapped)
+		expected_seq += 1;
+
+	/*
+	 * If we have a tail copy or are performing single page I/O we can
+	 * immediately look at the next page
+	 */
+	if (replace_page || (log->ra->flags & RESTART_SINGLE_PAGE_IO)) {
+		page_cnt = 2;
+		page_pos = 1;
+		goto check_valid;
+	}
+
+	if (page_pos != page_cnt)
+		goto check_valid;
+	/*
+	 * If the next page causes us to wrap to the beginning of the log
+	 * file then we know which page to check next.
+	 */
+	if (wrapped) {
+		page_cnt = 2;
+		page_pos = 1;
+		goto check_valid;
+	}
+
+	cur_pos = 2;
+
+next_test_page:
+	ntfs_free(tst_page);
+	tst_page = NULL;
+
+	/* Walk through the file, reading log pages */
+	err = read_log_page(log, nextpage_off, &tst_page, 1, 1, &usa_error);
+
+	/*
+	 * If we get a USA error then assume that we correctly found
+	 * the end of the original transfer
+	 */
+	if (usa_error)
+		goto file_is_valid;
+
+	/*
+	 * If we were able to read the page, we examine it to see if it
+	 * is the same or different Io block
+	 */
+	if (err)
+		goto next_test_page_1;
+
+	if (le16_to_cpu(tst_page->page_pos) == cur_pos &&
+	    check_subseq_log_page(log, tst_page, nextpage_off, expected_seq)) {
+		page_cnt = le16_to_cpu(tst_page->page_count) + 1;
+		page_pos = le16_to_cpu(tst_page->page_pos);
+		goto check_valid;
+	} else {
+		goto file_is_valid;
+	}
+
+next_test_page_1:
+
+	nextpage_off = next_page_off(log, curpage_off);
+	wrapped = nextpage_off == log->first_page;
+
+	if (wrapped) {
+		expected_seq += 1;
+		page_cnt = 2;
+		page_pos = 1;
+	}
+
+	cur_pos += 1;
+	part_io_count += 1;
+	if (!wrapped)
+		goto next_test_page;
+
+check_valid:
+	/* Skip over the remaining pages this transfer */
+	remain_pages = page_cnt - page_pos - 1;
+	part_io_count += remain_pages;
+
+	while (remain_pages--) {
+		nextpage_off = next_page_off(log, curpage_off);
+		wrapped = nextpage_off == log->first_page;
+
+		if (wrapped)
+			expected_seq += 1;
+	}
+
+	/* Call our routine to check this log page */
+	ntfs_free(tst_page);
+	tst_page = NULL;
+
+	err = read_log_page(log, nextpage_off, &tst_page, true, true,
+			    &usa_error);
+	if (!err && !usa_error &&
+	    check_subseq_log_page(log, tst_page, nextpage_off, expected_seq)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+file_is_valid:
+
+	/* We have a valid file */
+	if (page_off1 || tail_page) {
+		struct RECORD_PAGE_HDR *tmp_page;
+
+		if (sb_rdonly(log->ni->mi.sbi->sb)) {
+			err = -EROFS;
+			goto out;
+		}
+
+		if (page_off1) {
+			tmp_page = Add2Ptr(page_bufs, page_off1 - page_off);
+			tails -= (page_off1 - page_off) / log->page_size;
+			if (!tail_page)
+				tails -= 1;
+		} else {
+			tmp_page = tail_page;
+			tails = 1;
+		}
+
+		while (tails--) {
+			u64 off = hdr_file_off(log, tmp_page);
+
+			if (!page) {
+				page = ntfs_alloc(log->page_size, 0);
+				if (!page)
+					return -ENOMEM;
+			}
+
+			/*
+			 * Correct page and copy the data from this page
+			 * into it and flush it to disk
+			 */
+			memcpy(page, tmp_page, log->page_size);
+
+			/* Fill last flushed lsn value flush the page */
+			if (log->major_ver < 2)
+				page->rhdr.lsn = page->record_hdr.last_end_lsn;
+			else
+				page->file_off = 0;
+
+			page->page_pos = page->page_count = cpu_to_le16(1);
+
+			ntfs_fix_pre_write(&page->rhdr, log->page_size);
+
+			err = ntfs_sb_write_run(log->ni->mi.sbi,
+						&log->ni->file.run, off, page,
+						log->page_size);
+
+			if (err)
+				goto out;
+
+			if (part_io_count && second_off == off) {
+				second_off += log->page_size;
+				part_io_count -= 1;
+			}
+
+			tmp_page = Add2Ptr(tmp_page, log->page_size);
+		}
+	}
+
+	if (part_io_count) {
+		if (sb_rdonly(log->ni->mi.sbi->sb)) {
+			err = -EROFS;
+			goto out;
+		}
+	}
+
+out:
+	ntfs_free(second_tail);
+	ntfs_free(first_tail);
+	ntfs_free(page);
+	ntfs_free(tst_page);
+	ntfs_free(page_bufs);
+
+	return err;
+}
+
+/*
+ * read_log_rec_buf
+ *
+ * copies a log record from the file to a buffer
+ * The log record may span several log pages and may even wrap the file
+ */
+static int read_log_rec_buf(struct ntfs_log *log,
+			    const struct LFS_RECORD_HDR *rh, void *buffer)
+{
+	int err;
+	struct RECORD_PAGE_HDR *ph = NULL;
+	u64 lsn = le64_to_cpu(rh->this_lsn);
+	u32 vbo = lsn_to_vbo(log, lsn) & ~log->page_mask;
+	u32 off = lsn_to_page_off(log, lsn) + log->record_header_len;
+	u32 data_len = le32_to_cpu(rh->client_data_len);
+
+	/*
+	 * While there are more bytes to transfer,
+	 * we continue to attempt to perform the read
+	 */
+	for (;;) {
+		u32 tail = log->page_size - off;
+
+		if (tail >= data_len)
+			tail = data_len;
+
+		data_len -= tail;
+
+		err = read_log_page(log, vbo, &ph, false, true, NULL);
+		if (err)
+			goto out;
+
+		/*
+		 * The last lsn on this page better be greater or equal
+		 * to the lsn we are copying
+		 */
+		if (lsn > le64_to_cpu(ph->rhdr.lsn)) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		memcpy(buffer, Add2Ptr(ph, off), tail);
+
+		/* If there are no more bytes to transfer, we exit the loop */
+		if (!data_len) {
+			if (!is_log_record_end(ph) ||
+			    lsn > le64_to_cpu(ph->record_hdr.last_end_lsn)) {
+				err = -EINVAL;
+				goto out;
+			}
+			break;
+		}
+
+		if (ph->rhdr.lsn == ph->record_hdr.last_end_lsn ||
+		    lsn > le64_to_cpu(ph->rhdr.lsn)) {
+			err = -EINVAL;
+			goto out;
+		}
+
+		vbo = next_page_off(log, vbo);
+		off = log->data_off;
+
+		/*
+		 * adjust our pointer the user's buffer to transfer
+		 * the next block to
+		 */
+		buffer = Add2Ptr(buffer, tail);
+	}
+
+out:
+	ntfs_free(ph);
+	return err;
+}
+
+static int read_rst_area(struct ntfs_log *log, struct NTFS_RESTART **rst_,
+			 u64 *lsn)
+{
+	int err;
+	struct LFS_RECORD_HDR *rh = NULL;
+	const struct CLIENT_REC *cr =
+		Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off));
+	u64 lsnr, lsnc = le64_to_cpu(cr->restart_lsn);
+	u32 len;
+	struct NTFS_RESTART *rst;
+
+	*lsn = 0;
+	*rst_ = NULL;
+
+	/* If the client doesn't have a restart area, go ahead and exit now */
+	if (!lsnc)
+		return 0;
+
+	err = read_log_page(log, lsn_to_vbo(log, lsnc),
+			    (struct RECORD_PAGE_HDR **)&rh, false, false, NULL);
+	if (err)
+		return err;
+
+	rst = NULL;
+	lsnr = le64_to_cpu(rh->this_lsn);
+
+	if (lsnc != lsnr) {
+		/* If the lsn values don't match, then the disk is corrupt */
+		err = -EINVAL;
+		goto out;
+	}
+
+	*lsn = lsnr;
+	len = le32_to_cpu(rh->client_data_len);
+
+	if (!len) {
+		err = 0;
+		goto out;
+	}
+
+	if (len < sizeof(struct NTFS_RESTART)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	rst = ntfs_alloc(len, 0);
+	if (!rst) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	/* Copy the data into the 'rst' buffer */
+	err = read_log_rec_buf(log, rh, rst);
+	if (err)
+		goto out;
+
+	*rst_ = rst;
+	rst = NULL;
+
+out:
+	ntfs_free(rh);
+	ntfs_free(rst);
+
+	return err;
+}
+
+static int find_log_rec(struct ntfs_log *log, u64 lsn, struct lcb *lcb)
+{
+	int err;
+	struct LFS_RECORD_HDR *rh = lcb->lrh;
+	u32 rec_len, len;
+
+	/* Read the record header for this lsn */
+	if (!rh) {
+		err = read_log_page(log, lsn_to_vbo(log, lsn),
+				    (struct RECORD_PAGE_HDR **)&rh, false,
+				    false, NULL);
+
+		lcb->lrh = rh;
+		if (err)
+			return err;
+	}
+
+	/*
+	 * If the lsn the log record doesn't match the desired
+	 * lsn then the disk is corrupt
+	 */
+	if (lsn != le64_to_cpu(rh->this_lsn))
+		return -EINVAL;
+
+	len = le32_to_cpu(rh->client_data_len);
+
+	/*
+	 * check that the length field isn't greater than the total
+	 * available space the log file
+	 */
+	rec_len = len + log->record_header_len;
+	if (rec_len >= log->total_avail)
+		return -EINVAL;
+
+	/*
+	 * If the entire log record is on this log page,
+	 * put a pointer to the log record the context block
+	 */
+	if (rh->flags & LOG_RECORD_MULTI_PAGE) {
+		void *lr = ntfs_alloc(len, 0);
+
+		if (!lr)
+			return -ENOMEM;
+
+		lcb->log_rec = lr;
+		lcb->alloc = true;
+
+		/* Copy the data into the buffer returned */
+		err = read_log_rec_buf(log, rh, lr);
+		if (err)
+			return err;
+	} else {
+		/* If beyond the end of the current page -> an error */
+		u32 page_off = lsn_to_page_off(log, lsn);
+
+		if (page_off + len + log->record_header_len > log->page_size)
+			return -EINVAL;
+
+		lcb->log_rec = Add2Ptr(rh, sizeof(struct LFS_RECORD_HDR));
+		lcb->alloc = false;
+	}
+
+	return 0;
+}
+
+/*
+ * read_log_rec_lcb
+ *
+ * initiates the query operation.
+ */
+static int read_log_rec_lcb(struct ntfs_log *log, u64 lsn, u32 ctx_mode,
+			    struct lcb **lcb_)
+{
+	int err;
+	const struct CLIENT_REC *cr;
+	struct lcb *lcb;
+
+	switch (ctx_mode) {
+	case lcb_ctx_undo_next:
+	case lcb_ctx_prev:
+	case lcb_ctx_next:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* check that the given lsn is the legal range for this client */
+	cr = Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off));
+
+	if (!verify_client_lsn(log, cr, lsn))
+		return -EINVAL;
+
+	lcb = ntfs_alloc(sizeof(struct lcb), 1);
+	if (!lcb)
+		return -ENOMEM;
+	lcb->client = log->client_id;
+	lcb->ctx_mode = ctx_mode;
+
+	/* Find the log record indicated by the given lsn */
+	err = find_log_rec(log, lsn, lcb);
+	if (err)
+		goto out;
+
+	*lcb_ = lcb;
+	return 0;
+
+out:
+	lcb_put(lcb);
+	*lcb_ = NULL;
+	return err;
+}
+
+/*
+ * find_client_next_lsn
+ *
+ * attempt to find the next lsn to return to a client based on the context mode.
+ */
+static int find_client_next_lsn(struct ntfs_log *log, struct lcb *lcb, u64 *lsn)
+{
+	int err;
+	u64 next_lsn;
+	struct LFS_RECORD_HDR *hdr;
+
+	hdr = lcb->lrh;
+	*lsn = 0;
+
+	if (lcb_ctx_next != lcb->ctx_mode)
+		goto check_undo_next;
+
+	/* Loop as long as another lsn can be found */
+	for (;;) {
+		u64 current_lsn;
+
+		err = next_log_lsn(log, hdr, &current_lsn);
+		if (err)
+			goto out;
+
+		if (!current_lsn)
+			break;
+
+		if (hdr != lcb->lrh)
+			ntfs_free(hdr);
+
+		hdr = NULL;
+		err = read_log_page(log, lsn_to_vbo(log, current_lsn),
+				    (struct RECORD_PAGE_HDR **)&hdr, false,
+				    false, NULL);
+		if (err)
+			goto out;
+
+		if (memcmp(&hdr->client, &lcb->client,
+			   sizeof(struct CLIENT_ID))) {
+			/*err = -EINVAL; */
+		} else if (LfsClientRecord == hdr->record_type) {
+			ntfs_free(lcb->lrh);
+			lcb->lrh = hdr;
+			*lsn = current_lsn;
+			return 0;
+		}
+	}
+
+out:
+	if (hdr != lcb->lrh)
+		ntfs_free(hdr);
+	return err;
+
+check_undo_next:
+	if (lcb_ctx_undo_next == lcb->ctx_mode)
+		next_lsn = le64_to_cpu(hdr->client_undo_next_lsn);
+	else if (lcb_ctx_prev == lcb->ctx_mode)
+		next_lsn = le64_to_cpu(hdr->client_prev_lsn);
+	else
+		return 0;
+
+	if (!next_lsn)
+		return 0;
+
+	if (!verify_client_lsn(
+		    log, Add2Ptr(log->ra, le16_to_cpu(log->ra->client_off)),
+		    next_lsn))
+		return 0;
+
+	hdr = NULL;
+	err = read_log_page(log, lsn_to_vbo(log, next_lsn),
+			    (struct RECORD_PAGE_HDR **)&hdr, false, false,
+			    NULL);
+	if (err)
+		return err;
+	ntfs_free(lcb->lrh);
+	lcb->lrh = hdr;
+
+	*lsn = next_lsn;
+
+	return 0;
+}
+
+static int read_next_log_rec(struct ntfs_log *log, struct lcb *lcb, u64 *lsn)
+{
+	int err;
+
+	err = find_client_next_lsn(log, lcb, lsn);
+	if (err)
+		return err;
+
+	if (!*lsn)
+		return 0;
+
+	if (lcb->alloc)
+		ntfs_free(lcb->log_rec);
+
+	lcb->log_rec = NULL;
+	lcb->alloc = false;
+	ntfs_free(lcb->lrh);
+	lcb->lrh = NULL;
+
+	return find_log_rec(log, *lsn, lcb);
+}
+
+static inline bool check_index_header(const struct INDEX_HDR *hdr, size_t bytes)
+{
+	__le16 mask;
+	u32 min_de, de_off, used, total;
+	const struct NTFS_DE *e;
+
+	if (hdr_has_subnode(hdr)) {
+		min_de = sizeof(struct NTFS_DE) + sizeof(u64);
+		mask = NTFS_IE_HAS_SUBNODES;
+	} else {
+		min_de = sizeof(struct NTFS_DE);
+		mask = 0;
+	}
+
+	de_off = le32_to_cpu(hdr->de_off);
+	used = le32_to_cpu(hdr->used);
+	total = le32_to_cpu(hdr->total);
+
+	if (de_off > bytes - min_de || used > bytes || total > bytes ||
+	    de_off + min_de > used || used > total) {
+		return false;
+	}
+
+	e = Add2Ptr(hdr, de_off);
+	for (;;) {
+		u16 esize = le16_to_cpu(e->size);
+		struct NTFS_DE *next = Add2Ptr(e, esize);
+
+		if (esize < min_de || PtrOffset(hdr, next) > used ||
+		    (e->flags & NTFS_IE_HAS_SUBNODES) != mask) {
+			return false;
+		}
+
+		if (de_is_last(e))
+			break;
+
+		e = next;
+	}
+
+	return true;
+}
+
+static inline bool check_index_buffer(const struct INDEX_BUFFER *ib, u32 bytes)
+{
+	u16 fo;
+	const struct NTFS_RECORD_HEADER *r = &ib->rhdr;
+
+	if (r->sign != NTFS_INDX_SIGNATURE)
+		return false;
+
+	fo = (SECTOR_SIZE - ((bytes >> SECTOR_SHIFT) + 1) * sizeof(short));
+
+	if (le16_to_cpu(r->fix_off) > fo)
+		return false;
+
+	if ((le16_to_cpu(r->fix_num) - 1) * SECTOR_SIZE != bytes)
+		return false;
+
+	return check_index_header(&ib->ihdr,
+				  bytes - offsetof(struct INDEX_BUFFER, ihdr));
+}
+
+static inline bool check_index_root(const struct ATTRIB *attr,
+				    struct ntfs_sb_info *sbi)
+{
+	bool ret;
+	const struct INDEX_ROOT *root = resident_data(attr);
+	u8 index_bits =
+		le32_to_cpu(root->index_block_size) >= sbi->cluster_size ?
+			sbi->cluster_bits :
+			SECTOR_SHIFT;
+	u8 block_clst = root->index_block_clst;
+
+	if (le32_to_cpu(attr->res.data_size) < sizeof(struct INDEX_ROOT) ||
+	    (root->type != ATTR_NAME && root->type != ATTR_ZERO) ||
+	    (root->type == ATTR_NAME &&
+	     root->rule != NTFS_COLLATION_TYPE_FILENAME) ||
+	    (le32_to_cpu(root->index_block_size) !=
+	     (block_clst << index_bits)) ||
+	    (block_clst != 1 && block_clst != 2 && block_clst != 4 &&
+	     block_clst != 8 && block_clst != 0x10 && block_clst != 0x20 &&
+	     block_clst != 0x40 && block_clst != 0x80)) {
+		return false;
+	}
+
+	ret = check_index_header(&root->ihdr,
+				 le32_to_cpu(attr->res.data_size) -
+					 offsetof(struct INDEX_ROOT, ihdr));
+	return ret;
+}
+
+static inline bool check_attr(const struct MFT_REC *rec,
+			      const struct ATTRIB *attr,
+			      struct ntfs_sb_info *sbi)
+{
+	u32 asize = le32_to_cpu(attr->size);
+	u32 rsize = 0;
+	u64 dsize, svcn, evcn;
+	u16 run_off;
+
+	/* Check the fixed part of the attribute record header */
+	if (asize >= sbi->record_size ||
+	    asize + PtrOffset(rec, attr) >= sbi->record_size ||
+	    (attr->name_len &&
+	     le16_to_cpu(attr->name_off) + attr->name_len * sizeof(short) >
+		     asize)) {
+		return false;
+	}
+
+	/* Check the attribute fields */
+	switch (attr->non_res) {
+	case 0:
+		rsize = le32_to_cpu(attr->res.data_size);
+		if (rsize >= asize ||
+		    le16_to_cpu(attr->res.data_off) + rsize > asize) {
+			return false;
+		}
+		break;
+
+	case 1:
+		dsize = le64_to_cpu(attr->nres.data_size);
+		svcn = le64_to_cpu(attr->nres.svcn);
+		evcn = le64_to_cpu(attr->nres.evcn);
+		run_off = le16_to_cpu(attr->nres.run_off);
+
+		if (svcn > evcn + 1 || run_off >= asize ||
+		    le64_to_cpu(attr->nres.valid_size) > dsize ||
+		    dsize > le64_to_cpu(attr->nres.alloc_size)) {
+			return false;
+		}
+
+		if (run_unpack(NULL, sbi, 0, svcn, evcn, svcn,
+			       Add2Ptr(attr, run_off), asize - run_off) < 0) {
+			return false;
+		}
+
+		return true;
+
+	default:
+		return false;
+	}
+
+	switch (attr->type) {
+	case ATTR_NAME:
+		if (fname_full_size(Add2Ptr(
+			    attr, le16_to_cpu(attr->res.data_off))) > asize) {
+			return false;
+		}
+		break;
+
+	case ATTR_ROOT:
+		return check_index_root(attr, sbi);
+
+	case ATTR_STD:
+		if (rsize < sizeof(struct ATTR_STD_INFO5) &&
+		    rsize != sizeof(struct ATTR_STD_INFO)) {
+			return false;
+		}
+		break;
+
+	case ATTR_LIST:
+	case ATTR_ID:
+	case ATTR_SECURE:
+	case ATTR_LABEL:
+	case ATTR_VOL_INFO:
+	case ATTR_DATA:
+	case ATTR_ALLOC:
+	case ATTR_BITMAP:
+	case ATTR_REPARSE:
+	case ATTR_EA_INFO:
+	case ATTR_EA:
+	case ATTR_PROPERTYSET:
+	case ATTR_LOGGED_UTILITY_STREAM:
+		break;
+
+	default:
+		return false;
+	}
+
+	return true;
+}
+
+static inline bool check_file_record(const struct MFT_REC *rec,
+				     const struct MFT_REC *rec2,
+				     struct ntfs_sb_info *sbi)
+{
+	const struct ATTRIB *attr;
+	u16 fo = le16_to_cpu(rec->rhdr.fix_off);
+	u16 fn = le16_to_cpu(rec->rhdr.fix_num);
+	u16 ao = le16_to_cpu(rec->attr_off);
+	u32 rs = sbi->record_size;
+
+	/* check the file record header for consistency */
+	if (rec->rhdr.sign != NTFS_FILE_SIGNATURE ||
+	    fo > (SECTOR_SIZE - ((rs >> SECTOR_SHIFT) + 1) * sizeof(short)) ||
+	    (fn - 1) * SECTOR_SIZE != rs || ao < MFTRECORD_FIXUP_OFFSET_1 ||
+	    ao > sbi->record_size - SIZEOF_RESIDENT || !is_rec_inuse(rec) ||
+	    le32_to_cpu(rec->total) != rs) {
+		return false;
+	}
+
+	/* Loop to check all of the attributes */
+	for (attr = Add2Ptr(rec, ao); attr->type != ATTR_END;
+	     attr = Add2Ptr(attr, le32_to_cpu(attr->size))) {
+		if (check_attr(rec, attr, sbi))
+			continue;
+		return false;
+	}
+
+	return true;
+}
+
+static inline int check_lsn(const struct NTFS_RECORD_HEADER *hdr,
+			    const u64 *rlsn)
+{
+	u64 lsn;
+
+	if (!rlsn)
+		return true;
+
+	lsn = le64_to_cpu(hdr->lsn);
+
+	if (hdr->sign == NTFS_HOLE_SIGNATURE)
+		return false;
+
+	if (*rlsn > lsn)
+		return true;
+
+	return false;
+}
+
+static inline bool check_if_attr(const struct MFT_REC *rec,
+				 const struct LOG_REC_HDR *lrh)
+{
+	u16 ro = le16_to_cpu(lrh->record_off);
+	u16 o = le16_to_cpu(rec->attr_off);
+	const struct ATTRIB *attr = Add2Ptr(rec, o);
+
+	while (o < ro) {
+		u32 asize;
+
+		if (attr->type == ATTR_END)
+			break;
+
+		asize = le32_to_cpu(attr->size);
+		if (!asize)
+			break;
+
+		o += asize;
+		attr = Add2Ptr(attr, asize);
+	}
+
+	return o == ro;
+}
+
+static inline bool check_if_index_root(const struct MFT_REC *rec,
+				       const struct LOG_REC_HDR *lrh)
+{
+	u16 ro = le16_to_cpu(lrh->record_off);
+	u16 o = le16_to_cpu(rec->attr_off);
+	const struct ATTRIB *attr = Add2Ptr(rec, o);
+
+	while (o < ro) {
+		u32 asize;
+
+		if (attr->type == ATTR_END)
+			break;
+
+		asize = le32_to_cpu(attr->size);
+		if (!asize)
+			break;
+
+		o += asize;
+		attr = Add2Ptr(attr, asize);
+	}
+
+	return o == ro && attr->type == ATTR_ROOT;
+}
+
+static inline bool check_if_root_index(const struct ATTRIB *attr,
+				       const struct INDEX_HDR *hdr,
+				       const struct LOG_REC_HDR *lrh)
+{
+	u16 ao = le16_to_cpu(lrh->attr_off);
+	u32 de_off = le32_to_cpu(hdr->de_off);
+	u32 o = PtrOffset(attr, hdr) + de_off;
+	const struct NTFS_DE *e = Add2Ptr(hdr, de_off);
+	u32 asize = le32_to_cpu(attr->size);
+
+	while (o < ao) {
+		u16 esize;
+
+		if (o >= asize)
+			break;
+
+		esize = le16_to_cpu(e->size);
+		if (!esize)
+			break;
+
+		o += esize;
+		e = Add2Ptr(e, esize);
+	}
+
+	return o == ao;
+}
+
+static inline bool check_if_alloc_index(const struct INDEX_HDR *hdr,
+					u32 attr_off)
+{
+	u32 de_off = le32_to_cpu(hdr->de_off);
+	u32 o = offsetof(struct INDEX_BUFFER, ihdr) + de_off;
+	const struct NTFS_DE *e = Add2Ptr(hdr, de_off);
+	u32 used = le32_to_cpu(hdr->used);
+
+	while (o < attr_off) {
+		u16 esize;
+
+		if (de_off >= used)
+			break;
+
+		esize = le16_to_cpu(e->size);
+		if (!esize)
+			break;
+
+		o += esize;
+		de_off += esize;
+		e = Add2Ptr(e, esize);
