[PATCH v9 0/4] Prepare for upcoming Casefolding/Encryption patches

[PATCH v9 0/4] Prepare for upcoming Casefolding/Encryption patches

[Daniel Rosenberg]

This lays the ground work for enabling casefolding and encryption at the
same time for ext4 and f2fs. A future set of patches will enable that
functionality. These unify the highly similar dentry_operations that ext4
and f2fs both use for casefolding.

Daniel Rosenberg (4):
  unicode: Add utf8_casefold_hash
  fs: Add standard casefolding support
  f2fs: Use generic casefolding support
  ext4: Use generic casefolding support

 fs/ext4/dir.c           |  64 +------------------------
 fs/ext4/ext4.h          |  12 -----
 fs/ext4/hash.c          |   2 +-
 fs/ext4/namei.c         |  20 ++++----
 fs/ext4/super.c         |  12 ++---
 fs/f2fs/dir.c           |  84 ++++-----------------------------
 fs/f2fs/f2fs.h          |   4 --
 fs/f2fs/super.c         |  10 ++--
 fs/f2fs/sysfs.c         |  10 ++--
 fs/libfs.c              | 101 ++++++++++++++++++++++++++++++++++++++++
 fs/unicode/utf8-core.c  |  23 ++++++++-
 include/linux/f2fs_fs.h |   3 --
 include/linux/fs.h      |  22 +++++++++
 include/linux/unicode.h |   3 ++
 14 files changed, 186 insertions(+), 184 deletions(-)

-- 
2.27.0.111.gc72c7da667-goog

[PATCH v9 1/4] unicode: Add utf8_casefold_hash

[Daniel Rosenberg]

This adds a case insensitive hash function to allow taking the hash
without needing to allocate a casefolded copy of the string.

Signed-off-by: Daniel Rosenberg <drosen@google.com>
---
 fs/unicode/utf8-core.c  | 23 ++++++++++++++++++++++-
 include/linux/unicode.h |  3 +++
 2 files changed, 25 insertions(+), 1 deletion(-)

diff --git a/fs/unicode/utf8-core.c b/fs/unicode/utf8-core.c
index 2a878b739115d..90656b9980720 100644
--- a/fs/unicode/utf8-core.c
+++ b/fs/unicode/utf8-core.c
@@ -6,6 +6,7 @@
 #include <linux/parser.h>
 #include <linux/errno.h>
 #include <linux/unicode.h>
+#include <linux/stringhash.h>
 
 #include "utf8n.h"
 
@@ -122,9 +123,29 @@ int utf8_casefold(const struct unicode_map *um, const struct qstr *str,
 	}
 	return -EINVAL;
 }
-
 EXPORT_SYMBOL(utf8_casefold);
 
+int utf8_casefold_hash(const struct unicode_map *um, const void *salt,
+		       struct qstr *str)
+{
+	const struct utf8data *data = utf8nfdicf(um->version);
+	struct utf8cursor cur;
+	int c;
+	unsigned long hash = init_name_hash(salt);
+
+	if (utf8ncursor(&cur, data, str->name, str->len) < 0)
+		return -EINVAL;
+
+	while ((c = utf8byte(&cur))) {
+		if (c < 0)
+			return c;
+		hash = partial_name_hash((unsigned char)c, hash);
+	}
+	str->hash = end_name_hash(hash);
+	return 0;
+}
+EXPORT_SYMBOL(utf8_casefold_hash);
+
 int utf8_normalize(const struct unicode_map *um, const struct qstr *str,
 		   unsigned char *dest, size_t dlen)
 {
diff --git a/include/linux/unicode.h b/include/linux/unicode.h
index 990aa97d80496..74484d44c7554 100644
--- a/include/linux/unicode.h
+++ b/include/linux/unicode.h
@@ -27,6 +27,9 @@ int utf8_normalize(const struct unicode_map *um, const struct qstr *str,
 int utf8_casefold(const struct unicode_map *um, const struct qstr *str,
 		  unsigned char *dest, size_t dlen);
 
+int utf8_casefold_hash(const struct unicode_map *um, const void *salt,
+		       struct qstr *str);
+
 struct unicode_map *utf8_load(const char *version);
 void utf8_unload(struct unicode_map *um);
 
-- 
2.27.0.111.gc72c7da667-goog

Re: [PATCH v9 1/4] unicode: Add utf8_casefold_hash

[Gabriel Krisman Bertazi]

Daniel Rosenberg <drosen@google.com> writes:

> This adds a case insensitive hash function to allow taking the hash
> without needing to allocate a casefolded copy of the string.
>
> Signed-off-by: Daniel Rosenberg <drosen@google.com>
> ---
>  fs/unicode/utf8-core.c  | 23 ++++++++++++++++++++++-
>  include/linux/unicode.h |  3 +++
>  2 files changed, 25 insertions(+), 1 deletion(-)
>
> diff --git a/fs/unicode/utf8-core.c b/fs/unicode/utf8-core.c
> index 2a878b739115d..90656b9980720 100644
> --- a/fs/unicode/utf8-core.c
> +++ b/fs/unicode/utf8-core.c
> @@ -6,6 +6,7 @@
>  #include <linux/parser.h>
>  #include <linux/errno.h>
>  #include <linux/unicode.h>
> +#include <linux/stringhash.h>
>  
>  #include "utf8n.h"
>  
> @@ -122,9 +123,29 @@ int utf8_casefold(const struct unicode_map *um, const struct qstr *str,
>  	}
>  	return -EINVAL;
>  }
> -
>  EXPORT_SYMBOL(utf8_casefold);
>  
> +int utf8_casefold_hash(const struct unicode_map *um, const void *salt,
> +		       struct qstr *str)
> +{
> +	const struct utf8data *data = utf8nfdicf(um->version);
> +	struct utf8cursor cur;
> +	int c;
> +	unsigned long hash = init_name_hash(salt);
> +
> +	if (utf8ncursor(&cur, data, str->name, str->len) < 0)
> +		return -EINVAL;
> +
> +	while ((c = utf8byte(&cur))) {
> +		if (c < 0)
> +			return c;

Return -EINVAL here to match other unicode functions, since utf8byte
will return -1 on a binary blob, which doesn't make sense for this.

Other than that, looks good to me.

