[PATCH v5 00/55] xfs: online scrub/repair support

[PATCH v5 00/55] xfs: online scrub/repair support

[Darrick J. Wong]

Hi all,

This is the fifth revision of a patchset that adds to XFS kernel support
for online metadata scrubbing and repair.  There aren't any on-disk
format changes.  Changes since v4 include numerous bug fixes, somewhat
more aggressive log flushing so that on-disk metadata, and the ability
to distinguish between metadata that's obviously corrupt and metadata
that merely fails cross-referencing checks in the status that is sent
back to userspace.  I have also begun using it to check all my
development workstations, which has been useful for flushing out more
bugs.

Online scrub/repair support consists of four major pieces -- first, an
ioctl that maps physical extents to their owners; second, various
in-kernel metadata scrubbing ioctls to examine metadata records and
cross-reference them with other filesystem metadata; third, an in-kernel
mechanism for rebuilding damaged metadata objects and btrees; and
fourth, a userspace component to initiate kernel scrubbing, walk all
inodes and the directory tree, scrub data extents, and ask the kernel to
repair anything that is broken.

This new utility, xfs_scrub, is separate from the existing offline
xfs_repair tool.  Scrub has three main modes of operation -- in its most
powerful mode, it iterates all XFS metadata and asks the kernel to check
the metadata and repair it if necessary.  The second most powerful mode
can use certain VFS methods and XFS ioctls (BULKSTAT, GETBMAP, and
GETFSMAP) to check as much metadata as it reasonably can from userspace.
It cannot repair anything.  The least powerful mode uses only VFS
functions to access as much of the directory/file/xattr graph as
possible.  It has no mechanism to check internal metadata and also
cannot repair anything.  This is good enough for scrubbing non-XFS
filesystems, but the primary goal is first-class XFS support.

As usual, the first patches in this series are bug fixes for problems
discovered while running the code through rigorous fuzz testing.

The next few patches in this series implements the GETFSMAP ioctl that
maps a device number and physical extent either to filesystem metadata
or to a range of file blocks.  The initial implementation uses the
reverse-mapping B+tree to supply the mapping information, however a
fallback implementation based on the free space btrees is also provided.
The flexibility of having both implementations is important when it
comes to the userspace tool -- even without the owner/offset data, we
still have enough information to set up a read verification.  There's
also a patch to enable xfs_scrub to query the per-AG block reservations
so that the summary counters can be sanity-checked.

The next big chunk of patches implement in-kernel scrubbing.  This is
implemented as a new ioctl.  Pass in a metadata type and control data
such as an AG number or inode (when applicable); the kernel will examine
each record in that metadata structure looking for obvious logical
errors.  External corruption should be discoverable via the checksum
embedded in each (v5) filesystem metadata block.  When applicable, the
metadata record will be cross-referenced with the other metadata
structures to look for discrepancies.  Should any errors be found, an
error code is returned to userspace, which in the old days would require
the administrator to take the filesystem offline and repair it.  I've
hidden the new online scrubber behind CONFIG_XFS_DEBUG to keep it
disabled by default.

Last comes the online *repair* functionality, which largely uses the
redundancy between the new reverse-mapping feature introduced in 4.8 and
the existing storage space records (bno, cnt, ino, fino, and bmap) to
reconstruct primary metadata from the secondary, or secondary metadata
from the primaries.  That's right, we can regrow (some) of the XFS
metadata even if parts of the filesystem go bad!  Should the kernel
succeed, it is not necessary to take the filesystem offline for repair.

Finally, there's a patch that uses one of the new scrub features to
prevent mount-time deadlocks if the refcountbt is corrupt.

If you're going to start using this mess, you probably ought to just
pull from my github trees.  The kernel patches[1] should apply against
4.10-rc4.  xfsprogs[2] and xfstests[3] can be found in their usual
places.

The patches have survived all of the new tests in [3] that try to fuzz
every field in every data structure on disk, which has shaken out
several bugs in the scrubber and in other parts of XFS.

This is an extraordinary way to eat your data.  Enjoy! 
Comments and questions are, as always, welcome.

--D

[1] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfs-linux.git/log/?h=djwong-devel
[2] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfsprogs-dev.git/log/?h=djwong-devel
[3] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfstests-dev.git/log/?h=djwong-devel

[PATCH 01/55] xfs: fix toctou race when locking an inode to access the data map

[Darrick J. Wong]

We use di_format and if_flags to decide whether we're grabbing the ilock
in btree mode (btree extents not loaded) or shared mode (anything else),
but the state of those fields can be changed by other threads that are
also trying to load the btree extents -- IFEXTENTS gets set before the
_bmap_read_extents call and cleared if it fails.  Therefore, once we've
grabbed the shared ilock we have to re-check the fields to see if we
actually need to upgrade to the exclusive ilock in order to try loading
the extents.

Without this patch, we trigger ilock assert failures when a bunch of
threads try to access a btree format directory with a corrupt bmbt root
and corrupt the incore data structures, leading to a crash.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/xfs_inode.c |   12 ++++++++++++
 1 file changed, 12 insertions(+)


diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index b955779..b04bda8 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -125,6 +125,18 @@ xfs_ilock_data_map_shared(
 	    (ip->i_df.if_flags & XFS_IFEXTENTS) == 0)
 		lock_mode = XFS_ILOCK_EXCL;
 	xfs_ilock(ip, lock_mode);
+	/*
+	 * We can change if_flags under ilock if we try to read the
+	 * extents and fail.  Since we hadn't grabbed the ilock at check
+	 * time, we have to re-check and upgrade the lock now.
+	 */
+	if (lock_mode == XFS_ILOCK_SHARED &&
+	    ip->i_d.di_format == XFS_DINODE_FMT_BTREE &&
+	    (ip->i_df.if_flags & XFS_IFEXTENTS) == 0) {
+		xfs_iunlock(ip, lock_mode);
+		lock_mode = XFS_ILOCK_EXCL;
+		xfs_ilock(ip, lock_mode);
+	}
 	return lock_mode;
 }
 

[PATCH 02/55] xfs: fail _dir_open when readahead fails

[Darrick J. Wong]

When we open a directory, we try to readahead block 0 of the directory
on the assumption that we're going to need it soon.  If the bmbt is
corrupt, the directory will never be usable so we might as well prevent
the directory from being opened at all.. This prevents a subsequent read
or modify operation from hitting it and taking the fs offline.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_da_btree.c |    6 ++----
 fs/xfs/libxfs/xfs_da_btree.h |    2 +-
 fs/xfs/xfs_file.c            |    4 ++--
 3 files changed, 5 insertions(+), 7 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_da_btree.c b/fs/xfs/libxfs/xfs_da_btree.c
index f2dc1a9..1bdf288 100644
--- a/fs/xfs/libxfs/xfs_da_btree.c
+++ b/fs/xfs/libxfs/xfs_da_btree.c
@@ -2633,7 +2633,7 @@ xfs_da_read_buf(
 /*
  * Readahead the dir/attr block.
  */
-xfs_daddr_t
+int
 xfs_da_reada_buf(
 	struct xfs_inode	*dp,
 	xfs_dablk_t		bno,
@@ -2664,7 +2664,5 @@ xfs_da_reada_buf(
 	if (mapp != &map)
 		kmem_free(mapp);
 
-	if (error)
-		return -1;
-	return mappedbno;
+	return error;
 }
diff --git a/fs/xfs/libxfs/xfs_da_btree.h b/fs/xfs/libxfs/xfs_da_btree.h
index 98c75cb..4e29cb6 100644
--- a/fs/xfs/libxfs/xfs_da_btree.h
+++ b/fs/xfs/libxfs/xfs_da_btree.h
@@ -201,7 +201,7 @@ int	xfs_da_read_buf(struct xfs_trans *trans, struct xfs_inode *dp,
 			       xfs_dablk_t bno, xfs_daddr_t mappedbno,
 			       struct xfs_buf **bpp, int whichfork,
 			       const struct xfs_buf_ops *ops);
-xfs_daddr_t	xfs_da_reada_buf(struct xfs_inode *dp, xfs_dablk_t bno,
+int	xfs_da_reada_buf(struct xfs_inode *dp, xfs_dablk_t bno,
 				xfs_daddr_t mapped_bno, int whichfork,
 				const struct xfs_buf_ops *ops);
 int	xfs_da_shrink_inode(xfs_da_args_t *args, xfs_dablk_t dead_blkno,
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index bbb9eb6..4c87e60f 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -908,9 +908,9 @@ xfs_dir_open(
 	 */
 	mode = xfs_ilock_data_map_shared(ip);
 	if (ip->i_d.di_nextents > 0)
-		xfs_dir3_data_readahead(ip, 0, -1);
+		error = xfs_dir3_data_readahead(ip, 0, -1);
 	xfs_iunlock(ip, mode);
-	return 0;
+	return error;
 }
 
 STATIC int

[PATCH 03/55] xfs: filter out obviously bad btree pointers

[Darrick J. Wong]

Don't let anybody load an obviously bad btree pointer.  Since the values
come from disk, we must return an error, not just ASSERT.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_bmap.c  |    4 +---
 fs/xfs/libxfs/xfs_btree.c |    3 ++-
 fs/xfs/libxfs/xfs_btree.h |    2 +-
 3 files changed, 4 insertions(+), 5 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c
index 44773c9..3ac13be 100644
--- a/fs/xfs/libxfs/xfs_bmap.c
+++ b/fs/xfs/libxfs/xfs_bmap.c
@@ -1291,9 +1291,7 @@ xfs_bmap_read_extents(
 	ASSERT(level > 0);
 	pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes);
 	bno = be64_to_cpu(*pp);
-	ASSERT(bno != NULLFSBLOCK);
-	ASSERT(XFS_FSB_TO_AGNO(mp, bno) < mp->m_sb.sb_agcount);
-	ASSERT(XFS_FSB_TO_AGBNO(mp, bno) < mp->m_sb.sb_agblocks);
+
 	/*
 	 * Go down the tree until leaf level is reached, following the first
 	 * pointer (leftmost) at each level.
diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c
index 21e6a6a..2849d3f 100644
--- a/fs/xfs/libxfs/xfs_btree.c
+++ b/fs/xfs/libxfs/xfs_btree.c
@@ -810,7 +810,8 @@ xfs_btree_read_bufl(
 	xfs_daddr_t		d;		/* real disk block address */
 	int			error;
 
-	ASSERT(fsbno != NULLFSBLOCK);
+	if (!XFS_FSB_SANITY_CHECK(mp, fsbno))
+		return -EFSCORRUPTED;
 	d = XFS_FSB_TO_DADDR(mp, fsbno);
 	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, d,
 				   mp->m_bsize, lock, &bp, ops);
diff --git a/fs/xfs/libxfs/xfs_btree.h b/fs/xfs/libxfs/xfs_btree.h
index b69b947..33a8f86 100644
--- a/fs/xfs/libxfs/xfs_btree.h
+++ b/fs/xfs/libxfs/xfs_btree.h
@@ -456,7 +456,7 @@ static inline int xfs_btree_get_level(struct xfs_btree_block *block)
 #define	XFS_FILBLKS_MAX(a,b)	max_t(xfs_filblks_t, (a), (b))
 
 #define	XFS_FSB_SANITY_CHECK(mp,fsb)	\
-	(XFS_FSB_TO_AGNO(mp, fsb) < mp->m_sb.sb_agcount && \
+	(fsb && XFS_FSB_TO_AGNO(mp, fsb) < mp->m_sb.sb_agcount && \
 		XFS_FSB_TO_AGBNO(mp, fsb) < mp->m_sb.sb_agblocks)
 
 /*

[PATCH 04/55] xfs: check for obviously bad level values in the bmbt root

[Darrick J. Wong]

We can't handle a bmbt that's taller than BTREE_MAXLEVELS, and there's
no such thing as a zero-level bmbt (for that we have extents format),
so if we see this, send back an error code.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_inode_fork.c |    6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)


diff --git a/fs/xfs/libxfs/xfs_inode_fork.c b/fs/xfs/libxfs/xfs_inode_fork.c
index 222e103..84b3e51 100644
--- a/fs/xfs/libxfs/xfs_inode_fork.c
+++ b/fs/xfs/libxfs/xfs_inode_fork.c
@@ -26,6 +26,7 @@
 #include "xfs_inode.h"
 #include "xfs_trans.h"
 #include "xfs_inode_item.h"
+#include "xfs_btree.h"
 #include "xfs_bmap_btree.h"
 #include "xfs_bmap.h"
 #include "xfs_error.h"
@@ -429,11 +430,13 @@ xfs_iformat_btree(
 	/* REFERENCED */
 	int			nrecs;
 	int			size;
+	int			level;
 
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 	dfp = (xfs_bmdr_block_t *)XFS_DFORK_PTR(dip, whichfork);
 	size = XFS_BMAP_BROOT_SPACE(mp, dfp);
 	nrecs = be16_to_cpu(dfp->bb_numrecs);
+	level = be16_to_cpu(dfp->bb_level);
 
 	/*
 	 * blow out if -- fork has less extents than can fit in
@@ -446,7 +449,8 @@ xfs_iformat_btree(
 					XFS_IFORK_MAXEXT(ip, whichfork) ||
 		     XFS_BMDR_SPACE_CALC(nrecs) >
 					XFS_DFORK_SIZE(dip, mp, whichfork) ||
-		     XFS_IFORK_NEXTENTS(ip, whichfork) > ip->i_d.di_nblocks)) {
+		     XFS_IFORK_NEXTENTS(ip, whichfork) > ip->i_d.di_nblocks) ||
+		     level == 0 || level > XFS_BTREE_MAXLEVELS) {
 		xfs_warn(mp, "corrupt inode %Lu (btree).",
 					(unsigned long long) ip->i_ino);
 		XFS_CORRUPTION_ERROR("xfs_iformat_btree", XFS_ERRLEVEL_LOW,

[PATCH 05/55] xfs: verify free block header fields

[Darrick J. Wong]

Perform basic sanity checking of the directory free block header
fields so that we avoid hanging the system on invalid data.

(Granted that just means that now we shutdown on directory write,
but that seems better than hanging...)

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_dir2_node.c |   49 +++++++++++++++++++++++++++++++++++++++--
 1 file changed, 47 insertions(+), 2 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_dir2_node.c b/fs/xfs/libxfs/xfs_dir2_node.c
index 75a5574..6f04e40 100644
--- a/fs/xfs/libxfs/xfs_dir2_node.c
+++ b/fs/xfs/libxfs/xfs_dir2_node.c
@@ -155,6 +155,42 @@ const struct xfs_buf_ops xfs_dir3_free_buf_ops = {
 	.verify_write = xfs_dir3_free_write_verify,
 };
 
+/* Everything ok in the free block header? */
+static bool
+xfs_dir3_free_header_check(
+	struct xfs_inode	*dp,
+	xfs_dablk_t		fbno,
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = dp->i_mount;
+	unsigned int		firstdb;
+	int			maxbests;
+
+	maxbests = dp->d_ops->free_max_bests(mp->m_dir_geo);
+	firstdb = (fbno - xfs_dir2_byte_to_db(mp->m_dir_geo,
+					      XFS_DIR2_FREE_OFFSET)) *
+			maxbests;
+	if (xfs_sb_version_hascrc(&mp->m_sb)) {
+		struct xfs_dir3_free_hdr *hdr3 = bp->b_addr;
+
+		if (be32_to_cpu(hdr3->firstdb) != firstdb)
+			return false;
+		if (be32_to_cpu(hdr3->nvalid) > maxbests)
+			return false;
+		if (be32_to_cpu(hdr3->nvalid) < be32_to_cpu(hdr3->nused))
+			return false;
+	} else {
+		struct xfs_dir2_free_hdr *hdr = bp->b_addr;
+
+		if (be32_to_cpu(hdr->firstdb) != firstdb)
+			return false;
+		if (be32_to_cpu(hdr->nvalid) > maxbests)
+			return false;
+		if (be32_to_cpu(hdr->nvalid) < be32_to_cpu(hdr->nused))
+			return false;
+	}
+	return true;
+}
 
 static int
 __xfs_dir3_free_read(
@@ -168,11 +204,20 @@ __xfs_dir3_free_read(
 
 	err = xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
 				XFS_DATA_FORK, &xfs_dir3_free_buf_ops);
+	if (err || !(*bpp))
+		return err;
+
+	/* Check things that we can't do in the verifier. */
+	if (!xfs_dir3_free_header_check(dp, fbno, *bpp)) {
+		xfs_trans_brelse(tp, *bpp);
+		return -EFSCORRUPTED;
+	}
 
 	/* try read returns without an error or *bpp if it lands in a hole */
-	if (!err && tp && *bpp)
+	if (tp)
 		xfs_trans_buf_set_type(tp, *bpp, XFS_BLFT_DIR_FREE_BUF);
-	return err;
+
+	return 0;
 }
 
 int

[PATCH 06/55] xfs: plumb in needed functions for range querying of the freespace btrees

[Darrick J. Wong]

Plumb in the pieces (init_high_key, diff_two_keys) necessary to call
query_range on the free space btrees.  Remove the debugging asserts
so that we can make queries starting from block 0.

While we're at it, merge the redundant "if (btnum ==" hunks.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_alloc_btree.c |  162 +++++++++++++++++++++++++++++----------
 1 file changed, 119 insertions(+), 43 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_alloc_btree.c b/fs/xfs/libxfs/xfs_alloc_btree.c
index efb467b..ba3ec9c 100644
--- a/fs/xfs/libxfs/xfs_alloc_btree.c
+++ b/fs/xfs/libxfs/xfs_alloc_btree.c
@@ -205,19 +205,28 @@ xfs_allocbt_init_key_from_rec(
 	union xfs_btree_key	*key,
 	union xfs_btree_rec	*rec)
 {
-	ASSERT(rec->alloc.ar_startblock != 0);
-
 	key->alloc.ar_startblock = rec->alloc.ar_startblock;
 	key->alloc.ar_blockcount = rec->alloc.ar_blockcount;
 }
 
 STATIC void
+xfs_bnobt_init_high_key_from_rec(
+	union xfs_btree_key	*key,
+	union xfs_btree_rec	*rec)
+{
+	__u32			x;
+
+	x = be32_to_cpu(rec->alloc.ar_startblock);
+	x += be32_to_cpu(rec->alloc.ar_blockcount) - 1;
+	key->alloc.ar_startblock = cpu_to_be32(x);
+	key->alloc.ar_blockcount = 0;
+}
+
+STATIC void
 xfs_allocbt_init_rec_from_cur(
 	struct xfs_btree_cur	*cur,
 	union xfs_btree_rec	*rec)
 {
-	ASSERT(cur->bc_rec.a.ar_startblock != 0);
-
 	rec->alloc.ar_startblock = cpu_to_be32(cur->bc_rec.a.ar_startblock);
 	rec->alloc.ar_blockcount = cpu_to_be32(cur->bc_rec.a.ar_blockcount);
 }
@@ -236,18 +245,24 @@ xfs_allocbt_init_ptr_from_cur(
 }
 
 STATIC __int64_t
-xfs_allocbt_key_diff(
+xfs_bnobt_key_diff(
 	struct xfs_btree_cur	*cur,
 	union xfs_btree_key	*key)
 {
 	xfs_alloc_rec_incore_t	*rec = &cur->bc_rec.a;
 	xfs_alloc_key_t		*kp = &key->alloc;
-	__int64_t		diff;
 
-	if (cur->bc_btnum == XFS_BTNUM_BNO) {
-		return (__int64_t)be32_to_cpu(kp->ar_startblock) -
-				rec->ar_startblock;
-	}
+	return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
+}
+
+STATIC __int64_t
+xfs_cntbt_key_diff(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*key)
+{
+	xfs_alloc_rec_incore_t	*rec = &cur->bc_rec.a;
+	xfs_alloc_key_t		*kp = &key->alloc;
+	__int64_t		diff;
 
 	diff = (__int64_t)be32_to_cpu(kp->ar_blockcount) - rec->ar_blockcount;
 	if (diff)
@@ -256,6 +271,33 @@ xfs_allocbt_key_diff(
 	return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
 }
 
+STATIC __int64_t
+xfs_bnobt_diff_two_keys(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*k1,
+	union xfs_btree_key	*k2)
+{
+	return (__int64_t)be32_to_cpu(k1->alloc.ar_startblock) -
+			  be32_to_cpu(k2->alloc.ar_startblock);
+}
+
+STATIC __int64_t
+xfs_cntbt_diff_two_keys(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*k1,
+	union xfs_btree_key	*k2)
+{
+	__int64_t		diff;
+
+	diff =  be32_to_cpu(k1->alloc.ar_blockcount) -
+		be32_to_cpu(k2->alloc.ar_blockcount);
+	if (diff)
+		return diff;
+
+	return  be32_to_cpu(k1->alloc.ar_startblock) -
+		be32_to_cpu(k2->alloc.ar_startblock);
+}
+
 static bool
 xfs_allocbt_verify(
 	struct xfs_buf		*bp)
@@ -346,44 +388,78 @@ const struct xfs_buf_ops xfs_allocbt_buf_ops = {
 
 #if defined(DEBUG) || defined(XFS_WARN)
 STATIC int
-xfs_allocbt_keys_inorder(
+xfs_bnobt_keys_inorder(
 	struct xfs_btree_cur	*cur,
 	union xfs_btree_key	*k1,
 	union xfs_btree_key	*k2)
 {
-	if (cur->bc_btnum == XFS_BTNUM_BNO) {
-		return be32_to_cpu(k1->alloc.ar_startblock) <
-		       be32_to_cpu(k2->alloc.ar_startblock);
-	} else {
-		return be32_to_cpu(k1->alloc.ar_blockcount) <
-			be32_to_cpu(k2->alloc.ar_blockcount) ||
-			(k1->alloc.ar_blockcount == k2->alloc.ar_blockcount &&
-			 be32_to_cpu(k1->alloc.ar_startblock) <
-			 be32_to_cpu(k2->alloc.ar_startblock));
-	}
+	return be32_to_cpu(k1->alloc.ar_startblock) <
+	       be32_to_cpu(k2->alloc.ar_startblock);
 }
 
 STATIC int
-xfs_allocbt_recs_inorder(
+xfs_bnobt_recs_inorder(
 	struct xfs_btree_cur	*cur,
 	union xfs_btree_rec	*r1,
 	union xfs_btree_rec	*r2)
 {
-	if (cur->bc_btnum == XFS_BTNUM_BNO) {
-		return be32_to_cpu(r1->alloc.ar_startblock) +
-			be32_to_cpu(r1->alloc.ar_blockcount) <=
-			be32_to_cpu(r2->alloc.ar_startblock);
-	} else {
-		return be32_to_cpu(r1->alloc.ar_blockcount) <
-			be32_to_cpu(r2->alloc.ar_blockcount) ||
-			(r1->alloc.ar_blockcount == r2->alloc.ar_blockcount &&
-			 be32_to_cpu(r1->alloc.ar_startblock) <
-			 be32_to_cpu(r2->alloc.ar_startblock));
-	}
+	return be32_to_cpu(r1->alloc.ar_startblock) +
+		be32_to_cpu(r1->alloc.ar_blockcount) <=
+		be32_to_cpu(r2->alloc.ar_startblock);
+}
+
+STATIC int
+xfs_cntbt_keys_inorder(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*k1,
+	union xfs_btree_key	*k2)
+{
+	return be32_to_cpu(k1->alloc.ar_blockcount) <
+		be32_to_cpu(k2->alloc.ar_blockcount) ||
+		(k1->alloc.ar_blockcount == k2->alloc.ar_blockcount &&
+		 be32_to_cpu(k1->alloc.ar_startblock) <
+		 be32_to_cpu(k2->alloc.ar_startblock));
+}
+
+STATIC int
+xfs_cntbt_recs_inorder(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*r1,
+	union xfs_btree_rec	*r2)
+{
+	return be32_to_cpu(r1->alloc.ar_blockcount) <
+		be32_to_cpu(r2->alloc.ar_blockcount) ||
+		(r1->alloc.ar_blockcount == r2->alloc.ar_blockcount &&
+		 be32_to_cpu(r1->alloc.ar_startblock) <
+		 be32_to_cpu(r2->alloc.ar_startblock));
 }
-#endif	/* DEBUG */
+#endif /* DEBUG */
+
+static const struct xfs_btree_ops xfs_bnobt_ops = {
+	.rec_len		= sizeof(xfs_alloc_rec_t),
+	.key_len		= sizeof(xfs_alloc_key_t),
+
+	.dup_cursor		= xfs_allocbt_dup_cursor,
+	.set_root		= xfs_allocbt_set_root,
+	.alloc_block		= xfs_allocbt_alloc_block,
+	.free_block		= xfs_allocbt_free_block,
+	.update_lastrec		= xfs_allocbt_update_lastrec,
+	.get_minrecs		= xfs_allocbt_get_minrecs,
+	.get_maxrecs		= xfs_allocbt_get_maxrecs,
+	.init_key_from_rec	= xfs_allocbt_init_key_from_rec,
+	.init_high_key_from_rec	= xfs_bnobt_init_high_key_from_rec,
+	.init_rec_from_cur	= xfs_allocbt_init_rec_from_cur,
+	.init_ptr_from_cur	= xfs_allocbt_init_ptr_from_cur,
+	.key_diff		= xfs_bnobt_key_diff,
+	.buf_ops		= &xfs_allocbt_buf_ops,
+	.diff_two_keys		= xfs_bnobt_diff_two_keys,
+#if defined(DEBUG) || defined(XFS_WARN)
+	.keys_inorder		= xfs_bnobt_keys_inorder,
+	.recs_inorder		= xfs_bnobt_recs_inorder,
+#endif
+};
 
-static const struct xfs_btree_ops xfs_allocbt_ops = {
+static const struct xfs_btree_ops xfs_cntbt_ops = {
 	.rec_len		= sizeof(xfs_alloc_rec_t),
 	.key_len		= sizeof(xfs_alloc_key_t),
 
@@ -397,11 +473,12 @@ static const struct xfs_btree_ops xfs_allocbt_ops = {
 	.init_key_from_rec	= xfs_allocbt_init_key_from_rec,
 	.init_rec_from_cur	= xfs_allocbt_init_rec_from_cur,
 	.init_ptr_from_cur	= xfs_allocbt_init_ptr_from_cur,
-	.key_diff		= xfs_allocbt_key_diff,
+	.key_diff		= xfs_cntbt_key_diff,
 	.buf_ops		= &xfs_allocbt_buf_ops,
+	.diff_two_keys		= xfs_cntbt_diff_two_keys,
 #if defined(DEBUG) || defined(XFS_WARN)
-	.keys_inorder		= xfs_allocbt_keys_inorder,
-	.recs_inorder		= xfs_allocbt_recs_inorder,
+	.keys_inorder		= xfs_cntbt_keys_inorder,
+	.recs_inorder		= xfs_cntbt_recs_inorder,
 #endif
 };
 
@@ -427,16 +504,15 @@ xfs_allocbt_init_cursor(
 	cur->bc_mp = mp;
 	cur->bc_btnum = btnum;
 	cur->bc_blocklog = mp->m_sb.sb_blocklog;
-	cur->bc_ops = &xfs_allocbt_ops;
-	if (btnum == XFS_BTNUM_BNO)
-		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtb_2);
-	else
-		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtc_2);
 
 	if (btnum == XFS_BTNUM_CNT) {
+		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtc_2);
+		cur->bc_ops = &xfs_cntbt_ops;
 		cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]);
 		cur->bc_flags = XFS_BTREE_LASTREC_UPDATE;
 	} else {
+		cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_abtb_2);
+		cur->bc_ops = &xfs_bnobt_ops;
 		cur->bc_nlevels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]);
 	}
 

[PATCH 07/55] xfs: provide a query_range function for freespace btrees

[Darrick J. Wong]

Implement a query_range function for the bnobt and cntbt.  This will
be used for getfsmap fallback if there is no rmapbt and by the online
scrub and repair code.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_alloc.c |   42 ++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/libxfs/xfs_alloc.h |   10 ++++++++++
 2 files changed, 52 insertions(+)


diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
index 9f06a21..526df17 100644
--- a/fs/xfs/libxfs/xfs_alloc.c
+++ b/fs/xfs/libxfs/xfs_alloc.c
@@ -2875,3 +2875,45 @@ xfs_free_extent(
 	xfs_trans_brelse(tp, agbp);
 	return error;
 }
+
+struct xfs_alloc_query_range_info {
+	xfs_alloc_query_range_fn	fn;
+	void				*priv;
+};
+
+/* Format btree record and pass to our callback. */
+STATIC int
+xfs_alloc_query_range_helper(
+	struct xfs_btree_cur		*cur,
+	union xfs_btree_rec		*rec,
+	void				*priv)
+{
+	struct xfs_alloc_query_range_info	*query = priv;
+	struct xfs_alloc_rec_incore		irec;
+
+	irec.ar_startblock = be32_to_cpu(rec->alloc.ar_startblock);
+	irec.ar_blockcount = be32_to_cpu(rec->alloc.ar_blockcount);
+	return query->fn(cur, &irec, query->priv);
+}
+
+/* Find all free space within a given range of blocks. */
+int
+xfs_alloc_query_range(
+	struct xfs_btree_cur			*cur,
+	struct xfs_alloc_rec_incore		*low_rec,
+	struct xfs_alloc_rec_incore		*high_rec,
+	xfs_alloc_query_range_fn		fn,
+	void					*priv)
+{
+	union xfs_btree_irec			low_brec;
+	union xfs_btree_irec			high_brec;
+	struct xfs_alloc_query_range_info	query;
+
+	ASSERT(cur->bc_btnum == XFS_BTNUM_BNO);
+	low_brec.a = *low_rec;
+	high_brec.a = *high_rec;
+	query.priv = priv;
+	query.fn = fn;
+	return xfs_btree_query_range(cur, &low_brec, &high_brec,
+			xfs_alloc_query_range_helper, &query);
+}
diff --git a/fs/xfs/libxfs/xfs_alloc.h b/fs/xfs/libxfs/xfs_alloc.h
index 1d0f48a..f534998 100644
--- a/fs/xfs/libxfs/xfs_alloc.h
+++ b/fs/xfs/libxfs/xfs_alloc.h
@@ -223,4 +223,14 @@ int xfs_free_extent_fix_freelist(struct xfs_trans *tp, xfs_agnumber_t agno,
 
 xfs_extlen_t xfs_prealloc_blocks(struct xfs_mount *mp);
 
+typedef int (*xfs_alloc_query_range_fn)(
+	struct xfs_btree_cur		*cur,
+	struct xfs_alloc_rec_incore	*rec,
+	void				*priv);
+
+int xfs_alloc_query_range(struct xfs_btree_cur *cur,
+		struct xfs_alloc_rec_incore *low_rec,
+		struct xfs_alloc_rec_incore *high_rec,
+		xfs_alloc_query_range_fn fn, void *priv);
+
 #endif	/* __XFS_ALLOC_H__ */

[PATCH 08/55] xfs: create a function to query all records in a btree

[Darrick J. Wong]

Create a helper function that will query all records in a btree.
This will be used by the online repair functions to examine every
record in a btree to rebuild a second btree.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_alloc.c |   15 +++++++++++++++
 fs/xfs/libxfs/xfs_alloc.h |    2 ++
 fs/xfs/libxfs/xfs_btree.c |   15 +++++++++++++++
 fs/xfs/libxfs/xfs_btree.h |    2 ++
 fs/xfs/libxfs/xfs_rmap.c  |   28 +++++++++++++++++++++-------
 fs/xfs/libxfs/xfs_rmap.h  |    2 ++
 6 files changed, 57 insertions(+), 7 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
index 526df17..ba15f30 100644
--- a/fs/xfs/libxfs/xfs_alloc.c
+++ b/fs/xfs/libxfs/xfs_alloc.c
@@ -2917,3 +2917,18 @@ xfs_alloc_query_range(
 	return xfs_btree_query_range(cur, &low_brec, &high_brec,
 			xfs_alloc_query_range_helper, &query);
 }
+
+/* Find all free space records. */
+int
+xfs_alloc_query_all(
+	struct xfs_btree_cur			*cur,
+	xfs_alloc_query_range_fn		fn,
+	void					*priv)
+{
+	struct xfs_alloc_query_range_info	query;
+
+	ASSERT(cur->bc_btnum == XFS_BTNUM_BNO);
+	query.priv = priv;
+	query.fn = fn;
+	return xfs_btree_query_all(cur, xfs_alloc_query_range_helper, &query);
+}
diff --git a/fs/xfs/libxfs/xfs_alloc.h b/fs/xfs/libxfs/xfs_alloc.h
index f534998..a2101de 100644
--- a/fs/xfs/libxfs/xfs_alloc.h
+++ b/fs/xfs/libxfs/xfs_alloc.h
@@ -232,5 +232,7 @@ int xfs_alloc_query_range(struct xfs_btree_cur *cur,
 		struct xfs_alloc_rec_incore *low_rec,
 		struct xfs_alloc_rec_incore *high_rec,
 		xfs_alloc_query_range_fn fn, void *priv);
+int xfs_alloc_query_all(struct xfs_btree_cur *cur, xfs_alloc_query_range_fn fn,
+		void *priv);
 
 #endif	/* __XFS_ALLOC_H__ */
diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c
index 2849d3f..ca94b87 100644
--- a/fs/xfs/libxfs/xfs_btree.c
+++ b/fs/xfs/libxfs/xfs_btree.c
@@ -4823,6 +4823,21 @@ xfs_btree_query_range(
 			fn, priv);
 }
 
+/* Query a btree for all records. */
+int
+xfs_btree_query_all(
+	struct xfs_btree_cur		*cur,
+	xfs_btree_query_range_fn	fn,
+	void				*priv)
+{
+	union xfs_btree_irec		low_rec;
+	union xfs_btree_irec		high_rec;
+
+	memset(&low_rec, 0, sizeof(low_rec));
+	memset(&high_rec, 0xFF, sizeof(high_rec));
+	return xfs_btree_query_range(cur, &low_rec, &high_rec, fn, priv);
+}
+
 /*
  * Calculate the number of blocks needed to store a given number of records
  * in a short-format (per-AG metadata) btree.
diff --git a/fs/xfs/libxfs/xfs_btree.h b/fs/xfs/libxfs/xfs_btree.h
index 33a8f86..5114055 100644
--- a/fs/xfs/libxfs/xfs_btree.h
+++ b/fs/xfs/libxfs/xfs_btree.h
@@ -494,6 +494,8 @@ typedef int (*xfs_btree_query_range_fn)(struct xfs_btree_cur *cur,
 int xfs_btree_query_range(struct xfs_btree_cur *cur,
 		union xfs_btree_irec *low_rec, union xfs_btree_irec *high_rec,
 		xfs_btree_query_range_fn fn, void *priv);
+int xfs_btree_query_all(struct xfs_btree_cur *cur, xfs_btree_query_range_fn fn,
+		void *priv);
 
 typedef int (*xfs_btree_visit_blocks_fn)(struct xfs_btree_cur *cur, int level,
 		void *data);
diff --git a/fs/xfs/libxfs/xfs_rmap.c b/fs/xfs/libxfs/xfs_rmap.c
index 3a8cc71..3840556 100644
--- a/fs/xfs/libxfs/xfs_rmap.c
+++ b/fs/xfs/libxfs/xfs_rmap.c
@@ -2001,14 +2001,14 @@ xfs_rmap_query_range_helper(
 /* Find all rmaps between two keys. */
 int
 xfs_rmap_query_range(
-	struct xfs_btree_cur		*cur,
-	struct xfs_rmap_irec		*low_rec,
-	struct xfs_rmap_irec		*high_rec,
-	xfs_rmap_query_range_fn	fn,
-	void				*priv)
+	struct xfs_btree_cur			*cur,
+	struct xfs_rmap_irec			*low_rec,
+	struct xfs_rmap_irec			*high_rec,
+	xfs_rmap_query_range_fn			fn,
+	void					*priv)
 {
-	union xfs_btree_irec		low_brec;
-	union xfs_btree_irec		high_brec;
+	union xfs_btree_irec			low_brec;
+	union xfs_btree_irec			high_brec;
 	struct xfs_rmap_query_range_info	query;
 
 	low_brec.r = *low_rec;
@@ -2019,6 +2019,20 @@ xfs_rmap_query_range(
 			xfs_rmap_query_range_helper, &query);
 }
 
+/* Find all rmaps. */
+int
+xfs_rmap_query_all(
+	struct xfs_btree_cur			*cur,
+	xfs_rmap_query_range_fn			fn,
+	void					*priv)
+{
+	struct xfs_rmap_query_range_info	query;
+
+	query.priv = priv;
+	query.fn = fn;
+	return xfs_btree_query_all(cur, xfs_rmap_query_range_helper, &query);
+}
+
 /* Clean up after calling xfs_rmap_finish_one. */
 void
 xfs_rmap_finish_one_cleanup(
diff --git a/fs/xfs/libxfs/xfs_rmap.h b/fs/xfs/libxfs/xfs_rmap.h
index 7899305..faf2c1a 100644
--- a/fs/xfs/libxfs/xfs_rmap.h
+++ b/fs/xfs/libxfs/xfs_rmap.h
@@ -162,6 +162,8 @@ typedef int (*xfs_rmap_query_range_fn)(
 int xfs_rmap_query_range(struct xfs_btree_cur *cur,
 		struct xfs_rmap_irec *low_rec, struct xfs_rmap_irec *high_rec,
 		xfs_rmap_query_range_fn fn, void *priv);
+int xfs_rmap_query_all(struct xfs_btree_cur *cur, xfs_rmap_query_range_fn fn,
+		void *priv);
 
 enum xfs_rmap_intent_type {
 	XFS_RMAP_MAP,

[PATCH 09/55] xfs: introduce the XFS_IOC_GETFSMAP ioctl

[Darrick J. Wong]

Introduce a new ioctl that uses the reverse mapping btree to return
information about the physical layout of the filesystem.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile        |    1 
 fs/xfs/libxfs/xfs_fs.h |   95 +++++++
 fs/xfs/xfs_fsmap.c     |  698 ++++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_fsmap.h     |   51 ++++
 fs/xfs/xfs_ioctl.c     |  104 +++++++
 fs/xfs/xfs_ioctl32.c   |    1 
 fs/xfs/xfs_trace.h     |   85 ++++++
 fs/xfs/xfs_trans.c     |   22 ++
 fs/xfs/xfs_trans.h     |    2 
 9 files changed, 1059 insertions(+)
 create mode 100644 fs/xfs/xfs_fsmap.c
 create mode 100644 fs/xfs/xfs_fsmap.h


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index c7515d4..0e7ee30 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -80,6 +80,7 @@ xfs-y				+= xfs_aops.o \
 				   xfs_extent_busy.o \
 				   xfs_file.o \
 				   xfs_filestream.o \
+				   xfs_fsmap.o \
 				   xfs_fsops.o \
 				   xfs_globals.o \
 				   xfs_icache.o \
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index b72dc82..e62996f 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -93,6 +93,100 @@ struct getbmapx {
 #define BMV_OF_SHARED		0x8	/* segment shared with another file */
 
 /*
+ *	Structure for XFS_IOC_GETFSMAP.
+ *
+ *	The memory layout for this call are the scalar values defined in
+ *	struct fsmap_head, followed by two struct fsmap that describe
+ *	the lower and upper bound of mappings to return, followed by an
+ *	array of struct fsmap mappings.
+ *
+ *	fmh_iflags control the output of the call, whereas fmh_oflags report
+ *	on the overall record output.  fmh_count should be set to the
+ *	length of the fmh_recs array, and fmh_entries will be set to the
+ *	number of entries filled out during each call.  If fmh_count is
+ *	zero, the number of reverse mappings will be returned in
+ *	fmh_entries, though no mappings will be returned.  fmh_reserved
+ *	must be set to zero.
+ *
+ *	The two elements in the fmh_keys array are used to constrain the
+ *	output.  The first element in the array should represent the
+ *	lowest disk mapping ("low key") that the user wants to learn
+ *	about.  If this value is all zeroes, the filesystem will return
+ *	the first entry it knows about.  For a subsequent call, the
+ *	contents of fsmap_head.fmh_recs[fsmap_head.fmh_count - 1] should be
+ *	copied into fmh_keys[0] to have the kernel start where it left off.
+ *
+ *	The second element in the fmh_keys array should represent the
+ *	highest disk mapping ("high key") that the user wants to learn
+ *	about.  If this value is all ones, the filesystem will not stop
+ *	until it runs out of mapping to return or runs out of space in
+ *	fmh_recs.
+ *
+ *	fmr_device can be either a 32-bit cookie representing a device, or
+ *	a 32-bit dev_t if the FMH_OF_DEV_T flag is set.  fmr_physical,
+ *	fmr_offset, and fmr_length are expressed in units of bytes.
+ *	fmr_owner is either an inode number, or a special value if
+ *	FMR_OF_SPECIAL_OWNER is set in fmr_flags.
+ */
+#ifndef HAVE_GETFSMAP
+struct fsmap {
+	__u32		fmr_device;	/* device id */
+	__u32		fmr_flags;	/* mapping flags */
+	__u64		fmr_physical;	/* device offset of segment */
+	__u64		fmr_owner;	/* owner id */
+	__u64		fmr_offset;	/* file offset of segment */
+	__u64		fmr_length;	/* length of segment */
+	__u64		fmr_reserved[3];	/* must be zero */
+};
+
+struct fsmap_head {
+	__u32		fmh_iflags;	/* control flags */
+	__u32		fmh_oflags;	/* output flags */
+	__u32		fmh_count;	/* # of entries in array incl. input */
+	__u32		fmh_entries;	/* # of entries filled in (output). */
+	__u64		fmh_reserved[6];	/* must be zero */
+
+	struct fsmap	fmh_keys[2];	/* low and high keys for the mapping search */
+	struct fsmap	fmh_recs[];	/* returned records */
+};
+
+/* Size of an fsmap_head with room for nr records. */
+static inline size_t
+fsmap_sizeof(
+	unsigned int	nr)
+{
+	return sizeof(struct fsmap_head) + nr * sizeof(struct fsmap);
+}
+#endif
+
+/*	fmh_iflags values - set by XFS_IOC_GETFSMAP caller in the header. */
+/* no flags defined yet */
+#define FMH_IF_VALID		0
+
+/*	fmh_oflags values - returned in the header segment only. */
+#define FMH_OF_DEV_T		0x1	/* fmr_device values will be dev_t */
+
+/*	fmr_flags values - returned for each non-header segment */
+#define FMR_OF_PREALLOC		0x1	/* segment = unwritten pre-allocation */
+#define FMR_OF_ATTR_FORK	0x2	/* segment = attribute fork */
+#define FMR_OF_EXTENT_MAP	0x4	/* segment = extent map */
+#define FMR_OF_SHARED		0x8	/* segment = shared with another file */
+#define FMR_OF_SPECIAL_OWNER	0x10	/* owner is a special value */
+#define FMR_OF_LAST		0x20	/* segment is the last in the FS */
+
+/*	fmr_owner special values */
+#define FMR_OWN_FREE		(-1ULL)	/* free space */
+#define FMR_OWN_UNKNOWN		(-2ULL)	/* unknown owner */
+#define FMR_OWN_FS		(-3ULL)	/* static fs metadata */
+#define FMR_OWN_LOG		(-4ULL)	/* journalling log */
+#define FMR_OWN_AG		(-5ULL)	/* per-AG metadata */
+#define FMR_OWN_INOBT		(-6ULL)	/* inode btree blocks */
+#define FMR_OWN_INODES		(-7ULL)	/* inodes */
+#define FMR_OWN_REFC		(-8ULL) /* refcount tree */
+#define FMR_OWN_COW		(-9ULL) /* cow staging */
+#define FMR_OWN_DEFECTIVE	(-10ULL) /* bad blocks */
+
+/*
  * Structure for XFS_IOC_FSSETDM.
  * For use by backup and restore programs to set the XFS on-disk inode
  * fields di_dmevmask and di_dmstate.  These must be set to exactly and
@@ -502,6 +596,7 @@ typedef struct xfs_swapext
 #define XFS_IOC_GETBMAPX	_IOWR('X', 56, struct getbmap)
 #define XFS_IOC_ZERO_RANGE	_IOW ('X', 57, struct xfs_flock64)
 #define XFS_IOC_FREE_EOFBLOCKS	_IOR ('X', 58, struct xfs_fs_eofblocks)
+#define XFS_IOC_GETFSMAP	_IOWR('X', 59, struct fsmap_head)
 
 /*
  * ioctl commands that replace IRIX syssgi()'s
diff --git a/fs/xfs/xfs_fsmap.c b/fs/xfs/xfs_fsmap.c
new file mode 100644
index 0000000..985415b
--- /dev/null
+++ b/fs/xfs/xfs_fsmap.c
@@ -0,0 +1,698 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_log_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_sb.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_inode.h"
+#include "xfs_trans.h"
+#include "xfs_error.h"
+#include "xfs_btree.h"
+#include "xfs_rmap_btree.h"
+#include "xfs_trace.h"
+#include "xfs_log.h"
+#include "xfs_rmap.h"
+#include "xfs_alloc.h"
+#include "xfs_bit.h"
+#include "xfs_fsmap.h"
+
+/* Convert an xfs_fsmap to an fsmap. */
+void
+xfs_fsmap_from_internal(
+	struct fsmap		*dest,
+	struct xfs_fsmap	*src)
+{
+	dest->fmr_device = src->fmr_device;
+	dest->fmr_flags = src->fmr_flags;
+	dest->fmr_physical = BBTOB(src->fmr_physical);
+	dest->fmr_owner = src->fmr_owner;
+	dest->fmr_offset = BBTOB(src->fmr_offset);
+	dest->fmr_length = BBTOB(src->fmr_length);
+	dest->fmr_reserved[0] = 0;
+	dest->fmr_reserved[1] = 0;
+	dest->fmr_reserved[2] = 0;
+}
+
+/* Convert an fsmap to an xfs_fsmap. */
+void
+xfs_fsmap_to_internal(
+	struct xfs_fsmap	*dest,
+	struct fsmap		*src)
+{
+	dest->fmr_device = src->fmr_device;
+	dest->fmr_flags = src->fmr_flags;
+	dest->fmr_physical = BTOBBT(src->fmr_physical);
+	dest->fmr_owner = src->fmr_owner;
+	dest->fmr_offset = BTOBBT(src->fmr_offset);
+	dest->fmr_length = BTOBBT(src->fmr_length);
+}
+
+/* Convert an fsmap owner into an rmapbt owner. */
+static int
+xfs_fsmap_owner_to_rmap(
+	struct xfs_fsmap	*fmr,
+	struct xfs_rmap_irec	*rm)
+{
+	if (!(fmr->fmr_flags & FMR_OF_SPECIAL_OWNER)) {
+		if (XFS_RMAP_NON_INODE_OWNER(fmr->fmr_owner))
+			return -EINVAL;
+		rm->rm_owner = fmr->fmr_owner;
+		return 0;
+	}
+
+	switch (fmr->fmr_owner) {
+	case 0:			/* "lowest owner id possible" */
+	case FMR_OWN_FREE:
+	case FMR_OWN_UNKNOWN:
+	case FMR_OWN_FS:
+	case FMR_OWN_LOG:
+	case FMR_OWN_AG:
+	case FMR_OWN_INOBT:
+	case FMR_OWN_INODES:
+	case FMR_OWN_REFC:
+	case FMR_OWN_COW:
+		rm->rm_owner = fmr->fmr_owner;
+		return 0;
+	case FMR_OWN_DEFECTIVE:
+		/* fall through */
+	default:
+		return -EINVAL;
+	}
+}
+
+/* Convert an rmapbt owner into an fsmap owner. */
+static int
+xfs_fsmap_owner_from_rmap(
+	struct xfs_rmap_irec	*rm,
+	struct xfs_fsmap	*fmr)
+{
+	fmr->fmr_flags = 0;
+	if (!XFS_RMAP_NON_INODE_OWNER(rm->rm_owner)) {
+		fmr->fmr_owner = rm->rm_owner;
+		return 0;
+	}
+	fmr->fmr_flags |= FMR_OF_SPECIAL_OWNER;
+
+	switch (rm->rm_owner) {
+	case XFS_RMAP_OWN_FS:
+	case XFS_RMAP_OWN_LOG:
+	case XFS_RMAP_OWN_AG:
+	case XFS_RMAP_OWN_INOBT:
+	case XFS_RMAP_OWN_INODES:
+	case XFS_RMAP_OWN_REFC:
+	case XFS_RMAP_OWN_COW:
+		fmr->fmr_owner = rm->rm_owner;
+		return 0;
+	default:
+		return -EFSCORRUPTED;
+	}
+}
+
+/* getfsmap query state */
+struct xfs_getfsmap_info {
+	struct xfs_fsmap_head	*head;
+	struct xfs_fsmap	*rkey_low;	/* lowest key */
+	xfs_fsmap_format_t	formatter;	/* formatting fn */
+	void			*format_arg;	/* format buffer */
+	bool			last;		/* last extent? */
+	xfs_daddr_t		next_daddr;	/* next daddr we expect */
+	u32			dev;		/* device id */
+	u64			missing_owner;	/* owner of holes */
+
+	xfs_agnumber_t		agno;		/* AG number, if applicable */
+	struct xfs_buf		*agf_bp;	/* AGF, for refcount queries */
+	struct xfs_rmap_irec	low;		/* low rmap key */
+	struct xfs_rmap_irec	high;		/* high rmap key */
+};
+
+/* Associate a device with a getfsmap handler. */
+struct xfs_getfsmap_dev {
+	u32			dev;
+	int			(*fn)(struct xfs_trans *tp,
+				      struct xfs_fsmap *keys,
+				      struct xfs_getfsmap_info *info);
+};
+
+/* Compare two getfsmap device handlers. */
+static int
+xfs_getfsmap_dev_compare(
+	const void			*p1,
+	const void			*p2)
+{
+	const struct xfs_getfsmap_dev	*d1 = p1;
+	const struct xfs_getfsmap_dev	*d2 = p2;
+
+	return d1->dev - d2->dev;
+}
+
+/* Compare a record against our starting point */
+static bool
+xfs_getfsmap_rec_before_low_key(
+	struct xfs_getfsmap_info	*info,
+	struct xfs_rmap_irec		*rec)
+{
+	uint64_t			x, y;
+
+	if (rec->rm_startblock < info->low.rm_startblock)
+		return true;
+	if (rec->rm_startblock > info->low.rm_startblock)
+		return false;
+
+	if (rec->rm_owner < info->low.rm_owner)
+		return true;
+	if (rec->rm_owner > info->low.rm_owner)
+		return false;
+
+	x = xfs_rmap_irec_offset_pack(rec);
+	y = xfs_rmap_irec_offset_pack(&info->low);
+	if (x < y)
+		return true;
+	return false;
+}
+
+/*
+ * Format a reverse mapping for getfsmap, having translated rm_startblock
+ * into the appropriate daddr units.
+ */
+STATIC int
+xfs_getfsmap_helper(
+	struct xfs_mount		*mp,
+	struct xfs_getfsmap_info	*info,
+	struct xfs_rmap_irec		*rec,
+	xfs_daddr_t			rec_daddr)
+{
+	struct xfs_fsmap		fmr;
+	xfs_daddr_t			key_end;
+	int				error;
+
+	/*
+	 * Filter out records that start before our startpoint, if the
+	 * caller requested that.
+	 */
+	if (xfs_getfsmap_rec_before_low_key(info, rec)) {
+		rec_daddr += XFS_FSB_TO_BB(mp, rec->rm_blockcount);
+		if (info->next_daddr < rec_daddr)
+			info->next_daddr = rec_daddr;
+		return XFS_BTREE_QUERY_RANGE_CONTINUE;
+	}
+
+	/*
+	 * If the caller passed in a length with the low record and
+	 * the record represents a file data extent, we incremented
+	 * the offset in the low key by the length in the hopes of
+	 * finding reverse mappings for the physical blocks we just
+	 * saw.  We did /not/ increment next_daddr by the length
+	 * because the range query would not be able to find shared
+	 * extents within the same physical block range.
+	 *
+	 * However, the extent we've been fed could have a startblock
+	 * past the passed-in low record.  If this is the case,
+	 * advance next_daddr to the end of the passed-in low record
+	 * so we don't report the extent prior to this extent as
+	 * free.
+	 */
+	key_end = info->rkey_low->fmr_physical + info->rkey_low->fmr_length;
+	if (info->dev == info->rkey_low->fmr_device &&
+	    info->next_daddr < key_end && rec_daddr >= key_end)
+		info->next_daddr = key_end;
+
+	/* Are we just counting mappings? */
+	if (info->head->fmh_count == 0) {
+		if (rec_daddr > info->next_daddr)
+			info->head->fmh_entries++;
+
+		if (info->last)
+			return XFS_BTREE_QUERY_RANGE_CONTINUE;
+
+		info->head->fmh_entries++;
+
+		rec_daddr += XFS_FSB_TO_BB(mp, rec->rm_blockcount);
+		if (info->next_daddr < rec_daddr)
+			info->next_daddr = rec_daddr;
+		return XFS_BTREE_QUERY_RANGE_CONTINUE;
+	}
+
+	/*
+	 * If the record starts past the last physical block we saw,
+	 * then we've found some free space.  Report that too.
+	 */
+	if (rec_daddr > info->next_daddr) {
+		if (info->head->fmh_entries >= info->head->fmh_count)
+			return XFS_BTREE_QUERY_RANGE_ABORT;
+
+		trace_xfs_fsmap_mapping(mp, info->dev, info->agno,
+				XFS_DADDR_TO_FSB(mp, info->next_daddr),
+				XFS_DADDR_TO_FSB(mp, rec_daddr -
+						info->next_daddr),
+				info->missing_owner, 0);
+
+		fmr.fmr_device = info->dev;
+		fmr.fmr_physical = info->next_daddr;
+		fmr.fmr_owner = info->missing_owner;
+		fmr.fmr_offset = 0;
+		fmr.fmr_length = rec_daddr - info->next_daddr;
+		fmr.fmr_flags = FMR_OF_SPECIAL_OWNER;
+		error = info->formatter(&fmr, info->format_arg);
+		if (error)
+			return error;
+		info->head->fmh_entries++;
+	}
+
+	if (info->last)
+		goto out;
+
+	/* Fill out the extent we found */
+	if (info->head->fmh_entries >= info->head->fmh_count)
+		return XFS_BTREE_QUERY_RANGE_ABORT;
+
+	trace_xfs_fsmap_mapping(mp, info->dev, info->agno,
+			rec->rm_startblock, rec->rm_blockcount, rec->rm_owner,
+			rec->rm_offset);
+
+	fmr.fmr_device = info->dev;
+	fmr.fmr_physical = rec_daddr;
+	error = xfs_fsmap_owner_from_rmap(rec, &fmr);
+	if (error)
+		return error;
+	fmr.fmr_offset = XFS_FSB_TO_BB(mp, rec->rm_offset);
+	fmr.fmr_length = XFS_FSB_TO_BB(mp, rec->rm_blockcount);
+	if (rec->rm_flags & XFS_RMAP_UNWRITTEN)
+		fmr.fmr_flags |= FMR_OF_PREALLOC;
+	if (rec->rm_flags & XFS_RMAP_ATTR_FORK)
+		fmr.fmr_flags |= FMR_OF_ATTR_FORK;
+	if (rec->rm_flags & XFS_RMAP_BMBT_BLOCK)
+		fmr.fmr_flags |= FMR_OF_EXTENT_MAP;
+	error = info->formatter(&fmr, info->format_arg);
+	if (error)
+		return error;
+	info->head->fmh_entries++;
+
+out:
+	rec_daddr += XFS_FSB_TO_BB(mp, rec->rm_blockcount);
+	if (info->next_daddr < rec_daddr)
+		info->next_daddr = rec_daddr;
+	return XFS_BTREE_QUERY_RANGE_CONTINUE;
+}
+
+/* Transform a rmapbt irec into a fsmap */
+STATIC int
+xfs_getfsmap_datadev_helper(
+	struct xfs_btree_cur		*cur,
+	struct xfs_rmap_irec		*rec,
+	void				*priv)
+{
+	struct xfs_mount		*mp = cur->bc_mp;
+	struct xfs_getfsmap_info	*info = priv;
+	xfs_fsblock_t			fsb;
+	xfs_daddr_t			rec_daddr;
+
+	fsb = XFS_AGB_TO_FSB(mp, cur->bc_private.a.agno, rec->rm_startblock);
+	rec_daddr = XFS_FSB_TO_DADDR(mp, fsb);
+
+	return xfs_getfsmap_helper(mp, info, rec, rec_daddr);
+}
+
+/* Transform a absolute-startblock rmap (rtdev, logdev) into a fsmap */
+STATIC int
+xfs_getfsmap_rtdev_helper(
+	struct xfs_btree_cur		*cur,
+	struct xfs_rmap_irec		*rec,
+	void				*priv)
+{
+	struct xfs_mount		*mp = cur->bc_mp;
+	struct xfs_getfsmap_info	*info = priv;
+	xfs_daddr_t			rec_daddr;
+
+	rec_daddr = XFS_FSB_TO_BB(mp, rec->rm_startblock);
+
+	return xfs_getfsmap_helper(mp, info, rec, rec_daddr);
+}
+
+/* Set rmap flags based on the getfsmap flags */
+static void
+xfs_getfsmap_set_irec_flags(
+	struct xfs_rmap_irec	*irec,
+	struct xfs_fsmap	*fmr)
+{
+	irec->rm_flags = 0;
+	if (fmr->fmr_flags & FMR_OF_ATTR_FORK)
+		irec->rm_flags |= XFS_RMAP_ATTR_FORK;
+	if (fmr->fmr_flags & FMR_OF_EXTENT_MAP)
+		irec->rm_flags |= XFS_RMAP_BMBT_BLOCK;
+	if (fmr->fmr_flags & FMR_OF_PREALLOC)
+		irec->rm_flags |= XFS_RMAP_UNWRITTEN;
+}
+
+/* Execute a getfsmap query against the log device. */
+STATIC int
+xfs_getfsmap_logdev(
+	struct xfs_trans		*tp,
+	struct xfs_fsmap		*keys,
+	struct xfs_getfsmap_info	*info)
+{
+	struct xfs_mount		*mp = tp->t_mountp;
+	struct xfs_fsmap		*dkey_low = keys;
+	struct xfs_btree_cur		cur;
+	struct xfs_rmap_irec		rmap;
+	int				error;
+
+	/* Set up search keys */
+	info->low.rm_startblock = XFS_BB_TO_FSBT(mp, dkey_low->fmr_physical);
+	info->low.rm_offset = XFS_BB_TO_FSBT(mp, dkey_low->fmr_offset);
+	error = xfs_fsmap_owner_to_rmap(keys, &info->low);
+	if (error)
+		return error;
+	info->low.rm_blockcount = 0;
+	xfs_getfsmap_set_irec_flags(&info->low, dkey_low);
+
+	error = xfs_fsmap_owner_to_rmap(keys + 1, &info->high);
+	if (error)
+		return error;
+	info->high.rm_startblock = -1U;
+	info->high.rm_owner = ULLONG_MAX;
+	info->high.rm_offset = ULLONG_MAX;
+	info->high.rm_blockcount = 0;
+	info->high.rm_flags = XFS_RMAP_KEY_FLAGS | XFS_RMAP_REC_FLAGS;
+	info->missing_owner = FMR_OWN_FREE;
+
+	trace_xfs_fsmap_low_key(mp, info->dev, info->agno,
+			info->low.rm_startblock,
+			info->low.rm_blockcount,
+			info->low.rm_owner,
+			info->low.rm_offset);
+
+	trace_xfs_fsmap_high_key(mp, info->dev, info->agno,
+			info->high.rm_startblock,
+			info->high.rm_blockcount,
+			info->high.rm_owner,
+			info->high.rm_offset);
+
+
+	if (dkey_low->fmr_physical > 0)
+		return 0;
+
+	rmap.rm_startblock = 0;
+	rmap.rm_blockcount = mp->m_sb.sb_logblocks;
+	rmap.rm_owner = XFS_RMAP_OWN_LOG;
+	rmap.rm_offset = 0;
+	rmap.rm_flags = 0;
+
+	cur.bc_mp = mp;
+	return xfs_getfsmap_rtdev_helper(&cur, &rmap, info);
+}
+
+/* Execute a getfsmap query against the regular data device. */
+STATIC int
+xfs_getfsmap_datadev(
+	struct xfs_trans		*tp,
+	struct xfs_fsmap		*keys,
+	struct xfs_getfsmap_info	*info)
+{
+	struct xfs_mount		*mp = tp->t_mountp;
+	struct xfs_btree_cur		*bt_cur = NULL;
+	struct xfs_fsmap		*dkey_low;
+	struct xfs_fsmap		*dkey_high;
+	xfs_fsblock_t			start_fsb;
+	xfs_fsblock_t			end_fsb;
+	xfs_agnumber_t			start_ag;
+	xfs_agnumber_t			end_ag;
+	xfs_daddr_t			eofs;
+	int				error = 0;
+
+	dkey_low = keys;
+	dkey_high = keys + 1;
+	eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
+	if (dkey_low->fmr_physical >= eofs)
+		return 0;
+	if (dkey_high->fmr_physical >= eofs)
+		dkey_high->fmr_physical = eofs - 1;
+	start_fsb = XFS_DADDR_TO_FSB(mp, dkey_low->fmr_physical);
+	end_fsb = XFS_DADDR_TO_FSB(mp, dkey_high->fmr_physical);
+
+	/* Set up search keys */
+	info->low.rm_startblock = XFS_FSB_TO_AGBNO(mp, start_fsb);
+	info->low.rm_offset = XFS_BB_TO_FSBT(mp, dkey_low->fmr_offset);
+	error = xfs_fsmap_owner_to_rmap(dkey_low, &info->low);
+	if (error)
+		return error;
+	info->low.rm_blockcount = 0;
+	xfs_getfsmap_set_irec_flags(&info->low, dkey_low);
+
+	info->high.rm_startblock = -1U;
+	info->high.rm_owner = ULLONG_MAX;
+	info->high.rm_offset = ULLONG_MAX;
+	info->high.rm_blockcount = 0;
+	info->high.rm_flags = XFS_RMAP_KEY_FLAGS | XFS_RMAP_REC_FLAGS;
+	info->missing_owner = FMR_OWN_FREE;
+
+	start_ag = XFS_FSB_TO_AGNO(mp, start_fsb);
+	end_ag = XFS_FSB_TO_AGNO(mp, end_fsb);
+
+	/* Query each AG */
+	for (info->agno = start_ag; info->agno <= end_ag; info->agno++) {
+		if (info->agno == end_ag) {
+			info->high.rm_startblock = XFS_FSB_TO_AGBNO(mp,
+					end_fsb);
+			info->high.rm_offset = XFS_BB_TO_FSBT(mp,
+					dkey_high->fmr_offset);
+			error = xfs_fsmap_owner_to_rmap(dkey_high, &info->high);
+			if (error)
+				goto err;
+			xfs_getfsmap_set_irec_flags(&info->high, dkey_high);
+		}
+
+		if (bt_cur) {
+			xfs_btree_del_cursor(bt_cur, XFS_BTREE_NOERROR);
+			bt_cur = NULL;
+			info->agf_bp = NULL;
+		}
+
+		error = xfs_alloc_read_agf(mp, tp, info->agno, 0,
+				&info->agf_bp);
+		if (error)
+			goto err;
+
+		trace_xfs_fsmap_low_key(mp, info->dev, info->agno,
+				info->low.rm_startblock,
+				info->low.rm_blockcount,
+				info->low.rm_owner,
+				info->low.rm_offset);
+
+		trace_xfs_fsmap_high_key(mp, info->dev, info->agno,
+				info->high.rm_startblock,
+				info->high.rm_blockcount,
+				info->high.rm_owner,
+				info->high.rm_offset);
+
+		bt_cur = xfs_rmapbt_init_cursor(mp, tp, info->agf_bp,
+				info->agno);
+		error = xfs_rmap_query_range(bt_cur, &info->low, &info->high,
+				xfs_getfsmap_datadev_helper, info);
+		if (error)
+			goto err;
+
+		if (info->agno == start_ag) {
+			info->low.rm_startblock = 0;
+			info->low.rm_owner = 0;
+			info->low.rm_offset = 0;
+			info->low.rm_flags = 0;
+		}
+	}
+
+	/* Report any free space at the end of the AG */
+	info->last = true;
+	error = xfs_getfsmap_datadev_helper(bt_cur, &info->high, info);
+	if (error)
+		goto err;
+
+err:
+	if (bt_cur)
+		xfs_btree_del_cursor(bt_cur, error < 0 ? XFS_BTREE_ERROR :
+							 XFS_BTREE_NOERROR);
+	if (info->agf_bp)
+		info->agf_bp = NULL;
+
+	return error;
+}
+
+/* Do we recognize the device? */
+STATIC bool
+xfs_getfsmap_is_valid_device(
+	struct xfs_mount	*mp,
+	struct xfs_fsmap	*fm)
+{
+	if (fm->fmr_device == 0 || fm->fmr_device == UINT_MAX ||
+	    fm->fmr_device == new_encode_dev(mp->m_ddev_targp->bt_dev))
+		return true;
+	if (mp->m_logdev_targp &&
+	    fm->fmr_device == new_encode_dev(mp->m_logdev_targp->bt_dev))
+		return true;
+	return false;
+}
+
+/* Ensure that the low key is less than the high key. */
+STATIC bool
+xfs_getfsmap_check_keys(
+	struct xfs_fsmap		*low_key,
+	struct xfs_fsmap		*high_key)
+{
+	if (low_key->fmr_device > high_key->fmr_device)
+		return false;
+	if (low_key->fmr_device < high_key->fmr_device)
+		return true;
+
+	if (low_key->fmr_physical > high_key->fmr_physical)
+		return false;
+	if (low_key->fmr_physical < high_key->fmr_physical)
+		return true;
+
+	if (low_key->fmr_owner > high_key->fmr_owner)
+		return false;
+	if (low_key->fmr_owner < high_key->fmr_owner)
+		return true;
+
+	if (low_key->fmr_offset > high_key->fmr_offset)
+		return false;
+	if (low_key->fmr_offset < high_key->fmr_offset)
+		return true;
+
+	return false;
+}
+
+#define XFS_GETFSMAP_DEVS	3
+/*
+ * Get filesystem's extents as described in head, and format for
+ * output.  Calls formatter to fill the user's buffer until all
+ * extents are mapped, until the passed-in head->fmh_count slots have
+ * been filled, or until the formatter short-circuits the loop, if it
+ * is tracking filled-in extents on its own.
+ */
+int
+xfs_getfsmap(
+	struct xfs_mount		*mp,
+	struct xfs_fsmap_head		*head,
+	xfs_fsmap_format_t		formatter,
+	void				*arg)
+{
+	struct xfs_trans		*tp = NULL;
+	struct xfs_fsmap		*rkey_low;	/* request keys */
+	struct xfs_fsmap		*rkey_high;
+	struct xfs_fsmap		dkeys[2];	/* per-dev keys */
+	struct xfs_getfsmap_dev		handlers[XFS_GETFSMAP_DEVS];
+	struct xfs_getfsmap_info	info = {0};
+	int				i;
+	int				error = 0;
+
+	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+		return -EOPNOTSUPP;
+	if (head->fmh_iflags & ~FMH_IF_VALID)
+		return -EINVAL;
+	rkey_low = head->fmh_keys;
+	rkey_high = rkey_low + 1;
+	if (!xfs_getfsmap_is_valid_device(mp, rkey_low) ||
+	    !xfs_getfsmap_is_valid_device(mp, rkey_high))
+		return -EINVAL;
+
+	head->fmh_entries = 0;
+
+	/* Set up our device handlers. */
+	memset(handlers, 0, sizeof(handlers));
+	handlers[0].dev = new_encode_dev(mp->m_ddev_targp->bt_dev);
+	handlers[0].fn = xfs_getfsmap_datadev;
+	if (mp->m_logdev_targp != mp->m_ddev_targp) {
+		handlers[1].dev = new_encode_dev(mp->m_logdev_targp->bt_dev);
+		handlers[1].fn = xfs_getfsmap_logdev;
+	}
+
+	xfs_sort(handlers, XFS_GETFSMAP_DEVS, sizeof(struct xfs_getfsmap_dev),
+			xfs_getfsmap_dev_compare);
+
+	/*
+	 * Since we allow the user to copy the last mapping from a previous
+	 * call into the low key slot, we have to advance the low key by
+	 * whatever the reported length is.  If the offset field doesn't apply,
+	 * move up the start block to the next extent and start over with the
+	 * lowest owner/offset possible; otherwise it's file data, so move up
+	 * the offset only.
+	 */
+	dkeys[0] = *rkey_low;
+	if (dkeys[0].fmr_flags & (FMR_OF_SPECIAL_OWNER | FMR_OF_EXTENT_MAP)) {
+		dkeys[0].fmr_physical += dkeys[0].fmr_length;
+		dkeys[0].fmr_owner = 0;
+		dkeys[0].fmr_offset = 0;
+	} else
+		dkeys[0].fmr_offset += dkeys[0].fmr_length;
+	memset(&dkeys[1], 0xFF, sizeof(struct xfs_fsmap));
+
+	if (!xfs_getfsmap_check_keys(dkeys, rkey_high))
+		return -EINVAL;
+
+	info.rkey_low = rkey_low;
+	info.formatter = formatter;
+	info.format_arg = arg;
+	info.head = head;
+
+	/* For each device we support... */
+	for (i = 0; i < XFS_GETFSMAP_DEVS; i++) {
+		/* Is this device within the range the user asked for? */
+		if (!handlers[i].fn)
+			continue;
+		if (rkey_low->fmr_device > handlers[i].dev)
+			continue;
+		if (rkey_high->fmr_device < handlers[i].dev)
+			break;
+
+		/*
+		 * If this device number matches the high key, we have
+		 * to pass the high key to the handler to limit the
+		 * query results.  If the device number exceeds the
+		 * low key, zero out the low key so that we get
+		 * everything from the beginning.
+		 */
+		if (handlers[i].dev == rkey_high->fmr_device)
+			dkeys[1] = *rkey_high;
+		if (handlers[i].dev > rkey_low->fmr_device)
+			memset(&dkeys[0], 0, sizeof(struct xfs_fsmap));
+
+		error = xfs_trans_alloc_empty(mp, &tp);
+		if (error)
+			break;
+
+		info.next_daddr = dkeys[0].fmr_physical;
+		info.dev = handlers[i].dev;
+		info.last = false;
+		info.agno = NULLAGNUMBER;
+		error = handlers[i].fn(tp, dkeys, &info);
+		if (error)
+			break;
+		xfs_trans_cancel(tp);
+		tp = NULL;
+	}
+
+	if (tp)
+		xfs_trans_cancel(tp);
+	head->fmh_oflags = FMH_OF_DEV_T;
+	return error;
+}
diff --git a/fs/xfs/xfs_fsmap.h b/fs/xfs/xfs_fsmap.h
new file mode 100644
index 0000000..1943047
--- /dev/null
+++ b/fs/xfs/xfs_fsmap.h
@@ -0,0 +1,51 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#ifndef __XFS_FSMAP_H__
+#define	__XFS_FSMAP_H__
+
+/* internal fsmap representation */
+struct xfs_fsmap {
+	dev_t		fmr_device;	/* device id */
+	uint32_t	fmr_flags;	/* mapping flags */
+	uint64_t	fmr_physical;	/* device offset of segment */
+	uint64_t	fmr_owner;	/* owner id */
+	xfs_fileoff_t	fmr_offset;	/* file offset of segment */
+	xfs_filblks_t	fmr_length;	/* length of segment, blocks */
+};
+
+struct xfs_fsmap_head {
+	uint32_t	fmh_iflags;	/* control flags */
+	uint32_t	fmh_oflags;	/* output flags */
+	unsigned int	fmh_count;	/* # of entries in array incl. input */
+	unsigned int	fmh_entries;	/* # of entries filled in (output). */
+
+	struct xfs_fsmap fmh_keys[2];	/* low and high keys */
+};
+
+void xfs_fsmap_from_internal(struct fsmap *dest, struct xfs_fsmap *src);
+void xfs_fsmap_to_internal(struct xfs_fsmap *dest, struct fsmap *src);
+
+/* fsmap to userspace formatter - copy to user & advance pointer */
+typedef int (*xfs_fsmap_format_t)(struct xfs_fsmap *, void *);
+
+int xfs_getfsmap(struct xfs_mount *mp, struct xfs_fsmap_head *head,
+		xfs_fsmap_format_t formatter, void *arg);
+
+#endif /* __XFS_FSMAP_H__ */
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index c67cfb4..d39da5d 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -41,6 +41,8 @@
 #include "xfs_trans.h"
 #include "xfs_pnfs.h"
 #include "xfs_acl.h"
+#include "xfs_btree.h"
+#include "xfs_fsmap.h"
 
 #include <linux/capability.h>
 #include <linux/dcache.h>
@@ -1607,6 +1609,103 @@ xfs_ioc_getbmapx(
 	return 0;
 }
 