+	}
+
+	return o == attr_off;
+}
+
+static inline void change_attr_size(struct MFT_REC *rec, struct ATTRIB *attr,
+				    u32 nsize)
+{
+	u32 asize = le32_to_cpu(attr->size);
+	int dsize = nsize - asize;
+	u8 *next = Add2Ptr(attr, asize);
+	u32 used = le32_to_cpu(rec->used);
+
+	memmove(Add2Ptr(attr, nsize), next, used - PtrOffset(rec, next));
+
+	rec->used = cpu_to_le32(used + dsize);
+	attr->size = cpu_to_le32(nsize);
+}
+
+struct OpenAttr {
+	struct ATTRIB *attr;
+	struct runs_tree *run1;
+	struct runs_tree run0;
+	struct ntfs_inode *ni;
+	// CLST rno;
+};
+
+/* Returns 0 if 'attr' has the same type and name */
+static inline int cmp_type_and_name(const struct ATTRIB *a1,
+				    const struct ATTRIB *a2)
+{
+	return a1->type != a2->type || a1->name_len != a2->name_len ||
+	       (a1->name_len && memcmp(attr_name(a1), attr_name(a2),
+				       a1->name_len * sizeof(short)));
+}
+
+static struct OpenAttr *find_loaded_attr(struct ntfs_log *log,
+					 const struct ATTRIB *attr, CLST rno)
+{
+	struct OPEN_ATTR_ENRTY *oe = NULL;
+
+	while ((oe = enum_rstbl(log->open_attr_tbl, oe))) {
+		struct OpenAttr *op_attr;
+
+		if (ino_get(&oe->ref) != rno)
+			continue;
+
+		op_attr = (struct OpenAttr *)oe->ptr;
+		if (!cmp_type_and_name(op_attr->attr, attr))
+			return op_attr;
+	}
+	return NULL;
+}
+
+static struct ATTRIB *attr_create_nonres_log(struct ntfs_sb_info *sbi,
+					     enum ATTR_TYPE type, u64 size,
+					     const u16 *name, size_t name_len,
+					     __le16 flags)
+{
+	struct ATTRIB *attr;
+	u32 name_size = QuadAlign(name_len * sizeof(short));
+	bool is_ext = flags & (ATTR_FLAG_COMPRESSED | ATTR_FLAG_SPARSED);
+	u32 asize = name_size +
+		    (is_ext ? SIZEOF_NONRESIDENT_EX : SIZEOF_NONRESIDENT);
+
+	attr = ntfs_alloc(asize, 1);
+	if (!attr)
+		return NULL;
+
+	attr->type = type;
+	attr->size = cpu_to_le32(asize);
+	attr->flags = flags;
+	attr->non_res = 1;
+	attr->name_len = name_len;
+
+	attr->nres.evcn = cpu_to_le64((u64)bytes_to_cluster(sbi, size) - 1);
+	attr->nres.alloc_size = cpu_to_le64(ntfs_up_cluster(sbi, size));
+	attr->nres.data_size = cpu_to_le64(size);
+	attr->nres.valid_size = attr->nres.data_size;
+	if (is_ext) {
+		attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
+		if (is_attr_compressed(attr))
+			attr->nres.c_unit = COMPRESSION_UNIT;
+
+		attr->nres.run_off =
+			cpu_to_le16(SIZEOF_NONRESIDENT_EX + name_size);
+		memcpy(Add2Ptr(attr, SIZEOF_NONRESIDENT_EX), name,
+		       name_len * sizeof(short));
+	} else {
+		attr->name_off = SIZEOF_NONRESIDENT_LE;
+		attr->nres.run_off =
+			cpu_to_le16(SIZEOF_NONRESIDENT + name_size);
+		memcpy(Add2Ptr(attr, SIZEOF_NONRESIDENT), name,
+		       name_len * sizeof(short));
+	}
+
+	return attr;
+}
+
+/*
+ * do_action
+ *
+ * common routine for the Redo and Undo Passes
+ * If rlsn is NULL then undo
+ */
+static int do_action(struct ntfs_log *log, struct OPEN_ATTR_ENRTY *oe,
+		     const struct LOG_REC_HDR *lrh, u32 op, void *data,
+		     u32 dlen, u32 rec_len, const u64 *rlsn)
+{
+	int err = 0;
+	struct ntfs_sb_info *sbi = log->ni->mi.sbi;
+	struct inode *inode = NULL, *inode_parent;
+	struct mft_inode *mi = NULL, *mi2_child = NULL;
+	CLST rno = 0, rno_base = 0;
+	struct INDEX_BUFFER *ib = NULL;
+	struct MFT_REC *rec = NULL;
+	struct ATTRIB *attr = NULL, *attr2;
+	struct INDEX_HDR *hdr;
+	struct INDEX_ROOT *root;
+	struct NTFS_DE *e, *e1, *e2;
+	struct NEW_ATTRIBUTE_SIZES *new_sz;
+	struct ATTR_FILE_NAME *fname;
+	struct OpenAttr *oa, *oa2;
+	u32 nsize, t32, asize, used, esize, bmp_off, bmp_bits;
+	u16 t16, id, id2;
+	u32 record_size = sbi->record_size;
+	u64 t64;
+	u64 lco = 0;
+	u64 cbo = (u64)le16_to_cpu(lrh->cluster_off) << SECTOR_SHIFT;
+	u64 tvo = le64_to_cpu(lrh->target_vcn) << sbi->cluster_bits;
+	u64 vbo = cbo + tvo;
+	void *buffer_le = NULL;
+	u32 bytes = 0;
+	bool a_dirty = false;
+	u16 data_off;
+
+	oa = oe->ptr;
+
+	/* Big switch to prepare */
+	switch (op) {
+	/* ============================================================
+	 * Process MFT records, as described by the current log record
+	 * ============================================================
+	 */
+	case InitializeFileRecordSegment:
+	case DeallocateFileRecordSegment:
+	case WriteEndOfFileRecordSegment:
+	case CreateAttribute:
+	case DeleteAttribute:
+	case UpdateResidentValue:
+	case UpdateMappingPairs:
+	case SetNewAttributeSizes:
+	case AddIndexEntryRoot:
+	case DeleteIndexEntryRoot:
+	case SetIndexEntryVcnRoot:
+	case UpdateFileNameRoot:
+	case UpdateRecordDataRoot:
+	case ZeroEndOfFileRecord:
+
+		rno = vbo >> sbi->record_bits;
+		inode = ilookup(sbi->sb, rno);
+		if (inode) {
+			mi = &ntfs_i(inode)->mi;
+		} else if (op == InitializeFileRecordSegment) {
+			mi = ntfs_alloc(sizeof(struct mft_inode), 1);
+			if (!mi)
+				return -ENOMEM;
+			err = mi_format_new(mi, sbi, rno, 0, false);
+			if (err)
+				goto out;
+		} else {
+			/* read from disk */
+			err = mi_get(sbi, rno, &mi);
+			if (err)
+				return err;
+		}
+		rec = mi->mrec;
+
+		if (op == DeallocateFileRecordSegment)
+			goto skip_load_parent;
+
+		if (InitializeFileRecordSegment != op) {
+			if (rec->rhdr.sign == NTFS_BAAD_SIGNATURE)
+				goto dirty_vol;
+			if (!check_lsn(&rec->rhdr, rlsn))
+				goto out;
+			if (!check_file_record(rec, NULL, sbi))
+				goto dirty_vol;
+			attr = Add2Ptr(rec, le16_to_cpu(lrh->record_off));
+		}
+
+		if (is_rec_base(rec) || InitializeFileRecordSegment == op) {
+			rno_base = rno;
+			goto skip_load_parent;
+		}
+
+		rno_base = ino_get(&rec->parent_ref);
+
+		inode_parent = ntfs_iget5(sbi->sb, &rec->parent_ref, NULL);
+		if (IS_ERR(inode_parent))
+			goto skip_load_parent;
+
+		if (is_bad_inode(inode_parent)) {
+			iput(inode_parent);
+			goto skip_load_parent;
+		}
+
+		if (ni_load_mi_ex(ntfs_i(inode_parent), rno, &mi2_child)) {
+			iput(inode_parent);
+		} else {
+			if (mi2_child->mrec != mi->mrec)
+				memcpy(mi2_child->mrec, mi->mrec,
+				       sbi->record_size);
+
+			if (inode)
+				iput(inode);
+			else if (mi)
+				mi_put(mi);
+
+			inode = inode_parent;
+			mi = mi2_child;
+			rec = mi2_child->mrec;
+			attr = Add2Ptr(rec, le16_to_cpu(lrh->record_off));
+		}
+
+skip_load_parent:
+		inode_parent = NULL;
+		break;
+
+	/* ============================================================
+	 * Process attributes, as described by the current log record
+	 * ============================================================
+	 */
+	case UpdateNonresidentValue:
+	case AddIndexEntryAllocation:
+	case DeleteIndexEntryAllocation:
+	case WriteEndOfIndexBuffer:
+	case SetIndexEntryVcnAllocation:
+	case UpdateFileNameAllocation:
+	case SetBitsInNonresidentBitMap:
+	case ClearBitsInNonresidentBitMap:
+	case UpdateRecordDataAllocation:
+
+		attr = oa->attr;
+		bytes = UpdateNonresidentValue == op ? dlen : 0;
+		lco = (u64)le16_to_cpu(lrh->lcns_follow) << sbi->cluster_bits;
+
+		if (attr->type == ATTR_ALLOC) {
+			t32 = le32_to_cpu(oe->bytes_per_index);
+			if (bytes < t32)
+				bytes = t32;
+		}
+
+		if (!bytes)
+			bytes = lco - cbo;
+
+		bytes += le16_to_cpu(lrh->record_off);
+		if (attr->type == ATTR_ALLOC)
+			bytes = (bytes + 511) & ~511; // align
+
+		buffer_le = ntfs_alloc(bytes, 0);
+		if (!buffer_le)
+			return -ENOMEM;
+
+		err = ntfs_read_run_nb(sbi, oa->run1, vbo, buffer_le, bytes,
+				       NULL);
+		if (err)
+			goto out;
+
+		if (attr->type == ATTR_ALLOC && *(int *)buffer_le)
+			ntfs_fix_post_read(buffer_le, bytes, false);
+		break;
+
+	default:
+		WARN_ON(1);
+	}
+
+	/* Big switch to do operation */
+	switch (op) {
+	case InitializeFileRecordSegment:
+		t16 = le16_to_cpu(lrh->record_off);
+		if (t16 + dlen > record_size)
+			goto dirty_vol;
+
+		memcpy(Add2Ptr(rec, t16), data, dlen);
+		mi->dirty = true;
+		break;
+
+	case DeallocateFileRecordSegment:
+		clear_rec_inuse(rec);
+		le16_add_cpu(&rec->seq, 1);
+		mi->dirty = true;
+		break;
+
+	case WriteEndOfFileRecordSegment:
+		attr2 = (struct ATTRIB *)data;
+		t16 = le16_to_cpu(lrh->record_off);
+
+		if (!check_if_attr(rec, lrh) || t16 + dlen > record_size)
+			goto dirty_vol;
+
+		memmove(attr, attr2, dlen);
+		rec->used = cpu_to_le32(QuadAlign(t16 + dlen));
+
+		mi->dirty = true;
+		break;
+
+	case CreateAttribute:
+		attr2 = (struct ATTRIB *)data;
+		asize = le32_to_cpu(attr2->size);
+		used = le32_to_cpu(rec->used);
+		t16 = le16_to_cpu(lrh->record_off);
+
+		if (!check_if_attr(rec, lrh) || dlen < SIZEOF_RESIDENT ||
+		    !IsQuadAligned(asize) ||
+		    Add2Ptr(attr2, asize) > Add2Ptr(lrh, rec_len) ||
+		    dlen > record_size - used) {
+			goto dirty_vol;
+		}
+
+		memmove(Add2Ptr(attr, asize), attr, used - t16);
+		memcpy(attr, attr2, asize);
+
+		rec->used = cpu_to_le32(used + asize);
+		id = le16_to_cpu(rec->next_attr_id);
+		id2 = le16_to_cpu(attr2->id);
+		if (id <= id2)
+			rec->next_attr_id = cpu_to_le16(id2 + 1);
+		if (is_attr_indexed(attr))
+			le16_add_cpu(&rec->hard_links, 1);
+
+		oa2 = find_loaded_attr(log, attr, rno_base);
+		if (oa2) {
+			void *p2 = ntfs_memdup(attr, le32_to_cpu(attr->size));
+
+			if (p2) {
+				// run_close(oa2->run1);
+				ntfs_free(oa2->attr);
+				oa2->attr = p2;
+			}
+		}
+
+		mi->dirty = true;
+		break;
+
+	case DeleteAttribute:
+		asize = le32_to_cpu(attr->size);
+		used = le32_to_cpu(rec->used);
+		t16 = le16_to_cpu(lrh->record_off);
+
+		if (!check_if_attr(rec, lrh))
+			goto dirty_vol;
+
+		rec->used = cpu_to_le32(used - asize);
+		if (is_attr_indexed(attr))
+			le16_add_cpu(&rec->hard_links, -1);
+
+		memmove(attr, Add2Ptr(attr, asize), used - t16);
+
+		mi->dirty = true;
+		break;
+
+	case UpdateResidentValue:
+		t16 = le16_to_cpu(lrh->attr_off);
+		nsize = t16 + dlen;
+
+		if (!check_if_attr(rec, lrh))
+			goto dirty_vol;
+
+		asize = le32_to_cpu(attr->size);
+		used = le32_to_cpu(rec->used);
+
+		if (lrh->redo_len == lrh->undo_len) {
+			if (nsize > asize)
+				goto dirty_vol;
+			goto move_data;
+		}
+
+		if (nsize > asize && nsize - asize > record_size - used)
+			goto dirty_vol;
+
+		nsize = QuadAlign(nsize);
+		data_off = le16_to_cpu(attr->res.data_off);
+
+		if (nsize < asize) {
+			memmove(Add2Ptr(attr, t16), data, dlen);
+			data = NULL; // To skip below memmove
+		}
+
+		memmove(Add2Ptr(attr, nsize), Add2Ptr(attr, asize),
+			used - le16_to_cpu(lrh->record_off) - asize);
+
+		rec->used = cpu_to_le32(used + nsize - asize);
+		attr->size = cpu_to_le32(nsize);
+		attr->res.data_size = cpu_to_le32(t16 + dlen - data_off);
+
+move_data:
+		if (data)
+			memmove(Add2Ptr(attr, t16), data, dlen);
+
+		oa2 = find_loaded_attr(log, attr, rno_base);
+		if (oa2) {
+			void *p2 = ntfs_memdup(attr, le32_to_cpu(attr->size));
+
+			if (p2) {
+				// run_close(&oa2->run0);
+				oa2->run1 = &oa2->run0;
+				ntfs_free(oa2->attr);
+				oa2->attr = p2;
+			}
+		}
+
+		mi->dirty = true;
+		break;
+
+	case UpdateMappingPairs:
+		t16 = le16_to_cpu(lrh->attr_off);
+		nsize = t16 + dlen;
+		asize = le32_to_cpu(attr->size);
+		used = le32_to_cpu(rec->used);
+
+		if (!check_if_attr(rec, lrh) || !attr->non_res ||
+		    t16 < le16_to_cpu(attr->nres.run_off) || t16 > asize ||
+		    (nsize > asize && nsize - asize > record_size - used)) {
+			goto dirty_vol;
+		}
+
+		nsize = QuadAlign(nsize);
+
+		memmove(Add2Ptr(attr, nsize), Add2Ptr(attr, asize),
+			used - le16_to_cpu(lrh->record_off) - asize);
+		rec->used = cpu_to_le32(used + nsize - asize);
+		attr->size = cpu_to_le32(nsize);
+		memmove(Add2Ptr(attr, t16), data, dlen);
+
+		if (run_get_highest_vcn(le64_to_cpu(attr->nres.svcn),
+					attr_run(attr), &t64)) {
+			goto dirty_vol;
+		}
+
+		attr->nres.evcn = cpu_to_le64(t64);
+		oa2 = find_loaded_attr(log, attr, rno_base);
+		if (oa2 && oa2->attr->non_res)
+			oa2->attr->nres.evcn = attr->nres.evcn;
+
+		mi->dirty = true;
+		break;
+
+	case SetNewAttributeSizes:
+		new_sz = data;
+
+		if (!check_if_attr(rec, lrh) || !attr->non_res)
+			goto dirty_vol;
+
+		attr->nres.alloc_size = new_sz->alloc_size;
+		attr->nres.data_size = new_sz->data_size;
+		attr->nres.valid_size = new_sz->valid_size;
+
+		if (dlen >= sizeof(struct NEW_ATTRIBUTE_SIZES))
+			attr->nres.total_size = new_sz->total_size;
+
+		oa2 = find_loaded_attr(log, attr, rno_base);
+		if (oa2) {
+			void *p2 = ntfs_memdup(attr, le32_to_cpu(attr->size));
+
+			if (p2) {
+				ntfs_free(oa2->attr);
+				oa2->attr = p2;
+			}
+		}
+		mi->dirty = true;
+		break;
+
+	case AddIndexEntryRoot:
+		e = (struct NTFS_DE *)data;
+		esize = le16_to_cpu(e->size);
+		root = resident_data(attr);
+		hdr = &root->ihdr;
+		used = le32_to_cpu(hdr->used);
+
+		if (!check_if_index_root(rec, lrh) ||
+		    !check_if_root_index(attr, hdr, lrh) ||
+		    Add2Ptr(data, esize) > Add2Ptr(lrh, rec_len) ||
+		    esize > le32_to_cpu(rec->total) - le32_to_cpu(rec->used)) {
+			goto dirty_vol;
+		}
+
+		e1 = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
+
+		change_attr_size(rec, attr, le32_to_cpu(attr->size) + esize);
+
+		memmove(Add2Ptr(e1, esize), e1,
+			PtrOffset(e1, Add2Ptr(hdr, used)));
+		memmove(e1, e, esize);
+
+		le32_add_cpu(&attr->res.data_size, esize);
+		hdr->used = cpu_to_le32(used + esize);
+		le32_add_cpu(&hdr->total, esize);
+
+		mi->dirty = true;
+		break;
+
+	case DeleteIndexEntryRoot:
+		root = resident_data(attr);
+		hdr = &root->ihdr;
+		used = le32_to_cpu(hdr->used);
+
+		if (!check_if_index_root(rec, lrh) ||
+		    !check_if_root_index(attr, hdr, lrh)) {
+			goto dirty_vol;
+		}
+
+		e1 = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
+		esize = le16_to_cpu(e1->size);
+		e2 = Add2Ptr(e1, esize);
+
+		memmove(e1, e2, PtrOffset(e2, Add2Ptr(hdr, used)));
+
+		le32_sub_cpu(&attr->res.data_size, esize);
+		hdr->used = cpu_to_le32(used - esize);
+		le32_sub_cpu(&hdr->total, esize);
+
+		change_attr_size(rec, attr, le32_to_cpu(attr->size) - esize);
+
+		mi->dirty = true;
+		break;
+
+	case SetIndexEntryVcnRoot:
+		root = resident_data(attr);
+		hdr = &root->ihdr;
+
+		if (!check_if_index_root(rec, lrh) ||
+		    !check_if_root_index(attr, hdr, lrh)) {
+			goto dirty_vol;
+		}
+
+		e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
+
+		de_set_vbn_le(e, *(__le64 *)data);
+		mi->dirty = true;
+		break;
+
+	case UpdateFileNameRoot:
+		root = resident_data(attr);
+		hdr = &root->ihdr;
+
+		if (!check_if_index_root(rec, lrh) ||
+		    !check_if_root_index(attr, hdr, lrh)) {
+			goto dirty_vol;
+		}
+
+		e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
+		fname = (struct ATTR_FILE_NAME *)(e + 1);
+		memmove(&fname->dup, data, sizeof(fname->dup)); //
+		mi->dirty = true;
+		break;
+
+	case UpdateRecordDataRoot:
+		root = resident_data(attr);
+		hdr = &root->ihdr;
+
+		if (!check_if_index_root(rec, lrh) ||
+		    !check_if_root_index(attr, hdr, lrh)) {
+			goto dirty_vol;
+		}
+
+		e = Add2Ptr(attr, le16_to_cpu(lrh->attr_off));
+
+		memmove(Add2Ptr(e, le16_to_cpu(e->view.data_off)), data, dlen);
+
+		mi->dirty = true;
+		break;
+
+	case ZeroEndOfFileRecord:
+		t16 = le16_to_cpu(lrh->record_off);
+		if (t16 + dlen > record_size)
+			goto dirty_vol;
+
+		memset(attr, 0, dlen);
+		mi->dirty = true;
+		break;
+
+	case UpdateNonresidentValue:
+		t16 = le16_to_cpu(lrh->record_off);
+
+		if (lco < cbo + t16 + dlen)
+			goto dirty_vol;
+
+		memcpy(Add2Ptr(buffer_le, t16), data, dlen);
+
+		a_dirty = true;
+		if (attr->type == ATTR_ALLOC)
+			ntfs_fix_pre_write(buffer_le, bytes);
+		break;
+
+	case AddIndexEntryAllocation:
+		t16 = le16_to_cpu(lrh->record_off);
+		ib = Add2Ptr(buffer_le, t16);
+		hdr = &ib->ihdr;
+		e = data;
+		esize = le16_to_cpu(e->size);
+		t16 = le16_to_cpu(lrh->attr_off);
+		e1 = Add2Ptr(ib, t16);
+
+		if (is_baad(&ib->rhdr))
+			goto dirty_vol;
+		if (!check_lsn(&ib->rhdr, rlsn))
+			goto out;
+
+		used = le32_to_cpu(hdr->used);
+
+		if (!check_index_buffer(ib, bytes) ||
+		    !check_if_alloc_index(hdr, t16) ||
+		    Add2Ptr(e, esize) > Add2Ptr(lrh, rec_len) ||
+		    used + esize > le32_to_cpu(hdr->total)) {
+			goto dirty_vol;
+		}
+
+		memmove(Add2Ptr(e1, esize), e1,
+			PtrOffset(e1, Add2Ptr(hdr, used)));
+		memcpy(e1, e, esize);
+
+		hdr->used = cpu_to_le32(used + esize);
+
+		a_dirty = true;
+
+		ntfs_fix_pre_write(&ib->rhdr, bytes);
+		break;
+
+	case DeleteIndexEntryAllocation:
+		t16 = le16_to_cpu(lrh->record_off);
+		ib = Add2Ptr(buffer_le, t16);
+		hdr = &ib->ihdr;
+		t16 = le16_to_cpu(lrh->attr_off);
+		e = Add2Ptr(ib, t16);
+		esize = le16_to_cpu(e->size);
+
+		if (is_baad(&ib->rhdr))
+			goto dirty_vol;
+		if (!check_lsn(&ib->rhdr, rlsn))
+			goto out;
+
+		if (!check_index_buffer(ib, bytes) ||
+		    !check_if_alloc_index(hdr, t16)) {
+			goto dirty_vol;
+		}
+
+		e1 = Add2Ptr(e, esize);
+		nsize = esize;
+		used = le32_to_cpu(hdr->used);
+
+		memmove(e, e1, PtrOffset(e1, Add2Ptr(hdr, used)));
+
+		hdr->used = cpu_to_le32(used - nsize);
+
+		a_dirty = true;
+
+		ntfs_fix_pre_write(&ib->rhdr, bytes);
+		break;
+
+	case WriteEndOfIndexBuffer:
+		t16 = le16_to_cpu(lrh->record_off);
+		ib = Add2Ptr(buffer_le, t16);
+		hdr = &ib->ihdr;
+		t16 = le16_to_cpu(lrh->attr_off);
+		e = Add2Ptr(ib, t16);
+
+		if (is_baad(&ib->rhdr))
+			goto dirty_vol;
+		if (!check_lsn(&ib->rhdr, rlsn))
+			goto out;
+		if (!check_index_buffer(ib, bytes) ||
+		    !check_if_alloc_index(hdr, t16) ||
+		    t16 + dlen > offsetof(struct INDEX_BUFFER, ihdr) +
+					 le32_to_cpu(hdr->total)) {
+			goto dirty_vol;
+		}
+
+		hdr->used = cpu_to_le32(dlen + PtrOffset(hdr, e));
+		memmove(e, data, dlen);
+
+		a_dirty = true;
+		ntfs_fix_pre_write(&ib->rhdr, bytes);
+		break;
+
+	case SetIndexEntryVcnAllocation:
+		t16 = le16_to_cpu(lrh->record_off);
+		ib = Add2Ptr(buffer_le, t16);
+		hdr = &ib->ihdr;
+		t16 = le16_to_cpu(lrh->attr_off);
+		e = Add2Ptr(ib, t16);
+
+		if (is_baad(&ib->rhdr))
+			goto dirty_vol;
+
+		if (!check_lsn(&ib->rhdr, rlsn))
+			goto out;
+		if (!check_index_buffer(ib, bytes) ||
+		    !check_if_alloc_index(hdr, t16)) {
+			goto dirty_vol;
+		}
+
+		de_set_vbn_le(e, *(__le64 *)data);
+
+		a_dirty = true;
+		ntfs_fix_pre_write(&ib->rhdr, bytes);
+		break;
+
+	case UpdateFileNameAllocation:
+		t16 = le16_to_cpu(lrh->record_off);
+		ib = Add2Ptr(buffer_le, t16);
+		hdr = &ib->ihdr;
+		t16 = le16_to_cpu(lrh->attr_off);
+		e = Add2Ptr(ib, t16);
+
+		if (is_baad(&ib->rhdr))
+			goto dirty_vol;
+
+		if (!check_lsn(&ib->rhdr, rlsn))
+			goto out;
+		if (!check_index_buffer(ib, bytes) ||
+		    !check_if_alloc_index(hdr, t16)) {
+			goto dirty_vol;
+		}
+
+		fname = (struct ATTR_FILE_NAME *)(e + 1);
+		memmove(&fname->dup, data, sizeof(fname->dup));
+
+		a_dirty = true;
+		ntfs_fix_pre_write(&ib->rhdr, bytes);
+		break;
+
+	case SetBitsInNonresidentBitMap:
+		bmp_off =
+			le32_to_cpu(((struct BITMAP_RANGE *)data)->bitmap_off);
+		bmp_bits = le32_to_cpu(((struct BITMAP_RANGE *)data)->bits);
+		t16 = le16_to_cpu(lrh->record_off);
+
+		if (cbo + (bmp_off + 7) / 8 > lco ||
+		    cbo + ((bmp_off + bmp_bits + 7) / 8) > lco) {
+			goto dirty_vol;
+		}
+
+		__bitmap_set(Add2Ptr(buffer_le, t16), bmp_off, bmp_bits);
+		a_dirty = true;
+		break;
+
+	case ClearBitsInNonresidentBitMap:
+		bmp_off =
+			le32_to_cpu(((struct BITMAP_RANGE *)data)->bitmap_off);
+		bmp_bits = le32_to_cpu(((struct BITMAP_RANGE *)data)->bits);
+		t16 = le16_to_cpu(lrh->record_off);
+
+		if (cbo + (bmp_off + 7) / 8 > lco ||
+		    cbo + ((bmp_off + bmp_bits + 7) / 8) > lco) {
+			goto dirty_vol;
+		}
+
+		__bitmap_clear(Add2Ptr(buffer_le, t16), bmp_off, bmp_bits);
+		a_dirty = true;
+		break;
+
+	case UpdateRecordDataAllocation:
+		t16 = le16_to_cpu(lrh->record_off);
+		ib = Add2Ptr(buffer_le, t16);
+		hdr = &ib->ihdr;
+		t16 = le16_to_cpu(lrh->attr_off);
+		e = Add2Ptr(ib, t16);
+
+		if (is_baad(&ib->rhdr))
+			goto dirty_vol;
+
+		if (!check_lsn(&ib->rhdr, rlsn))
+			goto out;
+		if (!check_index_buffer(ib, bytes) ||
+		    !check_if_alloc_index(hdr, t16)) {
+			goto dirty_vol;
+		}
+
+		memmove(Add2Ptr(e, le16_to_cpu(e->view.data_off)), data, dlen);
+
+		a_dirty = true;
+		ntfs_fix_pre_write(&ib->rhdr, bytes);
+		break;
+
+	default:
+		WARN_ON(1);
+	}
+
+	if (rlsn) {
+		__le64 t64 = cpu_to_le64(*rlsn);
+
+		if (rec)
+			rec->rhdr.lsn = t64;
+		if (ib)
+			ib->rhdr.lsn = t64;
+	}
+
+	if (inode) {
+		err = _ni_write_inode(inode, 0);
+	} else if (mi && mi->dirty) {
+		err = mi_write(mi, 0);
+		if (err)
+			goto out;
+	}
+
+	if (a_dirty) {
+		attr = oa->attr;
+		err = ntfs_sb_write_run(sbi, oa->run1, vbo, buffer_le, bytes);
+		if (err)
+			goto out;
+	}
+
+out:
+
+	if (inode)
+		iput(inode);
+	else if (mi != mi2_child)
+		mi_put(mi);
+
+	ntfs_free(buffer_le);
+
+	return err;
+
+dirty_vol:
+	log->set_dirty = true;
+	goto out;
+}
+
+/*
+ * log_replay
+ *
+ * this function is called during mount operation
+ * it replays log and empties it
+ */
+int log_replay(struct ntfs_inode *ni)
+{
+	int err;
+	struct ntfs_sb_info *sbi = ni->mi.sbi;
+	struct ntfs_log *log;
+
+	struct restart_info rst_info, rst_info2;
+	u64 rec_lsn, ra_lsn, checkpt_lsn = 0, rlsn = 0;
+	struct ATTR_NAME_ENTRY *attr_names = NULL;
+	struct ATTR_NAME_ENTRY *ane;
+	struct RESTART_TABLE *dptbl = NULL;
+	struct RESTART_TABLE *trtbl = NULL;
+	const struct RESTART_TABLE *rt;
+	struct RESTART_TABLE *oatbl = NULL;
+	struct inode *inode;
+	struct OpenAttr *oa;
+	struct ntfs_inode *ni_oe;
+	struct ATTRIB *attr = NULL;
+	u64 size, vcn, undo_next_lsn;
+	CLST rno, lcn, lcn0, len0, clen;
+	void *data;
+	struct NTFS_RESTART *rst = NULL;
+	struct lcb *lcb = NULL;
+	struct OPEN_ATTR_ENRTY *oe;
+	struct TRANSACTION_ENTRY *tr;
+	struct DIR_PAGE_ENTRY *dp;
+	u32 i, bytes_per_attr_entry;
+	u32 l_size = ni->vfs_inode.i_size;
+	u32 orig_file_size = l_size;
+	u32 page_size, vbo, tail, off, dlen;
+	u32 saved_len, rec_len, transact_id;
+	bool use_second_page;
+	struct RESTART_AREA *ra2, *ra = NULL;
+	struct CLIENT_REC *ca, *cr;
+	__le16 client;
+	struct RESTART_HDR *rh;
+	const struct LFS_RECORD_HDR *frh;
+	const struct LOG_REC_HDR *lrh;
+	bool is_mapped;
+	bool is_ro = sb_rdonly(sbi->sb);
+	u64 t64;
+	u16 t16;
+	u32 t32;
+
+	/* Get the size of page. NOTE: To replay we can use default page */
+	page_size = norm_file_page(PAGE_SIZE, &l_size,
+				   PAGE_SIZE >= DefaultLogPageSize &&
+					   PAGE_SIZE <= DefaultLogPageSize * 2);
+	if (!page_size)
+		return -EINVAL;
+
+	log = ntfs_alloc(sizeof(struct ntfs_log), 1);
+	if (!log)
+		return -ENOMEM;
+
+	log->ni = ni;
+	log->l_size = l_size;
+	log->one_page_buf = ntfs_alloc(page_size, 0);
+
+	if (!log->one_page_buf) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	log->page_size = page_size;
+	log->page_mask = page_size - 1;
+	log->page_bits = blksize_bits(page_size);
+
+	/* Look for a restart area on the disk */
+	err = log_read_rst(log, l_size, true, &rst_info);
+	if (err)
+		goto out;
+
+	if (!rst_info.restart) {
+		if (rst_info.initialized) {
+			/* no restart area but the file is not initialized */
+			err = -EINVAL;
+			goto out;
+		}
+
+		log_init_pg_hdr(log, page_size, page_size, 1, 1);
+		log_create(log, l_size, 0, get_random_int(), false, false);
+
+		log->ra = ra;
+
+		ra = log_create_ra(log);
+		if (!ra) {
+			err = -ENOMEM;
+			goto out;
+		}
+		log->ra = ra;
+		log->init_ra = true;
+
+		goto process_log;
+	}
+
+	/*
+	 * If the restart offset above wasn't zero then we won't
+	 * look for a second restart
+	 */
+	if (rst_info.vbo)
+		goto check_restart_area;
+
+	err = log_read_rst(log, l_size, false, &rst_info2);
+
+	/* Determine which restart area to use */
+	if (!rst_info2.restart || rst_info2.last_lsn <= rst_info.last_lsn)
+		goto use_first_page;
+
+	use_second_page = true;
+
+	if (rst_info.chkdsk_was_run && page_size != rst_info.vbo) {
+		struct RECORD_PAGE_HDR *sp = NULL;
+
+		if (!read_log_page(log, page_size, &sp, true, true, NULL) &&
+		    sp->rhdr.sign == NTFS_CHKD_SIGNATURE) {
+			use_second_page = false;
+		}
+		ntfs_free(sp);
+	}
+
+	if (use_second_page) {
+		ntfs_free(rst_info.r_page);
+		memcpy(&rst_info, &rst_info2, sizeof(struct restart_info));
+		rst_info2.r_page = NULL;
+	}
+
+use_first_page:
+	ntfs_free(rst_info2.r_page);
+
+check_restart_area:
+	/* If the restart area is at offset 0, we want to write the second restart area first */
+	log->init_ra = !!rst_info.vbo;
+
+	/* If we have a valid page then grab a pointer to the restart area */
+	ra2 = rst_info.valid_page ?
+		      Add2Ptr(rst_info.r_page,
+			      le16_to_cpu(rst_info.r_page->ra_off)) :
+		      NULL;
+
+	if (rst_info.chkdsk_was_run ||
+	    (ra2 && ra2->client_idx[1] == LFS_NO_CLIENT_LE)) {
+		bool wrapped = false;
+		bool use_multi_page = false;
+		u32 open_log_count;
+
+		/* Do some checks based on whether we have a valid log page */
+		if (!rst_info.valid_page) {
+			open_log_count = get_random_int();
+			goto init_log_instance;
+		}
+		open_log_count = le32_to_cpu(ra2->open_log_count);
+
+		/*
+		 * If the restart page size isn't changing then we want to
+		 * check how much work we need to do
+		 */
+		if (page_size != le32_to_cpu(rst_info.r_page->sys_page_size))
+			goto init_log_instance;
+
+init_log_instance:
+		log_init_pg_hdr(log, page_size, page_size, 1, 1);
+
+		log_create(log, l_size, rst_info.last_lsn, open_log_count,
+			   wrapped, use_multi_page);
+
+		ra = log_create_ra(log);
+		if (!ra) {
+			err = -ENOMEM;
+			goto out;
+		}
+		log->ra = ra;
+
+		/* Put the restart areas and initialize the log file as required */
+		goto process_log;
+	}
+
+	if (!ra2) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/*
+	 * If the log page or the system page sizes have changed, we can't use the log file
+	 * We must use the system page size instead of the default size
+	 * if there is not a clean shutdown
+	 */
+	t32 = le32_to_cpu(rst_info.r_page->sys_page_size);
+	if (page_size != t32) {
+		l_size = orig_file_size;
+		page_size =
+			norm_file_page(t32, &l_size, t32 == DefaultLogPageSize);
+	}
+
+	if (page_size != t32 ||
+	    page_size != le32_to_cpu(rst_info.r_page->page_size)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* If the file size has shrunk then we won't mount it */
+	if (l_size < le64_to_cpu(ra2->l_size)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	log_init_pg_hdr(log, page_size, page_size,
+			le16_to_cpu(rst_info.r_page->major_ver),
+			le16_to_cpu(rst_info.r_page->minor_ver));
+
+	log->l_size = le64_to_cpu(ra2->l_size);
+	log->seq_num_bits = le32_to_cpu(ra2->seq_num_bits);
+	log->file_data_bits = sizeof(u64) * 8 - log->seq_num_bits;
+	log->seq_num_mask = (8 << log->file_data_bits) - 1;
+	log->last_lsn = le64_to_cpu(ra2->current_lsn);
+	log->seq_num = log->last_lsn >> log->file_data_bits;
+	log->ra_off = le16_to_cpu(rst_info.r_page->ra_off);
+	log->restart_size = log->sys_page_size - log->ra_off;
+	log->record_header_len = le16_to_cpu(ra2->rec_hdr_len);
+	log->ra_size = le16_to_cpu(ra2->ra_len);
+	log->data_off = le16_to_cpu(ra2->data_off);
+	log->data_size = log->page_size - log->data_off;
+	log->reserved = log->data_size - log->record_header_len;
+
+	vbo = lsn_to_vbo(log, log->last_lsn);
+
+	if (vbo < log->first_page) {
+		/* This is a pseudo lsn */
+		log->l_flags |= NTFSLOG_NO_LAST_LSN;
+		log->next_page = log->first_page;
+		goto find_oldest;
+	}
+
+	/* Find the end of this log record */
+	off = final_log_off(log, log->last_lsn,
+			    le32_to_cpu(ra2->last_lsn_data_len));
+
+	/* If we wrapped the file then increment the sequence number */
+	if (off <= vbo) {
+		log->seq_num += 1;
+		log->l_flags |= NTFSLOG_WRAPPED;
+	}
+
+	/* Now compute the next log page to use */
+	vbo &= ~log->sys_page_mask;
+	tail = log->page_size - (off & log->page_mask) - 1;
+
+	/* If we can fit another log record on the page, move back a page the log file */
+	if (tail >= log->record_header_len) {
+		log->l_flags |= NTFSLOG_REUSE_TAIL;
+		log->next_page = vbo;
+	} else {
+		log->next_page = next_page_off(log, vbo);
+	}
+
+find_oldest:
+	/* Find the oldest client lsn. Use the last flushed lsn as a starting point */
+	log->oldest_lsn = log->last_lsn;
+	oldest_client_lsn(Add2Ptr(ra2, le16_to_cpu(ra2->client_off)),
+			  ra2->client_idx[1], &log->oldest_lsn);
+	log->oldest_lsn_off = lsn_to_vbo(log, log->oldest_lsn);
+
+	if (log->oldest_lsn_off < log->first_page)
+		log->l_flags |= NTFSLOG_NO_OLDEST_LSN;
+
+	if (!(ra2->flags & RESTART_SINGLE_PAGE_IO))
+		log->l_flags |= NTFSLOG_WRAPPED | NTFSLOG_MULTIPLE_PAGE_IO;
+
+	log->current_openlog_count = le32_to_cpu(ra2->open_log_count);
+	log->total_avail_pages = log->l_size - log->first_page;
+	log->total_avail = log->total_avail_pages >> log->page_bits;
+	log->max_current_avail = log->total_avail * log->reserved;
+	log->total_avail = log->total_avail * log->data_size;
+
+	log->current_avail = current_log_avail(log);
+
+	ra = ntfs_alloc(log->restart_size, 1);
+	if (!ra) {
+		err = -ENOMEM;
+		goto out;
+	}
+	log->ra = ra;
+
+	t16 = le16_to_cpu(ra2->client_off);
+	if (t16 == offsetof(struct RESTART_AREA, clients)) {
+		memcpy(ra, ra2, log->ra_size);
+	} else {
+		memcpy(ra, ra2, offsetof(struct RESTART_AREA, clients));
+		memcpy(ra->clients, Add2Ptr(ra2, t16),
+		       le16_to_cpu(ra2->ra_len) - t16);
+
+		log->current_openlog_count = get_random_int();
+		ra->open_log_count = cpu_to_le32(log->current_openlog_count);
+		log->ra_size = offsetof(struct RESTART_AREA, clients) +
+			       sizeof(struct CLIENT_REC);
+		ra->client_off =
+			cpu_to_le16(offsetof(struct RESTART_AREA, clients));
+		ra->ra_len = cpu_to_le16(log->ra_size);
+	}
+
+	le32_add_cpu(&ra->open_log_count, 1);
+
+	/* Now we need to walk through looking for the last lsn */
+	err = last_log_lsn(log);
+	if (err == -EROFS)
+		goto out;
+
+	if (err)
+		goto out;
+
+	log->current_avail = current_log_avail(log);
+
+	/* Remember which restart area to write first */
+	log->init_ra = rst_info.vbo;
+
+process_log:
+	/* 1.0, 1.1, 2.0 log->major_ver/minor_ver - short values */
+	switch ((log->major_ver << 16) + log->minor_ver) {
+	case 0x10000:
+	case 0x10001:
+	case 0x20000:
+		break;
+	default:
+		ntfs_warn(sbi->sb, "$LogFile version %d.%d is not supported",
+			  log->major_ver, log->minor_ver);
+		err = -EOPNOTSUPP;
+		log->set_dirty = true;
+		goto out;
+	}
+
+	/* One client "NTFS" per logfile */
+	ca = Add2Ptr(ra, le16_to_cpu(ra->client_off));
+
+	for (client = ra->client_idx[1];; client = cr->next_client) {
+		if (client == LFS_NO_CLIENT_LE) {
+			/* Insert "NTFS" client LogFile */
+			client = ra->client_idx[0];
+			if (client == LFS_NO_CLIENT_LE)
+				return -EINVAL;
+
+			t16 = le16_to_cpu(client);
+			cr = ca + t16;
+
+			remove_client(ca, cr, &ra->client_idx[0]);
+
+			cr->restart_lsn = 0;
+			cr->oldest_lsn = cpu_to_le64(log->oldest_lsn);
+			cr->name_bytes = cpu_to_le32(8);
+			cr->name[0] = cpu_to_le16('N');
+			cr->name[1] = cpu_to_le16('T');
+			cr->name[2] = cpu_to_le16('F');
+			cr->name[3] = cpu_to_le16('S');
+
+			add_client(ca, t16, &ra->client_idx[1]);
+			break;
+		}
+
+		cr = ca + le16_to_cpu(client);
+
+		if (cpu_to_le32(8) == cr->name_bytes &&
+		    cpu_to_le16('N') == cr->name[0] &&
+		    cpu_to_le16('T') == cr->name[1] &&
+		    cpu_to_le16('F') == cr->name[2] &&
+		    cpu_to_le16('S') == cr->name[3])
+			break;
+	}
+
+	/* Update the client handle with the client block information */
+	log->client_id.seq_num = cr->seq_num;
+	log->client_id.client_idx = client;
+
+	err = read_rst_area(log, &rst, &ra_lsn);
+	if (err)
+		goto out;
+
+	if (!rst)
+		goto out;
+
+	bytes_per_attr_entry = !rst->major_ver ? 0x2C : 0x28;
+
+	checkpt_lsn = le64_to_cpu(rst->check_point_start);
+	if (!checkpt_lsn)
+		checkpt_lsn = ra_lsn;
+
+	/* Allocate and Read the Transaction Table */
+	if (!rst->transact_table_len)
+		goto check_dirty_page_table;
+
+	t64 = le64_to_cpu(rst->transact_table_lsn);
+	err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
+	if (err)
+		goto out;
+
+	lrh = lcb->log_rec;
+	frh = lcb->lrh;
+	rec_len = le32_to_cpu(frh->client_data_len);
+
+	if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
+			   bytes_per_attr_entry)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	t16 = le16_to_cpu(lrh->redo_off);
+
+	rt = Add2Ptr(lrh, t16);
+	t32 = rec_len - t16;
+
+	/* Now check that this is a valid restart table */
+	if (!check_rstbl(rt, t32)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	trtbl = ntfs_memdup(rt, t32);
+	if (!trtbl) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	lcb_put(lcb);
+	lcb = NULL;
+
+check_dirty_page_table:
+	/* The next record back should be the Dirty Pages Table */
+	if (!rst->dirty_pages_len)
+		goto check_attribute_names;
+
+	t64 = le64_to_cpu(rst->dirty_pages_table_lsn);
+	err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
+	if (err)
+		goto out;
+
+	lrh = lcb->log_rec;
+	frh = lcb->lrh;
+	rec_len = le32_to_cpu(frh->client_data_len);
+
+	if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
+			   bytes_per_attr_entry)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	t16 = le16_to_cpu(lrh->redo_off);
+
+	rt = Add2Ptr(lrh, t16);
+	t32 = rec_len - t16;
+
+	/* Now check that this is a valid restart table */
+	if (!check_rstbl(rt, t32)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	dptbl = ntfs_memdup(rt, t32);
+	if (!dptbl) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	/* Convert Ra version '0' into version '1' */
+	if (rst->major_ver)
+		goto end_conv_1;
+
+	dp = NULL;
+	while ((dp = enum_rstbl(dptbl, dp))) {
+		struct DIR_PAGE_ENTRY_32 *dp0 = (struct DIR_PAGE_ENTRY_32 *)dp;
+		// NOTE: Danger. Check for of boundary
+		memmove(&dp->vcn, &dp0->vcn_low,
+			2 * sizeof(u64) +
+				le32_to_cpu(dp->lcns_follow) * sizeof(u64));
+	}
+
+end_conv_1:
+	lcb_put(lcb);
+	lcb = NULL;
+
+	/* Go through the table and remove the duplicates, remembering the oldest lsn values */
+	if (sbi->cluster_size <= log->page_size)
+		goto trace_dp_table;
+
+	dp = NULL;
+	while ((dp = enum_rstbl(dptbl, dp))) {
+		struct DIR_PAGE_ENTRY *next = dp;
+
+		while ((next = enum_rstbl(dptbl, next))) {
+			if (next->target_attr == dp->target_attr &&
+			    next->vcn == dp->vcn) {
+				if (le64_to_cpu(next->oldest_lsn) <
+				    le64_to_cpu(dp->oldest_lsn)) {
+					dp->oldest_lsn = next->oldest_lsn;
+				}
+
+				free_rsttbl_idx(dptbl, PtrOffset(dptbl, next));
+			}
+		}
+	}
+trace_dp_table:
+check_attribute_names:
+	/* The next record should be the Attribute Names */
+	if (!rst->attr_names_len)
+		goto check_attr_table;
+
+	t64 = le64_to_cpu(rst->attr_names_lsn);
+	err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
+	if (err)
+		goto out;
+
+	lrh = lcb->log_rec;
+	frh = lcb->lrh;
+	rec_len = le32_to_cpu(frh->client_data_len);
+
+	if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
+			   bytes_per_attr_entry)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	t32 = lrh_length(lrh);
+	rec_len -= t32;
+
+	attr_names = ntfs_memdup(Add2Ptr(lrh, t32), rec_len);
+
+	lcb_put(lcb);
+	lcb = NULL;
+
+check_attr_table:
+	/* The next record should be the attribute Table */
+	if (!rst->open_attr_len)
+		goto check_attribute_names2;
+
+	t64 = le64_to_cpu(rst->open_attr_table_lsn);
+	err = read_log_rec_lcb(log, t64, lcb_ctx_prev, &lcb);
+	if (err)
+		goto out;
+
+	lrh = lcb->log_rec;
+	frh = lcb->lrh;
+	rec_len = le32_to_cpu(frh->client_data_len);
+
+	if (!check_log_rec(lrh, rec_len, le32_to_cpu(frh->transact_id),
+			   bytes_per_attr_entry)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	t16 = le16_to_cpu(lrh->redo_off);
+
+	rt = Add2Ptr(lrh, t16);
+	t32 = rec_len - t16;
+
+	if (!check_rstbl(rt, t32)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	oatbl = ntfs_memdup(rt, t32);
+	if (!oatbl) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	log->open_attr_tbl = oatbl;
+
+	/* Clear all of the Attr pointers */
+	oe = NULL;
+	while ((oe = enum_rstbl(oatbl, oe))) {
+		if (!rst->major_ver) {
+			struct OPEN_ATTR_ENRTY_32 oe0;
+
+			/* Really 'oe' points to OPEN_ATTR_ENRTY_32 */
+			memcpy(&oe0, oe, SIZEOF_OPENATTRIBUTEENTRY0);
+
+			oe->bytes_per_index = oe0.bytes_per_index;
+			oe->type = oe0.type;
+			oe->is_dirty_pages = oe0.is_dirty_pages;
+			oe->name_len = 0;
+			oe->ref = oe0.ref;
+			oe->open_record_lsn = oe0.open_record_lsn;
+		}
+
+		oe->is_attr_name = 0;
+		oe->ptr = NULL;
+	}
+
+	lcb_put(lcb);
+	lcb = NULL;
+
+check_attribute_names2:
+	if (!rst->attr_names_len)
+		goto trace_attribute_table;
+
+	ane = attr_names;
+	if (!oatbl)
+		goto trace_attribute_table;
+	while (ane->off) {
+		/* TODO: Clear table on exit! */
+		oe = Add2Ptr(oatbl, le16_to_cpu(ane->off));
+		t16 = le16_to_cpu(ane->name_bytes);
+		oe->name_len = t16 / sizeof(short);
+		oe->ptr = ane->name;
+		oe->is_attr_name = 2;
+		ane = Add2Ptr(ane, sizeof(struct ATTR_NAME_ENTRY) + t16);
+	}
+
+trace_attribute_table:
+	/*
+	 * If the checkpt_lsn is zero, then this is a freshly
+	 * formatted disk and we have no work to do
+	 */
+	if (!checkpt_lsn) {
+		err = 0;
+		goto out;
+	}
+
+	if (!oatbl) {
+		oatbl = init_rsttbl(bytes_per_attr_entry, 8);
+		if (!oatbl) {
+			err = -ENOMEM;
+			goto out;
+		}
+	}
+
+	log->open_attr_tbl = oatbl;
+
+	/* Start the analysis pass from the Checkpoint lsn. */
+	rec_lsn = checkpt_lsn;
+
+	/* Read the first lsn */
+	err = read_log_rec_lcb(log, checkpt_lsn, lcb_ctx_next, &lcb);
+	if (err)
+		goto out;
+
+	/* Loop to read all subsequent records to the end of the log file */
+next_log_record_analyze:
+	err = read_next_log_rec(log, lcb, &rec_lsn);
+	if (err)
+		goto out;
+
+	if (!rec_lsn)
+		goto end_log_records_enumerate;
+
+	frh = lcb->lrh;
+	transact_id = le32_to_cpu(frh->transact_id);
+	rec_len = le32_to_cpu(frh->client_data_len);
+	lrh = lcb->log_rec;
+
+	if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/*
+	 * The first lsn after the previous lsn remembered
+	 * the checkpoint is the first candidate for the rlsn
+	 */
+	if (!rlsn)
+		rlsn = rec_lsn;
+
+	if (LfsClientRecord != frh->record_type)
+		goto next_log_record_analyze;
+
+	/*
+	 * Now update the Transaction Table for this transaction
+	 * If there is no entry present or it is unallocated we allocate the entry
+	 */
+	if (!trtbl) {
+		trtbl = init_rsttbl(sizeof(struct TRANSACTION_ENTRY),
+				    INITIAL_NUMBER_TRANSACTIONS);
+		if (!trtbl) {