Reviewed-by: Gabriel Krisman Bertazi <krisman@collabora.com>

> +		hash = partial_name_hash((unsigned char)c, hash);
> +	}
> +	str->hash = end_name_hash(hash);
> +	return 0;
> +}
> +EXPORT_SYMBOL(utf8_casefold_hash);
> +
>  int utf8_normalize(const struct unicode_map *um, const struct qstr *str,
>  		   unsigned char *dest, size_t dlen)
>  {
> diff --git a/include/linux/unicode.h b/include/linux/unicode.h
> index 990aa97d80496..74484d44c7554 100644
> --- a/include/linux/unicode.h
> +++ b/include/linux/unicode.h
> @@ -27,6 +27,9 @@ int utf8_normalize(const struct unicode_map *um, const struct qstr *str,
>  int utf8_casefold(const struct unicode_map *um, const struct qstr *str,
>  		  unsigned char *dest, size_t dlen);
>  
> +int utf8_casefold_hash(const struct unicode_map *um, const void *salt,
> +		       struct qstr *str);
> +
>  struct unicode_map *utf8_load(const char *version);
>  void utf8_unload(struct unicode_map *um);

-- 
Gabriel Krisman Bertazi

Re: [PATCH v9 1/4] unicode: Add utf8_casefold_hash

[Eric Biggers]

On Tue, Jun 23, 2020 at 09:33:38PM -0700, Daniel Rosenberg wrote:
> This adds a case insensitive hash function to allow taking the hash
> without needing to allocate a casefolded copy of the string.

It would be helpful to add a few more details in this commit message.
Somewhat along the lines of: ->d_hash() for casefolding currently allocates
memory, it needs to use GFP_ATOMIC due to ->d_hash() being called in rcu-walk
mode, this is unreliable and inefficient, and this patch allows solving that
problem by removing the need to allocate memory.

- Eric

[PATCH v9 2/4] fs: Add standard casefolding support

[Daniel Rosenberg]

This adds general supporting functions for filesystems that use
utf8 casefolding. It provides standard dentry_operations and adds the
necessary structures in struct super_block to allow this standardization.

Ext4 and F2fs will switch to these common implementations.

Signed-off-by: Daniel Rosenberg <drosen@google.com>
---
 fs/libfs.c         | 101 +++++++++++++++++++++++++++++++++++++++++++++
 include/linux/fs.h |  22 ++++++++++
 2 files changed, 123 insertions(+)

diff --git a/fs/libfs.c b/fs/libfs.c
index 4d08edf19c782..f7345a5ed562f 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -20,6 +20,8 @@
 #include <linux/fs_context.h>
 #include <linux/pseudo_fs.h>
 #include <linux/fsnotify.h>
+#include <linux/unicode.h>
+#include <linux/fscrypt.h>
 
 #include <linux/uaccess.h>
 
@@ -1363,3 +1365,102 @@ bool is_empty_dir_inode(struct inode *inode)
 	return (inode->i_fop == &empty_dir_operations) &&
 		(inode->i_op == &empty_dir_inode_operations);
 }
+
+#ifdef CONFIG_UNICODE
+/**
+ * needs_casefold - generic helper to determine if a filename should be casefolded
+ * @dir: Parent directory
+ *
+ * Generic helper for filesystems to use to determine if the name of a dentry
+ * should be casefolded. It does not make sense to casefold the no-key token of
+ * an encrypted filename.
+ *
+ * Return: if names will need casefolding
+ */
+bool needs_casefold(const struct inode *dir)
+{
+	return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding &&
+			(!IS_ENCRYPTED(dir) || fscrypt_has_encryption_key(dir));
+}
+EXPORT_SYMBOL(needs_casefold);
+
+/**
+ * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
+ * @dentry:	dentry whose name we are checking against
+ * @len:	len of name of dentry
+ * @str:	str pointer to name of dentry
+ * @name:	Name to compare against
+ *
+ * Return: 0 if names match, 1 if mismatch, or -ERRNO
+ */
+int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
+			  const char *str, const struct qstr *name)
+{
+	const struct dentry *parent = READ_ONCE(dentry->d_parent);
+	const struct inode *inode = READ_ONCE(parent->d_inode);
+	const struct super_block *sb = dentry->d_sb;
+	const struct unicode_map *um = sb->s_encoding;
+	struct qstr qstr = QSTR_INIT(str, len);
+	char strbuf[DNAME_INLINE_LEN];
+	int ret;
+
+	if (!inode || !needs_casefold(inode))
+		goto fallback;
+	/*
+	 * If the dentry name is stored in-line, then it may be concurrently
+	 * modified by a rename.  If this happens, the VFS will eventually retry
+	 * the lookup, so it doesn't matter what ->d_compare() returns.
+	 * However, it's unsafe to call utf8_strncasecmp() with an unstable
+	 * string.  Therefore, we have to copy the name into a temporary buffer.
+	 */
+	if (len <= DNAME_INLINE_LEN - 1) {
+		memcpy(strbuf, str, len);
+		strbuf[len] = 0;
+		qstr.name = strbuf;
+		/* prevent compiler from optimizing out the temporary buffer */
+		barrier();
+	}
+	ret = utf8_strncasecmp(um, name, &qstr);
+	if (ret >= 0)
+		return ret;
+
+	if (sb_has_enc_strict_mode(sb))
+		return -EINVAL;
+fallback:
+	if (len != name->len)
+		return 1;
+	return !!memcmp(str, name->name, len);
+}
+EXPORT_SYMBOL(generic_ci_d_compare);
+
+/**
+ * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
+ * @dentry:	dentry whose name we are hashing
+ * @str:	qstr of name whose hash we should fill in
+ *
+ * Return: 0 if hash was successful, or -ERRNO
+ */
+int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
+{
+	const struct inode *inode = READ_ONCE(dentry->d_inode);
+	struct super_block *sb = dentry->d_sb;
+	const struct unicode_map *um = sb->s_encoding;
+	int ret = 0;
+
+	if (!inode || !needs_casefold(inode))
+		return 0;
+
+	ret = utf8_casefold_hash(um, dentry, str);
+	if (ret < 0)
+		goto err;
+
+	return 0;
+err:
+	if (sb_has_enc_strict_mode(sb))
+		ret = -EINVAL;
+	else
+		ret = 0;
+	return ret;
+}
+EXPORT_SYMBOL(generic_ci_d_hash);
+#endif
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 3f881a892ea74..261904e06873b 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1392,6 +1392,12 @@ extern int send_sigurg(struct fown_struct *fown);
 #define SB_ACTIVE	(1<<30)
 #define SB_NOUSER	(1<<31)
 
+/* These flags relate to encoding and casefolding */
+#define SB_ENC_STRICT_MODE_FL	(1 << 0)
+
+#define sb_has_enc_strict_mode(sb) \
+	(sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
+
 /*
  *	Umount options
  */
@@ -1461,6 +1467,10 @@ struct super_block {
 #endif
 #ifdef CONFIG_FS_VERITY
 	const struct fsverity_operations *s_vop;
+#endif
+#ifdef CONFIG_UNICODE
+	struct unicode_map *s_encoding;
+	__u16 s_encoding_flags;
 #endif
 	struct hlist_bl_head	s_roots;	/* alternate root dentries for NFS */
 	struct list_head	s_mounts;	/* list of mounts; _not_ for fs use */
@@ -3385,6 +3395,18 @@ extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
 
 extern int generic_check_addressable(unsigned, u64);
 