+struct getfsmap_info {
+	struct xfs_mount	*mp;
+	struct fsmap __user	*data;
+	__u32			last_flags;
+};
+
+STATIC int
+xfs_getfsmap_format(struct xfs_fsmap *xfm, void *priv)
+{
+	struct getfsmap_info	*info = priv;
+	struct fsmap		fm;
+
+	trace_xfs_getfsmap_mapping(info->mp, xfm->fmr_device, xfm->fmr_physical,
+			xfm->fmr_length, xfm->fmr_owner, xfm->fmr_offset,
+			xfm->fmr_flags);
+
+	info->last_flags = xfm->fmr_flags;
+	xfs_fsmap_from_internal(&fm, xfm);
+	if (copy_to_user(info->data, &fm, sizeof(struct fsmap)))
+		return -EFAULT;
+
+	info->data++;
+	return 0;
+}
+
+STATIC int
+xfs_ioc_getfsmap(
+	struct xfs_inode	*ip,
+	void			__user *arg)
+{
+	struct getfsmap_info	info;
+	struct xfs_fsmap_head	xhead = {0};
+	struct fsmap_head	head;
+	bool			aborted = false;
+	int			error;
+
+	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
+		return -EFAULT;
+	if (head.fmh_reserved[0] || head.fmh_reserved[1] ||
+	    head.fmh_reserved[2] || head.fmh_reserved[3] ||
+	    head.fmh_reserved[4] || head.fmh_reserved[5] ||
+	    head.fmh_keys[0].fmr_reserved[0] ||
+	    head.fmh_keys[0].fmr_reserved[1] ||
+	    head.fmh_keys[0].fmr_reserved[2] ||
+	    head.fmh_keys[1].fmr_reserved[0] ||
+	    head.fmh_keys[1].fmr_reserved[1] ||
+	    head.fmh_keys[1].fmr_reserved[2])
+		return -EINVAL;
+
+	xhead.fmh_iflags = head.fmh_iflags;
+	xhead.fmh_count = head.fmh_count;
+	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
+	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
+
+	trace_xfs_getfsmap_low_key(ip->i_mount,
+			xhead.fmh_keys[0].fmr_device,
+			xhead.fmh_keys[0].fmr_physical,
+			xhead.fmh_keys[0].fmr_length,
+			xhead.fmh_keys[0].fmr_owner,
+			xhead.fmh_keys[0].fmr_offset,
+			xhead.fmh_keys[0].fmr_flags);
+
+	trace_xfs_getfsmap_high_key(ip->i_mount,
+			xhead.fmh_keys[1].fmr_device,
+			xhead.fmh_keys[1].fmr_physical,
+			xhead.fmh_keys[1].fmr_length,
+			xhead.fmh_keys[1].fmr_owner,
+			xhead.fmh_keys[1].fmr_offset,
+			xhead.fmh_keys[1].fmr_flags);
+
+	info.mp = ip->i_mount;
+	info.data = ((__force struct fsmap_head *)arg)->fmh_recs;
+	error = xfs_getfsmap(ip->i_mount, &xhead, xfs_getfsmap_format, &info);
+	if (error == XFS_BTREE_QUERY_RANGE_ABORT) {
+		error = 0;
+		aborted = true;
+	} else if (error)
+		return error;
+
+	/* If we didn't abort, set the "last" flag in the last fmx */
+	if (!aborted && xhead.fmh_entries) {
+		info.data--;
+		info.last_flags |= FMR_OF_LAST;
+		if (copy_to_user(&info.data->fmr_flags, &info.last_flags,
+				sizeof(info.last_flags)))
+			return -EFAULT;
+	}
+
+	/* copy back header */
+	head.fmh_entries = xhead.fmh_entries;
+	head.fmh_oflags = xhead.fmh_oflags;
+	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
+		return -EFAULT;
+
+	return 0;
+}
+
 int
 xfs_ioc_swapext(
 	xfs_swapext_t	*sxp)
@@ -1787,6 +1886,11 @@ xfs_file_ioctl(
 	case XFS_IOC_GETBMAPX:
 		return xfs_ioc_getbmapx(ip, arg);
 
+	case XFS_IOC_GETFSMAP:
+		if (!capable(CAP_SYS_ADMIN))
+			return -EPERM;
+		return xfs_ioc_getfsmap(ip, arg);
+
 	case XFS_IOC_FD_TO_HANDLE:
 	case XFS_IOC_PATH_TO_HANDLE:
 	case XFS_IOC_PATH_TO_FSHANDLE: {
diff --git a/fs/xfs/xfs_ioctl32.c b/fs/xfs/xfs_ioctl32.c
index 7c49938..5ba41b7 100644
--- a/fs/xfs/xfs_ioctl32.c
+++ b/fs/xfs/xfs_ioctl32.c
@@ -554,6 +554,7 @@ xfs_file_compat_ioctl(
 	case XFS_IOC_GOINGDOWN:
 	case XFS_IOC_ERROR_INJECTION:
 	case XFS_IOC_ERROR_CLEARALL:
+	case XFS_IOC_GETFSMAP:
 		return xfs_file_ioctl(filp, cmd, p);
 #ifndef BROKEN_X86_ALIGNMENT
 	/* These are handled fine if no alignment issues */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 69c5bcd..09f4755 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3266,6 +3266,91 @@ DEFINE_INODE_IREC_EVENT(xfs_swap_extent_rmap_remap);
 DEFINE_INODE_IREC_EVENT(xfs_swap_extent_rmap_remap_piece);
 DEFINE_INODE_ERROR_EVENT(xfs_swap_extent_rmap_error);
 
+/* fsmap traces */
+DECLARE_EVENT_CLASS(xfs_fsmap_class,
+	TP_PROTO(struct xfs_mount *mp, u32 keydev, xfs_agnumber_t agno,
+		 xfs_fsblock_t bno, xfs_filblks_t len, __uint64_t owner,
+		 __uint64_t offset),
+	TP_ARGS(mp, keydev, agno, bno, len, owner, offset),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(dev_t, keydev)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_fsblock_t, bno)
+		__field(xfs_filblks_t, len)
+		__field(__uint64_t, owner)
+		__field(__uint64_t, offset)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->keydev = new_decode_dev(keydev);
+		__entry->agno = agno;
+		__entry->bno = bno;
+		__entry->len = len;
+		__entry->owner = owner;
+		__entry->offset = offset;
+	),
+	TP_printk("dev %d:%d keydev %d:%d agno %u bno %llu len %llu owner %lld offset 0x%llx\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  MAJOR(__entry->keydev), MINOR(__entry->keydev),
+		  __entry->agno,
+		  __entry->bno,
+		  __entry->len,
+		  __entry->owner,
+		  __entry->offset)
+)
+#define DEFINE_FSMAP_EVENT(name) \
+DEFINE_EVENT(xfs_fsmap_class, name, \
+	TP_PROTO(struct xfs_mount *mp, u32 keydev, xfs_agnumber_t agno, \
+		 xfs_fsblock_t bno, xfs_filblks_t len, __uint64_t owner, \
+		 __uint64_t offset), \
+	TP_ARGS(mp, keydev, agno, bno, len, owner, offset))
+DEFINE_FSMAP_EVENT(xfs_fsmap_low_key);
+DEFINE_FSMAP_EVENT(xfs_fsmap_high_key);
+DEFINE_FSMAP_EVENT(xfs_fsmap_mapping);
+
+DECLARE_EVENT_CLASS(xfs_getfsmap_class,
+	TP_PROTO(struct xfs_mount *mp, u32 keydev, xfs_daddr_t block,
+		 xfs_daddr_t len, __uint64_t owner, __uint64_t offset,
+		 __uint64_t flags),
+	TP_ARGS(mp, keydev, block, len, owner, offset, flags),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(dev_t, keydev)
+		__field(xfs_daddr_t, block)
+		__field(xfs_daddr_t, len)
+		__field(__uint64_t, owner)
+		__field(__uint64_t, offset)
+		__field(__uint64_t, flags)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->keydev = new_decode_dev(keydev);
+		__entry->block = block;
+		__entry->len = len;
+		__entry->owner = owner;
+		__entry->offset = offset;
+		__entry->flags = flags;
+	),
+	TP_printk("dev %d:%d keydev %d:%d block %llu len %llu owner %lld offset %llu flags 0x%llx\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  MAJOR(__entry->keydev), MINOR(__entry->keydev),
+		  __entry->block,
+		  __entry->len,
+		  __entry->owner,
+		  __entry->offset,
+		  __entry->flags)
+)
+#define DEFINE_GETFSMAP_EVENT(name) \
+DEFINE_EVENT(xfs_getfsmap_class, name, \
+	TP_PROTO(struct xfs_mount *mp, u32 keydev, xfs_daddr_t block, \
+		 xfs_daddr_t len, __uint64_t owner, __uint64_t offset, \
+		 __uint64_t flags), \
+	TP_ARGS(mp, keydev, block, len, owner, offset, flags))
+DEFINE_GETFSMAP_EVENT(xfs_getfsmap_low_key);
+DEFINE_GETFSMAP_EVENT(xfs_getfsmap_high_key);
+DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
+
 #endif /* _TRACE_XFS_H */
 
 #undef TRACE_INCLUDE_PATH
diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c
index 70f42ea..a280e12 100644
--- a/fs/xfs/xfs_trans.c
+++ b/fs/xfs/xfs_trans.c
@@ -263,6 +263,28 @@ xfs_trans_alloc(
 }
 
 /*
+ * Create an empty transaction with no reservation.  This is a defensive
+ * mechanism for routines that query metadata without actually modifying
+ * them -- if the metadata being queried is somehow cross-linked (think a
+ * btree block pointer that points higher in the tree), we risk deadlock.
+ * However, blocks grabbed as part of a transaction can be re-grabbed.
+ * The verifiers will notice the corrupt block and the operation will fail
+ * back to userspace without deadlocking.
+ *
+ * Note the zero-length reservation; this transaction MUST be cancelled
+ * without any dirty data.
+ */
+int
+xfs_trans_alloc_empty(
+	struct xfs_mount		*mp,
+	struct xfs_trans		**tpp)
+{
+	struct xfs_trans_res		resv = {0};
+
+	return xfs_trans_alloc(mp, &resv, 0, 0, XFS_TRANS_NO_WRITECOUNT, tpp);
+}
+
+/*
  * Record the indicated change to the given field for application
  * to the file system's superblock when the transaction commits.
  * For now, just store the change in the transaction structure.
diff --git a/fs/xfs/xfs_trans.h b/fs/xfs/xfs_trans.h
index 61b7fbd..98024cb 100644
--- a/fs/xfs/xfs_trans.h
+++ b/fs/xfs/xfs_trans.h
@@ -159,6 +159,8 @@ typedef struct xfs_trans {
 int		xfs_trans_alloc(struct xfs_mount *mp, struct xfs_trans_res *resp,
 			uint blocks, uint rtextents, uint flags,
 			struct xfs_trans **tpp);
+int		xfs_trans_alloc_empty(struct xfs_mount *mp,
+			struct xfs_trans **tpp);
 void		xfs_trans_mod_sb(xfs_trans_t *, uint, int64_t);
 
 struct xfs_buf	*xfs_trans_get_buf_map(struct xfs_trans *tp,

[PATCH 10/55] xfs: report shared extents in getfsmapx

[Darrick J. Wong]

Cross-reference the reverse mapping data with the refcount btree to find
out which extents are shared.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/xfs_fsmap.c |   46 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 46 insertions(+)


diff --git a/fs/xfs/xfs_fsmap.c b/fs/xfs/xfs_fsmap.c
index 985415b..d729581 100644
--- a/fs/xfs/xfs_fsmap.c
+++ b/fs/xfs/xfs_fsmap.c
@@ -37,6 +37,8 @@
 #include "xfs_alloc.h"
 #include "xfs_bit.h"
 #include "xfs_fsmap.h"
+#include "xfs_refcount.h"
+#include "xfs_refcount_btree.h"
 
 /* Convert an xfs_fsmap to an fsmap. */
 void
@@ -192,6 +194,42 @@ xfs_getfsmap_rec_before_low_key(
 	return false;
 }
 
+/* Decide if this mapping is shared. */
+STATIC int
+xfs_getfsmap_is_shared(
+	struct xfs_mount		*mp,
+	struct xfs_getfsmap_info	*info,
+	struct xfs_rmap_irec		*rec,
+	bool				*stat)
+{
+	struct xfs_btree_cur		*cur;
+	xfs_agblock_t			fbno;
+	xfs_extlen_t			flen;
+	int				error;
+
+	*stat = false;
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return 0;
+	/* rt files will have agno set to NULLAGNUMBER */
+	if (info->agno == NULLAGNUMBER)
+		return 0;
+
+	/* Are there any shared blocks here? */
+	flen = 0;
+	cur = xfs_refcountbt_init_cursor(mp, NULL, info->agf_bp,
+			info->agno, NULL);
+
+	error = xfs_refcount_find_shared(cur, rec->rm_startblock,
+			rec->rm_blockcount, &fbno, &flen, false);
+
+	xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
+	if (error)
+		return error;
+
+	*stat = flen > 0;
+	return 0;
+}
+
 /*
  * Format a reverse mapping for getfsmap, having translated rm_startblock
  * into the appropriate daddr units.
@@ -205,6 +243,7 @@ xfs_getfsmap_helper(
 {
 	struct xfs_fsmap		fmr;
 	xfs_daddr_t			key_end;
+	bool				shared;
 	int				error;
 
 	/*
@@ -304,6 +343,13 @@ xfs_getfsmap_helper(
 		fmr.fmr_flags |= FMR_OF_ATTR_FORK;
 	if (rec->rm_flags & XFS_RMAP_BMBT_BLOCK)
 		fmr.fmr_flags |= FMR_OF_EXTENT_MAP;
+	if (fmr.fmr_flags == 0) {
+		error = xfs_getfsmap_is_shared(mp, info, rec, &shared);
+		if (error)
+			return error;
+		if (shared)
+			fmr.fmr_flags |= FMR_OF_SHARED;
+	}
 	error = info->formatter(&fmr, info->format_arg);
 	if (error)
 		return error;

[PATCH 11/55] xfs: have getfsmap fall back to the freesp btrees when rmap is not present

[Darrick J. Wong]

If the reverse-mapping btree isn't available, fall back to the
free space btrees to provide partial reverse mapping information.
The online scrub tool can make use of even partial information to
speed up the data block scan.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/xfs_fsmap.c |  153 +++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 150 insertions(+), 3 deletions(-)


diff --git a/fs/xfs/xfs_fsmap.c b/fs/xfs/xfs_fsmap.c
index d729581..4e364f8 100644
--- a/fs/xfs/xfs_fsmap.c
+++ b/fs/xfs/xfs_fsmap.c
@@ -39,6 +39,7 @@
 #include "xfs_fsmap.h"
 #include "xfs_refcount.h"
 #include "xfs_refcount_btree.h"
+#include "xfs_alloc_btree.h"
 
 /* Convert an xfs_fsmap to an fsmap. */
 void
@@ -125,6 +126,7 @@ xfs_fsmap_owner_from_rmap(
 	case XFS_RMAP_OWN_INODES:
 	case XFS_RMAP_OWN_REFC:
 	case XFS_RMAP_OWN_COW:
+	case XFS_RMAP_OWN_NULL:	/* "free" */
 		fmr->fmr_owner = rm->rm_owner;
 		return 0;
 	default:
@@ -396,6 +398,31 @@ xfs_getfsmap_rtdev_helper(
 	return xfs_getfsmap_helper(mp, info, rec, rec_daddr);
 }
 
+/* Transform a bnobt irec into a fsmap */
+STATIC int
+xfs_getfsmap_datadev_bnobt_helper(
+	struct xfs_btree_cur		*cur,
+	struct xfs_alloc_rec_incore	*rec,
+	void				*priv)
+{
+	struct xfs_mount		*mp = cur->bc_mp;
+	struct xfs_getfsmap_info	*info = priv;
+	struct xfs_rmap_irec		irec;
+	xfs_fsblock_t			fsb;
+	xfs_daddr_t			rec_daddr;
+
+	fsb = XFS_AGB_TO_FSB(mp, cur->bc_private.a.agno, rec->ar_startblock);
+	rec_daddr = XFS_FSB_TO_DADDR(mp, fsb);
+
+	irec.rm_startblock = rec->ar_startblock;
+	irec.rm_blockcount = rec->ar_blockcount;
+	irec.rm_owner = XFS_RMAP_OWN_NULL;	/* "free" */
+	irec.rm_offset = 0;
+	irec.rm_flags = 0;
+
+	return xfs_getfsmap_helper(mp, info, &irec, rec_daddr);
+}
+
 /* Set rmap flags based on the getfsmap flags */
 static void
 xfs_getfsmap_set_irec_flags(
@@ -583,6 +610,125 @@ xfs_getfsmap_datadev(
 	return error;
 }
 
+/* Execute a getfsmap query against the regular data device's bnobt. */
+STATIC int
+xfs_getfsmap_datadev_bnobt(
+	struct xfs_trans		*tp,
+	struct xfs_fsmap		*keys,
+	struct xfs_getfsmap_info	*info)
+{
+	struct xfs_mount		*mp = tp->t_mountp;
+	struct xfs_btree_cur		*bt_cur = NULL;
+	struct xfs_fsmap		*dkey_low;
+	struct xfs_fsmap		*dkey_high;
+	struct xfs_alloc_rec_incore	alow;
+	struct xfs_alloc_rec_incore	ahigh;
+	xfs_fsblock_t			start_fsb;
+	xfs_fsblock_t			end_fsb;
+	xfs_agnumber_t			start_ag;
+	xfs_agnumber_t			end_ag;
+	xfs_daddr_t			eofs;
+	int				error = 0;
+
+	dkey_low = keys;
+	dkey_high = keys + 1;
+	eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
+	if (dkey_low->fmr_physical >= eofs)
+		return 0;
+	if (dkey_high->fmr_physical >= eofs)
+		dkey_high->fmr_physical = eofs - 1;
+	start_fsb = XFS_DADDR_TO_FSB(mp, dkey_low->fmr_physical);
+	end_fsb = XFS_DADDR_TO_FSB(mp, dkey_high->fmr_physical);
+
+	/* Set up search keys */
+	info->low.rm_startblock = XFS_FSB_TO_AGBNO(mp, start_fsb);
+	info->low.rm_offset = XFS_BB_TO_FSBT(mp, dkey_low->fmr_offset);
+	error = xfs_fsmap_owner_to_rmap(dkey_low, &info->low);
+	if (error)
+		return error;
+	info->low.rm_blockcount = 0;
+	xfs_getfsmap_set_irec_flags(&info->low, dkey_low);
+
+	info->high.rm_startblock = -1U;
+	info->high.rm_owner = ULLONG_MAX;
+	info->high.rm_offset = ULLONG_MAX;
+	info->high.rm_blockcount = 0;
+	info->high.rm_flags = XFS_RMAP_KEY_FLAGS | XFS_RMAP_REC_FLAGS;
+
+	info->missing_owner = FMR_OWN_UNKNOWN;
+
+	start_ag = XFS_FSB_TO_AGNO(mp, start_fsb);
+	end_ag = XFS_FSB_TO_AGNO(mp, end_fsb);
+
+	/* Query each AG */
+	for (info->agno = start_ag; info->agno <= end_ag; info->agno++) {
+		if (info->agno == end_ag) {
+			info->high.rm_startblock = XFS_FSB_TO_AGBNO(mp,
+					end_fsb);
+			info->high.rm_offset = XFS_BB_TO_FSBT(mp,
+					dkey_high->fmr_offset);
+			error = xfs_fsmap_owner_to_rmap(dkey_high, &info->high);
+			if (error)
+				goto err;
+			xfs_getfsmap_set_irec_flags(&info->high, dkey_high);
+		}
+
+		if (bt_cur) {
+			xfs_btree_del_cursor(bt_cur, XFS_BTREE_NOERROR);
+			bt_cur = NULL;
+			info->agf_bp = NULL;
+		}
+
+		error = xfs_alloc_read_agf(mp, tp, info->agno, 0,
+				&info->agf_bp);
+		if (error)
+			goto err;
+
+		trace_xfs_fsmap_low_key(mp, info->dev, info->agno,
+				info->low.rm_startblock,
+				info->low.rm_blockcount,
+				info->low.rm_owner,
+				info->low.rm_offset);
+
+		trace_xfs_fsmap_high_key(mp, info->dev, info->agno,
+				info->high.rm_startblock,
+				info->high.rm_blockcount,
+				info->high.rm_owner,
+				info->high.rm_offset);
+
+		bt_cur = xfs_allocbt_init_cursor(mp, tp, info->agf_bp,
+				info->agno, XFS_BTNUM_BNO);
+		alow.ar_startblock = info->low.rm_startblock;
+		ahigh.ar_startblock = info->high.rm_startblock;
+		error = xfs_alloc_query_range(bt_cur, &alow, &ahigh,
+				xfs_getfsmap_datadev_bnobt_helper, info);
+		if (error)
+			goto err;
+
+		if (info->agno == start_ag) {
+			info->low.rm_startblock = 0;
+			info->low.rm_owner = 0;
+			info->low.rm_offset = 0;
+			info->low.rm_flags = 0;
+		}
+	}
+
+	/* Report any free space at the end of the AG */
+	info->last = true;
+	error = xfs_getfsmap_datadev_bnobt_helper(bt_cur, &ahigh, info);
+	if (error)
+		goto err;
+
+err:
+	if (bt_cur)
+		xfs_btree_del_cursor(bt_cur, error < 0 ? XFS_BTREE_ERROR :
+							 XFS_BTREE_NOERROR);
+	if (info->agf_bp)
+		info->agf_bp = NULL;
+
+	return error;
+}
+
 /* Do we recognize the device? */
 STATIC bool
 xfs_getfsmap_is_valid_device(
@@ -651,8 +797,6 @@ xfs_getfsmap(
 	int				i;
 	int				error = 0;
 
-	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
-		return -EOPNOTSUPP;
 	if (head->fmh_iflags & ~FMH_IF_VALID)
 		return -EINVAL;
 	rkey_low = head->fmh_keys;
@@ -666,7 +810,10 @@ xfs_getfsmap(
 	/* Set up our device handlers. */
 	memset(handlers, 0, sizeof(handlers));
 	handlers[0].dev = new_encode_dev(mp->m_ddev_targp->bt_dev);
-	handlers[0].fn = xfs_getfsmap_datadev;
+	if (xfs_sb_version_hasrmapbt(&mp->m_sb))
+		handlers[0].fn = xfs_getfsmap_datadev;
+	else
+		handlers[0].fn = xfs_getfsmap_datadev_bnobt;
 	if (mp->m_logdev_targp != mp->m_ddev_targp) {
 		handlers[1].dev = new_encode_dev(mp->m_logdev_targp->bt_dev);
 		handlers[1].fn = xfs_getfsmap_logdev;

[PATCH 12/55] xfs: getfsmap should fall back to rtbitmap when rtrmapbt not present

[Darrick J. Wong]

Use the realtime bitmap to return freespace information when the
rtrmapbt isn't present.  Note that the rtrmapbt fsmap implementation
will show up later with the rtrmapbt patchset.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/xfs_fsmap.c   |  133 ++++++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_rtalloc.h |    2 +
 2 files changed, 135 insertions(+)


diff --git a/fs/xfs/xfs_fsmap.c b/fs/xfs/xfs_fsmap.c
index 4e364f8..ba5544b 100644
--- a/fs/xfs/xfs_fsmap.c
+++ b/fs/xfs/xfs_fsmap.c
@@ -40,6 +40,7 @@
 #include "xfs_refcount.h"
 #include "xfs_refcount_btree.h"
 #include "xfs_alloc_btree.h"
+#include "xfs_rtalloc.h"
 
 /* Convert an xfs_fsmap to an fsmap. */
 void
@@ -398,6 +399,29 @@ xfs_getfsmap_rtdev_helper(
 	return xfs_getfsmap_helper(mp, info, rec, rec_daddr);
 }
 
+/* Transform a rtbitmap "record" into a fsmap */
+STATIC int
+xfs_getfsmap_rtdev_rtbitmap_helper(
+	struct xfs_mount		*mp,
+	xfs_rtblock_t			start,
+	xfs_rtblock_t			end,
+	void				*priv)
+{
+	struct xfs_getfsmap_info	*info = priv;
+	struct xfs_rmap_irec		irec;
+	xfs_daddr_t			rec_daddr;
+
+	rec_daddr = XFS_FSB_TO_BB(mp, start);
+
+	irec.rm_startblock = start;
+	irec.rm_blockcount = end - start + 1;
+	irec.rm_owner = XFS_RMAP_OWN_NULL;	/* "free" */
+	irec.rm_offset = 0;
+	irec.rm_flags = 0;
+
+	return xfs_getfsmap_helper(mp, info, &irec, rec_daddr);
+}
+
 /* Transform a bnobt irec into a fsmap */
 STATIC int
 xfs_getfsmap_datadev_bnobt_helper(
@@ -496,6 +520,108 @@ xfs_getfsmap_logdev(
 	return xfs_getfsmap_rtdev_helper(&cur, &rmap, info);
 }
 
+/* Execute a getfsmap query against the realtime data device (rtbitmap). */
+STATIC int
+xfs_getfsmap_rtdev_rtbitmap(
+	struct xfs_trans		*tp,
+	struct xfs_fsmap		*keys,
+	struct xfs_getfsmap_info	*info)
+{
+	struct xfs_mount		*mp = tp->t_mountp;
+	struct xfs_fsmap		*dkey_low;
+	struct xfs_fsmap		*dkey_high;
+	xfs_fsblock_t			start_fsb;
+	xfs_fsblock_t			end_fsb;
+	xfs_rtblock_t			rtstart;
+	xfs_rtblock_t			rtend;
+	xfs_rtblock_t			rem;
+	xfs_daddr_t			eofs;
+	int				is_free;
+	int				error = 0;
+
+	dkey_low = keys;
+	dkey_high = keys + 1;
+	eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_rblocks);
+	if (dkey_low->fmr_physical >= eofs)
+		return 0;
+	if (dkey_high->fmr_physical >= eofs)
+		dkey_high->fmr_physical = eofs - 1;
+	start_fsb = XFS_BB_TO_FSBT(mp, dkey_low->fmr_physical);
+	end_fsb = XFS_BB_TO_FSB(mp, dkey_high->fmr_physical);
+
+	/* Set up search keys */
+	info->low.rm_startblock = start_fsb;
+	error = xfs_fsmap_owner_to_rmap(dkey_low, &info->low);
+	if (error)
+		return error;
+	info->low.rm_offset = XFS_BB_TO_FSBT(mp, dkey_low->fmr_offset);
+	info->low.rm_blockcount = 0;
+	xfs_getfsmap_set_irec_flags(&info->low, dkey_low);
+
+	info->high.rm_startblock = end_fsb;
+	error = xfs_fsmap_owner_to_rmap(dkey_high, &info->high);
+	if (error)
+		return error;
+	info->high.rm_offset = XFS_BB_TO_FSBT(mp, dkey_high->fmr_offset);
+	info->high.rm_blockcount = 0;
+	xfs_getfsmap_set_irec_flags(&info->high, dkey_high);
+
+	info->missing_owner = FMR_OWN_UNKNOWN;
+
+	trace_xfs_fsmap_low_key(mp, info->dev, info->agno,
+			info->low.rm_startblock,
+			info->low.rm_blockcount,
+			info->low.rm_owner,
+			info->low.rm_offset);
+
+	trace_xfs_fsmap_high_key(mp, info->dev, info->agno,
+			info->high.rm_startblock,
+			info->high.rm_blockcount,
+			info->high.rm_owner,
+			info->high.rm_offset);
+
+	xfs_ilock(mp->m_rbmip, XFS_ILOCK_SHARED | XFS_ILOCK_RTBITMAP);
+
+	/* Iterate the bitmap, looking for discrepancies. */
+	rtstart = 0;
+	rem = mp->m_sb.sb_rblocks;
+	while (rem) {
+		/* Is the first block free? */
+		error = xfs_rtcheck_range(mp, tp, rtstart, 1, 1, &rtend,
+				&is_free);
+		if (error)
+			goto out_unlock;
+
+		/* How long does the extent go for? */
+		error = xfs_rtfind_forw(mp, tp, rtstart,
+				mp->m_sb.sb_rblocks - 1, &rtend);
+		if (error)
+			goto out_unlock;
+
+		if (is_free) {
+			error = xfs_getfsmap_rtdev_rtbitmap_helper(mp,
+					rtstart, rtend, info);
+			if (error)
+				goto out_unlock;
+		}
+
+		rem -= rtend - rtstart + 1;
+		rtstart = rtend + 1;
+	}
+
+out_unlock:
+	xfs_iunlock(mp->m_rbmip, XFS_ILOCK_SHARED | XFS_ILOCK_RTBITMAP);
+
+	/* Report any free space at the end of the rtdev */
+	info->last = true;
+	error = xfs_getfsmap_rtdev_rtbitmap_helper(mp, end_fsb, 0, info);
+	if (error)
+		goto err;
+
+err:
+	return error;
+}
+
 /* Execute a getfsmap query against the regular data device. */
 STATIC int
 xfs_getfsmap_datadev(
@@ -741,6 +867,9 @@ xfs_getfsmap_is_valid_device(
 	if (mp->m_logdev_targp &&
 	    fm->fmr_device == new_encode_dev(mp->m_logdev_targp->bt_dev))
 		return true;
+	if (mp->m_rtdev_targp &&
+	    fm->fmr_device == new_encode_dev(mp->m_rtdev_targp->bt_dev))
+		return true;
 	return false;
 }
 
@@ -818,6 +947,10 @@ xfs_getfsmap(
 		handlers[1].dev = new_encode_dev(mp->m_logdev_targp->bt_dev);
 		handlers[1].fn = xfs_getfsmap_logdev;
 	}
+	if (mp->m_rtdev_targp) {
+		handlers[2].dev = new_encode_dev(mp->m_rtdev_targp->bt_dev);
+		handlers[2].fn = xfs_getfsmap_rtdev_rtbitmap;
+	}
 
 	xfs_sort(handlers, XFS_GETFSMAP_DEVS, sizeof(struct xfs_getfsmap_dev),
 			xfs_getfsmap_dev_compare);
diff --git a/fs/xfs/xfs_rtalloc.h b/fs/xfs/xfs_rtalloc.h
index 355dd9e..f798a3e 100644
--- a/fs/xfs/xfs_rtalloc.h
+++ b/fs/xfs/xfs_rtalloc.h
@@ -126,6 +126,8 @@ int xfs_rtfree_range(struct xfs_mount *mp, struct xfs_trans *tp,
 # define xfs_rtfree_extent(t,b,l)                       (ENOSYS)
 # define xfs_rtpick_extent(m,t,l,rb)                    (ENOSYS)
 # define xfs_growfs_rt(mp,in)                           (ENOSYS)
+# define xfs_rtcheck_range(...)                         (ENOSYS)
+# define xfs_rtfind_forw(...)                           (ENOSYS)
 static inline int		/* error */
 xfs_rtmount_init(
 	xfs_mount_t	*mp)	/* file system mount structure */

[PATCH 13/55] xfs: query the per-AG reservation counters

[Darrick J. Wong]

Establish an ioctl for userspace to query the original and current
per-AG reservation counts.  This will be used by xfs_scrub to
check that the vfs counters are at least somewhat sane.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_fs.h |   10 ++++++++++
 fs/xfs/xfs_fsops.c     |   29 +++++++++++++++++++++++++++++
 fs/xfs/xfs_fsops.h     |    2 ++
 fs/xfs/xfs_ioctl.c     |   16 ++++++++++++++++
 fs/xfs/xfs_ioctl32.c   |    1 +
 5 files changed, 58 insertions(+)


diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index e62996f..10aa908 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -555,6 +555,15 @@ typedef struct xfs_swapext
 #define XFS_FSOP_GOING_FLAGS_NOLOGFLUSH		0x2	/* don't flush log nor data */
 
 /*
+ * AG reserved block counters
+ */
+struct xfs_fsop_ag_resblks {
+	__u64 resblks;		/* blocks reserved now */
+	__u64 resblks_orig;	/* blocks reserved at mount time */
+	__u64 reserved[2];
+};
+
+/*
  * ioctl limits
  */
 #ifdef XATTR_LIST_MAX
@@ -629,6 +638,7 @@ typedef struct xfs_swapext
 #define XFS_IOC_ATTRMULTI_BY_HANDLE  _IOW ('X', 123, struct xfs_fsop_attrmulti_handlereq)
 #define XFS_IOC_FSGEOMETRY	     _IOR ('X', 124, struct xfs_fsop_geom)
 #define XFS_IOC_GOINGDOWN	     _IOR ('X', 125, __uint32_t)
+#define XFS_IOC_GET_AG_RESBLKS	     _IOR ('X', 126, struct xfs_fsop_ag_resblks)
 /*	XFS_IOC_GETFSUUID ---------- deprecated 140	 */
 
 
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index 242e809..70b50f2 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -44,6 +44,7 @@
 #include "xfs_filestream.h"
 #include "xfs_rmap.h"
 #include "xfs_ag_resv.h"
+#include "xfs_fs.h"
 
 /*
  * File system operations
@@ -1067,3 +1068,31 @@ xfs_fs_unreserve_ag_blocks(
 
 	return error;
 }
+
+/* Query the per-AG reservations to see how many blocks we have reserved. */
+int
+xfs_fs_get_ag_reserve_blocks(
+	struct xfs_mount		*mp,
+	struct xfs_fsop_ag_resblks	*out)
+{
+	struct xfs_ag_resv		*r;
+	struct xfs_perag		*pag;
+	xfs_agnumber_t			agno;
+
+	out->resblks = 0;
+	out->resblks_orig = 0;
+	out->reserved[0] = out->reserved[1] = 0;
+
+	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
+		pag = xfs_perag_get(mp, agno);
+		r = xfs_perag_resv(pag, XFS_AG_RESV_METADATA);
+		out->resblks += r->ar_reserved;
+		out->resblks_orig += r->ar_asked;
+		r = xfs_perag_resv(pag, XFS_AG_RESV_AGFL);
+		out->resblks += r->ar_reserved;
+		out->resblks_orig += r->ar_asked;
+		xfs_perag_put(pag);
+	}
+
+	return 0;
+}
diff --git a/fs/xfs/xfs_fsops.h b/fs/xfs/xfs_fsops.h
index f349158..91609ae 100644
--- a/fs/xfs/xfs_fsops.h
+++ b/fs/xfs/xfs_fsops.h
@@ -25,6 +25,8 @@ extern int xfs_fs_counts(xfs_mount_t *mp, xfs_fsop_counts_t *cnt);
 extern int xfs_reserve_blocks(xfs_mount_t *mp, __uint64_t *inval,
 				xfs_fsop_resblks_t *outval);
 extern int xfs_fs_goingdown(xfs_mount_t *mp, __uint32_t inflags);
+extern int xfs_fs_get_ag_reserve_blocks(struct xfs_mount *mp,
+		struct xfs_fsop_ag_resblks *out);
 
 extern int xfs_fs_reserve_ag_blocks(struct xfs_mount *mp);
 extern int xfs_fs_unreserve_ag_blocks(struct xfs_mount *mp);
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index d39da5d..347f726 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -1993,6 +1993,22 @@ xfs_file_ioctl(
 		return 0;
 	}
 
+	case XFS_IOC_GET_AG_RESBLKS: {
+		struct xfs_fsop_ag_resblks	out;
+
+		if (!capable(CAP_SYS_ADMIN))
+			return -EPERM;
+
+		error = xfs_fs_get_ag_reserve_blocks(mp, &out);
+		if (error)
+			return error;
+
+		if (copy_to_user(arg, &out, sizeof(out)))
+			return -EFAULT;
+
+		return 0;
+	}
+
 	case XFS_IOC_FSGROWFSDATA: {
 		xfs_growfs_data_t in;
 
diff --git a/fs/xfs/xfs_ioctl32.c b/fs/xfs/xfs_ioctl32.c
index 5ba41b7..da479e9 100644
--- a/fs/xfs/xfs_ioctl32.c
+++ b/fs/xfs/xfs_ioctl32.c
@@ -555,6 +555,7 @@ xfs_file_compat_ioctl(
 	case XFS_IOC_ERROR_INJECTION:
 	case XFS_IOC_ERROR_CLEARALL:
 	case XFS_IOC_GETFSMAP:
+	case XFS_IOC_GET_AG_RESBLKS:
 		return xfs_file_ioctl(filp, cmd, p);
 #ifndef BROKEN_X86_ALIGNMENT
 	/* These are handled fine if no alignment issues */

[PATCH 14/55] xfs: add scrub tracepoints

[Darrick J. Wong]

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_types.h |    5 +
 fs/xfs/xfs_trace.h        |  372 +++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 377 insertions(+)


diff --git a/fs/xfs/libxfs/xfs_types.h b/fs/xfs/libxfs/xfs_types.h
index 717909f..04145e5 100644
--- a/fs/xfs/libxfs/xfs_types.h
+++ b/fs/xfs/libxfs/xfs_types.h
@@ -94,6 +94,11 @@ typedef __int64_t	xfs_sfiloff_t;	/* signed block number in a file */
 #define	XFS_ATTR_FORK	1
 #define	XFS_COW_FORK	2
 
+#define XFS_FORK_DESC \
+	{ XFS_DATA_FORK,	"data" }, \
+	{ XFS_ATTR_FORK,	"attr" }, \
+	{ XFS_COW_FORK,		"CoW" }
+
 /*
  * Min numbers of data/attr fork btree root pointers.
  */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 09f4755..6427c70 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3351,6 +3351,378 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_low_key);
 DEFINE_GETFSMAP_EVENT(xfs_getfsmap_high_key);
 DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 
+/* scrub */
+#define XFS_SCRUB_TYPE_DESC \
+	{ 0, NULL }
+DECLARE_EVENT_CLASS(xfs_scrub_class,
+	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
+		 xfs_ino_t inum, unsigned int gen, unsigned int flags,
+		 int error),
+	TP_ARGS(ip, type, agno, inum, gen, flags, error),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_ino_t, ino)
+		__field(int, type)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_ino_t, inum)
+		__field(unsigned int, gen)
+		__field(unsigned int, flags)
+		__field(int, error)
+	),
+	TP_fast_assign(
+		__entry->dev = ip->i_mount->m_super->s_dev;
+		__entry->ino = ip->i_ino;
+		__entry->type = type;
+		__entry->agno = agno;
+		__entry->inum = inum;
+		__entry->gen = gen;
+		__entry->flags = flags;
+		__entry->error = error;
+	),
+	TP_printk("dev %d:%d ino %llu type %s agno %u inum %llu gen %u flags 0x%x error %d\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->ino,
+		  __print_symbolic(__entry->type, XFS_SCRUB_TYPE_DESC),
+		  __entry->agno,
+		  __entry->inum,
+		  __entry->gen,
+		  __entry->flags,
+		  __entry->error)
+)
+#define DEFINE_SCRUB_EVENT(name) \
+DEFINE_EVENT(xfs_scrub_class, name, \
+	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno, \
+		 xfs_ino_t inum, unsigned int gen, unsigned int flags, \
+		 int error), \
+	TP_ARGS(ip, type, agno, inum, gen, flags, error))
+
+DECLARE_EVENT_CLASS(xfs_scrub_sbtree_class,
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agblock_t bno,
+		 xfs_btnum_t btnum, int level, int nlevels, int ptr),
+	TP_ARGS(mp, agno, bno, btnum, level, nlevels, ptr),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_btnum_t, btnum)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_agblock_t, bno)
+		__field(int, level)
+		__field(int, nlevels)
+		__field(int, ptr)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->agno = agno;
+		__entry->btnum = btnum;
+		__entry->bno = bno;
+		__entry->level = level;
+		__entry->nlevels = nlevels;
+		__entry->ptr = ptr;
+	),
+	TP_printk("dev %d:%d agno %u agbno %u btnum %d level %d nlevels %d ptr %d\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->agno,
+		  __entry->bno,
+		  __entry->btnum,
+		  __entry->level,
+		  __entry->nlevels,
+		  __entry->ptr)
+)
+#define DEFINE_SCRUB_SBTREE_EVENT(name) \
+DEFINE_EVENT(xfs_scrub_sbtree_class, name, \
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agblock_t bno, \
+		 xfs_btnum_t btnum, int level, int nlevels, int ptr), \
+	TP_ARGS(mp, agno, bno, btnum, level, nlevels, ptr))
+
+DEFINE_SCRUB_EVENT(xfs_scrub);
+DEFINE_SCRUB_EVENT(xfs_scrub_done);
+DEFINE_SCRUB_SBTREE_EVENT(xfs_scrub_btree_rec);
+DEFINE_SCRUB_SBTREE_EVENT(xfs_scrub_btree_key);
+
+TRACE_EVENT(xfs_scrub_op_error,
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agblock_t bno,
+		 const char *type, int error, const char *func,
+		 int line),
+	TP_ARGS(mp, agno, bno, type, error, func, line),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_agblock_t, bno)
+		__string(type, type)
+		__field(int, error)
+		__string(func, func)
+		__field(int, line)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->agno = agno;
+		__entry->bno = bno;
+		__assign_str(type, type);
+		__entry->error = error;
+		__assign_str(func, func);
+		__entry->line = line;
+	),
+	TP_printk("dev %d:%d agno %u agbno %u type '%s' error %d fn %s:%d\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->agno,
+		  __entry->bno,
+		  __get_str(type),
+		  __entry->error,
+		  __get_str(func),
+		  __entry->line)
+);
+
+TRACE_EVENT(xfs_scrub_file_op_error,
+	TP_PROTO(struct xfs_inode *ip, int whichfork, xfs_fileoff_t offset,
+		 const char *type, int error, const char *func,
+		 int line),
+	TP_ARGS(ip, whichfork, offset, type, error, func, line),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_ino_t, ino)
+		__field(int, whichfork)
+		__field(xfs_fileoff_t, offset)
+		__string(type, type)
+		__field(int, error)
+		__string(func, func)
+		__field(int, line)
+	),
+	TP_fast_assign(
+		__entry->dev = ip->i_mount->m_super->s_dev;
+		__entry->ino = ip->i_ino;
+		__entry->whichfork = whichfork;
+		__entry->offset = offset;
+		__assign_str(type, type);
+		__entry->error = error;
+		__assign_str(func, func);
+		__entry->line = line;
+	),
+	TP_printk("dev %d:%d ino %llu %s offset %llu type '%s' error %d fn %s:%d\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->ino,
+		  __print_symbolic(__entry->whichfork, XFS_FORK_DESC),
+		  __entry->offset,
+		  __get_str(type),
+		  __entry->error,
+		  __get_str(func),
+		  __entry->line)
+);
+
+DECLARE_EVENT_CLASS(xfs_scrub_block_error_class,
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agblock_t bno,
+		 const char *type, const char *check, const char *func,
+		 int line),
+	TP_ARGS(mp, agno, bno, type, check, func, line),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_agblock_t, bno)
+		__string(type, type)
+		__string(check, check)
+		__string(func, func)
+		__field(int, line)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->agno = agno;
+		__entry->bno = bno;
+		__assign_str(type, type);
+		__assign_str(check, check);
+		__assign_str(func, func);
+		__entry->line = line;
+	),
+	TP_printk("dev %d:%d agno %u agbno %u type '%s' check '%s' fn %s:%d\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->agno,
+		  __entry->bno,
+		  __get_str(type),
+		  __get_str(check),
+		  __get_str(func),
+		  __entry->line)
+)
+
+#define DEFINE_SCRUB_BLOCK_ERROR_EVENT(name) \
+DEFINE_EVENT(xfs_scrub_block_error_class, name, \
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agblock_t bno, \
+		 const char *type, const char *check, const char *func, \
+		 int line), \
+	TP_ARGS(mp, agno, bno, type, check, func, line))
+
+DEFINE_SCRUB_BLOCK_ERROR_EVENT(xfs_scrub_block_error);
+DEFINE_SCRUB_BLOCK_ERROR_EVENT(xfs_scrub_block_preen);
+
+DECLARE_EVENT_CLASS(xfs_scrub_ino_error_class,
+	TP_PROTO(struct xfs_mount *mp, xfs_ino_t ino, xfs_agnumber_t agno, xfs_agblock_t bno,
+		 const char *type, const char *check, const char *func,
+		 int line),
+	TP_ARGS(mp, ino, agno, bno, type, check, func, line),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_ino_t, ino)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_agblock_t, bno)
+		__string(type, type)
+		__string(check, check)
+		__string(func, func)
+		__field(int, line)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->ino = ino;
+		__entry->agno = agno;
+		__entry->bno = bno;
+		__assign_str(type, type);
+		__assign_str(check, check);
+		__assign_str(func, func);
+		__entry->line = line;
+	),
+	TP_printk("dev %d:%d ino %llu agno %u agbno %u type '%s' check '%s' fn %s:%d\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->ino,
+		  __entry->agno,
+		  __entry->bno,
+		  __get_str(type),
+		  __get_str(check),
+		  __get_str(func),
+		  __entry->line)
+)
+
+#define DEFINE_SCRUB_INO_ERROR_EVENT(name) \
+DEFINE_EVENT(xfs_scrub_ino_error_class, name, \
+	TP_PROTO(struct xfs_mount *mp, xfs_ino_t ino, xfs_agnumber_t agno, xfs_agblock_t bno, \
+		 const char *type, const char *check, const char *func, \
+		 int line), \
+	TP_ARGS(mp, ino, agno, bno, type, check, func, line))
+
+DEFINE_SCRUB_INO_ERROR_EVENT(xfs_scrub_ino_error);
+DEFINE_SCRUB_INO_ERROR_EVENT(xfs_scrub_ino_preen);
+
+TRACE_EVENT(xfs_scrub_data_error,
+	TP_PROTO(struct xfs_inode *ip, int whichfork, xfs_fileoff_t offset,
+		 const char *type, const char *check, const char *func,
+		 int line),
+	TP_ARGS(ip, whichfork, offset, type, check, func, line),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_ino_t, ino)
+		__field(int, whichfork)
+		__field(xfs_fileoff_t, offset)
+		__string(type, type)
+		__string(check, check)
+		__string(func, func)
+		__field(int, line)
+	),
+	TP_fast_assign(
+		__entry->dev = ip->i_mount->m_super->s_dev;
+		__entry->ino = ip->i_ino;
+		__entry->whichfork = whichfork;
+		__entry->offset = offset;
+		__assign_str(type, type);
+		__assign_str(check, check);
+		__assign_str(func, func);
+		__entry->line = line;
+	),
+	TP_printk("dev %d:%d ino %llu %s fork offset %llu type '%s' check '%s' fn %s:%d\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->ino,
+		  __print_symbolic(__entry->whichfork, XFS_FORK_DESC),
+		  __entry->offset,
+		  __get_str(type),
+		  __get_str(check),
+		  __get_str(func),
+		  __entry->line)
+);
+
+TRACE_EVENT(xfs_scrub_xref_error,
+	TP_PROTO(struct xfs_mount *mp, const char *type, int error,
+		 const char *func, int line),
+	TP_ARGS(mp, type, error, func, line),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__string(type, type)
+		__field(int, error)
+		__string(func, func)
+		__field(int, line)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__assign_str(type, type);
+		__entry->error = error;
+		__assign_str(func, func);
+		__entry->line = line;
+	),
+	TP_printk("dev %d:%d btree %s xref error %d fn %s:%d\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __get_str(type),
+		  __entry->error,
+		  __get_str(func),
+		  __entry->line)
+);
+
+TRACE_EVENT(xfs_scrub_btree_error,
+	TP_PROTO(struct xfs_mount *mp, const char *bt_type, const char *bt_ptr,
+		 xfs_agnumber_t agno, xfs_agblock_t bno, const char *check,
+		 const char *func, int line),
+	TP_ARGS(mp, bt_type, bt_ptr, agno, bno, check, func, line),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__string(bt_type, bt_type)
+		__string(bt_ptr, bt_ptr)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_agblock_t, bno)
+		__string(check, check)
+		__string(func, func)
+		__field(int, line)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__assign_str(bt_type, bt_type);
+		__assign_str(bt_ptr, bt_ptr);
+		__entry->agno = agno;
+		__entry->bno = bno;
+		__assign_str(check, check);
+		__assign_str(func, func);
+		__entry->line = line;
+	),
+	TP_printk("dev %d:%d %s %s agno %u agbno %u check '%s' fn %s:%d\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __get_str(bt_type),
+		  __get_str(bt_ptr),
+		  __entry->agno,
+		  __entry->bno,
+		  __get_str(check),
+		  __get_str(func),
+		  __entry->line)
+);
+
+DECLARE_EVENT_CLASS(xfs_scrub_ag_lock_class,
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t max_ag,
+		 xfs_agnumber_t agno),
+	TP_ARGS(mp, max_ag, agno),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_agnumber_t, max_ag)
+		__field(xfs_agnumber_t, agno)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->max_ag = max_ag;
+		__entry->agno = agno;
+	),
+	TP_printk("dev %d:%d max_ag %u agno %u\n",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->max_ag,
+		  __entry->agno)
+)
+#define DEFINE_SCRUB_AG_LOCK_EVENT(name) \
+DEFINE_EVENT(xfs_scrub_ag_lock_class, name, \
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t max_ag, \
+		 xfs_agnumber_t agno), \
+	TP_ARGS(mp, max_ag, agno))
+
+DEFINE_SCRUB_AG_LOCK_EVENT(xfs_scrub_ag_can_lock);
+DEFINE_SCRUB_AG_LOCK_EVENT(xfs_scrub_ag_may_deadlock);
+DEFINE_SCRUB_AG_LOCK_EVENT(xfs_scrub_ag_lock_all);
+
 #endif /* _TRACE_XFS_H */
 
 #undef TRACE_INCLUDE_PATH

[PATCH 15/55] xfs: create an ioctl to scrub AG metadata

[Darrick J. Wong]

Create an ioctl that can be used to scrub internal filesystem metadata.
The new ioctl takes the metadata type, an (optional) AG number, an
(optional) inode number and generation, and a flags argument.  This will
be used by the upcoming XFS online scrub tool.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile          |    5 +
 fs/xfs/libxfs/xfs_fs.h   |   37 ++++
 fs/xfs/scrub/common.c    |  443 ++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.h    |  147 +++++++++++++++
 fs/xfs/scrub/xfs_scrub.h |   29 +++
 fs/xfs/xfs_ioctl.c       |   28 +++
 fs/xfs/xfs_ioctl32.c     |    1 
 fs/xfs/xfs_trace.h       |    2 
 8 files changed, 691 insertions(+), 1 deletion(-)
 create mode 100644 fs/xfs/scrub/common.c
 create mode 100644 fs/xfs/scrub/common.h
 create mode 100644 fs/xfs/scrub/xfs_scrub.h


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 0e7ee30..c3eb189 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -102,6 +102,11 @@ xfs-y				+= xfs_aops.o \
 				   kmem.o \
 				   uuid.o
 
+# online scrub/repair
+xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
+				   common.o \
+				   )
+
 # low-level transaction/log code
 xfs-y				+= xfs_log.o \
 				   xfs_log_cil.o \
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 10aa908..da3277a 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -554,6 +554,42 @@ typedef struct xfs_swapext
 #define XFS_FSOP_GOING_FLAGS_LOGFLUSH		0x1	/* flush log but not data */
 #define XFS_FSOP_GOING_FLAGS_NOLOGFLUSH		0x2	/* don't flush log nor data */
 
+/* metadata scrubbing */
+struct xfs_scrub_metadata {
+	__u32 sm_type;		/* What to check? */
+	__u32 sm_flags;		/* flags; see below. */
+	union {
+		__u32		__agno;
+		struct {
+			__u64	__ino;
+			__u32	__gen;
+		} i;
+		__u64		__reserved[7];	/* pad to 64 bytes */
+	} p;
+};
+#define sm_agno	p.__agno
+#define sm_ino	p.i.__ino
+#define sm_gen	p.i.__gen
+
+/*
+ * Metadata types and flags for scrub operation.
+ */
+#define XFS_SCRUB_TYPE_TEST	0	/* dummy to test ioctl */
+#define XFS_SCRUB_TYPE_MAX	0
+
+#define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
+#define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
+#define XFS_SCRUB_FLAG_PREEN	0x04	/* o: could be optimized */
+#define XFS_SCRUB_FLAG_XFAIL	0x08	/* o: incomplete xref */
+#define XFS_SCRUB_FLAG_XCORRUPT	0x10	/* o: corruption during xref */
+
+#define XFS_SCRUB_FLAGS_IN	(XFS_SCRUB_FLAG_REPAIR)
+#define XFS_SCRUB_FLAGS_OUT	(XFS_SCRUB_FLAG_CORRUPT | \
+				 XFS_SCRUB_FLAG_PREEN | \
+				 XFS_SCRUB_FLAG_XFAIL | \
+				 XFS_SCRUB_FLAG_XCORRUPT)
+#define XFS_SCRUB_FLAGS_ALL	(XFS_SCRUB_FLAGS_IN | XFS_SCRUB_FLAGS_OUT)
+
 /*
  * AG reserved block counters
  */
@@ -606,6 +642,7 @@ struct xfs_fsop_ag_resblks {
 #define XFS_IOC_ZERO_RANGE	_IOW ('X', 57, struct xfs_flock64)
 #define XFS_IOC_FREE_EOFBLOCKS	_IOR ('X', 58, struct xfs_fs_eofblocks)
 #define XFS_IOC_GETFSMAP	_IOWR('X', 59, struct fsmap_head)
+#define XFS_IOC_SCRUB_METADATA	_IOWR('X', 60, struct xfs_scrub_metadata)
 
 /*
  * ioctl commands that replace IRIX syssgi()'s
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
new file mode 100644
index 0000000..70afa11
--- /dev/null
+++ b/fs/xfs/scrub/common.c
@@ -0,0 +1,443 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_alloc.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_bmap.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_ialloc.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_refcount.h"
+#include "xfs_refcount_btree.h"
+#include "xfs_rmap.h"
+#include "xfs_rmap_btree.h"
+#include "scrub/xfs_scrub.h"
+#include "scrub/common.h"
+
+/*
+ * Online Scrub and Repair
+ *
+ * Traditionally, XFS (the kernel driver) did not know how to check or
+ * repair on-disk data structures.  That task was left to the xfs_check
+ * and xfs_repair tools, both of which require taking the filesystem
+ * offline for a thorough but time consuming examination.  Online
+ * scrub & repair, on the other hand, enables us to check the metadata
+ * for obvious errors while carefully stepping around the filesystem's
+ * ongoing operations, locking rules, etc.
+ *
+ * Given that most XFS metadata consist of records stored in a btree,
+ * most of the checking functions iterate the btree blocks themselves
+ * looking for irregularities.  When a record block is encountered, each
+ * record can be checked for obviously bad values.  Record values can
+ * also be cross-referenced against other btrees to look for potential
+ * misunderstandings between pieces of metadata.
+ *
+ * It is expected that the checkers responsible for per-AG metadata
+ * structures will lock the AG headers (AGI, AGF, AGFL), iterate the
+ * metadata structure, and perform any relevant cross-referencing before
+ * unlocking the AG and returning the results to userspace.  These
+ * scrubbers must not keep an AG locked for too long to avoid tying up
+ * the block and inode allocators.
+ *
+ * Block maps and b-trees rooted in an inode present a special challenge
+ * because they can involve extents from any AG.  The general scrubber
+ * structure of lock -> check -> xref -> unlock still holds, but AG
+ * locking order rules /must/ be obeyed to avoid deadlocks.  The
+ * ordering rule, of course, is that we must lock in increasing AG
+ * order.  Helper functions are provided to track which AG headers we've
+ * already locked.  If we detect an imminent locking order violation, we
+ * can signal a potential deadlock, in which case the scrubber can jump
+ * out to the top level, lock all the AGs in order, and retry the scrub.
+ *
+ * For file data (directories, extended attributes, symlinks) scrub, we
+ * can simply lock the inode and walk the data.  For btree data
+ * (directories and attributes) we follow the same btree-scrubbing
+ * strategy outlined previously to check the records.
+ *
+ * We use a bit of trickery with transactions to avoid buffer deadlocks
+ * if there is a cycle in the metadata.  The basic problem is that
+ * travelling down a btree involves locking the current buffer at each
+ * tree level.  If a pointer should somehow point back to a buffer that
+ * we've already examined, we will deadlock due to the second buffer
+ * locking attempt.  Note however that grabbing a buffer in transaction
+ * context links the locked buffer to the transaction.  If we try to
+ * re-grab the buffer in the context of the same transaction, we avoid
+ * the second lock attempt and continue.  Between the verifier and the
+ * scrubber, something will notice that something is amiss and report
+ * the corruption.  Therefore, each scrubber will allocate an empty
+ * transaction, attach buffers to it, and cancel the transaction at the
+ * end of the scrub run.  Cancelling a non-dirty transaction simply
+ * unlocks the buffers.
+ *
+ * There are four pieces of data that scrub can communicate to
+ * userspace.  The first is the error code (errno), which can be used to
+ * communicate operational errors in performing the scrub.  There are
+ * also three flags that can be set in the scrub context.  If the data
+ * structure itself is corrupt, the "corrupt" flag should be set.  If
+ * the metadata is correct but otherwise suboptimal, there's a "preen"
+ * flag to signal that.  Finally, if we were unable to access a data
+ * structure to perform cross-referencing, we can signal that as well.
+ */
+
+/* Check for operational errors. */
+bool
+xfs_scrub_op_ok(
+	struct xfs_scrub_context	*sc,
+	xfs_agnumber_t			agno,
+	xfs_agblock_t			bno,
+	const char			*type,
+	int				*error,
+	const char			*func,
+	int				line)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+
+	if (*error == 0)
+		return true;
+
+	trace_xfs_scrub_op_error(mp, agno, bno, type, *error, func, line);
+	if (*error == -EFSBADCRC || *error == -EFSCORRUPTED) {
+		sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+		*error = 0;
+	}
+	return false;
+}
+
+/* Check for operational errors for a file offset. */
+bool
+xfs_scrub_file_op_ok(
+	struct xfs_scrub_context	*sc,
+	int				whichfork,
+	xfs_fileoff_t			offset,
+	const char			*type,
+	int				*error,
+	const char			*func,
+	int				line)
+{
+	if (*error == 0)
+		return true;
+
+	trace_xfs_scrub_file_op_error(sc->ip, whichfork, offset, type, *error,
+			func, line);
+	if (*error == -EFSBADCRC || *error == -EFSCORRUPTED) {
+		sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+		*error = 0;
+	}
+	return false;
+}
+
+/* Check for metadata block optimization possibilities. */
+bool
+xfs_scrub_block_preen(
+	struct xfs_scrub_context	*sc,
+	struct xfs_buf			*bp,
+	const char			*type,
+	bool				fs_ok,
+	const char			*check,
+	const char			*func,
+	int				line)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	xfs_fsblock_t			fsbno;
+	xfs_agnumber_t			agno;
+	xfs_agblock_t			bno;
+
+	if (fs_ok)
+		return fs_ok;
+
+	fsbno = XFS_DADDR_TO_FSB(mp, bp->b_bn);
+	agno = XFS_FSB_TO_AGNO(mp, fsbno);
+	bno = XFS_FSB_TO_AGBNO(mp, fsbno);
+
+	sc->sm->sm_flags |= XFS_SCRUB_FLAG_PREEN;
+	trace_xfs_scrub_block_preen(mp, agno, bno, type, check, func, line);
+	return fs_ok;
+}
+
+/* Check for metadata block corruption. */
+bool
+xfs_scrub_block_ok(
+	struct xfs_scrub_context	*sc,
+	struct xfs_buf			*bp,
+	const char			*type,
+	bool				fs_ok,
+	const char			*check,
+	const char			*func,
+	int				line)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	xfs_fsblock_t			fsbno;
+	xfs_agnumber_t			agno;
+	xfs_agblock_t			bno;
+
+	if (fs_ok)
+		return fs_ok;
+
+	fsbno = XFS_DADDR_TO_FSB(mp, bp->b_bn);
+	agno = XFS_FSB_TO_AGNO(mp, fsbno);
+	bno = XFS_FSB_TO_AGBNO(mp, fsbno);
+
+	sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+	trace_xfs_scrub_block_error(mp, agno, bno, type, check, func, line);
+	return fs_ok;
+}
+
+/* Check for inode metadata corruption. */
+bool
+xfs_scrub_ino_ok(
+	struct xfs_scrub_context	*sc,
+	xfs_ino_t			ino,
+	struct xfs_buf			*bp,
+	const char			*type,
+	bool				fs_ok,
+	const char			*check,
+	const char			*func,
+	int				line)
+{
+	struct xfs_inode		*ip = sc->ip;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	xfs_fsblock_t			fsbno;
+	xfs_agnumber_t			agno;
+	xfs_agblock_t			bno;
+
+	if (fs_ok)
+		return fs_ok;
+
+	if (bp) {
+		fsbno = XFS_DADDR_TO_FSB(mp, bp->b_bn);
+		agno = XFS_FSB_TO_AGNO(mp, fsbno);
+		bno = XFS_FSB_TO_AGBNO(mp, fsbno);
+	} else {
+		agno = XFS_INO_TO_AGNO(mp, ip->i_ino);
+		bno = XFS_INO_TO_AGINO(mp, ip->i_ino);
+	}
+
+	sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+	trace_xfs_scrub_ino_error(mp, ino, agno, bno, type, check, func, line);
+	return fs_ok;
+}
+
+/* Check for inode metadata optimization possibilities. */
+bool
+xfs_scrub_ino_preen(
+	struct xfs_scrub_context	*sc,
+	struct xfs_buf			*bp,
+	const char			*type,
+	bool				fs_ok,
+	const char			*check,
+	const char			*func,
+	int				line)
+{
+	struct xfs_inode		*ip = sc->ip;
+	struct xfs_mount		*mp = ip->i_mount;
+	xfs_fsblock_t			fsbno;
+	xfs_agnumber_t			agno;
+	xfs_agblock_t			bno;
+
+	if (fs_ok)
+		return fs_ok;
+
+	if (bp) {
+		fsbno = XFS_DADDR_TO_FSB(mp, bp->b_bn);
+		agno = XFS_FSB_TO_AGNO(mp, fsbno);
+		bno = XFS_FSB_TO_AGBNO(mp, fsbno);
+	} else {
+		agno = XFS_INO_TO_AGNO(mp, ip->i_ino);
+		bno = XFS_INO_TO_AGINO(mp, ip->i_ino);
+	}
+
+	sc->sm->sm_flags |= XFS_SCRUB_FLAG_PREEN;
+	trace_xfs_scrub_ino_preen(mp, ip->i_ino, agno, bno, type, check,
+			func, line);
+	return fs_ok;
+}
+
+/* Check for file data block corruption. */
+bool
+xfs_scrub_data_ok(
+	struct xfs_scrub_context	*sc,
+	int				whichfork,
+	xfs_fileoff_t			offset,
+	const char			*type,
+	bool				fs_ok,
+	const char			*check,
+	const char			*func,
+	int				line)
+{
+	if (fs_ok)
+		return fs_ok;
+
+	sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+	trace_xfs_scrub_data_error(sc->ip, whichfork, offset, type, check,
+			func, line);
+	return fs_ok;
+}
+
+/* Dummy scrubber */
+
+STATIC int
+xfs_scrub_dummy(
+	struct xfs_scrub_context	*sc)
+{
+	if (sc->sm->sm_gen & XFS_SCRUB_FLAG_CORRUPT)
+		sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+	if (sc->sm->sm_gen & XFS_SCRUB_FLAG_PREEN)
+		sc->sm->sm_flags |= XFS_SCRUB_FLAG_PREEN;
+	if (sc->sm->sm_gen & XFS_SCRUB_FLAG_XFAIL)
+		sc->sm->sm_flags |= XFS_SCRUB_FLAG_XFAIL;
+	if (sc->sm->sm_gen & XFS_SCRUB_FLAG_XCORRUPT)
+		sc->sm->sm_flags |= XFS_SCRUB_FLAG_XCORRUPT;
+	if (sc->sm->sm_gen & ~XFS_SCRUB_FLAGS_OUT)
+		return -ENOENT;
+
+	return 0;
+}
+
+/* Scrub setup and teardown. */
+
+/* Free all the resources and finish the transactions. */
+int
+xfs_scrub_teardown(
+	struct xfs_scrub_context	*sc,
+	int				error)
+{
+	xfs_trans_cancel(sc->tp);
+	sc->tp = NULL;
+	return error;
+}
+
+/* Set us up with a transaction and an empty context. */
+int
+xfs_scrub_setup(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+
+	memset(sc, 0, sizeof(*sc));
+	sc->sm = sm;
+	return xfs_scrub_trans_alloc(sm, mp, &M_RES(mp)->tr_itruncate,
+			0, 0, 0, &sc->tp);
+}
+
+/* Scrubbing dispatch. */
+
+struct xfs_scrub_meta_fns {
+	int	(*setup)(struct xfs_scrub_context *, struct xfs_inode *,
+			 struct xfs_scrub_metadata *, bool);
+	int	(*scrub)(struct xfs_scrub_context *);
+	int	(*repair)(struct xfs_scrub_context *);
+	bool	(*has)(struct xfs_sb *);
+};
+
+static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
+	{xfs_scrub_setup, xfs_scrub_dummy, NULL, NULL},
+};
+
+/* Dispatch metadata scrubbing. */
+int
+xfs_scrub_metadata(
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm)
+{
+	struct xfs_scrub_context	sc;
+	struct xfs_mount		*mp = ip->i_mount;
+	const struct xfs_scrub_meta_fns	*fns;
+	bool				deadlocked = false;
+	int				error = 0;
+
+	trace_xfs_scrub(ip, sm->sm_type, sm->sm_agno, sm->sm_ino, sm->sm_gen,
+			sm->sm_flags, error);
+
+	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
+		return -ESHUTDOWN;
+
+	/* Check our inputs. */
+	error = -EINVAL;
+	sm->sm_flags &= ~XFS_SCRUB_FLAGS_OUT;
+	if (sm->sm_flags & ~XFS_SCRUB_FLAGS_IN)
+		goto out;
+	if (sm->sm_flags & XFS_SCRUB_FLAG_REPAIR)
+		goto out;
+	error = -ENOENT;
+	if (sm->sm_type > XFS_SCRUB_TYPE_MAX)
+		goto out;
+	fns = &meta_scrub_fns[sm->sm_type];
+	if (fns->scrub == NULL)
+		goto out;
+	error = -EOPNOTSUPP;
+
+	/* Do we even have this type of metadata? */
+	error = -ENOENT;
+	if (fns->has && !fns->has(&mp->m_sb))
+		goto out;
+
+	/* This isn't a stable feature.  Use with care. */
+	{
+		static bool warned;
+
+		if (!warned)
+			xfs_alert(mp,
+	"EXPERIMENTAL online scrub feature in use. Use at your own risk!");
+		warned = true;
+	}
+
+retry_op:
+	/* Set up for the operation. */
+	error = fns->setup(&sc, ip, sm, deadlocked);
+	if (error)
+		goto out;
+
+	/* Scrub for errors. */
+	error = fns->scrub(&sc);
+	if (!deadlocked && error == -EDEADLOCK) {
+		deadlocked = true;
+		error = xfs_scrub_teardown(&sc, error);
+		if (error != -EDEADLOCK)
+			goto out;
+		goto retry_op;
+	} else if (error)
+		goto out_teardown;
+
+	if (sm->sm_flags & XFS_SCRUB_FLAG_CORRUPT)
+		xfs_alert_ratelimited(mp, "Corruption detected during scrub.");
+
+out_teardown:
+	error = xfs_scrub_teardown(&sc, error);
+out:
+	trace_xfs_scrub_done(ip, sm->sm_type, sm->sm_agno, sm->sm_ino,
+			sm->sm_gen, sm->sm_flags, error);
+	return error;
+}
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
new file mode 100644
index 0000000..af88d67
--- /dev/null
+++ b/fs/xfs/scrub/common.h
@@ -0,0 +1,147 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#ifndef __XFS_REPAIR_COMMON_H__
+#define __XFS_REPAIR_COMMON_H__
+
+struct xfs_scrub_context {
+	/* General scrub state. */
+	struct xfs_scrub_metadata	*sm;
+	struct xfs_trans		*tp;
+	struct xfs_inode		*ip;
+};
+
+/* Should we end the scrub early? */
+static inline bool
+xfs_scrub_should_terminate(
+	int		*error)
+{
+	if (fatal_signal_pending(current)) {
+		if (*error == 0)
+			*error = -EAGAIN;
+		return true;
+	}
+	return false;
+}
+
+/*
+ * Grab a transaction.  If we're going to repair something, we need to
+ * ensure there's enough reservation to make all the changes.  If not,
+ * we can use an empty transaction.
+ */
+static inline int
+xfs_scrub_trans_alloc(
+	struct xfs_scrub_metadata	*sm,
+	struct xfs_mount		*mp,
+	struct xfs_trans_res		*resp,
+	uint				blocks,
+	uint				rtextents,
+	uint				flags,
+	struct xfs_trans		**tpp)
+{
+	return xfs_trans_alloc_empty(mp, tpp);
+}
+
+/* Check for operational errors. */
+bool xfs_scrub_op_ok(struct xfs_scrub_context *sc, xfs_agnumber_t agno,
+		     xfs_agblock_t bno, const char *type, int *error,
+		     const char	*func, int line);
+#define XFS_SCRUB_OP_ERROR_GOTO(sc, agno, bno, type, error, label) \
+	do { \
+		if (!xfs_scrub_op_ok((sc), (agno), (bno), (type), \
+				(error), __func__, __LINE__)) \
+			goto label; \
+	} while (0)
+
+/* Check for operational errors for a file offset. */
+bool xfs_scrub_file_op_ok(struct xfs_scrub_context *sc, int whichfork,
+			  xfs_fileoff_t offset, const char *type,
+			  int *error, const char *func, int line);
+#define XFS_SCRUB_FILE_OP_ERROR_GOTO(sc, which, off, type, error, label) \
+	do { \
+		if (!xfs_scrub_file_op_ok((sc), (which), (off), (type), \
+				(error), __func__, __LINE__)) \
+			goto label; \
+	} while (0)
+
+/* Check for metadata block optimization possibilities. */
+bool xfs_scrub_block_preen(struct xfs_scrub_context *sc, struct xfs_buf *bp,
+			   const char *type, bool fs_ok, const char *check,
+			   const char *func, int line);
+#define XFS_SCRUB_PREEN(sc, bp, type, fs_ok) \
+	xfs_scrub_block_preen((sc), (bp), (type), (fs_ok), #fs_ok, \
+			__func__, __LINE__)
+
+/* Check for inode metadata optimization possibilities. */
+bool xfs_scrub_ino_preen(struct xfs_scrub_context *sc, struct xfs_buf *bp,
+		      const char *type, bool fs_ok, const char *check,
+		      const char *func, int line);
+#define XFS_SCRUB_INO_PREEN(sc, bp, type, fs_ok) \
+	xfs_scrub_ino_preen((sc), (bp), (type), (fs_ok), #fs_ok, \
+			__func__, __LINE__)
+
+/* Check for metadata block corruption. */
+bool xfs_scrub_block_ok(struct xfs_scrub_context *sc, struct xfs_buf *bp,
+			const char *type, bool fs_ok, const char *check,
+			const char *func, int line);
+#define XFS_SCRUB_CHECK(sc, bp, type, fs_ok) \
+	xfs_scrub_block_ok((sc), (bp), (type), (fs_ok), #fs_ok, \
+			__func__, __LINE__)
+#define XFS_SCRUB_GOTO(sc, bp, type, fs_ok, label) \
+	do { \
+		if (!xfs_scrub_block_ok((sc), (bp), (type), (fs_ok), \
+				#fs_ok, __func__, __LINE__)) \
+			goto label; \
+	} while (0)
+
+/* Check for inode metadata corruption. */
+bool xfs_scrub_ino_ok(struct xfs_scrub_context *sc, xfs_ino_t ino,
+		      struct xfs_buf *bp, const char *type, bool fs_ok,
+		      const char *check, const char *func, int line);
+#define XFS_SCRUB_INO_CHECK(sc, ino, bp, type, fs_ok) \
+	xfs_scrub_ino_ok((sc), (ino), (bp), (type), (fs_ok), #fs_ok, \
+			__func__, __LINE__)
+#define XFS_SCRUB_INO_GOTO(sc, ino, bp, type, fs_ok, label) \
+	do { \
+		if (!xfs_scrub_ino_ok((sc), (ino), (bp), (type), (fs_ok), \
+				#fs_ok, __func__, __LINE__)) \
+			goto label; \
+	} while(0)
+
+/* Check for file data block corruption. */
+bool xfs_scrub_data_ok(struct xfs_scrub_context *sc, int whichfork,
+		       xfs_fileoff_t offset, const char *type, bool fs_ok,
+		       const char *check, const char *func, int line);
+#define XFS_SCRUB_DATA_CHECK(sc, whichfork, offset, type, fs_ok) \
+	xfs_scrub_data_ok((sc), (whichfork), (offset), (type), (fs_ok), \
+			#fs_ok, __func__, __LINE__)
+#define XFS_SCRUB_DATA_GOTO(sc, whichfork, offset, type, fs_ok, label) \
+	do { \
+		if (!xfs_scrub_data_ok((sc), (whichfork), (offset), \
+				(type), (fs_ok), #fs_ok, __func__, __LINE__)) \
+			goto label; \
+	} while(0)
+
+/* Setup functions */
+
+int xfs_scrub_teardown(struct xfs_scrub_context *sc, int error);
+int xfs_scrub_setup(struct xfs_scrub_context *sc, struct xfs_inode *ip,
+		    struct xfs_scrub_metadata *sm, bool retry_deadlocked);
+
+#endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/xfs_scrub.h b/fs/xfs/scrub/xfs_scrub.h
new file mode 100644
index 0000000..64e21b4
--- /dev/null
+++ b/fs/xfs/scrub/xfs_scrub.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#ifndef __XFS_SCRUB_H__
+#define __XFS_SCRUB_H__
+
+#ifndef CONFIG_XFS_DEBUG
+# define xfs_scrub_metadata(ip, sm)	(-ENOTTY)
+#else
+int xfs_scrub_metadata(struct xfs_inode *ip, struct xfs_scrub_metadata *sm);
+#endif /* CONFIG_XFS_DEBUG */
+
+#endif	/* __XFS_SCRUB_H__ */
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index 347f726..768bcb0 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -43,6 +43,7 @@
 #include "xfs_acl.h"
 #include "xfs_btree.h"
 #include "xfs_fsmap.h"
+#include "scrub/xfs_scrub.h"
 
 #include <linux/capability.h>
 #include <linux/dcache.h>
@@ -1706,6 +1707,30 @@ xfs_ioc_getfsmap(
 	return 0;
 }
 