+			err = -ENOMEM;
+			goto out;
+		}
+	}
+
+	tr = Add2Ptr(trtbl, transact_id);
+
+	if (transact_id >= bytes_per_rt(trtbl) ||
+	    tr->next != RESTART_ENTRY_ALLOCATED_LE) {
+		tr = alloc_rsttbl_from_idx(&trtbl, transact_id);
+		if (!tr) {
+			err = -ENOMEM;
+			goto out;
+		}
+		tr->transact_state = TransactionActive;
+		tr->first_lsn = cpu_to_le64(rec_lsn);
+	}
+
+	tr->prev_lsn = tr->undo_next_lsn = cpu_to_le64(rec_lsn);
+
+	/*
+	 * If this is a compensation log record, then change
+	 * the undo_next_lsn to be the undo_next_lsn of this record
+	 */
+	if (lrh->undo_op == cpu_to_le16(CompensationLogRecord))
+		tr->undo_next_lsn = frh->client_undo_next_lsn;
+
+	/* Dispatch to handle log record depending on type */
+	switch (le16_to_cpu(lrh->redo_op)) {
+	case InitializeFileRecordSegment:
+	case DeallocateFileRecordSegment:
+	case WriteEndOfFileRecordSegment:
+	case CreateAttribute:
+	case DeleteAttribute:
+	case UpdateResidentValue:
+	case UpdateNonresidentValue:
+	case UpdateMappingPairs:
+	case SetNewAttributeSizes:
+	case AddIndexEntryRoot:
+	case DeleteIndexEntryRoot:
+	case AddIndexEntryAllocation:
+	case DeleteIndexEntryAllocation:
+	case WriteEndOfIndexBuffer:
+	case SetIndexEntryVcnRoot:
+	case SetIndexEntryVcnAllocation:
+	case UpdateFileNameRoot:
+	case UpdateFileNameAllocation:
+	case SetBitsInNonresidentBitMap:
+	case ClearBitsInNonresidentBitMap:
+	case UpdateRecordDataRoot:
+	case UpdateRecordDataAllocation:
+	case ZeroEndOfFileRecord:
+		t16 = le16_to_cpu(lrh->target_attr);
+		t64 = le64_to_cpu(lrh->target_vcn);
+		dp = find_dp(dptbl, t16, t64);
+
+		if (dp)
+			goto copy_lcns;
+
+		/*
+		 * Calculate the number of clusters per page the system
+		 * which wrote the checkpoint, possibly creating the table
+		 */
+		if (dptbl) {
+			t32 = 1 + (le16_to_cpu(dptbl->size) -
+				   sizeof(struct DIR_PAGE_ENTRY)) /
+					  sizeof(u64);
+		} else {
+			t32 = log->clst_per_page;
+			ntfs_free(dptbl);
+			dptbl = init_rsttbl(sizeof(struct DIR_PAGE_ENTRY) +
+						    (t32 - 1) * sizeof(u64),
+					    32);
+			if (!dptbl) {
+				err = -ENOMEM;
+				goto out;
+			}
+		}
+
+		dp = alloc_rsttbl_idx(&dptbl);
+		dp->target_attr = cpu_to_le32(t16);
+		dp->transfer_len = cpu_to_le32(t32 << sbi->cluster_bits);
+		dp->lcns_follow = cpu_to_le32(t32);
+		dp->vcn = cpu_to_le64(t64 & ~((u64)t32 - 1));
+		dp->oldest_lsn = cpu_to_le64(rec_lsn);
+
+copy_lcns:
+		/*
+		 * Copy the Lcns from the log record into the Dirty Page Entry
+		 * TODO: for different page size support, must somehow make
+		 * whole routine a loop, case Lcns do not fit below
+		 */
+		t16 = le16_to_cpu(lrh->lcns_follow);
+		for (i = 0; i < t16; i++) {
+			size_t j = (size_t)(le64_to_cpu(lrh->target_vcn) -
+					    le64_to_cpu(dp->vcn));
+			dp->page_lcns[j + i] = lrh->page_lcns[i];
+		}
+
+		goto next_log_record_analyze;
+
+	case DeleteDirtyClusters: {
+		u32 range_count =
+			le16_to_cpu(lrh->redo_len) / sizeof(struct LCN_RANGE);
+		const struct LCN_RANGE *r =
+			Add2Ptr(lrh, le16_to_cpu(lrh->redo_off));
+
+		/* Loop through all of the Lcn ranges this log record */
+		for (i = 0; i < range_count; i++, r++) {
+			u64 lcn0 = le64_to_cpu(r->lcn);
+			u64 lcn_e = lcn0 + le64_to_cpu(r->len) - 1;
+
+			dp = NULL;
+			while ((dp = enum_rstbl(dptbl, dp))) {
+				u32 j;
+
+				t32 = le32_to_cpu(dp->lcns_follow);
+				for (j = 0; j < t32; j++) {
+					t64 = le64_to_cpu(dp->page_lcns[j]);
+					if (t64 >= lcn0 && t64 <= lcn_e)
+						dp->page_lcns[j] = 0;
+				}
+			}
+		}
+		goto next_log_record_analyze;
+		;
+	}
+
+	case OpenNonresidentAttribute:
+		t16 = le16_to_cpu(lrh->target_attr);
+		if (t16 >= bytes_per_rt(oatbl)) {
+			/*
+			 * Compute how big the table needs to be.
+			 * Add 10 extra entries for some cushion
+			 */
+			u32 new_e = t16 / le16_to_cpu(oatbl->size);
+
+			new_e += 10 - le16_to_cpu(oatbl->used);
+
+			oatbl = extend_rsttbl(oatbl, new_e, ~0u);
+			log->open_attr_tbl = oatbl;
+			if (!oatbl) {
+				err = -ENOMEM;
+				goto out;
+			}
+		}
+
+		/* Point to the entry being opened */
+		oe = alloc_rsttbl_from_idx(&oatbl, t16);
+		log->open_attr_tbl = oatbl;
+		if (!oe) {
+			err = -ENOMEM;
+			goto out;
+		}
+
+		/* Initialize this entry from the log record */
+		t16 = le16_to_cpu(lrh->redo_off);
+		if (!rst->major_ver) {
+			/* Convert version '0' into version '1' */
+			struct OPEN_ATTR_ENRTY_32 *oe0 = Add2Ptr(lrh, t16);
+
+			oe->bytes_per_index = oe0->bytes_per_index;
+			oe->type = oe0->type;
+			oe->is_dirty_pages = oe0->is_dirty_pages;
+			oe->name_len = 0; //oe0.name_len;
+			oe->ref = oe0->ref;
+			oe->open_record_lsn = oe0->open_record_lsn;
+		} else {
+			memcpy(oe, Add2Ptr(lrh, t16), bytes_per_attr_entry);
+		}
+
+		t16 = le16_to_cpu(lrh->undo_len);
+		if (t16) {
+			oe->ptr = ntfs_alloc(t16, 0);
+			if (!oe->ptr) {
+				err = -ENOMEM;
+				goto out;
+			}
+			oe->name_len = t16 / sizeof(short);
+			memcpy(oe->ptr,
+			       Add2Ptr(lrh, le16_to_cpu(lrh->undo_off)), t16);
+			oe->is_attr_name = 1;
+		} else {
+			oe->ptr = NULL;
+			oe->is_attr_name = 0;
+		}
+
+		goto next_log_record_analyze;
+
+	case HotFix:
+		t16 = le16_to_cpu(lrh->target_attr);
+		t64 = le64_to_cpu(lrh->target_vcn);
+		dp = find_dp(dptbl, t16, t64);
+		if (dp) {
+			size_t j = le64_to_cpu(lrh->target_vcn) -
+				   le64_to_cpu(dp->vcn);
+			if (dp->page_lcns[j])
+				dp->page_lcns[j] = lrh->page_lcns[0];
+		}
+		goto next_log_record_analyze;
+
+	case EndTopLevelAction:
+		tr = Add2Ptr(trtbl, transact_id);
+		tr->prev_lsn = cpu_to_le64(rec_lsn);
+		tr->undo_next_lsn = frh->client_undo_next_lsn;
+		goto next_log_record_analyze;
+
+	case PrepareTransaction:
+		tr = Add2Ptr(trtbl, transact_id);
+		tr->transact_state = TransactionPrepared;
+		goto next_log_record_analyze;
+
+	case CommitTransaction:
+		tr = Add2Ptr(trtbl, transact_id);
+		tr->transact_state = TransactionCommitted;
+		goto next_log_record_analyze;
+
+	case ForgetTransaction:
+		free_rsttbl_idx(trtbl, transact_id);
+		goto next_log_record_analyze;
+
+	case Noop:
+	case OpenAttributeTableDump:
+	case AttributeNamesDump:
+	case DirtyPageTableDump:
+	case TransactionTableDump:
+		/* The following cases require no action the Analysis Pass */
+		goto next_log_record_analyze;
+
+	default:
+		/*
+		 * All codes will be explicitly handled.
+		 * If we see a code we do not expect, then we are trouble
+		 */
+		goto next_log_record_analyze;
+	}
+
+end_log_records_enumerate:
+	lcb_put(lcb);
+	lcb = NULL;
+
+	/*
+	 * Scan the Dirty Page Table and Transaction Table for
+	 * the lowest lsn, and return it as the Redo lsn
+	 */
+	dp = NULL;
+	while ((dp = enum_rstbl(dptbl, dp))) {
+		t64 = le64_to_cpu(dp->oldest_lsn);
+		if (t64 && t64 < rlsn)
+			rlsn = t64;
+	}
+
+	tr = NULL;
+	while ((tr = enum_rstbl(trtbl, tr))) {
+		t64 = le64_to_cpu(tr->first_lsn);
+		if (t64 && t64 < rlsn)
+			rlsn = t64;
+	}
+
+	/* Only proceed if the Dirty Page Table or Transaction table are not empty */
+	if ((!dptbl || !dptbl->total) && (!trtbl || !trtbl->total))
+		goto end_reply;
+
+	sbi->flags |= NTFS_FLAGS_NEED_REPLAY;
+	if (is_ro)
+		goto out;
+
+	/* Reopen all of the attributes with dirty pages */
+	oe = NULL;
+next_open_attribute:
+
+	oe = enum_rstbl(oatbl, oe);
+	if (!oe) {
+		err = 0;
+		dp = NULL;
+		goto next_dirty_page;
+	}
+
+	oa = ntfs_alloc(sizeof(struct OpenAttr), 1);
+	if (!oa) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	inode = ntfs_iget5(sbi->sb, &oe->ref, NULL);
+	if (IS_ERR(inode))
+		goto fake_attr;
+
+	if (is_bad_inode(inode)) {
+		iput(inode);
+fake_attr:
+		if (oa->ni) {
+			iput(&oa->ni->vfs_inode);
+			oa->ni = NULL;
+		}
+
+		attr = attr_create_nonres_log(sbi, oe->type, 0, oe->ptr,
+					      oe->name_len, 0);
+		if (!attr) {
+			ntfs_free(oa);
+			err = -ENOMEM;
+			goto out;
+		}
+		oa->attr = attr;
+		oa->run1 = &oa->run0;
+		goto final_oe;
+	}
+
+	ni_oe = ntfs_i(inode);
+	oa->ni = ni_oe;
+
+	attr = ni_find_attr(ni_oe, NULL, NULL, oe->type, oe->ptr, oe->name_len,
+			    NULL, NULL);
+
+	if (!attr)
+		goto fake_attr;
+
+	t32 = le32_to_cpu(attr->size);
+	oa->attr = ntfs_memdup(attr, t32);
+	if (!oa->attr)
+		goto fake_attr;
+
+	if (!S_ISDIR(inode->i_mode)) {
+		if (attr->type == ATTR_DATA && !attr->name_len) {
+			oa->run1 = &ni_oe->file.run;
+			goto final_oe;
+		}
+	} else {
+		if (attr->type == ATTR_ALLOC &&
+		    attr->name_len == ARRAY_SIZE(I30_NAME) &&
+		    !memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME))) {
+			oa->run1 = &ni_oe->dir.alloc_run;
+			goto final_oe;
+		}
+	}
+
+	if (attr->non_res) {
+		u16 roff = le16_to_cpu(attr->nres.run_off);
+		CLST svcn = le64_to_cpu(attr->nres.svcn);
+
+		err = run_unpack(&oa->run0, sbi, inode->i_ino, svcn,
+				 le64_to_cpu(attr->nres.evcn), svcn,
+				 Add2Ptr(attr, roff), t32 - roff);
+		if (err < 0) {
+			ntfs_free(oa->attr);
+			oa->attr = NULL;
+			goto fake_attr;
+		}
+		err = 0;
+	}
+	oa->run1 = &oa->run0;
+	attr = oa->attr;
+
+final_oe:
+	if (oe->is_attr_name == 1)
+		ntfs_free(oe->ptr);
+	oe->is_attr_name = 0;
+	oe->ptr = oa;
+	oe->name_len = attr->name_len;
+
+	goto next_open_attribute;
+
+	/*
+	 * Now loop through the dirty page table to extract all of the Vcn/Lcn
+	 * Mapping that we have, and insert it into the appropriate run
+	 */
+next_dirty_page:
+	dp = enum_rstbl(dptbl, dp);
+	if (!dp)
+		goto do_redo_1;
+
+	oe = Add2Ptr(oatbl, le32_to_cpu(dp->target_attr));
+
+	if (oe->next != RESTART_ENTRY_ALLOCATED_LE)
+		goto next_dirty_page;
+
+	oa = oe->ptr;
+	if (!oa)
+		goto next_dirty_page;
+
+	i = -1;
+next_dirty_page_vcn:
+	i += 1;
+	if (i >= le32_to_cpu(dp->lcns_follow))
+		goto next_dirty_page;
+
+	vcn = le64_to_cpu(dp->vcn) + i;
+	size = (vcn + 1) << sbi->cluster_bits;
+
+	if (!dp->page_lcns[i])
+		goto next_dirty_page_vcn;
+
+	rno = ino_get(&oe->ref);
+	if (rno <= MFT_REC_MIRR &&
+	    size < (MFT_REC_VOL + 1) * sbi->record_size &&
+	    oe->type == ATTR_DATA) {
+		goto next_dirty_page_vcn;
+	}
+
+	lcn = le64_to_cpu(dp->page_lcns[i]);
+
+	if ((!run_lookup_entry(oa->run1, vcn, &lcn0, &len0, NULL) ||
+	     lcn0 != lcn) &&
+	    !run_add_entry(oa->run1, vcn, lcn, 1)) {
+		err = -ENOMEM;
+		goto out;
+	}
+	attr = oa->attr;
+	t64 = le64_to_cpu(attr->nres.alloc_size);
+	if (size > t64) {
+		attr->nres.valid_size = attr->nres.data_size =
+			attr->nres.alloc_size = cpu_to_le64(size);
+	}
+	goto next_dirty_page_vcn;
+
+do_redo_1:
+	/*
+	 * Perform the Redo Pass, to restore all of the dirty pages to the same
+	 * contents that they had immediately before the crash
+	 * If the dirty page table is empty, then we can skip the entire Redo Pass
+	 */
+	if (!dptbl || !dptbl->total)
+		goto do_undo_action;
+
+	rec_lsn = rlsn;
+
+	/*
+	 * Read the record at the Redo lsn, before falling
+	 * into common code to handle each record
+	 */
+	err = read_log_rec_lcb(log, rlsn, lcb_ctx_next, &lcb);
+	if (err)
+		goto out;
+
+	/*
+	 * Now loop to read all of our log records forwards,
+	 * until we hit the end of the file, cleaning up at the end
+	 */
+do_action_next:
+	frh = lcb->lrh;
+
+	if (LfsClientRecord != frh->record_type)
+		goto read_next_log_do_action;
+
+	transact_id = le32_to_cpu(frh->transact_id);
+	rec_len = le32_to_cpu(frh->client_data_len);
+	lrh = lcb->log_rec;
+
+	if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	/* Ignore log records that do not update pages */
+	if (lrh->lcns_follow)
+		goto find_dirty_page;
+
+	goto read_next_log_do_action;
+
+find_dirty_page:
+	t16 = le16_to_cpu(lrh->target_attr);
+	t64 = le64_to_cpu(lrh->target_vcn);
+	dp = find_dp(dptbl, t16, t64);
+
+	if (!dp)
+		goto read_next_log_do_action;
+
+	if (rec_lsn < le64_to_cpu(dp->oldest_lsn))
+		goto read_next_log_do_action;
+
+	t16 = le16_to_cpu(lrh->target_attr);
+	if (t16 >= bytes_per_rt(oatbl)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	oe = Add2Ptr(oatbl, t16);
+
+	if (oe->next != RESTART_ENTRY_ALLOCATED_LE) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	oa = oe->ptr;
+
+	if (!oa) {
+		err = -EINVAL;
+		goto out;
+	}
+	attr = oa->attr;
+
+	vcn = le64_to_cpu(lrh->target_vcn);
+
+	if (!run_lookup_entry(oa->run1, vcn, &lcn, NULL, NULL) ||
+	    lcn == SPARSE_LCN) {
+		goto read_next_log_do_action;
+	}
+
+	/* Point to the Redo data and get its length */
+	data = Add2Ptr(lrh, le16_to_cpu(lrh->redo_off));
+	dlen = le16_to_cpu(lrh->redo_len);
+
+	/* Shorten length by any Lcns which were deleted */
+	saved_len = dlen;
+
+	for (i = le16_to_cpu(lrh->lcns_follow); i; i--) {
+		size_t j;
+		u32 alen, voff;
+
+		voff = le16_to_cpu(lrh->record_off) +
+		       le16_to_cpu(lrh->attr_off);
+		voff += le16_to_cpu(lrh->cluster_off) << SECTOR_SHIFT;
+
+		/* If the Vcn question is allocated, we can just get out.*/
+		j = le64_to_cpu(lrh->target_vcn) - le64_to_cpu(dp->vcn);
+		if (dp->page_lcns[j + i - 1])
+			break;
+
+		if (!saved_len)
+			saved_len = 1;
+
+		/*
+		 * Calculate the allocated space left relative to the
+		 * log record Vcn, after removing this unallocated Vcn
+		 */
+		alen = (i - 1) << sbi->cluster_bits;
+
+		/*
+		 * If the update described this log record goes beyond
+		 * the allocated space, then we will have to reduce the length
+		 */
+		if (voff >= alen)
+			dlen = 0;
+		else if (voff + dlen > alen)
+			dlen = alen - voff;
+	}
+
+	/* If the resulting dlen from above is now zero, we can skip this log record */
+	if (!dlen && saved_len)
+		goto read_next_log_do_action;
+
+	t16 = le16_to_cpu(lrh->redo_op);
+	if (can_skip_action(t16))
+		goto read_next_log_do_action;
+
+	/* Apply the Redo operation a common routine */
+	err = do_action(log, oe, lrh, t16, data, dlen, rec_len, &rec_lsn);
+	if (err)
+		goto out;
+
+	/* Keep reading and looping back until end of file */
+read_next_log_do_action:
+	err = read_next_log_rec(log, lcb, &rec_lsn);
+	if (!err && rec_lsn)
+		goto do_action_next;
+
+	lcb_put(lcb);
+	lcb = NULL;
+
+do_undo_action:
+	/* Scan Transaction Table */
+	tr = NULL;
+transaction_table_next:
+	tr = enum_rstbl(trtbl, tr);
+	if (!tr)
+		goto undo_action_done;
+
+	if (TransactionActive != tr->transact_state || !tr->undo_next_lsn) {
+		free_rsttbl_idx(trtbl, PtrOffset(trtbl, tr));
+		goto transaction_table_next;
+	}
+
+	log->transaction_id = PtrOffset(trtbl, tr);
+	undo_next_lsn = le64_to_cpu(tr->undo_next_lsn);
+
+	/*
+	 * We only have to do anything if the transaction has
+	 * something its undo_next_lsn field
+	 */
+	if (!undo_next_lsn)
+		goto commit_undo;
+
+	/* Read the first record to be undone by this transaction */
+	err = read_log_rec_lcb(log, undo_next_lsn, lcb_ctx_undo_next, &lcb);
+	if (err)
+		goto out;
+
+	/*
+	 * Now loop to read all of our log records forwards,
+	 * until we hit the end of the file, cleaning up at the end
+	 */
+undo_action_next:
+
+	lrh = lcb->log_rec;
+	frh = lcb->lrh;
+	transact_id = le32_to_cpu(frh->transact_id);
+	rec_len = le32_to_cpu(frh->client_data_len);
+
+	if (!check_log_rec(lrh, rec_len, transact_id, bytes_per_attr_entry)) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	if (lrh->undo_op == cpu_to_le16(Noop))
+		goto read_next_log_undo_action;
+
+	oe = Add2Ptr(oatbl, le16_to_cpu(lrh->target_attr));
+	oa = oe->ptr;
+
+	t16 = le16_to_cpu(lrh->lcns_follow);
+	if (!t16)
+		goto add_allocated_vcns;
+
+	is_mapped = run_lookup_entry(oa->run1, le64_to_cpu(lrh->target_vcn),
+				     &lcn, &clen, NULL);
+
+	/*
+	 * If the mapping isn't already the table or the  mapping
+	 * corresponds to a hole the mapping, we need to make sure
+	 * there is no partial page already memory
+	 */
+	if (is_mapped && lcn != SPARSE_LCN && clen >= t16)
+		goto add_allocated_vcns;
+
+	vcn = le64_to_cpu(lrh->target_vcn);
+	vcn &= ~(log->clst_per_page - 1);
+
+add_allocated_vcns:
+	for (i = 0, vcn = le64_to_cpu(lrh->target_vcn),
+	    size = (vcn + 1) << sbi->cluster_bits;
+	     i < t16; i++, vcn += 1, size += sbi->cluster_size) {
+		attr = oa->attr;
+		if (!attr->non_res) {
+			if (size > le32_to_cpu(attr->res.data_size))
+				attr->res.data_size = cpu_to_le32(size);
+		} else {
+			if (size > le64_to_cpu(attr->nres.data_size))
+				attr->nres.valid_size = attr->nres.data_size =
+					attr->nres.alloc_size =
+						cpu_to_le64(size);
+		}
+	}
+
+	t16 = le16_to_cpu(lrh->undo_op);
+	if (can_skip_action(t16))
+		goto read_next_log_undo_action;
+
+	/* Point to the Redo data and get its length */
+	data = Add2Ptr(lrh, le16_to_cpu(lrh->undo_off));
+	dlen = le16_to_cpu(lrh->undo_len);
+
+	/* it is time to apply the undo action */
+	err = do_action(log, oe, lrh, t16, data, dlen, rec_len, NULL);
+
+read_next_log_undo_action:
+	/*
+	 * Keep reading and looping back until we have read the
+	 * last record for this transaction
+	 */
+	err = read_next_log_rec(log, lcb, &rec_lsn);
+	if (err)
+		goto out;
+
+	if (rec_lsn)
+		goto undo_action_next;
+
+commit_undo:
+	free_rsttbl_idx(trtbl, log->transaction_id);
+
+	log->transaction_id = 0;
+
+	goto transaction_table_next;
+
+undo_action_done:
+
+	ntfs_update_mftmirr(sbi, 0);
+
+	sbi->flags &= ~NTFS_FLAGS_NEED_REPLAY;
+
+end_reply:
+
+	err = 0;
+	if (is_ro)
+		goto out;
+
+	rh = ntfs_alloc(log->page_size, 1);
+	if (!rh) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	rh->rhdr.sign = NTFS_RSTR_SIGNATURE;
+	rh->rhdr.fix_off = cpu_to_le16(offsetof(struct RESTART_HDR, fixups));
+	t16 = (log->page_size >> SECTOR_SHIFT) + 1;
+	rh->rhdr.fix_num = cpu_to_le16(t16);
+	rh->sys_page_size = cpu_to_le32(log->page_size);
+	rh->page_size = cpu_to_le32(log->page_size);
+
+	t16 = QuadAlign(offsetof(struct RESTART_HDR, fixups) +
+			sizeof(short) * t16);
+	rh->ra_off = cpu_to_le16(t16);
+	rh->minor_ver = cpu_to_le16(1); // 0x1A:
+	rh->major_ver = cpu_to_le16(1); // 0x1C:
+
+	ra2 = Add2Ptr(rh, t16);
+	memcpy(ra2, ra, sizeof(struct RESTART_AREA));
+
+	ra2->client_idx[0] = 0;
+	ra2->client_idx[1] = LFS_NO_CLIENT_LE;
+	ra2->flags = cpu_to_le16(2);
+
+	le32_add_cpu(&ra2->open_log_count, 1);
+
+	ntfs_fix_pre_write(&rh->rhdr, log->page_size);
+
+	err = ntfs_sb_write_run(sbi, &ni->file.run, 0, rh, log->page_size);
+	if (!err)
+		err = ntfs_sb_write_run(sbi, &log->ni->file.run, log->page_size,
+					rh, log->page_size);
+
+	ntfs_free(rh);
+	if (err)
+		goto out;
+
+out:
+	ntfs_free(rst);
+	if (lcb)
+		lcb_put(lcb);
+
+	/* Scan the Open Attribute Table to close all of the open attributes */
+	oe = NULL;
+	while ((oe = enum_rstbl(oatbl, oe))) {
+		rno = ino_get(&oe->ref);
+
+		if (oe->is_attr_name == 1) {
+			ntfs_free(oe->ptr);
+			oe->ptr = NULL;
+			continue;
+		}
+
+		if (oe->is_attr_name)
+			continue;
+
+		oa = oe->ptr;
+		if (!oa)
+			continue;
+
+		run_close(&oa->run0);
+		ntfs_free(oa->attr);
+		if (oa->ni)
+			iput(&oa->ni->vfs_inode);
+		ntfs_free(oa);
+	}
+
+	ntfs_free(trtbl);
+	ntfs_free(oatbl);
+	ntfs_free(dptbl);
+	ntfs_free(attr_names);
+	ntfs_free(rst_info.r_page);
+
+	ntfs_free(ra);
+	ntfs_free(log->one_page_buf);
+
+	if (err)
+		sbi->flags |= NTFS_FLAGS_NEED_REPLAY;
+
+	if (err == -EROFS)
+		err = 0;
+	else if (log->set_dirty)
+		ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
+
+	ntfs_free(log);
+
+	return err;
+}
-- 
2.25.4