+#ifdef CONFIG_UNICODE
+extern int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str);
+extern int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
+				const char *str, const struct qstr *name);
+extern bool needs_casefold(const struct inode *dir);
+#else
+static inline bool needs_casefold(const struct inode *dir)
+{
+	return 0;
+}
+#endif
+
 #ifdef CONFIG_MIGRATION
 extern int buffer_migrate_page(struct address_space *,
 				struct page *, struct page *,
-- 
2.27.0.111.gc72c7da667-goog

Re: [PATCH v9 2/4] fs: Add standard casefolding support

[Gabriel Krisman Bertazi]

Daniel Rosenberg <drosen@google.com> writes:

> This adds general supporting functions for filesystems that use
> utf8 casefolding. It provides standard dentry_operations and adds the
> necessary structures in struct super_block to allow this standardization.
>
> Ext4 and F2fs will switch to these common implementations.
>
> Signed-off-by: Daniel Rosenberg <drosen@google.com>
> ---
>  fs/libfs.c         | 101 +++++++++++++++++++++++++++++++++++++++++++++
>  include/linux/fs.h |  22 ++++++++++
>  2 files changed, 123 insertions(+)
>
> diff --git a/fs/libfs.c b/fs/libfs.c
> index 4d08edf19c782..f7345a5ed562f 100644
> --- a/fs/libfs.c
> +++ b/fs/libfs.c
> @@ -20,6 +20,8 @@
>  #include <linux/fs_context.h>
>  #include <linux/pseudo_fs.h>
>  #include <linux/fsnotify.h>
> +#include <linux/unicode.h>
> +#include <linux/fscrypt.h>
>  
>  #include <linux/uaccess.h>
>  
> @@ -1363,3 +1365,102 @@ bool is_empty_dir_inode(struct inode *inode)
>  	return (inode->i_fop == &empty_dir_operations) &&
>  		(inode->i_op == &empty_dir_inode_operations);
>  }
> +
> +#ifdef CONFIG_UNICODE
> +/**
> + * needs_casefold - generic helper to determine if a filename should be casefolded
> + * @dir: Parent directory
> + *
> + * Generic helper for filesystems to use to determine if the name of a dentry
> + * should be casefolded. It does not make sense to casefold the no-key token of
> + * an encrypted filename.
> + *
> + * Return: if names will need casefolding
> + */
> +bool needs_casefold(const struct inode *dir)
> +{
> +	return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding &&
> +			(!IS_ENCRYPTED(dir) || fscrypt_has_encryption_key(dir));
> +}
> +EXPORT_SYMBOL(needs_casefold);
> +
> +/**
> + * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
> + * @dentry:	dentry whose name we are checking against
> + * @len:	len of name of dentry
> + * @str:	str pointer to name of dentry
> + * @name:	Name to compare against
> + *
> + * Return: 0 if names match, 1 if mismatch, or -ERRNO
> + */
> +int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
> +			  const char *str, const struct qstr *name)
> +{
> +	const struct dentry *parent = READ_ONCE(dentry->d_parent);
> +	const struct inode *inode = READ_ONCE(parent->d_inode);
> +	const struct super_block *sb = dentry->d_sb;
> +	const struct unicode_map *um = sb->s_encoding;
> +	struct qstr qstr = QSTR_INIT(str, len);
> +	char strbuf[DNAME_INLINE_LEN];
> +	int ret;
> +
> +	if (!inode || !needs_casefold(inode))
> +		goto fallback;
> +	/*
> +	 * If the dentry name is stored in-line, then it may be concurrently
> +	 * modified by a rename.  If this happens, the VFS will eventually retry
> +	 * the lookup, so it doesn't matter what ->d_compare() returns.
> +	 * However, it's unsafe to call utf8_strncasecmp() with an unstable
> +	 * string.  Therefore, we have to copy the name into a temporary buffer.
> +	 */
> +	if (len <= DNAME_INLINE_LEN - 1) {
> +		memcpy(strbuf, str, len);
> +		strbuf[len] = 0;
> +		qstr.name = strbuf;
> +		/* prevent compiler from optimizing out the temporary buffer */
> +		barrier();
> +	}
> +	ret = utf8_strncasecmp(um, name, &qstr);
> +	if (ret >= 0)
> +		return ret;
> +
> +	if (sb_has_enc_strict_mode(sb))
> +		return -EINVAL;
> +fallback:
> +	if (len != name->len)
> +		return 1;
> +	return !!memcmp(str, name->name, len);
> +}
> +EXPORT_SYMBOL(generic_ci_d_compare);
> +
> +/**
> + * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
> + * @dentry:	dentry whose name we are hashing
> + * @str:	qstr of name whose hash we should fill in
> + *
> + * Return: 0 if hash was successful, or -ERRNO
> + */
> +int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
> +{
> +	const struct inode *inode = READ_ONCE(dentry->d_inode);
> +	struct super_block *sb = dentry->d_sb;
> +	const struct unicode_map *um = sb->s_encoding;
> +	int ret = 0;
> +
> +	if (!inode || !needs_casefold(inode))
> +		return 0;
> +
> +	ret = utf8_casefold_hash(um, dentry, str);
> +	if (ret < 0)
> +		goto err;
> +
> +	return 0;
> +err:
> +	if (sb_has_enc_strict_mode(sb))
> +		ret = -EINVAL;
> +	else
> +		ret = 0;
> +	return ret;
> +}

Maybe drop the err label and simplify:

  ret = utf8_casefold_hash(um, dentry, str);
  if (ret < 0 && sb_has_enc_strict_mode(sb))
     return -EINVAL;
  return 0;