+STATIC int
+xfs_ioc_scrub_metadata(
+	struct xfs_inode		*ip,
+	void				__user *arg)
+{
+	struct xfs_scrub_metadata	scrub;
+	int				error;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	if (copy_from_user(&scrub, arg, sizeof(scrub)))
+		return -EFAULT;
+
+	error = xfs_scrub_metadata(ip, &scrub);
+	if (error)
+		return error;
+
+	if (copy_to_user(arg, &scrub, sizeof(scrub)))
+		return -EFAULT;
+
+	return 0;
+}
+
 int
 xfs_ioc_swapext(
 	xfs_swapext_t	*sxp)
@@ -1891,6 +1916,9 @@ xfs_file_ioctl(
 			return -EPERM;
 		return xfs_ioc_getfsmap(ip, arg);
 
+	case XFS_IOC_SCRUB_METADATA:
+		return xfs_ioc_scrub_metadata(ip, arg);
+
 	case XFS_IOC_FD_TO_HANDLE:
 	case XFS_IOC_PATH_TO_HANDLE:
 	case XFS_IOC_PATH_TO_FSHANDLE: {
diff --git a/fs/xfs/xfs_ioctl32.c b/fs/xfs/xfs_ioctl32.c
index da479e9..ba310ff 100644
--- a/fs/xfs/xfs_ioctl32.c
+++ b/fs/xfs/xfs_ioctl32.c
@@ -556,6 +556,7 @@ xfs_file_compat_ioctl(
 	case XFS_IOC_ERROR_CLEARALL:
 	case XFS_IOC_GETFSMAP:
 	case XFS_IOC_GET_AG_RESBLKS:
+	case XFS_IOC_SCRUB_METADATA:
 		return xfs_file_ioctl(filp, cmd, p);
 #ifndef BROKEN_X86_ALIGNMENT
 	/* These are handled fine if no alignment issues */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 6427c70..3e04690 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3353,7 +3353,7 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 
 /* scrub */
 #define XFS_SCRUB_TYPE_DESC \
-	{ 0, NULL }
+	{ XFS_SCRUB_TYPE_TEST,		"dummy" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 16/55] xfs: generic functions to scrub metadata and btrees

[Darrick J. Wong]

Create a function that walks a btree, checking the integrity of each
btree block (headers, keys, records) and calling back to the caller
to perform further checks on the records.  Add some helper functions
so that we report detailed scrub errors in a uniform manner in dmesg.
These are helper functions for subsequent patches.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile            |    1 
 fs/xfs/libxfs/xfs_alloc.c  |    2 
 fs/xfs/libxfs/xfs_alloc.h  |    2 
 fs/xfs/libxfs/xfs_btree.c  |   41 ++-
 fs/xfs/libxfs/xfs_btree.h  |   17 +
 fs/xfs/libxfs/xfs_format.h |    2 
 fs/xfs/scrub/btree.c       |  658 ++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/btree.h       |   95 ++++++
 fs/xfs/scrub/common.c      |  234 ++++++++++++++++
 fs/xfs/scrub/common.h      |   53 ++++
 10 files changed, 1095 insertions(+), 10 deletions(-)
 create mode 100644 fs/xfs/scrub/btree.c
 create mode 100644 fs/xfs/scrub/btree.h


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index c3eb189..39bcf8a 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -104,6 +104,7 @@ xfs-y				+= xfs_aops.o \
 
 # online scrub/repair
 xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
+				   btree.o \
 				   common.o \
 				   )
 
diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
index ba15f30..5384d88 100644
--- a/fs/xfs/libxfs/xfs_alloc.c
+++ b/fs/xfs/libxfs/xfs_alloc.c
@@ -602,7 +602,7 @@ const struct xfs_buf_ops xfs_agfl_buf_ops = {
 /*
  * Read in the allocation group free block array.
  */
-STATIC int				/* error */
+int					/* error */
 xfs_alloc_read_agfl(
 	xfs_mount_t	*mp,		/* mount point structure */
 	xfs_trans_t	*tp,		/* transaction pointer */
diff --git a/fs/xfs/libxfs/xfs_alloc.h b/fs/xfs/libxfs/xfs_alloc.h
index a2101de..1e7e2e2 100644
--- a/fs/xfs/libxfs/xfs_alloc.h
+++ b/fs/xfs/libxfs/xfs_alloc.h
@@ -217,6 +217,8 @@ xfs_alloc_get_rec(
 
 int xfs_read_agf(struct xfs_mount *mp, struct xfs_trans *tp,
 			xfs_agnumber_t agno, int flags, struct xfs_buf **bpp);
+int xfs_alloc_read_agfl(struct xfs_mount *mp, struct xfs_trans *tp,
+			xfs_agnumber_t agno, struct xfs_buf **bpp);
 int xfs_alloc_fix_freelist(struct xfs_alloc_arg *args, int flags);
 int xfs_free_extent_fix_freelist(struct xfs_trans *tp, xfs_agnumber_t agno,
 		struct xfs_buf **agbp);
diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c
index ca94b87..fecfd45 100644
--- a/fs/xfs/libxfs/xfs_btree.c
+++ b/fs/xfs/libxfs/xfs_btree.c
@@ -552,7 +552,7 @@ xfs_btree_ptr_offset(
 /*
  * Return a pointer to the n-th record in the btree block.
  */
-STATIC union xfs_btree_rec *
+union xfs_btree_rec *
 xfs_btree_rec_addr(
 	struct xfs_btree_cur	*cur,
 	int			n,
@@ -565,7 +565,7 @@ xfs_btree_rec_addr(
 /*
  * Return a pointer to the n-th key in the btree block.
  */
-STATIC union xfs_btree_key *
+union xfs_btree_key *
 xfs_btree_key_addr(
 	struct xfs_btree_cur	*cur,
 	int			n,
@@ -578,7 +578,7 @@ xfs_btree_key_addr(
 /*
  * Return a pointer to the n-th high key in the btree block.
  */
-STATIC union xfs_btree_key *
+union xfs_btree_key *
 xfs_btree_high_key_addr(
 	struct xfs_btree_cur	*cur,
 	int			n,
@@ -591,7 +591,7 @@ xfs_btree_high_key_addr(
 /*
  * Return a pointer to the n-th block pointer in the btree block.
  */
-STATIC union xfs_btree_ptr *
+union xfs_btree_ptr *
 xfs_btree_ptr_addr(
 	struct xfs_btree_cur	*cur,
 	int			n,
@@ -625,7 +625,7 @@ xfs_btree_get_iroot(
  * Retrieve the block pointer from the cursor at the given level.
  * This may be an inode btree root or from a buffer.
  */
-STATIC struct xfs_btree_block *		/* generic btree block pointer */
+struct xfs_btree_block *		/* generic btree block pointer */
 xfs_btree_get_block(
 	struct xfs_btree_cur	*cur,	/* btree cursor */
 	int			level,	/* level in btree */
@@ -1737,7 +1737,7 @@ xfs_btree_decrement(
 	return error;
 }
 
-STATIC int
+int
 xfs_btree_lookup_get_block(
 	struct xfs_btree_cur	*cur,	/* btree cursor */
 	int			level,	/* level in the btree */
@@ -4884,3 +4884,32 @@ xfs_btree_count_blocks(
 	return xfs_btree_visit_blocks(cur, xfs_btree_count_blocks_helper,
 			blocks);
 }
+
+/* If there's an extent, we're done. */
+STATIC int
+xfs_btree_has_record_helper(
+	struct xfs_btree_cur		*cur,
+	union xfs_btree_rec		*rec,
+	void				*priv)
+{
+	return XFS_BTREE_QUERY_RANGE_ABORT;
+}
+
+/* Is there a record covering a given range of keys? */
+int
+xfs_btree_has_record(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_irec	*low,
+	union xfs_btree_irec	*high,
+	bool			*exists)
+{
+	int			error;
+
+	error = xfs_btree_query_range(cur, low, high,
+			&xfs_btree_has_record_helper, NULL);
+	if (error && error != XFS_BTREE_QUERY_RANGE_ABORT)
+		return error;
+	*exists = error == XFS_BTREE_QUERY_RANGE_ABORT;
+
+	return 0;
+}
diff --git a/fs/xfs/libxfs/xfs_btree.h b/fs/xfs/libxfs/xfs_btree.h
index 5114055..f1b8241 100644
--- a/fs/xfs/libxfs/xfs_btree.h
+++ b/fs/xfs/libxfs/xfs_btree.h
@@ -161,7 +161,6 @@ struct xfs_btree_ops {
 
 	const struct xfs_buf_ops	*buf_ops;
 
-#if defined(DEBUG) || defined(XFS_WARN)
 	/* check that k1 is lower than k2 */
 	int	(*keys_inorder)(struct xfs_btree_cur *cur,
 				union xfs_btree_key *k1,
@@ -171,7 +170,6 @@ struct xfs_btree_ops {
 	int	(*recs_inorder)(struct xfs_btree_cur *cur,
 				union xfs_btree_rec *r1,
 				union xfs_btree_rec *r2);
-#endif
 };
 
 /*
@@ -504,4 +502,19 @@ int xfs_btree_visit_blocks(struct xfs_btree_cur *cur,
 
 int xfs_btree_count_blocks(struct xfs_btree_cur *cur, xfs_extlen_t *blocks);
 
+union xfs_btree_rec *xfs_btree_rec_addr(struct xfs_btree_cur *cur, int n,
+		struct xfs_btree_block *block);
+union xfs_btree_key *xfs_btree_key_addr(struct xfs_btree_cur *cur, int n,
+		struct xfs_btree_block *block);
+union xfs_btree_key *xfs_btree_high_key_addr(struct xfs_btree_cur *cur, int n,
+		struct xfs_btree_block *block);
+union xfs_btree_ptr *xfs_btree_ptr_addr(struct xfs_btree_cur *cur, int n,
+		struct xfs_btree_block *block);
+int xfs_btree_lookup_get_block(struct xfs_btree_cur *cur, int level,
+		union xfs_btree_ptr *pp, struct xfs_btree_block **blkp);
+struct xfs_btree_block *xfs_btree_get_block(struct xfs_btree_cur *cur,
+		int level, struct xfs_buf **bpp);
+int xfs_btree_has_record(struct xfs_btree_cur *cur, union xfs_btree_irec *low,
+		union xfs_btree_irec *high, bool *exists);
+
 #endif	/* __XFS_BTREE_H__ */
diff --git a/fs/xfs/libxfs/xfs_format.h b/fs/xfs/libxfs/xfs_format.h
index 6b7579e..301effc 100644
--- a/fs/xfs/libxfs/xfs_format.h
+++ b/fs/xfs/libxfs/xfs_format.h
@@ -518,7 +518,7 @@ static inline int xfs_sb_version_hasftype(struct xfs_sb *sbp)
 		 (sbp->sb_features2 & XFS_SB_VERSION2_FTYPE));
 }
 
-static inline int xfs_sb_version_hasfinobt(xfs_sb_t *sbp)
+static inline bool xfs_sb_version_hasfinobt(xfs_sb_t *sbp)
 {
 	return (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5) &&
 		(sbp->sb_features_ro_compat & XFS_SB_FEAT_RO_COMPAT_FINOBT);
diff --git a/fs/xfs/scrub/btree.c b/fs/xfs/scrub/btree.c
new file mode 100644
index 0000000..06607de
--- /dev/null
+++ b/fs/xfs/scrub/btree.c
@@ -0,0 +1,658 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_alloc.h"
+#include "scrub/common.h"
+#include "scrub/btree.h"
+
+/* btree scrubbing */
+
+const char * const btree_types[] = {
+	[XFS_BTNUM_BNO]		= "bnobt",
+	[XFS_BTNUM_CNT]		= "cntbt",
+	[XFS_BTNUM_RMAP]	= "rmapbt",
+	[XFS_BTNUM_BMAP]	= "bmapbt",
+	[XFS_BTNUM_INO]		= "inobt",
+	[XFS_BTNUM_FINO]	= "finobt",
+	[XFS_BTNUM_REFC]	= "refcountbt",
+};
+
+/* Format the trace parameters for the tree cursor. */
+static inline void
+xfs_scrub_btree_format(
+	struct xfs_btree_cur		*cur,
+	int				level,
+	char				*bt_type,
+	size_t				type_len,
+	char				*bt_ptr,
+	size_t				ptr_len,
+	xfs_fsblock_t			*fsbno)
+{
+	char				*type = NULL;
+	struct xfs_btree_block		*block;
+	struct xfs_buf			*bp;
+
+	switch (cur->bc_btnum) {
+	case XFS_BTNUM_BMAP:
+		switch (cur->bc_private.b.whichfork) {
+		case XFS_DATA_FORK:
+			type = "data";
+			break;
+		case XFS_ATTR_FORK:
+			type = "attr";
+			break;
+		case XFS_COW_FORK:
+			type = "CoW";
+			break;
+		}
+		snprintf(bt_type, type_len, "inode %llu %s fork",
+				(unsigned long long)cur->bc_private.b.ip->i_ino,
+				type);
+		break;
+	default:
+		strncpy(bt_type, btree_types[cur->bc_btnum], type_len);
+		break;
+	}
+
+	if (level < cur->bc_nlevels && cur->bc_ptrs[level] >= 1) {
+		block = xfs_btree_get_block(cur, level, &bp);
+		snprintf(bt_ptr, ptr_len, " %s %d/%d",
+				level == 0 ? "rec" : "ptr",
+				cur->bc_ptrs[level],
+				be16_to_cpu(block->bb_numrecs));
+	} else
+		bt_ptr[0] = 0;
+
+	if (level < cur->bc_nlevels && cur->bc_bufs[level])
+		*fsbno = XFS_DADDR_TO_FSB(cur->bc_mp,
+				cur->bc_bufs[level]->b_bn);
+	else if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
+		*fsbno = XFS_INO_TO_FSB(cur->bc_mp,
+				cur->bc_private.b.ip->i_ino);
+	else
+		*fsbno = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_private.a.agno, 0);
+}
+
+/* Check for btree corruption. */
+bool
+xfs_scrub_btree_ok(
+	struct xfs_scrub_context	*sc,
+	struct xfs_btree_cur		*cur,
+	int				level,
+	bool				fs_ok,
+	const char			*check,
+	const char			*func,
+	int				line)
+{
+	char				bt_ptr[24];
+	char				bt_type[48];
+	xfs_fsblock_t			fsbno;
+
+	if (fs_ok)
+		return fs_ok;
+
+	sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+	xfs_scrub_btree_format(cur, level, bt_type, 48, bt_ptr, 24, &fsbno);
+
+	trace_xfs_scrub_btree_error(cur->bc_mp, bt_type, bt_ptr,
+			XFS_FSB_TO_AGNO(cur->bc_mp, fsbno),
+			XFS_FSB_TO_AGBNO(cur->bc_mp, fsbno),
+			check, func, line);
+	return fs_ok;
+}
+
+/* Check for btree operation errors . */
+bool
+xfs_scrub_btree_op_ok(
+	struct xfs_scrub_context	*sc,
+	struct xfs_btree_cur		*cur,
+	int				level,
+	int				*error,
+	const char			*func,
+	int				line)
+{
+	char				bt_ptr[24];
+	char				bt_type[48];
+	xfs_fsblock_t			fsbno;
+
+	if (*error == 0)
+		return true;
+
+	xfs_scrub_btree_format(cur, level, bt_type, 48, bt_ptr, 24, &fsbno);
+
+	return xfs_scrub_op_ok(sc,
+			XFS_FSB_TO_AGNO(cur->bc_mp, fsbno),
+			XFS_FSB_TO_AGBNO(cur->bc_mp, fsbno),
+			bt_type, error, func, line);
+}
+
+/*
+ * Make sure this record is in order and doesn't stray outside of the parent
+ * keys.
+ */
+STATIC int
+xfs_scrub_btree_rec(
+	struct xfs_scrub_btree	*bs)
+{
+	struct xfs_btree_cur	*cur = bs->cur;
+	union xfs_btree_rec	*rec;
+	union xfs_btree_key	key;
+	union xfs_btree_key	hkey;
+	union xfs_btree_key	*keyp;
+	struct xfs_btree_block	*block;
+	struct xfs_btree_block	*keyblock;
+	struct xfs_buf		*bp;
+
+	block = xfs_btree_get_block(cur, 0, &bp);
+	rec = xfs_btree_rec_addr(cur, cur->bc_ptrs[0], block);
+
+	if (bp)
+		trace_xfs_scrub_btree_rec(cur->bc_mp,
+				XFS_FSB_TO_AGNO(cur->bc_mp,
+					XFS_DADDR_TO_FSB(cur->bc_mp, bp->b_bn)),
+				XFS_FSB_TO_AGBNO(cur->bc_mp,
+					XFS_DADDR_TO_FSB(cur->bc_mp, bp->b_bn)),
+				cur->bc_btnum, 0, cur->bc_nlevels,
+				cur->bc_ptrs[0]);
+	else if (cur->bc_flags & XFS_BTREE_ROOT_IN_INODE)
+		trace_xfs_scrub_btree_rec(cur->bc_mp,
+				XFS_INO_TO_AGNO(cur->bc_mp,
+					cur->bc_private.b.ip->i_ino),
+				XFS_INO_TO_AGBNO(cur->bc_mp,
+					cur->bc_private.b.ip->i_ino),
+				cur->bc_btnum, 0, cur->bc_nlevels,
+				cur->bc_ptrs[0]);
+	else
+		trace_xfs_scrub_btree_rec(cur->bc_mp,
+				NULLAGNUMBER, NULLAGBLOCK,
+				cur->bc_btnum, 0, cur->bc_nlevels,
+				cur->bc_ptrs[0]);
+
+	/* If this isn't the first record, are they in order? */
+	XFS_SCRUB_BTREC_CHECK(bs, bs->firstrec ||
+			cur->bc_ops->recs_inorder(cur, &bs->lastrec, rec));
+	bs->firstrec = false;
+	bs->lastrec = *rec;
+
+	if (cur->bc_nlevels == 1)
+		return 0;
+
+	/* Is this at least as large as the parent low key? */
+	cur->bc_ops->init_key_from_rec(&key, rec);
+	keyblock = xfs_btree_get_block(cur, 1, &bp);
+	keyp = xfs_btree_key_addr(cur, cur->bc_ptrs[1], keyblock);
+	XFS_SCRUB_BTKEY_CHECK(bs, 1,
+			cur->bc_ops->diff_two_keys(cur, &key, keyp) >= 0);
+
+	if (!(cur->bc_flags & XFS_BTREE_OVERLAPPING))
+		return 0;
+
+	/* Is this no larger than the parent high key? */
+	cur->bc_ops->init_high_key_from_rec(&hkey, rec);
+	keyp = xfs_btree_high_key_addr(cur, cur->bc_ptrs[1], keyblock);
+	XFS_SCRUB_BTKEY_CHECK(bs, 1,
+			cur->bc_ops->diff_two_keys(cur, keyp, &hkey) >= 0);
+
+	return 0;
+}
+
+/*
+ * Make sure this key is in order and doesn't stray outside of the parent
+ * keys.
+ */
+STATIC int
+xfs_scrub_btree_key(
+	struct xfs_scrub_btree	*bs,
+	int			level)
+{
+	struct xfs_btree_cur	*cur = bs->cur;
+	union xfs_btree_key	*key;
+	union xfs_btree_key	*keyp;
+	struct xfs_btree_block	*block;
+	struct xfs_btree_block	*keyblock;
+	struct xfs_buf		*bp;
+
+	block = xfs_btree_get_block(cur, level, &bp);
+	key = xfs_btree_key_addr(cur, cur->bc_ptrs[level], block);
+
+	if (bp)
+		trace_xfs_scrub_btree_key(cur->bc_mp,
+				XFS_FSB_TO_AGNO(cur->bc_mp,
+					XFS_DADDR_TO_FSB(cur->bc_mp, bp->b_bn)),
+				XFS_FSB_TO_AGBNO(cur->bc_mp,
+					XFS_DADDR_TO_FSB(cur->bc_mp, bp->b_bn)),
+				cur->bc_btnum, level, cur->bc_nlevels,
+				cur->bc_ptrs[level]);
+	else if (cur->bc_flags & XFS_BTREE_ROOT_IN_INODE)
+		trace_xfs_scrub_btree_key(cur->bc_mp,
+				XFS_INO_TO_AGNO(cur->bc_mp,
+					cur->bc_private.b.ip->i_ino),
+				XFS_INO_TO_AGBNO(cur->bc_mp,
+					cur->bc_private.b.ip->i_ino),
+				cur->bc_btnum, level, cur->bc_nlevels,
+				cur->bc_ptrs[level]);
+	else
+		trace_xfs_scrub_btree_key(cur->bc_mp,
+				NULLAGNUMBER, NULLAGBLOCK,
+				cur->bc_btnum, level, cur->bc_nlevels,
+				cur->bc_ptrs[level]);
+
+	/* If this isn't the first key, are they in order? */
+	XFS_SCRUB_BTKEY_CHECK(bs, level, bs->firstkey[level] ||
+			cur->bc_ops->keys_inorder(cur, &bs->lastkey[level],
+					key));
+	bs->firstkey[level] = false;
+	bs->lastkey[level] = *key;
+
+	if (level + 1 >= cur->bc_nlevels)
+		return 0;
+
+	/* Is this at least as large as the parent low key? */
+	keyblock = xfs_btree_get_block(cur, level + 1, &bp);
+	keyp = xfs_btree_key_addr(cur, cur->bc_ptrs[level + 1], keyblock);
+	XFS_SCRUB_BTKEY_CHECK(bs, level,
+			cur->bc_ops->diff_two_keys(cur, key, keyp) >= 0);
+
+	if (!(cur->bc_flags & XFS_BTREE_OVERLAPPING))
+		return 0;
+
+	/* Is this no larger than the parent high key? */
+	key = xfs_btree_high_key_addr(cur, cur->bc_ptrs[level], block);
+	keyp = xfs_btree_high_key_addr(cur, cur->bc_ptrs[level + 1], keyblock);
+	XFS_SCRUB_BTKEY_CHECK(bs, level,
+			cur->bc_ops->diff_two_keys(cur, keyp, key) >= 0);
+
+	return 0;
+}
+
+/* Check a btree pointer. */
+static int
+xfs_scrub_btree_ptr(
+	struct xfs_scrub_btree		*bs,
+	int				level,
+	union xfs_btree_ptr		*ptr)
+{
+	struct xfs_btree_cur		*cur = bs->cur;
+	xfs_daddr_t			daddr;
+	xfs_daddr_t			eofs;
+
+	if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
+			level == cur->bc_nlevels) {
+		if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
+			XFS_SCRUB_BTKEY_GOTO(bs, level, ptr->l == 0, corrupt);
+		} else {
+			XFS_SCRUB_BTKEY_GOTO(bs, level, ptr->s == 0, corrupt);
+		}
+		return 0;
+	}
+
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
+		XFS_SCRUB_BTKEY_GOTO(bs, level,
+				ptr->l != cpu_to_be64(NULLFSBLOCK), corrupt);
+
+		daddr = XFS_FSB_TO_DADDR(cur->bc_mp, be64_to_cpu(ptr->l));
+	} else {
+		XFS_SCRUB_BTKEY_GOTO(bs, level,
+				cur->bc_private.a.agno != NULLAGNUMBER, corrupt);
+		XFS_SCRUB_BTKEY_GOTO(bs, level,
+				ptr->s != cpu_to_be32(NULLAGBLOCK), corrupt);
+
+		daddr = XFS_AGB_TO_DADDR(cur->bc_mp, cur->bc_private.a.agno,
+				be32_to_cpu(ptr->s));
+	}
+	eofs = XFS_FSB_TO_BB(cur->bc_mp, cur->bc_mp->m_sb.sb_dblocks);
+	XFS_SCRUB_BTKEY_GOTO(bs, level, daddr != 0, corrupt);
+	XFS_SCRUB_BTKEY_GOTO(bs, level, daddr < eofs, corrupt);
+
+	return 0;
+
+corrupt:
+	return -EFSCORRUPTED;
+}
+
+/* Check the siblings of a large format btree block. */
+STATIC int
+xfs_scrub_btree_lblock_check_siblings(
+	struct xfs_scrub_btree		*bs,
+	struct xfs_btree_block		*block)
+{
+	struct xfs_btree_block		*pblock;
+	struct xfs_buf			*pbp;
+	struct xfs_btree_cur		*ncur = NULL;
+	union xfs_btree_ptr		*pp;
+	xfs_fsblock_t			leftsib;
+	xfs_fsblock_t			rightsib;
+	xfs_fsblock_t			fsbno;
+	int				level;
+	int				success;
+	int				error = 0;
+
+	leftsib = be64_to_cpu(block->bb_u.l.bb_leftsib);
+	rightsib = be64_to_cpu(block->bb_u.l.bb_rightsib);
+	level = xfs_btree_get_level(block);
+
+	/* Root block should never have siblings. */
+	if (level == bs->cur->bc_nlevels - 1) {
+		XFS_SCRUB_BTKEY_CHECK(bs, level, leftsib == NULLFSBLOCK);
+		XFS_SCRUB_BTKEY_CHECK(bs, level, rightsib == NULLFSBLOCK);
+		return error;
+	}
+
+	/* Does the left sibling match the parent level left block? */
+	if (leftsib != NULLFSBLOCK) {
+		error = xfs_btree_dup_cursor(bs->cur, &ncur);
+		if (error)
+			return error;
+		error = xfs_btree_decrement(ncur, level + 1, &success);
+		XFS_SCRUB_BTKEY_OP_ERROR_GOTO(bs, level + 1, &error, out_cur);
+		XFS_SCRUB_BTKEY_GOTO(bs, level, success, out_cur);
+
+		pblock = xfs_btree_get_block(ncur, level + 1, &pbp);
+		pp = xfs_btree_ptr_addr(ncur, ncur->bc_ptrs[level + 1], pblock);
+		if (!xfs_scrub_btree_ptr(bs, level + 1, pp)) {
+			fsbno = be64_to_cpu(pp->l);
+			XFS_SCRUB_BTKEY_CHECK(bs, level, fsbno == leftsib);
+		}
+
+		xfs_btree_del_cursor(ncur, XFS_BTREE_ERROR);
+		ncur = NULL;
+	}
+
+	/* Does the right sibling match the parent level right block? */
+	if (!error && rightsib != NULLFSBLOCK) {
+		error = xfs_btree_dup_cursor(bs->cur, &ncur);
+		if (error)
+			return error;
+		error = xfs_btree_increment(ncur, level + 1, &success);
+		XFS_SCRUB_BTKEY_OP_ERROR_GOTO(bs, level + 1, &error, out_cur);
+		XFS_SCRUB_BTKEY_GOTO(bs, level, success, out_cur);
+
+		pblock = xfs_btree_get_block(ncur, level + 1, &pbp);
+		pp = xfs_btree_ptr_addr(ncur, ncur->bc_ptrs[level + 1], pblock);
+		if (!xfs_scrub_btree_ptr(bs, level + 1, pp)) {
+			fsbno = be64_to_cpu(pp->l);
+			XFS_SCRUB_BTKEY_CHECK(bs, level, fsbno == rightsib);
+		}
+
+		xfs_btree_del_cursor(ncur, XFS_BTREE_ERROR);
+		ncur = NULL;
+	}
+
+out_cur:
+	if (ncur)
+		xfs_btree_del_cursor(ncur, XFS_BTREE_ERROR);
+	return error;
+}
+
+/* Check the siblings of a small format btree block. */
+STATIC int
+xfs_scrub_btree_sblock_check_siblings(
+	struct xfs_scrub_btree		*bs,
+	struct xfs_btree_block		*block)
+{
+	struct xfs_btree_block		*pblock;
+	struct xfs_buf			*pbp;
+	struct xfs_btree_cur		*ncur = NULL;
+	union xfs_btree_ptr		*pp;
+	xfs_agblock_t			leftsib;
+	xfs_agblock_t			rightsib;
+	xfs_agblock_t			agbno;
+	int				level;
+	int				success;
+	int				error = 0;
+
+	leftsib = be32_to_cpu(block->bb_u.s.bb_leftsib);
+	rightsib = be32_to_cpu(block->bb_u.s.bb_rightsib);
+	level = xfs_btree_get_level(block);
+
+	/* Root block should never have siblings. */
+	if (level == bs->cur->bc_nlevels - 1) {
+		XFS_SCRUB_BTKEY_CHECK(bs, level, leftsib == NULLAGBLOCK);
+		XFS_SCRUB_BTKEY_CHECK(bs, level, rightsib == NULLAGBLOCK);
+		return error;
+	}
+
+	/* Does the left sibling match the parent level left block? */
+	if (leftsib != NULLAGBLOCK) {
+		error = xfs_btree_dup_cursor(bs->cur, &ncur);
+		if (error)
+			return error;
+		error = xfs_btree_decrement(ncur, level + 1, &success);
+		XFS_SCRUB_BTKEY_OP_ERROR_GOTO(bs, level + 1, &error, out_cur);
+		XFS_SCRUB_BTKEY_GOTO(bs, level, success, verify_rightsib);
+
+		pblock = xfs_btree_get_block(ncur, level + 1, &pbp);
+		pp = xfs_btree_ptr_addr(ncur, ncur->bc_ptrs[level + 1], pblock);
+		if (!xfs_scrub_btree_ptr(bs, level + 1, pp)) {
+			agbno = be32_to_cpu(pp->s);
+			XFS_SCRUB_BTKEY_CHECK(bs, level, agbno == leftsib);
+		}
+
+		xfs_btree_del_cursor(ncur, XFS_BTREE_ERROR);
+		ncur = NULL;
+	}
+
+verify_rightsib:
+	if (ncur) {
+		xfs_btree_del_cursor(ncur, XFS_BTREE_ERROR);
+		ncur = NULL;
+	}
+
+	/* Does the right sibling match the parent level right block? */
+	if (rightsib != NULLAGBLOCK) {
+		error = xfs_btree_dup_cursor(bs->cur, &ncur);
+		if (error)
+			return error;
+		error = xfs_btree_increment(ncur, level + 1, &success);
+		XFS_SCRUB_BTKEY_OP_ERROR_GOTO(bs, level + 1, &error, out_cur);
+		XFS_SCRUB_BTKEY_GOTO(bs, level, success, out_cur);
+
+		pblock = xfs_btree_get_block(ncur, level + 1, &pbp);
+		pp = xfs_btree_ptr_addr(ncur, ncur->bc_ptrs[level + 1], pblock);
+		if (!xfs_scrub_btree_ptr(bs, level + 1, pp)) {
+			agbno = be32_to_cpu(pp->s);
+			XFS_SCRUB_BTKEY_CHECK(bs, level, agbno == rightsib);
+		}
+
+		xfs_btree_del_cursor(ncur, XFS_BTREE_ERROR);
+		ncur = NULL;
+	}
+
+out_cur:
+	if (ncur)
+		xfs_btree_del_cursor(ncur, XFS_BTREE_ERROR);
+	return error;
+}
+
+/* Grab and scrub a btree block. */
+STATIC int
+xfs_scrub_btree_block(
+	struct xfs_scrub_btree		*bs,
+	int				level,
+	union xfs_btree_ptr		*pp,
+	struct xfs_btree_block		**pblock,
+	struct xfs_buf			**pbp)
+{
+	int				error;
+
+	error = xfs_btree_lookup_get_block(bs->cur, level, pp, pblock);
+	if (error)
+		return error;
+
+	xfs_btree_get_block(bs->cur, level, pbp);
+	error = xfs_btree_check_block(bs->cur, *pblock, level, *pbp);
+	if (error)
+		return error;
+
+	return bs->check_siblings_fn(bs, *pblock);
+}
+
+/*
+ * Visit all nodes and leaves of a btree.  Check that all pointers and
+ * records are in order, that the keys reflect the records, and use a callback
+ * so that the caller can verify individual records.  The callback is the same
+ * as the one for xfs_btree_query_range, so therefore this function also
+ * returns XFS_BTREE_QUERY_RANGE_ABORT, zero, or a negative error code.
+ */
+int
+xfs_scrub_btree(
+	struct xfs_scrub_context	*sc,
+	struct xfs_btree_cur		*cur,
+	xfs_scrub_btree_rec_fn		scrub_fn,
+	struct xfs_owner_info		*oinfo,
+	void				*private)
+{
+	struct xfs_scrub_btree		bs = {0};
+	union xfs_btree_ptr		ptr;
+	union xfs_btree_ptr		*pp;
+	union xfs_btree_rec		*recp;
+	struct xfs_btree_block		*block;
+	int				level;
+	struct xfs_buf			*bp;
+	int				i;
+	int				error = 0;
+
+	/* Finish filling out the scrub state */
+	bs.cur = cur;
+	bs.scrub_rec = scrub_fn;
+	bs.oinfo = oinfo;
+	bs.firstrec = true;
+	bs.private = private;
+	bs.sc = sc;
+	for (i = 0; i < XFS_BTREE_MAXLEVELS; i++)
+		bs.firstkey[i] = true;
+	INIT_LIST_HEAD(&bs.to_check);
+
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
+		bs.check_siblings_fn = xfs_scrub_btree_lblock_check_siblings;
+	else
+		bs.check_siblings_fn = xfs_scrub_btree_sblock_check_siblings;
+
+	/* Don't try to check a tree with a height we can't handle. */
+	XFS_SCRUB_BTREC_GOTO(&bs, cur->bc_nlevels > 0, out_badcursor);
+	XFS_SCRUB_BTREC_GOTO(&bs, cur->bc_nlevels <= XFS_BTREE_MAXLEVELS,
+			out_badcursor);
+
+	/* Make sure the root isn't in the superblock. */
+	cur->bc_ops->init_ptr_from_cur(cur, &ptr);
+	error = xfs_scrub_btree_ptr(&bs, cur->bc_nlevels, &ptr);
+	XFS_SCRUB_BTKEY_OP_ERROR_GOTO(&bs, cur->bc_nlevels, &error,
+			out_badcursor);
+
+	/* Load the root of the btree. */
+	level = cur->bc_nlevels - 1;
+	cur->bc_ops->init_ptr_from_cur(cur, &ptr);
+	error = xfs_scrub_btree_block(&bs, level, &ptr, &block, &bp);
+	XFS_SCRUB_BTKEY_OP_ERROR_GOTO(&bs, level, &error, out);
+
+	cur->bc_ptrs[level] = 1;
+
+	while (level < cur->bc_nlevels) {
+		block = xfs_btree_get_block(cur, level, &bp);
+
+		if (level == 0) {
+			/* End of leaf, pop back towards the root. */
+			if (cur->bc_ptrs[level] >
+			    be16_to_cpu(block->bb_numrecs)) {
+				if (level < cur->bc_nlevels - 1)
+					cur->bc_ptrs[level + 1]++;
+				level++;
+				continue;
+			}
+
+			/* Records in order for scrub? */
+			error = xfs_scrub_btree_rec(&bs);
+			if (error)
+				goto out;
+			recp = xfs_btree_rec_addr(cur, cur->bc_ptrs[0], block);
+			error = bs.scrub_rec(&bs, recp);
+			if (error < 0 ||
+			    error == XFS_BTREE_QUERY_RANGE_ABORT)
+				break;
+			if (xfs_scrub_should_terminate(&error))
+				break;
+
+			cur->bc_ptrs[level]++;
+			continue;
+		}
+
+		/* End of node, pop back towards the root. */
+		if (cur->bc_ptrs[level] > be16_to_cpu(block->bb_numrecs)) {
+			if (level < cur->bc_nlevels - 1)
+				cur->bc_ptrs[level + 1]++;
+			level++;
+			continue;
+		}
+
+		/* Keys in order for scrub? */
+		error = xfs_scrub_btree_key(&bs, level);
+		if (error)
+			goto out;
+
+		/* Drill another level deeper. */
+		pp = xfs_btree_ptr_addr(cur, cur->bc_ptrs[level], block);
+		error = xfs_scrub_btree_ptr(&bs, level, pp);
+		if (error) {
+			error = 0;
+			cur->bc_ptrs[level]++;
+			continue;
+		}
+		level--;
+		error = xfs_scrub_btree_block(&bs, level, pp, &block, &bp);
+		XFS_SCRUB_BTKEY_OP_ERROR_GOTO(&bs, level, &error, out);
+
+		cur->bc_ptrs[level] = 1;
+	}
+
+out:
+	/*
+	 * If we don't end this function with the cursor pointing at a record
+	 * block, a subsequent non-error cursor deletion will not release
+	 * node-level buffers, causing a buffer leak.  This is quite possible
+	 * with a zero-results scrubbing run, so release the buffers if we
+	 * aren't pointing at a record.
+	 */
+	if (cur->bc_bufs[0] == NULL) {
+		for (i = 0; i < cur->bc_nlevels; i++) {
+			if (cur->bc_bufs[i]) {
+				xfs_trans_brelse(cur->bc_tp, cur->bc_bufs[i]);
+				cur->bc_bufs[i] = NULL;
+				cur->bc_ptrs[i] = 0;
+				cur->bc_ra[i] = 0;
+			}
+		}
+	}
+
+out_badcursor:
+	return error;
+}
diff --git a/fs/xfs/scrub/btree.h b/fs/xfs/scrub/btree.h
new file mode 100644
index 0000000..75e89b1
--- /dev/null
+++ b/fs/xfs/scrub/btree.h
@@ -0,0 +1,95 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#ifndef __XFS_REPAIR_BTREE_H__
+#define __XFS_REPAIR_BTREE_H__
+
+/* btree scrub */
+
+extern const char * const btree_types[];
+
+/* Check for btree corruption. */
+bool xfs_scrub_btree_ok(struct xfs_scrub_context *sc,
+			struct xfs_btree_cur *cur, int level, bool fs_ok,
+			const char *check, const char *func, int line);
+
+/* Check for btree operation errors. */
+bool xfs_scrub_btree_op_ok(struct xfs_scrub_context *sc,
+			   struct xfs_btree_cur *cur, int level, int *error,
+			   const char *func, int line);
+
+#define XFS_SCRUB_BTREC_CHECK(bs, fs_ok) \
+	xfs_scrub_btree_ok((bs)->sc, (bs)->cur, 0, (fs_ok), #fs_ok, \
+			__func__, __LINE__)
+#define XFS_SCRUB_BTREC_GOTO(bs, fs_ok, label) \
+	do { \
+		if (!xfs_scrub_btree_ok((bs)->sc, (bs)->cur, 0, (fs_ok), \
+				#fs_ok, __func__, __LINE__)) \
+			goto label; \
+	} while (0)
+#define XFS_SCRUB_BTREC_OP_ERROR_GOTO(bs, error, label) \
+	do { \
+		if (!xfs_scrub_btree_op_ok((bs)->sc, (bs)->cur, 0, \
+				(error), __func__, __LINE__)) \
+			goto label; \
+	} while (0)
+#define XFS_SCRUB_BTKEY_CHECK(bs, level, fs_ok) \
+	xfs_scrub_btree_ok((bs)->sc, (bs)->cur, (level), (fs_ok), #fs_ok, \
+			__func__, __LINE__)
+#define XFS_SCRUB_BTKEY_GOTO(bs, level, fs_ok, label) \
+	do { \
+		if (!xfs_scrub_btree_ok((bs)->sc, (bs)->cur, (level), (fs_ok), \
+				#fs_ok, __func__, __LINE__)) \
+			goto label; \
+	} while (0)
+#define XFS_SCRUB_BTKEY_OP_ERROR_GOTO(bs, level, error, label) \
+	do { \
+		if (!xfs_scrub_btree_op_ok((bs)->sc, (bs)->cur, (level), \
+				(error), __func__, __LINE__)) \
+			goto label; \
+	} while (0)
+
+struct xfs_scrub_btree;
+typedef int (*xfs_scrub_btree_rec_fn)(
+	struct xfs_scrub_btree	*bs,
+	union xfs_btree_rec	*rec);
+
+struct xfs_scrub_btree {
+	/* caller-provided scrub state */
+	struct xfs_scrub_context	*sc;
+	struct xfs_btree_cur		*cur;
+	xfs_scrub_btree_rec_fn		scrub_rec;
+	struct xfs_owner_info		*oinfo;
+	void				*private;
+
+	/* internal scrub state */
+	union xfs_btree_rec		lastrec;
+	bool				firstrec;
+	union xfs_btree_key		lastkey[XFS_BTREE_MAXLEVELS];
+	bool				firstkey[XFS_BTREE_MAXLEVELS];
+	struct list_head		to_check;
+	int				(*check_siblings_fn)(
+						struct xfs_scrub_btree *,
+						struct xfs_btree_block *);
+};
+int xfs_scrub_btree(struct xfs_scrub_context *sc, struct xfs_btree_cur *cur,
+		    xfs_scrub_btree_rec_fn scrub_fn,
+		    struct xfs_owner_info *oinfo, void *private);
+
+#endif /* __XFS_REPAIR_BTREE_H__ */
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 70afa11..c4dd16c 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -43,6 +43,7 @@
 #include "xfs_rmap_btree.h"
 #include "scrub/xfs_scrub.h"
 #include "scrub/common.h"
+#include "scrub/btree.h"
 
 /*
  * Online Scrub and Repair
@@ -303,6 +304,235 @@ xfs_scrub_data_ok(
 	return fs_ok;
 }
 
+/* AG scrubbing */
+
+/* Grab all the headers for an AG. */
+int
+xfs_scrub_ag_read_headers(
+	struct xfs_scrub_context	*sc,
+	xfs_agnumber_t			agno,
+	struct xfs_buf			**agi,
+	struct xfs_buf			**agf,
+	struct xfs_buf			**agfl)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	int				error;
+
+	error = xfs_ialloc_read_agi(mp, sc->tp, agno, agi);
+	if (error)
+		goto out;
+
+	error = xfs_alloc_read_agf(mp, sc->tp, agno, 0, agf);
+	if (error)
+		goto out;
+
+	error = xfs_alloc_read_agfl(mp, sc->tp, agno, agfl);
+	if (error)
+		goto out;
+
+out:
+	return error;
+}
+
+/* Release all the AG btree cursors. */
+STATIC void
+xfs_scrub_ag_btcur_free(
+	struct xfs_scrub_ag		*sa)
+{
+	if (sa->refc_cur)
+		xfs_btree_del_cursor(sa->refc_cur, XFS_BTREE_ERROR);
+	if (sa->rmap_cur)
+		xfs_btree_del_cursor(sa->rmap_cur, XFS_BTREE_ERROR);
+	if (sa->fino_cur)
+		xfs_btree_del_cursor(sa->fino_cur, XFS_BTREE_ERROR);
+	if (sa->ino_cur)
+		xfs_btree_del_cursor(sa->ino_cur, XFS_BTREE_ERROR);
+	if (sa->cnt_cur)
+		xfs_btree_del_cursor(sa->cnt_cur, XFS_BTREE_ERROR);
+	if (sa->bno_cur)
+		xfs_btree_del_cursor(sa->bno_cur, XFS_BTREE_ERROR);
+
+	sa->refc_cur = NULL;
+	sa->rmap_cur = NULL;
+	sa->fino_cur = NULL;
+	sa->ino_cur = NULL;
+	sa->bno_cur = NULL;
+	sa->cnt_cur = NULL;
+}
+
+/* Initialize all the btree cursors for an AG. */
+int
+xfs_scrub_ag_btcur_init(
+	struct xfs_scrub_context	*sc,
+	struct xfs_scrub_ag		*sa)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	xfs_agnumber_t			agno = sa->agno;
+
+	if (sa->agf_bp) {
+		/* Set up a bnobt cursor for cross-referencing. */
+		sa->bno_cur = xfs_allocbt_init_cursor(mp, sc->tp, sa->agf_bp,
+				agno, XFS_BTNUM_BNO);
+		if (!sa->bno_cur)
+			goto err;
+
+		/* Set up a cntbt cursor for cross-referencing. */
+		sa->cnt_cur = xfs_allocbt_init_cursor(mp, sc->tp, sa->agf_bp,
+				agno, XFS_BTNUM_CNT);
+		if (!sa->cnt_cur)
+			goto err;
+	}
+
+	/* Set up a inobt cursor for cross-referencing. */
+	if (sa->agi_bp) {
+		sa->ino_cur = xfs_inobt_init_cursor(mp, sc->tp, sa->agi_bp,
+					agno, XFS_BTNUM_INO);
+		if (!sa->ino_cur)
+			goto err;
+	}
+
+	/* Set up a finobt cursor for cross-referencing. */
+	if (sa->agi_bp && xfs_sb_version_hasfinobt(&mp->m_sb)) {
+		sa->fino_cur = xfs_inobt_init_cursor(mp, sc->tp, sa->agi_bp,
+				agno, XFS_BTNUM_FINO);
+		if (!sa->fino_cur)
+			goto err;
+	}
+
+	/* Set up a rmapbt cursor for cross-referencing. */
+	if (sa->agf_bp && xfs_sb_version_hasrmapbt(&mp->m_sb)) {
+		sa->rmap_cur = xfs_rmapbt_init_cursor(mp, sc->tp, sa->agf_bp,
+				agno);
+		if (!sa->rmap_cur)
+			goto err;
+	}
+
+	/* Set up a refcountbt cursor for cross-referencing. */
+	if (sa->agf_bp && xfs_sb_version_hasreflink(&mp->m_sb)) {
+		sa->refc_cur = xfs_refcountbt_init_cursor(mp, sc->tp,
+				sa->agf_bp, agno, NULL);
+		if (!sa->refc_cur)
+			goto err;
+	}
+
+	return 0;
+err:
+	return -ENOMEM;
+}
+
+/* Release the AG header context and btree cursors. */
+void
+xfs_scrub_ag_free(
+	struct xfs_scrub_ag		*sa)
+{
+	xfs_scrub_ag_btcur_free(sa);
+	sa->agno = NULLAGNUMBER;
+}
+
+/*
+ * For scrub, grab the AGI and the AGF headers, in that order.  Locking
+ * order requires us to get the AGI before the AGF.  We use the
+ * transaction to avoid deadlocking on crosslinked metadata buffers;
+ * either the caller passes one in (bmap scrub) or we have to create a
+ * transaction ourselves.
+ */
+int
+xfs_scrub_ag_init(
+	struct xfs_scrub_context	*sc,
+	xfs_agnumber_t			agno,
+	struct xfs_scrub_ag		*sa)
+{
+	int				error;
+
+	memset(sa, 0, sizeof(*sa));
+	sa->agno = agno;
+	error = xfs_scrub_ag_read_headers(sc, agno, &sa->agi_bp,
+			&sa->agf_bp, &sa->agfl_bp);
+	if (error)
+		goto err;
+
+	error = xfs_scrub_ag_btcur_init(sc, sa);
+	if (error)
+		goto err;
+
+	return error;
+err:
+	xfs_scrub_ag_free(sa);
+	return error;
+}
+
+/* Organize locking of multiple AGs for a scrub. */
+
+/* Initialize the AG lock handler. */
+void
+xfs_scrub_ag_lock_init(
+	struct xfs_mount		*mp,
+	struct xfs_scrub_ag_lock	*ag_lock)
+{
+	if (mp->m_sb.sb_agcount <= XFS_SCRUB_AGMASK_NR)
+		ag_lock->agmask = ag_lock->__agmask;
+	else
+		ag_lock->agmask = kmem_alloc(1 + (mp->m_sb.sb_agcount / NBBY),
+				KM_SLEEP | KM_NOFS);
+	ag_lock->max_ag = NULLAGNUMBER;
+}
+
+/* Can we lock the AG's headers without deadlocking? */
+bool
+xfs_scrub_ag_can_lock(
+	struct xfs_scrub_context	*sc,
+	xfs_agnumber_t			agno)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_scrub_ag_lock	*ag_lock = &sc->ag_lock;
+
+	ASSERT(agno < mp->m_sb.sb_agcount);
+
+	trace_xfs_scrub_ag_can_lock(mp, ag_lock->max_ag, agno);
+
+	/* Already locked? */
+	if (test_bit(agno, ag_lock->agmask))
+		return true;
+
+	/* If we can't lock the AG without violating locking order, bail out. */
+	if (ag_lock->max_ag != NULLAGNUMBER && agno < ag_lock->max_ag) {
+		trace_xfs_scrub_ag_may_deadlock(mp, ag_lock->max_ag, agno);
+		return false;
+	}
+
+	set_bit(agno, ag_lock->agmask);
+	ag_lock->max_ag = agno;
+	return true;
+}
+
+/* Read all AG headers and attach to this transaction. */
+int
+xfs_scrub_ag_lock_all(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_scrub_ag_lock	*ag_lock = &sc->ag_lock;
+	struct xfs_buf			*agi;
+	struct xfs_buf			*agf;
+	struct xfs_buf			*agfl;
+	xfs_agnumber_t			agno;
+	int				error = 0;
+
+	trace_xfs_scrub_ag_lock_all(mp, ag_lock->max_ag, mp->m_sb.sb_agcount);
+
+	ASSERT(ag_lock->max_ag == NULLAGNUMBER);
+	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
+		error = xfs_scrub_ag_read_headers(sc, agno, &agi, &agf,
+				&agfl);
+		if (error)
+			break;
+		set_bit(agno, ag_lock->agmask);
+		ag_lock->max_ag = agno;
+	}
+
+	return error;
+}
+
 /* Dummy scrubber */
 
 STATIC int
@@ -331,6 +561,10 @@ xfs_scrub_teardown(
 	struct xfs_scrub_context	*sc,
 	int				error)
 {
+	xfs_scrub_ag_free(&sc->sa);
+	if (sc->ag_lock.agmask != sc->ag_lock.__agmask)
+		kmem_free(sc->ag_lock.agmask);
+	sc->ag_lock.agmask = NULL;
 	xfs_trans_cancel(sc->tp);
 	sc->tp = NULL;
 	return error;
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index af88d67..fc38f67 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -20,11 +20,51 @@
 #ifndef __XFS_REPAIR_COMMON_H__
 #define __XFS_REPAIR_COMMON_H__
 
+/* Buffer pointers and btree cursors for an entire AG. */
+struct xfs_scrub_ag {
+	xfs_agnumber_t			agno;
+
+	/* AG btree roots */
+	struct xfs_buf			*agf_bp;
+	struct xfs_buf			*agfl_bp;
+	struct xfs_buf			*agi_bp;
+
+	/* AG btrees */
+	struct xfs_btree_cur		*bno_cur;
+	struct xfs_btree_cur		*cnt_cur;
+	struct xfs_btree_cur		*ino_cur;
+	struct xfs_btree_cur		*fino_cur;
+	struct xfs_btree_cur		*rmap_cur;
+	struct xfs_btree_cur		*refc_cur;
+};
+
+/*
+ * Track which AGs for which we've already locked the header buffers.
+ * This information helps us avoid deadlocks by ensuring locking order
+ * rule compliance.  max_ag is the highest AG number that we've locked;
+ * we can only re-lock an AG we've already locked, or lock a higher AG.
+ * If we try to lock a lower numbered AG, we must restart the operation
+ * with all AG headers locked from the beginning.
+ */
+#define XFS_SCRUB_AGMASK_NR		128
+struct xfs_scrub_ag_lock {
+	xfs_agnumber_t			max_ag;
+	unsigned long			*agmask;
+	unsigned long			__agmask[XFS_SCRUB_AGMASK_NR /
+						 sizeof(unsigned long)];
+};
+
 struct xfs_scrub_context {
 	/* General scrub state. */
 	struct xfs_scrub_metadata	*sm;
 	struct xfs_trans		*tp;
 	struct xfs_inode		*ip;
+
+	/* State tracking for multi-AG operations. */
+	struct xfs_scrub_ag_lock	ag_lock;
+
+	/* State tracking for single-AG operations. */
+	struct xfs_scrub_ag		sa;
 };
 
 /* Should we end the scrub early? */
@@ -138,6 +178,19 @@ bool xfs_scrub_data_ok(struct xfs_scrub_context *sc, int whichfork,
 			goto label; \
 	} while(0)
 
+bool xfs_scrub_ag_can_lock(struct xfs_scrub_context *sc, xfs_agnumber_t agno);
+int xfs_scrub_ag_lock_all(struct xfs_scrub_context *sc);
+void xfs_scrub_ag_lock_init(struct xfs_mount *mp,
+			    struct xfs_scrub_ag_lock *ag_lock);
+void xfs_scrub_ag_free(struct xfs_scrub_ag *sa);
+int xfs_scrub_ag_init(struct xfs_scrub_context *sc, xfs_agnumber_t agno,
+		      struct xfs_scrub_ag *sa);
+int xfs_scrub_ag_read_headers(struct xfs_scrub_context *sc, xfs_agnumber_t agno,
+			      struct xfs_buf **agi, struct xfs_buf **agf,
+			      struct xfs_buf **agfl);
+int xfs_scrub_ag_btcur_init(struct xfs_scrub_context *sc,
+			    struct xfs_scrub_ag *sa);
+
 /* Setup functions */
 
 int xfs_scrub_teardown(struct xfs_scrub_context *sc, int error);

[PATCH 17/55] xfs: scrub the backup superblocks

[Darrick J. Wong]

Ensure that the geometry presented in the backup superblocks matches
the primary superblock so that repair can recover the filesystem if
that primary gets corrupted.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile         |    1 
 fs/xfs/libxfs/xfs_fs.h  |    3 -
 fs/xfs/scrub/agheader.c |  198 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.c   |    1 
 fs/xfs/scrub/common.h   |    6 +
 fs/xfs/xfs_trace.h      |    3 -
 6 files changed, 210 insertions(+), 2 deletions(-)
 create mode 100644 fs/xfs/scrub/agheader.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 39bcf8a..ca7399a 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -104,6 +104,7 @@ xfs-y				+= xfs_aops.o \
 
 # online scrub/repair
 xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
+				   agheader.o \
 				   btree.o \
 				   common.o \
 				   )
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index da3277a..6fb1cc6 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -575,7 +575,8 @@ struct xfs_scrub_metadata {
  * Metadata types and flags for scrub operation.
  */
 #define XFS_SCRUB_TYPE_TEST	0	/* dummy to test ioctl */
-#define XFS_SCRUB_TYPE_MAX	0
+#define XFS_SCRUB_TYPE_SB	1	/* superblock */
+#define XFS_SCRUB_TYPE_MAX	1
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
new file mode 100644
index 0000000..98031e3
--- /dev/null
+++ b/fs/xfs/scrub/agheader.c
@@ -0,0 +1,198 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "scrub/common.h"
+
+/* Set us up to check an AG header. */
+int
+xfs_scrub_setup_ag(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+
+	if (sm->sm_agno >= mp->m_sb.sb_agcount)
+		return -EINVAL;
+	return xfs_scrub_setup(sc, ip, sm, retry_deadlocked);
+}
+
+/* Superblock */
+
+#define XFS_SCRUB_SB_CHECK(fs_ok) \
+	XFS_SCRUB_CHECK(sc, bp, "superblock", fs_ok)
+#define XFS_SCRUB_SB_PREEN(fs_ok) \
+	XFS_SCRUB_PREEN(sc, bp, "superblock", fs_ok)
+#define XFS_SCRUB_SB_OP_ERROR_GOTO(label) \
+	XFS_SCRUB_OP_ERROR_GOTO(sc, agno, 0, "superblock", &error, out)
+/* Scrub the filesystem superblock. */
+int
+xfs_scrub_superblock(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_buf			*bp;
+	struct xfs_sb			sb;
+	xfs_agnumber_t			agno;
+	uint32_t			v2_ok;
+	int				error;
+
+	agno = sc->sm->sm_agno;
+
+	error = xfs_trans_read_buf(mp, sc->tp, mp->m_ddev_targp,
+		  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
+		  XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_sb_buf_ops);
+	if (error) {
+		trace_xfs_scrub_block_error(mp, agno, XFS_SB_BLOCK(mp),
+				"superblock", "error != 0", __func__, __LINE__);
+		error = 0;
+		sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+		goto out;
+	}
+
+	/*
+	 * The in-core sb is a more up-to-date copy of AG 0's sb,
+	 * so there's no point in comparing the two.
+	 */
+	if (agno == 0)
+		goto out;
+
+	xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp));
+
+	/* Verify the geometries match. */
+#define XFS_SCRUB_SB_FIELD(fn) \
+		XFS_SCRUB_SB_CHECK(sb.sb_##fn == mp->m_sb.sb_##fn)
+#define XFS_PREEN_SB_FIELD(fn) \
+		XFS_SCRUB_SB_PREEN(sb.sb_##fn == mp->m_sb.sb_##fn)
+	XFS_SCRUB_SB_FIELD(blocksize);
+	XFS_SCRUB_SB_FIELD(dblocks);
+	XFS_SCRUB_SB_FIELD(rblocks);
+	XFS_SCRUB_SB_FIELD(rextents);
+	XFS_SCRUB_SB_PREEN(uuid_equal(&sb.sb_uuid, &mp->m_sb.sb_uuid));
+	XFS_SCRUB_SB_FIELD(logstart);
+	XFS_PREEN_SB_FIELD(rootino);
+	XFS_PREEN_SB_FIELD(rbmino);
+	XFS_PREEN_SB_FIELD(rsumino);
+	XFS_SCRUB_SB_FIELD(rextsize);
+	XFS_SCRUB_SB_FIELD(agblocks);
+	XFS_SCRUB_SB_FIELD(agcount);
+	XFS_SCRUB_SB_FIELD(rbmblocks);
+	XFS_SCRUB_SB_FIELD(logblocks);
+	XFS_SCRUB_SB_CHECK(!(sb.sb_versionnum & ~XFS_SB_VERSION_OKBITS));
+	XFS_SCRUB_SB_CHECK(XFS_SB_VERSION_NUM(&sb) ==
+			   XFS_SB_VERSION_NUM(&mp->m_sb));
+	XFS_SCRUB_SB_FIELD(sectsize);
+	XFS_SCRUB_SB_FIELD(inodesize);
+	XFS_SCRUB_SB_FIELD(inopblock);
+	XFS_SCRUB_SB_PREEN(memcmp(sb.sb_fname, mp->m_sb.sb_fname,
+			   sizeof(sb.sb_fname)) == 0);
+	XFS_SCRUB_SB_FIELD(blocklog);
+	XFS_SCRUB_SB_FIELD(sectlog);
+	XFS_SCRUB_SB_FIELD(inodelog);
+	XFS_SCRUB_SB_FIELD(inopblog);
+	XFS_SCRUB_SB_FIELD(agblklog);
+	XFS_SCRUB_SB_FIELD(rextslog);
+	XFS_PREEN_SB_FIELD(imax_pct);
+	XFS_PREEN_SB_FIELD(uquotino);
+	XFS_PREEN_SB_FIELD(gquotino);
+	XFS_SCRUB_SB_FIELD(shared_vn);
+	XFS_SCRUB_SB_FIELD(inoalignmt);
+	XFS_PREEN_SB_FIELD(unit);
+	XFS_PREEN_SB_FIELD(width);
+	XFS_SCRUB_SB_FIELD(dirblklog);
+	XFS_SCRUB_SB_FIELD(logsectlog);
+	XFS_SCRUB_SB_FIELD(logsectsize);
+	XFS_SCRUB_SB_FIELD(logsunit);
+	v2_ok = XFS_SB_VERSION2_OKBITS;
+	if (XFS_SB_VERSION_NUM(&sb) >= XFS_SB_VERSION_5)
+		v2_ok |= XFS_SB_VERSION2_CRCBIT;
+	XFS_SCRUB_SB_CHECK(!(sb.sb_features2 & ~v2_ok));
+	XFS_SCRUB_SB_PREEN(sb.sb_features2 == sb.sb_bad_features2);
+	XFS_SCRUB_SB_CHECK(!sb.sb_features2 ||
+			xfs_sb_version_hasmorebits(&mp->m_sb));
+	if (xfs_sb_version_hascrc(&mp->m_sb)) {
+		XFS_SCRUB_SB_CHECK(!xfs_sb_has_compat_feature(&sb,
+				XFS_SB_FEAT_COMPAT_UNKNOWN));
+		XFS_SCRUB_SB_CHECK(!xfs_sb_has_ro_compat_feature(&sb,
+				XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
+		XFS_SCRUB_SB_CHECK(!xfs_sb_has_incompat_feature(&sb,
+				XFS_SB_FEAT_INCOMPAT_UNKNOWN));
+		XFS_SCRUB_SB_CHECK(!xfs_sb_has_incompat_log_feature(&sb,
+				XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN));
+		XFS_SCRUB_SB_FIELD(spino_align);
+		XFS_PREEN_SB_FIELD(pquotino);
+	}
+	if (xfs_sb_version_hasmetauuid(&mp->m_sb)) {
+		XFS_SCRUB_SB_CHECK(uuid_equal(&sb.sb_meta_uuid,
+					&mp->m_sb.sb_meta_uuid));
+		XFS_SCRUB_SB_CHECK(uuid_equal(&sb.sb_uuid,
+					&mp->m_sb.sb_uuid));
+	} else
+		XFS_SCRUB_SB_CHECK(uuid_equal(&sb.sb_uuid,
+					&mp->m_sb.sb_meta_uuid));
+#undef XFS_SCRUB_SB_FIELD
+
+#define XFS_SCRUB_SB_FEAT(fn) \
+		XFS_SCRUB_SB_CHECK(xfs_sb_version_has##fn(&sb) == \
+		xfs_sb_version_has##fn(&mp->m_sb))
+	XFS_SCRUB_SB_FEAT(align);
+	XFS_SCRUB_SB_FEAT(dalign);
+	XFS_SCRUB_SB_FEAT(logv2);
+	XFS_SCRUB_SB_FEAT(extflgbit);
+	XFS_SCRUB_SB_FEAT(sector);
+	XFS_SCRUB_SB_FEAT(asciici);
+	XFS_SCRUB_SB_FEAT(morebits);
+	XFS_SCRUB_SB_FEAT(lazysbcount);
+	XFS_SCRUB_SB_FEAT(crc);
+	XFS_SCRUB_SB_FEAT(_pquotino);
+	XFS_SCRUB_SB_FEAT(ftype);
+	XFS_SCRUB_SB_FEAT(finobt);
+	XFS_SCRUB_SB_FEAT(sparseinodes);
+	XFS_SCRUB_SB_FEAT(metauuid);
+	XFS_SCRUB_SB_FEAT(rmapbt);
+	XFS_SCRUB_SB_FEAT(reflink);
+#undef XFS_SCRUB_SB_FEAT
+
+#define XFS_SCRUB_SB_FEAT_PREEN(fn) \
+		XFS_SCRUB_SB_PREEN(xfs_sb_version_has##fn(&sb) == \
+		xfs_sb_version_has##fn(&mp->m_sb))
+	XFS_SCRUB_SB_FEAT_PREEN(attr);
+	XFS_SCRUB_SB_FEAT_PREEN(attr2);
+#undef XFS_SCRUB_SB_FEAT_PREEN
+
+out:
+	return error;
+}
+#undef XFS_SCRUB_SB_OP_ERROR_GOTO
+#undef XFS_SCRUB_SB_CHECK
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index c4dd16c..924079d 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -598,6 +598,7 @@ struct xfs_scrub_meta_fns {
 
 static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup, xfs_scrub_dummy, NULL, NULL},
+	{xfs_scrub_setup_ag, xfs_scrub_superblock, NULL, NULL},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index fc38f67..bd2896b 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -196,5 +196,11 @@ int xfs_scrub_ag_btcur_init(struct xfs_scrub_context *sc,
 int xfs_scrub_teardown(struct xfs_scrub_context *sc, int error);
 int xfs_scrub_setup(struct xfs_scrub_context *sc, struct xfs_inode *ip,
 		    struct xfs_scrub_metadata *sm, bool retry_deadlocked);
+int xfs_scrub_setup_ag(struct xfs_scrub_context *sc, struct xfs_inode *ip,
+		       struct xfs_scrub_metadata *sm, bool retry_deadlocked);
+
+/* Metadata scrubbers */
+
+int xfs_scrub_superblock(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 3e04690..ddfe15f 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3353,7 +3353,8 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 
 /* scrub */
 #define XFS_SCRUB_TYPE_DESC \
-	{ XFS_SCRUB_TYPE_TEST,		"dummy" }
+	{ XFS_SCRUB_TYPE_TEST,		"dummy" }, \
+	{ XFS_SCRUB_TYPE_SB,		"superblock" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 18/55] xfs: scrub AGF and AGFL

[Darrick J. Wong]

Check the block references in the AGF and AGFL headers to make sure
they make sense.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_fs.h  |    4 +
 fs/xfs/scrub/agheader.c |  227 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.c   |   62 +++++++++++++
 fs/xfs/scrub/common.h   |    8 ++
 fs/xfs/xfs_trace.h      |    4 +
 5 files changed, 303 insertions(+), 2 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 6fb1cc6..7434f16 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -576,7 +576,9 @@ struct xfs_scrub_metadata {
  */
 #define XFS_SCRUB_TYPE_TEST	0	/* dummy to test ioctl */
 #define XFS_SCRUB_TYPE_SB	1	/* superblock */
-#define XFS_SCRUB_TYPE_MAX	1
+#define XFS_SCRUB_TYPE_AGF	2	/* AG free header */
+#define XFS_SCRUB_TYPE_AGFL	3	/* AG free list */
+#define XFS_SCRUB_TYPE_MAX	3
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index 98031e3..94f29ae 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -48,6 +48,72 @@ xfs_scrub_setup_ag(
 	return xfs_scrub_setup(sc, ip, sm, retry_deadlocked);
 }
 
+/* Find the size of the AG, in blocks. */
+static inline xfs_agblock_t
+xfs_scrub_ag_blocks(
+	struct xfs_mount	*mp,
+	xfs_agnumber_t		agno)
+{
+	ASSERT(agno < mp->m_sb.sb_agcount);
+
+	if (agno < mp->m_sb.sb_agcount - 1)
+		return mp->m_sb.sb_agblocks;
+	return mp->m_sb.sb_dblocks - (agno * mp->m_sb.sb_agblocks);
+}
+
+/* Walk all the blocks in the AGFL. */
+int
+xfs_scrub_walk_agfl(
+	struct xfs_scrub_context	*sc,
+	int				(*fn)(struct xfs_scrub_context *,
+					      xfs_agblock_t bno, void *),
+	void				*priv)
+{
+	struct xfs_agf			*agf;
+	__be32				*agfl_bno;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	unsigned int			flfirst;
+	unsigned int			fllast;
+	int				i;
+	int				error;
+
+	agf = XFS_BUF_TO_AGF(sc->sa.agf_bp);
+	agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, sc->sa.agfl_bp);
+	flfirst = be32_to_cpu(agf->agf_flfirst);
+	fllast = be32_to_cpu(agf->agf_fllast);
+
+	/* Skip an empty AGFL. */
+	if (agf->agf_flcount == cpu_to_be32(0))
+		return 0;
+
+	/* first to last is a consecutive list. */
+	if (fllast >= flfirst) {
+		for (i = flfirst; i <= fllast; i++) {
+			error = fn(sc, be32_to_cpu(agfl_bno[i]), priv);
+			if (error)
+				return error;
+		}
+
+		return 0;
+	}
+
+	/* first to the end */
+	for (i = flfirst; i < XFS_AGFL_SIZE(mp); i++) {
+		error = fn(sc, be32_to_cpu(agfl_bno[i]), priv);
+		if (error)
+			return error;
+	}
+
+	/* the start to last. */
+	for (i = 0; i <= fllast; i++) {
+		error = fn(sc, be32_to_cpu(agfl_bno[i]), priv);
+		if (error)
+			return error;
+	}
+
+	return 0;
+}
+
 /* Superblock */
 
 #define XFS_SCRUB_SB_CHECK(fs_ok) \
@@ -196,3 +262,164 @@ xfs_scrub_superblock(
 }
 #undef XFS_SCRUB_SB_OP_ERROR_GOTO
 #undef XFS_SCRUB_SB_CHECK
+
+/* AGF */
+
+#define XFS_SCRUB_AGF_CHECK(fs_ok) \
+	XFS_SCRUB_CHECK(sc, sc->sa.agf_bp, "AGF", fs_ok)
+#define XFS_SCRUB_AGF_OP_ERROR_GOTO(error, label) \
+	XFS_SCRUB_OP_ERROR_GOTO(sc, sc->sm->sm_agno, \
+			XFS_AGF_BLOCK(sc->tp->t_mountp), "AGF", error, label)
+/* Scrub the AGF. */
+int
+xfs_scrub_agf(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_agf			*agf;
+	xfs_daddr_t			daddr;
+	xfs_daddr_t			eofs;
+	xfs_agnumber_t			agno;
+	xfs_agblock_t			agbno;
+	xfs_agblock_t			eoag;
+	xfs_agblock_t			agfl_first;
+	xfs_agblock_t			agfl_last;
+	xfs_agblock_t			agfl_count;
+	xfs_agblock_t			fl_count;
+	int				level;
+	int				error = 0;
+
+	agno = sc->sm->sm_agno;
+	error = xfs_scrub_load_ag_headers(sc, agno, XFS_SCRUB_TYPE_AGF);
+	XFS_SCRUB_AGF_OP_ERROR_GOTO(&error, out);
+
+	agf = XFS_BUF_TO_AGF(sc->sa.agf_bp);
+	eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
+
+	/* Check the AG length */
+	eoag = be32_to_cpu(agf->agf_length);
+	XFS_SCRUB_AGF_CHECK(eoag == xfs_scrub_ag_blocks(mp, agno));
+
+	/* Check the AGF btree roots and levels */
+	agbno = be32_to_cpu(agf->agf_roots[XFS_BTNUM_BNO]);
+	daddr = XFS_AGB_TO_DADDR(mp, agno, agbno);
+	XFS_SCRUB_AGF_CHECK(agbno > XFS_AGI_BLOCK(mp));
+	XFS_SCRUB_AGF_CHECK(agbno < mp->m_sb.sb_agblocks);
+	XFS_SCRUB_AGF_CHECK(agbno < eoag);
+	XFS_SCRUB_AGF_CHECK(daddr < eofs);
+
+	agbno = be32_to_cpu(agf->agf_roots[XFS_BTNUM_CNT]);
+	daddr = XFS_AGB_TO_DADDR(mp, agno, agbno);
+	XFS_SCRUB_AGF_CHECK(agbno > XFS_AGI_BLOCK(mp));
+	XFS_SCRUB_AGF_CHECK(agbno < mp->m_sb.sb_agblocks);
+	XFS_SCRUB_AGF_CHECK(agbno < eoag);
+	XFS_SCRUB_AGF_CHECK(daddr < eofs);
+
+	level = be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]);
+	XFS_SCRUB_AGF_CHECK(level > 0);
+	XFS_SCRUB_AGF_CHECK(level <= XFS_BTREE_MAXLEVELS);
+
+	level = be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]);
+	XFS_SCRUB_AGF_CHECK(level > 0);
+	XFS_SCRUB_AGF_CHECK(level <= XFS_BTREE_MAXLEVELS);
+
+	if (xfs_sb_version_hasrmapbt(&mp->m_sb)) {
+		agbno = be32_to_cpu(agf->agf_roots[XFS_BTNUM_RMAP]);
+		daddr = XFS_AGB_TO_DADDR(mp, agno, agbno);
+		XFS_SCRUB_AGF_CHECK(agbno > XFS_AGI_BLOCK(mp));
+		XFS_SCRUB_AGF_CHECK(agbno < mp->m_sb.sb_agblocks);
+		XFS_SCRUB_AGF_CHECK(agbno < eoag);
+		XFS_SCRUB_AGF_CHECK(daddr < eofs);
+
+		level = be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]);
+		XFS_SCRUB_AGF_CHECK(level > 0);
+		XFS_SCRUB_AGF_CHECK(level <= XFS_BTREE_MAXLEVELS);
+	}
+
+	if (xfs_sb_version_hasreflink(&mp->m_sb)) {
+		agbno = be32_to_cpu(agf->agf_refcount_root);
+		daddr = XFS_AGB_TO_DADDR(mp, agno, agbno);
+		XFS_SCRUB_AGF_CHECK(agbno > XFS_AGI_BLOCK(mp));
+		XFS_SCRUB_AGF_CHECK(agbno < mp->m_sb.sb_agblocks);
+		XFS_SCRUB_AGF_CHECK(agbno < eoag);
+		XFS_SCRUB_AGF_CHECK(daddr < eofs);
+
+		level = be32_to_cpu(agf->agf_refcount_level);
+		XFS_SCRUB_AGF_CHECK(level > 0);
+		XFS_SCRUB_AGF_CHECK(level <= XFS_BTREE_MAXLEVELS);
+	}
+
+	/* Check the AGFL counters */
+	agfl_first = be32_to_cpu(agf->agf_flfirst);
+	agfl_last = be32_to_cpu(agf->agf_fllast);
+	agfl_count = be32_to_cpu(agf->agf_flcount);
+	if (agfl_last > agfl_first)
+		fl_count = agfl_last - agfl_first + 1;
+	else
+		fl_count = XFS_AGFL_SIZE(mp) - agfl_first + agfl_last + 1;
+	XFS_SCRUB_AGF_CHECK(agfl_count == 0 || fl_count == agfl_count);
+
+out:
+	return error;
+}
+#undef XFS_SCRUB_AGF_OP_ERROR_GOTO
+#undef XFS_SCRUB_AGF_CHECK
+
+/* AGFL */
+
+#define XFS_SCRUB_AGFL_CHECK(fs_ok) \
+	XFS_SCRUB_CHECK(sc, sc->sa.agfl_bp, "AGFL", fs_ok)
+struct xfs_scrub_agfl {
+	xfs_agblock_t			eoag;
+	xfs_daddr_t			eofs;
+};
+
+/* Scrub an AGFL block. */
+STATIC int
+xfs_scrub_agfl_block(
+	struct xfs_scrub_context	*sc,
+	xfs_agblock_t			agbno,
+	void				*priv)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	xfs_agnumber_t			agno = sc->sa.agno;
+	struct xfs_scrub_agfl		*sagfl = priv;
+
+	XFS_SCRUB_AGFL_CHECK(agbno > XFS_AGI_BLOCK(mp));
+	XFS_SCRUB_AGFL_CHECK(XFS_AGB_TO_DADDR(mp, agno, agbno) < sagfl->eofs);
+	XFS_SCRUB_AGFL_CHECK(agbno < mp->m_sb.sb_agblocks);
+	XFS_SCRUB_AGFL_CHECK(agbno < sagfl->eoag);
+
+	return 0;
+}
+
+#define XFS_SCRUB_AGFL_OP_ERROR_GOTO(error, label) \
+	XFS_SCRUB_OP_ERROR_GOTO(sc, sc->sm->sm_agno, \
+			XFS_AGFL_BLOCK(sc->tp->t_mountp), "AGFL", error, label)
+/* Scrub the AGFL. */
+int
+xfs_scrub_agfl(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_scrub_agfl		sagfl;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_agf			*agf;
+	int				error;
+
+	error = xfs_scrub_load_ag_headers(sc, sc->sm->sm_agno,
+			XFS_SCRUB_TYPE_AGFL);
+	XFS_SCRUB_AGFL_OP_ERROR_GOTO(&error, out);
+	if (!sc->sa.agf_bp)
+		return -EFSCORRUPTED;
+
+	agf = XFS_BUF_TO_AGF(sc->sa.agf_bp);
+	sagfl.eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
+	sagfl.eoag = be32_to_cpu(agf->agf_length);
+
+	/* Check the blocks in the AGFL. */
+	return xfs_scrub_walk_agfl(sc, xfs_scrub_agfl_block, &sagfl);
+out:
+	return error;
+}
+#undef XFS_SCRUB_AGFL_OP_ERROR_GOTO
+#undef XFS_SCRUB_AGFL_CHECK
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 924079d..a3d771c 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -586,6 +586,66 @@ xfs_scrub_setup(
 			0, 0, 0, &sc->tp);
 }
 