^ permalink raw reply related	[flat|nested] 21+ messages in thread

* [PATCH v11 08/10] fs/ntfs3: Add Kconfig, Makefile and doc
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
                   ` (6 preceding siblings ...)
  2020-10-30 15:02 ` [PATCH v11 07/10] fs/ntfs3: Add NTFS journal Konstantin Komarov
@ 2020-10-30 15:02 ` Konstantin Komarov
  2020-10-30 15:02 ` [PATCH v11 09/10] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile Konstantin Komarov
                   ` (4 subsequent siblings)
  12 siblings, 0 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This adds Kconfig, Makefile and doc

Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
---
 Documentation/filesystems/ntfs3.rst | 112 ++++++++++++++++++++++++++++
 fs/ntfs3/Kconfig                    |  23 ++++++
 fs/ntfs3/Makefile                   |  11 +++
 3 files changed, 146 insertions(+)
 create mode 100644 Documentation/filesystems/ntfs3.rst
 create mode 100644 fs/ntfs3/Kconfig
 create mode 100644 fs/ntfs3/Makefile

diff --git a/Documentation/filesystems/ntfs3.rst b/Documentation/filesystems/ntfs3.rst
new file mode 100644
index 000000000000..90b2125d1af2
--- /dev/null
+++ b/Documentation/filesystems/ntfs3.rst
@@ -0,0 +1,112 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+=====
+NTFS3
+=====
+
+
+Summary and Features
+====================
+
+NTFS3 is fully functional NTFS Read-Write driver. The driver works with
+NTFS versions up to 3.1, normal/compressed/sparse files
+and journal replaying. File system type to use on mount is 'ntfs3'.
+
+- This driver implements NTFS read/write support for normal, sparse and
+  compressed files.
+- Supports native journal replaying;
+- Supports extended attributes
+	Predefined extended attributes:
+	- 'system.ntfs_security' gets/sets security
+			descriptor (SECURITY_DESCRIPTOR_RELATIVE)
+	- 'system.ntfs_attrib' gets/sets ntfs file/dir attributes.
+		Note: applied to empty files, this allows to switch type between
+		sparse(0x200), compressed(0x800) and normal;
+- Supports NFS export of mounted NTFS volumes.
+
+Mount Options
+=============
+
+The list below describes mount options supported by NTFS3 driver in addition to
+generic ones.
+
+===============================================================================
+
+nls=name		This option informs the driver how to interpret path
+			strings and translate them to Unicode and back. If
+			this option is not set, the default codepage will be
+			used (CONFIG_NLS_DEFAULT).
+			Examples:
+				'nls=utf8'
+
+nls_alt=name		This option extends "nls". It will be used to translate
+			path string to Unicode if primary nls failed.
+			Examples:
+				'nls_alt=cp1251'
+
+uid=
+gid=
+umask=			Controls the default permissions for files/directories created
+			after the NTFS volume is mounted.
+
+fmask=
+dmask=			Instead of specifying umask which applies both to
+			files and directories, fmask applies only to files and
+			dmask only to directories.
+
+nohidden		Files with the Windows-specific HIDDEN (FILE_ATTRIBUTE_HIDDEN)
+			attribute will not be shown under Linux.
+
+sys_immutable		Files with the Windows-specific SYSTEM
+			(FILE_ATTRIBUTE_SYSTEM) attribute will be marked as system
+			immutable files.
+
+discard			Enable support of the TRIM command for improved performance
+			on delete operations, which is recommended for use with the
+			solid-state drives (SSD).
+
+force			Forces the driver to mount partitions even if 'dirty' flag
+			(volume dirty) is set. Not recommended for use.
+
+sparse			Create new files as "sparse".
+
+showmeta		Use this parameter to show all meta-files (System Files) on
+			a mounted NTFS partition.
+			By default, all meta-files are hidden.
+
+prealloc		Preallocate space for files excessively when file size is
+			increasing on writes. Decreases fragmentation in case of
+			parallel write operations to different files.
+
+no_acs_rules		"No access rules" mount option sets access rights for
+			files/folders to 777 and owner/group to root. This mount
+			option absorbs all other permissions:
+			- permissions change for files/folders will be reported
+				as successful, but they will remain 777;
+			- owner/group change will be reported as successful, but
+				they will stay as root
+
+acl			Support POSIX ACLs (Access Control Lists). Effective if
+			supported by Kernel. Not to be confused with NTFS ACLs.
+			The option specified as acl enables support for POSIX ACLs.
+
+noatime			All files and directories will not update their last access
+			time attribute if a partition is mounted with this parameter.
+			This option can speed up file system operation.
+
+===============================================================================
+
+ToDo list
+=========
+
+- Full journaling support (currently journal replaying is supported) over JBD.
+
+
+References
+==========
+https://www.paragon-software.com/home/ntfs-linux-professional/
+	- Commercial version of the NTFS driver for Linux.
+
+almaz.alexandrovich@paragon-software.com
+	- Direct e-mail address for feedback and requests on the NTFS3 implementation.
+
diff --git a/fs/ntfs3/Kconfig b/fs/ntfs3/Kconfig
new file mode 100644
index 000000000000..92a9c68008c8
--- /dev/null
+++ b/fs/ntfs3/Kconfig
@@ -0,0 +1,23 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config NTFS3_FS
+	tristate "NTFS Read-Write file system support"
+	select NLS
+	help
+	  Windows OS native file system (NTFS) support up to NTFS version 3.1.
+
+	  Y or M enables the NTFS3 driver with full features enabled (read,
+	  write, journal replaying, sparse/compressed files support).
+	  File system type to use on mount is "ntfs3". Module name (M option)
+	  is also "ntfs3".
+
+	  Documentation: <file:Documentation/filesystems/ntfs3.rst>
+
+config NTFS3_64BIT_CLUSTER
+	bool "64 bits per NTFS clusters"
+	depends on NTFS3_FS && 64BIT
+	help
+	  Windows implementation of ntfs.sys uses 32 bits per clusters.
+	  If activated 64 bits per clusters you will be able to use 4k cluster
+	  for 16T+ volumes. Windows will not be able to mount such volumes.
+
+	  It is recommended to say N here.
diff --git a/fs/ntfs3/Makefile b/fs/ntfs3/Makefile
new file mode 100644
index 000000000000..d99dd1af43aa
--- /dev/null
+++ b/fs/ntfs3/Makefile
@@ -0,0 +1,11 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# Makefile for the ntfs3 filesystem support.
+#
+
+obj-$(CONFIG_NTFS3_FS) += ntfs3.o
+
+ntfs3-y := bitfunc.o bitmap.o inode.o fsntfs.o frecord.o \
+	    index.o attrlist.o record.o attrib.o run.o xattr.o\
+	    upcase.o super.o file.o dir.o namei.o lznt.o\
+	    fslog.o
-- 
2.25.4


^ permalink raw reply related	[flat|nested] 21+ messages in thread

* [PATCH v11 09/10] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
                   ` (7 preceding siblings ...)
  2020-10-30 15:02 ` [PATCH v11 08/10] fs/ntfs3: Add Kconfig, Makefile and doc Konstantin Komarov
@ 2020-10-30 15:02 ` Konstantin Komarov
  2020-10-31  1:23   ` kernel test robot
                     ` (2 more replies)
  2020-10-30 15:02 ` [PATCH v11 10/10] fs/ntfs3: Add MAINTAINERS Konstantin Komarov
                   ` (3 subsequent siblings)
  12 siblings, 3 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This adds NTFS3 in fs/Kconfig and fs/Makefile

Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
---
 fs/Kconfig  | 1 +
 fs/Makefile | 1 +
 2 files changed, 2 insertions(+)

diff --git a/fs/Kconfig b/fs/Kconfig
index aa4c12282301..eae96d55ab67 100644
--- a/fs/Kconfig
+++ b/fs/Kconfig
@@ -145,6 +145,7 @@ menu "DOS/FAT/EXFAT/NT Filesystems"
 source "fs/fat/Kconfig"
 source "fs/exfat/Kconfig"
 source "fs/ntfs/Kconfig"
+source "fs/ntfs3/Kconfig"
 
 endmenu
 endif # BLOCK
diff --git a/fs/Makefile b/fs/Makefile
index 999d1a23f036..4f5242cdaee2 100644
--- a/fs/Makefile
+++ b/fs/Makefile
@@ -100,6 +100,7 @@ obj-$(CONFIG_SYSV_FS)		+= sysv/
 obj-$(CONFIG_CIFS)		+= cifs/
 obj-$(CONFIG_HPFS_FS)		+= hpfs/
 obj-$(CONFIG_NTFS_FS)		+= ntfs/
+obj-$(CONFIG_NTFS3_FS)		+= ntfs3/
 obj-$(CONFIG_UFS_FS)		+= ufs/
 obj-$(CONFIG_EFS_FS)		+= efs/
 obj-$(CONFIG_JFFS2_FS)		+= jffs2/
-- 
2.25.4


^ permalink raw reply related	[flat|nested] 21+ messages in thread

* [PATCH v11 10/10] fs/ntfs3: Add MAINTAINERS
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
                   ` (8 preceding siblings ...)
  2020-10-30 15:02 ` [PATCH v11 09/10] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile Konstantin Komarov
@ 2020-10-30 15:02 ` Konstantin Komarov
  2020-10-30 15:24 ` [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Pali Rohár
                   ` (2 subsequent siblings)
  12 siblings, 0 replies; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:02 UTC (permalink / raw)
  To: linux-fsdevel
  Cc: viro, linux-kernel, pali, dsterba, aaptel, willy, rdunlap, joe,
	mark, nborisov, linux-ntfs-dev, anton, Konstantin Komarov

This adds MAINTAINERS

Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
---
 MAINTAINERS | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/MAINTAINERS b/MAINTAINERS
index e73636b75f29..55922e8a4557 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -12494,6 +12494,13 @@ T:	git git://git.kernel.org/pub/scm/linux/kernel/git/aia21/ntfs.git
 F:	Documentation/filesystems/ntfs.rst
 F:	fs/ntfs/
 
+NTFS3 FILESYSTEM
+M:	Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
+S:	Supported
+W:	http://www.paragon-software.com/
+F:	Documentation/filesystems/ntfs3.rst
+F:	fs/ntfs3/
+
 NUBUS SUBSYSTEM
 M:	Finn Thain <fthain@telegraphics.com.au>
 L:	linux-m68k@lists.linux-m68k.org
-- 
2.25.4


^ permalink raw reply related	[flat|nested] 21+ messages in thread

* Re: [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
                   ` (9 preceding siblings ...)
  2020-10-30 15:02 ` [PATCH v11 10/10] fs/ntfs3: Add MAINTAINERS Konstantin Komarov
@ 2020-10-30 15:24 ` Pali Rohár
  2020-10-30 15:51   ` Konstantin Komarov
  2020-10-30 16:41 ` Pali Rohár
  2020-10-31  2:42 ` Eric Biggers
  12 siblings, 1 reply; 21+ messages in thread
From: Pali Rohár @ 2020-10-30 15:24 UTC (permalink / raw)
  To: Konstantin Komarov
  Cc: linux-fsdevel, viro, linux-kernel, dsterba, aaptel, willy,
	rdunlap, joe, mark, nborisov, linux-ntfs-dev, anton

Hello and thanks for update!

I have just two comments for the last v11 version.

I really do not like nls_alt mount option and I do not think we should
merge this mount option into ntfs kernel driver. Details I described in:
https://lore.kernel.org/linux-fsdevel/20201009154734.andv4es3azkkskm5@pali/

tl;dr it is not systematic solution and is incompatible with existing
in-kernel ntfs driver, also incompatible with in-kernel vfat, udf and
ext4 (with UNICODE support) drivers. In my opinion, all kernel fs
drivers which deals with UNICODE should handle it in similar way.

It would be really bad if userspace application need to behave
differently for this new ntfs driver and differently for all other
UNICODE drivers.

Second comment is simplification of usage nls_load() with UTF-8 parameter
which I described in older email:
https://lore.kernel.org/linux-fsdevel/948ac894450d494ea15496c2e5b8c906@paragon-software.com/

You wrote that you have applied it, but seems it was lost (maybe during
rebase?) as it is not present in the last v11 version.

I suggested to not use nls_load() with UTF-8 at all. Your version of
ntfs driver does not use kernel's nls utf8 module for UTF-8 support, so
trying to load it should be avoided. Also kernel can be compiled without
utf8 nls module (which is moreover broken) and with my above suggestion,
ntfs driver would work correctly. Without that suggestion, mounting
would fail.

^ permalink raw reply	[flat|nested] 21+ messages in thread

* RE: [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software
  2020-10-30 15:24 ` [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Pali Rohár
@ 2020-10-30 15:51   ` Konstantin Komarov
  2020-10-30 16:24     ` Pali Rohár
  0 siblings, 1 reply; 21+ messages in thread
From: Konstantin Komarov @ 2020-10-30 15:51 UTC (permalink / raw)
  To: Pali Rohár
  Cc: linux-fsdevel, viro, linux-kernel, dsterba, aaptel, willy,
	rdunlap, joe, mark, nborisov, linux-ntfs-dev, anton

From: Pali Rohár <pali@kernel.org>
Sent: Friday, October 30, 2020 6:25 PM
> To: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
> Cc: linux-fsdevel@vger.kernel.org; viro@zeniv.linux.org.uk; linux-kernel@vger.kernel.org; dsterba@suse.cz; aaptel@suse.com;
> willy@infradead.org; rdunlap@infradead.org; joe@perches.com; mark@harmstone.com; nborisov@suse.com; linux-ntfs-
> dev@lists.sourceforge.net; anton@tuxera.com
> Subject: Re: [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software
> 
> Hello and thanks for update!
> 
> I have just two comments for the last v11 version.
> 
> I really do not like nls_alt mount option and I do not think we should
> merge this mount option into ntfs kernel driver. Details I described in:
> https://lore.kernel.org/linux-fsdevel/20201009154734.andv4es3azkkskm5@pali/
> 
> tl;dr it is not systematic solution and is incompatible with existing
> in-kernel ntfs driver, also incompatible with in-kernel vfat, udf and
> ext4 (with UNICODE support) drivers. In my opinion, all kernel fs
> drivers which deals with UNICODE should handle it in similar way.
> 

Hello Pali! First of all, apologies for not providing a feedback on your previous
message regarding the 'nls_alt'. We had internal discussions on the topic and
overall conclusion is that: we do not want to compromise Kernel standards with
our submission. So we will remove the 'nls_alt' option in the next version.

However, there are still few points we have on the topic, please read below.

> It would be really bad if userspace application need to behave
> differently for this new ntfs driver and differently for all other
> UNICODE drivers.
> 

The option does not anyhow affect userspace applications. For the "default" example
of unzip/tar:
1 - if this option is not applied (e.g. "vfat case"), trying to unzip an archive with, e.g. CP-1251,
names inside to the target fs volume, will return error, and issued file(s) won't be unzipped;
2 - if this option is applied and "nls_alt" is set, the above case will result in unzipping all the files;

Also, this issue in general only applies to "non-native" filesystems. I.e. ext4 is not affected by it
in any case, as it just stores the name as bytes, no matter what those bytes are. The above case
won't give an unzip error on ext4. The only symptom of this would be, maybe, "incorrect encoding"
marking within the listing of such files (in File Manager or Terminal, e.g. in Ubuntu), but there won't
be an unzip process termination with incomplete unarchived fileset, unlike it is for vfat/exfat/ntfs
without "nls_alt".