> +EXPORT_SYMBOL(generic_ci_d_hash);
> +#endif
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 3f881a892ea74..261904e06873b 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -1392,6 +1392,12 @@ extern int send_sigurg(struct fown_struct *fown);
>  #define SB_ACTIVE	(1<<30)
>  #define SB_NOUSER	(1<<31)
>  
> +/* These flags relate to encoding and casefolding */
> +#define SB_ENC_STRICT_MODE_FL	(1 << 0)
> +
> +#define sb_has_enc_strict_mode(sb) \
> +	(sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
> +
>  /*
>   *	Umount options
>   */
> @@ -1461,6 +1467,10 @@ struct super_block {
>  #endif
>  #ifdef CONFIG_FS_VERITY
>  	const struct fsverity_operations *s_vop;
> +#endif
> +#ifdef CONFIG_UNICODE
> +	struct unicode_map *s_encoding;
> +	__u16 s_encoding_flags;
>  #endif
>  	struct hlist_bl_head	s_roots;	/* alternate root dentries for NFS */
>  	struct list_head	s_mounts;	/* list of mounts; _not_ for fs use */
> @@ -3385,6 +3395,18 @@ extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
>  
>  extern int generic_check_addressable(unsigned, u64);
>  
> +#ifdef CONFIG_UNICODE
> +extern int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str);
> +extern int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
> +				const char *str, const struct qstr *name);
> +extern bool needs_casefold(const struct inode *dir);
> +#else
> +static inline bool needs_casefold(const struct inode *dir)
> +{
> +	return 0;
> +}
> +#endif
> +
>  #ifdef CONFIG_MIGRATION
>  extern int buffer_migrate_page(struct address_space *,
>  				struct page *, struct page *,

-- 
Gabriel Krisman Bertazi

Re: [PATCH v9 2/4] fs: Add standard casefolding support

[Eric Biggers]

On Tue, Jun 23, 2020 at 09:33:39PM -0700, Daniel Rosenberg wrote:
> This adds general supporting functions for filesystems that use
> utf8 casefolding. It provides standard dentry_operations and adds the
> necessary structures in struct super_block to allow this standardization.
> 
> Ext4 and F2fs will switch to these common implementations.

It would be helpful to explicitly call out anything that's "new" in this commit,
i.e. anything that isn't simply moving code into the libfs with no behavior
changes.  There's the change of ->d_hash() to use utf8_casefold_hash() instead
of allocating memory; that's not present in the ext4 and f2fs versions.

There's also the change of needs_casefold() to be aware of encrypt+casefold.
(Maybe that small change would better belong in a later patchset that actually
introduces encrypt+casefold support?)

Anything else?

> +/**
> + * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
> + * @dentry:	dentry whose name we are hashing
> + * @str:	qstr of name whose hash we should fill in
> + *
> + * Return: 0 if hash was successful, or -ERRNO
> + */

It also returns 0 if the hashing was not done because it wants to fallback to
the standard hashing.

> +static inline bool needs_casefold(const struct inode *dir)
> +{
> +	return 0;
> +}

The return type is bool, so it should 'return false', not 'return 0'.

- Eric

Re: [PATCH v9 2/4] fs: Add standard casefolding support

[Eric Biggers]

On Tue, Jun 23, 2020 at 09:33:39PM -0700, Daniel Rosenberg wrote:
> This adds general supporting functions for filesystems that use
> utf8 casefolding. It provides standard dentry_operations and adds the
> necessary structures in struct super_block to allow this standardization.
> 
> Ext4 and F2fs will switch to these common implementations.
> 
> Signed-off-by: Daniel Rosenberg <drosen@google.com>
> ---
>  fs/libfs.c         | 101 +++++++++++++++++++++++++++++++++++++++++++++
>  include/linux/fs.h |  22 ++++++++++
>  2 files changed, 123 insertions(+)
> 
> diff --git a/fs/libfs.c b/fs/libfs.c
> index 4d08edf19c782..f7345a5ed562f 100644
> --- a/fs/libfs.c
> +++ b/fs/libfs.c
> @@ -20,6 +20,8 @@
>  #include <linux/fs_context.h>
>  #include <linux/pseudo_fs.h>
>  #include <linux/fsnotify.h>
> +#include <linux/unicode.h>
> +#include <linux/fscrypt.h>
>  
>  #include <linux/uaccess.h>
>  
> @@ -1363,3 +1365,102 @@ bool is_empty_dir_inode(struct inode *inode)
>  	return (inode->i_fop == &empty_dir_operations) &&
>  		(inode->i_op == &empty_dir_inode_operations);
>  }
> +
> +#ifdef CONFIG_UNICODE
> +/**
> + * needs_casefold - generic helper to determine if a filename should be casefolded
> + * @dir: Parent directory
> + *
> + * Generic helper for filesystems to use to determine if the name of a dentry
> + * should be casefolded. It does not make sense to casefold the no-key token of
> + * an encrypted filename.
> + *
> + * Return: if names will need casefolding
> + */
> +bool needs_casefold(const struct inode *dir)
> +{
> +	return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding &&
> +			(!IS_ENCRYPTED(dir) || fscrypt_has_encryption_key(dir));
> +}
> +EXPORT_SYMBOL(needs_casefold);

Note that the '!IS_ENCRYPTED(dir) || fscrypt_has_encryption_key(dir)' check can
be racy, because a process can be looking up a no-key token in a directory while
concurrently another process initializes the directory's ->i_crypt_info, causing
fscrypt_has_encryption_key(dir) to suddenly start returning true.

In my rework of filename handling in f2fs, I actually ended up removing all
calls to needs_casefold(), thus avoiding this race.  f2fs now decides whether
the name is going to need casefolding early on, in __f2fs_setup_filename(),
where it knows in a race-free way whether the filename is a no-key token or not.

Perhaps ext4 should work the same way?  It did look like there would be some
extra complexity due to how the ext4 directory hashing works in comparison to
f2fs's, but I haven't had a chance to properly investigate it.

- Eric

Re: [PATCH v9 2/4] fs: Add standard casefolding support

[Daniel Rosenberg]

On Tue, Jun 23, 2020 at 10:57 PM Eric Biggers <ebiggers@kernel.org> wrote:
>
> Note that the '!IS_ENCRYPTED(dir) || fscrypt_has_encryption_key(dir)' check can
> be racy, because a process can be looking up a no-key token in a directory while
> concurrently another process initializes the directory's ->i_crypt_info, causing
> fscrypt_has_encryption_key(dir) to suddenly start returning true.
>
> In my rework of filename handling in f2fs, I actually ended up removing all
> calls to needs_casefold(), thus avoiding this race.  f2fs now decides whether
> the name is going to need casefolding early on, in __f2fs_setup_filename(),
> where it knows in a race-free way whether the filename is a no-key token or not.
>
> Perhaps ext4 should work the same way?  It did look like there would be some
> extra complexity due to how the ext4 directory hashing works in comparison to
> f2fs's, but I haven't had a chance to properly investigate it.
>
> - Eric

Hm. I think I should be able to just check for DCACHE_ENCRYPTED_NAME
in the dentry here, right? I'm just trying to avoid casefolding the
no-key token, and that flag should indicate that.
I'll see if I can rework the ext4 patches to not need needs_casefold
as well, since then there'd be no need to export it.
-Daniel

Re: [PATCH v9 2/4] fs: Add standard casefolding support

[Eric Biggers]