+/*
+ * Load and verify an AG header for further AG header examination.
+ * If this header is not the target of the examination, don't return
+ * the buffer if a runtime or verifier error occurs.
+ */
+STATIC int
+xfs_scrub_load_ag_header(
+	struct xfs_scrub_context	*sc,
+	xfs_daddr_t			daddr,
+	struct xfs_buf			**bpp,
+	const struct xfs_buf_ops	*ops,
+	bool				is_target)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	int				error;
+
+	*bpp = NULL;
+	error = xfs_trans_read_buf(mp, sc->tp, mp->m_ddev_targp,
+			XFS_AG_DADDR(mp, sc->sa.agno, daddr),
+			XFS_FSS_TO_BB(mp, 1), 0, bpp, ops);
+	return is_target ? error : 0;
+}
+
+/*
+ * Load as many of the AG headers and btree cursors as we can for an
+ * examination and cross-reference of an AG header.
+ */
+int
+xfs_scrub_load_ag_headers(
+	struct xfs_scrub_context	*sc,
+	xfs_agnumber_t			agno,
+	unsigned int			type)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	int				error;
+
+	ASSERT(type == XFS_SCRUB_TYPE_AGF || type == XFS_SCRUB_TYPE_AGFL);
+	memset(&sc->sa, 0, sizeof(sc->sa));
+	sc->sa.agno = agno;
+
+	error = xfs_scrub_load_ag_header(sc, XFS_AGI_DADDR(mp),
+			&sc->sa.agi_bp, &xfs_agi_buf_ops, false);
+	if (error)
+		return error;
+
+	error = xfs_scrub_load_ag_header(sc, XFS_AGF_DADDR(mp),
+			&sc->sa.agf_bp, &xfs_agf_buf_ops,
+			type == XFS_SCRUB_TYPE_AGF);
+	if (error)
+		return error;
+
+	error = xfs_scrub_load_ag_header(sc, XFS_AGFL_DADDR(mp),
+			&sc->sa.agfl_bp, &xfs_agfl_buf_ops,
+			type == XFS_SCRUB_TYPE_AGFL);
+	if (error)
+		return error;
+
+	return 0;
+}
+
 /* Scrubbing dispatch. */
 
 struct xfs_scrub_meta_fns {
@@ -599,6 +659,8 @@ struct xfs_scrub_meta_fns {
 static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup, xfs_scrub_dummy, NULL, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_superblock, NULL, NULL},
+	{xfs_scrub_setup_ag, xfs_scrub_agf, NULL, NULL},
+	{xfs_scrub_setup_ag, xfs_scrub_agfl, NULL, NULL},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index bd2896b..2f7075d 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -190,6 +190,12 @@ int xfs_scrub_ag_read_headers(struct xfs_scrub_context *sc, xfs_agnumber_t agno,
 			      struct xfs_buf **agfl);
 int xfs_scrub_ag_btcur_init(struct xfs_scrub_context *sc,
 			    struct xfs_scrub_ag *sa);
+int xfs_scrub_load_ag_headers(struct xfs_scrub_context *sc, xfs_agnumber_t agno,
+			      unsigned int type);
+int xfs_scrub_walk_agfl(struct xfs_scrub_context *sc,
+			int (*fn)(struct xfs_scrub_context *, xfs_agblock_t bno,
+				  void *),
+			void *priv);
 
 /* Setup functions */
 
@@ -202,5 +208,7 @@ int xfs_scrub_setup_ag(struct xfs_scrub_context *sc, struct xfs_inode *ip,
 /* Metadata scrubbers */
 
 int xfs_scrub_superblock(struct xfs_scrub_context *sc);
+int xfs_scrub_agf(struct xfs_scrub_context *sc);
+int xfs_scrub_agfl(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index ddfe15f..b8b0028 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3354,7 +3354,9 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 /* scrub */
 #define XFS_SCRUB_TYPE_DESC \
 	{ XFS_SCRUB_TYPE_TEST,		"dummy" }, \
-	{ XFS_SCRUB_TYPE_SB,		"superblock" }
+	{ XFS_SCRUB_TYPE_SB,		"superblock" }, \
+	{ XFS_SCRUB_TYPE_AGF,		"AGF" }, \
+	{ XFS_SCRUB_TYPE_AGFL,		"AGFL" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 19/55] xfs: scrub the AGI

[Darrick J. Wong]

Add a forgotten check to the AGI verifier, then wire up the scrub
infrastructure to check the AGI contents.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_fs.h  |    3 +
 fs/xfs/scrub/agheader.c |   96 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.c   |    7 ++-
 fs/xfs/scrub/common.h   |    1 
 fs/xfs/xfs_trace.h      |    3 +
 5 files changed, 106 insertions(+), 4 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 7434f16..15c74e5 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -578,7 +578,8 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_SB	1	/* superblock */
 #define XFS_SCRUB_TYPE_AGF	2	/* AG free header */
 #define XFS_SCRUB_TYPE_AGFL	3	/* AG free list */
-#define XFS_SCRUB_TYPE_MAX	3
+#define XFS_SCRUB_TYPE_AGI	4	/* AG inode header */
+#define XFS_SCRUB_TYPE_MAX	4
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index 94f29ae..f3bbd89 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -423,3 +423,99 @@ xfs_scrub_agfl(
 }
 #undef XFS_SCRUB_AGFL_OP_ERROR_GOTO
 #undef XFS_SCRUB_AGFL_CHECK
+
+/* AGI */
+
+#define XFS_SCRUB_AGI_CHECK(fs_ok) \
+	XFS_SCRUB_CHECK(sc, sc->sa.agi_bp, "AGI", fs_ok)
+#define XFS_SCRUB_AGI_OP_ERROR_GOTO(error, label) \
+	XFS_SCRUB_OP_ERROR_GOTO(sc, sc->sm->sm_agno, \
+			XFS_AGI_BLOCK(sc->tp->t_mountp), "AGI", error, label)
+/* Scrub the AGI. */
+int
+xfs_scrub_agi(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_agi			*agi;
+	xfs_daddr_t			daddr;
+	xfs_daddr_t			eofs;
+	xfs_agnumber_t			agno;
+	xfs_agblock_t			agbno;
+	xfs_agblock_t			eoag;
+	xfs_agino_t			agino;
+	xfs_agino_t			first_agino;
+	xfs_agino_t			last_agino;
+	int				i;
+	int				level;
+	int				error = 0;
+
+	agno = sc->sm->sm_agno;
+	error = xfs_scrub_load_ag_headers(sc, agno, XFS_SCRUB_TYPE_AGI);
+	XFS_SCRUB_AGI_OP_ERROR_GOTO(&error, out);
+
+	agi = XFS_BUF_TO_AGI(sc->sa.agi_bp);
+	eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
+
+	/* Check the AG length */
+	eoag = be32_to_cpu(agi->agi_length);
+	XFS_SCRUB_AGI_CHECK(eoag == xfs_scrub_ag_blocks(mp, agno));
+
+	/* Check btree roots and levels */
+	agbno = be32_to_cpu(agi->agi_root);
+	daddr = XFS_AGB_TO_DADDR(mp, agno, agbno);
+	XFS_SCRUB_AGI_CHECK(agbno > XFS_AGI_BLOCK(mp));
+	XFS_SCRUB_AGI_CHECK(agbno < mp->m_sb.sb_agblocks);
+	XFS_SCRUB_AGI_CHECK(agbno < eoag);
+	XFS_SCRUB_AGI_CHECK(daddr < eofs);
+
+	level = be32_to_cpu(agi->agi_level);
+	XFS_SCRUB_AGI_CHECK(level > 0);
+	XFS_SCRUB_AGI_CHECK(level <= XFS_BTREE_MAXLEVELS);
+
+	if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
+		agbno = be32_to_cpu(agi->agi_free_root);
+		daddr = XFS_AGB_TO_DADDR(mp, agno, agbno);
+		XFS_SCRUB_AGI_CHECK(agbno > XFS_AGI_BLOCK(mp));
+		XFS_SCRUB_AGI_CHECK(agbno < mp->m_sb.sb_agblocks);
+		XFS_SCRUB_AGI_CHECK(agbno < eoag);
+		XFS_SCRUB_AGI_CHECK(daddr < eofs);
+
+		level = be32_to_cpu(agi->agi_free_level);
+		XFS_SCRUB_AGI_CHECK(level > 0);
+		XFS_SCRUB_AGI_CHECK(level <= XFS_BTREE_MAXLEVELS);
+	}
+
+	/* Check inode counters */
+	first_agino = XFS_OFFBNO_TO_AGINO(mp, XFS_AGI_BLOCK(mp) + 1, 0);
+	last_agino = XFS_OFFBNO_TO_AGINO(mp, eoag + 1, 0) - 1;
+	agino = be32_to_cpu(agi->agi_count);
+	XFS_SCRUB_AGI_CHECK(agino <= last_agino - first_agino + 1);
+	XFS_SCRUB_AGI_CHECK(agino >= be32_to_cpu(agi->agi_freecount));
+
+	/* Check inode pointers */
+	agino = be32_to_cpu(agi->agi_newino);
+	if (agino != NULLAGINO) {
+		XFS_SCRUB_AGI_CHECK(agino >= first_agino);
+		XFS_SCRUB_AGI_CHECK(agino <= last_agino);
+	}
+	agino = be32_to_cpu(agi->agi_dirino);
+	if (agino != NULLAGINO) {
+		XFS_SCRUB_AGI_CHECK(agino >= first_agino);
+		XFS_SCRUB_AGI_CHECK(agino <= last_agino);
+	}
+
+	/* Check unlinked inode buckets */
+	for (i = 0; i < XFS_AGI_UNLINKED_BUCKETS; i++) {
+		agino = be32_to_cpu(agi->agi_unlinked[i]);
+		if (agino == NULLAGINO)
+			continue;
+		XFS_SCRUB_AGI_CHECK(agino >= first_agino);
+		XFS_SCRUB_AGI_CHECK(agino <= last_agino);
+	}
+
+out:
+	return error;
+}
+#undef XFS_SCRUB_AGI_CHECK
+#undef XFS_SCRUB_AGI_OP_ERROR_GOTO
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index a3d771c..3b378d0 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -622,12 +622,14 @@ xfs_scrub_load_ag_headers(
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 	int				error;
 
-	ASSERT(type == XFS_SCRUB_TYPE_AGF || type == XFS_SCRUB_TYPE_AGFL);
+	ASSERT(type == XFS_SCRUB_TYPE_AGF || type == XFS_SCRUB_TYPE_AGFL ||
+	       type == XFS_SCRUB_TYPE_AGI);
 	memset(&sc->sa, 0, sizeof(sc->sa));
 	sc->sa.agno = agno;
 
 	error = xfs_scrub_load_ag_header(sc, XFS_AGI_DADDR(mp),
-			&sc->sa.agi_bp, &xfs_agi_buf_ops, false);
+			&sc->sa.agi_bp, &xfs_agi_buf_ops,
+			type == XFS_SCRUB_TYPE_AGI);
 	if (error)
 		return error;
 
@@ -661,6 +663,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag, xfs_scrub_superblock, NULL, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agf, NULL, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agfl, NULL, NULL},
+	{xfs_scrub_setup_ag, xfs_scrub_agi, NULL, NULL},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 2f7075d..ec4c230 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -210,5 +210,6 @@ int xfs_scrub_setup_ag(struct xfs_scrub_context *sc, struct xfs_inode *ip,
 int xfs_scrub_superblock(struct xfs_scrub_context *sc);
 int xfs_scrub_agf(struct xfs_scrub_context *sc);
 int xfs_scrub_agfl(struct xfs_scrub_context *sc);
+int xfs_scrub_agi(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index b8b0028..38118f5 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3356,7 +3356,8 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_TEST,		"dummy" }, \
 	{ XFS_SCRUB_TYPE_SB,		"superblock" }, \
 	{ XFS_SCRUB_TYPE_AGF,		"AGF" }, \
-	{ XFS_SCRUB_TYPE_AGFL,		"AGFL" }
+	{ XFS_SCRUB_TYPE_AGFL,		"AGFL" }, \
+	{ XFS_SCRUB_TYPE_AGI,		"AGI" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 20/55] xfs: support scrubbing free space btrees

[Darrick J. Wong]

Check the extent records free space btrees to ensure that the values
look sane.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile                 |    1 
 fs/xfs/libxfs/xfs_alloc_btree.c |    6 --
 fs/xfs/libxfs/xfs_fs.h          |    4 +
 fs/xfs/scrub/alloc.c            |  114 +++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.c           |    2 +
 fs/xfs/scrub/common.h           |    6 ++
 fs/xfs/xfs_trace.h              |    4 +
 7 files changed, 129 insertions(+), 8 deletions(-)
 create mode 100644 fs/xfs/scrub/alloc.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index ca7399a..3168655 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -105,6 +105,7 @@ xfs-y				+= xfs_aops.o \
 # online scrub/repair
 xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   agheader.o \
+				   alloc.o \
 				   btree.o \
 				   common.o \
 				   )
diff --git a/fs/xfs/libxfs/xfs_alloc_btree.c b/fs/xfs/libxfs/xfs_alloc_btree.c
index ba3ec9c..10d6c12 100644
--- a/fs/xfs/libxfs/xfs_alloc_btree.c
+++ b/fs/xfs/libxfs/xfs_alloc_btree.c
@@ -386,7 +386,6 @@ const struct xfs_buf_ops xfs_allocbt_buf_ops = {
 };
 
 
-#if defined(DEBUG) || defined(XFS_WARN)
 STATIC int
 xfs_bnobt_keys_inorder(
 	struct xfs_btree_cur	*cur,
@@ -433,7 +432,6 @@ xfs_cntbt_recs_inorder(
 		 be32_to_cpu(r1->alloc.ar_startblock) <
 		 be32_to_cpu(r2->alloc.ar_startblock));
 }
-#endif /* DEBUG */
 
 static const struct xfs_btree_ops xfs_bnobt_ops = {
 	.rec_len		= sizeof(xfs_alloc_rec_t),
@@ -453,10 +451,8 @@ static const struct xfs_btree_ops xfs_bnobt_ops = {
 	.key_diff		= xfs_bnobt_key_diff,
 	.buf_ops		= &xfs_allocbt_buf_ops,
 	.diff_two_keys		= xfs_bnobt_diff_two_keys,
-#if defined(DEBUG) || defined(XFS_WARN)
 	.keys_inorder		= xfs_bnobt_keys_inorder,
 	.recs_inorder		= xfs_bnobt_recs_inorder,
-#endif
 };
 
 static const struct xfs_btree_ops xfs_cntbt_ops = {
@@ -476,10 +472,8 @@ static const struct xfs_btree_ops xfs_cntbt_ops = {
 	.key_diff		= xfs_cntbt_key_diff,
 	.buf_ops		= &xfs_allocbt_buf_ops,
 	.diff_two_keys		= xfs_cntbt_diff_two_keys,
-#if defined(DEBUG) || defined(XFS_WARN)
 	.keys_inorder		= xfs_cntbt_keys_inorder,
 	.recs_inorder		= xfs_cntbt_recs_inorder,
-#endif
 };
 
 /*
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 15c74e5..446dd14 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -579,7 +579,9 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_AGF	2	/* AG free header */
 #define XFS_SCRUB_TYPE_AGFL	3	/* AG free list */
 #define XFS_SCRUB_TYPE_AGI	4	/* AG inode header */
-#define XFS_SCRUB_TYPE_MAX	4
+#define XFS_SCRUB_TYPE_BNOBT	5	/* freesp by block btree */
+#define XFS_SCRUB_TYPE_CNTBT	6	/* freesp by length btree */
+#define XFS_SCRUB_TYPE_MAX	6
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/scrub/alloc.c b/fs/xfs/scrub/alloc.c
new file mode 100644
index 0000000..2edf85d
--- /dev/null
+++ b/fs/xfs/scrub/alloc.c
@@ -0,0 +1,114 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_rmap.h"
+#include "scrub/common.h"
+#include "scrub/btree.h"
+
+/* Set us up with AG headers and btree cursors. */
+int
+xfs_scrub_setup_ag_header(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	int				error;
+
+	error = xfs_scrub_setup_ag(sc, ip, sm, retry_deadlocked);
+	if (error)
+		goto out;
+
+	error = xfs_scrub_ag_init(sc, sm->sm_agno, &sc->sa);
+	if (error)
+		xfs_trans_cancel(sc->tp);
+out:
+	return error;
+}
+
+/* Free space btree scrubber. */
+
+/* Scrub a bnobt/cntbt record. */
+STATIC int
+xfs_scrub_allocbt_helper(
+	struct xfs_scrub_btree		*bs,
+	union xfs_btree_rec		*rec)
+{
+	struct xfs_mount		*mp = bs->cur->bc_mp;
+	struct xfs_agf			*agf;
+	xfs_agblock_t			bno;
+	xfs_extlen_t			len;
+	int				error = 0;
+
+	bno = be32_to_cpu(rec->alloc.ar_startblock);
+	len = be32_to_cpu(rec->alloc.ar_blockcount);
+	agf = XFS_BUF_TO_AGF(bs->sc->sa.agf_bp);
+
+	XFS_SCRUB_BTREC_CHECK(bs, bno < mp->m_sb.sb_agblocks);
+	XFS_SCRUB_BTREC_CHECK(bs, bno < be32_to_cpu(agf->agf_length));
+	XFS_SCRUB_BTREC_CHECK(bs, bno < bno + len);
+	XFS_SCRUB_BTREC_CHECK(bs, (unsigned long long)bno + len <=
+			mp->m_sb.sb_agblocks);
+	XFS_SCRUB_BTREC_CHECK(bs, (unsigned long long)bno + len <=
+			be32_to_cpu(agf->agf_length));
+
+	return error;
+}
+
+/* Scrub the freespace btrees for some AG. */
+STATIC int
+xfs_scrub_allocbt(
+	struct xfs_scrub_context	*sc,
+	xfs_btnum_t			which)
+{
+	struct xfs_owner_info		oinfo;
+	struct xfs_btree_cur		*cur;
+
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_AG);
+	cur = which == XFS_BTNUM_BNO ? sc->sa.bno_cur : sc->sa.cnt_cur;
+	return xfs_scrub_btree(sc, cur, xfs_scrub_allocbt_helper,
+			&oinfo, NULL);
+}
+
+int
+xfs_scrub_bnobt(
+	struct xfs_scrub_context	*sc)
+{
+	return xfs_scrub_allocbt(sc, XFS_BTNUM_BNO);
+}
+
+int
+xfs_scrub_cntbt(
+	struct xfs_scrub_context	*sc)
+{
+	return xfs_scrub_allocbt(sc, XFS_BTNUM_CNT);
+}
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 3b378d0..bc7c7a0 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -664,6 +664,8 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag, xfs_scrub_agf, NULL, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agfl, NULL, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agi, NULL, NULL},
+	{xfs_scrub_setup_ag_header, xfs_scrub_bnobt, NULL, NULL},
+	{xfs_scrub_setup_ag_header, xfs_scrub_cntbt, NULL, NULL},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index ec4c230..b4a907f 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -204,6 +204,10 @@ int xfs_scrub_setup(struct xfs_scrub_context *sc, struct xfs_inode *ip,
 		    struct xfs_scrub_metadata *sm, bool retry_deadlocked);
 int xfs_scrub_setup_ag(struct xfs_scrub_context *sc, struct xfs_inode *ip,
 		       struct xfs_scrub_metadata *sm, bool retry_deadlocked);
+int xfs_scrub_setup_ag_header(struct xfs_scrub_context *sc,
+			      struct xfs_inode *ip,
+			      struct xfs_scrub_metadata *sm,
+			      bool retry_deadlocked);
 
 /* Metadata scrubbers */
 
@@ -211,5 +215,7 @@ int xfs_scrub_superblock(struct xfs_scrub_context *sc);
 int xfs_scrub_agf(struct xfs_scrub_context *sc);
 int xfs_scrub_agfl(struct xfs_scrub_context *sc);
 int xfs_scrub_agi(struct xfs_scrub_context *sc);
+int xfs_scrub_bnobt(struct xfs_scrub_context *sc);
+int xfs_scrub_cntbt(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 38118f5..5cc7fed 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3357,7 +3357,9 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_SB,		"superblock" }, \
 	{ XFS_SCRUB_TYPE_AGF,		"AGF" }, \
 	{ XFS_SCRUB_TYPE_AGFL,		"AGFL" }, \
-	{ XFS_SCRUB_TYPE_AGI,		"AGI" }
+	{ XFS_SCRUB_TYPE_AGI,		"AGI" }, \
+	{ XFS_SCRUB_TYPE_BNOBT,		"bnobt" }, \
+	{ XFS_SCRUB_TYPE_CNTBT,		"cntbt" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 21/55] xfs: support scrubbing inode btrees

[Darrick J. Wong]

Check the records of the inode btrees to make sure that the values
make sense given the inode records themselves.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile                  |    1 
 fs/xfs/libxfs/xfs_fs.h           |    4 
 fs/xfs/libxfs/xfs_ialloc.c       |   41 +++-
 fs/xfs/libxfs/xfs_ialloc.h       |    3 
 fs/xfs/libxfs/xfs_ialloc_btree.c |   32 +++
 fs/xfs/scrub/common.c            |    2 
 fs/xfs/scrub/common.h            |    7 +
 fs/xfs/scrub/ialloc.c            |  359 ++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_icache.c              |    4 
 fs/xfs/xfs_icache.h              |    7 -
 fs/xfs/xfs_trace.h               |    4 
 11 files changed, 437 insertions(+), 27 deletions(-)
 create mode 100644 fs/xfs/scrub/ialloc.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 3168655..0c98a98 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -108,6 +108,7 @@ xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   alloc.o \
 				   btree.o \
 				   common.o \
+				   ialloc.o \
 				   )
 
 # low-level transaction/log code
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 446dd14..5fc45f5 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -581,7 +581,9 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_AGI	4	/* AG inode header */
 #define XFS_SCRUB_TYPE_BNOBT	5	/* freesp by block btree */
 #define XFS_SCRUB_TYPE_CNTBT	6	/* freesp by length btree */
-#define XFS_SCRUB_TYPE_MAX	6
+#define XFS_SCRUB_TYPE_INOBT	7	/* inode btree */
+#define XFS_SCRUB_TYPE_FINOBT	8	/* free inode btree */
+#define XFS_SCRUB_TYPE_MAX	8
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/libxfs/xfs_ialloc.c b/fs/xfs/libxfs/xfs_ialloc.c
index f272abf..e2f93e6 100644
--- a/fs/xfs/libxfs/xfs_ialloc.c
+++ b/fs/xfs/libxfs/xfs_ialloc.c
@@ -99,24 +99,14 @@ xfs_inobt_update(
 	return xfs_btree_update(cur, &rec);
 }
 
-/*
- * Get the data from the pointed-to record.
- */
-int					/* error */
-xfs_inobt_get_rec(
-	struct xfs_btree_cur	*cur,	/* btree cursor */
-	xfs_inobt_rec_incore_t	*irec,	/* btree record */
-	int			*stat)	/* output: success/failure */
+void
+xfs_inobt_btrec_to_irec(
+	struct xfs_mount		*mp,
+	union xfs_btree_rec		*rec,
+	struct xfs_inobt_rec_incore	*irec)
 {
-	union xfs_btree_rec	*rec;
-	int			error;
-
-	error = xfs_btree_get_rec(cur, &rec, stat);
-	if (error || *stat == 0)
-		return error;
-
 	irec->ir_startino = be32_to_cpu(rec->inobt.ir_startino);
-	if (xfs_sb_version_hassparseinodes(&cur->bc_mp->m_sb)) {
+	if (xfs_sb_version_hassparseinodes(&mp->m_sb)) {
 		irec->ir_holemask = be16_to_cpu(rec->inobt.ir_u.sp.ir_holemask);
 		irec->ir_count = rec->inobt.ir_u.sp.ir_count;
 		irec->ir_freecount = rec->inobt.ir_u.sp.ir_freecount;
@@ -131,6 +121,25 @@ xfs_inobt_get_rec(
 				be32_to_cpu(rec->inobt.ir_u.f.ir_freecount);
 	}
 	irec->ir_free = be64_to_cpu(rec->inobt.ir_free);
+}
+
+/*
+ * Get the data from the pointed-to record.
+ */
+int					/* error */
+xfs_inobt_get_rec(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_inobt_rec_incore_t	*irec,	/* btree record */
+	int			*stat)	/* output: success/failure */
+{
+	union xfs_btree_rec	*rec;
+	int			error;
+
+	error = xfs_btree_get_rec(cur, &rec, stat);
+	if (error || *stat == 0)
+		return error;
+
+	xfs_inobt_btrec_to_irec(cur->bc_mp, rec, irec);
 
 	return 0;
 }
diff --git a/fs/xfs/libxfs/xfs_ialloc.h b/fs/xfs/libxfs/xfs_ialloc.h
index 0bb8966..8e5861d 100644
--- a/fs/xfs/libxfs/xfs_ialloc.h
+++ b/fs/xfs/libxfs/xfs_ialloc.h
@@ -168,5 +168,8 @@ int xfs_ialloc_inode_init(struct xfs_mount *mp, struct xfs_trans *tp,
 int xfs_read_agi(struct xfs_mount *mp, struct xfs_trans *tp,
 		xfs_agnumber_t agno, struct xfs_buf **bpp);
 
+union xfs_btree_rec;
+void xfs_inobt_btrec_to_irec(struct xfs_mount *mp, union xfs_btree_rec *rec,
+		struct xfs_inobt_rec_incore *irec);
 
 #endif	/* __XFS_IALLOC_H__ */
diff --git a/fs/xfs/libxfs/xfs_ialloc_btree.c b/fs/xfs/libxfs/xfs_ialloc_btree.c
index 0fd086d..09d8cb0 100644
--- a/fs/xfs/libxfs/xfs_ialloc_btree.c
+++ b/fs/xfs/libxfs/xfs_ialloc_btree.c
@@ -152,6 +152,18 @@ xfs_inobt_init_key_from_rec(
 }
 
 STATIC void
+xfs_inobt_init_high_key_from_rec(
+	union xfs_btree_key	*key,
+	union xfs_btree_rec	*rec)
+{
+	__u32			x;
+
+	x = be32_to_cpu(rec->inobt.ir_startino);
+	x += XFS_INODES_PER_CHUNK - 1;
+	key->inobt.ir_startino = cpu_to_be32(x);
+}
+
+STATIC void
 xfs_inobt_init_rec_from_cur(
 	struct xfs_btree_cur	*cur,
 	union xfs_btree_rec	*rec)
@@ -205,6 +217,16 @@ xfs_inobt_key_diff(
 			  cur->bc_rec.i.ir_startino;
 }
 
+STATIC __int64_t
+xfs_inobt_diff_two_keys(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*k1,
+	union xfs_btree_key	*k2)
+{
+	return (__int64_t)be32_to_cpu(k1->inobt.ir_startino) -
+			  be32_to_cpu(k2->inobt.ir_startino);
+}
+
 static int
 xfs_inobt_verify(
 	struct xfs_buf		*bp)
@@ -279,7 +301,6 @@ const struct xfs_buf_ops xfs_inobt_buf_ops = {
 	.verify_write = xfs_inobt_write_verify,
 };
 
-#if defined(DEBUG) || defined(XFS_WARN)
 STATIC int
 xfs_inobt_keys_inorder(
 	struct xfs_btree_cur	*cur,
@@ -299,7 +320,6 @@ xfs_inobt_recs_inorder(
 	return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
 		be32_to_cpu(r2->inobt.ir_startino);
 }
-#endif	/* DEBUG */
 
 static const struct xfs_btree_ops xfs_inobt_ops = {
 	.rec_len		= sizeof(xfs_inobt_rec_t),
@@ -312,14 +332,14 @@ static const struct xfs_btree_ops xfs_inobt_ops = {
 	.get_minrecs		= xfs_inobt_get_minrecs,
 	.get_maxrecs		= xfs_inobt_get_maxrecs,
 	.init_key_from_rec	= xfs_inobt_init_key_from_rec,
+	.init_high_key_from_rec	= xfs_inobt_init_high_key_from_rec,
 	.init_rec_from_cur	= xfs_inobt_init_rec_from_cur,
 	.init_ptr_from_cur	= xfs_inobt_init_ptr_from_cur,
 	.key_diff		= xfs_inobt_key_diff,
 	.buf_ops		= &xfs_inobt_buf_ops,
-#if defined(DEBUG) || defined(XFS_WARN)
+	.diff_two_keys		= xfs_inobt_diff_two_keys,
 	.keys_inorder		= xfs_inobt_keys_inorder,
 	.recs_inorder		= xfs_inobt_recs_inorder,
-#endif
 };
 
 static const struct xfs_btree_ops xfs_finobt_ops = {
@@ -333,14 +353,14 @@ static const struct xfs_btree_ops xfs_finobt_ops = {
 	.get_minrecs		= xfs_inobt_get_minrecs,
 	.get_maxrecs		= xfs_inobt_get_maxrecs,
 	.init_key_from_rec	= xfs_inobt_init_key_from_rec,
+	.init_high_key_from_rec	= xfs_inobt_init_high_key_from_rec,
 	.init_rec_from_cur	= xfs_inobt_init_rec_from_cur,
 	.init_ptr_from_cur	= xfs_finobt_init_ptr_from_cur,
 	.key_diff		= xfs_inobt_key_diff,
 	.buf_ops		= &xfs_inobt_buf_ops,
-#if defined(DEBUG) || defined(XFS_WARN)
+	.diff_two_keys		= xfs_inobt_diff_two_keys,
 	.keys_inorder		= xfs_inobt_keys_inorder,
 	.recs_inorder		= xfs_inobt_recs_inorder,
-#endif
 };
 
 /*
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index bc7c7a0..e57d775 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -666,6 +666,8 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag, xfs_scrub_agi, NULL, NULL},
 	{xfs_scrub_setup_ag_header, xfs_scrub_bnobt, NULL, NULL},
 	{xfs_scrub_setup_ag_header, xfs_scrub_cntbt, NULL, NULL},
+	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_inobt, NULL, NULL},
+	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_finobt, NULL, xfs_sb_version_hasfinobt},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index b4a907f..69177c8 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -59,6 +59,7 @@ struct xfs_scrub_context {
 	struct xfs_scrub_metadata	*sm;
 	struct xfs_trans		*tp;
 	struct xfs_inode		*ip;
+	bool				retry;
 
 	/* State tracking for multi-AG operations. */
 	struct xfs_scrub_ag_lock	ag_lock;
@@ -208,6 +209,10 @@ int xfs_scrub_setup_ag_header(struct xfs_scrub_context *sc,
 			      struct xfs_inode *ip,
 			      struct xfs_scrub_metadata *sm,
 			      bool retry_deadlocked);
+int xfs_scrub_setup_ag_iallocbt(struct xfs_scrub_context *sc,
+				struct xfs_inode *ip,
+				struct xfs_scrub_metadata *sm,
+				bool retry_deadlocked);
 
 /* Metadata scrubbers */
 
@@ -217,5 +222,7 @@ int xfs_scrub_agfl(struct xfs_scrub_context *sc);
 int xfs_scrub_agi(struct xfs_scrub_context *sc);
 int xfs_scrub_bnobt(struct xfs_scrub_context *sc);
 int xfs_scrub_cntbt(struct xfs_scrub_context *sc);
+int xfs_scrub_inobt(struct xfs_scrub_context *sc);
+int xfs_scrub_finobt(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/ialloc.c b/fs/xfs/scrub/ialloc.c
new file mode 100644
index 0000000..0c654c9
--- /dev/null
+++ b/fs/xfs/scrub/ialloc.c
@@ -0,0 +1,359 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_ialloc.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_icache.h"
+#include "xfs_rmap.h"
+#include "xfs_log.h"
+#include "xfs_trans_priv.h"
+#include "scrub/common.h"
+#include "scrub/btree.h"
+
+/*
+ * Set us up with AG headers and btree cursors.  Push everything out
+ * of the log so that we can correlate inodes to inobt.
+ */
+int
+xfs_scrub_setup_ag_iallocbt(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+	int				error;
+
+	/* Push everything out of the log onto disk prior to checking. */
+	if (retry_deadlocked) {
+		error = _xfs_log_force(mp, XFS_LOG_SYNC, NULL);
+		if (error)
+			goto out;
+		xfs_ail_push_all_sync(mp->m_ail);
+	}
+
+	error = xfs_scrub_setup_ag_header(sc, ip, sm, retry_deadlocked);
+	if (error)
+		goto out;
+	sc->retry = retry_deadlocked;
+out:
+	return error;
+}
+
+/* Inode btree scrubber. */
+
+/* Scrub a chunk of an inobt record. */
+STATIC int
+xfs_scrub_iallocbt_chunk(
+	struct xfs_scrub_btree		*bs,
+	struct xfs_inobt_rec_incore	*irec,
+	xfs_agino_t			agino,
+	xfs_extlen_t			len,
+	bool				*keep_scanning)
+{
+	struct xfs_mount		*mp = bs->cur->bc_mp;
+	struct xfs_agf			*agf;
+	xfs_agblock_t			eoag;
+	xfs_agblock_t			bno;
+	int				error = 0;
+
+	agf = XFS_BUF_TO_AGF(bs->sc->sa.agf_bp);
+	eoag = be32_to_cpu(agf->agf_length);
+	bno = XFS_AGINO_TO_AGBNO(mp, agino);
+
+	*keep_scanning = true;
+	XFS_SCRUB_BTREC_CHECK(bs, bno < mp->m_sb.sb_agblocks);
+	XFS_SCRUB_BTREC_CHECK(bs, bno < eoag);
+	XFS_SCRUB_BTREC_CHECK(bs, bno < bno + len);
+	XFS_SCRUB_BTREC_CHECK(bs, (unsigned long long)bno + len <=
+			mp->m_sb.sb_agblocks);
+	XFS_SCRUB_BTREC_CHECK(bs, (unsigned long long)bno + len <=
+			eoag);
+	if (error) {
+		*keep_scanning = false;
+		goto out;
+	}
+
+out:
+	return error;
+}
+
+/* Count the number of free inodes. */
+static unsigned int
+xfs_scrub_iallocbt_freecount(
+	xfs_inofree_t			freemask)
+{
+	int				bits = XFS_INODES_PER_CHUNK;
+	unsigned int			ret = 0;
+
+	while (bits--) {
+		if (freemask & 1)
+			ret++;
+		freemask >>= 1;
+	}
+
+	return ret;
+}
+
+/* Check a particular inode with ir_free. */
+STATIC int
+xfs_scrub_iallocbt_check_cluster_freemask(
+	struct xfs_scrub_btree		*bs,
+	xfs_ino_t			fsino,
+	xfs_agino_t			chunkino,
+	xfs_agino_t			clusterino,
+	struct xfs_inobt_rec_incore	*irec,
+	struct xfs_buf			*bp)
+{
+	struct xfs_dinode		*dip;
+	struct xfs_inode		*ip;
+	struct xfs_mount		*mp = bs->cur->bc_mp;
+	bool				freemask_ok;
+	int				error;
+
+	dip = xfs_buf_offset(bp, clusterino * mp->m_sb.sb_inodesize);
+	XFS_SCRUB_BTREC_GOTO(bs,
+			be16_to_cpu(dip->di_magic) == XFS_DINODE_MAGIC,
+			out);
+	XFS_SCRUB_BTREC_GOTO(bs,
+			dip->di_version < 3 || be64_to_cpu(dip->di_ino) ==
+				fsino + clusterino,
+			out);
+	freemask_ok = !!(irec->ir_free & XFS_INOBT_MASK(chunkino + clusterino));
+	error = xfs_iget(mp, bs->cur->bc_tp, fsino + clusterino,
+			XFS_IGET_HITONLY, 0, &ip);
+	if (error == -ENOENT) {
+		; /* i_mode is 0 */
+	} else if (!error && ip) {
+		freemask_ok ^= !!(VFS_I(ip)->i_mode);
+		IRELE(ip);
+	} else {
+		freemask_ok ^= !!(dip->di_mode);
+		if (!bs->sc->retry && !freemask_ok)
+			return -EDEADLOCK;
+	}
+	XFS_SCRUB_BTREC_CHECK(bs, freemask_ok);
+out:
+	return 0;
+}
+
+/* Make sure the free mask is consistent with what the inodes think. */
+STATIC int
+xfs_scrub_iallocbt_check_freemask(
+	struct xfs_scrub_btree		*bs,
+	struct xfs_inobt_rec_incore	*irec)
+{
+	struct xfs_owner_info		oinfo;
+	struct xfs_imap			imap;
+	struct xfs_mount		*mp = bs->cur->bc_mp;
+	struct xfs_dinode		*dip;
+	struct xfs_buf			*bp;
+	xfs_ino_t			fsino;
+	xfs_agino_t			nr_inodes;
+	xfs_agino_t			agino;
+	xfs_agino_t			chunkino;
+	xfs_agino_t			clusterino;
+	xfs_agblock_t			agbno;
+	int				blks_per_cluster;
+	__uint16_t			holemask;
+	__uint16_t			ir_holemask;
+	int				error = 0;
+
+	/* Make sure the freemask matches the inode records. */
+	blks_per_cluster = xfs_icluster_size_fsb(mp);
+	nr_inodes = XFS_OFFBNO_TO_AGINO(mp, blks_per_cluster, 0);
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INODES);
+
+	for (agino = irec->ir_startino;
+	     agino < irec->ir_startino + XFS_INODES_PER_CHUNK;
+	     agino += blks_per_cluster * mp->m_sb.sb_inopblock) {
+		fsino = XFS_AGINO_TO_INO(mp, bs->cur->bc_private.a.agno, agino);
+		chunkino = agino - irec->ir_startino;
+		agbno = XFS_AGINO_TO_AGBNO(mp, agino);
+
+		/* Compute the holemask mask for this cluster. */
+		for (clusterino = 0, holemask = 0; clusterino < nr_inodes;
+		     clusterino += XFS_INODES_PER_HOLEMASK_BIT)
+			holemask |= XFS_INOBT_MASK((chunkino + clusterino) /
+					XFS_INODES_PER_HOLEMASK_BIT);
+
+		/* The whole cluster must be a hole or not a hole. */
+		ir_holemask = (irec->ir_holemask & holemask);
+		XFS_SCRUB_BTREC_CHECK(bs, ir_holemask == holemask ||
+				ir_holemask == 0);
+
+		/* If any part of this is a hole, skip it. */
+		if (ir_holemask)
+			continue;
+
+		/* Grab the inode cluster buffer. */
+		imap.im_blkno = XFS_AGB_TO_DADDR(mp, bs->cur->bc_private.a.agno,
+				agbno);
+		imap.im_len = XFS_FSB_TO_BB(mp, blks_per_cluster);
+		imap.im_boffset = 0;
+
+		error = xfs_imap_to_bp(mp, bs->cur->bc_tp, &imap,
+				&dip, &bp, 0, 0);
+		XFS_SCRUB_BTREC_OP_ERROR_GOTO(bs, &error, next_cluster);
+
+		/* Which inodes are free? */
+		for (clusterino = 0; clusterino < nr_inodes; clusterino++) {
+			error = xfs_scrub_iallocbt_check_cluster_freemask(bs,
+					fsino, chunkino, clusterino, irec, bp);
+			if (error) {
+				xfs_trans_brelse(bs->cur->bc_tp, bp);
+				return error;
+			}
+		}
+
+		xfs_trans_brelse(bs->cur->bc_tp, bp);
+next_cluster:
+		;
+	}
+
+	return error;
+}
+
+/* Scrub an inobt/finobt record. */
+STATIC int
+xfs_scrub_iallocbt_helper(
+	struct xfs_scrub_btree		*bs,
+	union xfs_btree_rec		*rec)
+{
+	struct xfs_mount		*mp = bs->cur->bc_mp;
+	struct xfs_agi			*agi;
+	struct xfs_inobt_rec_incore	irec;
+	uint64_t			holes;
+	xfs_agino_t			agino;
+	xfs_agblock_t			agbno;
+	xfs_extlen_t			len;
+	bool				keep_scanning;
+	int				holecount;
+	int				i;
+	int				error = 0;
+	int				err2 = 0;
+	unsigned int			real_freecount;
+	__uint16_t			holemask;
+
+	xfs_inobt_btrec_to_irec(mp, rec, &irec);
+
+	XFS_SCRUB_BTREC_CHECK(bs, irec.ir_count <= XFS_INODES_PER_CHUNK);
+	XFS_SCRUB_BTREC_CHECK(bs, irec.ir_freecount <= XFS_INODES_PER_CHUNK);
+	real_freecount = irec.ir_freecount +
+			(XFS_INODES_PER_CHUNK - irec.ir_count);
+	XFS_SCRUB_BTREC_CHECK(bs, real_freecount ==
+			xfs_scrub_iallocbt_freecount(irec.ir_free));
+	agi = XFS_BUF_TO_AGI(bs->sc->sa.agi_bp);
+	agino = irec.ir_startino;
+	agbno = XFS_AGINO_TO_AGBNO(mp, irec.ir_startino);
+	XFS_SCRUB_BTREC_GOTO(bs, agbno < be32_to_cpu(agi->agi_length), out);
+
+	/* Handle non-sparse inodes */
+	if (!xfs_inobt_issparse(irec.ir_holemask)) {
+		len = XFS_B_TO_FSB(mp,
+				XFS_INODES_PER_CHUNK * mp->m_sb.sb_inodesize);
+		XFS_SCRUB_BTREC_CHECK(bs,
+				irec.ir_count == XFS_INODES_PER_CHUNK);
+
+		error = xfs_scrub_iallocbt_chunk(bs, &irec, agino, len,
+				&keep_scanning);
+		if (error)
+			goto out;
+		goto check_freemask;
+	}
+
+	/* Check each chunk of a sparse inode cluster. */
+	holemask = irec.ir_holemask;
+	holecount = 0;
+	len = XFS_B_TO_FSB(mp,
+			XFS_INODES_PER_HOLEMASK_BIT * mp->m_sb.sb_inodesize);
+	holes = ~xfs_inobt_irec_to_allocmask(&irec);
+	XFS_SCRUB_BTREC_CHECK(bs, (holes & irec.ir_free) == holes);
+	XFS_SCRUB_BTREC_CHECK(bs, irec.ir_freecount <= irec.ir_count);
+
+	for (i = 0; i < XFS_INOBT_HOLEMASK_BITS; holemask >>= 1,
+			i++, agino += XFS_INODES_PER_HOLEMASK_BIT) {
+		if (holemask & 1) {
+			holecount += XFS_INODES_PER_HOLEMASK_BIT;
+			continue;
+		}
+
+		err2 = xfs_scrub_iallocbt_chunk(bs, &irec, agino, len,
+				&keep_scanning);
+		if (!error && err2)
+			error = err2;
+		if (!keep_scanning)
+			break;
+	}
+
+	XFS_SCRUB_BTREC_CHECK(bs, holecount <= XFS_INODES_PER_CHUNK);
+	XFS_SCRUB_BTREC_CHECK(bs, holecount + irec.ir_count ==
+			XFS_INODES_PER_CHUNK);
+
+check_freemask:
+	error = xfs_scrub_iallocbt_check_freemask(bs, &irec);
+	if (error)
+		goto out;
+
+out:
+	return error;
+}
+
+/* Scrub the inode btrees for some AG. */
+STATIC int
+xfs_scrub_iallocbt(
+	struct xfs_scrub_context	*sc,
+	xfs_btnum_t			which)
+{
+	struct xfs_btree_cur		*cur;
+	struct xfs_owner_info		oinfo;
+
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INOBT);
+	cur = which == XFS_BTNUM_INO ? sc->sa.ino_cur : sc->sa.fino_cur;
+	return xfs_scrub_btree(sc, cur, xfs_scrub_iallocbt_helper,
+			&oinfo, NULL);
+}
+
+int
+xfs_scrub_inobt(
+	struct xfs_scrub_context	*sc)
+{
+	return xfs_scrub_iallocbt(sc, XFS_BTNUM_INO);
+}
+
+int
+xfs_scrub_finobt(
+	struct xfs_scrub_context	*sc)
+{
+	return xfs_scrub_iallocbt(sc, XFS_BTNUM_FINO);
+}
diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
index 70ca4f6..0ea0403 100644
--- a/fs/xfs/xfs_icache.c
+++ b/fs/xfs/xfs_icache.c
@@ -583,6 +583,8 @@ xfs_iget(
 		error = xfs_iget_cache_hit(pag, ip, ino, flags, lock_flags);
 		if (error)
 			goto out_error_or_again;
+	} else if (flags & XFS_IGET_HITONLY) {
+		rcu_read_unlock();
 	} else {
 		rcu_read_unlock();
 		XFS_STATS_INC(mp, xs_ig_missed);
@@ -595,6 +597,8 @@ xfs_iget(
 	xfs_perag_put(pag);
 
 	*ipp = ip;
+	if (!ip)
+		return 0;
 
 	/*
 	 * If we have a real type for an on-disk inode, we can setup the inode
diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h
index a1e02f4..e926f38 100644
--- a/fs/xfs/xfs_icache.h
+++ b/fs/xfs/xfs_icache.h
@@ -45,9 +45,10 @@ struct xfs_eofblocks {
 /*
  * Flags for xfs_iget()
  */
-#define XFS_IGET_CREATE		0x1
-#define XFS_IGET_UNTRUSTED	0x2
-#define XFS_IGET_DONTCACHE	0x4
+#define XFS_IGET_CREATE		0x1	/* initialize free inodes */
+#define XFS_IGET_UNTRUSTED	0x2	/* check the inode number */
+#define XFS_IGET_DONTCACHE	0x4	/* don't keep the inode cached */
+#define XFS_IGET_HITONLY	0x8	/* only return cached inodes */
 
 int xfs_iget(struct xfs_mount *mp, struct xfs_trans *tp, xfs_ino_t ino,
 	     uint flags, uint lock_flags, xfs_inode_t **ipp);
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 5cc7fed..ab50df5 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3359,7 +3359,9 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_AGFL,		"AGFL" }, \
 	{ XFS_SCRUB_TYPE_AGI,		"AGI" }, \
 	{ XFS_SCRUB_TYPE_BNOBT,		"bnobt" }, \
-	{ XFS_SCRUB_TYPE_CNTBT,		"cntbt" }
+	{ XFS_SCRUB_TYPE_CNTBT,		"cntbt" }, \
+	{ XFS_SCRUB_TYPE_INOBT,		"inobt" }, \
+	{ XFS_SCRUB_TYPE_FINOBT,	"finobt" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 22/55] xfs: support scrubbing rmap btree

[Darrick J. Wong]

Check the reverse mapping records to make sure that the contents
make sense.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile                |    1 
 fs/xfs/libxfs/xfs_fs.h         |    3 +
 fs/xfs/libxfs/xfs_rmap.c       |    3 +
 fs/xfs/libxfs/xfs_rmap.h       |    3 +
 fs/xfs/libxfs/xfs_rmap_btree.c |    4 -
 fs/xfs/scrub/common.c          |    1 
 fs/xfs/scrub/common.h          |    1 
 fs/xfs/scrub/rmap.c            |  116 ++++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_trace.h             |    3 +
 9 files changed, 128 insertions(+), 7 deletions(-)
 create mode 100644 fs/xfs/scrub/rmap.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 0c98a98..e8875c0 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -109,6 +109,7 @@ xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   btree.o \
 				   common.o \
 				   ialloc.o \
+				   rmap.o \
 				   )
 
 # low-level transaction/log code
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 5fc45f5..324ce66 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -583,7 +583,8 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_CNTBT	6	/* freesp by length btree */
 #define XFS_SCRUB_TYPE_INOBT	7	/* inode btree */
 #define XFS_SCRUB_TYPE_FINOBT	8	/* free inode btree */
-#define XFS_SCRUB_TYPE_MAX	8
+#define XFS_SCRUB_TYPE_RMAPBT	9	/* reverse mapping btree */
+#define XFS_SCRUB_TYPE_MAX	9
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/libxfs/xfs_rmap.c b/fs/xfs/libxfs/xfs_rmap.c
index 3840556..c7d5102 100644
--- a/fs/xfs/libxfs/xfs_rmap.c
+++ b/fs/xfs/libxfs/xfs_rmap.c
@@ -179,7 +179,8 @@ xfs_rmap_delete(
 	return error;
 }
 
-static int
+/* Convert an internal btree record to an rmap record. */
+int
 xfs_rmap_btrec_to_irec(
 	union xfs_btree_rec	*rec,
 	struct xfs_rmap_irec	*irec)
diff --git a/fs/xfs/libxfs/xfs_rmap.h b/fs/xfs/libxfs/xfs_rmap.h
index faf2c1a..3fa4559 100644
--- a/fs/xfs/libxfs/xfs_rmap.h
+++ b/fs/xfs/libxfs/xfs_rmap.h
@@ -214,5 +214,8 @@ int xfs_rmap_find_left_neighbor(struct xfs_btree_cur *cur, xfs_agblock_t bno,
 int xfs_rmap_lookup_le_range(struct xfs_btree_cur *cur, xfs_agblock_t bno,
 		uint64_t owner, uint64_t offset, unsigned int flags,
 		struct xfs_rmap_irec *irec, int	*stat);
+union xfs_btree_rec;
+int xfs_rmap_btrec_to_irec(union xfs_btree_rec *rec,
+		struct xfs_rmap_irec *irec);
 
 #endif	/* __XFS_RMAP_H__ */
diff --git a/fs/xfs/libxfs/xfs_rmap_btree.c b/fs/xfs/libxfs/xfs_rmap_btree.c
index 74e5a54..b342cc8 100644
--- a/fs/xfs/libxfs/xfs_rmap_btree.c
+++ b/fs/xfs/libxfs/xfs_rmap_btree.c
@@ -377,7 +377,6 @@ const struct xfs_buf_ops xfs_rmapbt_buf_ops = {
 	.verify_write		= xfs_rmapbt_write_verify,
 };
 
-#if defined(DEBUG) || defined(XFS_WARN)
 STATIC int
 xfs_rmapbt_keys_inorder(
 	struct xfs_btree_cur	*cur,
@@ -437,7 +436,6 @@ xfs_rmapbt_recs_inorder(
 		return 1;
 	return 0;
 }
-#endif	/* DEBUG */
 
 static const struct xfs_btree_ops xfs_rmapbt_ops = {
 	.rec_len		= sizeof(struct xfs_rmap_rec),
@@ -456,10 +454,8 @@ static const struct xfs_btree_ops xfs_rmapbt_ops = {
 	.key_diff		= xfs_rmapbt_key_diff,
 	.buf_ops		= &xfs_rmapbt_buf_ops,
 	.diff_two_keys		= xfs_rmapbt_diff_two_keys,
-#if defined(DEBUG) || defined(XFS_WARN)
 	.keys_inorder		= xfs_rmapbt_keys_inorder,
 	.recs_inorder		= xfs_rmapbt_recs_inorder,
-#endif
 };
 
 /*
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index e57d775..15c76c5 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -668,6 +668,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag_header, xfs_scrub_cntbt, NULL, NULL},
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_inobt, NULL, NULL},
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_finobt, NULL, xfs_sb_version_hasfinobt},
+	{xfs_scrub_setup_ag_header, xfs_scrub_rmapbt, NULL, xfs_sb_version_hasrmapbt},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 69177c8..7f160e4 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -224,5 +224,6 @@ int xfs_scrub_bnobt(struct xfs_scrub_context *sc);
 int xfs_scrub_cntbt(struct xfs_scrub_context *sc);
 int xfs_scrub_inobt(struct xfs_scrub_context *sc);
 int xfs_scrub_finobt(struct xfs_scrub_context *sc);
+int xfs_scrub_rmapbt(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/rmap.c b/fs/xfs/scrub/rmap.c
new file mode 100644
index 0000000..bb94260
--- /dev/null
+++ b/fs/xfs/scrub/rmap.c
@@ -0,0 +1,116 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_rmap.h"
+#include "scrub/common.h"
+#include "scrub/btree.h"
+
+/* Reverse-mapping scrubber. */
+
+/* Scrub an rmapbt record. */
+STATIC int
+xfs_scrub_rmapbt_helper(
+	struct xfs_scrub_btree		*bs,
+	union xfs_btree_rec		*rec)
+{
+	struct xfs_mount		*mp = bs->cur->bc_mp;
+	struct xfs_agf			*agf;
+	struct xfs_rmap_irec		irec;
+	xfs_agblock_t			eoag;
+	bool				non_inode;
+	bool				is_unwritten;
+	bool				is_bmbt;
+	bool				is_attr;
+	int				error;
+
+	error = xfs_rmap_btrec_to_irec(rec, &irec);
+	XFS_SCRUB_BTREC_OP_ERROR_GOTO(bs, &error, out);
+
+	/* Check extent. */
+	agf = XFS_BUF_TO_AGF(bs->sc->sa.agf_bp);
+	eoag = be32_to_cpu(agf->agf_length);
+	XFS_SCRUB_BTREC_CHECK(bs, irec.rm_startblock < mp->m_sb.sb_agblocks);
+	XFS_SCRUB_BTREC_CHECK(bs, irec.rm_startblock < eoag);
+	XFS_SCRUB_BTREC_CHECK(bs, irec.rm_startblock < irec.rm_startblock +
+			irec.rm_blockcount);
+	XFS_SCRUB_BTREC_CHECK(bs, irec.rm_startblock + irec.rm_blockcount <=
+			mp->m_sb.sb_agblocks);
+	XFS_SCRUB_BTREC_CHECK(bs, irec.rm_startblock + irec.rm_blockcount <=
+			eoag);
+
+	/* Check flags. */
+	non_inode = XFS_RMAP_NON_INODE_OWNER(irec.rm_owner);
+	is_bmbt = irec.rm_flags & XFS_RMAP_BMBT_BLOCK;
+	is_attr = irec.rm_flags & XFS_RMAP_ATTR_FORK;
+	is_unwritten = irec.rm_flags & XFS_RMAP_UNWRITTEN;
+
+	XFS_SCRUB_BTREC_CHECK(bs, !is_bmbt || irec.rm_offset == 0);
+	XFS_SCRUB_BTREC_CHECK(bs, !non_inode || irec.rm_offset == 0);
+	XFS_SCRUB_BTREC_CHECK(bs, !is_unwritten || !(is_bmbt || non_inode ||
+			is_attr));
+	XFS_SCRUB_BTREC_CHECK(bs, !non_inode || !(is_bmbt || is_unwritten ||
+			is_attr));
+
+	/* Owner inode within an AG? */
+	XFS_SCRUB_BTREC_CHECK(bs, non_inode ||
+			(XFS_INO_TO_AGNO(mp, irec.rm_owner) <
+							mp->m_sb.sb_agcount &&
+			 XFS_AGINO_TO_AGBNO(mp,
+				XFS_INO_TO_AGINO(mp, irec.rm_owner)) <
+							mp->m_sb.sb_agblocks));
+	/* Owner inode within the FS? */
+	XFS_SCRUB_BTREC_CHECK(bs, non_inode ||
+			XFS_AGB_TO_DADDR(mp,
+				XFS_INO_TO_AGNO(mp, irec.rm_owner),
+				XFS_AGINO_TO_AGBNO(mp,
+					XFS_INO_TO_AGINO(mp, irec.rm_owner))) <
+			XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks));
+
+	/* Non-inode owner within the magic values? */
+	XFS_SCRUB_BTREC_CHECK(bs, !non_inode ||
+			(irec.rm_owner > XFS_RMAP_OWN_MIN &&
+			 irec.rm_owner <= XFS_RMAP_OWN_FS));
+out:
+	return error;
+}
+
+/* Scrub the rmap btree for some AG. */
+int
+xfs_scrub_rmapbt(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_owner_info		oinfo;
+
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_AG);
+	return xfs_scrub_btree(sc, sc->sa.rmap_cur, xfs_scrub_rmapbt_helper,
+			&oinfo, NULL);
+}
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index ab50df5..e752f68 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3361,7 +3361,8 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_BNOBT,		"bnobt" }, \
 	{ XFS_SCRUB_TYPE_CNTBT,		"cntbt" }, \
 	{ XFS_SCRUB_TYPE_INOBT,		"inobt" }, \
-	{ XFS_SCRUB_TYPE_FINOBT,	"finobt" }
+	{ XFS_SCRUB_TYPE_FINOBT,	"finobt" }, \
+	{ XFS_SCRUB_TYPE_RMAPBT,	"rmapbt" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 23/55] xfs: support scrubbing refcount btree

[Darrick J. Wong]

Plumb in the pieces necessary to check the refcount btree.  If rmap is
available, check the reference count by performing an interval query
against the rmapbt.

v2: Handle the case where the rmap records are not all at least the
length of the refcount extent.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile                    |    1 
 fs/xfs/libxfs/xfs_fs.h             |    3 +
 fs/xfs/libxfs/xfs_refcount_btree.c |    4 --
 fs/xfs/scrub/common.c              |    1 
 fs/xfs/scrub/common.h              |    1 
 fs/xfs/scrub/refcount.c            |   85 ++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_trace.h                 |    3 +
 7 files changed, 92 insertions(+), 6 deletions(-)
 create mode 100644 fs/xfs/scrub/refcount.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index e8875c0..252d8c6 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -109,6 +109,7 @@ xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   btree.o \
 				   common.o \
 				   ialloc.o \
+				   refcount.o \
 				   rmap.o \
 				   )
 
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 324ce66..8dc3e7b 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -584,7 +584,8 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_INOBT	7	/* inode btree */
 #define XFS_SCRUB_TYPE_FINOBT	8	/* free inode btree */
 #define XFS_SCRUB_TYPE_RMAPBT	9	/* reverse mapping btree */
-#define XFS_SCRUB_TYPE_MAX	9
+#define XFS_SCRUB_TYPE_REFCNTBT	10	/* reference count btree */
+#define XFS_SCRUB_TYPE_MAX	10
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/libxfs/xfs_refcount_btree.c b/fs/xfs/libxfs/xfs_refcount_btree.c
index 50add52..cae24bd 100644
--- a/fs/xfs/libxfs/xfs_refcount_btree.c
+++ b/fs/xfs/libxfs/xfs_refcount_btree.c
@@ -285,7 +285,6 @@ const struct xfs_buf_ops xfs_refcountbt_buf_ops = {
 	.verify_write		= xfs_refcountbt_write_verify,
 };
 
-#if defined(DEBUG) || defined(XFS_WARN)
 STATIC int
 xfs_refcountbt_keys_inorder(
 	struct xfs_btree_cur	*cur,
@@ -306,7 +305,6 @@ xfs_refcountbt_recs_inorder(
 		be32_to_cpu(r1->refc.rc_blockcount) <=
 		be32_to_cpu(r2->refc.rc_startblock);
 }
-#endif
 
 static const struct xfs_btree_ops xfs_refcountbt_ops = {
 	.rec_len		= sizeof(struct xfs_refcount_rec),
@@ -325,10 +323,8 @@ static const struct xfs_btree_ops xfs_refcountbt_ops = {
 	.key_diff		= xfs_refcountbt_key_diff,
 	.buf_ops		= &xfs_refcountbt_buf_ops,
 	.diff_two_keys		= xfs_refcountbt_diff_two_keys,
-#if defined(DEBUG) || defined(XFS_WARN)
 	.keys_inorder		= xfs_refcountbt_keys_inorder,
 	.recs_inorder		= xfs_refcountbt_recs_inorder,
-#endif
 };
 
 /*
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 15c76c5..4438482 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -669,6 +669,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_inobt, NULL, NULL},
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_finobt, NULL, xfs_sb_version_hasfinobt},
 	{xfs_scrub_setup_ag_header, xfs_scrub_rmapbt, NULL, xfs_sb_version_hasrmapbt},
+	{xfs_scrub_setup_ag_header, xfs_scrub_refcountbt, NULL, xfs_sb_version_hasreflink},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 7f160e4..7dea98a 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -225,5 +225,6 @@ int xfs_scrub_cntbt(struct xfs_scrub_context *sc);
 int xfs_scrub_inobt(struct xfs_scrub_context *sc);
 int xfs_scrub_finobt(struct xfs_scrub_context *sc);
 int xfs_scrub_rmapbt(struct xfs_scrub_context *sc);
+int xfs_scrub_refcountbt(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/refcount.c b/fs/xfs/scrub/refcount.c
new file mode 100644
index 0000000..f7ceba9
--- /dev/null
+++ b/fs/xfs/scrub/refcount.c
@@ -0,0 +1,85 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_rmap.h"
+#include "scrub/common.h"
+#include "scrub/btree.h"
+
+/* Reference count btree scrubber. */
+
+/* Scrub a refcountbt record. */
+STATIC int
+xfs_scrub_refcountbt_helper(
+	struct xfs_scrub_btree		*bs,
+	union xfs_btree_rec		*rec)
+{
+	struct xfs_mount		*mp = bs->cur->bc_mp;
+	struct xfs_agf			*agf;
+	struct xfs_refcount_irec	irec;
+	xfs_agblock_t			eoag;
+	bool				has_cowflag;
+	int				error = 0;
+
+	irec.rc_startblock = be32_to_cpu(rec->refc.rc_startblock);
+	irec.rc_blockcount = be32_to_cpu(rec->refc.rc_blockcount);
+	irec.rc_refcount = be32_to_cpu(rec->refc.rc_refcount);
+	agf = XFS_BUF_TO_AGF(bs->sc->sa.agf_bp);
+	eoag = be32_to_cpu(agf->agf_length);
+
+	has_cowflag = !!(irec.rc_startblock & XFS_REFC_COW_START);
+	XFS_SCRUB_BTREC_CHECK(bs, (irec.rc_refcount == 1 && has_cowflag) ||
+				  (irec.rc_refcount != 1 && !has_cowflag));
+	irec.rc_startblock &= ~XFS_REFC_COW_START;
+	XFS_SCRUB_BTREC_CHECK(bs, irec.rc_startblock < mp->m_sb.sb_agblocks);
+	XFS_SCRUB_BTREC_CHECK(bs, irec.rc_startblock < eoag);
+	XFS_SCRUB_BTREC_CHECK(bs, irec.rc_startblock < irec.rc_startblock +
+			irec.rc_blockcount);
+	XFS_SCRUB_BTREC_CHECK(bs, (unsigned long long)irec.rc_startblock +
+			irec.rc_blockcount <= mp->m_sb.sb_agblocks);
+	XFS_SCRUB_BTREC_CHECK(bs, (unsigned long long)irec.rc_startblock +
+			irec.rc_blockcount <= eoag);
+	XFS_SCRUB_BTREC_CHECK(bs, irec.rc_refcount >= 1);
+
+	return error;
+}
+
+/* Scrub the refcount btree for some AG. */
+int
+xfs_scrub_refcountbt(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_owner_info		oinfo;
+
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_REFC);
+	return xfs_scrub_btree(sc, sc->sa.refc_cur, xfs_scrub_refcountbt_helper,
+			&oinfo, NULL);
+}
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index e752f68..4757fea 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3362,7 +3362,8 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_CNTBT,		"cntbt" }, \
 	{ XFS_SCRUB_TYPE_INOBT,		"inobt" }, \
 	{ XFS_SCRUB_TYPE_FINOBT,	"finobt" }, \
-	{ XFS_SCRUB_TYPE_RMAPBT,	"rmapbt" }
+	{ XFS_SCRUB_TYPE_RMAPBT,	"rmapbt" }, \
+	{ XFS_SCRUB_TYPE_REFCNTBT,	"refcountbt" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 24/55] xfs: scrub inodes

[Darrick J. Wong]

Scrub the fields within an inode.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile               |    1 
 fs/xfs/libxfs/xfs_fs.h        |    3 
 fs/xfs/libxfs/xfs_inode_buf.c |    2 
 fs/xfs/libxfs/xfs_inode_buf.h |    3 
 fs/xfs/scrub/common.c         |   14 +-
 fs/xfs/scrub/common.h         |   12 +
 fs/xfs/scrub/inode.c          |  379 +++++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_itable.c           |    2 
 fs/xfs/xfs_itable.h           |    5 +
 fs/xfs/xfs_trace.h            |    3 
 10 files changed, 417 insertions(+), 7 deletions(-)
 create mode 100644 fs/xfs/scrub/inode.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 252d8c6..83fc8d3 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -109,6 +109,7 @@ xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   btree.o \
 				   common.o \
 				   ialloc.o \
+				   inode.o \
 				   refcount.o \
 				   rmap.o \
 				   )
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 8dc3e7b..5cafbf6 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -585,7 +585,8 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_FINOBT	8	/* free inode btree */
 #define XFS_SCRUB_TYPE_RMAPBT	9	/* reverse mapping btree */
 #define XFS_SCRUB_TYPE_REFCNTBT	10	/* reference count btree */
-#define XFS_SCRUB_TYPE_MAX	10
+#define XFS_SCRUB_TYPE_INODE	11	/* inode record */
+#define XFS_SCRUB_TYPE_MAX	11
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/libxfs/xfs_inode_buf.c b/fs/xfs/libxfs/xfs_inode_buf.c
index d93f9d9..3752bac 100644
--- a/fs/xfs/libxfs/xfs_inode_buf.c
+++ b/fs/xfs/libxfs/xfs_inode_buf.c
@@ -381,7 +381,7 @@ xfs_log_dinode_to_disk(
 	}
 }
 
-static bool
+bool
 xfs_dinode_verify(
 	struct xfs_mount	*mp,
 	xfs_ino_t		ino,
diff --git a/fs/xfs/libxfs/xfs_inode_buf.h b/fs/xfs/libxfs/xfs_inode_buf.h
index 6848a0a..988fd67 100644
--- a/fs/xfs/libxfs/xfs_inode_buf.h
+++ b/fs/xfs/libxfs/xfs_inode_buf.h
@@ -82,4 +82,7 @@ void	xfs_inobp_check(struct xfs_mount *, struct xfs_buf *);
 #define	xfs_inobp_check(mp, bp)
 #endif /* DEBUG */
 
+bool	xfs_dinode_verify(struct xfs_mount *mp, xfs_ino_t ino,
+			  struct xfs_dinode *dip);
+
 #endif	/* __XFS_INODE_BUF_H__ */
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 4438482..a5fecae 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -559,6 +559,7 @@ xfs_scrub_dummy(
 int
 xfs_scrub_teardown(
 	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip_in,
 	int				error)
 {
 	xfs_scrub_ag_free(&sc->sa);
@@ -567,6 +568,14 @@ xfs_scrub_teardown(
 	sc->ag_lock.agmask = NULL;
 	xfs_trans_cancel(sc->tp);
 	sc->tp = NULL;
+	if (sc->ip != NULL) {
+		xfs_iunlock(sc->ip, XFS_ILOCK_EXCL);
+		xfs_iunlock(sc->ip, XFS_IOLOCK_EXCL);
+		xfs_iunlock(sc->ip, XFS_MMAPLOCK_EXCL);
+		if (sc->ip != ip_in)
+			IRELE(sc->ip);
+		sc->ip = NULL;
+	}
 	return error;
 }
 
@@ -670,6 +679,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_finobt, NULL, xfs_sb_version_hasfinobt},
 	{xfs_scrub_setup_ag_header, xfs_scrub_rmapbt, NULL, xfs_sb_version_hasrmapbt},
 	{xfs_scrub_setup_ag_header, xfs_scrub_refcountbt, NULL, xfs_sb_version_hasreflink},
+	{xfs_scrub_setup_inode_raw, xfs_scrub_inode, NULL, NULL},
 };
 
 /* Dispatch metadata scrubbing. */
@@ -730,7 +740,7 @@ xfs_scrub_metadata(
 	error = fns->scrub(&sc);
 	if (!deadlocked && error == -EDEADLOCK) {
 		deadlocked = true;
-		error = xfs_scrub_teardown(&sc, error);
+		error = xfs_scrub_teardown(&sc, ip, error);
 		if (error != -EDEADLOCK)
 			goto out;
 		goto retry_op;
@@ -741,7 +751,7 @@ xfs_scrub_metadata(
 		xfs_alert_ratelimited(mp, "Corruption detected during scrub.");
 
 out_teardown:
-	error = xfs_scrub_teardown(&sc, error);
+	error = xfs_scrub_teardown(&sc, ip, error);
 out:
 	trace_xfs_scrub_done(ip, sm->sm_type, sm->sm_agno, sm->sm_ino,
 			sm->sm_gen, sm->sm_flags, error);
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 7dea98a..ae4bee5 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -200,7 +200,8 @@ int xfs_scrub_walk_agfl(struct xfs_scrub_context *sc,
 
 /* Setup functions */
 
-int xfs_scrub_teardown(struct xfs_scrub_context *sc, int error);
+int xfs_scrub_teardown(struct xfs_scrub_context *sc, struct xfs_inode *ip_in,
+		       int error);
 int xfs_scrub_setup(struct xfs_scrub_context *sc, struct xfs_inode *ip,
 		    struct xfs_scrub_metadata *sm, bool retry_deadlocked);
 int xfs_scrub_setup_ag(struct xfs_scrub_context *sc, struct xfs_inode *ip,
@@ -213,6 +214,14 @@ int xfs_scrub_setup_ag_iallocbt(struct xfs_scrub_context *sc,
 				struct xfs_inode *ip,
 				struct xfs_scrub_metadata *sm,
 				bool retry_deadlocked);
+int xfs_scrub_setup_inode(struct xfs_scrub_context *sc,
+			  struct xfs_inode *ip,
+			  struct xfs_scrub_metadata *sm,
+			  bool retry_deadlocked);
+int xfs_scrub_setup_inode_raw(struct xfs_scrub_context *sc,
+			      struct xfs_inode *ip,
+			      struct xfs_scrub_metadata *sm,
+			      bool retry_deadlocked);
 