> Second comment is simplification of usage nls_load() with UTF-8 parameter
> which I described in older email:
> https://lore.kernel.org/linux-fsdevel/948ac894450d494ea15496c2e5b8c906@paragon-software.com/
> 
> You wrote that you have applied it, but seems it was lost (maybe during
> rebase?) as it is not present in the last v11 version.
> 
> I suggested to not use nls_load() with UTF-8 at all. Your version of
> ntfs driver does not use kernel's nls utf8 module for UTF-8 support, so
> trying to load it should be avoided. Also kernel can be compiled without
> utf8 nls module (which is moreover broken) and with my above suggestion,
> ntfs driver would work correctly. Without that suggestion, mounting
> would fail.

Thanks for pointing that out. It is likely the "nls_load()" fixes were lost during rebase.
Will recheck it and return them to the v12.

Best regards!

^ permalink raw reply	[flat|nested] 21+ messages in thread

* Re: [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software
  2020-10-30 15:51   ` Konstantin Komarov
@ 2020-10-30 16:24     ` Pali Rohár
  0 siblings, 0 replies; 21+ messages in thread
From: Pali Rohár @ 2020-10-30 16:24 UTC (permalink / raw)
  To: Konstantin Komarov
  Cc: linux-fsdevel, viro, linux-kernel, dsterba, aaptel, willy,
	rdunlap, joe, mark, nborisov, linux-ntfs-dev, anton

Hello!

On Friday 30 October 2020 15:51:10 Konstantin Komarov wrote:
> From: Pali Rohár <pali@kernel.org>
> Sent: Friday, October 30, 2020 6:25 PM
> > To: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
> > Cc: linux-fsdevel@vger.kernel.org; viro@zeniv.linux.org.uk; linux-kernel@vger.kernel.org; dsterba@suse.cz; aaptel@suse.com;
> > willy@infradead.org; rdunlap@infradead.org; joe@perches.com; mark@harmstone.com; nborisov@suse.com; linux-ntfs-
> > dev@lists.sourceforge.net; anton@tuxera.com
> > Subject: Re: [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software
> > 
> > Hello and thanks for update!
> > 
> > I have just two comments for the last v11 version.
> > 
> > I really do not like nls_alt mount option and I do not think we should
> > merge this mount option into ntfs kernel driver. Details I described in:
> > https://lore.kernel.org/linux-fsdevel/20201009154734.andv4es3azkkskm5@pali/
> > 
> > tl;dr it is not systematic solution and is incompatible with existing
> > in-kernel ntfs driver, also incompatible with in-kernel vfat, udf and
> > ext4 (with UNICODE support) drivers. In my opinion, all kernel fs
> > drivers which deals with UNICODE should handle it in similar way.
> > 
> 
> Hello Pali! First of all, apologies for not providing a feedback on your previous
> message regarding the 'nls_alt'. We had internal discussions on the topic and
> overall conclusion is that: we do not want to compromise Kernel standards with
> our submission. So we will remove the 'nls_alt' option in the next version.
> 
> However, there are still few points we have on the topic, please read below.
> 
> > It would be really bad if userspace application need to behave
> > differently for this new ntfs driver and differently for all other
> > UNICODE drivers.
> > 
> 
> The option does not anyhow affect userspace applications. For the "default" example
> of unzip/tar:
> 1 - if this option is not applied (e.g. "vfat case"), trying to unzip an archive with, e.g. CP-1251,
> names inside to the target fs volume, will return error, and issued file(s) won't be unzipped;
> 2 - if this option is applied and "nls_alt" is set, the above case will result in unzipping all the files;

I understand what is the point and I'm not against discussion how to fix
it. But it should be implemented for all filesystems with UNICODE
semantic, so behavior would be same.

For user application point of view, behavior of vfat, ntfs, udf and ext4
(with UNICODE support; see below) in handling file names should be very
similar (or exactly same if fs tech details allows it).

> Also, this issue in general only applies to "non-native" filesystems. I.e. ext4 is not affected by it
> in any case, as it just stores the name as bytes, no matter what those bytes are. The above case
> won't give an unzip error on ext4. The only symptom of this would be, maybe, "incorrect encoding"
> marking within the listing of such files (in File Manager or Terminal, e.g. in Ubuntu), but there won't
> be an unzip process termination with incomplete unarchived fileset, unlike it is for vfat/exfat/ntfs
> without "nls_alt".

When using ext4 in default mode then it really does not apply here. But
I wrote that it applies for ext4 with UNICODE support. This mode needs
to be first enabled for directory, it is relatively new feature and I do
not know if there are users of it and how many people tried different
crazy test scenarios with normalization, etc...

> > Second comment is simplification of usage nls_load() with UTF-8 parameter
> > which I described in older email:
> > https://lore.kernel.org/linux-fsdevel/948ac894450d494ea15496c2e5b8c906@paragon-software.com/
> > 
> > You wrote that you have applied it, but seems it was lost (maybe during
> > rebase?) as it is not present in the last v11 version.
> > 
> > I suggested to not use nls_load() with UTF-8 at all. Your version of
> > ntfs driver does not use kernel's nls utf8 module for UTF-8 support, so
> > trying to load it should be avoided. Also kernel can be compiled without
> > utf8 nls module (which is moreover broken) and with my above suggestion,
> > ntfs driver would work correctly. Without that suggestion, mounting
> > would fail.
> 
> Thanks for pointing that out. It is likely the "nls_load()" fixes were lost during rebase.
> Will recheck it and return them to the v12.

OK!

^ permalink raw reply	[flat|nested] 21+ messages in thread

* Re: [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
                   ` (10 preceding siblings ...)
  2020-10-30 15:24 ` [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Pali Rohár
@ 2020-10-30 16:41 ` Pali Rohár
  2020-10-31  8:51   ` Christoph Hellwig
  2020-10-31  2:42 ` Eric Biggers
  12 siblings, 1 reply; 21+ messages in thread
From: Pali Rohár @ 2020-10-30 16:41 UTC (permalink / raw)
  To: Konstantin Komarov, Christoph Hellwig
  Cc: linux-fsdevel, viro, linux-kernel, dsterba, aaptel, willy,
	rdunlap, joe, mark, nborisov, linux-ntfs-dev, anton

Hello!

On Friday 30 October 2020 18:02:29 Konstantin Komarov wrote:
>  27 files changed, 28364 insertions(+)
>  create mode 100644 Documentation/filesystems/ntfs3.rst
>  create mode 100644 fs/ntfs3/Kconfig
>  create mode 100644 fs/ntfs3/Makefile
>  create mode 100644 fs/ntfs3/attrib.c
>  create mode 100644 fs/ntfs3/attrlist.c
>  create mode 100644 fs/ntfs3/bitfunc.c
>  create mode 100644 fs/ntfs3/bitmap.c
>  create mode 100644 fs/ntfs3/debug.h
>  create mode 100644 fs/ntfs3/dir.c
>  create mode 100644 fs/ntfs3/file.c
>  create mode 100644 fs/ntfs3/frecord.c
>  create mode 100644 fs/ntfs3/fslog.c
>  create mode 100644 fs/ntfs3/fsntfs.c
>  create mode 100644 fs/ntfs3/index.c
>  create mode 100644 fs/ntfs3/inode.c
>  create mode 100644 fs/ntfs3/lznt.c
>  create mode 100644 fs/ntfs3/namei.c
>  create mode 100644 fs/ntfs3/ntfs.h
>  create mode 100644 fs/ntfs3/ntfs_fs.h
>  create mode 100644 fs/ntfs3/record.c
>  create mode 100644 fs/ntfs3/run.c
>  create mode 100644 fs/ntfs3/super.c
>  create mode 100644 fs/ntfs3/upcase.c
>  create mode 100644 fs/ntfs3/xattr.c

I would like to open discussion about two ntfs kernel drivers. Do we
really need two drivers (one read only - current version and one
read/write - this new version)?

What other people think?

I remember that Christoph (added to loop) had in past a good argument
about old staging exfat driver (it had support also for fat32/vfat),
that it would cause problems if two filesystem drivers would provide
support for same filesystem.

^ permalink raw reply	[flat|nested] 21+ messages in thread

* Re: [PATCH v11 09/10] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile
  2020-10-30 15:02 ` [PATCH v11 09/10] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile Konstantin Komarov
@ 2020-10-31  1:23   ` kernel test robot
  2020-11-02  8:36   ` [kbuild] " Dan Carpenter
  2020-11-03  3:06   ` kernel test robot
  2 siblings, 0 replies; 21+ messages in thread
From: kernel test robot @ 2020-10-31  1:23 UTC (permalink / raw)
  To: Konstantin Komarov, linux-fsdevel
  Cc: kbuild-all, viro, linux-kernel, pali, dsterba, aaptel, willy,
	rdunlap, joe, mark

[-- Attachment #1: Type: text/plain, Size: 12070 bytes --]

Hi Konstantin,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on linus/master]
[also build test WARNING on v5.10-rc1 next-20201030]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Konstantin-Komarov/NTFS-read-write-driver-GPL-implementation-by-Paragon-Software/20201030-230756
base:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 07e0887302450a62f51dba72df6afb5fabb23d1c
config: h8300-randconfig-s031-20201030 (attached as .config)
compiler: h8300-linux-gcc (GCC) 9.3.0
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # apt-get install sparse
        # sparse version: v0.6.3-68-g49c98aa3-dirty
        # https://github.com/0day-ci/linux/commit/7c34316b6c7f9af2046f8343d3b010c37340ef1d
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Konstantin-Komarov/NTFS-read-write-driver-GPL-implementation-by-Paragon-Software/20201030-230756
        git checkout 7c34316b6c7f9af2046f8343d3b010c37340ef1d
        # save the attached .config to linux build tree
        COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' ARCH=h8300 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>


"sparse warnings: (new ones prefixed by >>)"
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast to restricted __le32
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast from restricted __le64
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast to restricted __le32
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast from restricted __le64
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast to restricted __le32
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast from restricted __le64
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast to restricted __le32
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast from restricted __le64
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast to restricted __le32
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast from restricted __le64
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast to restricted __le32
>> fs/ntfs3/frecord.c:569:33: sparse: sparse: cast from restricted __le64

vim +569 fs/ntfs3/frecord.c

cbd4257e6d85149 Konstantin Komarov 2020-10-30  532  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  533  /*
cbd4257e6d85149 Konstantin Komarov 2020-10-30  534   * random write access to sparsed or compressed file may result to
cbd4257e6d85149 Konstantin Komarov 2020-10-30  535   * not optimized packed runs.
cbd4257e6d85149 Konstantin Komarov 2020-10-30  536   * Here it is the place to optimize it
cbd4257e6d85149 Konstantin Komarov 2020-10-30  537   */
cbd4257e6d85149 Konstantin Komarov 2020-10-30  538  static int ni_repack(struct ntfs_inode *ni)
cbd4257e6d85149 Konstantin Komarov 2020-10-30  539  {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  540  	int err = 0;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  541  	struct ntfs_sb_info *sbi = ni->mi.sbi;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  542  	struct mft_inode *mi, *mi_p = NULL;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  543  	struct ATTRIB *attr = NULL, *attr_p;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  544  	struct ATTR_LIST_ENTRY *le = NULL, *le_p;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  545  	CLST alloc = 0;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  546  	u8 cluster_bits = sbi->cluster_bits;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  547  	CLST svcn, evcn = 0, svcn_p, evcn_p, next_svcn;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  548  	u32 roff, rs = sbi->record_size;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  549  	struct runs_tree run;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  550  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  551  	run_init(&run);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  552  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  553  	while ((attr = ni_enum_attr_ex(ni, attr, &le))) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  554  		if (!attr->non_res)
cbd4257e6d85149 Konstantin Komarov 2020-10-30  555  			continue;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  556  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  557  		if (ni_load_mi(ni, le, &mi)) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  558  			err = -EINVAL;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  559  			break;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  560  		}
cbd4257e6d85149 Konstantin Komarov 2020-10-30  561  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  562  		svcn = le64_to_cpu(attr->nres.svcn);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  563  		if (svcn != le64_to_cpu(le->vcn)) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  564  			err = -EINVAL;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  565  			break;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  566  		}
cbd4257e6d85149 Konstantin Komarov 2020-10-30  567  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  568  		if (!svcn) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30 @569  			alloc = le32_to_cpu(attr->nres.alloc_size) >>
cbd4257e6d85149 Konstantin Komarov 2020-10-30  570  				cluster_bits;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  571  			mi_p = NULL;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  572  		} else if (svcn != evcn + 1) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  573  			err = -EINVAL;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  574  			break;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  575  		}
cbd4257e6d85149 Konstantin Komarov 2020-10-30  576  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  577  		evcn = le64_to_cpu(attr->nres.evcn);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  578  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  579  		if (svcn > evcn + 1) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  580  			err = -EINVAL;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  581  			break;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  582  		}
cbd4257e6d85149 Konstantin Komarov 2020-10-30  583  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  584  		if (!mi_p) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  585  			/* do not try if too little free space */
cbd4257e6d85149 Konstantin Komarov 2020-10-30  586  			if (le32_to_cpu(mi->mrec->used) + 8 >= rs)
cbd4257e6d85149 Konstantin Komarov 2020-10-30  587  				continue;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  588  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  589  			/* do not try if last attribute segment */
cbd4257e6d85149 Konstantin Komarov 2020-10-30  590  			if (evcn + 1 == alloc)
cbd4257e6d85149 Konstantin Komarov 2020-10-30  591  				continue;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  592  			run_close(&run);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  593  		}
cbd4257e6d85149 Konstantin Komarov 2020-10-30  594  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  595  		roff = le16_to_cpu(attr->nres.run_off);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  596  		err = run_unpack(&run, sbi, ni->mi.rno, svcn, evcn, svcn,
cbd4257e6d85149 Konstantin Komarov 2020-10-30  597  				 Add2Ptr(attr, roff),
cbd4257e6d85149 Konstantin Komarov 2020-10-30  598  				 le32_to_cpu(attr->size) - roff);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  599  		if (err < 0)
cbd4257e6d85149 Konstantin Komarov 2020-10-30  600  			break;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  601  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  602  		if (!mi_p) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  603  			mi_p = mi;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  604  			attr_p = attr;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  605  			svcn_p = svcn;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  606  			evcn_p = evcn;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  607  			le_p = le;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  608  			continue;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  609  		}
cbd4257e6d85149 Konstantin Komarov 2020-10-30  610  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  611  		/*
cbd4257e6d85149 Konstantin Komarov 2020-10-30  612  		 * run contains data from two records: mi_p and mi
cbd4257e6d85149 Konstantin Komarov 2020-10-30  613  		 * try to pack in one
cbd4257e6d85149 Konstantin Komarov 2020-10-30  614  		 */
cbd4257e6d85149 Konstantin Komarov 2020-10-30  615  		err = mi_pack_runs(mi_p, attr_p, &run, evcn + 1 - svcn_p);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  616  		if (err)
cbd4257e6d85149 Konstantin Komarov 2020-10-30  617  			break;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  618  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  619  		next_svcn = le64_to_cpu(attr_p->nres.evcn) + 1;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  620  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  621  		if (next_svcn >= evcn + 1) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  622  			/* we can remove this attribute segment */
cbd4257e6d85149 Konstantin Komarov 2020-10-30  623  			al_remove_le(ni, le);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  624  			mi_remove_attr(mi, attr);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  625  			le = le_p;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  626  			continue;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  627  		}
cbd4257e6d85149 Konstantin Komarov 2020-10-30  628  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  629  		attr->nres.svcn = le->vcn = cpu_to_le64(next_svcn);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  630  		mi->dirty = true;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  631  		ni->attr_list.dirty = true;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  632  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  633  		if (evcn + 1 == alloc) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  634  			err = mi_pack_runs(mi, attr, &run,
cbd4257e6d85149 Konstantin Komarov 2020-10-30  635  					   evcn + 1 - next_svcn);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  636  			if (err)
cbd4257e6d85149 Konstantin Komarov 2020-10-30  637  				break;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  638  			mi_p = NULL;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  639  		} else {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  640  			mi_p = mi;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  641  			attr_p = attr;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  642  			svcn_p = next_svcn;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  643  			evcn_p = evcn;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  644  			le_p = le;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  645  			run_truncate_head(&run, next_svcn);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  646  		}
cbd4257e6d85149 Konstantin Komarov 2020-10-30  647  	}
cbd4257e6d85149 Konstantin Komarov 2020-10-30  648  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  649  	if (err) {
cbd4257e6d85149 Konstantin Komarov 2020-10-30  650  		ntfs_inode_warn(&ni->vfs_inode, "there is a problem");
cbd4257e6d85149 Konstantin Komarov 2020-10-30  651  		ntfs_set_state(sbi, NTFS_DIRTY_ERROR);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  652  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  653  		/* Pack loaded but not packed runs */
cbd4257e6d85149 Konstantin Komarov 2020-10-30  654  		if (mi_p)
cbd4257e6d85149 Konstantin Komarov 2020-10-30  655  			mi_pack_runs(mi_p, attr_p, &run, evcn_p + 1 - svcn_p);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  656  	}
cbd4257e6d85149 Konstantin Komarov 2020-10-30  657  
cbd4257e6d85149 Konstantin Komarov 2020-10-30  658  	run_close(&run);
cbd4257e6d85149 Konstantin Komarov 2020-10-30  659  	return err;
cbd4257e6d85149 Konstantin Komarov 2020-10-30  660  }
cbd4257e6d85149 Konstantin Komarov 2020-10-30  661  

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 17639 bytes --]

^ permalink raw reply	[flat|nested] 21+ messages in thread

* Re: [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software
  2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
                   ` (11 preceding siblings ...)
  2020-10-30 16:41 ` Pali Rohár
@ 2020-10-31  2:42 ` Eric Biggers
  2020-10-31  2:48   ` Eric Biggers
  12 siblings, 1 reply; 21+ messages in thread
From: Eric Biggers @ 2020-10-31  2:42 UTC (permalink / raw)
  To: Konstantin Komarov
  Cc: linux-fsdevel, viro, linux-kernel, pali, dsterba, aaptel, willy,
	rdunlap, joe, mark, nborisov, linux-ntfs-dev, anton

Hi Konstantin,

On Fri, Oct 30, 2020 at 06:02:29PM +0300, Konstantin Komarov wrote:
> This patch adds NTFS Read-Write driver to fs/ntfs3.
> 
> Having decades of expertise in commercial file systems development and huge
> test coverage, we at Paragon Software GmbH want to make our contribution to
> the Open Source Community by providing implementation of NTFS Read-Write
> driver for the Linux Kernel.
> 
> This is fully functional NTFS Read-Write driver. Current version works with
> NTFS(including v3.1) and normal/compressed/sparse files and supports journal replaying.
> 
> We plan to support this version after the codebase once merged, and add new
> features and fix bugs. For example, full journaling support over JBD will be
> added in later updates.
> 

Have you tried testing this filesystem using some of the kernel debugging
options (lockdep, KASAN, etc.?).  I tried a basic test just for fun, and I
immediately got a lockdep report:

mkfs.ntfs -f /dev/vdb
mount /dev/vdb /mnt -t ntfs3
echo foo > /mnt/foo

======================================================
WARNING: possible circular locking dependency detected
5.10.0-rc1-00275-ga34a2c322380 #33 Not tainted
------------------------------------------------------
bash/160 is trying to acquire lock:
ffff888011e68108 (&ni->ni_lock){+.+.}-{3:3}, at: ni_lock fs/ntfs3/ntfs_fs.h:959 [inline]
ffff888011e68108 (&ni->ni_lock){+.+.}-{3:3}, at: ntfs_set_size+0xee/0x210 fs/ntfs3/inode.c:880

but task is already holding lock:
ffff888011e68370 (&sb->s_type->i_mutex_key#11){+.+.}-{3:3}, at: inode_trylock include/linux/fs.h:794 [inline]
ffff888011e68370 (&sb->s_type->i_mutex_key#11){+.+.}-{3:3}, at: ntfs_file_write_iter+0x1bc/0x4e0 fs/ntfs3/file.c:1040

which lock already depends on the new lock.


the existing dependency chain (in reverse order) is:

-> #1 (&sb->s_type->i_mutex_key#11){+.+.}-{3:3}:
       __lock_acquire+0x7b7/0x1060 kernel/locking/lockdep.c:4837
       lock_acquire.part.0+0x13b/0x2a0 kernel/locking/lockdep.c:5442
       lock_acquire+0xb4/0x1d0 kernel/locking/lockdep.c:5415
       down_write+0x93/0x150 kernel/locking/rwsem.c:1531
       inode_lock include/linux/fs.h:774 [inline]
       ntfs_set_state+0x192/0x650 fs/ntfs3/fsntfs.c:996
       ntfs_create_inode+0x318/0x4590 fs/ntfs3/inode.c:1254
       ntfs_atomic_open+0x350/0x450 fs/ntfs3/namei.c:521
       atomic_open+0x16b/0x420 fs/namei.c:2970
       lookup_open+0x7d1/0xdc0 fs/namei.c:3076
       open_last_lookups+0x277/0x1040 fs/namei.c:3178
       path_openat+0x161/0x310 fs/namei.c:3366
       do_filp_open+0x17d/0x3b0 fs/namei.c:3396
       do_sys_openat2+0x120/0x3d0 fs/open.c:1168
       do_sys_open fs/open.c:1184 [inline]
       __do_sys_openat fs/open.c:1200 [inline]
       __se_sys_openat fs/open.c:1195 [inline]
       __x64_sys_openat+0x124/0x200 fs/open.c:1195
       do_syscall_64+0x32/0x50 arch/x86/entry/common.c:46
       entry_SYSCALL_64_after_hwframe+0x44/0xa9

-> #0 (&ni->ni_lock){+.+.}-{3:3}:
       check_prev_add+0x162/0x20b0 kernel/locking/lockdep.c:2864
       check_prevs_add kernel/locking/lockdep.c:2989 [inline]
       validate_chain+0xbff/0x1620 kernel/locking/lockdep.c:3607
       __lock_acquire+0x7b7/0x1060 kernel/locking/lockdep.c:4837
       lock_acquire.part.0+0x13b/0x2a0 kernel/locking/lockdep.c:5442
       lock_acquire+0xb4/0x1d0 kernel/locking/lockdep.c:5415
       __mutex_lock_common kernel/locking/mutex.c:956 [inline]
       __mutex_lock+0x142/0x12d0 kernel/locking/mutex.c:1103
       mutex_lock_nested+0x16/0x20 kernel/locking/mutex.c:1118
       ni_lock fs/ntfs3/ntfs_fs.h:959 [inline]
       ntfs_set_size+0xee/0x210 fs/ntfs3/inode.c:880
       ntfs_extend_ex+0x3b1/0x5c0 fs/ntfs3/file.c:466
       ntfs_file_write_iter+0x244/0x4e0 fs/ntfs3/file.c:1050
       call_write_iter include/linux/fs.h:1887 [inline]
       new_sync_write+0x348/0x6d0 fs/read_write.c:518
       vfs_write+0x408/0x5b0 fs/read_write.c:605
       ksys_write+0x11d/0x220 fs/read_write.c:658
       __do_sys_write fs/read_write.c:670 [inline]
       __se_sys_write fs/read_write.c:667 [inline]
       __x64_sys_write+0x6e/0xb0 fs/read_write.c:667
       do_syscall_64+0x32/0x50 arch/x86/entry/common.c:46
       entry_SYSCALL_64_after_hwframe+0x44/0xa9

other info that might help us debug this:

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&sb->s_type->i_mutex_key#11);
                               lock(&ni->ni_lock);
                               lock(&sb->s_type->i_mutex_key#11);
  lock(&ni->ni_lock);