On Thu, Jul 02, 2020 at 06:01:37PM -0700, Daniel Rosenberg wrote:
> On Tue, Jun 23, 2020 at 10:57 PM Eric Biggers <ebiggers@kernel.org> wrote:
> >
> > Note that the '!IS_ENCRYPTED(dir) || fscrypt_has_encryption_key(dir)' check can
> > be racy, because a process can be looking up a no-key token in a directory while
> > concurrently another process initializes the directory's ->i_crypt_info, causing
> > fscrypt_has_encryption_key(dir) to suddenly start returning true.
> >
> > In my rework of filename handling in f2fs, I actually ended up removing all
> > calls to needs_casefold(), thus avoiding this race.  f2fs now decides whether
> > the name is going to need casefolding early on, in __f2fs_setup_filename(),
> > where it knows in a race-free way whether the filename is a no-key token or not.
> >
> > Perhaps ext4 should work the same way?  It did look like there would be some
> > extra complexity due to how the ext4 directory hashing works in comparison to
> > f2fs's, but I haven't had a chance to properly investigate it.
> >
> > - Eric
> 
> Hm. I think I should be able to just check for DCACHE_ENCRYPTED_NAME
> in the dentry here, right? I'm just trying to avoid casefolding the
> no-key token, and that flag should indicate that.

Ideally yes, but currently the 'struct dentry' isn't always available.  See how
fscrypt_setup_filename(), f2fs_setup_filename(), f2fs_find_entry(),
ext4_find_entry(), etc. take a 'struct qstr', not a 'struct dentry'.

At some point we should fix that by passing down the dentry whenever it's
available, so that we reliably know whether the name is a no-key name or not.

So even my new f2fs code is still racy.  But it at least handles each filename
in a consistent way within each directory operation.  In comparison, your
proposed ext4 code can treat a filename as a no-key name while matching one
dir_entry and then as a regular filename while matching the next.  I think the
f2fs way is more on the right track, both correctness-wise and efficiency-wise.

- Eric

[PATCH v9 3/4] f2fs: Use generic casefolding support

[Daniel Rosenberg]

This switches f2fs over to the generic support provided in
commit 5f829feca774 ("fs: Add standard casefolding support")

Signed-off-by: Daniel Rosenberg <drosen@google.com>
---
 fs/f2fs/dir.c           | 84 +++++------------------------------------
 fs/f2fs/f2fs.h          |  4 --
 fs/f2fs/super.c         | 10 ++---
 fs/f2fs/sysfs.c         | 10 +++--
 include/linux/f2fs_fs.h |  3 --
 5 files changed, 21 insertions(+), 90 deletions(-)

diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
index d35976785e8c5..b59c2673daa09 100644
--- a/fs/f2fs/dir.c
+++ b/fs/f2fs/dir.c
@@ -75,21 +75,22 @@ int f2fs_init_casefolded_name(const struct inode *dir,
 			      struct f2fs_filename *fname)
 {
 #ifdef CONFIG_UNICODE
-	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
+	struct super_block *sb = dir->i_sb;
+	struct f2fs_sb_info *sbi = F2FS_SB(sb);
 
 	if (IS_CASEFOLDED(dir)) {
 		fname->cf_name.name = f2fs_kmalloc(sbi, F2FS_NAME_LEN,
 						   GFP_NOFS);
 		if (!fname->cf_name.name)
 			return -ENOMEM;
-		fname->cf_name.len = utf8_casefold(sbi->s_encoding,
+		fname->cf_name.len = utf8_casefold(sb->s_encoding,
 						   fname->usr_fname,
 						   fname->cf_name.name,
 						   F2FS_NAME_LEN);
 		if ((int)fname->cf_name.len <= 0) {
 			kfree(fname->cf_name.name);
 			fname->cf_name.name = NULL;
-			if (f2fs_has_strict_mode(sbi))
+			if (sb_has_enc_strict_mode(sb))
 				return -EINVAL;
 			/* fall back to treating name as opaque byte sequence */
 		}
@@ -215,8 +216,9 @@ static struct f2fs_dir_entry *find_in_block(struct inode *dir,
 static bool f2fs_match_ci_name(const struct inode *dir, const struct qstr *name,
 			       const u8 *de_name, u32 de_name_len)
 {
-	const struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
-	const struct unicode_map *um = sbi->s_encoding;
+	const struct super_block *sb = dir->i_sb;
+	const struct f2fs_sb_info *sbi = F2FS_SB(sb);
+	const struct unicode_map *um = sb->s_encoding;
 	struct qstr entry = QSTR_INIT(de_name, de_name_len);
 	int res;
 
@@ -226,7 +228,7 @@ static bool f2fs_match_ci_name(const struct inode *dir, const struct qstr *name,
 		 * In strict mode, ignore invalid names.  In non-strict mode,
 		 * fall back to treating them as opaque byte sequences.
 		 */
-		if (f2fs_has_strict_mode(sbi) || name->len != entry.len)
+		if (sb_has_enc_strict_mode(sb) || name->len != entry.len)
 			return false;
 		return !memcmp(name->name, entry.name, name->len);
 	}
@@ -1107,75 +1109,9 @@ const struct file_operations f2fs_dir_operations = {
 };
 
 #ifdef CONFIG_UNICODE
-static int f2fs_d_compare(const struct dentry *dentry, unsigned int len,
-			  const char *str, const struct qstr *name)
-{
-	const struct dentry *parent = READ_ONCE(dentry->d_parent);
-	const struct inode *dir = READ_ONCE(parent->d_inode);
-	const struct f2fs_sb_info *sbi = F2FS_SB(dentry->d_sb);
-	struct qstr entry = QSTR_INIT(str, len);
-	char strbuf[DNAME_INLINE_LEN];
-	int res;
-
-	if (!dir || !IS_CASEFOLDED(dir))
-		goto fallback;
-
-	/*
-	 * If the dentry name is stored in-line, then it may be concurrently
-	 * modified by a rename.  If this happens, the VFS will eventually retry
-	 * the lookup, so it doesn't matter what ->d_compare() returns.
-	 * However, it's unsafe to call utf8_strncasecmp() with an unstable
-	 * string.  Therefore, we have to copy the name into a temporary buffer.
-	 */
-	if (len <= DNAME_INLINE_LEN - 1) {
-		memcpy(strbuf, str, len);
-		strbuf[len] = 0;
-		entry.name = strbuf;
-		/* prevent compiler from optimizing out the temporary buffer */
-		barrier();
-	}
-
-	res = utf8_strncasecmp(sbi->s_encoding, name, &entry);
-	if (res >= 0)
-		return res;
-
-	if (f2fs_has_strict_mode(sbi))
-		return -EINVAL;
-fallback:
-	if (len != name->len)
-		return 1;
-	return !!memcmp(str, name->name, len);
-}
-
-static int f2fs_d_hash(const struct dentry *dentry, struct qstr *str)
-{
-	struct f2fs_sb_info *sbi = F2FS_SB(dentry->d_sb);
-	const struct unicode_map *um = sbi->s_encoding;
-	const struct inode *inode = READ_ONCE(dentry->d_inode);
-	unsigned char *norm;
-	int len, ret = 0;
-
-	if (!inode || !IS_CASEFOLDED(inode))
-		return 0;
-
-	norm = f2fs_kmalloc(sbi, PATH_MAX, GFP_ATOMIC);
-	if (!norm)
-		return -ENOMEM;
-
-	len = utf8_casefold(um, str, norm, PATH_MAX);
-	if (len < 0) {
-		if (f2fs_has_strict_mode(sbi))
-			ret = -EINVAL;
-		goto out;
-	}
-	str->hash = full_name_hash(dentry, norm, len);
-out:
-	kvfree(norm);
-	return ret;
-}
 