 /* Metadata scrubbers */
 
@@ -226,5 +235,6 @@ int xfs_scrub_inobt(struct xfs_scrub_context *sc);
 int xfs_scrub_finobt(struct xfs_scrub_context *sc);
 int xfs_scrub_rmapbt(struct xfs_scrub_context *sc);
 int xfs_scrub_refcountbt(struct xfs_scrub_context *sc);
+int xfs_scrub_inode(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/inode.c b/fs/xfs/scrub/inode.c
new file mode 100644
index 0000000..b9e7060
--- /dev/null
+++ b/fs/xfs/scrub/inode.c
@@ -0,0 +1,379 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_icache.h"
+#include "xfs_itable.h"
+#include "xfs_inode_buf.h"
+#include "xfs_inode_fork.h"
+#include "xfs_ialloc.h"
+#include "xfs_log.h"
+#include "xfs_trans_priv.h"
+#include "scrub/common.h"
+
+/*
+ * Given an inode and the scrub control structure, return either the
+ * inode referenced in the control structure or the inode passed in.
+ * The inode is not locked.
+ */
+STATIC struct xfs_inode *
+xfs_scrub_get_inode(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+	struct xfs_inode		*ips = NULL;
+	int				error;
+
+	if (sc->sm->sm_gen && !sc->sm->sm_ino)
+		return ERR_PTR(-EINVAL);
+
+	if (sc->sm->sm_ino && sc->sm->sm_ino != ip->i_ino) {
+		if (xfs_internal_inum(mp, sc->sm->sm_ino))
+			return ERR_PTR(-ENOENT);
+		error = xfs_iget(mp, NULL, sc->sm->sm_ino, XFS_IGET_UNTRUSTED,
+				0, &ips);
+		if (error) {
+			trace_xfs_scrub_op_error(mp,
+					XFS_INO_TO_AGNO(mp, sc->sm->sm_ino),
+					XFS_INO_TO_AGBNO(mp, sc->sm->sm_ino),
+					"inode", error, __func__, __LINE__);
+			goto out_err;
+		}
+		if (VFS_I(ips)->i_generation != sc->sm->sm_gen) {
+			IRELE(ips);
+			return ERR_PTR(-ENOENT);
+		}
+
+		return ips;
+	}
+
+	return ip;
+out_err:
+	return ERR_PTR(error);
+}
+
+/* Set us up with an inode. */
+int
+xfs_scrub_setup_inode(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+	int				error;
+
+	memset(sc, 0, sizeof(*sc));
+	sc->sm = sm;
+	sc->ip = xfs_scrub_get_inode(sc, ip);
+	if (IS_ERR(sc->ip))
+		return PTR_ERR(sc->ip);
+	else if (sc->ip == NULL)
+		return -ENOENT;
+
+	xfs_ilock(sc->ip, XFS_IOLOCK_EXCL);
+	xfs_ilock(sc->ip, XFS_MMAPLOCK_EXCL);
+	error = xfs_scrub_trans_alloc(sm, mp, &M_RES(mp)->tr_itruncate,
+			0, 0, 0, &sc->tp);
+	if (error)
+		goto out_unlock;
+	xfs_ilock(sc->ip, XFS_ILOCK_EXCL);
+
+	xfs_scrub_ag_lock_init(mp, &sc->ag_lock);
+	return error;
+out_unlock:
+	xfs_iunlock(sc->ip, XFS_IOLOCK_EXCL);
+	xfs_iunlock(sc->ip, XFS_MMAPLOCK_EXCL);
+	if (sc->ip != ip)
+		IRELE(sc->ip);
+	return error;
+}
+
+/* Try to get the in-core inode.  If we can't, we'll just have to do it raw. */
+int
+xfs_scrub_setup_inode_raw(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+	int				error;
+
+	if (sm->sm_ino && xfs_internal_inum(mp, sm->sm_ino))
+		return -ENOENT;
+
+	error = xfs_scrub_setup_inode(sc, ip, sm, retry_deadlocked);
+	if (error) {
+		memset(sc, 0, sizeof(*sc));
+		sc->ip = NULL;
+		sc->sm = sm;
+
+		/* Push everything out of the log onto disk prior to check. */
+		error = _xfs_log_force(mp, XFS_LOG_SYNC, NULL);
+		if (error)
+			return error;
+		xfs_ail_push_all_sync(mp->m_ail);
+
+		return xfs_scrub_trans_alloc(sm, mp,
+				&M_RES(mp)->tr_itruncate, 0, 0, 0, &sc->tp);
+	}
+	return 0;
+}
+
+/* Inode core */
+
+#define XFS_SCRUB_INODE_CHECK(fs_ok) \
+	XFS_SCRUB_INO_CHECK(sc, ino, bp, "inode", fs_ok);
+#define XFS_SCRUB_INODE_GOTO(fs_ok, label) \
+	XFS_SCRUB_INO_GOTO(sc, ino, bp, "inode", fs_ok, label);
+#define XFS_SCRUB_INODE_OP_ERROR_GOTO(label) \
+	XFS_SCRUB_OP_ERROR_GOTO(sc, XFS_INO_TO_AGNO(mp, ino), \
+			XFS_INO_TO_AGBNO(mp, ino), "inode", &error, label);
+#define XFS_SCRUB_INODE_PREEN(fs_ok) \
+	XFS_SCRUB_INO_PREEN(sc, bp, "inode", fs_ok);
+/* Scrub an inode. */
+int
+xfs_scrub_inode(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_imap			imap;
+	struct xfs_dinode		di;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_ifork		*ifp;
+	struct xfs_buf			*bp = NULL;
+	struct xfs_dinode		*dip;
+	xfs_ino_t			ino;
+	unsigned long long		isize;
+	uint64_t			flags2;
+	uint32_t			nextents;
+	uint32_t			extsize;
+	uint32_t			cowextsize;
+	uint16_t			flags;
+	uint16_t			mode;
+	int				error = 0;
+
+	/* Did we get the in-core inode, or are we doing this manually? */
+	if (sc->ip) {
+		ino = sc->ip->i_ino;
+		xfs_inode_to_disk(sc->ip, &di, 0);
+		dip = &di;
+	} else {
+		/* Map & read inode. */
+		ino = sc->sm->sm_ino;
+		error = xfs_imap(mp, sc->tp, ino, &imap, XFS_IGET_UNTRUSTED);
+		XFS_SCRUB_INODE_OP_ERROR_GOTO(out);
+
+		error = xfs_trans_read_buf(mp, sc->tp, mp->m_ddev_targp,
+				imap.im_blkno, imap.im_len, XBF_UNMAPPED, &bp,
+				NULL);
+		XFS_SCRUB_INODE_OP_ERROR_GOTO(out);
+
+		/* Is this really the inode we want? */
+		bp->b_ops = &xfs_inode_buf_ops;
+		dip = xfs_buf_offset(bp, imap.im_boffset);
+		error = xfs_dinode_verify(mp, ino, dip) ? 0 : -EFSCORRUPTED;
+		XFS_SCRUB_INODE_OP_ERROR_GOTO(out);
+		XFS_SCRUB_INODE_GOTO(
+				xfs_dinode_good_version(mp, dip->di_version),
+				out);
+		if (be32_to_cpu(dip->di_gen) != sc->sm->sm_gen) {
+			error = -EINVAL;
+			goto out;
+		}
+	}
+
+	flags = be16_to_cpu(dip->di_flags);
+	if (dip->di_version >= 3)
+		flags2 = be64_to_cpu(dip->di_flags2);
+	else
+		flags2 = 0;
+
+	/* di_mode */
+	mode = be16_to_cpu(dip->di_mode);
+	XFS_SCRUB_INODE_CHECK(!(mode & ~(S_IALLUGO | S_IFMT)));
+
+	/* v1/v2 fields */
+	switch (dip->di_version) {
+	case 1:
+		XFS_SCRUB_INODE_CHECK(dip->di_nlink == 0);
+		XFS_SCRUB_INODE_CHECK(dip->di_mode || !sc->ip);
+		XFS_SCRUB_INODE_CHECK(dip->di_projid_lo == 0);
+		XFS_SCRUB_INODE_CHECK(dip->di_projid_hi == 0);
+		break;
+	case 2:
+	case 3:
+		XFS_SCRUB_INODE_CHECK(dip->di_onlink == 0);
+		XFS_SCRUB_INODE_CHECK(dip->di_mode || !sc->ip);
+		XFS_SCRUB_INODE_CHECK(dip->di_projid_hi == 0 ||
+				xfs_sb_version_hasprojid32bit(&mp->m_sb));
+		break;
+	default:
+		ASSERT(0);
+		break;
+	}
+
+	/* di_format */
+	switch (dip->di_format) {
+	case XFS_DINODE_FMT_DEV:
+		XFS_SCRUB_INODE_CHECK(S_ISCHR(mode) || S_ISBLK(mode) ||
+				      S_ISFIFO(mode) || S_ISSOCK(mode));
+		break;
+	case XFS_DINODE_FMT_LOCAL:
+		XFS_SCRUB_INODE_CHECK(S_ISDIR(mode) || S_ISLNK(mode));
+		break;
+	case XFS_DINODE_FMT_EXTENTS:
+		XFS_SCRUB_INODE_CHECK(S_ISREG(mode) || S_ISDIR(mode) ||
+				      S_ISLNK(mode));
+		break;
+	case XFS_DINODE_FMT_BTREE:
+		XFS_SCRUB_INODE_CHECK(S_ISREG(mode) || S_ISDIR(mode));
+		break;
+	case XFS_DINODE_FMT_UUID:
+	default:
+		XFS_SCRUB_INODE_CHECK(false);
+		break;
+	}
+
+	/* di_size */
+	isize = be64_to_cpu(dip->di_size);
+	XFS_SCRUB_INODE_CHECK(!(isize & (1ULL << 63)));
+	if (!S_ISDIR(mode) && !S_ISREG(mode) && !S_ISLNK(mode))
+		XFS_SCRUB_INODE_CHECK(isize == 0);
+
+	/* di_nblocks */
+	if (flags2 & XFS_DIFLAG2_REFLINK) {
+		; /* nblocks can exceed dblocks */
+	} else if (flags & XFS_DIFLAG_REALTIME) {
+		XFS_SCRUB_INODE_CHECK(be64_to_cpu(dip->di_nblocks) <
+				mp->m_sb.sb_dblocks + mp->m_sb.sb_rblocks);
+	} else {
+		XFS_SCRUB_INODE_CHECK(be64_to_cpu(dip->di_nblocks) <
+				mp->m_sb.sb_dblocks);
+	}
+
+	/* di_extsize */
+	if (flags & XFS_DIFLAG_EXTSIZE) {
+		extsize = be32_to_cpu(dip->di_extsize);
+		XFS_SCRUB_INODE_CHECK(extsize > 0);
+		XFS_SCRUB_INODE_CHECK(extsize <= MAXEXTLEN);
+		XFS_SCRUB_INODE_CHECK(extsize <= mp->m_sb.sb_agblocks / 2 ||
+				(flags & XFS_DIFLAG_REALTIME));
+	}
+
+	/* di_flags */
+	XFS_SCRUB_INODE_CHECK(!(flags & XFS_DIFLAG_IMMUTABLE) ||
+			      !(flags & XFS_DIFLAG_APPEND));
+
+	XFS_SCRUB_INODE_CHECK(!(flags & XFS_DIFLAG_FILESTREAM) ||
+			      !(flags & XFS_DIFLAG_REALTIME));
+
+	/* di_nextents */
+	nextents = be32_to_cpu(dip->di_nextents);
+	switch (dip->di_format) {
+	case XFS_DINODE_FMT_EXTENTS:
+		XFS_SCRUB_INODE_CHECK(nextents <=
+			XFS_DFORK_DSIZE(dip, mp) / sizeof(struct xfs_bmbt_rec));
+		break;
+	case XFS_DINODE_FMT_BTREE:
+		XFS_SCRUB_INODE_CHECK(nextents >
+			XFS_DFORK_DSIZE(dip, mp) / sizeof(struct xfs_bmbt_rec));
+		break;
+	case XFS_DINODE_FMT_LOCAL:
+	case XFS_DINODE_FMT_DEV:
+	case XFS_DINODE_FMT_UUID:
+	default:
+		XFS_SCRUB_INODE_CHECK(nextents == 0);
+		break;
+	}
+
+	/* di_anextents */
+	nextents = be16_to_cpu(dip->di_anextents);
+	switch (dip->di_aformat) {
+	case XFS_DINODE_FMT_EXTENTS:
+		XFS_SCRUB_INODE_CHECK(nextents <=
+			XFS_DFORK_ASIZE(dip, mp) / sizeof(struct xfs_bmbt_rec));
+		break;
+	case XFS_DINODE_FMT_BTREE:
+		XFS_SCRUB_INODE_CHECK(nextents >
+			XFS_DFORK_ASIZE(dip, mp) / sizeof(struct xfs_bmbt_rec));
+		break;
+	case XFS_DINODE_FMT_LOCAL:
+	case XFS_DINODE_FMT_DEV:
+	case XFS_DINODE_FMT_UUID:
+	default:
+		XFS_SCRUB_INODE_CHECK(nextents == 0);
+		break;
+	}
+
+	/* di_forkoff */
+	XFS_SCRUB_INODE_CHECK(XFS_DFORK_APTR(dip) <
+			(char *)dip + mp->m_sb.sb_inodesize);
+	XFS_SCRUB_INODE_CHECK(dip->di_anextents == 0 || dip->di_forkoff);
+
+	/* di_aformat */
+	XFS_SCRUB_INODE_CHECK(dip->di_aformat == XFS_DINODE_FMT_LOCAL ||
+			      dip->di_aformat == XFS_DINODE_FMT_EXTENTS ||
+			      dip->di_aformat == XFS_DINODE_FMT_BTREE);
+
+	/* di_cowextsize */
+	if (flags2 & XFS_DIFLAG2_COWEXTSIZE) {
+		cowextsize = be32_to_cpu(dip->di_cowextsize);
+		XFS_SCRUB_INODE_CHECK(xfs_sb_version_hasreflink(&mp->m_sb));
+		XFS_SCRUB_INODE_CHECK(cowextsize > 0);
+		XFS_SCRUB_INODE_CHECK(cowextsize <= MAXEXTLEN);
+		XFS_SCRUB_INODE_CHECK(cowextsize <= mp->m_sb.sb_agblocks / 2);
+	}
+
+	/* Now let's do the things that require a live inode. */
+	if (!sc->ip)
+		goto out;
+
+	/*
+	 * If this is a reflink inode with no CoW in progress, maybe we
+	 * can turn off the reflink flag?
+	 */
+	if (xfs_is_reflink_inode(sc->ip)) {
+		ifp = XFS_IFORK_PTR(sc->ip, XFS_COW_FORK);
+		XFS_SCRUB_INODE_PREEN(ifp->if_bytes > 0);
+	}
+
+out:
+	if (bp)
+		xfs_trans_brelse(sc->tp, bp);
+	return error;
+}
+#undef XFS_SCRUB_INODE_PREEN
+#undef XFS_SCRUB_INODE_OP_ERROR_GOTO
+#undef XFS_SCRUB_INODE_GOTO
+#undef XFS_SCRUB_INODE_CHECK
diff --git a/fs/xfs/xfs_itable.c b/fs/xfs/xfs_itable.c
index 66e8817..4fd5fe1 100644
--- a/fs/xfs/xfs_itable.c
+++ b/fs/xfs/xfs_itable.c
@@ -31,7 +31,7 @@
 #include "xfs_trace.h"
 #include "xfs_icache.h"
 
-STATIC int
+int
 xfs_internal_inum(
 	xfs_mount_t	*mp,
 	xfs_ino_t	ino)
diff --git a/fs/xfs/xfs_itable.h b/fs/xfs/xfs_itable.h
index 6ea8b39..dd2427b 100644
--- a/fs/xfs/xfs_itable.h
+++ b/fs/xfs/xfs_itable.h
@@ -96,4 +96,9 @@ xfs_inumbers(
 	void			__user *buffer, /* buffer with inode info */
 	inumbers_fmt_pf		formatter);
 
+int
+xfs_internal_inum(
+	xfs_mount_t	*mp,
+	xfs_ino_t	ino);
+
 #endif	/* __XFS_ITABLE_H__ */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 4757fea..2cf344e 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3363,7 +3363,8 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_INOBT,		"inobt" }, \
 	{ XFS_SCRUB_TYPE_FINOBT,	"finobt" }, \
 	{ XFS_SCRUB_TYPE_RMAPBT,	"rmapbt" }, \
-	{ XFS_SCRUB_TYPE_REFCNTBT,	"refcountbt" }
+	{ XFS_SCRUB_TYPE_REFCNTBT,	"refcountbt" }, \
+	{ XFS_SCRUB_TYPE_INODE,		"inode" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 25/55] xfs: scrub inode block mappings

[Darrick J. Wong]

Scrub an individual inode's block mappings to make sure they make sense.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile                |    1 
 fs/xfs/libxfs/xfs_bmap_btree.c |   26 +++
 fs/xfs/libxfs/xfs_fs.h         |    5 -
 fs/xfs/scrub/bmap.c            |  351 ++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.c          |    3 
 fs/xfs/scrub/common.h          |    7 +
 fs/xfs/xfs_bmap_util.c         |  105 +++++++-----
 fs/xfs/xfs_bmap_util.h         |    4 
 fs/xfs/xfs_trace.h             |    5 -
 9 files changed, 459 insertions(+), 48 deletions(-)
 create mode 100644 fs/xfs/scrub/bmap.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 83fc8d3..7fad1d8 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -106,6 +106,7 @@ xfs-y				+= xfs_aops.o \
 xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   agheader.o \
 				   alloc.o \
+				   bmap.o \
 				   btree.o \
 				   common.o \
 				   ialloc.o \
diff --git a/fs/xfs/libxfs/xfs_bmap_btree.c b/fs/xfs/libxfs/xfs_bmap_btree.c
index d9be241..fde4a29 100644
--- a/fs/xfs/libxfs/xfs_bmap_btree.c
+++ b/fs/xfs/libxfs/xfs_bmap_btree.c
@@ -622,6 +622,16 @@ xfs_bmbt_init_key_from_rec(
 }
 
 STATIC void
+xfs_bmbt_init_high_key_from_rec(
+	union xfs_btree_key	*key,
+	union xfs_btree_rec	*rec)
+{
+	key->bmbt.br_startoff = cpu_to_be64(
+			xfs_bmbt_disk_get_startoff(&rec->bmbt) +
+			xfs_bmbt_disk_get_blockcount(&rec->bmbt) - 1);
+}
+
+STATIC void
 xfs_bmbt_init_rec_from_cur(
 	struct xfs_btree_cur	*cur,
 	union xfs_btree_rec	*rec)
@@ -646,6 +656,16 @@ xfs_bmbt_key_diff(
 				      cur->bc_rec.b.br_startoff;
 }
 
+STATIC __int64_t
+xfs_bmbt_diff_two_keys(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*k1,
+	union xfs_btree_key	*k2)
+{
+	return (__int64_t)be64_to_cpu(k1->bmbt.br_startoff) -
+			  be64_to_cpu(k2->bmbt.br_startoff);
+}
+
 static bool
 xfs_bmbt_verify(
 	struct xfs_buf		*bp)
@@ -736,7 +756,6 @@ const struct xfs_buf_ops xfs_bmbt_buf_ops = {
 };
 
 
-#if defined(DEBUG) || defined(XFS_WARN)
 STATIC int
 xfs_bmbt_keys_inorder(
 	struct xfs_btree_cur	*cur,
@@ -757,7 +776,6 @@ xfs_bmbt_recs_inorder(
 		xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
 		xfs_bmbt_disk_get_startoff(&r2->bmbt);
 }
-#endif	/* DEBUG */
 
 static const struct xfs_btree_ops xfs_bmbt_ops = {
 	.rec_len		= sizeof(xfs_bmbt_rec_t),
@@ -771,14 +789,14 @@ static const struct xfs_btree_ops xfs_bmbt_ops = {
 	.get_minrecs		= xfs_bmbt_get_minrecs,
 	.get_dmaxrecs		= xfs_bmbt_get_dmaxrecs,
 	.init_key_from_rec	= xfs_bmbt_init_key_from_rec,
+	.init_high_key_from_rec	= xfs_bmbt_init_high_key_from_rec,
 	.init_rec_from_cur	= xfs_bmbt_init_rec_from_cur,
 	.init_ptr_from_cur	= xfs_bmbt_init_ptr_from_cur,
 	.key_diff		= xfs_bmbt_key_diff,
+	.diff_two_keys		= xfs_bmbt_diff_two_keys,
 	.buf_ops		= &xfs_bmbt_buf_ops,
-#if defined(DEBUG) || defined(XFS_WARN)
 	.keys_inorder		= xfs_bmbt_keys_inorder,
 	.recs_inorder		= xfs_bmbt_recs_inorder,
-#endif
 };
 
 /*
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 5cafbf6..eb32870 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -586,7 +586,10 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_RMAPBT	9	/* reverse mapping btree */
 #define XFS_SCRUB_TYPE_REFCNTBT	10	/* reference count btree */
 #define XFS_SCRUB_TYPE_INODE	11	/* inode record */
-#define XFS_SCRUB_TYPE_MAX	11
+#define XFS_SCRUB_TYPE_BMBTD	12	/* data fork block mapping */
+#define XFS_SCRUB_TYPE_BMBTA	13	/* attr fork block mapping */
+#define XFS_SCRUB_TYPE_BMBTC	14	/* CoW fork block mapping */
+#define XFS_SCRUB_TYPE_MAX	14
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/scrub/bmap.c b/fs/xfs/scrub/bmap.c
new file mode 100644
index 0000000..87ee945
--- /dev/null
+++ b/fs/xfs/scrub/bmap.c
@@ -0,0 +1,351 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_inode_fork.h"
+#include "xfs_bmap.h"
+#include "xfs_bmap_util.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_rmap.h"
+#include "scrub/common.h"
+#include "scrub/btree.h"
+
+/* Set us up with an inode and AG headers, if needed. */
+int
+xfs_scrub_setup_inode_bmap(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	int				error;
+
+	error = xfs_scrub_setup_inode(sc, ip, sm, retry_deadlocked);
+	if (error || !retry_deadlocked)
+		return error;
+
+	error = xfs_scrub_ag_lock_all(sc);
+	if (error)
+		goto err;
+	sc->retry = retry_deadlocked;
+	return 0;
+err:
+	return xfs_scrub_teardown(sc, ip, error);
+}
+
+/*
+ * Inode fork block mapping (BMBT) scrubber.
+ * More complex than the others because we have to scrub
+ * all the extents regardless of whether or not the fork
+ * is in btree format.
+ */
+
+struct xfs_scrub_bmap_info {
+	struct xfs_scrub_context	*sc;
+	const char			*type;
+	xfs_daddr_t			eofs;
+	xfs_fileoff_t			lastoff;
+	bool				is_rt;
+	bool				is_shared;
+	int				whichfork;
+};
+
+#define XFS_SCRUB_BMAP_CHECK(fs_ok) \
+	XFS_SCRUB_INO_CHECK(info->sc, info->sc->ip->i_ino, bp, info->type, fs_ok)
+#define XFS_SCRUB_BMAP_GOTO(fs_ok, label) \
+	XFS_SCRUB_INO_GOTO(info->sc, info->sc->ip->i_ino, bp, info->type, fs_ok, label)
+#define XFS_SCRUB_BMAP_OP_ERROR_GOTO(label) \
+	XFS_SCRUB_OP_ERROR_GOTO(info->sc, agno, 0, "bmap", &error, label);
+/* Scrub a single extent record. */
+STATIC int
+xfs_scrub_bmap_extent(
+	struct xfs_inode		*ip,
+	struct xfs_btree_cur		*cur,
+	struct xfs_scrub_bmap_info	*info,
+	struct xfs_bmbt_irec		*irec)
+{
+	struct xfs_scrub_ag		sa = {0};
+	struct xfs_mount		*mp = ip->i_mount;
+	struct xfs_buf			*bp = NULL;
+	xfs_daddr_t			daddr;
+	xfs_daddr_t			dlen;
+	xfs_agnumber_t			agno;
+	int				error = 0;
+
+	if (cur)
+		xfs_btree_get_block(cur, 0, &bp);
+
+	XFS_SCRUB_BMAP_CHECK(irec->br_startoff >= info->lastoff);
+	XFS_SCRUB_BMAP_CHECK(irec->br_startblock != HOLESTARTBLOCK);
+	XFS_SCRUB_BMAP_CHECK(!isnullstartblock(irec->br_startblock));
+
+	/* Actual mapping, so check the block ranges. */
+	if (info->is_rt) {
+		daddr = XFS_FSB_TO_BB(mp, irec->br_startblock);
+		agno = NULLAGNUMBER;
+	} else {
+		daddr = XFS_FSB_TO_DADDR(mp, irec->br_startblock);
+		agno = XFS_FSB_TO_AGNO(mp, irec->br_startblock);
+		XFS_SCRUB_BMAP_GOTO(agno < mp->m_sb.sb_agcount, out);
+	}
+	dlen = XFS_FSB_TO_BB(mp, irec->br_blockcount);
+	XFS_SCRUB_BMAP_CHECK(agno < mp->m_sb.sb_agcount);
+	XFS_SCRUB_BMAP_CHECK(bno < mp->m_sb.sb_agblocks);
+	XFS_SCRUB_BMAP_CHECK(irec->br_blockcount > 0);
+	XFS_SCRUB_BMAP_CHECK(irec->br_blockcount <= MAXEXTLEN);
+	XFS_SCRUB_BMAP_CHECK(daddr < info->eofs);
+	XFS_SCRUB_BMAP_CHECK(daddr + dlen < info->eofs);
+	XFS_SCRUB_BMAP_CHECK(irec->br_state != XFS_EXT_UNWRITTEN ||
+			xfs_sb_version_hasextflgbit(&mp->m_sb));
+	if (error)
+		goto out;
+
+	/* Set ourselves up for cross-referencing later. */
+	if (!info->is_rt) {
+		if (!xfs_scrub_ag_can_lock(info->sc, agno))
+			return -EDEADLOCK;
+		error = xfs_scrub_ag_init(info->sc, agno, &sa);
+		XFS_SCRUB_BMAP_OP_ERROR_GOTO(out);
+	}
+
+	xfs_scrub_ag_free(&sa);
+out:
+	info->lastoff = irec->br_startoff + irec->br_blockcount;
+	return error;
+}
+#undef XFS_SCRUB_BMAP_OP_ERROR_GOTO
+#undef XFS_SCRUB_BMAP_GOTO
+
+/* Scrub a bmbt record. */
+STATIC int
+xfs_scrub_bmapbt_helper(
+	struct xfs_scrub_btree		*bs,
+	union xfs_btree_rec		*rec)
+{
+	struct xfs_bmbt_rec_host	ihost;
+	struct xfs_bmbt_irec		irec;
+	struct xfs_scrub_bmap_info	*info = bs->private;
+	struct xfs_inode		*ip = bs->cur->bc_private.b.ip;
+	struct xfs_buf			*bp = NULL;
+	struct xfs_btree_block		*block;
+	uint64_t			owner;
+	int				i;
+
+	/*
+	 * Check the owners of the btree blocks up to the level below
+	 * the root since the verifiers don't do that.
+	 */
+	if (xfs_sb_version_hascrc(&bs->cur->bc_mp->m_sb) &&
+	    bs->cur->bc_ptrs[0] == 1) {
+		for (i = 0; i < bs->cur->bc_nlevels - 1; i++) {
+			block = xfs_btree_get_block(bs->cur, i, &bp);
+			owner = be64_to_cpu(block->bb_u.l.bb_owner);
+			XFS_SCRUB_BMAP_CHECK(owner == ip->i_ino);
+		}
+	}
+
+	/* Set up the in-core record and scrub it. */
+	ihost.l0 = be64_to_cpu(rec->bmbt.l0);
+	ihost.l1 = be64_to_cpu(rec->bmbt.l1);
+	xfs_bmbt_get_all(&ihost, &irec);
+	return xfs_scrub_bmap_extent(ip, bs->cur, info, &irec);
+}
+#undef XFS_SCRUB_BMAP_CHECK
+
+#define XFS_SCRUB_FORK_CHECK(fs_ok) \
+	XFS_SCRUB_INO_CHECK(sc, ip->i_ino, NULL, info.type, fs_ok);
+#define XFS_SCRUB_FORK_GOTO(fs_ok, label) \
+	XFS_SCRUB_INO_GOTO(sc, ip->i_ino, NULL, info.type, fs_ok, label);
+#define XFS_SCRUB_FORK_OP_ERROR_GOTO(label) \
+	XFS_SCRUB_OP_ERROR_GOTO(sc, \
+			XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino), \
+			XFS_INO_TO_AGBNO(ip->i_mount, ip->i_ino), \
+			info.type, &error, label)
+/* Scrub an inode fork's block mappings. */
+STATIC int
+xfs_scrub_bmap(
+	struct xfs_scrub_context	*sc,
+	int				whichfork)
+{
+	struct xfs_bmbt_irec		irec;
+	struct xfs_scrub_bmap_info	info = {0};
+	struct xfs_owner_info		oinfo;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_inode		*ip = sc->ip;
+	struct xfs_ifork		*ifp;
+	struct xfs_btree_cur		*cur;
+	xfs_fileoff_t			off;
+	xfs_fileoff_t			endoff;
+	int				nmaps;
+	int				flags = 0;
+	int				error = 0;
+	int				err2 = 0;
+
+	switch (whichfork) {
+	case XFS_DATA_FORK:
+		info.type = "data fork";
+		break;
+	case XFS_ATTR_FORK:
+		info.type = "attr fork";
+		break;
+	case XFS_COW_FORK:
+		info.type = "CoW fork";
+		break;
+	}
+	ifp = XFS_IFORK_PTR(ip, whichfork);
+
+	info.is_rt = whichfork == XFS_DATA_FORK && XFS_IS_REALTIME_INODE(ip);
+	info.eofs = XFS_FSB_TO_BB(mp, info.is_rt ? mp->m_sb.sb_rblocks :
+					      mp->m_sb.sb_dblocks);
+	info.whichfork = whichfork;
+	info.is_shared = whichfork == XFS_DATA_FORK && xfs_is_reflink_inode(ip);
+	info.sc = sc;
+
+	switch (whichfork) {
+	case XFS_COW_FORK:
+		/* Non-existent CoW forks are ignorable. */
+		if (!ifp)
+			goto out_unlock;
+		/* No CoW forks on non-reflink inodes/filesystems. */
+		XFS_SCRUB_FORK_GOTO(xfs_is_reflink_inode(ip), out_unlock);
+		break;
+	case XFS_ATTR_FORK:
+		if (!ifp)
+			goto out_unlock;
+		XFS_SCRUB_FORK_CHECK(xfs_sb_version_hasattr(&mp->m_sb) ||
+				     xfs_sb_version_hasattr2(&mp->m_sb));
+		break;
+	}
+
+	/* Check the fork values */
+	switch (XFS_IFORK_FORMAT(ip, whichfork)) {
+	case XFS_DINODE_FMT_UUID:
+	case XFS_DINODE_FMT_DEV:
+	case XFS_DINODE_FMT_LOCAL:
+		/* No mappings to check. */
+		goto out_unlock;
+	case XFS_DINODE_FMT_EXTENTS:
+		XFS_SCRUB_FORK_GOTO(ifp->if_flags & XFS_IFEXTENTS, out_unlock);
+		break;
+	case XFS_DINODE_FMT_BTREE:
+		XFS_SCRUB_FORK_CHECK(whichfork != XFS_COW_FORK);
+		/* Scan the btree records. */
+		cur = xfs_bmbt_init_cursor(mp, sc->tp, ip, whichfork);
+		xfs_rmap_ino_bmbt_owner(&oinfo, ip->i_ino, whichfork);
+		err2 = xfs_scrub_btree(sc, cur, xfs_scrub_bmapbt_helper,
+				&oinfo, &info);
+		xfs_btree_del_cursor(cur, err2 ? XFS_BTREE_ERROR :
+						 XFS_BTREE_NOERROR);
+		if (err2 == -EDEADLOCK)
+			return err2;
+		else if (err2)
+			goto out_unlock;
+		break;
+	default:
+		XFS_SCRUB_FORK_GOTO(false, out_unlock);
+		break;
+	}
+
+	/* Extent data is in memory, so scrub that. */
+	switch (whichfork) {
+	case XFS_ATTR_FORK:
+		flags |= XFS_BMAPI_ATTRFORK;
+		break;
+	case XFS_COW_FORK:
+		flags |= XFS_BMAPI_COWFORK;
+		break;
+	default:
+		break;
+	}
+
+	/* Find the offset of the last extent in the mapping. */
+	error = xfs_bmap_last_offset(ip, &endoff, whichfork);
+	XFS_SCRUB_FORK_OP_ERROR_GOTO(out_unlock);
+
+	/* Scrub extent records. */
+	off = 0;
+	info.lastoff = 0;
+	while (true) {
+		nmaps = 1;
+		err2 = xfs_bmapi_read(ip, off, endoff - off, &irec,
+				&nmaps, flags);
+		if (err2 || nmaps == 0 || irec.br_startoff > endoff)
+			break;
+		/* Scrub non-hole extent. */
+		if (irec.br_startblock != HOLESTARTBLOCK &&
+		    irec.br_startblock != DELAYSTARTBLOCK) {
+			err2 = xfs_scrub_bmap_extent(ip, NULL, &info, &irec);
+			if (err2 == -EDEADLOCK)
+				return err2;
+			else if (!error && err2)
+				error = err2;
+			if (xfs_scrub_should_terminate(&error))
+				break;
+		}
+
+		off += irec.br_blockcount;
+	}
+
+out_unlock:
+	if (error == 0 && err2 != 0)
+		error = err2;
+	return error;
+}
+#undef XFS_SCRUB_FORK_CHECK
+#undef XFS_SCRUB_FORK_GOTO
+
+/* Scrub an inode's data fork. */
+int
+xfs_scrub_bmap_data(
+	struct xfs_scrub_context	*sc)
+{
+	return xfs_scrub_bmap(sc, XFS_DATA_FORK);
+}
+
+/* Scrub an inode's attr fork. */
+int
+xfs_scrub_bmap_attr(
+	struct xfs_scrub_context	*sc)
+{
+	return xfs_scrub_bmap(sc, XFS_ATTR_FORK);
+}
+
+/* Scrub an inode's CoW fork. */
+int
+xfs_scrub_bmap_cow(
+	struct xfs_scrub_context	*sc)
+{
+	if (!xfs_is_reflink_inode(sc->ip))
+		return -ENOENT;
+
+	return xfs_scrub_bmap(sc, XFS_COW_FORK);
+}
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index a5fecae..545c46a 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -680,6 +680,9 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag_header, xfs_scrub_rmapbt, NULL, xfs_sb_version_hasrmapbt},
 	{xfs_scrub_setup_ag_header, xfs_scrub_refcountbt, NULL, xfs_sb_version_hasreflink},
 	{xfs_scrub_setup_inode_raw, xfs_scrub_inode, NULL, NULL},
+	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_data, NULL, NULL},
+	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_attr, NULL, NULL},
+	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_cow, NULL, NULL},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index ae4bee5..8c30e78 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -222,6 +222,10 @@ int xfs_scrub_setup_inode_raw(struct xfs_scrub_context *sc,
 			      struct xfs_inode *ip,
 			      struct xfs_scrub_metadata *sm,
 			      bool retry_deadlocked);
+int xfs_scrub_setup_inode_bmap(struct xfs_scrub_context *sc,
+			       struct xfs_inode *ip,
+			       struct xfs_scrub_metadata *sm,
+			       bool retry_deadlocked);
 
 /* Metadata scrubbers */
 
@@ -236,5 +240,8 @@ int xfs_scrub_finobt(struct xfs_scrub_context *sc);
 int xfs_scrub_rmapbt(struct xfs_scrub_context *sc);
 int xfs_scrub_refcountbt(struct xfs_scrub_context *sc);
 int xfs_scrub_inode(struct xfs_scrub_context *sc);
+int xfs_scrub_bmap_data(struct xfs_scrub_context *sc);
+int xfs_scrub_bmap_attr(struct xfs_scrub_context *sc);
+int xfs_scrub_bmap_cow(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/xfs_bmap_util.c b/fs/xfs/xfs_bmap_util.c
index b9abce5..d983f28 100644
--- a/fs/xfs/xfs_bmap_util.c
+++ b/fs/xfs/xfs_bmap_util.c
@@ -231,7 +231,7 @@ xfs_bmap_count_leaves(
 	xfs_ifork_t		*ifp,
 	xfs_extnum_t		idx,
 	int			numrecs,
-	int			*count)
+	unsigned long long	*count)
 {
 	int		b;
 
@@ -250,7 +250,7 @@ xfs_bmap_disk_count_leaves(
 	struct xfs_mount	*mp,
 	struct xfs_btree_block	*block,
 	int			numrecs,
-	int			*count)
+	unsigned long long	*count)
 {
 	int		b;
 	xfs_bmbt_rec_t	*frp;
@@ -265,17 +265,18 @@ xfs_bmap_disk_count_leaves(
  * Recursively walks each level of a btree
  * to count total fsblocks in use.
  */
-STATIC int                                     /* error */
+STATIC int
 xfs_bmap_count_tree(
-	xfs_mount_t     *mp,            /* file system mount point */
-	xfs_trans_t     *tp,            /* transaction pointer */
-	xfs_ifork_t	*ifp,		/* inode fork pointer */
-	xfs_fsblock_t   blockno,	/* file system block number */
-	int             levelin,	/* level in btree */
-	int		*count)		/* Count of blocks */
+	struct xfs_mount	*mp,
+	struct xfs_trans	*tp,
+	struct xfs_ifork	*ifp,
+	xfs_fsblock_t		blockno,
+	int			levelin,
+	unsigned int		*nextents,
+	unsigned long long	*count)
 {
 	int			error;
-	xfs_buf_t		*bp, *nbp;
+	struct xfs_buf		*bp, *nbp;
 	int			level = levelin;
 	__be64			*pp;
 	xfs_fsblock_t           bno = blockno;
@@ -308,8 +309,9 @@ xfs_bmap_count_tree(
 		/* Dive to the next level */
 		pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]);
 		bno = be64_to_cpu(*pp);
-		if (unlikely((error =
-		     xfs_bmap_count_tree(mp, tp, ifp, bno, level, count)) < 0)) {
+		error = xfs_bmap_count_tree(mp, tp, ifp, bno, level, nextents,
+				count);
+		if (error) {
 			xfs_trans_brelse(tp, bp);
 			XFS_ERROR_REPORT("xfs_bmap_count_tree(1)",
 					 XFS_ERRLEVEL_LOW, mp);
@@ -321,6 +323,7 @@ xfs_bmap_count_tree(
 		for (;;) {
 			nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
 			numrecs = be16_to_cpu(block->bb_numrecs);
+			(*nextents) += numrecs;
 			xfs_bmap_disk_count_leaves(mp, block, numrecs, count);
 			xfs_trans_brelse(tp, bp);
 			if (nextbno == NULLFSBLOCK)
@@ -341,44 +344,61 @@ xfs_bmap_count_tree(
 /*
  * Count fsblocks of the given fork.
  */
-static int					/* error */
+int
 xfs_bmap_count_blocks(
-	xfs_trans_t		*tp,		/* transaction pointer */
-	xfs_inode_t		*ip,		/* incore inode */
-	int			whichfork,	/* data or attr fork */
-	int			*count)		/* out: count of blocks */
+	struct xfs_trans	*tp,
+	struct xfs_inode	*ip,
+	int			whichfork,
+	unsigned int		*nextents,
+	unsigned long long	*count)
 {
 	struct xfs_btree_block	*block;	/* current btree block */
 	xfs_fsblock_t		bno;	/* block # of "block" */
-	xfs_ifork_t		*ifp;	/* fork structure */
+	struct xfs_ifork	*ifp;	/* fork structure */
 	int			level;	/* btree level, for checking */
-	xfs_mount_t		*mp;	/* file system mount structure */
+	struct xfs_mount	*mp;	/* file system mount structure */
 	__be64			*pp;	/* pointer to block address */
+	int			error;
 
 	bno = NULLFSBLOCK;
 	mp = ip->i_mount;
+	*nextents = 0;
 	ifp = XFS_IFORK_PTR(ip, whichfork);
-	if ( XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_EXTENTS ) {
-		xfs_bmap_count_leaves(ifp, 0, xfs_iext_count(ifp), count);
+	if (!ifp)
 		return 0;
-	}
 
-	/*
-	 * Root level must use BMAP_BROOT_PTR_ADDR macro to get ptr out.
-	 */
-	block = ifp->if_broot;
-	level = be16_to_cpu(block->bb_level);
-	ASSERT(level > 0);
-	pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes);
-	bno = be64_to_cpu(*pp);
-	ASSERT(bno != NULLFSBLOCK);
-	ASSERT(XFS_FSB_TO_AGNO(mp, bno) < mp->m_sb.sb_agcount);
-	ASSERT(XFS_FSB_TO_AGBNO(mp, bno) < mp->m_sb.sb_agblocks);
-
-	if (unlikely(xfs_bmap_count_tree(mp, tp, ifp, bno, level, count) < 0)) {
-		XFS_ERROR_REPORT("xfs_bmap_count_blocks(2)", XFS_ERRLEVEL_LOW,
-				 mp);
-		return -EFSCORRUPTED;
+	switch (XFS_IFORK_FORMAT(ip, whichfork)) {
+	case XFS_DINODE_FMT_EXTENTS:
+		*nextents = xfs_iext_count(ifp);
+		xfs_bmap_count_leaves(ifp, 0, (*nextents), count);
+		return 0;
+	case XFS_DINODE_FMT_BTREE:
+		if (!(ifp->if_flags & XFS_IFEXTENTS)) {
+			error = xfs_iread_extents(tp, ip, whichfork);
+			if (error)
+				return error;
+		}
+
+		/*
+		 * Root level must use BMAP_BROOT_PTR_ADDR macro to get ptr out.
+		 */
+		block = ifp->if_broot;
+		level = be16_to_cpu(block->bb_level);
+		ASSERT(level > 0);
+		pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes);
+		bno = be64_to_cpu(*pp);
+		ASSERT(bno != NULLFSBLOCK);
+		ASSERT(XFS_FSB_TO_AGNO(mp, bno) < mp->m_sb.sb_agcount);
+		ASSERT(XFS_FSB_TO_AGBNO(mp, bno) < mp->m_sb.sb_agblocks);
+
+		error = xfs_bmap_count_tree(mp, tp, ifp, bno, level,
+				nextents, count);
+		if (error) {
+			XFS_ERROR_REPORT("xfs_bmap_count_blocks(2)",
+					XFS_ERRLEVEL_LOW, mp);
+			return -EFSCORRUPTED;
+		}
+		return 0;
 	}
 
 	return 0;
@@ -1788,8 +1808,9 @@ xfs_swap_extent_forks(
 	int			*target_log_flags)
 {
 	struct xfs_ifork	tempifp, *ifp, *tifp;
-	int			aforkblks = 0;
-	int			taforkblks = 0;
+	unsigned long long	aforkblks = 0;
+	unsigned long long	taforkblks = 0;
+	unsigned int		junk;
 	xfs_extnum_t		nextents;
 	__uint64_t		tmp;
 	int			error;
@@ -1799,14 +1820,14 @@ xfs_swap_extent_forks(
 	 */
 	if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
 	     (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
-		error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK,
+		error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &junk,
 				&aforkblks);
 		if (error)
 			return error;
 	}
 	if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
 	     (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
-		error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
+		error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK, &junk,
 				&taforkblks);
 		if (error)
 			return error;
diff --git a/fs/xfs/xfs_bmap_util.h b/fs/xfs/xfs_bmap_util.h
index 68a621a..73a03c3 100644
--- a/fs/xfs/xfs_bmap_util.h
+++ b/fs/xfs/xfs_bmap_util.h
@@ -71,4 +71,8 @@ int	xfs_swap_extents(struct xfs_inode *ip, struct xfs_inode *tip,
 
 xfs_daddr_t xfs_fsb_to_db(struct xfs_inode *ip, xfs_fsblock_t fsb);
 
+int xfs_bmap_count_blocks(struct xfs_trans *tp, struct xfs_inode *ip,
+			  int whichfork, unsigned int *nextents,
+			  unsigned long long *count);
+
 #endif	/* __XFS_BMAP_UTIL_H__ */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 2cf344e..141a39f 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3364,7 +3364,10 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_FINOBT,	"finobt" }, \
 	{ XFS_SCRUB_TYPE_RMAPBT,	"rmapbt" }, \
 	{ XFS_SCRUB_TYPE_REFCNTBT,	"refcountbt" }, \
-	{ XFS_SCRUB_TYPE_INODE,		"inode" }
+	{ XFS_SCRUB_TYPE_INODE,		"inode" }, \
+	{ XFS_SCRUB_TYPE_BMBTD, 	"bmapbtd" }, \
+	{ XFS_SCRUB_TYPE_BMBTA,		"bmapbta" }, \
+	{ XFS_SCRUB_TYPE_BMBTC,		"bmapbtc" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 26/55] xfs: scrub directory/attribute btrees

[Darrick J. Wong]

Provide a way to check the shape and scrub the hashes and records
in a directory or extended attribute btree.  These are helper functions
for the directory & attribute scrubbers in subsequent patches.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile               |    1 
 fs/xfs/libxfs/xfs_dir2_node.c |   28 ++
 fs/xfs/libxfs/xfs_dir2_priv.h |    2 
 fs/xfs/scrub/dabtree.c        |  472 +++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/dabtree.h        |   62 +++++
 5 files changed, 565 insertions(+)
 create mode 100644 fs/xfs/scrub/dabtree.c
 create mode 100644 fs/xfs/scrub/dabtree.h


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 7fad1d8..1b1ed40 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -109,6 +109,7 @@ xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   bmap.o \
 				   btree.o \
 				   common.o \
+				   dabtree.o \
 				   ialloc.o \
 				   inode.o \
 				   refcount.o \
diff --git a/fs/xfs/libxfs/xfs_dir2_node.c b/fs/xfs/libxfs/xfs_dir2_node.c
index 6f04e40..7415161 100644
--- a/fs/xfs/libxfs/xfs_dir2_node.c
+++ b/fs/xfs/libxfs/xfs_dir2_node.c
@@ -526,6 +526,34 @@ xfs_dir2_free_hdr_check(
  * Stale entries are ok.
  */
 xfs_dahash_t					/* hash value */
+xfs_dir2_leaf1_lasthash(
+	struct xfs_inode *dp,
+	struct xfs_buf	*bp,			/* leaf buffer */
+	int		*count)			/* count of entries in leaf */
+{
+	struct xfs_dir2_leaf	*leaf = bp->b_addr;
+	struct xfs_dir2_leaf_entry *ents;
+	struct xfs_dir3_icleaf_hdr leafhdr;
+
+	dp->d_ops->leaf_hdr_from_disk(&leafhdr, leaf);
+
+	ASSERT(leafhdr.magic == XFS_DIR2_LEAF1_MAGIC ||
+	       leafhdr.magic == XFS_DIR3_LEAF1_MAGIC);
+
+	if (count)
+		*count = leafhdr.count;
+	if (!leafhdr.count)
+		return 0;
+
+	ents = dp->d_ops->leaf_ents_p(leaf);
+	return be32_to_cpu(ents[leafhdr.count - 1].hashval);
+}
+
+/*
+ * Return the last hash value in the leaf.
+ * Stale entries are ok.
+ */
+xfs_dahash_t					/* hash value */
 xfs_dir2_leafn_lasthash(
 	struct xfs_inode *dp,
 	struct xfs_buf	*bp,			/* leaf buffer */
diff --git a/fs/xfs/libxfs/xfs_dir2_priv.h b/fs/xfs/libxfs/xfs_dir2_priv.h
index d04547f..1abd314 100644
--- a/fs/xfs/libxfs/xfs_dir2_priv.h
+++ b/fs/xfs/libxfs/xfs_dir2_priv.h
@@ -93,6 +93,8 @@ extern bool xfs_dir3_leaf_check_int(struct xfs_mount *mp, struct xfs_inode *dp,
 /* xfs_dir2_node.c */
 extern int xfs_dir2_leaf_to_node(struct xfs_da_args *args,
 		struct xfs_buf *lbp);
+extern xfs_dahash_t xfs_dir2_leaf1_lasthash(struct xfs_inode *dp,
+		struct xfs_buf *bp, int *count);
 extern xfs_dahash_t xfs_dir2_leafn_lasthash(struct xfs_inode *dp,
 		struct xfs_buf *bp, int *count);
 extern int xfs_dir2_leafn_lookup_int(struct xfs_buf *bp,
diff --git a/fs/xfs/scrub/dabtree.c b/fs/xfs/scrub/dabtree.c
new file mode 100644
index 0000000..3090074
--- /dev/null
+++ b/fs/xfs/scrub/dabtree.c
@@ -0,0 +1,472 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_inode_fork.h"
+#include "xfs_da_format.h"
+#include "xfs_da_btree.h"
+#include "xfs_dir2.h"
+#include "xfs_dir2_priv.h"
+#include "xfs_attr_leaf.h"
+#include "scrub/common.h"
+#include "scrub/dabtree.h"
+
+/* Directory/Attribute Btree */
+
+/* Find an entry at a certain level in a da btree. */
+STATIC void *
+xfs_scrub_da_btree_entry(
+	struct xfs_scrub_da_btree	*ds,
+	int				level,
+	int				rec)
+{
+	char				*ents;
+	void 				*(*fn)(void *);
+	size_t				sz;
+	struct xfs_da_state_blk		*blk;
+
+	/* Dispatch the entry finding function. */
+	blk = &ds->state->path.blk[level];
+	switch (blk->magic) {
+	case XFS_ATTR_LEAF_MAGIC:
+	case XFS_ATTR3_LEAF_MAGIC:
+		fn = (xfs_da_leaf_ents_fn)xfs_attr3_leaf_entryp;
+		sz = sizeof(struct xfs_attr_leaf_entry);
+		break;
+	case XFS_DIR2_LEAFN_MAGIC:
+	case XFS_DIR3_LEAFN_MAGIC:
+		fn = (xfs_da_leaf_ents_fn)ds->dargs.dp->d_ops->leaf_ents_p;
+		sz = sizeof(struct xfs_dir2_leaf_entry);
+		break;
+	case XFS_DIR2_LEAF1_MAGIC:
+	case XFS_DIR3_LEAF1_MAGIC:
+		fn = (xfs_da_leaf_ents_fn)ds->dargs.dp->d_ops->leaf_ents_p;
+		sz = sizeof(struct xfs_dir2_leaf_entry);
+		break;
+	case XFS_DA_NODE_MAGIC:
+	case XFS_DA3_NODE_MAGIC:
+		fn = (xfs_da_leaf_ents_fn)ds->dargs.dp->d_ops->node_tree_p;
+		sz = sizeof(struct xfs_da_node_entry);
+		break;
+	default:
+		return NULL;
+	}
+
+	ents = fn(blk->bp->b_addr);
+	return ents + (sz * rec);
+}
+
+/* Scrub a da btree hash (key). */
+int
+xfs_scrub_da_btree_hash(
+	struct xfs_scrub_da_btree	*ds,
+	int				level,
+	__be32				*hashp)
+{
+	struct xfs_da_state_blk		*blks;
+	struct xfs_da_node_entry	*btree;
+	xfs_dahash_t			hash;
+	xfs_dahash_t			parent_hash;
+	int				error = 0;
+
+	/* Is this hash in order? */
+	hash = be32_to_cpu(*hashp);
+	XFS_SCRUB_DA_CHECK(ds, hash >= ds->hashes[level]);
+	ds->hashes[level] = hash;
+
+	if (level == 0)
+		return error;
+
+	/* Is this hash no larger than the parent hash? */
+	blks = ds->state->path.blk;
+	btree = xfs_scrub_da_btree_entry(ds, level - 1, blks[level - 1].index);
+	parent_hash = be32_to_cpu(btree->hashval);
+	XFS_SCRUB_DA_CHECK(ds, hash <= parent_hash);
+
+	return error;
+}
+
+/* Scrub a da btree pointer. */
+STATIC int
+xfs_scrub_da_btree_ptr(
+	struct xfs_scrub_da_btree	*ds,
+	int				level,
+	xfs_dablk_t			blkno)
+{
+	int				error = 0;
+
+	XFS_SCRUB_DA_CHECK(ds, blkno >= ds->lowest);
+	XFS_SCRUB_DA_CHECK(ds, ds->highest == 0 || blkno < ds->highest);
+
+	return error;
+}
+
+/*
+ * The da btree scrubber can handle leaf1 blocks as a degenerate
+ * form of da btree.  Since the regular da code doesn't handle
+ * leaf1, we must multiplex the verifiers.
+ */
+static void
+xfs_scrub_da_btree_read_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_da_blkinfo	*info = bp->b_addr;
+
+	switch (be16_to_cpu(info->magic)) {
+	case XFS_DIR2_LEAF1_MAGIC:
+	case XFS_DIR3_LEAF1_MAGIC:
+		bp->b_ops = &xfs_dir3_leaf1_buf_ops;
+		bp->b_ops->verify_read(bp);
+		return;
+	default:
+		bp->b_ops = &xfs_da3_node_buf_ops;
+		bp->b_ops->verify_read(bp);
+		return;
+	}
+}
+static void
+xfs_scrub_da_btree_write_verify(
+	struct xfs_buf	*bp)
+{
+	struct xfs_da_blkinfo	*info = bp->b_addr;
+
+	switch (be16_to_cpu(info->magic)) {
+	case XFS_DIR2_LEAF1_MAGIC:
+	case XFS_DIR3_LEAF1_MAGIC:
+		bp->b_ops = &xfs_dir3_leaf1_buf_ops;
+		bp->b_ops->verify_write(bp);
+		return;
+	default:
+		bp->b_ops = &xfs_da3_node_buf_ops;
+		bp->b_ops->verify_write(bp);
+		return;
+	}
+}
+
+const static struct xfs_buf_ops xfs_scrub_da_btree_buf_ops = {
+	.name = "xfs_scrub_da_btree",
+	.verify_read = xfs_scrub_da_btree_read_verify,
+	.verify_write = xfs_scrub_da_btree_write_verify,
+};
+
+/* Check a block's sibling pointers. */
+STATIC int
+xfs_scrub_da_btree_block_check_siblings(
+	struct xfs_scrub_da_btree	*ds,
+	int				level,
+	struct xfs_da_blkinfo		*hdr)
+{
+	xfs_dablk_t			forw;
+	xfs_dablk_t			back;
+	int				retval;
+	int				error = 0;
+
+	forw = be32_to_cpu(hdr->forw);
+	back = be32_to_cpu(hdr->back);
+
+	/* Top level blocks should not have sibling pointers. */
+	if (level == 0) {
+		XFS_SCRUB_DA_CHECK(ds, forw == 0);
+		XFS_SCRUB_DA_CHECK(ds, back == 0);
+		return error;
+	}
+
+	/* Check back (left) pointer. */
+	if (back != 0) {
+		/* Move the alternate cursor back one block. */
+		ds->state->altpath = ds->state->path;
+		error = xfs_da3_path_shift(ds->state, &ds->state->altpath,
+				0, false, &retval);
+		XFS_SCRUB_DA_OP_ERROR_GOTO(ds, &error, out);
+		XFS_SCRUB_DA_GOTO(ds, retval == 0, verify_forw);
+		XFS_SCRUB_DA_CHECK(ds,
+				ds->state->altpath.blk[level].blkno == back);
+	}
+
+verify_forw:
+	/* Check forw (right) pointer. */
+	if (!error && forw != 0) {
+		/* Move the alternate cursor forward one block. */
+		ds->state->altpath = ds->state->path;
+		error = xfs_da3_path_shift(ds->state, &ds->state->altpath,
+				1, false, &retval);
+		XFS_SCRUB_DA_OP_ERROR_GOTO(ds, &error, out);
+		XFS_SCRUB_DA_GOTO(ds, retval == 0, out);
+		XFS_SCRUB_DA_CHECK(ds,
+				ds->state->altpath.blk[level].blkno == forw);
+	}
+out:
+	memset(&ds->state->altpath, 0, sizeof(ds->state->altpath));
+	return error;
+}
+
+/* Load a dir/attribute block from a btree. */
+STATIC int
+xfs_scrub_da_btree_block(
+	struct xfs_scrub_da_btree	*ds,
+	int				level,
+	xfs_dablk_t			blkno)
+{
+	struct xfs_da_state_blk		*blk;
+	struct xfs_da_intnode		*node;
+	struct xfs_da_node_entry	*btree;
+	struct xfs_da3_blkinfo		*hdr3;
+	struct xfs_da_args		*dargs = &ds->dargs;
+	struct xfs_inode		*ip = ds->dargs.dp;
+	xfs_ino_t			owner;
+	int				*pmaxrecs;
+	struct xfs_da3_icnode_hdr 	nodehdr;
+	int				error;
+
+	blk = &ds->state->path.blk[level];
+	ds->state->path.active = level + 1;
+
+	/* Release old block. */
+	if (blk->bp) {
+		xfs_trans_brelse(dargs->trans, blk->bp);
+		blk->bp = NULL;
+	}
+
+	/* Check the pointer. */
+	blk->blkno = blkno;
+	error = xfs_scrub_da_btree_ptr(ds, level, blkno);
+	if (error) {
+		blk->blkno = 0;
+		goto out;
+	}
+
+	/* Read the buffer. */
+	error = xfs_da_read_buf(dargs->trans, dargs->dp, blk->blkno, -2,
+			&blk->bp, dargs->whichfork,
+			&xfs_scrub_da_btree_buf_ops);
+	XFS_SCRUB_DA_OP_ERROR_GOTO(ds, &error, out_nobuf);
+
+	/* It's ok for a directory not to have a da btree in it. */
+	if (ds->dargs.whichfork == XFS_DATA_FORK && level == 0 &&
+			blk->bp == NULL)
+		goto out_nobuf;
+	XFS_SCRUB_DA_GOTO(ds, blk->bp != NULL, out_nobuf);
+
+	hdr3 = blk->bp->b_addr;
+	blk->magic = be16_to_cpu(hdr3->hdr.magic);
+	pmaxrecs = &ds->maxrecs[level];
+
+	/* Check the owner. */
+	if (xfs_sb_version_hascrc(&ip->i_mount->m_sb)) {
+		owner = be64_to_cpu(hdr3->owner);
+		error = -EFSCORRUPTED;
+		XFS_SCRUB_DA_GOTO(ds, owner == ip->i_ino, out);
+	}
+
+	/* Check the siblings. */
+	error = xfs_scrub_da_btree_block_check_siblings(ds, level, &hdr3->hdr);
+	if (error)
+		goto out;
+
+	/* Interpret the buffer. */
+	error = -EFSCORRUPTED;
+	switch (blk->magic) {
+	case XFS_ATTR_LEAF_MAGIC:
+	case XFS_ATTR3_LEAF_MAGIC:
+		xfs_trans_buf_set_type(dargs->trans, blk->bp,
+				XFS_BLFT_ATTR_LEAF_BUF);
+		blk->magic = XFS_ATTR_LEAF_MAGIC;
+		blk->hashval = xfs_attr_leaf_lasthash(blk->bp, pmaxrecs);
+		XFS_SCRUB_DA_CHECK(ds, ds->tree_level == 0);
+		break;
+	case XFS_DIR2_LEAFN_MAGIC:
+	case XFS_DIR3_LEAFN_MAGIC:
+		xfs_trans_buf_set_type(dargs->trans, blk->bp,
+				XFS_BLFT_DIR_LEAFN_BUF);
+		blk->magic = XFS_DIR2_LEAFN_MAGIC;
+		blk->hashval = xfs_dir2_leafn_lasthash(ip, blk->bp, pmaxrecs);
+		XFS_SCRUB_DA_CHECK(ds, ds->tree_level == 0);
+		break;
+	case XFS_DIR2_LEAF1_MAGIC:
+	case XFS_DIR3_LEAF1_MAGIC:
+		xfs_trans_buf_set_type(dargs->trans, blk->bp,
+				XFS_BLFT_DIR_LEAF1_BUF);
+		blk->magic = XFS_DIR2_LEAF1_MAGIC;
+		blk->hashval = xfs_dir2_leaf1_lasthash(ip, blk->bp, pmaxrecs);
+		XFS_SCRUB_DA_CHECK(ds, ds->tree_level == 0);
+		break;
+	case XFS_DA_NODE_MAGIC:
+	case XFS_DA3_NODE_MAGIC:
+		xfs_trans_buf_set_type(dargs->trans, blk->bp,
+				XFS_BLFT_DA_NODE_BUF);
+		blk->magic = XFS_DA_NODE_MAGIC;
+		node = blk->bp->b_addr;
+		ip->d_ops->node_hdr_from_disk(&nodehdr, node);
+		btree = ip->d_ops->node_tree_p(node);
+		*pmaxrecs = nodehdr.count;
+		blk->hashval = be32_to_cpu(btree[*pmaxrecs - 1].hashval);
+		if (level == 0) {
+			XFS_SCRUB_DA_GOTO(ds,
+					nodehdr.level < XFS_DA_NODE_MAXDEPTH,
+					out);
+			ds->tree_level = nodehdr.level;
+		} else
+			XFS_SCRUB_DA_GOTO(ds, ds->tree_level == nodehdr.level,
+					out);
+		break;
+	default:
+		XFS_SCRUB_DA_CHECK(ds, false);
+		xfs_trans_brelse(dargs->trans, blk->bp);
+		blk->bp = NULL;
+		blk->blkno = 0;
+		break;
+	}
+	error = 0;
+
+out:
+	return error;
+out_nobuf:
+	blk->blkno = 0;
+	return error;
+}
+
+/* Visit all nodes and leaves of a da btree. */
+int
+xfs_scrub_da_btree(
+	struct xfs_scrub_context	*sc,
+	int				whichfork,
+	xfs_scrub_da_btree_rec_fn	scrub_fn)
+{
+	struct xfs_scrub_da_btree	ds;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_da_state_blk		*blks;
+	struct xfs_da_node_entry	*btree;
+	void				*rec;
+	xfs_dablk_t			blkno;
+	bool				is_attr;
+	int				level;
+	int				error;
+
+	memset(&ds, 0, sizeof(ds));
+	/* Skip short format data structures; no btree to scan. */
+	if (XFS_IFORK_FORMAT(sc->ip, whichfork) != XFS_DINODE_FMT_EXTENTS &&
+	    XFS_IFORK_FORMAT(sc->ip, whichfork) != XFS_DINODE_FMT_BTREE)
+		return 0;
+
+	/* Set up initial da state. */
+	is_attr = whichfork == XFS_ATTR_FORK;
+	ds.dargs.geo = is_attr ? mp->m_attr_geo : mp->m_dir_geo;
+	ds.dargs.dp = sc->ip;
+	ds.dargs.whichfork = whichfork;
+	ds.dargs.trans = sc->tp;
+	ds.dargs.op_flags = XFS_DA_OP_OKNOENT;
+	ds.state = xfs_da_state_alloc();
+	ds.state->args = &ds.dargs;
+	ds.state->mp = sc->ip->i_mount;
+	ds.type = is_attr ? "attr" : "dir";
+	ds.sc = sc;
+	blkno = ds.lowest = is_attr ? 0 : ds.dargs.geo->leafblk;
+	ds.highest = is_attr ? 0 : ds.dargs.geo->freeblk;
+	level = 0;
+
+	/* Find the root of the da tree, if present. */
+	blks = ds.state->path.blk;
+	error = xfs_scrub_da_btree_block(&ds, level, blkno);
+	if (error)
+		goto out_state;
+	if (blks[level].bp == NULL)
+		goto out_state;
+
+	blks[level].index = 0;
+	while (level >= 0 && level < XFS_DA_NODE_MAXDEPTH) {
+		/* Handle leaf block. */
+		if (blks[level].magic != XFS_DA_NODE_MAGIC) {
+			/* End of leaf, pop back towards the root. */
+			if (blks[level].index >= ds.maxrecs[level]) {
+				if (level > 0)
+					blks[level - 1].index++;
+				ds.tree_level++;
+				level--;
+				continue;
+			}
+
+			/* Dispatch record scrubbing. */
+			rec = xfs_scrub_da_btree_entry(&ds, level,
+					blks[level].index);
+			error = scrub_fn(&ds, level, rec);
+			if (error < 0 ||
+			    error == XFS_BTREE_QUERY_RANGE_ABORT)
+				break;
+			if (xfs_scrub_should_terminate(&error))
+				break;
+
+			blks[level].index++;
+			continue;
+		}
+
+		btree = xfs_scrub_da_btree_entry(&ds, level, blks[level].index);
+
+		/* End of node, pop back towards the root. */
+		if (blks[level].index >= ds.maxrecs[level]) {
+			if (level > 0)
+				blks[level - 1].index++;
+			ds.tree_level++;
+			level--;
+			continue;
+		}
+
+		/* Hashes in order for scrub? */
+		error = xfs_scrub_da_btree_hash(&ds, level, &btree->hashval);
+		if (error)
+			goto out;
+
+		/* Drill another level deeper. */
+		blkno = be32_to_cpu(btree->before);
+		level++;
+		ds.tree_level--;
+		error = xfs_scrub_da_btree_block(&ds, level, blkno);
+		if (error)
+			goto out;
+		if (blks[level].bp == NULL)
+			goto out;
+
+		blks[level].index = 0;
+	}
+
+out:
+	/* Release all the buffers we're tracking. */
+	for (level = 0; level < XFS_DA_NODE_MAXDEPTH; level++) {
+		if (blks[level].bp == NULL)
+			continue;
+		xfs_trans_brelse(sc->tp, blks[level].bp);
+		blks[level].bp = NULL;
+	}
+
+out_state:
+	xfs_da_state_free(ds.state);
+	return error;
+}
diff --git a/fs/xfs/scrub/dabtree.h b/fs/xfs/scrub/dabtree.h
new file mode 100644
index 0000000..1302d67
--- /dev/null
+++ b/fs/xfs/scrub/dabtree.h
@@ -0,0 +1,62 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#ifndef __XFS_REPAIR_DABTREE_H__
+#define __XFS_REPAIR_DABTREE_H__
+
+/* dir/attr btree */
+
+struct xfs_scrub_da_btree {
+	struct xfs_da_args		dargs;
+	xfs_dahash_t			hashes[XFS_DA_NODE_MAXDEPTH];
+	int				maxrecs[XFS_DA_NODE_MAXDEPTH];
+	struct xfs_da_state		*state;
+	const char			*type;
+	struct xfs_scrub_context	*sc;
+	xfs_dablk_t			lowest;
+	xfs_dablk_t			highest;
+	int				tree_level;
+};
+
+typedef void *(*xfs_da_leaf_ents_fn)(void *);
+typedef int (*xfs_scrub_da_btree_rec_fn)(struct xfs_scrub_da_btree *ds,
+		int level, void *rec);
+
+#define XFS_SCRUB_DA_CHECK(ds, fs_ok) \
+	XFS_SCRUB_DATA_CHECK((ds)->sc, (ds)->dargs.whichfork, \
+			xfs_dir2_da_to_db((ds)->dargs.geo, \
+			(ds)->state->path.blk[level].blkno), (ds)->type, \
+			fs_ok)
+#define XFS_SCRUB_DA_GOTO(ds, fs_ok, label) \
+	XFS_SCRUB_DATA_GOTO((ds)->sc, (ds)->dargs.whichfork, \
+			xfs_dir2_da_to_db((ds)->dargs.geo, \
+			(ds)->state->path.blk[level].blkno), (ds)->type, \
+			fs_ok, label)
+#define XFS_SCRUB_DA_OP_ERROR_GOTO(ds, error, label) \
+	XFS_SCRUB_FILE_OP_ERROR_GOTO((ds)->sc, (ds)->dargs.whichfork, \
+			xfs_dir2_da_to_db((ds)->dargs.geo, \
+			(ds)->state->path.blk[level].blkno), (ds)->type, \
+			(error), label)
+
+int xfs_scrub_da_btree_hash(struct xfs_scrub_da_btree *ds, int level,
+			    __be32 *hashp);
+int xfs_scrub_da_btree(struct xfs_scrub_context *sc, int whichfork,
+		       xfs_scrub_da_btree_rec_fn scrub_fn);
+
+#endif /* __XFS_REPAIR_DABTREE_H__ */

[PATCH 27/55] xfs: scrub directory metadata

[Darrick J. Wong]

Scrub the hash tree and all the entries in a directory.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile               |    1 
 fs/xfs/libxfs/xfs_dir2_priv.h |    4 -
 fs/xfs/libxfs/xfs_fs.h        |    3 
 fs/xfs/scrub/common.c         |    1 
 fs/xfs/scrub/common.h         |    1 
 fs/xfs/scrub/dir.c            |  268 +++++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_dir2_readdir.c     |   19 ++-
 fs/xfs/xfs_file.c             |    2 
 fs/xfs/xfs_trace.h            |    3 
 9 files changed, 291 insertions(+), 11 deletions(-)
 create mode 100644 fs/xfs/scrub/dir.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 1b1ed40..c05f503 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -110,6 +110,7 @@ xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   btree.o \
 				   common.o \
 				   dabtree.o \
+				   dir.o \
 				   ialloc.o \
 				   inode.o \
 				   refcount.o \
diff --git a/fs/xfs/libxfs/xfs_dir2_priv.h b/fs/xfs/libxfs/xfs_dir2_priv.h
index 1abd314..4cf2956 100644
--- a/fs/xfs/libxfs/xfs_dir2_priv.h
+++ b/fs/xfs/libxfs/xfs_dir2_priv.h
@@ -129,7 +129,7 @@ extern int xfs_dir2_sf_removename(struct xfs_da_args *args);
 extern int xfs_dir2_sf_replace(struct xfs_da_args *args);
 
 /* xfs_dir2_readdir.c */
-extern int xfs_readdir(struct xfs_inode *dp, struct dir_context *ctx,
-		       size_t bufsize);
+extern int xfs_readdir(struct xfs_trans *tp, struct xfs_inode *dp,
+		       struct dir_context *ctx, size_t bufsize);
 