 *** DEADLOCK ***

2 locks held by bash/160:
 #0: ffff8880084ac458 (sb_writers#11){.+.+}-{0:0}, at: file_start_write include/linux/fs.h:2759 [inline]
 #0: ffff8880084ac458 (sb_writers#11){.+.+}-{0:0}, at: vfs_write+0x3bc/0x5b0 fs/read_write.c:601
 #1: ffff888011e68370 (&sb->s_type->i_mutex_key#11){+.+.}-{3:3}, at: inode_trylock include/linux/fs.h:794 [inline]
 #1: ffff888011e68370 (&sb->s_type->i_mutex_key#11){+.+.}-{3:3}, at: ntfs_file_write_iter+0x1bc/0x4e0 fs/ntfs3/file.c:1040

stack backtrace:
CPU: 1 PID: 160 Comm: bash Not tainted 5.10.0-rc1-00275-ga34a2c322380 #33
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS ArchLinux 1.14.0-1 04/01/2014
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0xa4/0xd9 lib/dump_stack.c:118
 print_circular_bug.cold+0x34/0x39 kernel/locking/lockdep.c:1994
 check_noncircular+0x277/0x350 kernel/locking/lockdep.c:2115
 check_prev_add+0x162/0x20b0 kernel/locking/lockdep.c:2864
 check_prevs_add kernel/locking/lockdep.c:2989 [inline]
 validate_chain+0xbff/0x1620 kernel/locking/lockdep.c:3607
 __lock_acquire+0x7b7/0x1060 kernel/locking/lockdep.c:4837
 lock_acquire.part.0+0x13b/0x2a0 kernel/locking/lockdep.c:5442
 lock_acquire+0xb4/0x1d0 kernel/locking/lockdep.c:5415
 __mutex_lock_common kernel/locking/mutex.c:956 [inline]
 __mutex_lock+0x142/0x12d0 kernel/locking/mutex.c:1103
 mutex_lock_nested+0x16/0x20 kernel/locking/mutex.c:1118
 ni_lock fs/ntfs3/ntfs_fs.h:959 [inline]
 ntfs_set_size+0xee/0x210 fs/ntfs3/inode.c:880
 ntfs_extend_ex+0x3b1/0x5c0 fs/ntfs3/file.c:466
 ntfs_file_write_iter+0x244/0x4e0 fs/ntfs3/file.c:1050
 call_write_iter include/linux/fs.h:1887 [inline]
 new_sync_write+0x348/0x6d0 fs/read_write.c:518
 vfs_write+0x408/0x5b0 fs/read_write.c:605
 ksys_write+0x11d/0x220 fs/read_write.c:658
 __do_sys_write fs/read_write.c:670 [inline]
 __se_sys_write fs/read_write.c:667 [inline]
 __x64_sys_write+0x6e/0xb0 fs/read_write.c:667
 do_syscall_64+0x32/0x50 arch/x86/entry/common.c:46
 entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x7fad69e95f67
Code: 0d 00 f7 d8 64 89 02 48 c7 c0 ff ff ff ff eb b7 0f 1f 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 51 c3 48 83 ec 28 48 89 54 24 18 48 89 74 24
RSP: 002b:00007ffc66bcfaa8 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
RAX: ffffffffffffffda RBX: 0000000000000004 RCX: 00007fad69e95f67
RDX: 0000000000000004 RSI: 0000555d79dccba0 RDI: 0000000000000001
RBP: 0000555d79dccba0 R08: 000000000000000a R09: 00007fad69f67a60
R10: 0000000000000080 R11: 0000000000000246 R12: 0000000000000004
R13: 00007fad69f68520 R14: 0000000000000004 R15: 00007fad69f68720

^ permalink raw reply	[flat|nested] 21+ messages in thread

* Re: [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software
  2020-10-31  2:42 ` Eric Biggers
@ 2020-10-31  2:48   ` Eric Biggers
  0 siblings, 0 replies; 21+ messages in thread
From: Eric Biggers @ 2020-10-31  2:48 UTC (permalink / raw)
  To: Konstantin Komarov
  Cc: linux-fsdevel, viro, linux-kernel, pali, dsterba, aaptel, willy,
	rdunlap, joe, mark, nborisov, linux-ntfs-dev, anton

On Fri, Oct 30, 2020 at 07:42:09PM -0700, Eric Biggers wrote:
> Hi Konstantin,
> 
> On Fri, Oct 30, 2020 at 06:02:29PM +0300, Konstantin Komarov wrote:
> > This patch adds NTFS Read-Write driver to fs/ntfs3.
> > 
> > Having decades of expertise in commercial file systems development and huge
> > test coverage, we at Paragon Software GmbH want to make our contribution to
> > the Open Source Community by providing implementation of NTFS Read-Write
> > driver for the Linux Kernel.
> > 
> > This is fully functional NTFS Read-Write driver. Current version works with
> > NTFS(including v3.1) and normal/compressed/sparse files and supports journal replaying.
> > 
> > We plan to support this version after the codebase once merged, and add new
> > features and fix bugs. For example, full journaling support over JBD will be
> > added in later updates.
> > 
> 
> Have you tried testing this filesystem using some of the kernel debugging
> options (lockdep, KASAN, etc.?).  I tried a basic test just for fun, and I
> immediately got a lockdep report:
> 
> mkfs.ntfs -f /dev/vdb
> mount /dev/vdb /mnt -t ntfs3
> echo foo > /mnt/foo
> 
> ======================================================
> WARNING: possible circular locking dependency detected
> 5.10.0-rc1-00275-ga34a2c322380 #33 Not tainted
> ------------------------------------------------------
> bash/160 is trying to acquire lock:
> ffff888011e68108 (&ni->ni_lock){+.+.}-{3:3}, at: ni_lock fs/ntfs3/ntfs_fs.h:959 [inline]
> ffff888011e68108 (&ni->ni_lock){+.+.}-{3:3}, at: ntfs_set_size+0xee/0x210 fs/ntfs3/inode.c:880
> 
> but task is already holding lock:
> ffff888011e68370 (&sb->s_type->i_mutex_key#11){+.+.}-{3:3}, at: inode_trylock include/linux/fs.h:794 [inline]
> ffff888011e68370 (&sb->s_type->i_mutex_key#11){+.+.}-{3:3}, at: ntfs_file_write_iter+0x1bc/0x4e0 fs/ntfs3/file.c:1040
> 
> which lock already depends on the new lock.

Also trying to create a symlink causes a stack out-of-bounds access:

$ mkfs.ntfs -f /dev/vdb
$ mount /dev/vdb /mnt -t ntfs3
$ ln -s target /mnt/symlink

BUG: KASAN: stack-out-of-bounds in memcpy include/linux/string.h:399 [inline]
BUG: KASAN: stack-out-of-bounds in hdr_insert_de+0x224/0x4d0 fs/ntfs3/index.c:851
Read of size 32 at addr ffffc900007b7b10 by task ln/181

CPU: 1 PID: 181 Comm: ln Not tainted 5.10.0-rc1-00275-ga34a2c322380 #33
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS ArchLinux 1.14.0-1 04/01/2014
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0xa4/0xd9 lib/dump_stack.c:118
 print_address_description.constprop.0+0x1f/0x160 mm/kasan/report.c:385
 __kasan_report.cold+0x37/0x7f mm/kasan/report.c:545
 kasan_report+0x3e/0x60 mm/kasan/report.c:562
 check_memory_region_inline mm/kasan/generic.c:186 [inline]
 check_memory_region+0xfb/0x1d0 mm/kasan/generic.c:192
 memcpy+0x24/0x60 mm/kasan/common.c:105
 memcpy include/linux/string.h:399 [inline]
 hdr_insert_de+0x224/0x4d0 fs/ntfs3/index.c:851
 indx_insert_into_root+0x5d0/0x1d10 fs/ntfs3/index.c:1585
 indx_insert_entry+0x299/0x4e0 fs/ntfs3/index.c:1936
 ntfs_insert_reparse+0x133/0x1b0 fs/ntfs3/fsntfs.c:2425
 ntfs_create_inode+0x28ec/0x4590 fs/ntfs3/inode.c:1511
 ntfs_symlink+0xb1/0xf0 fs/ntfs3/namei.c:198
 vfs_symlink fs/namei.c:3960 [inline]
 vfs_symlink+0x237/0x380 fs/namei.c:3946
 do_symlinkat+0x125/0x220 fs/namei.c:3987
 __do_sys_symlinkat fs/namei.c:4001 [inline]
 __se_sys_symlinkat fs/namei.c:3998 [inline]
 __x64_sys_symlinkat+0x6e/0xb0 fs/namei.c:3998
 do_syscall_64+0x32/0x50 arch/x86/entry/common.c:46
 entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x7fdaf2fb2b2b
Code: 73 01 c3 48 8b 0d 45 f3 0c 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa b8 0a 01 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 15 f3 0c 00 f7 d8 64 89 01 48
RSP: 002b:00007ffe428da378 EFLAGS: 00000246 ORIG_RAX: 000000000000010a
RAX: ffffffffffffffda RBX: 00007ffe428da598 RCX: 00007fdaf2fb2b2b
RDX: 00007ffe428db812 RSI: 00000000ffffff9c RDI: 00007ffe428db80b
RBP: 00007ffe428da588 R08: 00000000ffffffff R09: 0000000000000000
R10: 00005647f9c2b340 R11: 0000000000000246 R12: 0000000000000002
R13: 0000000000000001 R14: 0000000000000000 R15: 0000000000000000


addr ffffc900007b7b10 is located in stack of task ln/181 at offset 32 in frame:
 ntfs_insert_reparse+0x0/0x1b0 fs/ntfs3/fsntfs.c:2387

this frame has 1 object:
 [32, 60) 're'

Memory state around the buggy address:
 ffffc900007b7a00: 00 00 00 00 00 f1 f1 f1 f1 f1 f1 04 f2 00 f3 f3
 ffffc900007b7a80: f3 00 00 00 00 00 00 00 00 00 00 00 00 00 f1 f1
>ffffc900007b7b00: f1 f1 00 00 00 04 f3 f3 f3 f3 00 00 00 00 00 00
                                  ^
 ffffc900007b7b80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 ffffc900007b7c00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 f1 f1
==================================================================

^ permalink raw reply	[flat|nested] 21+ messages in thread

* Re: [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software
  2020-10-30 16:41 ` Pali Rohár
@ 2020-10-31  8:51   ` Christoph Hellwig
  0 siblings, 0 replies; 21+ messages in thread
From: Christoph Hellwig @ 2020-10-31  8:51 UTC (permalink / raw)
  To: Pali Rohár
  Cc: Konstantin Komarov, Christoph Hellwig, linux-fsdevel, viro,
	linux-kernel, dsterba, aaptel, willy, rdunlap, joe, mark,
	nborisov, linux-ntfs-dev, anton

On Fri, Oct 30, 2020 at 05:41:22PM +0100, Pali Rohár wrote:
> I would like to open discussion about two ntfs kernel drivers. Do we
> really need two drivers (one read only - current version and one
> read/write - this new version)?
> 
> What other people think?
> 
> I remember that Christoph (added to loop) had in past a good argument
> about old staging exfat driver (it had support also for fat32/vfat),
> that it would cause problems if two filesystem drivers would provide
> support for same filesystem.

Yes, we really should not have two drivers normally.  I think Konstantin
and Anton need to have a chat on how to go forard.  Without knowing the
details read-write support sounds like a killer feature we'd really want
if there aren't any code quality or other feature regression problems,
but I haven't had a chance to look at the code at all yet.

^ permalink raw reply	[flat|nested] 21+ messages in thread

* [kbuild] Re: [PATCH v11 09/10] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile
  2020-10-30 15:02 ` [PATCH v11 09/10] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile Konstantin Komarov
  2020-10-31  1:23   ` kernel test robot
@ 2020-11-02  8:36   ` Dan Carpenter
  2020-11-03  3:06   ` kernel test robot
  2 siblings, 0 replies; 21+ messages in thread
From: Dan Carpenter @ 2020-11-02  8:36 UTC (permalink / raw)
  To: kbuild, Konstantin Komarov, linux-fsdevel
  Cc: lkp, Dan Carpenter, kbuild-all, viro, linux-kernel, pali,
	dsterba, aaptel, willy, rdunlap, joe, mark

[-- Attachment #1: Type: text/plain, Size: 11858 bytes --]

Hi Konstantin,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on linus/master]
[also build test WARNING on v5.10-rc2 next-20201030]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch ]

url:    https://github.com/0day-ci/linux/commits/Konstantin-Komarov/NTFS-read-write-driver-GPL-implementation-by-Paragon-Software/20201031-220904 
base:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git  5fc6b075e165f641fbc366b58b578055762d5f8c
config: i386-randconfig-m021-20201101 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-15) 9.3.0

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>

smatch warnings:
fs/ntfs3/attrib.c:331 attr_set_size_res() error: dereferencing freed memory 'attr_s'
fs/ntfs3/attrib.c:1267 attr_allocate_frame() error: uninitialized symbol 'hint'.
fs/ntfs3/attrib.c:1393 attr_allocate_frame() error: we previously assumed 'attr_b' could be null (see line 1306)
fs/ntfs3/namei.c:438 ntfs_rename() warn: variable dereferenced before check 'old_inode' (see line 296)
fs/ntfs3/fsntfs.c:844 ntfs_clear_mft_tail() error: uninitialized symbol 'err'.
fs/ntfs3/fsntfs.c:1294 ntfs_read_run_nb() error: uninitialized symbol 'idx'.
fs/ntfs3/frecord.c:166 ni_load_mi_ex() error: we previously assumed 'r' could be null (see line 159)
fs/ntfs3/frecord.c:505 ni_ins_new_attr() error: we previously assumed 'le' could be null (see line 490)
fs/ntfs3/frecord.c:658 ni_repack() warn: 'run.runs_' double freed
fs/ntfs3/frecord.c:1439 ni_insert_nonresident() warn: potential memory corrupting cast 8 vs 2 bytes
fs/ntfs3/frecord.c:2214 ni_read_frame() warn: ignoring unreachable code.
fs/ntfs3/xattr.c:514 ntfs_get_acl_ex() warn: passing zero to 'ERR_PTR'
fs/ntfs3/index.c:1133 indx_find() warn: variable dereferenced before check 'fnd' (see line 1117)
fs/ntfs3/index.c:1371 indx_find_raw() error: we previously assumed 'n' could be null (see line 1349)
fs/ntfs3/index.c:1404 indx_create_allocate() warn: should '1 << indx->index_bits' be a 64 bit type?
fs/ntfs3/index.c:1755 indx_insert_into_root() warn: possible memory leak of 're'
fs/ntfs3/index.c:549 hdr_find_split() warn: variable dereferenced before check 'e' (see line 547)
fs/ntfs3/inode.c:687 ntfs_readpage() warn: should 'page->index << 12' be a 64 bit type?
fs/ntfs3/fslog.c:2205 last_log_lsn() warn: possible memory leak of 'page_bufs'
fs/ntfs3/fslog.c:2418 find_log_rec() error: we previously assumed 'rh' could be null (see line 2404)
fs/ntfs3/fslog.c:2551 find_client_next_lsn() error: double free of 'lcb->lrh'
fs/ntfs3/fslog.c:639 enum_rstbl() error: we previously assumed 't' could be null (see line 628)
fs/ntfs3/fslog.c:3158 do_action() warn: variable dereferenced before check 'mi' (see line 3118)
fs/ntfs3/fslog.c:3913 log_replay() error: dereferencing freed memory 'rst_info.r_page'

vim +/attr_s +331 fs/ntfs3/attrib.c

e3a1cdcc648083 Konstantin Komarov 2020-10-30  241  static int attr_set_size_res(struct ntfs_inode *ni, struct ATTRIB *attr,
e3a1cdcc648083 Konstantin Komarov 2020-10-30  242  			     struct ATTR_LIST_ENTRY *le, struct mft_inode *mi,
e3a1cdcc648083 Konstantin Komarov 2020-10-30  243  			     u64 new_size, struct runs_tree *run,
e3a1cdcc648083 Konstantin Komarov 2020-10-30  244  			     struct ATTRIB **ins_attr)
e3a1cdcc648083 Konstantin Komarov 2020-10-30  245  {
e3a1cdcc648083 Konstantin Komarov 2020-10-30  246  	int err = 0;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  247  	struct ntfs_sb_info *sbi = mi->sbi;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  248  	struct MFT_REC *rec = mi->mrec;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  249  	u32 used = le32_to_cpu(rec->used);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  250  	u32 asize = le32_to_cpu(attr->size);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  251  	u32 aoff = PtrOffset(rec, attr);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  252  	u32 rsize = le32_to_cpu(attr->res.data_size);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  253  	u32 tail = used - aoff - asize;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  254  	char *next = Add2Ptr(attr, asize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  255  	int dsize = QuadAlign(new_size) - QuadAlign(rsize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  256  	CLST len, alen;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  257  	struct ATTRIB *attr_s = NULL;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  258  	bool is_ext;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  259  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  260  	if (dsize < 0) {
e3a1cdcc648083 Konstantin Komarov 2020-10-30  261  		memmove(next + dsize, next, tail);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  262  	} else if (dsize > 0) {
e3a1cdcc648083 Konstantin Komarov 2020-10-30  263  		if (used + dsize > sbi->max_bytes_per_attr)
e3a1cdcc648083 Konstantin Komarov 2020-10-30  264  			goto resident2nonresident;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  265  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  266  		memmove(next + dsize, next, tail);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  267  		memset(next, 0, dsize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  268  	}
e3a1cdcc648083 Konstantin Komarov 2020-10-30  269  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  270  	rec->used = cpu_to_le32(used + dsize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  271  	attr->size = cpu_to_le32(asize + dsize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  272  	attr->res.data_size = cpu_to_le32(new_size);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  273  	mi->dirty = true;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  274  	*ins_attr = attr;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  275  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  276  	return 0;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  277  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  278  resident2nonresident:
e3a1cdcc648083 Konstantin Komarov 2020-10-30  279  	len = bytes_to_cluster(sbi, rsize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  280  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  281  	run_init(run);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  282  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  283  	is_ext = is_attr_ext(attr);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  284  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  285  	if (!len) {
e3a1cdcc648083 Konstantin Komarov 2020-10-30  286  		alen = 0;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  287  	} else if (is_ext) {
e3a1cdcc648083 Konstantin Komarov 2020-10-30  288  		if (!run_add_entry(run, 0, SPARSE_LCN, len)) {
e3a1cdcc648083 Konstantin Komarov 2020-10-30  289  			err = -ENOMEM;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  290  			goto out;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  291  		}
e3a1cdcc648083 Konstantin Komarov 2020-10-30  292  		alen = len;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  293  	} else {
e3a1cdcc648083 Konstantin Komarov 2020-10-30  294  		err = attr_allocate_clusters(sbi, run, 0, 0, len, NULL,
e3a1cdcc648083 Konstantin Komarov 2020-10-30  295  					     ALLOCATE_DEF, &alen, 0, NULL);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  296  		if (err)
e3a1cdcc648083 Konstantin Komarov 2020-10-30  297  			goto out;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  298  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  299  		err = ntfs_sb_write_run(sbi, run, 0, resident_data(attr),
e3a1cdcc648083 Konstantin Komarov 2020-10-30  300  					rsize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  301  		if (err)
e3a1cdcc648083 Konstantin Komarov 2020-10-30  302  			goto out;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  303  	}
e3a1cdcc648083 Konstantin Komarov 2020-10-30  304  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  305  	attr_s = ntfs_memdup(attr, asize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  306  	if (!attr_s) {
e3a1cdcc648083 Konstantin Komarov 2020-10-30  307  		err = -ENOMEM;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  308  		goto out;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  309  	}
e3a1cdcc648083 Konstantin Komarov 2020-10-30  310  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  311  	/*verify(mi_remove_attr(mi, attr));*/
e3a1cdcc648083 Konstantin Komarov 2020-10-30  312  	used -= asize;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  313  	memmove(attr, Add2Ptr(attr, asize), used - aoff);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  314  	rec->used = cpu_to_le32(used);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  315  	mi->dirty = true;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  316  	if (le)
e3a1cdcc648083 Konstantin Komarov 2020-10-30  317  		al_remove_le(ni, le);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  318  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  319  	err = ni_insert_nonresident(ni, attr_s->type, attr_name(attr_s),
e3a1cdcc648083 Konstantin Komarov 2020-10-30  320  				    attr_s->name_len, run, 0, alen,
e3a1cdcc648083 Konstantin Komarov 2020-10-30  321  				    attr_s->flags, &attr, NULL);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  322  	if (err)
e3a1cdcc648083 Konstantin Komarov 2020-10-30  323  		goto out;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  324  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  325  	ntfs_free(attr_s);
                                                        ^^^^^^^^^^^^^^^^^
Freed.

e3a1cdcc648083 Konstantin Komarov 2020-10-30  326  	attr->nres.data_size = cpu_to_le64(rsize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  327  	attr->nres.valid_size = attr->nres.data_size;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  328  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  329  	*ins_attr = attr;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  330  
e3a1cdcc648083 Konstantin Komarov 2020-10-30 @331  	if (attr_s->type == ATTR_DATA && !attr_s->name_len &&
                                                            ^^^^^^^^^^^^                  ^^^^^^^^^^^^^^^^
Dereferenced after a free.

e3a1cdcc648083 Konstantin Komarov 2020-10-30  332  	    run == &ni->file.run) {
e3a1cdcc648083 Konstantin Komarov 2020-10-30  333  		ni->ni_flags &= ~NI_FLAG_RESIDENT;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  334  	}
e3a1cdcc648083 Konstantin Komarov 2020-10-30  335  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  336  	/* Resident attribute becomes non resident */
e3a1cdcc648083 Konstantin Komarov 2020-10-30  337  	return 0;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  338  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  339  out:
e3a1cdcc648083 Konstantin Komarov 2020-10-30  340  	/* undo: do not trim new allocated clusters */
e3a1cdcc648083 Konstantin Komarov 2020-10-30  341  	run_deallocate(sbi, run, false);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  342  	run_close(run);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  343  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  344  	if (attr_s) {
e3a1cdcc648083 Konstantin Komarov 2020-10-30  345  		memmove(next, Add2Ptr(rec, aoff), used - aoff);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  346  		memcpy(Add2Ptr(rec, aoff), attr_s, asize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  347  		rec->used = cpu_to_le32(used + asize);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  348  		mi->dirty = true;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  349  		ntfs_free(attr_s);
e3a1cdcc648083 Konstantin Komarov 2020-10-30  350  		/*reinsert le*/
e3a1cdcc648083 Konstantin Komarov 2020-10-30  351  	}
e3a1cdcc648083 Konstantin Komarov 2020-10-30  352  
e3a1cdcc648083 Konstantin Komarov 2020-10-30  353  	return err;
e3a1cdcc648083 Konstantin Komarov 2020-10-30  354  }

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org 

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 32884 bytes --]

[-- Attachment #3: Type: text/plain, Size: 149 bytes --]

_______________________________________________
kbuild mailing list -- kbuild@lists.01.org
To unsubscribe send an email to kbuild-leave@lists.01.org

^ permalink raw reply	[flat|nested] 21+ messages in thread

* Re: [PATCH v11 09/10] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile
  2020-10-30 15:02 ` [PATCH v11 09/10] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile Konstantin Komarov
  2020-10-31  1:23   ` kernel test robot
  2020-11-02  8:36   ` [kbuild] " Dan Carpenter
@ 2020-11-03  3:06   ` kernel test robot
  2 siblings, 0 replies; 21+ messages in thread
From: kernel test robot @ 2020-11-03  3:06 UTC (permalink / raw)
  To: Konstantin Komarov, linux-fsdevel
  Cc: kbuild-all, viro, linux-kernel, pali, dsterba, aaptel, willy,
	rdunlap, joe, mark

[-- Attachment #1: Type: text/plain, Size: 18919 bytes --]