 const struct dentry_operations f2fs_dentry_ops = {
-	.d_hash = f2fs_d_hash,
-	.d_compare = f2fs_d_compare,
+	.d_hash = generic_ci_d_hash,
+	.d_compare = generic_ci_d_compare,
 };
 #endif
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index b35a50f4953c5..d11ffe26bfded 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -1402,10 +1402,6 @@ struct f2fs_sb_info {
 	int valid_super_block;			/* valid super block no */
 	unsigned long s_flag;				/* flags for sbi */
 	struct mutex writepages;		/* mutex for writepages() */
-#ifdef CONFIG_UNICODE
-	struct unicode_map *s_encoding;
-	__u16 s_encoding_flags;
-#endif
 
 #ifdef CONFIG_BLK_DEV_ZONED
 	unsigned int blocks_per_blkz;		/* F2FS blocks per zone */
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 20e56b0fa46a9..cca7a83ffa089 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1257,7 +1257,7 @@ static void f2fs_put_super(struct super_block *sb)
 	for (i = 0; i < NR_PAGE_TYPE; i++)
 		kvfree(sbi->write_io[i]);
 #ifdef CONFIG_UNICODE
-	utf8_unload(sbi->s_encoding);
+	utf8_unload(sb->s_encoding);
 #endif
 	kvfree(sbi);
 }
@@ -3278,7 +3278,7 @@ static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
 {
 #ifdef CONFIG_UNICODE
-	if (f2fs_sb_has_casefold(sbi) && !sbi->s_encoding) {
+	if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
 		const struct f2fs_sb_encodings *encoding_info;
 		struct unicode_map *encoding;
 		__u16 encoding_flags;
@@ -3309,8 +3309,8 @@ static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
 			 "%s-%s with flags 0x%hx", encoding_info->name,
 			 encoding_info->version?:"\b", encoding_flags);
 
-		sbi->s_encoding = encoding;
-		sbi->s_encoding_flags = encoding_flags;
+		sbi->sb->s_encoding = encoding;
+		sbi->sb->s_encoding_flags = encoding_flags;
 		sbi->sb->s_d_op = &f2fs_dentry_ops;
 	}
 #else
@@ -3806,7 +3806,7 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 		kvfree(sbi->write_io[i]);
 
 #ifdef CONFIG_UNICODE
-	utf8_unload(sbi->s_encoding);
+	utf8_unload(sb->s_encoding);
 #endif
 free_options:
 #ifdef CONFIG_QUOTA
diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c
index e877c59b9fdb4..8bee99ab3978e 100644
--- a/fs/f2fs/sysfs.c
+++ b/fs/f2fs/sysfs.c
@@ -176,12 +176,14 @@ static ssize_t encoding_show(struct f2fs_attr *a,
 		struct f2fs_sb_info *sbi, char *buf)
 {
 #ifdef CONFIG_UNICODE
+	struct super_block *sb = sbi->sb;
+
 	if (f2fs_sb_has_casefold(sbi))
 		return snprintf(buf, PAGE_SIZE, "%s (%d.%d.%d)\n",
-			sbi->s_encoding->charset,
-			(sbi->s_encoding->version >> 16) & 0xff,
-			(sbi->s_encoding->version >> 8) & 0xff,
-			sbi->s_encoding->version & 0xff);
+			sb->s_encoding->charset,
+			(sb->s_encoding->version >> 16) & 0xff,
+			(sb->s_encoding->version >> 8) & 0xff,
+			sb->s_encoding->version & 0xff);
 #endif
 	return sprintf(buf, "(none)");
 }
diff --git a/include/linux/f2fs_fs.h b/include/linux/f2fs_fs.h
index 3c383ddd92ddd..a5dbb57a687fb 100644
--- a/include/linux/f2fs_fs.h
+++ b/include/linux/f2fs_fs.h
@@ -38,9 +38,6 @@
 #define F2FS_MAX_QUOTAS		3
 
 #define F2FS_ENC_UTF8_12_1	1
-#define F2FS_ENC_STRICT_MODE_FL	(1 << 0)
-#define f2fs_has_strict_mode(sbi) \
-	(sbi->s_encoding_flags & F2FS_ENC_STRICT_MODE_FL)
 
 #define F2FS_IO_SIZE(sbi)	(1 << F2FS_OPTION(sbi).write_io_size_bits) /* Blocks */
 #define F2FS_IO_SIZE_KB(sbi)	(1 << (F2FS_OPTION(sbi).write_io_size_bits + 2)) /* KB */
-- 
2.27.0.111.gc72c7da667-goog

[PATCH v9 4/4] ext4: Use generic casefolding support

[Daniel Rosenberg]

This switches ext4 over to the generic support provided in
commit 5f829feca774 ("fs: Add standard casefolding support")

Signed-off-by: Daniel Rosenberg <drosen@google.com>
---
 fs/ext4/dir.c   | 64 ++-----------------------------------------------
 fs/ext4/ext4.h  | 12 ----------
 fs/ext4/hash.c  |  2 +-
 fs/ext4/namei.c | 20 +++++++---------
 fs/ext4/super.c | 12 +++++-----
 5 files changed, 17 insertions(+), 93 deletions(-)

diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c
index 1d82336b1cd45..b437120f0b3f5 100644
--- a/fs/ext4/dir.c
+++ b/fs/ext4/dir.c
@@ -669,68 +669,8 @@ const struct file_operations ext4_dir_operations = {
 };
 