 #endif /* __XFS_DIR2_PRIV_H__ */
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index eb32870..2a9b959 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -589,7 +589,8 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_BMBTD	12	/* data fork block mapping */
 #define XFS_SCRUB_TYPE_BMBTA	13	/* attr fork block mapping */
 #define XFS_SCRUB_TYPE_BMBTC	14	/* CoW fork block mapping */
-#define XFS_SCRUB_TYPE_MAX	14
+#define XFS_SCRUB_TYPE_DIR	15	/* directory */
+#define XFS_SCRUB_TYPE_MAX	15
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 545c46a..a570d64 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -683,6 +683,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_data, NULL, NULL},
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_attr, NULL, NULL},
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_cow, NULL, NULL},
+	{xfs_scrub_setup_inode, xfs_scrub_directory, NULL, NULL},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 8c30e78..080596b 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -243,5 +243,6 @@ int xfs_scrub_inode(struct xfs_scrub_context *sc);
 int xfs_scrub_bmap_data(struct xfs_scrub_context *sc);
 int xfs_scrub_bmap_attr(struct xfs_scrub_context *sc);
 int xfs_scrub_bmap_cow(struct xfs_scrub_context *sc);
+int xfs_scrub_directory(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/dir.c b/fs/xfs/scrub/dir.c
new file mode 100644
index 0000000..ed29bd1
--- /dev/null
+++ b/fs/xfs/scrub/dir.c
@@ -0,0 +1,268 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_icache.h"
+#include "xfs_itable.h"
+#include "xfs_da_format.h"
+#include "xfs_da_btree.h"
+#include "xfs_dir2.h"
+#include "xfs_dir2_priv.h"
+#include "scrub/common.h"
+#include "scrub/dabtree.h"
+
+/* Directories */
+
+/* Scrub a directory entry. */
+
+struct xfs_scrub_dir_ctx {
+	struct dir_context		dc;
+	struct xfs_scrub_context	*sc;
+};
+
+#define XFS_SCRUB_DIR_CHECK(fs_ok) \
+	XFS_SCRUB_DATA_CHECK(sdc->sc, XFS_DATA_FORK, offset, "dir", fs_ok)
+#define XFS_SCRUB_DIR_GOTO(fs_ok, label) \
+	XFS_SCRUB_DATA_GOTO(sdc->sc, XFS_DATA_FORK, offset, "dir", fs_ok, label)
+#define XFS_SCRUB_DIR_OP_ERROR_GOTO(label) \
+	XFS_SCRUB_FILE_OP_ERROR_GOTO(sdc->sc, XFS_DATA_FORK, offset, "dir", &error, label)
+/* Check that an inode's mode matches a given DT_ type. */
+STATIC int
+xfs_scrub_dir_check_ftype(
+	struct xfs_scrub_dir_ctx	*sdc,
+	xfs_fileoff_t			offset,
+	xfs_ino_t			inum,
+	int				dtype)
+{
+	struct xfs_mount		*mp = sdc->sc->ip->i_mount;
+	struct xfs_inode		*ip;
+	int				ino_dtype;
+	int				error = 0;
+
+	if (!xfs_sb_version_hasftype(&mp->m_sb)) {
+		XFS_SCRUB_DIR_CHECK(dtype == DT_UNKNOWN || dtype == DT_DIR);
+		goto out;
+	}
+
+	error = xfs_iget(mp, sdc->sc->tp, inum, 0, 0, &ip);
+	XFS_SCRUB_OP_ERROR_GOTO(sdc->sc,
+			XFS_INO_TO_AGNO(mp, inum),
+			XFS_INO_TO_AGBNO(mp, inum),
+			"inode", &error, out);
+	/* Convert mode to the DT_* values that dir_emit uses. */
+	ino_dtype = (VFS_I(ip)->i_mode & S_IFMT) >> 12;
+	XFS_SCRUB_DIR_CHECK(ino_dtype == dtype);
+	IRELE(ip);
+out:
+	return error;
+}
+
+/* Scrub a single directory entry. */
+STATIC int
+xfs_scrub_dir_actor(
+	struct dir_context		*dc,
+	const char			*name,
+	int				namelen,
+	loff_t				pos,
+	u64				ino,
+	unsigned			type)
+{
+	struct xfs_mount		*mp;
+	struct xfs_inode		*ip;
+	struct xfs_scrub_dir_ctx	*sdc;
+	struct xfs_name			xname;
+	xfs_ino_t			lookup_ino;
+	xfs_dablk_t			offset;
+	int				error = 0;
+
+	sdc = container_of(dc, struct xfs_scrub_dir_ctx, dc);
+	ip = sdc->sc->ip;
+	mp = ip->i_mount;
+	offset = xfs_dir2_db_to_da(mp->m_dir_geo,
+			xfs_dir2_dataptr_to_db(mp->m_dir_geo, pos));
+
+	/* Does this inode number make sense? */
+	XFS_SCRUB_DIR_GOTO(xfs_dir_ino_validate(mp, ino) == 0, out);
+	XFS_SCRUB_DIR_GOTO(!xfs_internal_inum(mp, ino), out);
+
+	/* Verify that we can look up this name by hash. */
+	xname.name = name;
+	xname.len = namelen;
+	xname.type = XFS_DIR3_FT_UNKNOWN;
+
+	error = xfs_dir_lookup(sdc->sc->tp, ip, &xname, &lookup_ino, NULL);
+	XFS_SCRUB_DIR_OP_ERROR_GOTO(fail_xref);
+	XFS_SCRUB_DIR_GOTO(lookup_ino == ino, out);
+
+	if (!memcmp(".", name, namelen)) {
+		/* If this is "." then check that the inum matches the dir. */
+		if (xfs_sb_version_hasftype(&mp->m_sb))
+			XFS_SCRUB_DIR_CHECK(type == DT_DIR);
+		XFS_SCRUB_DIR_CHECK(ino == ip->i_ino);
+	} else if (!memcmp("..", name, namelen)) {
+		/*
+		 * If this is ".." in the root inode, check that the inum
+		 * matches this dir.
+		 */
+		if (xfs_sb_version_hasftype(&mp->m_sb))
+			XFS_SCRUB_DIR_CHECK(type == DT_DIR);
+		if (ip->i_ino == mp->m_sb.sb_rootino)
+			XFS_SCRUB_DIR_CHECK(ino == ip->i_ino);
+	}
+	if (error)
+		goto out;
+
+	/* Verify the file type. */
+	error = xfs_scrub_dir_check_ftype(sdc, offset, lookup_ino, type);
+	if (error)
+		goto out;
+out:
+	return error;
+fail_xref:
+	return error ? error : -EFSCORRUPTED;
+}
+#undef XFS_SCRUB_DIR_OP_ERROR_GOTO
+#undef XFS_SCRUB_DIR_GOTO
+#undef XFS_SCRUB_DIR_CHECK
+
+#define XFS_SCRUB_DIRENT_CHECK(fs_ok) \
+	XFS_SCRUB_DATA_CHECK(ds->sc, XFS_DATA_FORK, rec_bno, "dir", fs_ok)
+#define XFS_SCRUB_DIRENT_GOTO(fs_ok, label) \
+	XFS_SCRUB_DATA_GOTO(ds->sc, XFS_DATA_FORK, rec_bno, "dir", fs_ok, label)
+#define XFS_SCRUB_DIRENT_OP_ERROR_GOTO(label) \
+	XFS_SCRUB_FILE_OP_ERROR_GOTO(ds->sc, XFS_DATA_FORK, rec_bno, "dir", &error, label)
+/* Scrub a directory btree record. */
+STATIC int
+xfs_scrub_dir_rec(
+	struct xfs_scrub_da_btree	*ds,
+	int				level,
+	void				*rec)
+{
+	struct xfs_mount		*mp = ds->state->mp;
+	struct xfs_dir2_leaf_entry	*ent = rec;
+	struct xfs_inode		*dp = ds->dargs.dp;
+	struct xfs_dir2_data_entry	*dent;
+	struct xfs_buf			*bp;
+	xfs_ino_t			ino;
+	xfs_dablk_t			rec_bno;
+	xfs_dir2_db_t			db;
+	xfs_dir2_data_aoff_t		off;
+	xfs_dir2_dataptr_t		ptr;
+	xfs_dahash_t			calc_hash;
+	xfs_dahash_t			hash;
+	unsigned int			tag;
+	int				error;
+
+	/* Check the hash of the entry. */
+	error = xfs_scrub_da_btree_hash(ds, level, &ent->hashval);
+	if (error)
+		goto out;
+
+	/* Valid hash pointer? */
+	ptr = be32_to_cpu(ent->address);
+	if (ptr == 0)
+		return 0;
+
+	/* Find the directory entry's location. */
+	db = xfs_dir2_dataptr_to_db(mp->m_dir_geo, ptr);
+	off = xfs_dir2_dataptr_to_off(mp->m_dir_geo, ptr);
+	rec_bno = xfs_dir2_db_to_da(mp->m_dir_geo, db);
+
+	XFS_SCRUB_DA_GOTO(ds, rec_bno < mp->m_dir_geo->leafblk, out);
+	error = xfs_dir3_data_read(ds->dargs.trans, dp, rec_bno, -2, &bp);
+	XFS_SCRUB_DIRENT_OP_ERROR_GOTO(out);
+	XFS_SCRUB_DIRENT_GOTO(bp != NULL, out);
+
+	/* Retrieve the entry and check it. */
+	dent = (struct xfs_dir2_data_entry *)(((char *)bp->b_addr) + off);
+	ino = be64_to_cpu(dent->inumber);
+	hash = be32_to_cpu(ent->hashval);
+	tag = be16_to_cpup(dp->d_ops->data_entry_tag_p(dent));
+	XFS_SCRUB_DIRENT_CHECK(xfs_dir_ino_validate(mp, ino) == 0);
+	XFS_SCRUB_DIRENT_CHECK(!xfs_internal_inum(mp, ino));
+	XFS_SCRUB_DIRENT_CHECK(tag == off);
+	XFS_SCRUB_DIRENT_GOTO(dent->namelen < MAXNAMELEN, out_relse);
+	calc_hash = xfs_da_hashname(dent->name, dent->namelen);
+	XFS_SCRUB_DIRENT_CHECK(calc_hash == hash);
+
+out_relse:
+	xfs_trans_brelse(ds->dargs.trans, bp);
+out:
+	return error;
+}
+#undef XFS_SCRUB_DIRENT_OP_ERROR_GOTO
+#undef XFS_SCRUB_DIRENT_GOTO
+#undef XFS_SCRUB_DIRENT_CHECK
+
+/* Scrub a whole directory. */
+int
+xfs_scrub_directory(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_scrub_dir_ctx	sdc = {
+		.dc.actor = xfs_scrub_dir_actor,
+		.dc.pos = 0,
+	};
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	size_t				bufsize;
+	loff_t				oldpos;
+	int				error;
+
+	if (!S_ISDIR(VFS_I(sc->ip)->i_mode))
+		return -ENOENT;
+
+	/* Check directory tree structure */
+	error = xfs_scrub_da_btree(sc, XFS_DATA_FORK, xfs_scrub_dir_rec);
+	if (error)
+		return error;
+
+	/* Check that every dirent we see can also be looked up by hash. */
+	bufsize = (size_t)min_t(loff_t, 32768, sc->ip->i_d.di_size);
+	sdc.sc = sc;
+
+	oldpos = 0;
+	xfs_iunlock(sc->ip, XFS_ILOCK_EXCL);
+	while (true) {
+		error = xfs_readdir(sc->tp, sc->ip, &sdc.dc, bufsize);
+		XFS_SCRUB_OP_ERROR_GOTO(sc,
+				XFS_INO_TO_AGNO(mp, sc->ip->i_ino),
+				XFS_INO_TO_AGBNO(mp, sc->ip->i_ino),
+				"inode", &error, out_unlock);
+		if (oldpos == sdc.dc.pos)
+			break;
+		oldpos = sdc.dc.pos;
+	}
+
+out_unlock:
+	xfs_ilock(sc->ip, XFS_ILOCK_EXCL);
+	return error;
+}
diff --git a/fs/xfs/xfs_dir2_readdir.c b/fs/xfs/xfs_dir2_readdir.c
index 003a99b..0b3b636 100644
--- a/fs/xfs/xfs_dir2_readdir.c
+++ b/fs/xfs/xfs_dir2_readdir.c
@@ -181,7 +181,7 @@ xfs_dir2_block_getdents(
 		return 0;
 
 	lock_mode = xfs_ilock_data_map_shared(dp);
-	error = xfs_dir3_block_read(NULL, dp, &bp);
+	error = xfs_dir3_block_read(args->trans, dp, &bp);
 	xfs_iunlock(dp, lock_mode);
 	if (error)
 		return error;
@@ -239,7 +239,7 @@ xfs_dir2_block_getdents(
 		if (!dir_emit(ctx, (char *)dep->name, dep->namelen,
 			    be64_to_cpu(dep->inumber),
 			    xfs_dir3_get_dtype(dp->i_mount, filetype))) {
-			xfs_trans_brelse(NULL, bp);
+			xfs_trans_brelse(args->trans, bp);
 			return 0;
 		}
 	}
@@ -250,7 +250,7 @@ xfs_dir2_block_getdents(
 	 */
 	ctx->pos = xfs_dir2_db_off_to_dataptr(geo, geo->datablk + 1, 0) &
 								0x7fffffff;
-	xfs_trans_brelse(NULL, bp);
+	xfs_trans_brelse(args->trans, bp);
 	return 0;
 }
 
@@ -386,7 +386,7 @@ xfs_dir2_leaf_readbuf(
 	 * Read the directory block starting at the first mapping.
 	 */
 	mip->curdb = xfs_dir2_da_to_db(geo, map->br_startoff);
-	error = xfs_dir3_data_read(NULL, dp, map->br_startoff,
+	error = xfs_dir3_data_read(args->trans, dp, map->br_startoff,
 			map->br_blockcount >= geo->fsbcount ?
 			    XFS_FSB_TO_DADDR(dp->i_mount, map->br_startblock) :
 			    -1, &bp);
@@ -535,7 +535,7 @@ xfs_dir2_leaf_getdents(
 			bool	trim_map = false;
 
 			if (bp) {
-				xfs_trans_brelse(NULL, bp);
+				xfs_trans_brelse(args->trans, bp);
 				bp = NULL;
 				trim_map = true;
 			}
@@ -649,15 +649,21 @@ xfs_dir2_leaf_getdents(
 		ctx->pos = xfs_dir2_byte_to_dataptr(curoff) & 0x7fffffff;
 	kmem_free(map_info);
 	if (bp)
-		xfs_trans_brelse(NULL, bp);
+		xfs_trans_brelse(args->trans, bp);
 	return error;
 }
 
 /*
  * Read a directory.
+ *
+ * If supplied, the transaction collects locked dir buffers to avoid
+ * nested buffer deadlocks.  This function does not dirty the
+ * transaction.  The caller should ensure that the inode is locked
+ * before calling this function.
  */
 int
 xfs_readdir(
+	struct xfs_trans	*tp,
 	struct xfs_inode	*dp,
 	struct dir_context	*ctx,
 	size_t			bufsize)
@@ -676,6 +682,7 @@ xfs_readdir(
 
 	args.dp = dp;
 	args.geo = dp->i_mount->m_dir_geo;
+	args.trans = tp;
 
 	if (dp->i_d.di_format == XFS_DINODE_FMT_LOCAL)
 		rval = xfs_dir2_sf_getdents(&args, ctx);
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 4c87e60f..8e1ab59d 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -944,7 +944,7 @@ xfs_file_readdir(
 	 */
 	bufsize = (size_t)min_t(loff_t, 32768, ip->i_d.di_size);
 
-	return xfs_readdir(ip, ctx, bufsize);
+	return xfs_readdir(NULL, ip, ctx, bufsize);
 }
 
 /*
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 141a39f..18b211f 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3367,7 +3367,8 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_INODE,		"inode" }, \
 	{ XFS_SCRUB_TYPE_BMBTD, 	"bmapbtd" }, \
 	{ XFS_SCRUB_TYPE_BMBTA,		"bmapbta" }, \
-	{ XFS_SCRUB_TYPE_BMBTC,		"bmapbtc" }
+	{ XFS_SCRUB_TYPE_BMBTC,		"bmapbtc" }, \
+	{ XFS_SCRUB_TYPE_DIR,		"dir" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 28/55] xfs: scrub directory freespace

[Darrick J. Wong]

Check the free space information in a directory.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_dir2_leaf.c |    2 
 fs/xfs/libxfs/xfs_dir2_priv.h |    2 
 fs/xfs/scrub/dir.c            |  318 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 321 insertions(+), 1 deletion(-)


diff --git a/fs/xfs/libxfs/xfs_dir2_leaf.c b/fs/xfs/libxfs/xfs_dir2_leaf.c
index b887fb2..e7ca764 100644
--- a/fs/xfs/libxfs/xfs_dir2_leaf.c
+++ b/fs/xfs/libxfs/xfs_dir2_leaf.c
@@ -256,7 +256,7 @@ const struct xfs_buf_ops xfs_dir3_leafn_buf_ops = {
 	.verify_write = xfs_dir3_leafn_write_verify,
 };
 
-static int
+int
 xfs_dir3_leaf_read(
 	struct xfs_trans	*tp,
 	struct xfs_inode	*dp,
diff --git a/fs/xfs/libxfs/xfs_dir2_priv.h b/fs/xfs/libxfs/xfs_dir2_priv.h
index 4cf2956..c586ade 100644
--- a/fs/xfs/libxfs/xfs_dir2_priv.h
+++ b/fs/xfs/libxfs/xfs_dir2_priv.h
@@ -58,6 +58,8 @@ extern int xfs_dir3_data_init(struct xfs_da_args *args, xfs_dir2_db_t blkno,
 		struct xfs_buf **bpp);
 
 /* xfs_dir2_leaf.c */
+extern int xfs_dir3_leaf_read(struct xfs_trans *tp, struct xfs_inode *dp,
+		xfs_dablk_t fbno, xfs_daddr_t mappedbno, struct xfs_buf **bpp);
 extern int xfs_dir3_leafn_read(struct xfs_trans *tp, struct xfs_inode *dp,
 		xfs_dablk_t fbno, xfs_daddr_t mappedbno, struct xfs_buf **bpp);
 extern int xfs_dir2_block_to_leaf(struct xfs_da_args *args,
diff --git a/fs/xfs/scrub/dir.c b/fs/xfs/scrub/dir.c
index ed29bd1..2bac798 100644
--- a/fs/xfs/scrub/dir.c
+++ b/fs/xfs/scrub/dir.c
@@ -223,6 +223,320 @@ xfs_scrub_dir_rec(
 #undef XFS_SCRUB_DIRENT_GOTO
 #undef XFS_SCRUB_DIRENT_CHECK
 
+#define XFS_SCRUB_DIR_BLOCK_CHECK(fs_ok) \
+	XFS_SCRUB_DATA_CHECK(sc, XFS_DATA_FORK, \
+		lblk << mp->m_sb.sb_blocklog, "dir", fs_ok)
+#define XFS_SCRUB_DIR_BLOCK_GOTO(fs_ok, label) \
+	XFS_SCRUB_DATA_GOTO(sc, XFS_DATA_FORK, \
+		lblk << mp->m_sb.sb_blocklog, "dir", fs_ok, label)
+#define XFS_SCRUB_DIR_BLOCK_OP_ERROR_GOTO(label) \
+	XFS_SCRUB_FILE_OP_ERROR_GOTO(sc, XFS_DATA_FORK, \
+		lblk << mp->m_sb.sb_blocklog, "dir", &error, label)
+/* Is this free entry either in the bestfree or smaller than all of them? */
+static inline bool
+xfs_scrub_directory_check_free_entry(
+	struct xfs_dir2_data_free	*bf,
+	struct xfs_dir2_data_unused	*dup)
+{
+	struct xfs_dir2_data_free	*dfp;
+	unsigned int			smallest;
+
+	smallest = -1U;
+	for (dfp = &bf[0]; dfp < &bf[XFS_DIR2_DATA_FD_COUNT]; dfp++) {
+		if (dfp->offset &&
+		    be16_to_cpu(dfp->length) == be16_to_cpu(dup->length))
+			return true;
+		if (smallest < be16_to_cpu(dfp->length))
+			smallest = be16_to_cpu(dfp->length);
+	}
+
+	return be16_to_cpu(dup->length) <= smallest;
+}
+
+/* Check free space info in a directory data block. */
+STATIC int
+xfs_scrub_directory_data_bestfree(
+	struct xfs_scrub_context	*sc,
+	xfs_dablk_t			lblk,
+	bool				is_block)
+{
+	struct xfs_dir2_data_unused	*dup;
+	struct xfs_dir2_data_free	*dfp;
+	struct xfs_buf			*bp;
+	struct xfs_dir2_data_free	*bf;
+	struct xfs_mount		*mp = sc->ip->i_mount;
+	char				*ptr;
+	char				*endptr;
+	u16				tag;
+	int				newlen;
+	int				offset;
+	int				error;
+
+	if (is_block) {
+		/* dir block format */
+		XFS_SCRUB_DIR_BLOCK_CHECK(lblk ==
+				XFS_B_TO_FSBT(mp, XFS_DIR2_DATA_OFFSET));
+		error = xfs_dir3_block_read(sc->tp, sc->ip, &bp);
+	} else {
+		/* dir data format */
+		error = xfs_dir3_data_read(sc->tp, sc->ip, lblk,
+				-1, &bp);
+	}
+	XFS_SCRUB_DIR_BLOCK_OP_ERROR_GOTO(out);
+
+	/* Do the bestfrees correspond to actual free space? */
+	bf = sc->ip->d_ops->data_bestfree_p(bp->b_addr);
+	for (dfp = &bf[0]; dfp < &bf[XFS_DIR2_DATA_FD_COUNT]; dfp++) {
+		offset = be16_to_cpu(dfp->offset);
+		XFS_SCRUB_DIR_BLOCK_GOTO(offset < BBTOB(bp->b_length), nextloop);
+		if (!offset)
+			continue;
+		dup = (struct xfs_dir2_data_unused *)(bp->b_addr + offset);
+		tag = be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup));
+
+		XFS_SCRUB_DIR_BLOCK_CHECK(dup->freetag ==
+				cpu_to_be16(XFS_DIR2_DATA_FREE_TAG));
+		XFS_SCRUB_DIR_BLOCK_CHECK(be16_to_cpu(dup->length) ==
+				be16_to_cpu(dfp->length));
+		XFS_SCRUB_DIR_BLOCK_CHECK(tag ==
+				((char *)dup - (char *)bp->b_addr));
+nextloop:;
+	}
+
+	/* Make sure the bestfrees are actually the best free spaces. */
+	ptr = (char *)sc->ip->d_ops->data_entry_p(bp->b_addr);
+	if (is_block) {
+		struct xfs_dir2_block_tail	*btp;
+
+		btp = xfs_dir2_block_tail_p(sc->ip->i_mount->m_dir_geo,
+				bp->b_addr);
+		endptr = (char *)xfs_dir2_block_leaf_p(btp);
+	} else
+		endptr = (char *)bp->b_addr + BBTOB(bp->b_length);
+	while (ptr < endptr) {
+		dup = (struct xfs_dir2_data_unused *)ptr;
+		/* Skip real entries */
+		if (dup->freetag != cpu_to_be16(XFS_DIR2_DATA_FREE_TAG)) {
+			struct xfs_dir2_data_entry	*dep;
+
+			dep = (struct xfs_dir2_data_entry *)ptr;
+			newlen = sc->ip->d_ops->data_entsize(dep->namelen);
+			XFS_SCRUB_DIR_BLOCK_GOTO(newlen > 0, out_buf);
+			ptr += newlen;
+			XFS_SCRUB_DIR_BLOCK_CHECK(ptr <= endptr);
+			continue;
+		}
+
+		/* Spot check this free entry */
+		tag = be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup));
+		XFS_SCRUB_DIR_BLOCK_CHECK(tag ==
+				((char *)dup - (char *)bp->b_addr));
+
+		/*
+		 * Either this entry is a bestfree or it's smaller than
+		 * any of the bestfrees.
+		 */
+		XFS_SCRUB_DIR_BLOCK_CHECK(
+				xfs_scrub_directory_check_free_entry(bf, dup));
+
+		/* Move on. */
+		newlen = be16_to_cpu(dup->length);
+		XFS_SCRUB_DIR_BLOCK_GOTO(newlen > 0, out_buf);
+		ptr += newlen;
+		XFS_SCRUB_DIR_BLOCK_CHECK(ptr <= endptr);
+	}
+out_buf:
+	xfs_trans_brelse(sc->tp, bp);
+out:
+	return error;
+}
+
+/* Is this the longest free entry in the block? */
+static inline bool
+xfs_scrub_directory_check_freesp(
+	struct xfs_inode		*dp,
+	struct xfs_buf			*dbp,
+	unsigned int			len)
+{
+	struct xfs_dir2_data_free	*bf;
+	struct xfs_dir2_data_free	*dfp;
+	unsigned int			longest = 0;
+	int				offset;
+
+	bf = dp->d_ops->data_bestfree_p(dbp->b_addr);
+	for (dfp = &bf[0]; dfp < &bf[XFS_DIR2_DATA_FD_COUNT]; dfp++) {
+		offset = be16_to_cpu(dfp->offset);
+		if (!offset)
+			continue;
+		if (longest < be16_to_cpu(dfp->length))
+			longest = be16_to_cpu(dfp->length);
+	}
+
+	return longest == len;
+}
+
+/* Check free space info in a directory leaf1 block. */
+STATIC int
+xfs_scrub_directory_leaf1_bestfree(
+	struct xfs_scrub_context	*sc,
+	xfs_dablk_t			lblk)
+{
+	struct xfs_dir2_leaf_tail	*ltp;
+	struct xfs_buf			*dbp;
+	struct xfs_buf			*bp;
+	struct xfs_mount		*mp = sc->ip->i_mount;
+	__be16				*plen;
+	int				i;
+	int				error;
+
+	/* Read the free space block */
+	error = xfs_dir3_leaf_read(sc->tp, sc->ip, lblk, -1, &bp);
+	XFS_SCRUB_DIR_BLOCK_OP_ERROR_GOTO(out);
+
+	/* Check all the entries. */
+	ltp = xfs_dir2_leaf_tail_p(sc->ip->i_mount->m_dir_geo, bp->b_addr);
+	plen = xfs_dir2_leaf_bests_p(ltp);
+	for (i = 0; i < be32_to_cpu(ltp->bestcount); i++, plen++) {
+		error = xfs_dir3_data_read(sc->tp, sc->ip, i, -1, &dbp);
+		XFS_SCRUB_DIR_BLOCK_OP_ERROR_GOTO(out_buf);
+		XFS_SCRUB_DIR_BLOCK_CHECK(
+				xfs_scrub_directory_check_freesp(sc->ip, dbp,
+					be16_to_cpu(*plen)));
+		xfs_trans_brelse(sc->tp, dbp);
+	}
+out_buf:
+	xfs_trans_brelse(sc->tp, bp);
+out:
+	return error;
+}
+
+/* Check free space info in a directory freespace block. */
+STATIC int
+xfs_scrub_directory_free_bestfree(
+	struct xfs_scrub_context	*sc,
+	xfs_dablk_t			lblk)
+{
+	struct xfs_dir3_icfree_hdr	freehdr;
+	struct xfs_buf			*dbp;
+	struct xfs_buf			*bp;
+	struct xfs_mount		*mp = sc->ip->i_mount;
+	__be16				*plen;
+	int				i;
+	int				error;
+
+	/* Read the free space block */
+	error = xfs_dir2_free_read(sc->tp, sc->ip, lblk, &bp);
+	XFS_SCRUB_DIR_BLOCK_OP_ERROR_GOTO(out);
+
+	/* Check all the entries. */
+	sc->ip->d_ops->free_hdr_from_disk(&freehdr, bp->b_addr);
+	plen = sc->ip->d_ops->free_bests_p(bp->b_addr);
+	for (i = 0; i < freehdr.nvalid; i++, plen++) {
+		error = xfs_dir3_data_read(sc->tp, sc->ip, freehdr.firstdb + i,
+				-1, &dbp);
+		XFS_SCRUB_DIR_BLOCK_OP_ERROR_GOTO(out_buf);
+		XFS_SCRUB_DIR_BLOCK_CHECK(
+				xfs_scrub_directory_check_freesp(sc->ip, dbp,
+					be16_to_cpu(*plen)));
+		xfs_trans_brelse(sc->tp, dbp);
+	}
+out_buf:
+	xfs_trans_brelse(sc->tp, bp);
+out:
+	return error;
+}
+
+/* Check free space information in directories. */
+STATIC int
+xfs_scrub_directory_blocks(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_bmbt_irec		got;
+	struct xfs_da_args		args;
+	struct xfs_ifork		*ifp;
+	struct xfs_mount		*mp = sc->ip->i_mount;
+	xfs_fileoff_t			leaf_lblk;
+	xfs_fileoff_t			free_lblk;
+	xfs_fileoff_t			lblk;
+	xfs_extnum_t			idx;
+	bool				found;
+	int				is_block = 0;
+	int				error;
+
+	/* Ignore local format directories. */
+	if (sc->ip->i_d.di_format != XFS_DINODE_FMT_EXTENTS &&
+	    sc->ip->i_d.di_format != XFS_DINODE_FMT_BTREE)
+		return 0;
+
+	/* Is this a block dir? */
+	lblk = 0;
+	args.dp = sc->ip;
+	args.geo = sc->ip->i_mount->m_dir_geo;
+	args.trans = sc->tp;
+	error = xfs_dir2_isblock(&args, &is_block);
+	XFS_SCRUB_DIR_BLOCK_OP_ERROR_GOTO(out);
+
+	/* Iterate all the data extents in the directory... */
+	ifp = XFS_IFORK_PTR(sc->ip, XFS_DATA_FORK);
+	leaf_lblk = XFS_DIR2_LEAF_OFFSET >> mp->m_sb.sb_blocklog;
+	free_lblk = XFS_DIR2_FREE_OFFSET >> mp->m_sb.sb_blocklog;
+	found = xfs_iext_lookup_extent(sc->ip, ifp, lblk, &idx, &got);
+	while (found) {
+		/* No more data blocks... */
+		lblk = got.br_startoff;
+		if (lblk >= leaf_lblk)
+			break;
+
+		for (lblk = got.br_startoff;
+		     lblk < got.br_startoff + got.br_blockcount;
+		     lblk += args.geo->fsbcount) {
+			/* Check each data block's bestfree data */
+			error = xfs_scrub_directory_data_bestfree(sc, lblk,
+					is_block);
+			if (error)
+				goto out;
+		}
+		found = xfs_iext_get_extent(ifp, ++idx, &got);
+	}
+
+	/* Look for a leaf1 block, which has free info. */
+	if (xfs_iext_lookup_extent(sc->ip, ifp, leaf_lblk, &idx, &got) &&
+	    got.br_startoff == leaf_lblk &&
+	    got.br_blockcount == args.geo->fsbcount &&
+	    !xfs_iext_get_extent(ifp, ++idx, &got)) {
+		XFS_SCRUB_DIR_BLOCK_GOTO(!is_block, not_leaf1);
+		error = xfs_scrub_directory_leaf1_bestfree(sc, lblk);
+		if (error)
+			goto out;
+	}
+not_leaf1:
+
+	/* Scan for free blocks */
+	lblk = free_lblk;
+	found = xfs_iext_lookup_extent(sc->ip, ifp, lblk, &idx, &got);
+	while (found) {
+		lblk = got.br_startoff;
+
+		XFS_SCRUB_DIR_BLOCK_GOTO(!is_block, nextfree);
+
+		/* For each dir free block in the extent... */
+		for (lblk = got.br_startoff;
+		     lblk < got.br_startoff + got.br_blockcount;
+		     lblk += args.geo->fsbcount) {
+			/* Check each data block's bestfree data */
+			error = xfs_scrub_directory_free_bestfree(sc, lblk);
+			if (error)
+				goto out;
+		}
+nextfree:
+		found = xfs_iext_get_extent(ifp, ++idx, &got);
+	}
+out:
+	return error;
+}
+#undef XFS_SCRUB_DIR_BLOCK_OP_ERROR_GOTO
+#undef XFS_SCRUB_DIR_BLOCK_CHECK
+
 /* Scrub a whole directory. */
 int
 xfs_scrub_directory(
@@ -264,5 +578,9 @@ xfs_scrub_directory(
 
 out_unlock:
 	xfs_ilock(sc->ip, XFS_ILOCK_EXCL);
+
+	/* Check the freespace. */
+	error = xfs_scrub_directory_blocks(sc);
+
 	return error;
 }

[PATCH 29/55] xfs: scrub extended attributes

[Darrick J. Wong]

Scrub the hash tree, keys, and values in an extended attribute structure.
Refactor the attribute code to use the transaction if the caller supplied
one to avoid buffer deadocks.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile                 |    1 
 fs/xfs/libxfs/xfs_attr.c        |   26 +++--
 fs/xfs/libxfs/xfs_attr_remote.c |    5 +
 fs/xfs/libxfs/xfs_fs.h          |    3 -
 fs/xfs/scrub/attr.c             |  216 +++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.c           |    5 +
 fs/xfs/scrub/common.h           |    6 +
 fs/xfs/xfs_attr.h               |    2 
 fs/xfs/xfs_attr_list.c          |   28 +++--
 fs/xfs/xfs_trace.h              |    3 -
 10 files changed, 269 insertions(+), 26 deletions(-)
 create mode 100644 fs/xfs/scrub/attr.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index c05f503..d723ebc 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -106,6 +106,7 @@ xfs-y				+= xfs_aops.o \
 xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   agheader.o \
 				   alloc.o \
+				   attr.o \
 				   bmap.o \
 				   btree.o \
 				   common.o \
diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index af1ecb1..b4e1686 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -114,6 +114,23 @@ xfs_inode_hasattr(
  * Overall external interface routines.
  *========================================================================*/
 
+/* Retrieve an extended attribute and its value.  Must have iolock. */
+int
+xfs_attr_get_locked(
+	struct xfs_inode	*ip,
+	struct xfs_da_args	*args)
+{
+	if (!xfs_inode_hasattr(ip))
+		return -ENOATTR;
+	else if (ip->i_d.di_aformat == XFS_DINODE_FMT_LOCAL)
+		return xfs_attr_shortform_getvalue(args);
+	else if (xfs_bmap_one_block(ip, XFS_ATTR_FORK))
+		return xfs_attr_leaf_get(args);
+	else
+		return xfs_attr_node_get(args);
+}
+
+/* Retrieve an extended attribute by name, and its value. */
 int
 xfs_attr_get(
 	struct xfs_inode	*ip,
@@ -144,14 +161,7 @@ xfs_attr_get(
 	args.op_flags = XFS_DA_OP_OKNOENT;
 
 	lock_mode = xfs_ilock_attr_map_shared(ip);
-	if (!xfs_inode_hasattr(ip))
-		error = -ENOATTR;
-	else if (ip->i_d.di_aformat == XFS_DINODE_FMT_LOCAL)
-		error = xfs_attr_shortform_getvalue(&args);
-	else if (xfs_bmap_one_block(ip, XFS_ATTR_FORK))
-		error = xfs_attr_leaf_get(&args);
-	else
-		error = xfs_attr_node_get(&args);
+	error = xfs_attr_get_locked(ip, &args);
 	xfs_iunlock(ip, lock_mode);
 
 	*valuelenp = args.valuelen;
diff --git a/fs/xfs/libxfs/xfs_attr_remote.c b/fs/xfs/libxfs/xfs_attr_remote.c
index d52f525..76958b4 100644
--- a/fs/xfs/libxfs/xfs_attr_remote.c
+++ b/fs/xfs/libxfs/xfs_attr_remote.c
@@ -386,7 +386,8 @@ xfs_attr_rmtval_get(
 			       (map[i].br_startblock != HOLESTARTBLOCK));
 			dblkno = XFS_FSB_TO_DADDR(mp, map[i].br_startblock);
 			dblkcnt = XFS_FSB_TO_BB(mp, map[i].br_blockcount);
-			error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
+			error = xfs_trans_read_buf(mp, args->trans,
+						   mp->m_ddev_targp,
 						   dblkno, dblkcnt, 0, &bp,
 						   &xfs_attr3_rmt_buf_ops);
 			if (error)
@@ -395,7 +396,7 @@ xfs_attr_rmtval_get(
 			error = xfs_attr_rmtval_copyout(mp, bp, args->dp->i_ino,
 							&offset, &valuelen,
 							&dst);
-			xfs_buf_relse(bp);
+			xfs_trans_brelse(args->trans, bp);
 			if (error)
 				return error;
 
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 2a9b959..6d5bfbd 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -590,7 +590,8 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_BMBTA	13	/* attr fork block mapping */
 #define XFS_SCRUB_TYPE_BMBTC	14	/* CoW fork block mapping */
 #define XFS_SCRUB_TYPE_DIR	15	/* directory */
-#define XFS_SCRUB_TYPE_MAX	15
+#define XFS_SCRUB_TYPE_XATTR	16	/* extended attribute */
+#define XFS_SCRUB_TYPE_MAX	16
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/scrub/attr.c b/fs/xfs/scrub/attr.c
new file mode 100644
index 0000000..a81ece0
--- /dev/null
+++ b/fs/xfs/scrub/attr.c
@@ -0,0 +1,216 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_da_format.h"
+#include "xfs_da_btree.h"
+#include "xfs_dir2.h"
+#include "xfs_attr.h"
+#include "xfs_attr_leaf.h"
+#include "scrub/common.h"
+#include "scrub/dabtree.h"
+
+#include <linux/posix_acl_xattr.h>
+#include <linux/xattr.h>
+
+/* Set us up with an inode and a buffer for reading xattr values. */
+int
+xfs_scrub_setup_inode_xattr(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	void				*buf;
+	int				error;
+
+	/* Allocate the buffer without the inode lock held. */
+	buf = kmem_zalloc_large(XATTR_SIZE_MAX, KM_SLEEP);
+	if (!buf)
+		return -ENOMEM;
+
+	error = xfs_scrub_setup_inode(sc, ip, sm, retry_deadlocked);
+	if (error) {
+		kmem_free(buf);
+		return error;
+	}
+
+	sc->buf = buf;
+	return 0;
+}
+
+/* Extended Attributes */
+
+struct xfs_scrub_xattr {
+	struct xfs_attr_list_context	context;
+	struct xfs_scrub_context	*sc;
+};
+
+#define XFS_SCRUB_ATTR_CHECK(fs_ok) \
+	XFS_SCRUB_DATA_CHECK(sx->sc, XFS_ATTR_FORK, args.blkno, "attr", fs_ok)
+#define XFS_SCRUB_ATTR_OP_ERROR_GOTO(label) \
+	XFS_SCRUB_FILE_OP_ERROR_GOTO(sx->sc, XFS_ATTR_FORK, args.blkno, "attr", &error, label)
+/* Check that an extended attribute key can be looked up by hash. */
+static void
+xfs_scrub_xattr_listent(
+	struct xfs_attr_list_context	*context,
+	int				flags,
+	unsigned char			*name,
+	int				namelen,
+	int				valuelen)
+{
+	struct xfs_scrub_xattr		*sx;
+	struct xfs_da_args		args = {0};
+	int				error = 0;
+
+	sx = container_of(context, struct xfs_scrub_xattr, context);
+
+	args.flags = ATTR_KERNOTIME;
+	if (flags & XFS_ATTR_ROOT)
+		args.flags |= ATTR_ROOT;
+	else if (flags & XFS_ATTR_SECURE)
+		args.flags |= ATTR_SECURE;
+	args.geo = context->dp->i_mount->m_attr_geo;
+	args.whichfork = XFS_ATTR_FORK;
+	args.dp = context->dp;
+	args.name = name;
+	args.namelen = namelen;
+	args.hashval = xfs_da_hashname(args.name, args.namelen);
+	args.trans = context->tp;
+	args.value = sx->sc->buf;
+	args.valuelen = XATTR_SIZE_MAX;
+
+	error = xfs_attr_get_locked(context->dp, &args);
+	if (error == -EEXIST)
+		error = 0;
+	XFS_SCRUB_ATTR_OP_ERROR_GOTO(fail_xref);
+	XFS_SCRUB_ATTR_CHECK(args.valuelen == valuelen);
+
+fail_xref:
+	return;
+}
+#undef XFS_SCRUB_ATTR_OP_ERROR_GOTO
+#undef XFS_SCRUB_ATTR_CHECK
+
+/* Scrub a attribute btree record. */
+STATIC int
+xfs_scrub_xattr_rec(
+	struct xfs_scrub_da_btree	*ds,
+	int				level,
+	void				*rec)
+{
+	struct xfs_mount		*mp = ds->state->mp;
+	struct xfs_attr_leaf_entry	*ent = rec;
+	struct xfs_da_state_blk		*blk;
+	struct xfs_attr_leaf_name_local	*lentry;
+	struct xfs_attr_leaf_name_remote	*rentry;
+	struct xfs_buf			*bp;
+	xfs_dahash_t			calc_hash;
+	xfs_dahash_t			hash;
+	int				nameidx;
+	int				hdrsize;
+	unsigned int			badflags;
+	int				error;
+
+	blk = &ds->state->path.blk[level];
+
+	/* Check the hash of the entry. */
+	error = xfs_scrub_da_btree_hash(ds, level, &ent->hashval);
+	if (error)
+		goto out;
+
+	/* Find the attr entry's location. */
+	bp = blk->bp;
+	hdrsize = xfs_attr3_leaf_hdr_size(bp->b_addr);
+	nameidx = be16_to_cpu(ent->nameidx);
+	XFS_SCRUB_DA_GOTO(ds, nameidx >= hdrsize, out);
+	XFS_SCRUB_DA_GOTO(ds, nameidx < mp->m_attr_geo->blksize, out);
+
+	/* Retrieve the entry and check it. */
+	hash = be32_to_cpu(ent->hashval);
+	badflags = ~(XFS_ATTR_LOCAL | XFS_ATTR_ROOT | XFS_ATTR_SECURE |
+			XFS_ATTR_INCOMPLETE);
+	XFS_SCRUB_DA_CHECK(ds, (ent->flags & badflags) == 0);
+	if (ent->flags & XFS_ATTR_LOCAL) {
+		lentry = (struct xfs_attr_leaf_name_local *)
+				(((char *)bp->b_addr) + nameidx);
+		XFS_SCRUB_DA_GOTO(ds, lentry->namelen < MAXNAMELEN, out);
+		calc_hash = xfs_da_hashname(lentry->nameval, lentry->namelen);
+	} else {
+		rentry = (struct xfs_attr_leaf_name_remote *)
+				(((char *)bp->b_addr) + nameidx);
+		XFS_SCRUB_DA_GOTO(ds, rentry->namelen < MAXNAMELEN, out);
+		calc_hash = xfs_da_hashname(rentry->name, rentry->namelen);
+	}
+	XFS_SCRUB_DA_CHECK(ds, calc_hash == hash);
+
+out:
+	return error;
+}
+
+/* Scrub the extended attribute metadata. */
+int
+xfs_scrub_xattr(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_scrub_xattr		sx = { 0 };
+	struct attrlist_cursor_kern	cursor = { 0 };
+	struct xfs_mount		*mp = sc->ip->i_mount;
+	int				error = 0;
+
+	if (!xfs_inode_hasattr(sc->ip))
+		return -ENOENT;
+
+	/* Check attribute tree structure */
+	error = xfs_scrub_da_btree(sc, XFS_ATTR_FORK, xfs_scrub_xattr_rec);
+	if (error)
+		goto out;
+
+	/* Check that every attr key can also be looked up by hash. */
+	sx.context.dp = sc->ip;
+	sx.context.cursor = &cursor;
+	sx.context.resynch = 1;
+	sx.context.put_listent = xfs_scrub_xattr_listent;
+	sx.context.tp = sc->tp;
+	sx.sc = sc;
+
+	xfs_iunlock(sc->ip, XFS_ILOCK_EXCL);
+	error = xfs_attr_list_int(&sx.context);
+	xfs_ilock(sc->ip, XFS_ILOCK_EXCL);
+
+	XFS_SCRUB_OP_ERROR_GOTO(sc,
+			XFS_INO_TO_AGNO(mp, sc->ip->i_ino),
+			XFS_INO_TO_AGBNO(mp, sc->ip->i_ino),
+			"inode", &error, out);
+out:
+	return error;
+}
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index a570d64..b68dd74 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -576,6 +576,10 @@ xfs_scrub_teardown(
 			IRELE(sc->ip);
 		sc->ip = NULL;
 	}
+	if (sc->buf) {
+		kmem_free(sc->buf);
+		sc->buf = NULL;
+	}
 	return error;
 }
 
@@ -684,6 +688,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_attr, NULL, NULL},
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_cow, NULL, NULL},
 	{xfs_scrub_setup_inode, xfs_scrub_directory, NULL, NULL},
+	{xfs_scrub_setup_inode_xattr, xfs_scrub_xattr, NULL, NULL},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 080596b..0f3ffd7 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -59,6 +59,7 @@ struct xfs_scrub_context {
 	struct xfs_scrub_metadata	*sm;
 	struct xfs_trans		*tp;
 	struct xfs_inode		*ip;
+	void				*buf;
 	bool				retry;
 
 	/* State tracking for multi-AG operations. */
@@ -226,6 +227,10 @@ int xfs_scrub_setup_inode_bmap(struct xfs_scrub_context *sc,
 			       struct xfs_inode *ip,
 			       struct xfs_scrub_metadata *sm,
 			       bool retry_deadlocked);
+int xfs_scrub_setup_inode_xattr(struct xfs_scrub_context *sc,
+				struct xfs_inode *ip,
+				struct xfs_scrub_metadata *sm,
+				bool retry_deadlocked);
 
 /* Metadata scrubbers */
 
@@ -244,5 +249,6 @@ int xfs_scrub_bmap_data(struct xfs_scrub_context *sc);
 int xfs_scrub_bmap_attr(struct xfs_scrub_context *sc);
 int xfs_scrub_bmap_cow(struct xfs_scrub_context *sc);
 int xfs_scrub_directory(struct xfs_scrub_context *sc);
+int xfs_scrub_xattr(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/xfs_attr.h b/fs/xfs/xfs_attr.h
index d14691a..24093f4 100644
--- a/fs/xfs/xfs_attr.h
+++ b/fs/xfs/xfs_attr.h
@@ -117,6 +117,7 @@ typedef void (*put_listent_func_t)(struct xfs_attr_list_context *, int,
 			      unsigned char *, int, int);
 
 typedef struct xfs_attr_list_context {
+	struct xfs_trans		*tp;
 	struct xfs_inode		*dp;		/* inode */
 	struct attrlist_cursor_kern	*cursor;	/* position in list */
 	char				*alist;		/* output buffer */
@@ -142,6 +143,7 @@ typedef struct xfs_attr_list_context {
 int xfs_attr_inactive(struct xfs_inode *dp);
 int xfs_attr_list_int(struct xfs_attr_list_context *);
 int xfs_inode_hasattr(struct xfs_inode *ip);
+int xfs_attr_get_locked(struct xfs_inode *ip, struct xfs_da_args *args);
 int xfs_attr_get(struct xfs_inode *ip, const unsigned char *name,
 		 unsigned char *value, int *valuelenp, int flags);
 int xfs_attr_set(struct xfs_inode *dp, const unsigned char *name,
diff --git a/fs/xfs/xfs_attr_list.c b/fs/xfs/xfs_attr_list.c
index 97c45b6..42bd26d 100644
--- a/fs/xfs/xfs_attr_list.c
+++ b/fs/xfs/xfs_attr_list.c
@@ -230,7 +230,7 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 	 */
 	bp = NULL;
 	if (cursor->blkno > 0) {
-		error = xfs_da3_node_read(NULL, dp, cursor->blkno, -1,
+		error = xfs_da3_node_read(context->tp, dp, cursor->blkno, -1,
 					      &bp, XFS_ATTR_FORK);
 		if ((error != 0) && (error != -EFSCORRUPTED))
 			return error;
@@ -242,7 +242,7 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 			case XFS_DA_NODE_MAGIC:
 			case XFS_DA3_NODE_MAGIC:
 				trace_xfs_attr_list_wrong_blk(context);
-				xfs_trans_brelse(NULL, bp);
+				xfs_trans_brelse(context->tp, bp);
 				bp = NULL;
 				break;
 			case XFS_ATTR_LEAF_MAGIC:
@@ -254,18 +254,18 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 				if (cursor->hashval > be32_to_cpu(
 						entries[leafhdr.count - 1].hashval)) {
 					trace_xfs_attr_list_wrong_blk(context);
-					xfs_trans_brelse(NULL, bp);
+					xfs_trans_brelse(context->tp, bp);
 					bp = NULL;
 				} else if (cursor->hashval <= be32_to_cpu(
 						entries[0].hashval)) {
 					trace_xfs_attr_list_wrong_blk(context);
-					xfs_trans_brelse(NULL, bp);
+					xfs_trans_brelse(context->tp, bp);
 					bp = NULL;
 				}
 				break;
 			default:
 				trace_xfs_attr_list_wrong_blk(context);
-				xfs_trans_brelse(NULL, bp);
+				xfs_trans_brelse(context->tp, bp);
 				bp = NULL;
 			}
 		}
@@ -281,7 +281,7 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 		for (;;) {
 			__uint16_t magic;
 
-			error = xfs_da3_node_read(NULL, dp,
+			error = xfs_da3_node_read(context->tp, dp,
 						      cursor->blkno, -1, &bp,
 						      XFS_ATTR_FORK);
 			if (error)
@@ -297,7 +297,7 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 						     XFS_ERRLEVEL_LOW,
 						     context->dp->i_mount,
 						     node);
-				xfs_trans_brelse(NULL, bp);
+				xfs_trans_brelse(context->tp, bp);
 				return -EFSCORRUPTED;
 			}
 
@@ -313,10 +313,10 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 				}
 			}
 			if (i == nodehdr.count) {
-				xfs_trans_brelse(NULL, bp);
+				xfs_trans_brelse(context->tp, bp);
 				return 0;
 			}
-			xfs_trans_brelse(NULL, bp);
+			xfs_trans_brelse(context->tp, bp);
 		}
 	}
 	ASSERT(bp != NULL);
@@ -333,12 +333,12 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 		if (context->seen_enough || leafhdr.forw == 0)
 			break;
 		cursor->blkno = leafhdr.forw;
-		xfs_trans_brelse(NULL, bp);
-		error = xfs_attr3_leaf_read(NULL, dp, cursor->blkno, -1, &bp);
+		xfs_trans_brelse(context->tp, bp);
+		error = xfs_attr3_leaf_read(context->tp, dp, cursor->blkno, -1, &bp);
 		if (error)
 			return error;
 	}
-	xfs_trans_brelse(NULL, bp);
+	xfs_trans_brelse(context->tp, bp);
 	return 0;
 }
 
@@ -448,12 +448,12 @@ xfs_attr_leaf_list(xfs_attr_list_context_t *context)
 	trace_xfs_attr_leaf_list(context);
 
 	context->cursor->blkno = 0;
-	error = xfs_attr3_leaf_read(NULL, context->dp, 0, -1, &bp);
+	error = xfs_attr3_leaf_read(context->tp, context->dp, 0, -1, &bp);
 	if (error)
 		return error;
 
 	xfs_attr3_leaf_list_int(bp, context);
-	xfs_trans_brelse(NULL, bp);
+	xfs_trans_brelse(context->tp, bp);
 	return 0;
 }
 
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 18b211f..760552d 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3368,7 +3368,8 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_BMBTD, 	"bmapbtd" }, \
 	{ XFS_SCRUB_TYPE_BMBTA,		"bmapbta" }, \
 	{ XFS_SCRUB_TYPE_BMBTC,		"bmapbtc" }, \
-	{ XFS_SCRUB_TYPE_DIR,		"dir" }
+	{ XFS_SCRUB_TYPE_DIR,		"dir" }, \
+	{ XFS_SCRUB_TYPE_XATTR,		"xattr" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 30/55] xfs: scrub symbolic links

[Darrick J. Wong]

Create the infrastructure to scrub symbolic link data.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile        |    1 
 fs/xfs/libxfs/xfs_fs.h |    3 +
 fs/xfs/scrub/common.c  |    1 
 fs/xfs/scrub/common.h  |    5 ++
 fs/xfs/scrub/symlink.c |  107 ++++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_trace.h     |    3 +
 6 files changed, 118 insertions(+), 2 deletions(-)
 create mode 100644 fs/xfs/scrub/symlink.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index d723ebc..5035ee0 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -116,6 +116,7 @@ xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   inode.o \
 				   refcount.o \
 				   rmap.o \
+				   symlink.o \
 				   )
 
 # low-level transaction/log code
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 6d5bfbd..5484c26 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -591,7 +591,8 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_BMBTC	14	/* CoW fork block mapping */
 #define XFS_SCRUB_TYPE_DIR	15	/* directory */
 #define XFS_SCRUB_TYPE_XATTR	16	/* extended attribute */
-#define XFS_SCRUB_TYPE_MAX	16
+#define XFS_SCRUB_TYPE_SYMLINK	17	/* symbolic link */
+#define XFS_SCRUB_TYPE_MAX	17
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index b68dd74..0b55933 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -689,6 +689,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_cow, NULL, NULL},
 	{xfs_scrub_setup_inode, xfs_scrub_directory, NULL, NULL},
 	{xfs_scrub_setup_inode_xattr, xfs_scrub_xattr, NULL, NULL},
+	{xfs_scrub_setup_inode_symlink, xfs_scrub_symlink, NULL, NULL},
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 0f3ffd7..123dc1d 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -231,6 +231,10 @@ int xfs_scrub_setup_inode_xattr(struct xfs_scrub_context *sc,
 				struct xfs_inode *ip,
 				struct xfs_scrub_metadata *sm,
 				bool retry_deadlocked);
+int xfs_scrub_setup_inode_symlink(struct xfs_scrub_context *sc,
+				  struct xfs_inode *ip,
+				  struct xfs_scrub_metadata *sm,
+				  bool retry_deadlocked);
 
 /* Metadata scrubbers */
 
@@ -250,5 +254,6 @@ int xfs_scrub_bmap_attr(struct xfs_scrub_context *sc);
 int xfs_scrub_bmap_cow(struct xfs_scrub_context *sc);
 int xfs_scrub_directory(struct xfs_scrub_context *sc);
 int xfs_scrub_xattr(struct xfs_scrub_context *sc);
+int xfs_scrub_symlink(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/symlink.c b/fs/xfs/scrub/symlink.c
new file mode 100644
index 0000000..80a1e70
--- /dev/null
+++ b/fs/xfs/scrub/symlink.c
@@ -0,0 +1,107 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_inode_fork.h"
+#include "xfs_symlink.h"
+#include "scrub/common.h"
+
+/* Set us up with an inode and a buffer for reading symlink targets. */
+int
+xfs_scrub_setup_inode_symlink(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	void				*buf;
+	int				error;
+
+	/* Allocate the buffer without the inode lock held. */
+	buf = kmem_zalloc_large(MAXPATHLEN + 1, KM_SLEEP);
+	if (!buf)
+		return -ENOMEM;
+
+	error = xfs_scrub_setup_inode(sc, ip, sm, retry_deadlocked);
+	if (error) {
+		kmem_free(buf);
+		return error;
+	}
+
+	sc->buf = buf;
+	return 0;
+}
+
+/* Symbolic links. */
+
+#define XFS_SCRUB_SYMLINK_CHECK(fs_ok) \
+	XFS_SCRUB_INO_CHECK(sc, ip->i_ino, NULL, "symlink", fs_ok)
+#define XFS_SCRUB_SYMLINK_GOTO(fs_ok, label) \
+	XFS_SCRUB_INO_GOTO(sc, ip->i_ino, NULL, "symlink", fs_ok, label)
+int
+xfs_scrub_symlink(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_inode		*ip = sc->ip;
+	struct xfs_ifork		*ifp;
+	loff_t				len;
+	int				error = 0;
+
+	if (!S_ISLNK(VFS_I(ip)->i_mode))
+		return -ENOENT;
+	ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
+	len = ip->i_d.di_size;
+
+	/* Plausible size? */
+	XFS_SCRUB_SYMLINK_GOTO(len <= MAXPATHLEN, out);
+
+	/* Inline symlink? */
+	if (ifp->if_flags & XFS_IFINLINE) {
+		XFS_SCRUB_SYMLINK_GOTO(len > 0, out);
+		XFS_SCRUB_SYMLINK_CHECK(len <= XFS_IFORK_DSIZE(ip));
+		XFS_SCRUB_SYMLINK_CHECK(len <= strnlen(ifp->if_u1.if_data,
+				XFS_IFORK_DSIZE(ip)));
+		goto out;
+	}
+
+	/* Remote symlink; must read. */
+	xfs_iunlock(sc->ip, XFS_ILOCK_EXCL);
+	error = xfs_readlink(sc->ip, sc->buf);
+	xfs_ilock(sc->ip, XFS_ILOCK_EXCL);
+	XFS_SCRUB_FILE_OP_ERROR_GOTO(sc, XFS_DATA_FORK, 0, "symlink",
+			&error, out);
+	XFS_SCRUB_SYMLINK_CHECK(len <= strnlen(sc->buf, MAXPATHLEN));
+out:
+	return error;
+}
+#undef XFS_SCRUB_SYMLINK_GOTO
+#undef XFS_SCRUB_SYMLINK_CHECK
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 760552d..87fd942 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3369,7 +3369,8 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_BMBTA,		"bmapbta" }, \
 	{ XFS_SCRUB_TYPE_BMBTC,		"bmapbtc" }, \
 	{ XFS_SCRUB_TYPE_DIR,		"dir" }, \
-	{ XFS_SCRUB_TYPE_XATTR,		"xattr" }
+	{ XFS_SCRUB_TYPE_XATTR,		"xattr" }, \
+	{ XFS_SCRUB_TYPE_SYMLINK,	"symlink" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 31/55] xfs: scrub realtime bitmap/summary

[Darrick J. Wong]

Perform simple tests of the realtime bitmap and summary.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile              |    5 ++
 fs/xfs/libxfs/xfs_format.h   |    5 ++
 fs/xfs/libxfs/xfs_fs.h       |    4 +
 fs/xfs/libxfs/xfs_rtbitmap.c |    2 -
 fs/xfs/scrub/agheader.c      |    1 
 fs/xfs/scrub/common.c        |    7 ++
 fs/xfs/scrub/common.h        |    6 ++
 fs/xfs/scrub/rtbitmap.c      |  131 ++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_rtalloc.h         |    3 +
 fs/xfs/xfs_trace.h           |    4 +
 10 files changed, 165 insertions(+), 3 deletions(-)
 create mode 100644 fs/xfs/scrub/rtbitmap.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 5035ee0..a22b027 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -118,6 +118,11 @@ xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   rmap.o \
 				   symlink.o \
 				   )
+ifeq ($(CONFIG_XFS_DEBUG)$(CONFIG_XFS_RT),yy)
+xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
+				   rtbitmap.o \
+				   )
+endif
 
 # low-level transaction/log code
 xfs-y				+= xfs_log.o \
diff --git a/fs/xfs/libxfs/xfs_format.h b/fs/xfs/libxfs/xfs_format.h
index 301effc..cb00017 100644
--- a/fs/xfs/libxfs/xfs_format.h
+++ b/fs/xfs/libxfs/xfs_format.h
@@ -315,6 +315,11 @@ static inline bool xfs_sb_good_version(struct xfs_sb *sbp)
 	return false;
 }
 
+static inline bool xfs_sb_version_hasrealtime(struct xfs_sb *sbp)
+{
+	return sbp->sb_rblocks > 0;
+}
+
 /*
  * Detect a mismatched features2 field.  Older kernels read/wrote
  * this into the wrong slot, so to be safe we keep them in sync.
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 5484c26..bcb3a69 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -592,7 +592,9 @@ struct xfs_scrub_metadata {
 #define XFS_SCRUB_TYPE_DIR	15	/* directory */
 #define XFS_SCRUB_TYPE_XATTR	16	/* extended attribute */
 #define XFS_SCRUB_TYPE_SYMLINK	17	/* symbolic link */
-#define XFS_SCRUB_TYPE_MAX	17
+#define XFS_SCRUB_TYPE_RTBITMAP	18	/* realtime bitmap */
+#define XFS_SCRUB_TYPE_RTSUM	19	/* realtime summary */
+#define XFS_SCRUB_TYPE_MAX	19
 
 #define XFS_SCRUB_FLAG_REPAIR	0x01	/* i: repair this metadata */
 #define XFS_SCRUB_FLAG_CORRUPT	0x02	/* o: needs repair */
diff --git a/fs/xfs/libxfs/xfs_rtbitmap.c b/fs/xfs/libxfs/xfs_rtbitmap.c
index ea45584..f4b68c0 100644
--- a/fs/xfs/libxfs/xfs_rtbitmap.c
+++ b/fs/xfs/libxfs/xfs_rtbitmap.c
@@ -70,7 +70,7 @@ const struct xfs_buf_ops xfs_rtbuf_ops = {
  * Get a buffer for the bitmap or summary file block specified.
  * The buffer is returned read and locked.
  */
-static int
+int
 xfs_rtbuf_get(
 	xfs_mount_t	*mp,		/* file system mount structure */
 	xfs_trans_t	*tp,		/* transaction pointer */
diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index f3bbd89..374157f 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -248,6 +248,7 @@ xfs_scrub_superblock(
 	XFS_SCRUB_SB_FEAT(metauuid);
 	XFS_SCRUB_SB_FEAT(rmapbt);
 	XFS_SCRUB_SB_FEAT(reflink);
+	XFS_SCRUB_SB_FEAT(realtime);
 #undef XFS_SCRUB_SB_FEAT
 
 #define XFS_SCRUB_SB_FEAT_PREEN(fn) \
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 0b55933..b7ac141 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -690,6 +690,13 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_inode, xfs_scrub_directory, NULL, NULL},
 	{xfs_scrub_setup_inode_xattr, xfs_scrub_xattr, NULL, NULL},
 	{xfs_scrub_setup_inode_symlink, xfs_scrub_symlink, NULL, NULL},
+#ifdef CONFIG_XFS_RT
+	{xfs_scrub_setup_rt, xfs_scrub_rtbitmap, NULL, xfs_sb_version_hasrealtime},
+	{xfs_scrub_setup_rt, xfs_scrub_rtsummary, NULL, xfs_sb_version_hasrealtime},
+#else
+	{NULL, NULL, NULL, NULL},
+	{NULL, NULL, NULL, NULL},
+#endif
 };
 
 /* Dispatch metadata scrubbing. */
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 123dc1d..8dc68b9 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -235,6 +235,10 @@ int xfs_scrub_setup_inode_symlink(struct xfs_scrub_context *sc,
 				  struct xfs_inode *ip,
 				  struct xfs_scrub_metadata *sm,
 				  bool retry_deadlocked);
+int xfs_scrub_setup_rt(struct xfs_scrub_context *sc,
+		       struct xfs_inode *ip,
+		       struct xfs_scrub_metadata *sm,
+		       bool retry_deadlocked);
 
 /* Metadata scrubbers */
 
@@ -255,5 +259,7 @@ int xfs_scrub_bmap_cow(struct xfs_scrub_context *sc);
 int xfs_scrub_directory(struct xfs_scrub_context *sc);
 int xfs_scrub_xattr(struct xfs_scrub_context *sc);
 int xfs_scrub_symlink(struct xfs_scrub_context *sc);
+int xfs_scrub_rtbitmap(struct xfs_scrub_context *sc);
+int xfs_scrub_rtsummary(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/rtbitmap.c b/fs/xfs/scrub/rtbitmap.c
new file mode 100644
index 0000000..2a5a546f
--- /dev/null
+++ b/fs/xfs/scrub/rtbitmap.c
@@ -0,0 +1,131 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_alloc.h"
+#include "xfs_rtalloc.h"
+#include "xfs_inode.h"
+#include "scrub/common.h"
+
+/* Set us up with the realtime metadata locked. */
+int
+xfs_scrub_setup_rt(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+	int				lockmode;
+	int				error = 0;
+
+	if (sm->sm_agno || sm->sm_ino || sm->sm_gen)
+		return -EINVAL;
+
+	error = xfs_scrub_setup(sc, ip, sm, retry_deadlocked);
+	if (error)
+		return error;
+
+	lockmode = XFS_ILOCK_EXCL | XFS_ILOCK_RTBITMAP;
+	xfs_ilock(mp->m_rbmip, lockmode);
+	xfs_trans_ijoin(sc->tp, mp->m_rbmip, lockmode);
+
+	return 0;
+}
+
+/* Realtime bitmap. */
+
+#define XFS_SCRUB_RTBITMAP_CHECK(fs_ok) \
+	XFS_SCRUB_CHECK(sc, bp, "rtbitmap", fs_ok);
+#define XFS_SCRUB_RTBITMAP_OP_ERROR_GOTO(error, label) \
+	XFS_SCRUB_OP_ERROR_GOTO(sc, 0, 0, "rtbitmap", error, label)
+/* Scrub the realtime bitmap. */
+int
+xfs_scrub_rtbitmap(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_buf			*bp = NULL;
+	xfs_rtblock_t			rtstart;
+	xfs_rtblock_t			rtend = 0;
+	xfs_rtblock_t			block;
+	xfs_rtblock_t			rem;
+	int				is_free = 0;
+	int				error = 0;
+	int				err2 = 0;
+
+	/* Iterate the bitmap, looking for discrepancies. */
+	rtstart = 0;
+	rem = mp->m_sb.sb_rblocks;
+	while (rem) {
+		if (xfs_scrub_should_terminate(&error))
+			break;
+
+		/* Is the first block free? */
+		err2 = xfs_rtcheck_range(mp, sc->tp, rtstart, 1, 1, &rtend,
+				&is_free);
+		XFS_SCRUB_RTBITMAP_OP_ERROR_GOTO(&err2, out);
+
+		/* How long does the extent go for? */
+		err2 = xfs_rtfind_forw(mp, sc->tp, rtstart,
+				mp->m_sb.sb_rblocks - 1, &rtend);
+		XFS_SCRUB_RTBITMAP_OP_ERROR_GOTO(&err2, out);
+
+		/* Find the buffer for error reporting. */
+		block = XFS_BITTOBLOCK(mp, rtstart);
+		err2 = xfs_rtbuf_get(mp, sc->tp, block, 0, &bp);
+		XFS_SCRUB_RTBITMAP_OP_ERROR_GOTO(&err2, out);
+		XFS_SCRUB_RTBITMAP_CHECK(rtend >= rtstart);
+
+		xfs_trans_brelse(sc->tp, bp);
+		bp = NULL;
+		rem -= rtend - rtstart + 1;
+		rtstart = rtend + 1;
+	}
+
+out:
+	if (bp)
+		xfs_trans_brelse(sc->tp, bp);
+	if (!error && err2)
+		error = err2;
+	return error;
+}
+#undef XFS_SCRUB_RTBITMAP_OP_ERROR_GOTO
+#undef XFS_SCRUB_RTBITMAP_CHECK
+
+/* Scrub the realtime summary. */
+int
+xfs_scrub_rtsummary(
+	struct xfs_scrub_context	*sc)
+{
+	/* XXX: implement this some day */
+	return -ENOENT;
+}
diff --git a/fs/xfs/xfs_rtalloc.h b/fs/xfs/xfs_rtalloc.h
index f798a3e..3036349 100644
--- a/fs/xfs/xfs_rtalloc.h
+++ b/fs/xfs/xfs_rtalloc.h
@@ -98,6 +98,8 @@ xfs_growfs_rt(
 /*
  * From xfs_rtbitmap.c
  */
+int xfs_rtbuf_get(struct xfs_mount *mp, struct xfs_trans *tp,
+		  xfs_rtblock_t block, int issum, struct xfs_buf **bpp);
 int xfs_rtcheck_range(struct xfs_mount *mp, struct xfs_trans *tp,
 		      xfs_rtblock_t start, xfs_extlen_t len, int val,
 		      xfs_rtblock_t *new, int *stat);
@@ -128,6 +130,7 @@ int xfs_rtfree_range(struct xfs_mount *mp, struct xfs_trans *tp,
 # define xfs_growfs_rt(mp,in)                           (ENOSYS)
 # define xfs_rtcheck_range(...)                         (ENOSYS)
 # define xfs_rtfind_forw(...)                           (ENOSYS)
+# define xfs_rtbuf_get(m,t,b,i,p)                       (ENOSYS)
 static inline int		/* error */
 xfs_rtmount_init(
 	xfs_mount_t	*mp)	/* file system mount structure */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 87fd942..c52e3a5 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3370,7 +3370,9 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
 	{ XFS_SCRUB_TYPE_BMBTC,		"bmapbtc" }, \
 	{ XFS_SCRUB_TYPE_DIR,		"dir" }, \
 	{ XFS_SCRUB_TYPE_XATTR,		"xattr" }, \
-	{ XFS_SCRUB_TYPE_SYMLINK,	"symlink" }
+	{ XFS_SCRUB_TYPE_SYMLINK,	"symlink" }, \
+	{ XFS_SCRUB_TYPE_RTBITMAP,	"rtbitmap" }, \
+	{ XFS_SCRUB_TYPE_RTSUM,		"rtsummary" }
 DECLARE_EVENT_CLASS(xfs_scrub_class,
 	TP_PROTO(struct xfs_inode *ip, int type, xfs_agnumber_t agno,
 		 xfs_ino_t inum, unsigned int gen, unsigned int flags,

[PATCH 32/55] xfs: set up cross-referencing helpers

[Darrick J. Wong]

Create some helper functions that we'll use later to cross reference
metadata with other metadata.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/btree.c  |  102 +++++++++++++++++++++++++++++++++++++++++++++++--
 fs/xfs/scrub/btree.h  |   47 +++++++++++++++++++----
 fs/xfs/scrub/common.c |   63 +++++++++++++++++++++++++-----
 fs/xfs/scrub/common.h |   84 ++++++++++++++++++++++++++++++++++------
 4 files changed, 261 insertions(+), 35 deletions(-)


diff --git a/fs/xfs/scrub/btree.c b/fs/xfs/scrub/btree.c
index 06607de..9ea0fbe 100644
--- a/fs/xfs/scrub/btree.c
+++ b/fs/xfs/scrub/btree.c
@@ -112,7 +112,8 @@ xfs_scrub_btree_ok(
 	bool				fs_ok,
 	const char			*check,
 	const char			*func,
-	int				line)
+	int				line,
+	bool				xref)
 {
 	char				bt_ptr[24];
 	char				bt_type[48];
@@ -121,7 +122,7 @@ xfs_scrub_btree_ok(
 	if (fs_ok)
 		return fs_ok;
 
-	sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+	sc->sm->sm_flags |= CORRUPT_FLAG(xref);
 	xfs_scrub_btree_format(cur, level, bt_type, 48, bt_ptr, 24, &fsbno);
 
 	trace_xfs_scrub_btree_error(cur->bc_mp, bt_type, bt_ptr,
@@ -139,7 +140,8 @@ xfs_scrub_btree_op_ok(
 	int				level,
 	int				*error,
 	const char			*func,
-	int				line)
+	int				line,
+	bool				xref)
 {
 	char				bt_ptr[24];
 	char				bt_type[48];
@@ -153,7 +155,7 @@ xfs_scrub_btree_op_ok(
 	return xfs_scrub_op_ok(sc,
 			XFS_FSB_TO_AGNO(cur->bc_mp, fsbno),
 			XFS_FSB_TO_AGBNO(cur->bc_mp, fsbno),
-			bt_type, error, func, line);
+			bt_type, error, func, line, xref);
 }
 
 /*
@@ -494,6 +496,84 @@ xfs_scrub_btree_sblock_check_siblings(
 	return error;
 }
 
+struct check_owner {
+	struct list_head	list;
+	xfs_fsblock_t		fsb;
+};
+
+/*
+ * Make sure this btree block isn't in the free list and that there's
+ * an rmap record for it.
+ */
+STATIC int
+xfs_scrub_btree_check_block_owner(
+	struct xfs_scrub_btree		*bs,
+	xfs_fsblock_t			fsb)
+{
+	struct xfs_scrub_ag		sa;
+	struct xfs_scrub_ag		*psa;
+	xfs_agnumber_t			agno;
+	xfs_agblock_t			bno;
+	int				error = 0;
+
+	agno = XFS_FSB_TO_AGNO(bs->cur->bc_mp, fsb);
+	bno = XFS_FSB_TO_AGBNO(bs->cur->bc_mp, fsb);
+
+	if (bs->cur->bc_flags & XFS_BTREE_LONG_PTRS) {
+		if (!xfs_scrub_ag_can_lock(bs->sc, agno))
+			return -EDEADLOCK;
+		error = xfs_scrub_ag_init(bs->sc, agno, &sa);
+		if (error)
+			return error;
+		psa = &sa;
+	} else
+		psa = &bs->sc->sa;
+
+	if (bs->cur->bc_flags & XFS_BTREE_LONG_PTRS)
+		xfs_scrub_ag_free(&sa);
+
+	return error;
+}
+
+/* Check the owner of a btree block. */
+STATIC int
+xfs_scrub_btree_check_owner(
+	struct xfs_scrub_btree		*bs,
+	struct xfs_buf			*bp)
+{
+	struct xfs_btree_cur		*cur = bs->cur;
+	struct check_owner		*co;
+	xfs_fsblock_t			fsbno;
+	xfs_agnumber_t			agno;
+
+	if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) && bp == NULL)
+		return 0;
+
+	fsbno = XFS_DADDR_TO_FSB(cur->bc_mp, bp->b_bn);
+	agno = XFS_FSB_TO_AGNO(cur->bc_mp, fsbno);
+
+	/* Turn back if we could deadlock. */
+	if ((bs->cur->bc_flags & XFS_BTREE_LONG_PTRS) &&
+	    !xfs_scrub_ag_can_lock(bs->sc, agno))
+		return -EDEADLOCK;
+
+	/*
+	 * We want to cross-reference each btree block with the bnobt
+	 * and the rmapbt.  We cannot cross-reference the bnobt or
+	 * rmapbt while scanning the bnobt or rmapbt, respectively,
+	 * because that would trash the cursor state.  Therefore, save
+	 * the block numbers for later scanning.
+	 */
+	if (cur->bc_btnum == XFS_BTNUM_BNO || cur->bc_btnum == XFS_BTNUM_RMAP) {
+		co = kmem_alloc(sizeof(struct check_owner), KM_SLEEP | KM_NOFS);
+		co->fsb = fsbno;
+		list_add_tail(&co->list, &bs->to_check);
+		return 0;
+	}
+
+	return xfs_scrub_btree_check_block_owner(bs, fsbno);
+}
+
 /* Grab and scrub a btree block. */
 STATIC int
 xfs_scrub_btree_block(
@@ -514,6 +594,10 @@ xfs_scrub_btree_block(
 	if (error)
 		return error;
 
+	error = xfs_scrub_btree_check_owner(bs, *pbp);
+	if (error)
+		return error;
+
 	return bs->check_siblings_fn(bs, *pblock);
 }
 
@@ -539,6 +623,8 @@ xfs_scrub_btree(
 	struct xfs_btree_block		*block;
 	int				level;
 	struct xfs_buf			*bp;
+	struct check_owner		*co;
+	struct check_owner		*n;
 	int				i;
 	int				error = 0;
 
@@ -653,6 +739,14 @@ xfs_scrub_btree(
 		}
 	}
 
+	/* Process deferred owner checks on btree blocks. */
+	list_for_each_entry_safe(co, n, &bs.to_check, list) {
+		if (!error)
+			error = xfs_scrub_btree_check_block_owner(&bs, co->fsb);
+		list_del(&co->list);
+		kmem_free(co);
+	}
+
 out_badcursor:
 	return error;
 }
diff --git a/fs/xfs/scrub/btree.h b/fs/xfs/scrub/btree.h
index 75e89b1..d8c8186 100644
--- a/fs/xfs/scrub/btree.h
+++ b/fs/xfs/scrub/btree.h
@@ -27,41 +27,72 @@ extern const char * const btree_types[];
 /* Check for btree corruption. */
 bool xfs_scrub_btree_ok(struct xfs_scrub_context *sc,
 			struct xfs_btree_cur *cur, int level, bool fs_ok,
-			const char *check, const char *func, int line);
+			const char *check, const char *func, int line,
+			bool xref);
 
 /* Check for btree operation errors. */
 bool xfs_scrub_btree_op_ok(struct xfs_scrub_context *sc,
 			   struct xfs_btree_cur *cur, int level, int *error,
-			   const char *func, int line);
+			   const char *func, int line, bool xref);
 
 #define XFS_SCRUB_BTREC_CHECK(bs, fs_ok) \
 	xfs_scrub_btree_ok((bs)->sc, (bs)->cur, 0, (fs_ok), #fs_ok, \
-			__func__, __LINE__)
+			__func__, __LINE__, false)
 #define XFS_SCRUB_BTREC_GOTO(bs, fs_ok, label) \
 	do { \
 		if (!xfs_scrub_btree_ok((bs)->sc, (bs)->cur, 0, (fs_ok), \
-				#fs_ok, __func__, __LINE__)) \
+				#fs_ok, __func__, __LINE__, false)) \
 			goto label; \
 	} while (0)
 #define XFS_SCRUB_BTREC_OP_ERROR_GOTO(bs, error, label) \
 	do { \
 		if (!xfs_scrub_btree_op_ok((bs)->sc, (bs)->cur, 0, \
-				(error), __func__, __LINE__)) \
+				(error), __func__, __LINE__, false)) \
 			goto label; \
 	} while (0)
 #define XFS_SCRUB_BTKEY_CHECK(bs, level, fs_ok) \
 	xfs_scrub_btree_ok((bs)->sc, (bs)->cur, (level), (fs_ok), #fs_ok, \
-			__func__, __LINE__)
+			__func__, __LINE__, false)
 #define XFS_SCRUB_BTKEY_GOTO(bs, level, fs_ok, label) \
 	do { \
 		if (!xfs_scrub_btree_ok((bs)->sc, (bs)->cur, (level), (fs_ok), \
-				#fs_ok, __func__, __LINE__)) \
+				#fs_ok, __func__, __LINE__, false)) \
 			goto label; \
 	} while (0)
 #define XFS_SCRUB_BTKEY_OP_ERROR_GOTO(bs, level, error, label) \
 	do { \
 		if (!xfs_scrub_btree_op_ok((bs)->sc, (bs)->cur, (level), \
-				(error), __func__, __LINE__)) \
+				(error), __func__, __LINE__, false)) \
+			goto label; \
+	} while (0)
+#define XFS_SCRUB_BTREC_XCHECK(bs, fs_ok) \
+	xfs_scrub_btree_ok((bs)->sc, (bs)->cur, 0, (fs_ok), #fs_ok, \
+			__func__, __LINE__, true)
+#define XFS_SCRUB_BTREC_XGOTO(bs, fs_ok, label) \
+	do { \
+		if (!xfs_scrub_btree_ok((bs)->sc, (bs)->cur, 0, (fs_ok), \
+				#fs_ok, __func__, __LINE__, true)) \
+			goto label; \
+	} while (0)
+#define XFS_SCRUB_BTREC_OP_ERROR_XGOTO(bs, error, label) \
+	do { \
+		if (!xfs_scrub_btree_op_ok((bs)->sc, (bs)->cur, 0, \
+				(error), __func__, __LINE__, true)) \
+			goto label; \
+	} while (0)
+#define XFS_SCRUB_BTKEY_XCHECK(bs, level, fs_ok) \
+	xfs_scrub_btree_ok((bs)->sc, (bs)->cur, (level), (fs_ok), #fs_ok, \
+			__func__, __LINE__, true)
+#define XFS_SCRUB_BTKEY_XGOTO(bs, level, fs_ok, label) \
+	do { \
+		if (!xfs_scrub_btree_ok((bs)->sc, (bs)->cur, (level), (fs_ok), \
+				#fs_ok, __func__, __LINE__, true)) \
+			goto label; \
+	} while (0)
+#define XFS_SCRUB_BTKEY_OP_ERROR_XGOTO(bs, level, error, label) \
+	do { \
+		if (!xfs_scrub_btree_op_ok((bs)->sc, (bs)->cur, (level), \
+				(error), __func__, __LINE__, true)) \
 			goto label; \
 	} while (0)
 