Hi Konstantin,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on linus/master]
[also build test ERROR on v5.10-rc2 next-20201102]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Konstantin-Komarov/NTFS-read-write-driver-GPL-implementation-by-Paragon-Software/20201031-220904
base:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 5fc6b075e165f641fbc366b58b578055762d5f8c
config: alpha-allmodconfig (attached as .config)
compiler: alpha-linux-gcc (GCC) 9.3.0
reproduce (this is a W=1 build):
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # https://github.com/0day-ci/linux/commit/0868596d71532d31ce16ee71c5e73c154878416e
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Konstantin-Komarov/NTFS-read-write-driver-GPL-implementation-by-Paragon-Software/20201031-220904
        git checkout 0868596d71532d31ce16ee71c5e73c154878416e
        # save the attached .config to linux build tree
        COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross ARCH=alpha 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

   In file included from fs/ntfs3/frecord.c:15:
   fs/ntfs3/ntfs.h:430:1: warning: 'inline' is not at beginning of declaration [-Wold-style-declaration]
     430 | static const inline __le16 *attr_name(const struct ATTRIB *attr)
         | ^~~~~~
   fs/ntfs3/ntfs.h:547:1: warning: 'inline' is not at beginning of declaration [-Wold-style-declaration]
     547 | static const inline __le16 *le_name(const struct ATTR_LIST_ENTRY *le)
         | ^~~~~~
   fs/ntfs3/frecord.c: In function 'ni_read_frame':
>> fs/ntfs3/frecord.c:2247:17: error: implicit declaration of function 'vmap'; did you mean 'kmap'? [-Werror=implicit-function-declaration]
    2247 |  frame_ondisk = vmap(pages_disk, npages_disk, VM_MAP, PAGE_KERNEL);
         |                 ^~~~
         |                 kmap
>> fs/ntfs3/frecord.c:2247:47: error: 'VM_MAP' undeclared (first use in this function); did you mean 'VM_MTE'?
    2247 |  frame_ondisk = vmap(pages_disk, npages_disk, VM_MAP, PAGE_KERNEL);
         |                                               ^~~~~~
         |                                               VM_MTE
   fs/ntfs3/frecord.c:2247:47: note: each undeclared identifier is reported only once for each function it appears in
>> fs/ntfs3/frecord.c:2278:2: error: implicit declaration of function 'vunmap'; did you mean 'kunmap'? [-Werror=implicit-function-declaration]
    2278 |  vunmap(frame_mem);
         |  ^~~~~~
         |  kunmap
   fs/ntfs3/frecord.c: In function 'ni_write_frame':
   fs/ntfs3/frecord.c:2375:51: error: 'VM_MAP' undeclared (first use in this function); did you mean 'VM_MTE'?
    2375 |  frame_ondisk = vmap(pages_disk, pages_per_frame, VM_MAP, PAGE_KERNEL);
         |                                                   ^~~~~~
         |                                                   VM_MTE
   cc1: some warnings being treated as errors

vim +2247 fs/ntfs3/frecord.c

8f609ab22280366 Konstantin Komarov 2020-10-30  2099  
8f609ab22280366 Konstantin Komarov 2020-10-30  2100  /*
8f609ab22280366 Konstantin Komarov 2020-10-30  2101   * ni_read_frame
8f609ab22280366 Konstantin Komarov 2020-10-30  2102   *
8f609ab22280366 Konstantin Komarov 2020-10-30  2103   * pages - array of locked pages
8f609ab22280366 Konstantin Komarov 2020-10-30  2104   */
8f609ab22280366 Konstantin Komarov 2020-10-30  2105  int ni_read_frame(struct ntfs_inode *ni, u64 frame_vbo, struct page **pages,
8f609ab22280366 Konstantin Komarov 2020-10-30  2106  		  u32 pages_per_frame)
8f609ab22280366 Konstantin Komarov 2020-10-30  2107  {
8f609ab22280366 Konstantin Komarov 2020-10-30  2108  	int err;
8f609ab22280366 Konstantin Komarov 2020-10-30  2109  	struct ntfs_sb_info *sbi = ni->mi.sbi;
8f609ab22280366 Konstantin Komarov 2020-10-30  2110  	char *frame_ondisk = NULL;
8f609ab22280366 Konstantin Komarov 2020-10-30  2111  	char *frame_mem = NULL;
8f609ab22280366 Konstantin Komarov 2020-10-30  2112  	struct page **pages_disk = NULL;
8f609ab22280366 Konstantin Komarov 2020-10-30  2113  	u64 valid_size = ni->i_valid;
8f609ab22280366 Konstantin Komarov 2020-10-30  2114  	u32 frame_size;
8f609ab22280366 Konstantin Komarov 2020-10-30  2115  	size_t unc_size;
8f609ab22280366 Konstantin Komarov 2020-10-30  2116  	u32 ondisk_size, i, npages_disk;
8f609ab22280366 Konstantin Komarov 2020-10-30  2117  	struct page *pg;
8f609ab22280366 Konstantin Komarov 2020-10-30  2118  	struct ATTRIB *attr;
8f609ab22280366 Konstantin Komarov 2020-10-30  2119  	CLST frame, clst_data;
8f609ab22280366 Konstantin Komarov 2020-10-30  2120  
8f609ab22280366 Konstantin Komarov 2020-10-30  2121  	attr = ni_find_attr(ni, NULL, NULL, ATTR_DATA, NULL, 0, NULL, NULL);
8f609ab22280366 Konstantin Komarov 2020-10-30  2122  	if (!attr) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2123  		err = -ENOENT;
8f609ab22280366 Konstantin Komarov 2020-10-30  2124  		goto out;
8f609ab22280366 Konstantin Komarov 2020-10-30  2125  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2126  
8f609ab22280366 Konstantin Komarov 2020-10-30  2127  	if (!attr->non_res) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2128  		u32 data_size = le32_to_cpu(attr->res.data_size);
8f609ab22280366 Konstantin Komarov 2020-10-30  2129  		void *kaddr = kmap_atomic(pages[0]);
8f609ab22280366 Konstantin Komarov 2020-10-30  2130  
8f609ab22280366 Konstantin Komarov 2020-10-30  2131  		memset(kaddr, 0, PAGE_SIZE);
8f609ab22280366 Konstantin Komarov 2020-10-30  2132  		if (frame_vbo < data_size) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2133  			err = data_size - frame_vbo;
8f609ab22280366 Konstantin Komarov 2020-10-30  2134  			if (err > PAGE_SIZE)
8f609ab22280366 Konstantin Komarov 2020-10-30  2135  				err = PAGE_SIZE;
8f609ab22280366 Konstantin Komarov 2020-10-30  2136  			memcpy(kaddr, resident_data(attr) + frame_vbo, err);
8f609ab22280366 Konstantin Komarov 2020-10-30  2137  		}
8f609ab22280366 Konstantin Komarov 2020-10-30  2138  		flush_dcache_page(pages[0]);
8f609ab22280366 Konstantin Komarov 2020-10-30  2139  		kunmap_atomic(kaddr);
8f609ab22280366 Konstantin Komarov 2020-10-30  2140  		err = 0;
8f609ab22280366 Konstantin Komarov 2020-10-30  2141  		goto out;
8f609ab22280366 Konstantin Komarov 2020-10-30  2142  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2143  
8f609ab22280366 Konstantin Komarov 2020-10-30  2144  	if (frame_vbo >= valid_size) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2145  all_zero:
8f609ab22280366 Konstantin Komarov 2020-10-30  2146  		for (i = 0; i < pages_per_frame; i++)
8f609ab22280366 Konstantin Komarov 2020-10-30  2147  			zero_user_segment(pages[i], 0, PAGE_SIZE);
8f609ab22280366 Konstantin Komarov 2020-10-30  2148  		err = 0;
8f609ab22280366 Konstantin Komarov 2020-10-30  2149  		goto out;
8f609ab22280366 Konstantin Komarov 2020-10-30  2150  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2151  
8f609ab22280366 Konstantin Komarov 2020-10-30  2152  	if (ni->ni_flags & NI_FLAG_COMPRESSED_MASK) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2153  		u32 cmpr = ((ni->ni_flags & NI_FLAG_COMPRESSED_MASK) >> 8) - 1;
8f609ab22280366 Konstantin Komarov 2020-10-30  2154  
8f609ab22280366 Konstantin Komarov 2020-10-30  2155  		switch (cmpr) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2156  		case WOF_COMPRESSION_XPRESS4K:
8f609ab22280366 Konstantin Komarov 2020-10-30  2157  			/* xpress 4k*/
8f609ab22280366 Konstantin Komarov 2020-10-30  2158  			frame_size = 0x1000;
8f609ab22280366 Konstantin Komarov 2020-10-30  2159  			break;
8f609ab22280366 Konstantin Komarov 2020-10-30  2160  		case WOF_COMPRESSION_LZX:
8f609ab22280366 Konstantin Komarov 2020-10-30  2161  			/* lzx 32k*/
8f609ab22280366 Konstantin Komarov 2020-10-30  2162  			frame_size = 0x8000;
8f609ab22280366 Konstantin Komarov 2020-10-30  2163  			break;
8f609ab22280366 Konstantin Komarov 2020-10-30  2164  		case WOF_COMPRESSION_XPRESS8K:
8f609ab22280366 Konstantin Komarov 2020-10-30  2165  			/* xpress 8k*/
8f609ab22280366 Konstantin Komarov 2020-10-30  2166  			frame_size = 0x2000;
8f609ab22280366 Konstantin Komarov 2020-10-30  2167  			break;
8f609ab22280366 Konstantin Komarov 2020-10-30  2168  		case WOF_COMPRESSION_XPRESS16K:
8f609ab22280366 Konstantin Komarov 2020-10-30  2169  			/* xpress 16k*/
8f609ab22280366 Konstantin Komarov 2020-10-30  2170  			frame_size = 0x4000;
8f609ab22280366 Konstantin Komarov 2020-10-30  2171  			break;
8f609ab22280366 Konstantin Komarov 2020-10-30  2172  		}
8f609ab22280366 Konstantin Komarov 2020-10-30  2173  		/* TODO: port lzx/xpress */
8f609ab22280366 Konstantin Komarov 2020-10-30  2174  		err = -EOPNOTSUPP;
8f609ab22280366 Konstantin Komarov 2020-10-30  2175  		goto out;
8f609ab22280366 Konstantin Komarov 2020-10-30  2176  	} else if (is_attr_compressed(attr)) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2177  		/* lznt compression*/
8f609ab22280366 Konstantin Komarov 2020-10-30  2178  		if (sbi->cluster_size > NTFS_LZNT_MAX_CLUSTER) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2179  			err = -EOPNOTSUPP;
8f609ab22280366 Konstantin Komarov 2020-10-30  2180  			goto out;
8f609ab22280366 Konstantin Komarov 2020-10-30  2181  		}
8f609ab22280366 Konstantin Komarov 2020-10-30  2182  
8f609ab22280366 Konstantin Komarov 2020-10-30  2183  		if (attr->nres.c_unit != NTFS_LZNT_CUNIT) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2184  			err = -EOPNOTSUPP;
8f609ab22280366 Konstantin Komarov 2020-10-30  2185  			goto out;
8f609ab22280366 Konstantin Komarov 2020-10-30  2186  		}
8f609ab22280366 Konstantin Komarov 2020-10-30  2187  
8f609ab22280366 Konstantin Komarov 2020-10-30  2188  		down_write(&ni->file.run_lock);
8f609ab22280366 Konstantin Komarov 2020-10-30  2189  		run_truncate_around(&ni->file.run,
8f609ab22280366 Konstantin Komarov 2020-10-30  2190  				    le64_to_cpu(attr->nres.svcn));
8f609ab22280366 Konstantin Komarov 2020-10-30  2191  		frame = frame_vbo >> (sbi->cluster_bits + NTFS_LZNT_CUNIT);
8f609ab22280366 Konstantin Komarov 2020-10-30  2192  		err = attr_is_frame_compressed(ni, attr, frame, &clst_data);
8f609ab22280366 Konstantin Komarov 2020-10-30  2193  		up_write(&ni->file.run_lock);
8f609ab22280366 Konstantin Komarov 2020-10-30  2194  		if (err)
8f609ab22280366 Konstantin Komarov 2020-10-30  2195  			goto out;
8f609ab22280366 Konstantin Komarov 2020-10-30  2196  
8f609ab22280366 Konstantin Komarov 2020-10-30  2197  		if (!clst_data)
8f609ab22280366 Konstantin Komarov 2020-10-30  2198  			goto all_zero;
8f609ab22280366 Konstantin Komarov 2020-10-30  2199  
8f609ab22280366 Konstantin Komarov 2020-10-30  2200  		frame_size = sbi->cluster_size << NTFS_LZNT_CUNIT;
8f609ab22280366 Konstantin Komarov 2020-10-30  2201  		ondisk_size = clst_data << sbi->cluster_bits;
8f609ab22280366 Konstantin Komarov 2020-10-30  2202  
8f609ab22280366 Konstantin Komarov 2020-10-30  2203  		if (clst_data >= NTFS_LZNT_CLUSTERS) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2204  			/* frame is not compressed */
8f609ab22280366 Konstantin Komarov 2020-10-30  2205  			down_read(&ni->file.run_lock);
8f609ab22280366 Konstantin Komarov 2020-10-30  2206  			err = ntfs_bio_pages(sbi, &ni->file.run, pages,
8f609ab22280366 Konstantin Komarov 2020-10-30  2207  					     pages_per_frame, frame_vbo,
8f609ab22280366 Konstantin Komarov 2020-10-30  2208  					     ondisk_size, REQ_OP_READ);
8f609ab22280366 Konstantin Komarov 2020-10-30  2209  			up_read(&ni->file.run_lock);
8f609ab22280366 Konstantin Komarov 2020-10-30  2210  			goto out;
8f609ab22280366 Konstantin Komarov 2020-10-30  2211  		}
8f609ab22280366 Konstantin Komarov 2020-10-30  2212  	} else {
8f609ab22280366 Konstantin Komarov 2020-10-30  2213  		__builtin_unreachable();
8f609ab22280366 Konstantin Komarov 2020-10-30  2214  		err = -EINVAL;
8f609ab22280366 Konstantin Komarov 2020-10-30  2215  		goto out;
8f609ab22280366 Konstantin Komarov 2020-10-30  2216  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2217  
8f609ab22280366 Konstantin Komarov 2020-10-30  2218  	npages_disk = (ondisk_size + PAGE_SIZE - 1) >> PAGE_SHIFT;
8f609ab22280366 Konstantin Komarov 2020-10-30  2219  	pages_disk = ntfs_alloc(npages_disk * sizeof(struct page *), 1);
8f609ab22280366 Konstantin Komarov 2020-10-30  2220  	if (!pages_disk) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2221  		err = -ENOMEM;
8f609ab22280366 Konstantin Komarov 2020-10-30  2222  		goto out;
8f609ab22280366 Konstantin Komarov 2020-10-30  2223  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2224  
8f609ab22280366 Konstantin Komarov 2020-10-30  2225  	for (i = 0; i < npages_disk; i++) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2226  		pg = alloc_page(GFP_KERNEL);
8f609ab22280366 Konstantin Komarov 2020-10-30  2227  		if (!pg) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2228  			err = -ENOMEM;
8f609ab22280366 Konstantin Komarov 2020-10-30  2229  			goto out1;
8f609ab22280366 Konstantin Komarov 2020-10-30  2230  		}
8f609ab22280366 Konstantin Komarov 2020-10-30  2231  		pages_disk[i] = pg;
8f609ab22280366 Konstantin Komarov 2020-10-30  2232  		lock_page(pg);
8f609ab22280366 Konstantin Komarov 2020-10-30  2233  		kmap(pg);
8f609ab22280366 Konstantin Komarov 2020-10-30  2234  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2235  
8f609ab22280366 Konstantin Komarov 2020-10-30  2236  	/* read 'ondisk_size' bytes from disk */
8f609ab22280366 Konstantin Komarov 2020-10-30  2237  	down_read(&ni->file.run_lock);
8f609ab22280366 Konstantin Komarov 2020-10-30  2238  	err = ntfs_bio_pages(sbi, &ni->file.run, pages_disk, npages_disk,
8f609ab22280366 Konstantin Komarov 2020-10-30  2239  			     frame_vbo, ondisk_size, REQ_OP_READ);
8f609ab22280366 Konstantin Komarov 2020-10-30  2240  	up_read(&ni->file.run_lock);
8f609ab22280366 Konstantin Komarov 2020-10-30  2241  	if (err)
8f609ab22280366 Konstantin Komarov 2020-10-30  2242  		goto out1;
8f609ab22280366 Konstantin Komarov 2020-10-30  2243  
8f609ab22280366 Konstantin Komarov 2020-10-30  2244  	/*
8f609ab22280366 Konstantin Komarov 2020-10-30  2245  	 * To simplify decompress algorithm do vmap for source and target pages
8f609ab22280366 Konstantin Komarov 2020-10-30  2246  	 */
8f609ab22280366 Konstantin Komarov 2020-10-30 @2247  	frame_ondisk = vmap(pages_disk, npages_disk, VM_MAP, PAGE_KERNEL);
8f609ab22280366 Konstantin Komarov 2020-10-30  2248  	if (!frame_ondisk) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2249  		err = -ENOMEM;
8f609ab22280366 Konstantin Komarov 2020-10-30  2250  		goto out1;
8f609ab22280366 Konstantin Komarov 2020-10-30  2251  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2252  
8f609ab22280366 Konstantin Komarov 2020-10-30  2253  	for (i = 0; i < pages_per_frame; i++)
8f609ab22280366 Konstantin Komarov 2020-10-30  2254  		kmap(pages[i]);
8f609ab22280366 Konstantin Komarov 2020-10-30  2255  
8f609ab22280366 Konstantin Komarov 2020-10-30  2256  	frame_mem = vmap(pages, pages_per_frame, VM_MAP, PAGE_KERNEL);
8f609ab22280366 Konstantin Komarov 2020-10-30  2257  	if (!frame_mem) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2258  		err = -ENOMEM;
8f609ab22280366 Konstantin Komarov 2020-10-30  2259  		goto out2;
8f609ab22280366 Konstantin Komarov 2020-10-30  2260  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2261  
8f609ab22280366 Konstantin Komarov 2020-10-30  2262  	/* decompress: frame_ondisk -> frame_mem */
8f609ab22280366 Konstantin Komarov 2020-10-30  2263  	unc_size = decompress_lznt(frame_ondisk, ondisk_size, frame_mem,
8f609ab22280366 Konstantin Komarov 2020-10-30  2264  				   frame_size);
8f609ab22280366 Konstantin Komarov 2020-10-30  2265  
8f609ab22280366 Konstantin Komarov 2020-10-30  2266  	if ((ssize_t)unc_size < 0) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2267  		err = unc_size;
8f609ab22280366 Konstantin Komarov 2020-10-30  2268  	} else if (!unc_size || unc_size > frame_size) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2269  		err = -EINVAL;
8f609ab22280366 Konstantin Komarov 2020-10-30  2270  	} else {
8f609ab22280366 Konstantin Komarov 2020-10-30  2271  		if (valid_size < frame_vbo + frame_size) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2272  			size_t ok = valid_size - frame_vbo;
8f609ab22280366 Konstantin Komarov 2020-10-30  2273  
8f609ab22280366 Konstantin Komarov 2020-10-30  2274  			memset(frame_mem + ok, 0, frame_size - ok);
8f609ab22280366 Konstantin Komarov 2020-10-30  2275  		}
8f609ab22280366 Konstantin Komarov 2020-10-30  2276  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2277  
8f609ab22280366 Konstantin Komarov 2020-10-30 @2278  	vunmap(frame_mem);
8f609ab22280366 Konstantin Komarov 2020-10-30  2279  out2:
8f609ab22280366 Konstantin Komarov 2020-10-30  2280  	for (i = 0; i < pages_per_frame; i++) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2281  		pg = pages[i];
8f609ab22280366 Konstantin Komarov 2020-10-30  2282  		kunmap(pg);
8f609ab22280366 Konstantin Komarov 2020-10-30  2283  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2284  	vunmap(frame_ondisk);
8f609ab22280366 Konstantin Komarov 2020-10-30  2285  out1:
8f609ab22280366 Konstantin Komarov 2020-10-30  2286  	for (i = 0; i < npages_disk; i++) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2287  		pg = pages_disk[i];
8f609ab22280366 Konstantin Komarov 2020-10-30  2288  		if (pg) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2289  			kunmap(pg);
8f609ab22280366 Konstantin Komarov 2020-10-30  2290  			unlock_page(pg);
8f609ab22280366 Konstantin Komarov 2020-10-30  2291  			put_page(pg);
8f609ab22280366 Konstantin Komarov 2020-10-30  2292  		}
8f609ab22280366 Konstantin Komarov 2020-10-30  2293  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2294  	ntfs_free(pages_disk);
8f609ab22280366 Konstantin Komarov 2020-10-30  2295  out:
8f609ab22280366 Konstantin Komarov 2020-10-30  2296  	for (i = 0; i < pages_per_frame; i++) {
8f609ab22280366 Konstantin Komarov 2020-10-30  2297  		pg = pages[i];
8f609ab22280366 Konstantin Komarov 2020-10-30  2298  		ClearPageError(pg);
8f609ab22280366 Konstantin Komarov 2020-10-30  2299  		SetPageUptodate(pg);
8f609ab22280366 Konstantin Komarov 2020-10-30  2300  	}
8f609ab22280366 Konstantin Komarov 2020-10-30  2301  
8f609ab22280366 Konstantin Komarov 2020-10-30  2302  	return err;
8f609ab22280366 Konstantin Komarov 2020-10-30  2303  }
8f609ab22280366 Konstantin Komarov 2020-10-30  2304  

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 66534 bytes --]

^ permalink raw reply	[flat|nested] 21+ messages in thread

end of thread, other threads:[~2020-11-03  3:07 UTC | newest]

Thread overview: 21+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-10-30 15:02 [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Konstantin Komarov
2020-10-30 15:02 ` [PATCH v11 01/10] fs/ntfs3: Add headers and misc files Konstantin Komarov
2020-10-30 15:02 ` [PATCH v11 02/10] fs/ntfs3: Add initialization of super block Konstantin Komarov
2020-10-30 15:02 ` [PATCH v11 03/10] fs/ntfs3: Add bitmap Konstantin Komarov
2020-10-30 15:02 ` [PATCH v11 04/10] fs/ntfs3: Add file operations and implementation Konstantin Komarov
2020-10-30 15:02 ` [PATCH v11 05/10] fs/ntfs3: Add attrib operations Konstantin Komarov
2020-10-30 15:02 ` [PATCH v11 06/10] fs/ntfs3: Add compression Konstantin Komarov
2020-10-30 15:02 ` [PATCH v11 07/10] fs/ntfs3: Add NTFS journal Konstantin Komarov
2020-10-30 15:02 ` [PATCH v11 08/10] fs/ntfs3: Add Kconfig, Makefile and doc Konstantin Komarov
2020-10-30 15:02 ` [PATCH v11 09/10] fs/ntfs3: Add NTFS3 in fs/Kconfig and fs/Makefile Konstantin Komarov
2020-10-31  1:23   ` kernel test robot
2020-11-02  8:36   ` [kbuild] " Dan Carpenter
2020-11-03  3:06   ` kernel test robot
2020-10-30 15:02 ` [PATCH v11 10/10] fs/ntfs3: Add MAINTAINERS Konstantin Komarov
2020-10-30 15:24 ` [PATCH v11 00/10] NTFS read-write driver GPL implementation by Paragon Software Pali Rohár
2020-10-30 15:51   ` Konstantin Komarov
2020-10-30 16:24     ` Pali Rohár
2020-10-30 16:41 ` Pali Rohár
2020-10-31  8:51   ` Christoph Hellwig
2020-10-31  2:42 ` Eric Biggers
2020-10-31  2:48   ` Eric Biggers

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).