 #ifdef CONFIG_UNICODE
-static int ext4_d_compare(const struct dentry *dentry, unsigned int len,
-			  const char *str, const struct qstr *name)
-{
-	struct qstr qstr = {.name = str, .len = len };
-	const struct dentry *parent = READ_ONCE(dentry->d_parent);
-	const struct inode *inode = READ_ONCE(parent->d_inode);
-	char strbuf[DNAME_INLINE_LEN];
-
-	if (!inode || !IS_CASEFOLDED(inode) ||
-	    !EXT4_SB(inode->i_sb)->s_encoding) {
-		if (len != name->len)
-			return -1;
-		return memcmp(str, name->name, len);
-	}
-
-	/*
-	 * If the dentry name is stored in-line, then it may be concurrently
-	 * modified by a rename.  If this happens, the VFS will eventually retry
-	 * the lookup, so it doesn't matter what ->d_compare() returns.
-	 * However, it's unsafe to call utf8_strncasecmp() with an unstable
-	 * string.  Therefore, we have to copy the name into a temporary buffer.
-	 */
-	if (len <= DNAME_INLINE_LEN - 1) {
-		memcpy(strbuf, str, len);
-		strbuf[len] = 0;
-		qstr.name = strbuf;
-		/* prevent compiler from optimizing out the temporary buffer */
-		barrier();
-	}
-
-	return ext4_ci_compare(inode, name, &qstr, false);
-}
-
-static int ext4_d_hash(const struct dentry *dentry, struct qstr *str)
-{
-	const struct ext4_sb_info *sbi = EXT4_SB(dentry->d_sb);
-	const struct unicode_map *um = sbi->s_encoding;
-	const struct inode *inode = READ_ONCE(dentry->d_inode);
-	unsigned char *norm;
-	int len, ret = 0;
-
-	if (!inode || !IS_CASEFOLDED(inode) || !um)
-		return 0;
-
-	norm = kmalloc(PATH_MAX, GFP_ATOMIC);
-	if (!norm)
-		return -ENOMEM;
-
-	len = utf8_casefold(um, str, norm, PATH_MAX);
-	if (len < 0) {
-		if (ext4_has_strict_mode(sbi))
-			ret = -EINVAL;
-		goto out;
-	}
-	str->hash = full_name_hash(dentry, norm, len);
-out:
-	kfree(norm);
-	return ret;
-}
-
 const struct dentry_operations ext4_dentry_ops = {
-	.d_hash = ext4_d_hash,
-	.d_compare = ext4_d_compare,
+	.d_hash = generic_ci_d_hash,
+	.d_compare = generic_ci_d_compare,
 };
 #endif
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 42f5060f3cdf1..5cd8be24a4fd9 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1393,14 +1393,6 @@ struct ext4_super_block {
 
 #define EXT4_ENC_UTF8_12_1	1
 
-/*
- * Flags for ext4_sb_info.s_encoding_flags.
- */
-#define EXT4_ENC_STRICT_MODE_FL	(1 << 0)
-
-#define ext4_has_strict_mode(sbi) \
-	(sbi->s_encoding_flags & EXT4_ENC_STRICT_MODE_FL)
-
 /*
  * fourth extended-fs super-block data in memory
  */
@@ -1450,10 +1442,6 @@ struct ext4_sb_info {
 	struct kobject s_kobj;
 	struct completion s_kobj_unregister;
 	struct super_block *s_sb;
-#ifdef CONFIG_UNICODE
-	struct unicode_map *s_encoding;
-	__u16 s_encoding_flags;
-#endif
 
 	/* Journaling */
 	struct journal_s *s_journal;
diff --git a/fs/ext4/hash.c b/fs/ext4/hash.c
index 3e133793a5a34..143b0073b3f46 100644
--- a/fs/ext4/hash.c
+++ b/fs/ext4/hash.c
@@ -275,7 +275,7 @@ int ext4fs_dirhash(const struct inode *dir, const char *name, int len,
 		   struct dx_hash_info *hinfo)
 {
 #ifdef CONFIG_UNICODE
-	const struct unicode_map *um = EXT4_SB(dir->i_sb)->s_encoding;
+	const struct unicode_map *um = dir->i_sb->s_encoding;
 	int r, dlen;
 	unsigned char *buff;
 	struct qstr qstr = {.name = name, .len = len };
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index 56738b538ddf4..7e9fb77fd2cc7 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -1286,8 +1286,8 @@ static void dx_insert_block(struct dx_frame *frame, u32 hash, ext4_lblk_t block)
 int ext4_ci_compare(const struct inode *parent, const struct qstr *name,
 		    const struct qstr *entry, bool quick)
 {
-	const struct ext4_sb_info *sbi = EXT4_SB(parent->i_sb);
-	const struct unicode_map *um = sbi->s_encoding;
+	const struct super_block *sb = parent->i_sb;
+	const struct unicode_map *um = sb->s_encoding;
 	int ret;
 
 	if (quick)
@@ -1299,7 +1299,7 @@ int ext4_ci_compare(const struct inode *parent, const struct qstr *name,
 		/* Handle invalid character sequence as either an error
 		 * or as an opaque byte sequence.
 		 */
-		if (ext4_has_strict_mode(sbi))
+		if (sb_has_enc_strict_mode(sb))
 			return -EINVAL;
 
 		if (name->len != entry->len)
@@ -1316,7 +1316,7 @@ void ext4_fname_setup_ci_filename(struct inode *dir, const struct qstr *iname,
 {
 	int len;
 
-	if (!IS_CASEFOLDED(dir) || !EXT4_SB(dir->i_sb)->s_encoding) {
+	if (!needs_casefold(dir)) {
 		cf_name->name = NULL;
 		return;
 	}
@@ -1325,7 +1325,7 @@ void ext4_fname_setup_ci_filename(struct inode *dir, const struct qstr *iname,
 	if (!cf_name->name)
 		return;
 
-	len = utf8_casefold(EXT4_SB(dir->i_sb)->s_encoding,
+	len = utf8_casefold(dir->i_sb->s_encoding,
 			    iname, cf_name->name,
 			    EXT4_NAME_LEN);
 	if (len <= 0) {
@@ -1362,7 +1362,7 @@ static inline bool ext4_match(const struct inode *parent,
 #endif
 
 #ifdef CONFIG_UNICODE
-	if (EXT4_SB(parent->i_sb)->s_encoding && IS_CASEFOLDED(parent)) {
+	if (needs_casefold(parent)) {
 		if (fname->cf_name.name) {
 			struct qstr cf = {.name = fname->cf_name.name,
 					  .len = fname->cf_name.len};
@@ -2171,9 +2171,6 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 	struct buffer_head *bh = NULL;
 	struct ext4_dir_entry_2 *de;
 	struct super_block *sb;
-#ifdef CONFIG_UNICODE
-	struct ext4_sb_info *sbi;
-#endif
 	struct ext4_filename fname;
 	int	retval;
 	int	dx_fallback=0;
@@ -2190,9 +2187,8 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 		return -EINVAL;
 