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index b7ac141..1c20006 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -119,7 +119,8 @@ xfs_scrub_op_ok(
 	const char			*type,
 	int				*error,
 	const char			*func,
-	int				line)
+	int				line,
+	bool				xref)
 {
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 
@@ -128,7 +129,7 @@ xfs_scrub_op_ok(
 
 	trace_xfs_scrub_op_error(mp, agno, bno, type, *error, func, line);
 	if (*error == -EFSBADCRC || *error == -EFSCORRUPTED) {
-		sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+		sc->sm->sm_flags |= CORRUPT_FLAG(xref);
 		*error = 0;
 	}
 	return false;
@@ -143,7 +144,8 @@ xfs_scrub_file_op_ok(
 	const char			*type,
 	int				*error,
 	const char			*func,
-	int				line)
+	int				line,
+	bool				xref)
 {
 	if (*error == 0)
 		return true;
@@ -151,7 +153,7 @@ xfs_scrub_file_op_ok(
 	trace_xfs_scrub_file_op_error(sc->ip, whichfork, offset, type, *error,
 			func, line);
 	if (*error == -EFSBADCRC || *error == -EFSCORRUPTED) {
-		sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+		sc->sm->sm_flags |= CORRUPT_FLAG(xref);
 		*error = 0;
 	}
 	return false;
@@ -194,7 +196,8 @@ xfs_scrub_block_ok(
 	bool				fs_ok,
 	const char			*check,
 	const char			*func,
-	int				line)
+	int				line,
+	bool				xref)
 {
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 	xfs_fsblock_t			fsbno;
@@ -208,7 +211,7 @@ xfs_scrub_block_ok(
 	agno = XFS_FSB_TO_AGNO(mp, fsbno);
 	bno = XFS_FSB_TO_AGBNO(mp, fsbno);
 
-	sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+	sc->sm->sm_flags |= CORRUPT_FLAG(xref);
 	trace_xfs_scrub_block_error(mp, agno, bno, type, check, func, line);
 	return fs_ok;
 }
@@ -223,7 +226,8 @@ xfs_scrub_ino_ok(
 	bool				fs_ok,
 	const char			*check,
 	const char			*func,
-	int				line)
+	int				line,
+	bool				xref)
 {
 	struct xfs_inode		*ip = sc->ip;
 	struct xfs_mount		*mp = sc->tp->t_mountp;
@@ -243,7 +247,7 @@ xfs_scrub_ino_ok(
 		bno = XFS_INO_TO_AGINO(mp, ip->i_ino);
 	}
 
-	sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+	sc->sm->sm_flags |= CORRUPT_FLAG(xref);
 	trace_xfs_scrub_ino_error(mp, ino, agno, bno, type, check, func, line);
 	return fs_ok;
 }
@@ -293,12 +297,13 @@ xfs_scrub_data_ok(
 	bool				fs_ok,
 	const char			*check,
 	const char			*func,
-	int				line)
+	int				line,
+	bool				xref)
 {
 	if (fs_ok)
 		return fs_ok;
 
-	sc->sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+	sc->sm->sm_flags |= CORRUPT_FLAG(xref);
 	trace_xfs_scrub_data_error(sc->ip, whichfork, offset, type, check,
 			func, line);
 	return fs_ok;
@@ -533,6 +538,44 @@ xfs_scrub_ag_lock_all(
 	return error;
 }
 
+/*
+ * Predicate that decides if we need to evaluate the cross-reference check.
+ * If there was an error accessing the cross-reference btree, just delete
+ * the cursor and skip the check.
+ */
+bool
+__xfs_scrub_should_xref(
+	struct xfs_scrub_context	*sc,
+	int				error,
+	struct xfs_btree_cur		**curpp,
+	const char			*func,
+	int				line)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+
+	/* If not a btree cross-reference, just check the error code. */
+	if (curpp == NULL) {
+		if (error == 0)
+			return true;
+		trace_xfs_scrub_xref_error(mp, "unknown", error, func, line);
+		return false;
+	}
+
+	ASSERT(*curpp != NULL);
+	/* If no error or we've already given up on xref, just bail out. */
+	if (error == 0 || *curpp == NULL)
+		return true;
+
+	/* xref error, delete cursor and bail out. */
+	sc->sm->sm_flags |= XFS_SCRUB_FLAG_XFAIL;
+	trace_xfs_scrub_xref_error(mp, btree_types[(*curpp)->bc_btnum],
+			error, func, line);
+	xfs_btree_del_cursor(*curpp, XFS_BTREE_ERROR);
+	*curpp = NULL;
+
+	return false;
+}
+
 /* Dummy scrubber */
 
 STATIC int
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 8dc68b9..4704b38 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -100,25 +100,44 @@ xfs_scrub_trans_alloc(
 	return xfs_trans_alloc_empty(mp, tpp);
 }
 
+static inline __u32
+CORRUPT_FLAG(
+	bool				xref)
+{
+	return xref ? XFS_SCRUB_FLAG_XCORRUPT : XFS_SCRUB_FLAG_CORRUPT;
+}
+
 /* Check for operational errors. */
 bool xfs_scrub_op_ok(struct xfs_scrub_context *sc, xfs_agnumber_t agno,
 		     xfs_agblock_t bno, const char *type, int *error,
-		     const char	*func, int line);
+		     const char	*func, int line, bool xref);
 #define XFS_SCRUB_OP_ERROR_GOTO(sc, agno, bno, type, error, label) \
 	do { \
 		if (!xfs_scrub_op_ok((sc), (agno), (bno), (type), \
-				(error), __func__, __LINE__)) \
+				(error), __func__, __LINE__, false)) \
+			goto label; \
+	} while (0)
+#define XFS_SCRUB_OP_ERROR_XGOTO(sc, agno, bno, type, error, label) \
+	do { \
+		if (!xfs_scrub_op_ok((sc), (agno), (bno), (type), \
+				(error), __func__, __LINE__, true)) \
 			goto label; \
 	} while (0)
 
 /* Check for operational errors for a file offset. */
 bool xfs_scrub_file_op_ok(struct xfs_scrub_context *sc, int whichfork,
 			  xfs_fileoff_t offset, const char *type,
-			  int *error, const char *func, int line);
+			  int *error, const char *func, int line, bool xref);
 #define XFS_SCRUB_FILE_OP_ERROR_GOTO(sc, which, off, type, error, label) \
 	do { \
 		if (!xfs_scrub_file_op_ok((sc), (which), (off), (type), \
-				(error), __func__, __LINE__)) \
+				(error), __func__, __LINE__, false)) \
+			goto label; \
+	} while (0)
+#define XFS_SCRUB_FILE_OP_ERROR_XGOTO(sc, which, off, type, error, label) \
+	do { \
+		if (!xfs_scrub_file_op_ok((sc), (which), (off), (type), \
+				(error), __func__, __LINE__, true)) \
 			goto label; \
 	} while (0)
 
@@ -141,44 +160,83 @@ bool xfs_scrub_ino_preen(struct xfs_scrub_context *sc, struct xfs_buf *bp,
 /* Check for metadata block corruption. */
 bool xfs_scrub_block_ok(struct xfs_scrub_context *sc, struct xfs_buf *bp,
 			const char *type, bool fs_ok, const char *check,
-			const char *func, int line);
+			const char *func, int line, bool xref);
 #define XFS_SCRUB_CHECK(sc, bp, type, fs_ok) \
 	xfs_scrub_block_ok((sc), (bp), (type), (fs_ok), #fs_ok, \
-			__func__, __LINE__)
+			__func__, __LINE__, false)
 #define XFS_SCRUB_GOTO(sc, bp, type, fs_ok, label) \
 	do { \
 		if (!xfs_scrub_block_ok((sc), (bp), (type), (fs_ok), \
-				#fs_ok, __func__, __LINE__)) \
+				#fs_ok, __func__, __LINE__, false)) \
+			goto label; \
+	} while (0)
+#define XFS_SCRUB_XCHECK(sc, bp, type, fs_ok) \
+	xfs_scrub_block_ok((sc), (bp), (type), (fs_ok), #fs_ok, \
+			__func__, __LINE__, true)
+#define XFS_SCRUB_XGOTO(sc, bp, type, fs_ok, label) \
+	do { \
+		if (!xfs_scrub_block_ok((sc), (bp), (type), (fs_ok), \
+				#fs_ok, __func__, __LINE__, true)) \
 			goto label; \
 	} while (0)
 
 /* Check for inode metadata corruption. */
 bool xfs_scrub_ino_ok(struct xfs_scrub_context *sc, xfs_ino_t ino,
 		      struct xfs_buf *bp, const char *type, bool fs_ok,
-		      const char *check, const char *func, int line);
+		      const char *check, const char *func, int line,
+		      bool xref);
 #define XFS_SCRUB_INO_CHECK(sc, ino, bp, type, fs_ok) \
 	xfs_scrub_ino_ok((sc), (ino), (bp), (type), (fs_ok), #fs_ok, \
-			__func__, __LINE__)
+			__func__, __LINE__, false)
 #define XFS_SCRUB_INO_GOTO(sc, ino, bp, type, fs_ok, label) \
 	do { \
 		if (!xfs_scrub_ino_ok((sc), (ino), (bp), (type), (fs_ok), \
-				#fs_ok, __func__, __LINE__)) \
+				#fs_ok, __func__, __LINE__, false)) \
+			goto label; \
+	} while(0)
+#define XFS_SCRUB_INO_XCHECK(sc, ino, bp, type, fs_ok) \
+	xfs_scrub_ino_ok((sc), (ino), (bp), (type), (fs_ok), #fs_ok, \
+			__func__, __LINE__, true)
+#define XFS_SCRUB_INO_XGOTO(sc, ino, bp, type, fs_ok, label) \
+	do { \
+		if (!xfs_scrub_ino_ok((sc), (ino), (bp), (type), (fs_ok), \
+				#fs_ok, __func__, __LINE__, true)) \
 			goto label; \
 	} while(0)
 
 /* Check for file data block corruption. */
 bool xfs_scrub_data_ok(struct xfs_scrub_context *sc, int whichfork,
 		       xfs_fileoff_t offset, const char *type, bool fs_ok,
-		       const char *check, const char *func, int line);
+		       const char *check, const char *func, int line,
+		       bool xref);
 #define XFS_SCRUB_DATA_CHECK(sc, whichfork, offset, type, fs_ok) \
 	xfs_scrub_data_ok((sc), (whichfork), (offset), (type), (fs_ok), \
-			#fs_ok, __func__, __LINE__)
+			#fs_ok, __func__, __LINE__, false)
 #define XFS_SCRUB_DATA_GOTO(sc, whichfork, offset, type, fs_ok, label) \
 	do { \
 		if (!xfs_scrub_data_ok((sc), (whichfork), (offset), \
-				(type), (fs_ok), #fs_ok, __func__, __LINE__)) \
+				(type), (fs_ok), #fs_ok, __func__, __LINE__, \
+				false)) \
 			goto label; \
 	} while(0)
+#define XFS_SCRUB_DATA_XCHECK(sc, whichfork, offset, type, fs_ok) \
+	xfs_scrub_data_ok((sc), (whichfork), (offset), (type), (fs_ok), \
+			#fs_ok, __func__, __LINE__, true)
+#define XFS_SCRUB_DATA_XGOTO(sc, whichfork, offset, type, fs_ok, label) \
+	do { \
+		if (!xfs_scrub_data_ok((sc), (whichfork), (offset), \
+				(type), (fs_ok), #fs_ok, __func__, __LINE__, \
+				true)) \
+			goto label; \
+	} while(0)
+
+bool __xfs_scrub_should_xref(struct xfs_scrub_context *sc, int error,
+			     struct xfs_btree_cur **curpp, const char *func,
+			     int line);
+#define xfs_scrub_should_xref(sc, error, curpp) \
+	__xfs_scrub_should_xref((sc), (error), (curpp), __func__, __LINE__)
+#define xfs_scrub_btree_should_xref(bs, error, curpp) \
+	__xfs_scrub_should_xref((bs)->sc, (error), (curpp), __func__, __LINE__)
 
 bool xfs_scrub_ag_can_lock(struct xfs_scrub_context *sc, xfs_agnumber_t agno);
 int xfs_scrub_ag_lock_all(struct xfs_scrub_context *sc);

[PATCH 33/55] xfs: scrub should cross-reference with the bnobt

[Darrick J. Wong]

When we're scrubbing various btrees, cross-reference the records with
the bnobt to ensure that we don't also think the space is free.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_alloc.c |   19 +++++++
 fs/xfs/libxfs/xfs_alloc.h |    3 +
 fs/xfs/scrub/agheader.c   |  117 +++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/bmap.c       |   21 ++++++++
 fs/xfs/scrub/btree.c      |    9 +++
 fs/xfs/scrub/ialloc.c     |   13 +++++
 fs/xfs/scrub/refcount.c   |   17 +++++++
 fs/xfs/scrub/rmap.c       |   18 +++++++
 8 files changed, 214 insertions(+), 3 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
index 5384d88..2c471bb 100644
--- a/fs/xfs/libxfs/xfs_alloc.c
+++ b/fs/xfs/libxfs/xfs_alloc.c
@@ -2932,3 +2932,22 @@ xfs_alloc_query_all(
 	query.fn = fn;
 	return xfs_btree_query_all(cur, xfs_alloc_query_range_helper, &query);
 }
+
+/* Is there a record covering a given extent? */
+int
+xfs_alloc_has_record(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	xfs_extlen_t		len,
+	bool			*exists)
+{
+	union xfs_btree_irec	low;
+	union xfs_btree_irec	high;
+
+	memset(&low, 0, sizeof(low));
+	low.a.ar_startblock = bno;
+	memset(&high, 0xFF, sizeof(high));
+	high.a.ar_startblock = bno + len - 1;
+
+	return xfs_btree_has_record(cur, &low, &high, exists);
+}
diff --git a/fs/xfs/libxfs/xfs_alloc.h b/fs/xfs/libxfs/xfs_alloc.h
index 1e7e2e2..c24fe03 100644
--- a/fs/xfs/libxfs/xfs_alloc.h
+++ b/fs/xfs/libxfs/xfs_alloc.h
@@ -237,4 +237,7 @@ int xfs_alloc_query_range(struct xfs_btree_cur *cur,
 int xfs_alloc_query_all(struct xfs_btree_cur *cur, xfs_alloc_query_range_fn fn,
 		void *priv);
 
+int xfs_alloc_has_record(struct xfs_btree_cur *cur, xfs_agblock_t bno,
+		xfs_extlen_t len, bool *exist);
+
 #endif	/* __XFS_ALLOC_H__ */
diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index 374157f..eda4abd 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -31,6 +31,7 @@
 #include "xfs_trace.h"
 #include "xfs_sb.h"
 #include "xfs_inode.h"
+#include "xfs_alloc.h"
 #include "scrub/common.h"
 
 /* Set us up to check an AG header. */
@@ -122,6 +123,8 @@ xfs_scrub_walk_agfl(
 	XFS_SCRUB_PREEN(sc, bp, "superblock", fs_ok)
 #define XFS_SCRUB_SB_OP_ERROR_GOTO(label) \
 	XFS_SCRUB_OP_ERROR_GOTO(sc, agno, 0, "superblock", &error, out)
+#define XFS_SCRUB_SB_XCHECK(fs_ok) \
+	XFS_SCRUB_XCHECK(sc, bp, "superblock", fs_ok)
 /* Scrub the filesystem superblock. */
 int
 xfs_scrub_superblock(
@@ -129,10 +132,13 @@ xfs_scrub_superblock(
 {
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 	struct xfs_buf			*bp;
+	struct xfs_scrub_ag		*psa;
 	struct xfs_sb			sb;
 	xfs_agnumber_t			agno;
 	uint32_t			v2_ok;
+	bool				is_freesp;
 	int				error;
+	int				err2;
 
 	agno = sc->sm->sm_agno;
 
@@ -152,7 +158,7 @@ xfs_scrub_superblock(
 	 * so there's no point in comparing the two.
 	 */
 	if (agno == 0)
-		goto out;
+		goto btree_xref;
 
 	xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp));
 
@@ -258,19 +264,53 @@ xfs_scrub_superblock(
 	XFS_SCRUB_SB_FEAT_PREEN(attr2);
 #undef XFS_SCRUB_SB_FEAT_PREEN
 
+	if (error)
+		goto out;
+
+btree_xref:
+
+	err2 = xfs_scrub_ag_init(sc, agno, &sc->sa);
+	if (!xfs_scrub_should_xref(sc, err2, NULL))
+		goto out;
+
+	psa = &sc->sa;
+	/* Cross-reference with bnobt. */
+	if (psa->bno_cur) {
+		err2 = xfs_alloc_has_record(psa->bno_cur, XFS_SB_BLOCK(mp),
+				1, &is_freesp);
+		if (xfs_scrub_should_xref(sc, err2, &psa->bno_cur))
+			XFS_SCRUB_SB_XCHECK(!is_freesp);
+	}
+
 out:
 	return error;
 }
+#undef XFS_SCRUB_SB_XCHECK
 #undef XFS_SCRUB_SB_OP_ERROR_GOTO
 #undef XFS_SCRUB_SB_CHECK
 
 /* AGF */
 
+/* Tally freespace record lengths. */
+STATIC int
+xfs_scrub_agf_record_bno_lengths(
+	struct xfs_btree_cur		*cur,
+	struct xfs_alloc_rec_incore	*rec,
+	void				*priv)
+{
+	xfs_extlen_t			*blocks = priv;
+
+	(*blocks) += rec->ar_blockcount;
+	return 0;
+}
+
 #define XFS_SCRUB_AGF_CHECK(fs_ok) \
 	XFS_SCRUB_CHECK(sc, sc->sa.agf_bp, "AGF", fs_ok)
 #define XFS_SCRUB_AGF_OP_ERROR_GOTO(error, label) \
 	XFS_SCRUB_OP_ERROR_GOTO(sc, sc->sm->sm_agno, \
 			XFS_AGF_BLOCK(sc->tp->t_mountp), "AGF", error, label)
+#define XFS_SCRUB_AGF_XCHECK(fs_ok) \
+	XFS_SCRUB_XCHECK(sc, sc->sa.agf_bp, "AGF", fs_ok)
 /* Scrub the AGF. */
 int
 xfs_scrub_agf(
@@ -278,6 +318,7 @@ xfs_scrub_agf(
 {
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 	struct xfs_agf			*agf;
+	struct xfs_scrub_ag		*psa;
 	xfs_daddr_t			daddr;
 	xfs_daddr_t			eofs;
 	xfs_agnumber_t			agno;
@@ -287,8 +328,11 @@ xfs_scrub_agf(
 	xfs_agblock_t			agfl_last;
 	xfs_agblock_t			agfl_count;
 	xfs_agblock_t			fl_count;
+	xfs_extlen_t			blocks;
+	bool				is_freesp;
 	int				level;
 	int				error = 0;
+	int				err2;
 
 	agno = sc->sm->sm_agno;
 	error = xfs_scrub_load_ag_headers(sc, agno, XFS_SCRUB_TYPE_AGF);
@@ -360,9 +404,35 @@ xfs_scrub_agf(
 		fl_count = XFS_AGFL_SIZE(mp) - agfl_first + agfl_last + 1;
 	XFS_SCRUB_AGF_CHECK(agfl_count == 0 || fl_count == agfl_count);
 
+	/* Load btrees for xref if the AGF is ok. */
+	psa = &sc->sa;
+	if (error || (sc->sm->sm_flags & XFS_SCRUB_FLAG_CORRUPT))
+		goto out;
+	error = xfs_scrub_ag_btcur_init(sc, psa);
+	if (error)
+		goto out;
+
+	/* Cross-reference with the bnobt. */
+	if (psa->bno_cur) {
+		err2 = xfs_alloc_has_record(psa->bno_cur, XFS_AGF_BLOCK(mp),
+				1, &is_freesp);
+		if (!xfs_scrub_should_xref(sc, err2, &psa->bno_cur))
+			goto skip_bnobt;
+		XFS_SCRUB_AGF_XCHECK(!is_freesp);
+
+		blocks = 0;
+		err2 = xfs_alloc_query_all(psa->bno_cur,
+				xfs_scrub_agf_record_bno_lengths, &blocks);
+		if (!xfs_scrub_should_xref(sc, err2, &psa->bno_cur))
+			goto skip_bnobt;
+		XFS_SCRUB_AGF_XCHECK(blocks == be32_to_cpu(agf->agf_freeblks));
+	}
+skip_bnobt:
+
 out:
 	return error;
 }
+#undef XFS_SCRUB_AGF_XCHECK
 #undef XFS_SCRUB_AGF_OP_ERROR_GOTO
 #undef XFS_SCRUB_AGF_CHECK
 
@@ -370,6 +440,8 @@ xfs_scrub_agf(
 
 #define XFS_SCRUB_AGFL_CHECK(fs_ok) \
 	XFS_SCRUB_CHECK(sc, sc->sa.agfl_bp, "AGFL", fs_ok)
+#define XFS_SCRUB_AGFL_XCHECK(fs_ok) \
+	XFS_SCRUB_XCHECK(sc, sc->sa.agfl_bp, "AGFL", fs_ok)
 struct xfs_scrub_agfl {
 	xfs_agblock_t			eoag;
 	xfs_daddr_t			eofs;
@@ -385,12 +457,22 @@ xfs_scrub_agfl_block(
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 	xfs_agnumber_t			agno = sc->sa.agno;
 	struct xfs_scrub_agfl		*sagfl = priv;
+	bool				is_freesp;
+	int				err2;
 
 	XFS_SCRUB_AGFL_CHECK(agbno > XFS_AGI_BLOCK(mp));
 	XFS_SCRUB_AGFL_CHECK(XFS_AGB_TO_DADDR(mp, agno, agbno) < sagfl->eofs);
 	XFS_SCRUB_AGFL_CHECK(agbno < mp->m_sb.sb_agblocks);
 	XFS_SCRUB_AGFL_CHECK(agbno < sagfl->eoag);
 
+	/* Cross-reference with the bnobt. */
+	if (sc->sa.bno_cur) {
+		err2 = xfs_alloc_has_record(sc->sa.bno_cur, agbno,
+				1, &is_freesp);
+		if (xfs_scrub_should_xref(sc, err2, &sc->sa.bno_cur))
+			XFS_SCRUB_AGFL_XCHECK(!is_freesp);
+	}
+
 	return 0;
 }
 
@@ -405,7 +487,9 @@ xfs_scrub_agfl(
 	struct xfs_scrub_agfl		sagfl;
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 	struct xfs_agf			*agf;
+	bool				is_freesp;
 	int				error;
+	int				err2;
 
 	error = xfs_scrub_load_ag_headers(sc, sc->sm->sm_agno,
 			XFS_SCRUB_TYPE_AGFL);
@@ -417,12 +501,21 @@ xfs_scrub_agfl(
 	sagfl.eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
 	sagfl.eoag = be32_to_cpu(agf->agf_length);
 
+	/* Cross-reference with the bnobt. */
+	if (sc->sa.bno_cur) {
+		err2 = xfs_alloc_has_record(sc->sa.bno_cur, XFS_AGFL_BLOCK(mp),
+				1, &is_freesp);
+		if (xfs_scrub_should_xref(sc, err2, &sc->sa.bno_cur))
+			XFS_SCRUB_AGFL_XCHECK(!is_freesp);
+	}
+
 	/* Check the blocks in the AGFL. */
 	return xfs_scrub_walk_agfl(sc, xfs_scrub_agfl_block, &sagfl);
 out:
 	return error;
 }
 #undef XFS_SCRUB_AGFL_OP_ERROR_GOTO
+#undef XFS_SCRUB_AGFL_XCHECK
 #undef XFS_SCRUB_AGFL_CHECK
 
 /* AGI */
@@ -432,6 +525,8 @@ xfs_scrub_agfl(
 #define XFS_SCRUB_AGI_OP_ERROR_GOTO(error, label) \
 	XFS_SCRUB_OP_ERROR_GOTO(sc, sc->sm->sm_agno, \
 			XFS_AGI_BLOCK(sc->tp->t_mountp), "AGI", error, label)
+#define XFS_SCRUB_AGI_XCHECK(fs_ok) \
+	XFS_SCRUB_XCHECK(sc, sc->sa.agi_bp, "AGI", fs_ok)
 /* Scrub the AGI. */
 int
 xfs_scrub_agi(
@@ -439,6 +534,7 @@ xfs_scrub_agi(
 {
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 	struct xfs_agi			*agi;
+	struct xfs_scrub_ag		*psa;
 	xfs_daddr_t			daddr;
 	xfs_daddr_t			eofs;
 	xfs_agnumber_t			agno;
@@ -447,9 +543,11 @@ xfs_scrub_agi(
 	xfs_agino_t			agino;
 	xfs_agino_t			first_agino;
 	xfs_agino_t			last_agino;
+	bool				is_freesp;
 	int				i;
 	int				level;
 	int				error = 0;
+	int				err2;
 
 	agno = sc->sm->sm_agno;
 	error = xfs_scrub_load_ag_headers(sc, agno, XFS_SCRUB_TYPE_AGI);
@@ -515,8 +613,25 @@ xfs_scrub_agi(
 		XFS_SCRUB_AGI_CHECK(agino <= last_agino);
 	}
 
+	/* Load btrees for xref if the AGI is ok. */
+	psa = &sc->sa;
+	if (error || (sc->sm->sm_flags & XFS_SCRUB_FLAG_CORRUPT))
+		goto out;
+	error = xfs_scrub_ag_btcur_init(sc, &sc->sa);
+	if (error)
+		goto out;
+
+	/* Cross-reference with bnobt. */
+	if (psa->bno_cur) {
+		err2 = xfs_alloc_has_record(psa->bno_cur, XFS_AGI_BLOCK(mp),
+				1, &is_freesp);
+		if (xfs_scrub_should_xref(sc, err2, &psa->bno_cur))
+			XFS_SCRUB_AGI_XCHECK(!is_freesp);
+	}
+
 out:
 	return error;
 }
+#undef XFS_SCRUB_AGI_XCHECK
 #undef XFS_SCRUB_AGI_CHECK
 #undef XFS_SCRUB_AGI_OP_ERROR_GOTO
diff --git a/fs/xfs/scrub/bmap.c b/fs/xfs/scrub/bmap.c
index 87ee945..0f860ea 100644
--- a/fs/xfs/scrub/bmap.c
+++ b/fs/xfs/scrub/bmap.c
@@ -36,6 +36,7 @@
 #include "xfs_bmap_util.h"
 #include "xfs_bmap_btree.h"
 #include "xfs_rmap.h"
+#include "xfs_alloc.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -85,6 +86,10 @@ struct xfs_scrub_bmap_info {
 	XFS_SCRUB_INO_GOTO(info->sc, info->sc->ip->i_ino, bp, info->type, fs_ok, label)
 #define XFS_SCRUB_BMAP_OP_ERROR_GOTO(label) \
 	XFS_SCRUB_OP_ERROR_GOTO(info->sc, agno, 0, "bmap", &error, label);
+#define XFS_SCRUB_BMAP_OP_ERROR_XGOTO(label) \
+	XFS_SCRUB_OP_ERROR_XGOTO(info->sc, agno, 0, "bmap", &error, label);
+#define XFS_SCRUB_BMAP_XCHECK(fs_ok) \
+	XFS_SCRUB_INO_XCHECK(info->sc, info->sc->ip->i_ino, bp, info->type, fs_ok)
 /* Scrub a single extent record. */
 STATIC int
 xfs_scrub_bmap_extent(
@@ -99,7 +104,10 @@ xfs_scrub_bmap_extent(
 	xfs_daddr_t			daddr;
 	xfs_daddr_t			dlen;
 	xfs_agnumber_t			agno;
+	xfs_fsblock_t			bno;
+	bool				is_freesp;
 	int				error = 0;
+	int				err2 = 0;
 
 	if (cur)
 		xfs_btree_get_block(cur, 0, &bp);
@@ -112,10 +120,12 @@ xfs_scrub_bmap_extent(
 	if (info->is_rt) {
 		daddr = XFS_FSB_TO_BB(mp, irec->br_startblock);
 		agno = NULLAGNUMBER;
+		bno = irec->br_startblock;
 	} else {
 		daddr = XFS_FSB_TO_DADDR(mp, irec->br_startblock);
 		agno = XFS_FSB_TO_AGNO(mp, irec->br_startblock);
 		XFS_SCRUB_BMAP_GOTO(agno < mp->m_sb.sb_agcount, out);
+		bno = XFS_FSB_TO_AGBNO(mp, irec->br_startblock);
 	}
 	dlen = XFS_FSB_TO_BB(mp, irec->br_blockcount);
 	XFS_SCRUB_BMAP_CHECK(agno < mp->m_sb.sb_agcount);
@@ -134,7 +144,15 @@ xfs_scrub_bmap_extent(
 		if (!xfs_scrub_ag_can_lock(info->sc, agno))
 			return -EDEADLOCK;
 		error = xfs_scrub_ag_init(info->sc, agno, &sa);
-		XFS_SCRUB_BMAP_OP_ERROR_GOTO(out);
+		XFS_SCRUB_BMAP_OP_ERROR_XGOTO(out);
+	}
+
+	/* Cross-reference with the bnobt. */
+	if (sa.bno_cur) {
+		err2 = xfs_alloc_has_record(sa.bno_cur, bno,
+				irec->br_blockcount, &is_freesp);
+		if (xfs_scrub_should_xref(info->sc, err2, &sa.bno_cur))
+			XFS_SCRUB_BMAP_XCHECK(!is_freesp);
 	}
 
 	xfs_scrub_ag_free(&sa);
@@ -179,6 +197,7 @@ xfs_scrub_bmapbt_helper(
 	xfs_bmbt_get_all(&ihost, &irec);
 	return xfs_scrub_bmap_extent(ip, bs->cur, info, &irec);
 }
+#undef XFS_SCRUB_BMAP_XCHECK
 #undef XFS_SCRUB_BMAP_CHECK
 
 #define XFS_SCRUB_FORK_CHECK(fs_ok) \
diff --git a/fs/xfs/scrub/btree.c b/fs/xfs/scrub/btree.c
index 9ea0fbe..30360e0 100644
--- a/fs/xfs/scrub/btree.c
+++ b/fs/xfs/scrub/btree.c
@@ -514,7 +514,9 @@ xfs_scrub_btree_check_block_owner(
 	struct xfs_scrub_ag		*psa;
 	xfs_agnumber_t			agno;
 	xfs_agblock_t			bno;
+	bool				is_freesp;
 	int				error = 0;
+	int				err2;
 
 	agno = XFS_FSB_TO_AGNO(bs->cur->bc_mp, fsb);
 	bno = XFS_FSB_TO_AGBNO(bs->cur->bc_mp, fsb);
@@ -529,6 +531,13 @@ xfs_scrub_btree_check_block_owner(
 	} else
 		psa = &bs->sc->sa;
 
+	/* Cross-reference with the bnobt. */
+	if (psa->bno_cur) {
+		err2 = xfs_alloc_has_record(psa->bno_cur, bno, 1, &is_freesp);
+		if (xfs_scrub_btree_should_xref(bs, err2, NULL))
+			XFS_SCRUB_BTREC_XCHECK(bs, !is_freesp);
+	}
+
 	if (bs->cur->bc_flags & XFS_BTREE_LONG_PTRS)
 		xfs_scrub_ag_free(&sa);
 
diff --git a/fs/xfs/scrub/ialloc.c b/fs/xfs/scrub/ialloc.c
index 0c654c9..69d9e77 100644
--- a/fs/xfs/scrub/ialloc.c
+++ b/fs/xfs/scrub/ialloc.c
@@ -37,6 +37,7 @@
 #include "xfs_rmap.h"
 #include "xfs_log.h"
 #include "xfs_trans_priv.h"
+#include "xfs_alloc.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -83,9 +84,12 @@ xfs_scrub_iallocbt_chunk(
 {
 	struct xfs_mount		*mp = bs->cur->bc_mp;
 	struct xfs_agf			*agf;
+	struct xfs_scrub_ag		*psa;
 	xfs_agblock_t			eoag;
 	xfs_agblock_t			bno;
+	bool				is_freesp;
 	int				error = 0;
+	int				err2;
 
 	agf = XFS_BUF_TO_AGF(bs->sc->sa.agf_bp);
 	eoag = be32_to_cpu(agf->agf_length);
@@ -104,6 +108,15 @@ xfs_scrub_iallocbt_chunk(
 		goto out;
 	}
 
+	psa = &bs->sc->sa;
+	/* Cross-reference with the bnobt. */
+	if (psa->bno_cur) {
+		err2 = xfs_alloc_has_record(psa->bno_cur, bno, len,
+				&is_freesp);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->bno_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, !is_freesp);
+	}
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/refcount.c b/fs/xfs/scrub/refcount.c
index f7ceba9..b371a58 100644
--- a/fs/xfs/scrub/refcount.c
+++ b/fs/xfs/scrub/refcount.c
@@ -31,6 +31,7 @@
 #include "xfs_trace.h"
 #include "xfs_sb.h"
 #include "xfs_rmap.h"
+#include "xfs_alloc.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -44,10 +45,13 @@ xfs_scrub_refcountbt_helper(
 {
 	struct xfs_mount		*mp = bs->cur->bc_mp;
 	struct xfs_agf			*agf;
+	struct xfs_scrub_ag		*psa;
 	struct xfs_refcount_irec	irec;
 	xfs_agblock_t			eoag;
 	bool				has_cowflag;
+	bool				is_freesp;
 	int				error = 0;
+	int				err2;
 
 	irec.rc_startblock = be32_to_cpu(rec->refc.rc_startblock);
 	irec.rc_blockcount = be32_to_cpu(rec->refc.rc_blockcount);
@@ -69,6 +73,19 @@ xfs_scrub_refcountbt_helper(
 			irec.rc_blockcount <= eoag);
 	XFS_SCRUB_BTREC_CHECK(bs, irec.rc_refcount >= 1);
 
+	if (error)
+		goto out;
+
+	psa = &bs->sc->sa;
+	/* Cross-reference with the bnobt. */
+	if (psa->bno_cur) {
+		err2 = xfs_alloc_has_record(psa->bno_cur, irec.rc_startblock,
+				irec.rc_blockcount, &is_freesp);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->bno_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, !is_freesp);
+	}
+
+out:
 	return error;
 }
 
diff --git a/fs/xfs/scrub/rmap.c b/fs/xfs/scrub/rmap.c
index bb94260..f92ddad 100644
--- a/fs/xfs/scrub/rmap.c
+++ b/fs/xfs/scrub/rmap.c
@@ -31,6 +31,7 @@
 #include "xfs_trace.h"
 #include "xfs_sb.h"
 #include "xfs_rmap.h"
+#include "xfs_alloc.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -44,13 +45,16 @@ xfs_scrub_rmapbt_helper(
 {
 	struct xfs_mount		*mp = bs->cur->bc_mp;
 	struct xfs_agf			*agf;
+	struct xfs_scrub_ag		*psa;
 	struct xfs_rmap_irec		irec;
 	xfs_agblock_t			eoag;
+	bool				is_freesp;
 	bool				non_inode;
 	bool				is_unwritten;
 	bool				is_bmbt;
 	bool				is_attr;
-	int				error;
+	int				error = 0;
+	int				err2;
 
 	error = xfs_rmap_btrec_to_irec(rec, &irec);
 	XFS_SCRUB_BTREC_OP_ERROR_GOTO(bs, &error, out);
@@ -99,6 +103,18 @@ xfs_scrub_rmapbt_helper(
 	XFS_SCRUB_BTREC_CHECK(bs, !non_inode ||
 			(irec.rm_owner > XFS_RMAP_OWN_MIN &&
 			 irec.rm_owner <= XFS_RMAP_OWN_FS));
+	if (error)
+		goto out;
+
+	psa = &bs->sc->sa;
+	/* Cross-reference with the bnobt. */
+	if (psa->bno_cur) {
+		err2 = xfs_alloc_has_record(psa->bno_cur, irec.rm_startblock,
+				irec.rm_blockcount, &is_freesp);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->bno_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, !is_freesp);
+	}
+
 out:
 	return error;
 }

[PATCH 34/55] xfs: cross-reference bnobt records with cntbt

[Darrick J. Wong]

Scrub should make sure that each bnobt record has a corresponding
cntbt record.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_alloc.c |    2 +-
 fs/xfs/libxfs/xfs_alloc.h |    7 +++++++
 fs/xfs/scrub/agheader.c   |   20 ++++++++++++++++++++
 fs/xfs/scrub/alloc.c      |   31 +++++++++++++++++++++++++++++++
 4 files changed, 59 insertions(+), 1 deletion(-)


diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
index 2c471bb..24227cd 100644
--- a/fs/xfs/libxfs/xfs_alloc.c
+++ b/fs/xfs/libxfs/xfs_alloc.c
@@ -166,7 +166,7 @@ xfs_alloc_lookup_ge(
  * Lookup the first record less than or equal to [bno, len]
  * in the btree given by cur.
  */
-static int				/* error */
+int					/* error */
 xfs_alloc_lookup_le(
 	struct xfs_btree_cur	*cur,	/* btree cursor */
 	xfs_agblock_t		bno,	/* starting block of extent */
diff --git a/fs/xfs/libxfs/xfs_alloc.h b/fs/xfs/libxfs/xfs_alloc.h
index c24fe03..0f75db4 100644
--- a/fs/xfs/libxfs/xfs_alloc.h
+++ b/fs/xfs/libxfs/xfs_alloc.h
@@ -202,6 +202,13 @@ xfs_free_extent(
 	enum xfs_ag_resv_type	type);	/* block reservation type */
 
 int				/* error */
+xfs_alloc_lookup_le(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agblock_t		bno,	/* starting block of extent */
+	xfs_extlen_t		len,	/* length of extent */
+	int			*stat);	/* success/failure */
+
+int				/* error */
 xfs_alloc_lookup_ge(
 	struct xfs_btree_cur	*cur,	/* btree cursor */
 	xfs_agblock_t		bno,	/* starting block of extent */
diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index eda4abd..d05a403 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -330,6 +330,7 @@ xfs_scrub_agf(
 	xfs_agblock_t			fl_count;
 	xfs_extlen_t			blocks;
 	bool				is_freesp;
+	int				have;
 	int				level;
 	int				error = 0;
 	int				err2;
@@ -429,6 +430,25 @@ xfs_scrub_agf(
 	}
 skip_bnobt:
 
+	/* Cross-reference with the cntbt. */
+	if (psa->cnt_cur) {
+		err2 = xfs_alloc_lookup_le(psa->cnt_cur, 0, -1U, &have);
+		if (!xfs_scrub_should_xref(sc, err2, &psa->cnt_cur))
+			goto skip_cntbt;
+		if (!have) {
+			XFS_SCRUB_AGF_XCHECK(agf->agf_freeblks ==
+					be32_to_cpu(0));
+			goto skip_cntbt;
+		}
+		err2 = xfs_alloc_get_rec(psa->cnt_cur, &agbno, &blocks, &have);
+		if (!xfs_scrub_should_xref(sc, err2, &psa->cnt_cur))
+			goto skip_cntbt;
+		XFS_SCRUB_AGF_XCHECK(have);
+		XFS_SCRUB_AGF_XCHECK(!have ||
+				blocks == be32_to_cpu(agf->agf_longest));
+	}
+skip_cntbt:
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/alloc.c b/fs/xfs/scrub/alloc.c
index 2edf85d..770e82c 100644
--- a/fs/xfs/scrub/alloc.c
+++ b/fs/xfs/scrub/alloc.c
@@ -31,6 +31,7 @@
 #include "xfs_trace.h"
 #include "xfs_sb.h"
 #include "xfs_rmap.h"
+#include "xfs_alloc.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -65,9 +66,15 @@ xfs_scrub_allocbt_helper(
 {
 	struct xfs_mount		*mp = bs->cur->bc_mp;
 	struct xfs_agf			*agf;
+	struct xfs_btree_cur		**xcur;
+	struct xfs_scrub_ag		*psa;
+	xfs_agblock_t			fbno;
 	xfs_agblock_t			bno;
+	xfs_extlen_t			flen;
 	xfs_extlen_t			len;
+	int				has_otherrec;
 	int				error = 0;
+	int				err2;
 
 	bno = be32_to_cpu(rec->alloc.ar_startblock);
 	len = be32_to_cpu(rec->alloc.ar_blockcount);
@@ -81,6 +88,30 @@ xfs_scrub_allocbt_helper(
 	XFS_SCRUB_BTREC_CHECK(bs, (unsigned long long)bno + len <=
 			be32_to_cpu(agf->agf_length));
 
+	if (error)
+		goto out;
+
+	psa = &bs->sc->sa;
+	/*
+	 * Ensure there's a corresponding cntbt/bnobt record matching
+	 * this bnobt/cntbt record, respectively.
+	 */
+	xcur = bs->cur == psa->bno_cur ? &psa->cnt_cur : &psa->bno_cur;
+	if (*xcur) {
+		err2 = xfs_alloc_lookup_le(*xcur, bno, len, &has_otherrec);
+		if (xfs_scrub_btree_should_xref(bs, err2, xcur)) {
+			XFS_SCRUB_BTREC_XGOTO(bs, has_otherrec, out);
+			err2 = xfs_alloc_get_rec(*xcur, &fbno, &flen,
+					&has_otherrec);
+			if (xfs_scrub_btree_should_xref(bs, err2, xcur)) {
+				XFS_SCRUB_BTREC_XGOTO(bs, has_otherrec, out);
+				XFS_SCRUB_BTREC_XCHECK(bs, fbno == bno);
+				XFS_SCRUB_BTREC_XCHECK(bs, flen == len);
+			}
+		}
+	}
+
+out:
 	return error;
 }
 

[PATCH 35/55] xfs: cross-reference extents with AG header

[Darrick J. Wong]

Ensure that none of the AG btree records overlap the AG sb/agf/agfl/agi
headers except for the XFS_RMAP_OWN_FS rmap.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/agheader.c |   27 +++++++++++++++++++++++++++
 fs/xfs/scrub/alloc.c    |    4 ++++
 fs/xfs/scrub/bmap.c     |    5 +++++
 fs/xfs/scrub/common.h   |    2 ++
 fs/xfs/scrub/ialloc.c   |    4 ++++
 fs/xfs/scrub/refcount.c |    4 ++++
 fs/xfs/scrub/rmap.c     |    5 +++++
 7 files changed, 51 insertions(+)


diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index d05a403..e5af267 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -115,6 +115,30 @@ xfs_scrub_walk_agfl(
 	return 0;
 }
 
+/* Does this AG extent cover the AG headers? */
+bool
+xfs_scrub_extent_covers_ag_head(
+	struct xfs_mount	*mp,
+	xfs_agblock_t		agbno,
+	xfs_extlen_t		len)
+{
+	xfs_agblock_t		bno;
+
+	bno = XFS_SB_BLOCK(mp);
+	if (bno >= agbno && bno < agbno + len)
+		return true;
+	bno = XFS_AGF_BLOCK(mp);
+	if (bno >= agbno && bno < agbno + len)
+		return true;
+	bno = XFS_AGFL_BLOCK(mp);
+	if (bno >= agbno && bno < agbno + len)
+		return true;
+	bno = XFS_AGI_BLOCK(mp);
+	if (bno >= agbno && bno < agbno + len)
+		return true;
+	return false;
+}
+
 /* Superblock */
 
 #define XFS_SCRUB_SB_CHECK(fs_ok) \
@@ -485,6 +509,9 @@ xfs_scrub_agfl_block(
 	XFS_SCRUB_AGFL_CHECK(agbno < mp->m_sb.sb_agblocks);
 	XFS_SCRUB_AGFL_CHECK(agbno < sagfl->eoag);
 
+	/* Cross-reference with the AG headers. */
+	XFS_SCRUB_AGFL_XCHECK(!xfs_scrub_extent_covers_ag_head(mp, agbno, 1));
+
 	/* Cross-reference with the bnobt. */
 	if (sc->sa.bno_cur) {
 		err2 = xfs_alloc_has_record(sc->sa.bno_cur, agbno,
diff --git a/fs/xfs/scrub/alloc.c b/fs/xfs/scrub/alloc.c
index 770e82c..af5565e 100644
--- a/fs/xfs/scrub/alloc.c
+++ b/fs/xfs/scrub/alloc.c
@@ -91,6 +91,10 @@ xfs_scrub_allocbt_helper(
 	if (error)
 		goto out;
 
+	/* Cross-reference with the AG headers. */
+	XFS_SCRUB_BTREC_XCHECK(bs,
+			!xfs_scrub_extent_covers_ag_head(mp, bno, len));
+
 	psa = &bs->sc->sa;
 	/*
 	 * Ensure there's a corresponding cntbt/bnobt record matching
diff --git a/fs/xfs/scrub/bmap.c b/fs/xfs/scrub/bmap.c
index 0f860ea..ac27852 100644
--- a/fs/xfs/scrub/bmap.c
+++ b/fs/xfs/scrub/bmap.c
@@ -147,6 +147,11 @@ xfs_scrub_bmap_extent(
 		XFS_SCRUB_BMAP_OP_ERROR_XGOTO(out);
 	}
 
+	/* Cross-reference with the AG headers. */
+	if (!info->is_rt)
+		XFS_SCRUB_BMAP_XCHECK(!xfs_scrub_extent_covers_ag_head(mp,
+				bno, irec->br_blockcount));
+
 	/* Cross-reference with the bnobt. */
 	if (sa.bno_cur) {
 		err2 = xfs_alloc_has_record(sa.bno_cur, bno,
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 4704b38..11c04d5 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -256,6 +256,8 @@ int xfs_scrub_walk_agfl(struct xfs_scrub_context *sc,
 			int (*fn)(struct xfs_scrub_context *, xfs_agblock_t bno,
 				  void *),
 			void *priv);
+bool xfs_scrub_extent_covers_ag_head(struct xfs_mount *mp, xfs_agblock_t agbno,
+				     xfs_extlen_t len);
 
 /* Setup functions */
 
diff --git a/fs/xfs/scrub/ialloc.c b/fs/xfs/scrub/ialloc.c
index 69d9e77..e96e500 100644
--- a/fs/xfs/scrub/ialloc.c
+++ b/fs/xfs/scrub/ialloc.c
@@ -108,6 +108,10 @@ xfs_scrub_iallocbt_chunk(
 		goto out;
 	}
 
+	/* Cross-reference with the AG headers. */
+	XFS_SCRUB_BTREC_XCHECK(bs,
+			!xfs_scrub_extent_covers_ag_head(mp, bno, len));
+
 	psa = &bs->sc->sa;
 	/* Cross-reference with the bnobt. */
 	if (psa->bno_cur) {
diff --git a/fs/xfs/scrub/refcount.c b/fs/xfs/scrub/refcount.c
index b371a58..76b1e22 100644
--- a/fs/xfs/scrub/refcount.c
+++ b/fs/xfs/scrub/refcount.c
@@ -76,6 +76,10 @@ xfs_scrub_refcountbt_helper(
 	if (error)
 		goto out;
 
+	/* Cross-reference with the AG headers. */
+	XFS_SCRUB_BTREC_XCHECK(bs, !xfs_scrub_extent_covers_ag_head(mp,
+			irec.rc_startblock, irec.rc_blockcount));
+
 	psa = &bs->sc->sa;
 	/* Cross-reference with the bnobt. */
 	if (psa->bno_cur) {
diff --git a/fs/xfs/scrub/rmap.c b/fs/xfs/scrub/rmap.c
index f92ddad..d536704 100644
--- a/fs/xfs/scrub/rmap.c
+++ b/fs/xfs/scrub/rmap.c
@@ -106,6 +106,11 @@ xfs_scrub_rmapbt_helper(
 	if (error)
 		goto out;
 
+	/* Cross-reference with the AG headers. */
+	XFS_SCRUB_BTREC_XCHECK(bs, irec.rm_owner == XFS_RMAP_OWN_FS ||
+			!xfs_scrub_extent_covers_ag_head(mp, irec.rm_startblock,
+				irec.rm_blockcount));
+
 	psa = &bs->sc->sa;
 	/* Cross-reference with the bnobt. */
 	if (psa->bno_cur) {

[PATCH 36/55] xfs: cross-reference inode btrees during scrub

[Darrick J. Wong]

Cross-reference the inode btrees with the other metadata when we
scrub the filesystem.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_ialloc.c |   99 ++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/libxfs/xfs_ialloc.h |    6 +++
 fs/xfs/scrub/agheader.c    |   98 ++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/alloc.c       |   18 ++++++++
 fs/xfs/scrub/bmap.c        |   20 +++++++++
 fs/xfs/scrub/ialloc.c      |   16 +++++++
 fs/xfs/scrub/refcount.c    |   20 +++++++++
 fs/xfs/scrub/rmap.c        |   24 +++++++++++
 8 files changed, 301 insertions(+)


diff --git a/fs/xfs/libxfs/xfs_ialloc.c b/fs/xfs/libxfs/xfs_ialloc.c
index e2f93e6..0fb7ba0 100644
--- a/fs/xfs/libxfs/xfs_ialloc.c
+++ b/fs/xfs/libxfs/xfs_ialloc.c
@@ -2668,3 +2668,102 @@ xfs_ialloc_pagi_init(
 		xfs_trans_brelse(tp, bp);
 	return 0;
 }
+
+/* Is there an inode record covering a given range of inode numbers? */
+int
+xfs_ialloc_has_inode_record(
+	struct xfs_btree_cur	*cur,
+	xfs_agino_t		low,
+	xfs_agino_t		high,
+	bool			*exists)
+{
+	struct xfs_inobt_rec_incore	irec;
+	xfs_agino_t		agino;
+	__uint16_t		holemask;
+	int			has;
+	int			i;
+	int			error;
+
+	*exists = false;
+	error = xfs_inobt_lookup(cur, low, XFS_LOOKUP_LE, &has);
+	while (error == 0 && has) {
+		error = xfs_inobt_get_rec(cur, &irec, &has);
+		if (error || irec.ir_startino > high)
+			break;
+
+		agino = irec.ir_startino;
+		holemask = irec.ir_holemask;
+		for (i = 0; i < XFS_INOBT_HOLEMASK_BITS; holemask >>= 1,
+				i++, agino += XFS_INODES_PER_HOLEMASK_BIT) {
+			if (holemask & 1)
+				continue;
+			if (agino + XFS_INODES_PER_HOLEMASK_BIT > low &&
+					agino <= high) {
+				*exists = true;
+				goto out;
+			}
+		}
+
+		error = xfs_btree_increment(cur, 0, &has);
+	}
+out:
+	return error;
+}
+
+/* Is there an inode record covering a given extent? */
+int
+xfs_ialloc_has_inodes_at_extent(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	xfs_extlen_t		len,
+	bool			*exists)
+{
+	xfs_agino_t		low;
+	xfs_agino_t		high;
+
+	low = XFS_OFFBNO_TO_AGINO(cur->bc_mp, bno, 0);
+	high = XFS_OFFBNO_TO_AGINO(cur->bc_mp, bno + len, 0) - 1;
+
+	return xfs_ialloc_has_inode_record(cur, low, high, exists);
+}
+
+struct xfs_ialloc_count_inodes {
+	xfs_agino_t			count;
+	xfs_agino_t			freecount;
+};
+
+/* Record inode counts across all inobt records. */
+STATIC int
+xfs_ialloc_count_inodes_helper(
+	struct xfs_btree_cur		*cur,
+	union xfs_btree_rec		*rec,
+	void				*priv)
+{
+	struct xfs_inobt_rec_incore	irec;
+	struct xfs_ialloc_count_inodes	*ci = priv;
+
+	xfs_inobt_btrec_to_irec(cur->bc_mp, rec, &irec);
+	ci->count += irec.ir_count;
+	ci->freecount += irec.ir_freecount;
+
+	return 0;
+}
+
+/* Count allocated and free inodes under an inobt. */
+int
+xfs_ialloc_count_inodes(
+	struct xfs_btree_cur		*cur,
+	xfs_agino_t			*count,
+	xfs_agino_t			*freecount)
+{
+	struct xfs_ialloc_count_inodes	ci = {0};
+	int				error;
+
+	ASSERT(cur->bc_btnum == XFS_BTNUM_INO);
+	error = xfs_btree_query_all(cur, xfs_ialloc_count_inodes_helper, &ci);
+	if (!error) {
+		*count = ci.count;
+		*freecount = ci.freecount;
+	}
+	return error;
+}
diff --git a/fs/xfs/libxfs/xfs_ialloc.h b/fs/xfs/libxfs/xfs_ialloc.h
index 8e5861d..17f0f1b 100644
--- a/fs/xfs/libxfs/xfs_ialloc.h
+++ b/fs/xfs/libxfs/xfs_ialloc.h
@@ -171,5 +171,11 @@ int xfs_read_agi(struct xfs_mount *mp, struct xfs_trans *tp,
 union xfs_btree_rec;
 void xfs_inobt_btrec_to_irec(struct xfs_mount *mp, union xfs_btree_rec *rec,
 		struct xfs_inobt_rec_incore *irec);
+int xfs_ialloc_has_inodes_at_extent(struct xfs_btree_cur *cur,
+		xfs_agblock_t bno, xfs_extlen_t len, bool *exists);
+int xfs_ialloc_has_inode_record(struct xfs_btree_cur *cur, xfs_agino_t low,
+		xfs_agino_t high, bool *exists);
+int xfs_ialloc_count_inodes(struct xfs_btree_cur *cur, xfs_agino_t *count,
+		xfs_agino_t *freecount);
 
 #endif	/* __XFS_IALLOC_H__ */
diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index e5af267..505b75f 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -32,6 +32,7 @@
 #include "xfs_sb.h"
 #include "xfs_inode.h"
 #include "xfs_alloc.h"
+#include "xfs_ialloc.h"
 #include "scrub/common.h"
 
 /* Set us up to check an AG header. */
@@ -161,6 +162,7 @@ xfs_scrub_superblock(
 	xfs_agnumber_t			agno;
 	uint32_t			v2_ok;
 	bool				is_freesp;
+	bool				has_inodes;
 	int				error;
 	int				err2;
 
@@ -306,6 +308,22 @@ xfs_scrub_superblock(
 			XFS_SCRUB_SB_XCHECK(!is_freesp);
 	}
 
+	/* Cross-reference with inobt. */
+	if (psa->ino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->ino_cur,
+				XFS_SB_BLOCK(mp), 1, &has_inodes);
+		if (xfs_scrub_should_xref(sc, err2, &psa->ino_cur))
+			XFS_SCRUB_SB_XCHECK(!has_inodes);
+	}
+
+	/* Cross-reference with finobt. */
+	if (psa->fino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->fino_cur,
+				XFS_SB_BLOCK(mp), 1, &has_inodes);
+		if (xfs_scrub_should_xref(sc, err2, &psa->fino_cur))
+			XFS_SCRUB_SB_XCHECK(!has_inodes);
+	}
+
 out:
 	return error;
 }
@@ -354,6 +372,7 @@ xfs_scrub_agf(
 	xfs_agblock_t			fl_count;
 	xfs_extlen_t			blocks;
 	bool				is_freesp;
+	bool				has_inodes;
 	int				have;
 	int				level;
 	int				error = 0;
@@ -473,6 +492,22 @@ xfs_scrub_agf(
 	}
 skip_cntbt:
 
+	/* Cross-reference with inobt. */
+	if (psa->ino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->ino_cur,
+				XFS_AGF_BLOCK(mp), 1, &has_inodes);
+		if (xfs_scrub_should_xref(sc, err2, &psa->ino_cur))
+			XFS_SCRUB_AGF_XCHECK(!has_inodes);
+	}
+
+	/* Cross-reference with finobt. */
+	if (psa->fino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->fino_cur,
+				XFS_AGF_BLOCK(mp), 1, &has_inodes);
+		if (xfs_scrub_should_xref(sc, err2, &psa->fino_cur))
+			XFS_SCRUB_AGF_XCHECK(!has_inodes);
+	}
+
 out:
 	return error;
 }
@@ -502,6 +537,7 @@ xfs_scrub_agfl_block(
 	xfs_agnumber_t			agno = sc->sa.agno;
 	struct xfs_scrub_agfl		*sagfl = priv;
 	bool				is_freesp;
+	bool				has_inodes;
 	int				err2;
 
 	XFS_SCRUB_AGFL_CHECK(agbno > XFS_AGI_BLOCK(mp));
@@ -520,6 +556,22 @@ xfs_scrub_agfl_block(
 			XFS_SCRUB_AGFL_XCHECK(!is_freesp);
 	}
 
+	/* Cross-reference with inobt. */
+	if (sc->sa.ino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(sc->sa.ino_cur,
+				agbno, 1, &has_inodes);
+		if (xfs_scrub_should_xref(sc, err2, &sc->sa.ino_cur))
+			XFS_SCRUB_AGFL_XCHECK(!has_inodes);
+	}
+
+	/* Cross-reference with finobt. */
+	if (sc->sa.fino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(sc->sa.fino_cur,
+				agbno, 1, &has_inodes);
+		if (xfs_scrub_should_xref(sc, err2, &sc->sa.fino_cur))
+			XFS_SCRUB_AGFL_XCHECK(!has_inodes);
+	}
+
 	return 0;
 }
 
@@ -535,6 +587,7 @@ xfs_scrub_agfl(
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 	struct xfs_agf			*agf;
 	bool				is_freesp;
+	bool				has_inodes;
 	int				error;
 	int				err2;
 
@@ -556,6 +609,22 @@ xfs_scrub_agfl(
 			XFS_SCRUB_AGFL_XCHECK(!is_freesp);
 	}
 
+	/* Cross-reference with inobt. */
+	if (sc->sa.ino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(sc->sa.ino_cur,
+			XFS_AGFL_BLOCK(mp), 1, &has_inodes);
+		if (xfs_scrub_should_xref(sc, err2, &sc->sa.ino_cur))
+			XFS_SCRUB_AGFL_XCHECK(!has_inodes);
+	}
+
+	/* Cross-reference with finobt. */
+	if (sc->sa.fino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(sc->sa.fino_cur,
+				XFS_AGFL_BLOCK(mp), 1, &has_inodes);
+		if (xfs_scrub_should_xref(sc, err2, &sc->sa.fino_cur))
+			XFS_SCRUB_AGFL_XCHECK(!has_inodes);
+	}
+
 	/* Check the blocks in the AGFL. */
 	return xfs_scrub_walk_agfl(sc, xfs_scrub_agfl_block, &sagfl);
 out:
@@ -590,7 +659,10 @@ xfs_scrub_agi(
 	xfs_agino_t			agino;
 	xfs_agino_t			first_agino;
 	xfs_agino_t			last_agino;
+	xfs_agino_t			count;
+	xfs_agino_t			freecount;
 	bool				is_freesp;
+	bool				has_inodes;
 	int				i;
 	int				level;
 	int				error = 0;
@@ -676,6 +748,32 @@ xfs_scrub_agi(
 			XFS_SCRUB_AGI_XCHECK(!is_freesp);
 	}
 
+	/* Cross-reference with inobt. */
+	if (psa->ino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->ino_cur,
+				XFS_AGI_BLOCK(mp), 1, &has_inodes);
+		if (!xfs_scrub_should_xref(sc, err2, &psa->ino_cur))
+			goto skip_inobt_xref;
+		XFS_SCRUB_AGI_XCHECK(!has_inodes);
+		err2 = xfs_ialloc_count_inodes(psa->ino_cur, &count,
+				&freecount);
+		if (xfs_scrub_should_xref(sc, err2, &psa->ino_cur)) {
+			XFS_SCRUB_AGI_XCHECK(be32_to_cpu(agi->agi_count) ==
+					count);
+			XFS_SCRUB_AGI_XCHECK(be32_to_cpu(agi->agi_freecount) ==
+					freecount);
+		}
+	}
+
+skip_inobt_xref:
+	/* Cross-reference with finobt. */
+	if (psa->fino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->fino_cur,
+				XFS_AGI_BLOCK(mp), 1, &has_inodes);
+		if (xfs_scrub_should_xref(sc, err2, &psa->fino_cur))
+			XFS_SCRUB_AGI_XCHECK(!has_inodes);
+	}
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/alloc.c b/fs/xfs/scrub/alloc.c
index af5565e..90b988d 100644
--- a/fs/xfs/scrub/alloc.c
+++ b/fs/xfs/scrub/alloc.c
@@ -32,6 +32,7 @@
 #include "xfs_sb.h"
 #include "xfs_rmap.h"
 #include "xfs_alloc.h"
+#include "xfs_ialloc.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -72,6 +73,7 @@ xfs_scrub_allocbt_helper(
 	xfs_agblock_t			bno;
 	xfs_extlen_t			flen;
 	xfs_extlen_t			len;
+	bool				has_inodes;
 	int				has_otherrec;
 	int				error = 0;
 	int				err2;
@@ -115,6 +117,22 @@ xfs_scrub_allocbt_helper(
 		}
 	}
 
+	/* Cross-reference with inobt. */
+	if (psa->ino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->ino_cur, bno,
+				len, &has_inodes);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->ino_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, !has_inodes);
+	}
+
+	/* Cross-reference with finobt. */
+	if (psa->fino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->fino_cur, bno,
+				len, &has_inodes);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->fino_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, !has_inodes);
+	}
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/bmap.c b/fs/xfs/scrub/bmap.c
index ac27852..263967a 100644
--- a/fs/xfs/scrub/bmap.c
+++ b/fs/xfs/scrub/bmap.c
@@ -37,6 +37,7 @@
 #include "xfs_bmap_btree.h"
 #include "xfs_rmap.h"
 #include "xfs_alloc.h"
+#include "xfs_ialloc.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -106,6 +107,7 @@ xfs_scrub_bmap_extent(
 	xfs_agnumber_t			agno;
 	xfs_fsblock_t			bno;
 	bool				is_freesp;
+	bool				has_inodes;
 	int				error = 0;
 	int				err2 = 0;
 
@@ -160,6 +162,24 @@ xfs_scrub_bmap_extent(
 			XFS_SCRUB_BMAP_XCHECK(!is_freesp);
 	}
 
+	/* Cross-reference with inobt. */
+	if (sa.ino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(sa.ino_cur,
+				irec->br_startblock, irec->br_blockcount,
+				&has_inodes);
+		if (xfs_scrub_should_xref(info->sc, err2, &sa.ino_cur))
+			XFS_SCRUB_BMAP_XCHECK(!has_inodes);
+	}
+
+	/* Cross-reference with finobt. */
+	if (sa.fino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(sa.fino_cur,
+				irec->br_startblock, irec->br_blockcount,
+				&has_inodes);
+		if (xfs_scrub_should_xref(info->sc, err2, &sa.fino_cur))
+			XFS_SCRUB_BMAP_XCHECK(!has_inodes);
+	}
+
 	xfs_scrub_ag_free(&sa);
 out:
 	info->lastoff = irec->br_startoff + irec->br_blockcount;
diff --git a/fs/xfs/scrub/ialloc.c b/fs/xfs/scrub/ialloc.c
index e96e500..585e4c3 100644
--- a/fs/xfs/scrub/ialloc.c
+++ b/fs/xfs/scrub/ialloc.c
@@ -85,9 +85,11 @@ xfs_scrub_iallocbt_chunk(
 	struct xfs_mount		*mp = bs->cur->bc_mp;
 	struct xfs_agf			*agf;
 	struct xfs_scrub_ag		*psa;
+	struct xfs_btree_cur		**xcur;
 	xfs_agblock_t			eoag;
 	xfs_agblock_t			bno;
 	bool				is_freesp;
+	bool				has_inodes;
 	int				error = 0;
 	int				err2;
 
@@ -121,6 +123,20 @@ xfs_scrub_iallocbt_chunk(
 			XFS_SCRUB_BTREC_XCHECK(bs, !is_freesp);
 	}
 
+	/* If we have a finobt, cross-reference with it. */
+	if (bs->cur == psa->fino_cur)
+		xcur = &psa->ino_cur;
+	else if (bs->cur == psa->ino_cur && irec->ir_freecount > 0)
+		xcur = &psa->fino_cur;
+	else
+		xcur = NULL;
+	if (xcur && *xcur) {
+		err2 = xfs_ialloc_has_inode_record(*xcur,
+				agino, agino, &has_inodes);
+		if (xfs_scrub_btree_should_xref(bs, err2, xcur))
+			XFS_SCRUB_BTREC_XCHECK(bs, has_inodes);
+	}
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/refcount.c b/fs/xfs/scrub/refcount.c
index 76b1e22..31a531f 100644
--- a/fs/xfs/scrub/refcount.c
+++ b/fs/xfs/scrub/refcount.c
@@ -32,6 +32,7 @@
 #include "xfs_sb.h"
 #include "xfs_rmap.h"
 #include "xfs_alloc.h"
+#include "xfs_ialloc.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -50,6 +51,7 @@ xfs_scrub_refcountbt_helper(
 	xfs_agblock_t			eoag;
 	bool				has_cowflag;
 	bool				is_freesp;
+	bool				has_inodes;
 	int				error = 0;
 	int				err2;
 
@@ -89,6 +91,24 @@ xfs_scrub_refcountbt_helper(
 			XFS_SCRUB_BTREC_XCHECK(bs, !is_freesp);
 	}
 
+	/* Cross-reference with inobt. */
+	if (psa->ino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->ino_cur,
+				irec.rc_startblock, irec.rc_blockcount,
+				&has_inodes);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->ino_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, !has_inodes);
+	}
+
+	/* Cross-reference with finobt. */
+	if (psa->fino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->fino_cur,
+				irec.rc_startblock, irec.rc_blockcount,
+				&has_inodes);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->fino_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, !has_inodes);
+	}
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/rmap.c b/fs/xfs/scrub/rmap.c
index d536704..30ec492 100644
--- a/fs/xfs/scrub/rmap.c
+++ b/fs/xfs/scrub/rmap.c
@@ -32,6 +32,7 @@
 #include "xfs_sb.h"
 #include "xfs_rmap.h"
 #include "xfs_alloc.h"
+#include "xfs_ialloc.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -53,6 +54,7 @@ xfs_scrub_rmapbt_helper(
 	bool				is_unwritten;
 	bool				is_bmbt;
 	bool				is_attr;
+	bool				has_inodes;
 	int				error = 0;
 	int				err2;
 
@@ -120,6 +122,28 @@ xfs_scrub_rmapbt_helper(
 			XFS_SCRUB_BTREC_XCHECK(bs, !is_freesp);
 	}
 
+	/* Cross-reference with inobt. */
+	if (psa->ino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->ino_cur,
+				irec.rm_startblock, irec.rm_blockcount,
+				&has_inodes);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->ino_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs,
+					irec.rm_owner == XFS_RMAP_OWN_INODES ||
+					!has_inodes);
+	}
+
+	/* Cross-reference with finobt. */
+	if (psa->fino_cur) {
+		err2 = xfs_ialloc_has_inodes_at_extent(psa->fino_cur,
+				irec.rm_startblock, irec.rm_blockcount,
+				&has_inodes);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->fino_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs,
+					irec.rm_owner == XFS_RMAP_OWN_INODES ||
+					!has_inodes);
+	}
+
 out:
 	return error;
 }

[PATCH 37/55] xfs: cross-reference reverse-mapping btree

[Darrick J. Wong]

When scrubbing various btrees, we should cross-reference the records
with the reverse mapping btree.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_rmap.c |   58 ++++++++++++++
 fs/xfs/libxfs/xfs_rmap.h |    5 +
 fs/xfs/scrub/agheader.c  |   78 +++++++++++++++++++
 fs/xfs/scrub/alloc.c     |    9 ++
 fs/xfs/scrub/bmap.c      |   92 +++++++++++++++++++++++
 fs/xfs/scrub/btree.c     |   10 ++
 fs/xfs/scrub/ialloc.c    |   30 +++++++
 fs/xfs/scrub/inode.c     |   30 +++++++
 fs/xfs/scrub/refcount.c  |  188 ++++++++++++++++++++++++++++++++++++++++++++++
 9 files changed, 500 insertions(+)


diff --git a/fs/xfs/libxfs/xfs_rmap.c b/fs/xfs/libxfs/xfs_rmap.c
index c7d5102..cce51cb 100644
--- a/fs/xfs/libxfs/xfs_rmap.c
+++ b/fs/xfs/libxfs/xfs_rmap.c
@@ -2306,3 +2306,61 @@ xfs_rmap_free_extent(
 	return __xfs_rmap_add(mp, dfops, XFS_RMAP_FREE, owner,
 			XFS_DATA_FORK, &bmap);
 }
+
+/* Is there a record covering a given extent? */
+int
+xfs_rmap_has_record(
+	struct xfs_btree_cur	*cur,
+	xfs_fsblock_t		bno,
+	xfs_filblks_t		len,
+	bool			*exists)
+{
+	union xfs_btree_irec	low;
+	union xfs_btree_irec	high;
+
+	memset(&low, 0, sizeof(low));
+	low.r.rm_startblock = bno;
+	memset(&high, 0xFF, sizeof(high));
+	high.r.rm_startblock = bno + len - 1;
+
+	return xfs_btree_has_record(cur, &low, &high, exists);
+}
+
+/* Is there a record covering a given extent? */
+int
+xfs_rmap_record_exists(
+	struct xfs_btree_cur	*cur,
+	xfs_fsblock_t		bno,
+	xfs_filblks_t		len,
+	struct xfs_owner_info	*oinfo,
+	bool			*has_rmap)
+{
+	uint64_t		owner;
+	uint64_t		offset;
+	unsigned int		flags;
+	int			stat;
+	struct xfs_rmap_irec	irec;
+	int			error;
+
+	xfs_owner_info_unpack(oinfo, &owner, &offset, &flags);
+
+	error = xfs_rmap_lookup_le(cur, bno, len, owner, offset, flags, &stat);
+	if (error)
+		return error;
+	if (!stat) {
+		*has_rmap = false;
+		return 0;
+	}
+
+	error = xfs_rmap_get_rec(cur, &irec, &stat);
+	if (error)
+		return error;
+	if (!stat) {
+		*has_rmap = false;
+		return 0;
+	}
+
+	*has_rmap = (irec.rm_owner == owner && irec.rm_startblock <= bno &&
+		     irec.rm_startblock + irec.rm_blockcount >= bno + len);
+	return 0;
+}
diff --git a/fs/xfs/libxfs/xfs_rmap.h b/fs/xfs/libxfs/xfs_rmap.h
index 3fa4559..ea359ab 100644
--- a/fs/xfs/libxfs/xfs_rmap.h
+++ b/fs/xfs/libxfs/xfs_rmap.h
@@ -217,5 +217,10 @@ int xfs_rmap_lookup_le_range(struct xfs_btree_cur *cur, xfs_agblock_t bno,
 union xfs_btree_rec;
 int xfs_rmap_btrec_to_irec(union xfs_btree_rec *rec,
 		struct xfs_rmap_irec *irec);
+int xfs_rmap_has_record(struct xfs_btree_cur *cur, xfs_fsblock_t bno,
+		xfs_filblks_t len, bool *exists);
+int xfs_rmap_record_exists(struct xfs_btree_cur *cur, xfs_fsblock_t bno,
+		xfs_filblks_t len, struct xfs_owner_info *oinfo,
+		bool *has_rmap);
 
 #endif	/* __XFS_RMAP_H__ */
diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index 505b75f..cf0093a 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -33,6 +33,7 @@
 #include "xfs_inode.h"
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
+#include "xfs_rmap.h"
 #include "scrub/common.h"
 
 /* Set us up to check an AG header. */
@@ -159,10 +160,12 @@ xfs_scrub_superblock(
 	struct xfs_buf			*bp;
 	struct xfs_scrub_ag		*psa;
 	struct xfs_sb			sb;
+	struct xfs_owner_info		oinfo;
 	xfs_agnumber_t			agno;
 	uint32_t			v2_ok;
 	bool				is_freesp;
 	bool				has_inodes;
+	bool				has_rmap;
 	int				error;
 	int				err2;
 
@@ -324,6 +327,15 @@ xfs_scrub_superblock(
 			XFS_SCRUB_SB_XCHECK(!has_inodes);
 	}
 
+	/* Cross-reference with the rmapbt. */
+	if (psa->rmap_cur) {
+		xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_FS);
+		err2 = xfs_rmap_record_exists(psa->rmap_cur, XFS_SB_BLOCK(mp),
+				1, &oinfo, &has_rmap);
+		if (xfs_scrub_should_xref(sc, err2, &psa->rmap_cur))
+			XFS_SCRUB_SB_XCHECK(has_rmap);
+	}
+
 out:
 	return error;
 }
@@ -358,6 +370,7 @@ int
 xfs_scrub_agf(
 	struct xfs_scrub_context	*sc)
 {
+	struct xfs_owner_info		oinfo;
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 	struct xfs_agf			*agf;
 	struct xfs_scrub_ag		*psa;
@@ -371,8 +384,10 @@ xfs_scrub_agf(
 	xfs_agblock_t			agfl_count;
 	xfs_agblock_t			fl_count;
 	xfs_extlen_t			blocks;
+	xfs_extlen_t			btreeblks = 0;
 	bool				is_freesp;
 	bool				has_inodes;
+	bool				has_rmap;
 	int				have;
 	int				level;
 	int				error = 0;
@@ -508,6 +523,37 @@ xfs_scrub_agf(
 			XFS_SCRUB_AGF_XCHECK(!has_inodes);
 	}
 