 #ifdef CONFIG_UNICODE
-	sbi = EXT4_SB(sb);
-	if (ext4_has_strict_mode(sbi) && IS_CASEFOLDED(dir) &&
-	    sbi->s_encoding && utf8_validate(sbi->s_encoding, &dentry->d_name))
+	if (sb_has_enc_strict_mode(sb) && IS_CASEFOLDED(dir) &&
+	    sb->s_encoding && utf8_validate(sb->s_encoding, &dentry->d_name))
 		return -EINVAL;
 #endif
 
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 330957ed1f05c..d097771a374f8 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -1102,7 +1102,7 @@ static void ext4_put_super(struct super_block *sb)
 	fs_put_dax(sbi->s_daxdev);
 	fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
 #ifdef CONFIG_UNICODE
-	utf8_unload(sbi->s_encoding);
+	utf8_unload(sb->s_encoding);
 #endif
 	kfree(sbi);
 }
@@ -4035,7 +4035,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 		goto failed_mount;
 
 #ifdef CONFIG_UNICODE
-	if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
+	if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
 		const struct ext4_sb_encodings *encoding_info;
 		struct unicode_map *encoding;
 		__u16 encoding_flags;
@@ -4066,8 +4066,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 			 "%s-%s with flags 0x%hx", encoding_info->name,
 			 encoding_info->version?:"\b", encoding_flags);
 
-		sbi->s_encoding = encoding;
-		sbi->s_encoding_flags = encoding_flags;
+		sb->s_encoding = encoding;
+		sb->s_encoding_flags = encoding_flags;
 	}
 #endif
 
@@ -4678,7 +4678,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	}
 
 #ifdef CONFIG_UNICODE
-	if (sbi->s_encoding)
+	if (sb->s_encoding)
 		sb->s_d_op = &ext4_dentry_ops;
 #endif
 
@@ -4873,7 +4873,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 		crypto_free_shash(sbi->s_chksum_driver);
 
 #ifdef CONFIG_UNICODE
-	utf8_unload(sbi->s_encoding);
+	utf8_unload(sb->s_encoding);
 #endif
 
 #ifdef CONFIG_QUOTA
-- 
2.27.0.111.gc72c7da667-goog

Re: [PATCH v9 4/4] ext4: Use generic casefolding support

[Gabriel Krisman Bertazi]

Daniel Rosenberg <drosen@google.com> writes:

> -
>  const struct dentry_operations ext4_dentry_ops = {
> -	.d_hash = ext4_d_hash,
> -	.d_compare = ext4_d_compare,
> +	.d_hash = generic_ci_d_hash,
> +	.d_compare = generic_ci_d_compare,
>  };
>  #endif

Can you make the structure generic since it is the same for f2fs and
ext4, which let you drop the code guards?  Unless that becomes a problem for
d_revalidate with fscrypt, it is fine like this.

>  #ifdef CONFIG_UNICODE
> -	sbi = EXT4_SB(sb);
> -	if (ext4_has_strict_mode(sbi) && IS_CASEFOLDED(dir) &&
> -	    sbi->s_encoding && utf8_validate(sbi->s_encoding, &dentry->d_name))
> +	if (sb_has_enc_strict_mode(sb) && IS_CASEFOLDED(dir) &&

I keep reading the 'enc' in sb_has_enc_strict_mode() as 'encryption'.  What do
you think about renaming it to sb_has_strict_encoding()?

These comments apply equally to patches 3 and 4.  Other than that,

Reviewed-by: Gabriel Krisman Bertazi <krisman@collabora.com>

-- 
Gabriel Krisman Bertazi

Re: [PATCH v9 4/4] ext4: Use generic casefolding support

[Daniel Rosenberg]

On Tue, Jun 23, 2020 at 10:43 PM Gabriel Krisman Bertazi
<krisman@collabora.com> wrote:
>
> Daniel Rosenberg <drosen@google.com> writes:
>
> > -
> >  const struct dentry_operations ext4_dentry_ops = {
> > -     .d_hash = ext4_d_hash,
> > -     .d_compare = ext4_d_compare,
> > +     .d_hash = generic_ci_d_hash,
> > +     .d_compare = generic_ci_d_compare,
> >  };
> >  #endif
>
> Can you make the structure generic since it is the same for f2fs and
> ext4, which let you drop the code guards?  Unless that becomes a problem for
> d_revalidate with fscrypt, it is fine like this.
>
> --
> Gabriel Krisman Bertazi

I unify them in a later patch, since I end up having to deal with
fscrypt's d_revalidate. With that patch I'd end up undoing the export
I'd add for this, so I'll skip that for the moment.

-Daniel

Re: [PATCH v9 4/4] ext4: Use generic casefolding support

[Eric Biggers]

On Tue, Jun 23, 2020 at 09:33:41PM -0700, Daniel Rosenberg wrote:
> This switches ext4 over to the generic support provided in
> commit 5f829feca774 ("fs: Add standard casefolding support")

Commit IDs aren't determined until the patches are applied.  It's possible for
the person applying the patches to fix them, but in general people will forget,
so it's better not to include non-stable commit IDs.

Also, a sentence explaining *why* this change is good would be helpful.
Commit messages should always have a *why* unless it's obvious.

Likewise for the f2fs commit.

- Eric

Re: [PATCH v9 0/4] Prepare for upcoming Casefolding/Encryption patches

[Eric Biggers]

On Tue, Jun 23, 2020 at 09:33:37PM -0700, Daniel Rosenberg wrote:
> This lays the ground work for enabling casefolding and encryption at the
> same time for ext4 and f2fs. A future set of patches will enable that
> functionality. These unify the highly similar dentry_operations that ext4
> and f2fs both use for casefolding.

I think this undersells this patchset a bit.  This patchset makes ext4 and f2fs
share the casefolded ->d_compare() and ->d_hash() implementations, which
eliminates duplicated code.  That's a good thing regardless of whether we're
going to add encrypt+casefold support or not.

It also changes the casefolded ->d_hash() implementation to not have to allocate
memory (with GFP_ATOMIC, no less), which was a big problem with the old
implementation as it's unreliable and inefficient.

So yes, this prepares for supporting encrypt+casefold.  But these changes make
sense on their own too as an improvement of the casefold feature.  Except for
the one line of code in needs_casefold() that is specific to encrypt+casefold;
maybe that should be left out for now.

(Side note: I think you could drop linux-doc and linux-mtd from Cc, as this
patchset isn't really relevant to those mailing lists.)

- Eric