+	/* Cross-reference with the rmapbt. */
+	if (psa->rmap_cur) {
+		xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_FS);
+		err2 = xfs_rmap_record_exists(psa->rmap_cur, XFS_AGF_BLOCK(mp),
+				1, &oinfo, &has_rmap);
+		if (xfs_scrub_should_xref(sc, err2, &psa->rmap_cur))
+			XFS_SCRUB_AGF_XCHECK(has_rmap);
+	}
+	if (psa->rmap_cur) {
+		err2 = xfs_btree_count_blocks(psa->rmap_cur, &blocks);
+		if (xfs_scrub_should_xref(sc, err2, &psa->rmap_cur)) {
+			btreeblks = blocks - 1;
+			XFS_SCRUB_AGF_XCHECK(blocks == be32_to_cpu(
+					agf->agf_rmap_blocks));
+		}
+	}
+
+	/* Check btreeblks */
+	if ((!xfs_sb_version_hasrmapbt(&mp->m_sb) || psa->rmap_cur) &&
+	    psa->bno_cur && psa->cnt_cur) {
+		err2 = xfs_btree_count_blocks(psa->bno_cur, &blocks);
+		if (xfs_scrub_should_xref(sc, err2, &psa->bno_cur))
+			btreeblks += blocks - 1;
+		err2 = xfs_btree_count_blocks(psa->cnt_cur, &blocks);
+		if (xfs_scrub_should_xref(sc, err2, &psa->cnt_cur))
+			btreeblks += blocks - 1;
+		if (psa->bno_cur && psa->cnt_cur)
+			XFS_SCRUB_AGF_XCHECK(btreeblks == be32_to_cpu(
+					agf->agf_btreeblks));
+	}
+
 out:
 	return error;
 }
@@ -522,6 +568,7 @@ xfs_scrub_agf(
 #define XFS_SCRUB_AGFL_XCHECK(fs_ok) \
 	XFS_SCRUB_XCHECK(sc, sc->sa.agfl_bp, "AGFL", fs_ok)
 struct xfs_scrub_agfl {
+	struct xfs_owner_info		oinfo;
 	xfs_agblock_t			eoag;
 	xfs_daddr_t			eofs;
 };
@@ -538,6 +585,7 @@ xfs_scrub_agfl_block(
 	struct xfs_scrub_agfl		*sagfl = priv;
 	bool				is_freesp;
 	bool				has_inodes;
+	bool				has_rmap;
 	int				err2;
 
 	XFS_SCRUB_AGFL_CHECK(agbno > XFS_AGI_BLOCK(mp));
@@ -572,6 +620,14 @@ xfs_scrub_agfl_block(
 			XFS_SCRUB_AGFL_XCHECK(!has_inodes);
 	}
 
+	/* Cross-reference with the rmapbt. */
+	if (sc->sa.rmap_cur) {
+		err2 = xfs_rmap_record_exists(sc->sa.rmap_cur, agbno, 1,
+				&sagfl->oinfo, &has_rmap);
+		if (xfs_scrub_should_xref(sc, err2, &sc->sa.rmap_cur))
+			XFS_SCRUB_AGFL_XCHECK(has_rmap);
+	}
+
 	return 0;
 }
 
@@ -588,6 +644,7 @@ xfs_scrub_agfl(
 	struct xfs_agf			*agf;
 	bool				is_freesp;
 	bool				has_inodes;
+	bool				has_rmap;
 	int				error;
 	int				err2;
 
@@ -625,7 +682,17 @@ xfs_scrub_agfl(
 			XFS_SCRUB_AGFL_XCHECK(!has_inodes);
 	}
 
+	/* Set up cross-reference with rmapbt. */
+	if (sc->sa.rmap_cur) {
+		xfs_rmap_ag_owner(&sagfl.oinfo, XFS_RMAP_OWN_FS);
+		err2 = xfs_rmap_record_exists(sc->sa.rmap_cur,
+				XFS_AGFL_BLOCK(mp), 1, &sagfl.oinfo, &has_rmap);
+		if (xfs_scrub_should_xref(sc, err2, &sc->sa.rmap_cur))
+			XFS_SCRUB_AGFL_XCHECK(has_rmap);
+	}
+
 	/* Check the blocks in the AGFL. */
+	xfs_rmap_ag_owner(&sagfl.oinfo, XFS_RMAP_OWN_AG);
 	return xfs_scrub_walk_agfl(sc, xfs_scrub_agfl_block, &sagfl);
 out:
 	return error;
@@ -648,6 +715,7 @@ int
 xfs_scrub_agi(
 	struct xfs_scrub_context	*sc)
 {
+	struct xfs_owner_info		oinfo;
 	struct xfs_mount		*mp = sc->tp->t_mountp;
 	struct xfs_agi			*agi;
 	struct xfs_scrub_ag		*psa;
@@ -663,6 +731,7 @@ xfs_scrub_agi(
 	xfs_agino_t			freecount;
 	bool				is_freesp;
 	bool				has_inodes;
+	bool				has_rmap;
 	int				i;
 	int				level;
 	int				error = 0;
@@ -774,6 +843,15 @@ xfs_scrub_agi(
 			XFS_SCRUB_AGI_XCHECK(!has_inodes);
 	}
 
+	/* Cross-reference with the rmapbt. */
+	if (psa->rmap_cur) {
+		xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_FS);
+		err2 = xfs_rmap_record_exists(psa->rmap_cur, XFS_AGI_BLOCK(mp),
+				1, &oinfo, &has_rmap);
+		if (xfs_scrub_should_xref(sc, err2, &psa->rmap_cur))
+			XFS_SCRUB_AGI_XCHECK(has_rmap);
+	}
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/alloc.c b/fs/xfs/scrub/alloc.c
index 90b988d..2cad0e4 100644
--- a/fs/xfs/scrub/alloc.c
+++ b/fs/xfs/scrub/alloc.c
@@ -73,6 +73,7 @@ xfs_scrub_allocbt_helper(
 	xfs_agblock_t			bno;
 	xfs_extlen_t			flen;
 	xfs_extlen_t			len;
+	bool				has_rmap;
 	bool				has_inodes;
 	int				has_otherrec;
 	int				error = 0;
@@ -133,6 +134,14 @@ xfs_scrub_allocbt_helper(
 			XFS_SCRUB_BTREC_XCHECK(bs, !has_inodes);
 	}
 
+	/* Cross-reference with the rmapbt. */
+	if (psa->rmap_cur) {
+		err2 = xfs_rmap_has_record(psa->rmap_cur, bno, len,
+				&has_rmap);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->rmap_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, !has_rmap);
+	}
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/bmap.c b/fs/xfs/scrub/bmap.c
index 263967a..843fb3c 100644
--- a/fs/xfs/scrub/bmap.c
+++ b/fs/xfs/scrub/bmap.c
@@ -91,6 +91,8 @@ struct xfs_scrub_bmap_info {
 	XFS_SCRUB_OP_ERROR_XGOTO(info->sc, agno, 0, "bmap", &error, label);
 #define XFS_SCRUB_BMAP_XCHECK(fs_ok) \
 	XFS_SCRUB_INO_XCHECK(info->sc, info->sc->ip->i_ino, bp, info->type, fs_ok)
+#define XFS_SCRUB_BMAP_XGOTO(fs_ok, label) \
+	XFS_SCRUB_INO_XGOTO(info->sc, info->sc->ip->i_ino, bp, info->type, fs_ok, label)
 /* Scrub a single extent record. */
 STATIC int
 xfs_scrub_bmap_extent(
@@ -106,8 +108,13 @@ xfs_scrub_bmap_extent(
 	xfs_daddr_t			dlen;
 	xfs_agnumber_t			agno;
 	xfs_fsblock_t			bno;
+	struct xfs_rmap_irec		rmap;
+	uint64_t			owner;
+	xfs_fileoff_t			offset;
 	bool				is_freesp;
 	bool				has_inodes;
+	unsigned int			rflags;
+	int				has_rmap;
 	int				error = 0;
 	int				err2 = 0;
 
@@ -180,11 +187,96 @@ xfs_scrub_bmap_extent(
 			XFS_SCRUB_BMAP_XCHECK(!has_inodes);
 	}
 
+	/* Cross-reference with rmapbt. */
+	if (sa.rmap_cur) {
+		if (info->whichfork == XFS_COW_FORK) {
+			owner = XFS_RMAP_OWN_COW;
+			offset = 0;
+		} else {
+			owner = ip->i_ino;
+			offset = irec->br_startoff;
+		}
+
+		/* Look for a corresponding rmap. */
+		rflags = 0;
+		if (info->whichfork == XFS_ATTR_FORK)
+			rflags |= XFS_RMAP_ATTR_FORK;
+
+		if (info->is_shared) {
+			err2 = xfs_rmap_lookup_le_range(sa.rmap_cur, bno, owner,
+					offset, rflags, &rmap,
+					&has_rmap);
+			if (xfs_scrub_should_xref(info->sc, err2,
+					&sa.rmap_cur)) {
+				XFS_SCRUB_BMAP_XGOTO(has_rmap, skip_rmap_xref);
+			} else
+				goto skip_rmap_xref;
+		} else {
+			err2 = xfs_rmap_lookup_le(sa.rmap_cur, bno, 0, owner,
+					offset, rflags, &has_rmap);
+			if (xfs_scrub_should_xref(info->sc, err2,
+					&sa.rmap_cur)) {
+				XFS_SCRUB_BMAP_XGOTO(has_rmap, skip_rmap_xref);
+			} else
+				goto skip_rmap_xref;
+
+			err2 = xfs_rmap_get_rec(sa.rmap_cur, &rmap,
+					&has_rmap);
+			if (xfs_scrub_should_xref(info->sc, err2,
+					&sa.rmap_cur)) {
+				XFS_SCRUB_BMAP_XGOTO(has_rmap, skip_rmap_xref);
+			} else
+				goto skip_rmap_xref;
+		}
+
+		/* Check the rmap. */
+		XFS_SCRUB_BMAP_XCHECK(rmap.rm_startblock <= bno);
+		XFS_SCRUB_BMAP_XCHECK(rmap.rm_startblock <
+				rmap.rm_startblock + rmap.rm_blockcount);
+		XFS_SCRUB_BMAP_XCHECK(bno + irec->br_blockcount <=
+				rmap.rm_startblock + rmap.rm_blockcount);
+		if (owner != XFS_RMAP_OWN_COW) {
+			XFS_SCRUB_BMAP_XCHECK(rmap.rm_offset <= offset);
+			XFS_SCRUB_BMAP_XCHECK(rmap.rm_offset <
+					rmap.rm_offset + rmap.rm_blockcount);
+			XFS_SCRUB_BMAP_XCHECK(offset + irec->br_blockcount <=
+					rmap.rm_offset + rmap.rm_blockcount);
+		}
+		XFS_SCRUB_BMAP_XCHECK(rmap.rm_owner == owner);
+		switch (irec->br_state) {
+		case XFS_EXT_UNWRITTEN:
+			XFS_SCRUB_BMAP_XCHECK(
+					rmap.rm_flags & XFS_RMAP_UNWRITTEN);
+			break;
+		case XFS_EXT_NORM:
+			XFS_SCRUB_BMAP_XCHECK(
+					!(rmap.rm_flags & XFS_RMAP_UNWRITTEN));
+			break;
+		default:
+			break;
+		}
+		switch (info->whichfork) {
+		case XFS_ATTR_FORK:
+			XFS_SCRUB_BMAP_XCHECK(
+					rmap.rm_flags & XFS_RMAP_ATTR_FORK);
+			break;
+		case XFS_DATA_FORK:
+		case XFS_COW_FORK:
+			XFS_SCRUB_BMAP_XCHECK(
+					!(rmap.rm_flags & XFS_RMAP_ATTR_FORK));
+			break;
+		}
+		XFS_SCRUB_BMAP_XCHECK(!(rmap.rm_flags & XFS_RMAP_BMBT_BLOCK));
+skip_rmap_xref:
+		;
+	}
+
 	xfs_scrub_ag_free(&sa);
 out:
 	info->lastoff = irec->br_startoff + irec->br_blockcount;
 	return error;
 }
+#undef XFS_SCRUB_BMAP_XGOTO
 #undef XFS_SCRUB_BMAP_OP_ERROR_GOTO
 #undef XFS_SCRUB_BMAP_GOTO
 
diff --git a/fs/xfs/scrub/btree.c b/fs/xfs/scrub/btree.c
index 30360e0..78e7ec4 100644
--- a/fs/xfs/scrub/btree.c
+++ b/fs/xfs/scrub/btree.c
@@ -32,6 +32,7 @@
 #include "xfs_sb.h"
 #include "xfs_inode.h"
 #include "xfs_alloc.h"
+#include "xfs_rmap.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -515,6 +516,7 @@ xfs_scrub_btree_check_block_owner(
 	xfs_agnumber_t			agno;
 	xfs_agblock_t			bno;
 	bool				is_freesp;
+	bool				has_rmap;
 	int				error = 0;
 	int				err2;
 
@@ -538,6 +540,14 @@ xfs_scrub_btree_check_block_owner(
 			XFS_SCRUB_BTREC_XCHECK(bs, !is_freesp);
 	}
 
+	/* Check that there's an rmap for this. */
+	if (psa->rmap_cur) {
+		err2 = xfs_rmap_record_exists(psa->rmap_cur, bno, 1, bs->oinfo,
+				&has_rmap);
+		if (xfs_scrub_btree_should_xref(bs, err2, NULL))
+			XFS_SCRUB_BTREC_XCHECK(bs, has_rmap);
+	}
+
 	if (bs->cur->bc_flags & XFS_BTREE_LONG_PTRS)
 		xfs_scrub_ag_free(&sa);
 
diff --git a/fs/xfs/scrub/ialloc.c b/fs/xfs/scrub/ialloc.c
index 585e4c3..772d97b 100644
--- a/fs/xfs/scrub/ialloc.c
+++ b/fs/xfs/scrub/ialloc.c
@@ -86,16 +86,19 @@ xfs_scrub_iallocbt_chunk(
 	struct xfs_agf			*agf;
 	struct xfs_scrub_ag		*psa;
 	struct xfs_btree_cur		**xcur;
+	struct xfs_owner_info		oinfo;
 	xfs_agblock_t			eoag;
 	xfs_agblock_t			bno;
 	bool				is_freesp;
 	bool				has_inodes;
+	bool				has_rmap;
 	int				error = 0;
 	int				err2;
 
 	agf = XFS_BUF_TO_AGF(bs->sc->sa.agf_bp);
 	eoag = be32_to_cpu(agf->agf_length);
 	bno = XFS_AGINO_TO_AGBNO(mp, agino);
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INODES);
 
 	*keep_scanning = true;
 	XFS_SCRUB_BTREC_CHECK(bs, bno < mp->m_sb.sb_agblocks);
@@ -137,6 +140,14 @@ xfs_scrub_iallocbt_chunk(
 			XFS_SCRUB_BTREC_XCHECK(bs, has_inodes);
 	}
 
+	/* Cross-reference with rmapbt. */
+	if (psa->rmap_cur) {
+		err2 = xfs_rmap_record_exists(psa->rmap_cur, bno,
+				len, &oinfo, &has_rmap);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->rmap_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, has_rmap);
+	}
+
 out:
 	return error;
 }
@@ -211,6 +222,7 @@ xfs_scrub_iallocbt_check_freemask(
 	struct xfs_mount		*mp = bs->cur->bc_mp;
 	struct xfs_dinode		*dip;
 	struct xfs_buf			*bp;
+	struct xfs_scrub_ag		*psa;
 	xfs_ino_t			fsino;
 	xfs_agino_t			nr_inodes;
 	xfs_agino_t			agino;
@@ -220,12 +232,15 @@ xfs_scrub_iallocbt_check_freemask(
 	int				blks_per_cluster;
 	__uint16_t			holemask;
 	__uint16_t			ir_holemask;
+	bool				has;
 	int				error = 0;
+	int				err2;
 
 	/* Make sure the freemask matches the inode records. */
 	blks_per_cluster = xfs_icluster_size_fsb(mp);
 	nr_inodes = XFS_OFFBNO_TO_AGINO(mp, blks_per_cluster, 0);
 	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INODES);
+	psa = &bs->sc->sa;
 
 	for (agino = irec->ir_startino;
 	     agino < irec->ir_startino + XFS_INODES_PER_CHUNK;
@@ -245,6 +260,21 @@ xfs_scrub_iallocbt_check_freemask(
 		XFS_SCRUB_BTREC_CHECK(bs, ir_holemask == holemask ||
 				ir_holemask == 0);
 
+		/* Does the rmap agree that we have inodes here? */
+		if (psa->rmap_cur) {
+			err2 = xfs_rmap_record_exists(psa->rmap_cur, agbno,
+					blks_per_cluster, &oinfo, &has);
+			if (!xfs_scrub_btree_should_xref(bs, err2,
+					&psa->rmap_cur))
+				goto skip_xref;
+			if (has)
+				XFS_SCRUB_BTREC_XCHECK(bs, ir_holemask == 0);
+			else
+				XFS_SCRUB_BTREC_XCHECK(bs,
+						ir_holemask == holemask);
+		}
+
+skip_xref:
 		/* If any part of this is a hole, skip it. */
 		if (ir_holemask)
 			continue;
diff --git a/fs/xfs/scrub/inode.c b/fs/xfs/scrub/inode.c
index b9e7060..8b625b5 100644
--- a/fs/xfs/scrub/inode.c
+++ b/fs/xfs/scrub/inode.c
@@ -38,6 +38,7 @@
 #include "xfs_ialloc.h"
 #include "xfs_log.h"
 #include "xfs_trans_priv.h"
+#include "xfs_rmap.h"
 #include "scrub/common.h"
 
 /*
@@ -162,6 +163,11 @@ xfs_scrub_setup_inode_raw(
 			XFS_INO_TO_AGBNO(mp, ino), "inode", &error, label);
 #define XFS_SCRUB_INODE_PREEN(fs_ok) \
 	XFS_SCRUB_INO_PREEN(sc, bp, "inode", fs_ok);
+#define XFS_SCRUB_INODE_XCHECK(fs_ok) \
+	XFS_SCRUB_INO_XCHECK(sc, ino, bp, "inode", fs_ok);
+#define XFS_SCRUB_INODE_OP_ERROR_XGOTO(label) \
+	XFS_SCRUB_OP_ERROR_XGOTO(sc, XFS_INO_TO_AGNO(mp, ino), \
+			XFS_INO_TO_AGBNO(mp, ino), "inode", &error, label);
 /* Scrub an inode. */
 int
 xfs_scrub_inode(
@@ -182,6 +188,7 @@ xfs_scrub_inode(
 	uint16_t			flags;
 	uint16_t			mode;
 	int				error = 0;
+	int				err2;
 
 	/* Did we get the in-core inode, or are we doing this manually? */
 	if (sc->ip) {
@@ -368,11 +375,34 @@ xfs_scrub_inode(
 		XFS_SCRUB_INODE_PREEN(ifp->if_bytes > 0);
 	}
 
+	/* Make sure the rmap thinks there's an inode here. */
+	if (xfs_sb_version_hasrmapbt(&mp->m_sb)) {
+		struct xfs_owner_info		oinfo;
+		struct xfs_scrub_ag		sa = {0};
+		xfs_agnumber_t			agno;
+		xfs_agblock_t			agbno;
+		bool				has_rmap;
+
+		agno = XFS_INO_TO_AGNO(mp, ino);
+		agbno = XFS_INO_TO_AGBNO(mp, ino);
+		xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INODES);
+		error = xfs_scrub_ag_init(sc, agno, &sa);
+		XFS_SCRUB_INODE_OP_ERROR_XGOTO(out);
+
+		err2 = xfs_rmap_record_exists(sa.rmap_cur, agbno,
+				1, &oinfo, &has_rmap);
+		if (xfs_scrub_should_xref(sc, err2, &sa.rmap_cur))
+			XFS_SCRUB_INODE_XCHECK(has_rmap);
+		xfs_scrub_ag_free(&sa);
+	}
+
 out:
 	if (bp)
 		xfs_trans_brelse(sc->tp, bp);
 	return error;
 }
+#undef XFS_SCRUB_INODE_OP_ERROR_XGOTO
+#undef XFS_SCRUB_INODE_XCHECK
 #undef XFS_SCRUB_INODE_PREEN
 #undef XFS_SCRUB_INODE_OP_ERROR_GOTO
 #undef XFS_SCRUB_INODE_GOTO
diff --git a/fs/xfs/scrub/refcount.c b/fs/xfs/scrub/refcount.c
index 31a531f..2194be9 100644
--- a/fs/xfs/scrub/refcount.c
+++ b/fs/xfs/scrub/refcount.c
@@ -38,6 +38,163 @@
 
 /* Reference count btree scrubber. */
 
+struct xfs_scrub_refcountbt_fragment {
+	struct xfs_rmap_irec		rm;
+	struct list_head		list;
+};
+
+struct xfs_scrub_refcountbt_rmap_check_info {
+	struct xfs_scrub_btree		*bs;
+	xfs_nlink_t			nr;
+	struct xfs_refcount_irec	rc;
+	struct list_head		fragments;
+};
+
+/*
+ * Decide if the given rmap is large enough that we can redeem it
+ * towards refcount verification now, or if it's a fragment, in
+ * which case we'll hang onto it in the hopes that we'll later
+ * discover that we've collected exactly the correct number of
+ * fragments as the refcountbt says we should have.
+ */
+STATIC int
+xfs_scrub_refcountbt_rmap_check(
+	struct xfs_btree_cur		*cur,
+	struct xfs_rmap_irec		*rec,
+	void				*priv)
+{
+	struct xfs_scrub_refcountbt_rmap_check_info	*rsrci = priv;
+	struct xfs_scrub_refcountbt_fragment		*frag;
+	xfs_agblock_t			rm_last;
+	xfs_agblock_t			rc_last;
+
+	rm_last = rec->rm_startblock + rec->rm_blockcount;
+	rc_last = rsrci->rc.rc_startblock + rsrci->rc.rc_blockcount;
+	XFS_SCRUB_BTREC_XCHECK(rsrci->bs, rsrci->rc.rc_refcount != 1 ||
+			rec->rm_owner == XFS_RMAP_OWN_COW);
+	if (rec->rm_startblock <= rsrci->rc.rc_startblock && rm_last >= rc_last)
+		rsrci->nr++;
+	else {
+		frag = kmem_zalloc(sizeof(struct xfs_scrub_refcountbt_fragment),
+				KM_SLEEP);
+		frag->rm = *rec;
+		list_add_tail(&frag->list, &rsrci->fragments);
+	}
+
+	return 0;
+}
+
+/*
+ * Given a bunch of rmap fragments, iterate through them, keeping
+ * a running tally of the refcount.  If this ever deviates from
+ * what we expect (which is the refcountbt's refcount minus the
+ * number of extents that totally covered the refcountbt extent),
+ * we have a refcountbt error.
+ */
+STATIC void
+xfs_scrub_refcountbt_process_rmap_fragments(
+	struct xfs_mount				*mp,
+	struct xfs_scrub_refcountbt_rmap_check_info	*rsrci)
+{
+	struct list_head				worklist;
+	struct xfs_scrub_refcountbt_fragment		*cur;
+	struct xfs_scrub_refcountbt_fragment		*n;
+	xfs_agblock_t					bno;
+	xfs_agblock_t					rbno;
+	xfs_agblock_t					next_rbno;
+	xfs_nlink_t					nr;
+	xfs_nlink_t					target_nr;
+
+	target_nr = rsrci->rc.rc_refcount - rsrci->nr;
+	if (target_nr == 0)
+		return;
+
+	/*
+	 * There are (rsrci->rc.rc_refcount - rsrci->nr refcount)
+	 * references we haven't found yet.  Pull that many off the
+	 * fragment list and figure out where the smallest rmap ends
+	 * (and therefore the next rmap should start).  All the rmaps
+	 * we pull off should start at or before the beginning of the
+	 * refcount record's range.
+	 */
+	INIT_LIST_HEAD(&worklist);
+	rbno = NULLAGBLOCK;
+	nr = 1;
+	list_for_each_entry_safe(cur, n, &rsrci->fragments, list) {
+		if (cur->rm.rm_startblock > rsrci->rc.rc_startblock)
+			goto fail;
+		bno = cur->rm.rm_startblock + cur->rm.rm_blockcount;
+		if (rbno > bno)
+			rbno = bno;
+		list_del(&cur->list);
+		list_add_tail(&cur->list, &worklist);
+		if (nr == target_nr)
+			break;
+		nr++;
+	}
+
+	if (nr != target_nr)
+		goto fail;
+
+	while (!list_empty(&rsrci->fragments)) {
+		/* Discard any fragments ending at rbno. */
+		nr = 0;
+		next_rbno = NULLAGBLOCK;
+		list_for_each_entry_safe(cur, n, &worklist, list) {
+			bno = cur->rm.rm_startblock + cur->rm.rm_blockcount;
+			if (bno != rbno) {
+				if (next_rbno > bno)
+					next_rbno = bno;
+				continue;
+			}
+			list_del(&cur->list);
+			kmem_free(cur);
+			nr++;
+		}
+
+		/* Empty list?  We're done. */
+		if (list_empty(&rsrci->fragments))
+			break;
+
+		/* Try to add nr rmaps starting at rbno to the worklist. */
+		list_for_each_entry_safe(cur, n, &rsrci->fragments, list) {
+			bno = cur->rm.rm_startblock + cur->rm.rm_blockcount;
+			if (cur->rm.rm_startblock != rbno)
+				goto fail;
+			list_del(&cur->list);
+			list_add_tail(&cur->list, &worklist);
+			if (next_rbno > bno)
+				next_rbno = bno;
+			nr--;
+			if (nr == 0)
+				break;
+		}
+
+		rbno = next_rbno;
+	}
+
+	/*
+	 * Make sure the last extent we processed ends at or beyond
+	 * the end of the refcount extent.
+	 */
+	if (rbno < rsrci->rc.rc_startblock + rsrci->rc.rc_blockcount)
+		goto fail;
+
+	rsrci->nr = rsrci->rc.rc_refcount;
+fail:
+	/* Delete fragments and work list. */
+	list_for_each_entry_safe(cur, n, &worklist, list) {
+		list_del(&cur->list);
+		kmem_free(cur);
+	}
+	list_for_each_entry_safe(cur, n, &rsrci->fragments, list) {
+		cur = list_first_entry(&rsrci->fragments,
+				struct xfs_scrub_refcountbt_fragment, list);
+		list_del(&cur->list);
+		kmem_free(cur);
+	}
+}
+
 /* Scrub a refcountbt record. */
 STATIC int
 xfs_scrub_refcountbt_helper(
@@ -48,6 +205,11 @@ xfs_scrub_refcountbt_helper(
 	struct xfs_agf			*agf;
 	struct xfs_scrub_ag		*psa;
 	struct xfs_refcount_irec	irec;
+	struct xfs_rmap_irec		low;
+	struct xfs_rmap_irec		high;
+	struct xfs_scrub_refcountbt_rmap_check_info	rsrci;
+	struct xfs_scrub_refcountbt_fragment		*cur;
+	struct xfs_scrub_refcountbt_fragment		*n;
 	xfs_agblock_t			eoag;
 	bool				has_cowflag;
 	bool				is_freesp;
@@ -109,6 +271,32 @@ xfs_scrub_refcountbt_helper(
 			XFS_SCRUB_BTREC_XCHECK(bs, !has_inodes);
 	}
 
+	/* Cross-reference with the rmapbt to confirm the refcount. */
+	if (psa->rmap_cur) {
+		memset(&low, 0, sizeof(low));
+		low.rm_startblock = irec.rc_startblock;
+		memset(&high, 0xFF, sizeof(high));
+		high.rm_startblock = irec.rc_startblock +
+				irec.rc_blockcount - 1;
+
+		rsrci.bs = bs;
+		rsrci.nr = 0;
+		rsrci.rc = irec;
+		INIT_LIST_HEAD(&rsrci.fragments);
+		err2 = xfs_rmap_query_range(psa->rmap_cur, &low, &high,
+				&xfs_scrub_refcountbt_rmap_check, &rsrci);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->rmap_cur)) {
+			xfs_scrub_refcountbt_process_rmap_fragments(mp, &rsrci);
+			XFS_SCRUB_BTREC_XCHECK(bs,
+					irec.rc_refcount == rsrci.nr);
+		}
+
+		list_for_each_entry_safe(cur, n, &rsrci.fragments, list) {
+			list_del(&cur->list);
+			kmem_free(cur);
+		}
+	}
+
 out:
 	return error;
 }

[PATCH 38/55] xfs: cross-reference refcount btree during scrub

[Darrick J. Wong]

During metadata btree scrub, we should cross-reference with the
reference counts.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_refcount.c |   19 +++++++++++++
 fs/xfs/libxfs/xfs_refcount.h |    3 ++
 fs/xfs/scrub/agheader.c      |   52 ++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/alloc.c         |   10 +++++++
 fs/xfs/scrub/bmap.c          |   57 ++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/ialloc.c        |   10 +++++++
 fs/xfs/scrub/rmap.c          |   60 ++++++++++++++++++++++++++++++++++++++++++
 7 files changed, 211 insertions(+)


diff --git a/fs/xfs/libxfs/xfs_refcount.c b/fs/xfs/libxfs/xfs_refcount.c
index b177ef3..c6c875d 100644
--- a/fs/xfs/libxfs/xfs_refcount.c
+++ b/fs/xfs/libxfs/xfs_refcount.c
@@ -1696,3 +1696,22 @@ xfs_refcount_recover_cow_leftovers(
 	xfs_trans_cancel(tp);
 	goto out_free;
 }
+
+/* Is there a record covering a given extent? */
+int
+xfs_refcount_has_record(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	xfs_extlen_t		len,
+	bool			*exists)
+{
+	union xfs_btree_irec	low;
+	union xfs_btree_irec	high;
+
+	memset(&low, 0, sizeof(low));
+	low.rc.rc_startblock = bno;
+	memset(&high, 0xFF, sizeof(high));
+	high.rc.rc_startblock = bno + len - 1;
+
+	return xfs_btree_has_record(cur, &low, &high, exists);
+}
diff --git a/fs/xfs/libxfs/xfs_refcount.h b/fs/xfs/libxfs/xfs_refcount.h
index 098dc66..78cb142 100644
--- a/fs/xfs/libxfs/xfs_refcount.h
+++ b/fs/xfs/libxfs/xfs_refcount.h
@@ -67,4 +67,7 @@ extern int xfs_refcount_free_cow_extent(struct xfs_mount *mp,
 extern int xfs_refcount_recover_cow_leftovers(struct xfs_mount *mp,
 		xfs_agnumber_t agno);
 
+extern int xfs_refcount_has_record(struct xfs_btree_cur *cur,
+		xfs_agblock_t bno, xfs_extlen_t len, bool *exists);
+
 #endif	/* __XFS_REFCOUNT_H__ */
diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index cf0093a..d94129c 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -34,6 +34,7 @@
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
 #include "xfs_rmap.h"
+#include "xfs_refcount.h"
 #include "scrub/common.h"
 
 /* Set us up to check an AG header. */
@@ -166,6 +167,7 @@ xfs_scrub_superblock(
 	bool				is_freesp;
 	bool				has_inodes;
 	bool				has_rmap;
+	bool				has_refcount;
 	int				error;
 	int				err2;
 
@@ -336,6 +338,14 @@ xfs_scrub_superblock(
 			XFS_SCRUB_SB_XCHECK(has_rmap);
 	}
 
+	/* Cross-reference with the refcountbt. */
+	if (psa->refc_cur) {
+		err2 = xfs_refcount_has_record(psa->refc_cur, XFS_SB_BLOCK(mp),
+				1, &has_refcount);
+		if (xfs_scrub_should_xref(sc, err2, &psa->refc_cur))
+			XFS_SCRUB_SB_XCHECK(!has_refcount);
+	}
+
 out:
 	return error;
 }
@@ -388,6 +398,7 @@ xfs_scrub_agf(
 	bool				is_freesp;
 	bool				has_inodes;
 	bool				has_rmap;
+	bool				has_refcount;
 	int				have;
 	int				level;
 	int				error = 0;
@@ -554,6 +565,20 @@ xfs_scrub_agf(
 					agf->agf_btreeblks));
 	}
 
+	/* Cross-reference with the refcountbt. */
+	if (psa->refc_cur) {
+		err2 = xfs_refcount_has_record(psa->refc_cur, XFS_AGF_BLOCK(mp),
+				1, &has_refcount);
+		if (xfs_scrub_should_xref(sc, err2, &psa->refc_cur))
+			XFS_SCRUB_AGF_XCHECK(!has_refcount);
+	}
+	if (psa->refc_cur) {
+		err2 = xfs_btree_count_blocks(psa->refc_cur, &blocks);
+		if (xfs_scrub_should_xref(sc, err2, &psa->refc_cur))
+			XFS_SCRUB_AGF_XCHECK(blocks == be32_to_cpu(
+					agf->agf_refcount_blocks));
+	}
+
 out:
 	return error;
 }
@@ -586,6 +611,7 @@ xfs_scrub_agfl_block(
 	bool				is_freesp;
 	bool				has_inodes;
 	bool				has_rmap;
+	bool				has_refcount;
 	int				err2;
 
 	XFS_SCRUB_AGFL_CHECK(agbno > XFS_AGI_BLOCK(mp));
@@ -628,6 +654,14 @@ xfs_scrub_agfl_block(
 			XFS_SCRUB_AGFL_XCHECK(has_rmap);
 	}
 
+	/* Cross-reference with the refcountbt. */
+	if (sc->sa.refc_cur) {
+		err2 = xfs_refcount_has_record(sc->sa.refc_cur, agbno, 1,
+				&has_refcount);
+		if (xfs_scrub_should_xref(sc, err2, &sc->sa.refc_cur))
+			XFS_SCRUB_AGFL_XCHECK(!has_refcount);
+	}
+
 	return 0;
 }
 
@@ -645,6 +679,7 @@ xfs_scrub_agfl(
 	bool				is_freesp;
 	bool				has_inodes;
 	bool				has_rmap;
+	bool				has_refcount;
 	int				error;
 	int				err2;
 
@@ -691,6 +726,14 @@ xfs_scrub_agfl(
 			XFS_SCRUB_AGFL_XCHECK(has_rmap);
 	}
 
+	/* Set up cross-reference with refcountbt. */
+	if (sc->sa.refc_cur) {
+		err2 = xfs_refcount_has_record(sc->sa.refc_cur,
+				XFS_AGFL_BLOCK(mp), 1, &has_refcount);
+		if (xfs_scrub_should_xref(sc, err2, &sc->sa.refc_cur))
+			XFS_SCRUB_AGFL_XCHECK(!has_refcount);
+	}
+
 	/* Check the blocks in the AGFL. */
 	xfs_rmap_ag_owner(&sagfl.oinfo, XFS_RMAP_OWN_AG);
 	return xfs_scrub_walk_agfl(sc, xfs_scrub_agfl_block, &sagfl);
@@ -732,6 +775,7 @@ xfs_scrub_agi(
 	bool				is_freesp;
 	bool				has_inodes;
 	bool				has_rmap;
+	bool				has_refcount;
 	int				i;
 	int				level;
 	int				error = 0;
@@ -852,6 +896,14 @@ xfs_scrub_agi(
 			XFS_SCRUB_AGI_XCHECK(has_rmap);
 	}
 
+	/* Cross-reference with the refcountbt. */
+	if (psa->refc_cur) {
+		err2 = xfs_refcount_has_record(psa->refc_cur, XFS_AGI_BLOCK(mp),
+				1, &has_refcount);
+		if (xfs_scrub_should_xref(sc, err2, &psa->refc_cur))
+			XFS_SCRUB_AGI_XCHECK(!has_refcount);
+	}
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/alloc.c b/fs/xfs/scrub/alloc.c
index 2cad0e4..8d81875 100644
--- a/fs/xfs/scrub/alloc.c
+++ b/fs/xfs/scrub/alloc.c
@@ -33,6 +33,7 @@
 #include "xfs_rmap.h"
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
+#include "xfs_refcount.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -75,6 +76,7 @@ xfs_scrub_allocbt_helper(
 	xfs_extlen_t			len;
 	bool				has_rmap;
 	bool				has_inodes;
+	bool				has_refcount;
 	int				has_otherrec;
 	int				error = 0;
 	int				err2;
@@ -142,6 +144,14 @@ xfs_scrub_allocbt_helper(
 			XFS_SCRUB_BTREC_XCHECK(bs, !has_rmap);
 	}
 
+	/* Cross-reference with the refcountbt. */
+	if (psa->refc_cur) {
+		err2 = xfs_refcount_has_record(psa->refc_cur, bno, len,
+				&has_refcount);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->refc_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, !has_refcount);
+	}
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/bmap.c b/fs/xfs/scrub/bmap.c
index 843fb3c..953f77b 100644
--- a/fs/xfs/scrub/bmap.c
+++ b/fs/xfs/scrub/bmap.c
@@ -38,6 +38,7 @@
 #include "xfs_rmap.h"
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
+#include "xfs_refcount.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -109,12 +110,17 @@ xfs_scrub_bmap_extent(
 	xfs_agnumber_t			agno;
 	xfs_fsblock_t			bno;
 	struct xfs_rmap_irec		rmap;
+	struct xfs_refcount_irec	rc;
 	uint64_t			owner;
 	xfs_fileoff_t			offset;
+	xfs_agblock_t			fbno;
+	xfs_extlen_t			flen;
 	bool				is_freesp;
 	bool				has_inodes;
+	bool				has_cowflag;
 	unsigned int			rflags;
 	int				has_rmap;
+	int				has_refcount;
 	int				error = 0;
 	int				err2 = 0;
 
@@ -271,6 +277,57 @@ xfs_scrub_bmap_extent(
 		;
 	}
 
+	/*
+	 * If this is a non-shared file on a reflink filesystem,
+	 * check the refcountbt to see if the flag is wrong.
+	 */
+	if (sa.refc_cur) {
+		if (info->whichfork == XFS_COW_FORK) {
+			/* Check this CoW staging extent. */
+			err2 = xfs_refcount_lookup_le(sa.refc_cur,
+					bno + XFS_REFC_COW_START,
+					&has_refcount);
+			if (xfs_scrub_should_xref(info->sc, err2,
+					&sa.refc_cur)) {
+				XFS_SCRUB_BMAP_XGOTO(has_refcount,
+						skip_refc_xref);
+			} else
+				goto skip_refc_xref;
+
+			err2 = xfs_refcount_get_rec(sa.refc_cur, &rc,
+					&has_refcount);
+			if (xfs_scrub_should_xref(info->sc, err2,
+					&sa.refc_cur)) {
+				XFS_SCRUB_BMAP_XGOTO(has_refcount,
+						skip_refc_xref);
+			} else
+				goto skip_refc_xref;
+
+			has_cowflag = !!(rc.rc_startblock & XFS_REFC_COW_START);
+			XFS_SCRUB_BMAP_XCHECK(
+					(rc.rc_refcount == 1 && has_cowflag) ||
+					(rc.rc_refcount != 1 && !has_cowflag));
+			rc.rc_startblock &= ~XFS_REFC_COW_START;
+			XFS_SCRUB_BMAP_XCHECK(rc.rc_startblock <= bno);
+			XFS_SCRUB_BMAP_XCHECK(rc.rc_startblock <
+					rc.rc_startblock + rc.rc_blockcount);
+			XFS_SCRUB_BMAP_XCHECK(bno + irec->br_blockcount <=
+					rc.rc_startblock + rc.rc_blockcount);
+			XFS_SCRUB_BMAP_XCHECK(rc.rc_refcount == 1);
+		} else {
+			/* If this is shared, the inode flag must be set. */
+			err2 = xfs_refcount_find_shared(sa.refc_cur, bno,
+					irec->br_blockcount, &fbno, &flen,
+					false);
+			if (xfs_scrub_should_xref(info->sc, err2,
+					&sa.refc_cur))
+				XFS_SCRUB_BMAP_XCHECK(flen == 0 ||
+						xfs_is_reflink_inode(ip));
+		}
+skip_refc_xref:
+		;
+	}
+
 	xfs_scrub_ag_free(&sa);
 out:
 	info->lastoff = irec->br_startoff + irec->br_blockcount;
diff --git a/fs/xfs/scrub/ialloc.c b/fs/xfs/scrub/ialloc.c
index 772d97b..7fa4ebf 100644
--- a/fs/xfs/scrub/ialloc.c
+++ b/fs/xfs/scrub/ialloc.c
@@ -38,6 +38,7 @@
 #include "xfs_log.h"
 #include "xfs_trans_priv.h"
 #include "xfs_alloc.h"
+#include "xfs_refcount.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -92,6 +93,7 @@ xfs_scrub_iallocbt_chunk(
 	bool				is_freesp;
 	bool				has_inodes;
 	bool				has_rmap;
+	bool				has_refcount;
 	int				error = 0;
 	int				err2;
 
@@ -148,6 +150,14 @@ xfs_scrub_iallocbt_chunk(
 			XFS_SCRUB_BTREC_XCHECK(bs, has_rmap);
 	}
 
+	/* Cross-reference with the refcountbt. */
+	if (psa->refc_cur) {
+		err2 = xfs_refcount_has_record(psa->refc_cur, bno,
+				len, &has_refcount);
+		if (xfs_scrub_btree_should_xref(bs, err2, &psa->refc_cur))
+			XFS_SCRUB_BTREC_XCHECK(bs, !has_refcount);
+	}
+
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/rmap.c b/fs/xfs/scrub/rmap.c
index 30ec492..9b7ed8b 100644
--- a/fs/xfs/scrub/rmap.c
+++ b/fs/xfs/scrub/rmap.c
@@ -33,6 +33,7 @@
 #include "xfs_rmap.h"
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
+#include "xfs_refcount.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -48,13 +49,18 @@ xfs_scrub_rmapbt_helper(
 	struct xfs_agf			*agf;
 	struct xfs_scrub_ag		*psa;
 	struct xfs_rmap_irec		irec;
+	struct xfs_refcount_irec	crec;
 	xfs_agblock_t			eoag;
+	xfs_agblock_t			fbno;
+	xfs_extlen_t			flen;
 	bool				is_freesp;
 	bool				non_inode;
 	bool				is_unwritten;
 	bool				is_bmbt;
 	bool				is_attr;
 	bool				has_inodes;
+	bool				has_cowflag;
+	int				has_refcount;
 	int				error = 0;
 	int				err2;
 
@@ -144,6 +150,60 @@ xfs_scrub_rmapbt_helper(
 					!has_inodes);
 	}
 
+	/* Cross-reference with the refcount btree. */
+	if (psa->refc_cur) {
+		if (irec.rm_owner == XFS_RMAP_OWN_COW) {
+			/* Check this CoW staging extent. */
+			err2 = xfs_refcount_lookup_le(psa->refc_cur,
+					irec.rm_startblock + XFS_REFC_COW_START,
+					&has_refcount);
+			if (xfs_scrub_btree_should_xref(bs, err2,
+					&psa->refc_cur)) {
+				XFS_SCRUB_BTREC_XGOTO(bs, has_refcount,
+						skip_refc_xref);
+			} else
+				goto skip_refc_xref;
+
+			err2 = xfs_refcount_get_rec(psa->refc_cur, &crec,
+					&has_refcount);
+			if (xfs_scrub_btree_should_xref(bs, err2,
+					&psa->refc_cur)) {
+				XFS_SCRUB_BTREC_XGOTO(bs, has_refcount,
+						skip_refc_xref);
+			} else
+				goto skip_refc_xref;
+
+			has_cowflag = !!(crec.rc_startblock & XFS_REFC_COW_START);
+			XFS_SCRUB_BTREC_XCHECK(bs,
+					(crec.rc_refcount == 1 && has_cowflag) ||
+					(crec.rc_refcount != 1 && !has_cowflag));
+			crec.rc_startblock &= ~XFS_REFC_COW_START;
+			XFS_SCRUB_BTREC_XCHECK(bs, crec.rc_startblock <=
+					irec.rm_startblock);
+			XFS_SCRUB_BTREC_XCHECK(bs, crec.rc_startblock +
+					crec.rc_blockcount >
+					crec.rc_startblock);
+			XFS_SCRUB_BTREC_XCHECK(bs, crec.rc_startblock +
+					crec.rc_blockcount >=
+					irec.rm_startblock +
+					irec.rm_blockcount);
+			XFS_SCRUB_BTREC_XCHECK(bs,
+					crec.rc_refcount == 1);
+		} else {
+			/* If this is shared, the inode flag must be set. */
+			err2 = xfs_refcount_find_shared(psa->refc_cur,
+					irec.rm_startblock, irec.rm_blockcount,
+					&fbno, &flen, false);
+			if (xfs_scrub_btree_should_xref(bs, err2,
+					&psa->refc_cur))
+				XFS_SCRUB_BTREC_XCHECK(bs, flen == 0 ||
+						(!non_inode && !is_attr &&
+						 !is_bmbt && !is_unwritten));
+		}
+skip_refc_xref:
+		;
+	}
+
 out:
 	return error;
 }

[PATCH 39/55] xfs: scrub should cross-reference the realtime bitmap

[Darrick J. Wong]

While we're scrubbing various btrees, cross-reference the records
with the other metadata.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_rtbitmap.c |   30 ++++++++++++++++++++++++++++++
 fs/xfs/scrub/bmap.c          |   10 ++++++++++
 fs/xfs/xfs_rtalloc.h         |    3 +++
 3 files changed, 43 insertions(+)


diff --git a/fs/xfs/libxfs/xfs_rtbitmap.c b/fs/xfs/libxfs/xfs_rtbitmap.c
index f4b68c0..4b8457c 100644
--- a/fs/xfs/libxfs/xfs_rtbitmap.c
+++ b/fs/xfs/libxfs/xfs_rtbitmap.c
@@ -1016,3 +1016,33 @@ xfs_rtfree_extent(
 	}
 	return 0;
 }
+
+/* Is the given extent all free? */
+int
+xfs_rtbitmap_extent_is_free(
+	struct xfs_mount		*mp,
+	struct xfs_trans		*tp,
+	xfs_rtblock_t			start,
+	xfs_rtblock_t			len,
+	bool				*is_free)
+{
+	xfs_rtblock_t			end;
+	xfs_extlen_t			clen;
+	int				matches;
+	int				error;
+
+	*is_free = false;
+	while (len) {
+		clen = len > ~0U ? ~0U : len;
+		error = xfs_rtcheck_range(mp, tp, start, clen, 1, &end,
+				&matches);
+		if (error || !matches || end < start + clen)
+			return error;
+
+		len -= end - start;
+		start = end + 1;
+	}
+
+	*is_free = true;
+	return error;
+}
diff --git a/fs/xfs/scrub/bmap.c b/fs/xfs/scrub/bmap.c
index 953f77b..96dbe66 100644
--- a/fs/xfs/scrub/bmap.c
+++ b/fs/xfs/scrub/bmap.c
@@ -39,6 +39,7 @@
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
 #include "xfs_refcount.h"
+#include "xfs_rtalloc.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -118,6 +119,7 @@ xfs_scrub_bmap_extent(
 	bool				is_freesp;
 	bool				has_inodes;
 	bool				has_cowflag;
+	bool				is_free = false;
 	unsigned int			rflags;
 	int				has_rmap;
 	int				has_refcount;
@@ -173,6 +175,14 @@ xfs_scrub_bmap_extent(
 				irec->br_blockcount, &is_freesp);
 		if (xfs_scrub_should_xref(info->sc, err2, &sa.bno_cur))
 			XFS_SCRUB_BMAP_XCHECK(!is_freesp);
+	} else {
+		xfs_ilock(mp->m_rbmip, XFS_ILOCK_SHARED | XFS_ILOCK_RTBITMAP);
+		err2 = xfs_rtbitmap_extent_is_free(mp, info->sc->tp,
+				irec->br_startblock, irec->br_blockcount,
+				&is_free);
+		if (xfs_scrub_should_xref(info->sc, err2, NULL))
+			XFS_SCRUB_BMAP_XCHECK(!is_free);
+		xfs_iunlock(mp->m_rbmip, XFS_ILOCK_SHARED | XFS_ILOCK_RTBITMAP);
 	}
 
 	/* Cross-reference with inobt. */
diff --git a/fs/xfs/xfs_rtalloc.h b/fs/xfs/xfs_rtalloc.h
index 3036349..bd1c6a9 100644
--- a/fs/xfs/xfs_rtalloc.h
+++ b/fs/xfs/xfs_rtalloc.h
@@ -121,6 +121,8 @@ int xfs_rtmodify_summary(struct xfs_mount *mp, struct xfs_trans *tp, int log,
 int xfs_rtfree_range(struct xfs_mount *mp, struct xfs_trans *tp,
 		     xfs_rtblock_t start, xfs_extlen_t len,
 		     struct xfs_buf **rbpp, xfs_fsblock_t *rsb);
+int xfs_rtbitmap_extent_is_free(struct xfs_mount *mp, struct xfs_trans *tp,
+		xfs_rtblock_t start, xfs_rtblock_t len, bool *is_free);
 
 
 #else
@@ -131,6 +133,7 @@ int xfs_rtfree_range(struct xfs_mount *mp, struct xfs_trans *tp,
 # define xfs_rtcheck_range(...)                         (ENOSYS)
 # define xfs_rtfind_forw(...)                           (ENOSYS)
 # define xfs_rtbuf_get(m,t,b,i,p)                       (ENOSYS)
+# define xfs_rtbitmap_extent_is_free(m,t,s,l,i)         (ENOSYS)
 static inline int		/* error */
 xfs_rtmount_init(
 	xfs_mount_t	*mp)	/* file system mount structure */

[PATCH 40/55] xfs: cross-reference the block mappings when possible

[Darrick J. Wong]

Check inode field contents against the block mappings when possible.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/inode.c |   43 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 43 insertions(+)


diff --git a/fs/xfs/scrub/inode.c b/fs/xfs/scrub/inode.c
index 8b625b5..e79fe5b 100644
--- a/fs/xfs/scrub/inode.c
+++ b/fs/xfs/scrub/inode.c
@@ -39,6 +39,8 @@
 #include "xfs_log.h"
 #include "xfs_trans_priv.h"
 #include "xfs_rmap.h"
+#include "xfs_bmap.h"
+#include "xfs_bmap_util.h"
 #include "scrub/common.h"
 
 /*
@@ -181,6 +183,7 @@ xfs_scrub_inode(
 	struct xfs_dinode		*dip;
 	xfs_ino_t			ino;
 	unsigned long long		isize;
+	unsigned long long		count;
 	uint64_t			flags2;
 	uint32_t			nextents;
 	uint32_t			extsize;
@@ -396,6 +399,46 @@ xfs_scrub_inode(
 		xfs_scrub_ag_free(&sa);
 	}
 
+	/* Walk all the extents to check nextents/naextents/nblocks. */
+	count = 0;
+	err2 = xfs_bmap_count_blocks(sc->tp, sc->ip, XFS_DATA_FORK,
+			&nextents, &count);
+	if (!xfs_scrub_should_xref(sc, err2, NULL))
+		goto skip_block_check;
+	XFS_SCRUB_INODE_XCHECK(nextents >= be32_to_cpu(dip->di_nextents));
+
+	err2 = xfs_bmap_count_blocks(sc->tp, sc->ip, XFS_ATTR_FORK,
+			&nextents, &count);
+	if (!xfs_scrub_should_xref(sc, err2, NULL))
+		goto skip_block_check;
+	XFS_SCRUB_INODE_XCHECK(nextents == be16_to_cpu(dip->di_anextents));
+
+	/* Check nblocks, taking any delalloc extents into account. */
+	if (sc->ip)
+		count -= sc->ip->i_delayed_blks;
+	XFS_SCRUB_INODE_XCHECK(count == be64_to_cpu(dip->di_nblocks));
+
+skip_block_check:
+	/* Make sure we don't have any written extents after EOF. */
+	if (S_ISREG(mode) && !(flags & XFS_DIFLAG_PREALLOC) &&
+	    (dip->di_format == XFS_DINODE_FMT_EXTENTS ||
+	     dip->di_format == XFS_DINODE_FMT_BTREE)) {
+		struct xfs_bmbt_irec		got;
+		struct xfs_ifork		*ifp;
+		xfs_fileoff_t			lblk;
+		xfs_extnum_t			idx;
+		bool				found;
+
+		lblk = XFS_B_TO_FSB(mp, i_size_read(VFS_I(sc->ip)));
+		ifp = XFS_IFORK_PTR(sc->ip, XFS_DATA_FORK);
+		found = xfs_iext_lookup_extent(sc->ip, ifp, lblk, &idx, &got);
+		while (found) {
+			XFS_SCRUB_INODE_PREEN(got.br_startoff < lblk ||
+					got.br_state != XFS_EXT_NORM);
+			lblk = got.br_startoff + got.br_blockcount;
+			found = xfs_iext_get_extent(ifp, ++idx, &got);
+		}
+	}
 out:
 	if (bp)
 		xfs_trans_brelse(sc->tp, bp);

[PATCH 41/55] xfs: shut off scrub-related error and corruption messages

[Darrick J. Wong]

While we're scrubbing and repairing, silence errors and corruption
messages to avoid flooding the logs.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/common.c |    3 +++
 fs/xfs/xfs_buf.c      |    3 +++
 fs/xfs/xfs_error.c    |    9 +++++++++
 fs/xfs/xfs_mount.h    |    1 +
 fs/xfs/xfs_super.c    |    1 +
 5 files changed, 17 insertions(+)


diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 1c20006..b0d0ed2 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -790,6 +790,8 @@ xfs_scrub_metadata(
 		warned = true;
 	}
 
+	atomic_inc(&mp->m_scrubbers);
+
 retry_op:
 	/* Set up for the operation. */
 	error = fns->setup(&sc, ip, sm, deadlocked);
@@ -813,6 +815,7 @@ xfs_scrub_metadata(
 out_teardown:
 	error = xfs_scrub_teardown(&sc, ip, error);
 out:
+	atomic_dec(&mp->m_scrubbers);
 	trace_xfs_scrub_done(ip, sm->sm_type, sm->sm_agno, sm->sm_ino,
 			sm->sm_gen, sm->sm_flags, error);
 	return error;
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index 7f0a01f..e8d2108 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -41,6 +41,7 @@
 #include "xfs_mount.h"
 #include "xfs_trace.h"
 #include "xfs_log.h"
+#include "xfs_error.h"
 
 static kmem_zone_t *xfs_buf_zone;
 
@@ -561,6 +562,8 @@ _xfs_buf_find(
 	 */
 	eofs = XFS_FSB_TO_BB(btp->bt_mount, btp->bt_mount->m_sb.sb_dblocks);
 	if (cmap.bm_bn < 0 || cmap.bm_bn >= eofs) {
+		if (atomic_read(&btp->bt_mount->m_scrubbers) > 0)
+			return NULL;
 		/*
 		 * XXX (dgc): we should really be returning -EFSCORRUPTED here,
 		 * but none of the higher level infrastructure supports
diff --git a/fs/xfs/xfs_error.c b/fs/xfs/xfs_error.c
index ed7ee4e..955c4cc 100644
--- a/fs/xfs/xfs_error.c
+++ b/fs/xfs/xfs_error.c
@@ -132,6 +132,9 @@ xfs_error_report(
 	int			linenum,
 	void			*ra)
 {
+	if (atomic_read(&mp->m_scrubbers) > 0)
+		return;
+
 	if (level <= xfs_error_level) {
 		xfs_alert_tag(mp, XFS_PTAG_ERROR_REPORT,
 		"Internal error %s at line %d of file %s.  Caller %pS",
@@ -151,6 +154,9 @@ xfs_corruption_error(
 	int			linenum,
 	void			*ra)
 {
+	if (atomic_read(&mp->m_scrubbers) > 0)
+		return;
+
 	if (level <= xfs_error_level)
 		xfs_hex_dump(p, 64);
 	xfs_error_report(tag, level, mp, filename, linenum, ra);
@@ -167,6 +173,9 @@ xfs_verifier_error(
 {
 	struct xfs_mount *mp = bp->b_target->bt_mount;
 
+	if (atomic_read(&mp->m_scrubbers) > 0)
+		return;
+
 	xfs_alert(mp, "Metadata %s detected at %pF, %s block 0x%llx",
 		  bp->b_error == -EFSBADCRC ? "CRC error" : "corruption",
 		  __return_address, bp->b_ops->name, bp->b_bn);
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 84f7852..3883757 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -205,6 +205,7 @@ typedef struct xfs_mount {
 	 */
 	bool			m_fail_writes;
 #endif
+	atomic_t		m_scrubbers;	/* # of active scrub processes */
 } xfs_mount_t;
 
 /*
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index eecbaac..950d268 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -1538,6 +1538,7 @@ xfs_fs_fill_super(
 	spin_lock_init(&mp->m_sb_lock);
 	mutex_init(&mp->m_growlock);
 	atomic_set(&mp->m_active_trans, 0);
+	atomic_set(&mp->m_scrubbers, 0);
 	INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
 	INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
 	INIT_DELAYED_WORK(&mp->m_cowblocks_work, xfs_cowblocks_worker);

[PATCH 42/55] xfs: create tracepoints for online repair

[Darrick J. Wong]

These tracepoints will be used to debug the online repair routines.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/xfs_trace.h |  147 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 147 insertions(+)


diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index c52e3a5..becdb7a 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -3742,6 +3742,153 @@ DEFINE_SCRUB_AG_LOCK_EVENT(xfs_scrub_ag_can_lock);
 DEFINE_SCRUB_AG_LOCK_EVENT(xfs_scrub_ag_may_deadlock);
 DEFINE_SCRUB_AG_LOCK_EVENT(xfs_scrub_ag_lock_all);
 
+/* repair tracepoints */
+DEFINE_SCRUB_EVENT(xfs_repair_attempt);
+DEFINE_SCRUB_EVENT(xfs_repair_done);
+DEFINE_BUSY_EVENT(xfs_repair_free_or_unmap_extent);
+DEFINE_BUSY_EVENT(xfs_repair_collect_btree_extent);
+TRACE_EVENT(xfs_repair_init_btblock,
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agblock_t agbno,
+		 uint32_t magic),
+	TP_ARGS(mp, agno, agbno, magic),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_agblock_t, agbno)
+		__field(uint32_t, magic)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->agno = agno;
+		__entry->agbno = agbno;
+		__entry->magic = magic;
+	),
+	TP_printk("dev %d:%d agno %u agbno %u magic 0x%x",
+		  MAJOR(__entry->dev), MINOR(__entry->dev), __entry->agno,
+		  __entry->agbno, __entry->magic)
+)
+TRACE_EVENT(xfs_repair_find_ag_btree_roots_helper,
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, xfs_agblock_t agbno,
+		 uint32_t magic, uint16_t level),
+	TP_ARGS(mp, agno, agbno, magic, level),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_agblock_t, agbno)
+		__field(uint32_t, magic)
+		__field(uint16_t, level)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->agno = agno;
+		__entry->agbno = agbno;
+		__entry->magic = magic;
+		__entry->level = level;
+	),
+	TP_printk("dev %d:%d agno %u agbno %u magic 0x%x level %u",
+		  MAJOR(__entry->dev), MINOR(__entry->dev), __entry->agno,
+		  __entry->agbno, __entry->magic, __entry->level)
+)
+TRACE_EVENT(xfs_repair_calc_ag_resblks,
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno,
+		 xfs_agino_t icount, xfs_agblock_t aglen, xfs_agblock_t freelen,
+		 xfs_agblock_t usedlen),
+	TP_ARGS(mp, agno, icount, aglen, freelen, usedlen),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_agino_t, icount)
+		__field(xfs_agblock_t, aglen)
+		__field(xfs_agblock_t, freelen)
+		__field(xfs_agblock_t, usedlen)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->agno = agno;
+		__entry->icount = icount;
+		__entry->aglen = aglen;
+		__entry->freelen = freelen;
+		__entry->usedlen = usedlen;
+	),
+	TP_printk("dev %d:%d agno %d icount %u aglen %u freelen %u usedlen %u",
+		  MAJOR(__entry->dev), MINOR(__entry->dev), __entry->agno,
+		  __entry->icount, __entry->aglen, __entry->freelen,
+		  __entry->usedlen)
+)
+TRACE_EVENT(xfs_repair_calc_ag_resblks_btsize,
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno,
+		 xfs_agblock_t bnobt_sz, xfs_agblock_t inobt_sz,
+		 xfs_agblock_t rmapbt_sz, xfs_agblock_t refcbt_sz),
+	TP_ARGS(mp, agno, bnobt_sz, inobt_sz, rmapbt_sz, refcbt_sz),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_agblock_t, bnobt_sz)
+		__field(xfs_agblock_t, inobt_sz)
+		__field(xfs_agblock_t, rmapbt_sz)
+		__field(xfs_agblock_t, refcbt_sz)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->agno = agno;
+		__entry->bnobt_sz = bnobt_sz;
+		__entry->inobt_sz = inobt_sz;
+		__entry->rmapbt_sz = rmapbt_sz;
+		__entry->refcbt_sz = refcbt_sz;
+	),
+	TP_printk("dev %d:%d agno %d bno %u ino %u rmap %u refcount %u",
+		  MAJOR(__entry->dev), MINOR(__entry->dev), __entry->agno,
+		  __entry->bnobt_sz, __entry->inobt_sz, __entry->rmapbt_sz,
+		  __entry->refcbt_sz)
+)
+TRACE_EVENT(xfs_repair_reset_counters,
+	TP_PROTO(struct xfs_mount *mp),
+	TP_ARGS(mp),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+	),
+	TP_printk("dev %d:%d",
+		  MAJOR(__entry->dev), MINOR(__entry->dev))
+)
+
+DEFINE_BUSY_EVENT(xfs_repair_agfl_insert);
+DEFINE_RMAPBT_EVENT(xfs_repair_alloc_extent_fn);
+DEFINE_RMAPBT_EVENT(xfs_repair_ialloc_extent_fn);
+TRACE_EVENT(xfs_repair_ialloc_insert,
+	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno,
+		 xfs_agino_t startino, uint16_t holemask, uint8_t count,
+		 uint8_t freecount, uint64_t freemask),
+	TP_ARGS(mp, agno, startino, holemask, count, freecount, freemask),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_agnumber_t, agno)
+		__field(xfs_agino_t, startino)
+		__field(uint16_t, holemask)
+		__field(uint8_t, count)
+		__field(uint8_t, freecount)
+		__field(uint64_t, freemask)
+	),
+	TP_fast_assign(
+		__entry->dev = mp->m_super->s_dev;
+		__entry->agno = agno;
+		__entry->startino = startino;
+		__entry->holemask = holemask;
+		__entry->count = count;
+		__entry->freecount = freecount;
+		__entry->freemask = freemask;
+	),
+	TP_printk("dev %d:%d agno %d startino %u holemask 0x%x count %u freecount %u freemask 0x%llx",
+		  MAJOR(__entry->dev), MINOR(__entry->dev), __entry->agno,
+		  __entry->startino, __entry->holemask, __entry->count,
+		  __entry->freecount, __entry->freemask)
+)
+DEFINE_RMAPBT_EVENT(xfs_repair_rmap_extent_fn);
+DEFINE_REFCOUNT_EXTENT_EVENT(xfs_repair_refcount_extent_fn);
+DEFINE_RMAPBT_EVENT(xfs_repair_bmap_extent_fn);
+
 #endif /* _TRACE_XFS_H */
 
 #undef TRACE_INCLUDE_PATH

[PATCH 43/55] xfs: implement the metadata repair ioctl flag

[Darrick J. Wong]

Plumb in the pieces necessary to make the "scrub" subfunction of
the scrub ioctl actually work.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/common.c |  101 +++++++++++++++++++++++++++++++++++++++++++++++--
 fs/xfs/xfs_error.h    |    4 +-
 2 files changed, 99 insertions(+), 6 deletions(-)


diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index b0d0ed2..8b6384f 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -41,6 +41,7 @@
 #include "xfs_refcount_btree.h"
 #include "xfs_rmap.h"
 #include "xfs_rmap_btree.h"
+#include "xfs_error.h"
 #include "scrub/xfs_scrub.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
@@ -108,8 +109,43 @@
  * the metadata is correct but otherwise suboptimal, there's a "preen"
  * flag to signal that.  Finally, if we were unable to access a data
  * structure to perform cross-referencing, we can signal that as well.
+ *
+ * If a piece of metadata proves corrupt or suboptimal, the userspace
+ * program can ask the kernel to apply some tender loving care (TLC) to
+ * the metadata object.  "Corruption" is defined by metadata violating
+ * the on-disk specification; operations cannot continue if the
+ * violation is left untreated.  It is possible for XFS to continue if
+ * an object is "suboptimal", however performance may be degraded.
+ * Repairs are usually performed by rebuilding the metadata entirely out
+ * of redundant metadata.  Optimizing, on the other hand, can sometimes
+ * be done without rebuilding entire structures.
+ *
+ * Generally speaking, the repair code has the following code structure:
+ * Lock -> scrub -> repair -> commit -> re-lock -> re-scrub -> unlock.
+ * The first check helps us figure out if we need to rebuild or simply
+ * optimize the structure so that the rebuild knows what to do.  The
+ * second check evaluates the completeness of the repair; that is what
+ * is reported to userspace.
  */
 
+/* Fix something if errors were detected and the user asked for repair. */
+static inline bool
+xfs_scrub_should_fix(
+	struct xfs_scrub_metadata	*sm)
+{
+	return (sm->sm_flags & XFS_SCRUB_FLAG_REPAIR) &&
+	       (sm->sm_flags & (XFS_SCRUB_FLAG_CORRUPT | XFS_SCRUB_FLAG_PREEN |
+				XFS_SCRUB_FLAG_XCORRUPT));
+}
+
+/* Clear the corruption status flags. */
+static inline bool
+xfs_scrub_reset_corruption_flags(
+	struct xfs_scrub_metadata	*sm)
+{
+	return sm->sm_flags &= ~XFS_SCRUB_FLAGS_OUT;
+}
+
 /* Check for operational errors. */
 bool
 xfs_scrub_op_ok(
@@ -609,7 +645,10 @@ xfs_scrub_teardown(
 	if (sc->ag_lock.agmask != sc->ag_lock.__agmask)
 		kmem_free(sc->ag_lock.agmask);
 	sc->ag_lock.agmask = NULL;
-	xfs_trans_cancel(sc->tp);
+	if (error == 0 && (sc->sm->sm_flags & XFS_SCRUB_FLAG_REPAIR))
+		error = xfs_trans_commit(sc->tp);
+	else
+		xfs_trans_cancel(sc->tp);
 	sc->tp = NULL;
 	if (sc->ip != NULL) {
 		xfs_iunlock(sc->ip, XFS_ILOCK_EXCL);
@@ -752,6 +791,8 @@ xfs_scrub_metadata(
 	struct xfs_mount		*mp = ip->i_mount;
 	const struct xfs_scrub_meta_fns	*fns;
 	bool				deadlocked = false;
+	bool				already_fixed = false;
+	bool				was_corrupt = false;
 	int				error = 0;
 
 	trace_xfs_scrub(ip, sm->sm_type, sm->sm_agno, sm->sm_ino, sm->sm_gen,
@@ -765,8 +806,6 @@ xfs_scrub_metadata(
 	sm->sm_flags &= ~XFS_SCRUB_FLAGS_OUT;
 	if (sm->sm_flags & ~XFS_SCRUB_FLAGS_IN)
 		goto out;
-	if (sm->sm_flags & XFS_SCRUB_FLAG_REPAIR)
-		goto out;
 	error = -ENOENT;
 	if (sm->sm_type > XFS_SCRUB_TYPE_MAX)
 		goto out;
@@ -774,6 +813,14 @@ xfs_scrub_metadata(
 	if (fns->scrub == NULL)
 		goto out;
 	error = -EOPNOTSUPP;
+	if ((sm->sm_flags & XFS_SCRUB_FLAG_REPAIR) &&
+	    (fns->repair == NULL || !xfs_sb_version_hascrc(&mp->m_sb)))
+		goto out;
+
+	error = -EROFS;
+	if ((sm->sm_flags & XFS_SCRUB_FLAG_REPAIR) &&
+	    (mp->m_flags & XFS_MOUNT_RDONLY))
+		goto out;
 
 	/* Do we even have this type of metadata? */
 	error = -ENOENT;
@@ -809,8 +856,52 @@ xfs_scrub_metadata(
 	} else if (error)
 		goto out_teardown;
 
-	if (sm->sm_flags & XFS_SCRUB_FLAG_CORRUPT)
-		xfs_alert_ratelimited(mp, "Corruption detected during scrub.");
+	/* Let debug users force us into the repair routines. */
+	if ((sm->sm_flags & XFS_SCRUB_FLAG_REPAIR) && !already_fixed &&
+	    XFS_TEST_ERROR(false, mp,
+			XFS_ERRTAG_FORCE_SCRUB_REPAIR,
+			XFS_RANDOM_FORCE_SCRUB_REPAIR)) {
+		sm->sm_flags |= XFS_SCRUB_FLAG_CORRUPT;
+	}
+	if (!already_fixed)
+		was_corrupt = (sm->sm_flags & (XFS_SCRUB_FLAG_CORRUPT |
+					       XFS_SCRUB_FLAG_XCORRUPT));
+
+	if (!already_fixed && xfs_scrub_should_fix(sm)) {
+		xfs_scrub_ag_btcur_free(&sc.sa);
+
+		/* Ok, something's wrong.  Repair it. */
+		trace_xfs_repair_attempt(ip, sm->sm_type, sm->sm_agno,
+			sm->sm_ino, sm->sm_gen, sm->sm_flags, error);
+		error = fns->repair(&sc);
+		trace_xfs_repair_done(ip, sm->sm_type, sm->sm_agno,
+			sm->sm_ino, sm->sm_gen, sm->sm_flags, error);
+		if (error)
+			goto out_teardown;
+
+		/*
+		 * Commit the fixes and perform a second dry-run scrub
+		 * so that we can tell userspace if we fixed the problem.
+		 */
+		error = xfs_scrub_teardown(&sc, ip, error);
+		if (error)
+			goto out;
+		xfs_scrub_reset_corruption_flags(sm);
+		already_fixed = true;
+		goto retry_op;
+	}
+
+	if (sm->sm_flags & (XFS_SCRUB_FLAG_CORRUPT | XFS_SCRUB_FLAG_XCORRUPT)) {
+		char	*errstr;
+
+		if (sm->sm_flags & XFS_SCRUB_FLAG_REPAIR)
+			errstr = "Corruption not fixed during online repair.  "
+				 "Unmount and run xfs_repair.";
+		else
+			errstr = "Corruption detected during scrub.";
+		xfs_alert_ratelimited(mp, errstr);
+	} else if (already_fixed && was_corrupt)
+		xfs_alert_ratelimited(mp, "Corruption repaired during scrub.");
 
 out_teardown:
 	error = xfs_scrub_teardown(&sc, ip, error);
diff --git a/fs/xfs/xfs_error.h b/fs/xfs/xfs_error.h
index 05f8666..4c22d9a 100644
--- a/fs/xfs/xfs_error.h
+++ b/fs/xfs/xfs_error.h
@@ -96,7 +96,8 @@ extern void xfs_verifier_error(struct xfs_buf *bp);
 #define XFS_ERRTAG_REFCOUNT_FINISH_ONE			25
 #define XFS_ERRTAG_BMAP_FINISH_ONE			26
 #define XFS_ERRTAG_AG_RESV_CRITICAL			27
-#define XFS_ERRTAG_MAX					28
+#define XFS_ERRTAG_FORCE_SCRUB_REPAIR			28
+#define XFS_ERRTAG_MAX					29
 
 /*
  * Random factors for above tags, 1 means always, 2 means 1/2 time, etc.
@@ -129,6 +130,7 @@ extern void xfs_verifier_error(struct xfs_buf *bp);
 #define XFS_RANDOM_REFCOUNT_FINISH_ONE			1
 #define XFS_RANDOM_BMAP_FINISH_ONE			1
 #define XFS_RANDOM_AG_RESV_CRITICAL			4
+#define XFS_RANDOM_FORCE_SCRUB_REPAIR			1
 
 #ifdef DEBUG
 extern int xfs_error_test_active;

[PATCH 44/55] xfs: add helper routines for the repair code

[Darrick J. Wong]

Add some helper functions for repair functions that will help us to
allocate and initialize new metadata blocks for btrees that we're
rebuilding.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/Makefile                  |    1 
 fs/xfs/libxfs/xfs_alloc_btree.c  |    9 
 fs/xfs/libxfs/xfs_alloc_btree.h  |    2 
 fs/xfs/libxfs/xfs_bmap_btree.c   |    9 
 fs/xfs/libxfs/xfs_bmap_btree.h   |    3 
 fs/xfs/libxfs/xfs_btree.c        |    4 
 fs/xfs/libxfs/xfs_btree.h        |    2 
 fs/xfs/libxfs/xfs_ialloc_btree.c |    9 
 fs/xfs/libxfs/xfs_ialloc_btree.h |    3 
 fs/xfs/libxfs/xfs_rmap.c         |   51 +++
 fs/xfs/libxfs/xfs_rmap.h         |    3 
 fs/xfs/scrub/common.c            |    8 
 fs/xfs/scrub/common.h            |   50 ++
 fs/xfs/scrub/repair.c            |  763 ++++++++++++++++++++++++++++++++++++++
 14 files changed, 913 insertions(+), 4 deletions(-)
 create mode 100644 fs/xfs/scrub/repair.c


diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index a22b027..5453e3c 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -115,6 +115,7 @@ xfs-$(CONFIG_XFS_DEBUG)		+= $(addprefix scrub/, \
 				   ialloc.o \
 				   inode.o \
 				   refcount.o \
+				   repair.o \
 				   rmap.o \
 				   symlink.o \
 				   )
diff --git a/fs/xfs/libxfs/xfs_alloc_btree.c b/fs/xfs/libxfs/xfs_alloc_btree.c
index 10d6c12..80b89c9 100644
--- a/fs/xfs/libxfs/xfs_alloc_btree.c
+++ b/fs/xfs/libxfs/xfs_alloc_btree.c
@@ -534,3 +534,12 @@ xfs_allocbt_maxrecs(
 		return blocklen / sizeof(xfs_alloc_rec_t);
 	return blocklen / (sizeof(xfs_alloc_key_t) + sizeof(xfs_alloc_ptr_t));
 }
+
+/* Calculate the freespace btree size for some records. */
+xfs_extlen_t
+xfs_allocbt_calc_size(
+	struct xfs_mount	*mp,
+	unsigned long long	len)
+{
+	return xfs_btree_calc_size(mp, mp->m_alloc_mnr, len);
+}
diff --git a/fs/xfs/libxfs/xfs_alloc_btree.h b/fs/xfs/libxfs/xfs_alloc_btree.h
index 45e189e..2fd5472 100644
--- a/fs/xfs/libxfs/xfs_alloc_btree.h
+++ b/fs/xfs/libxfs/xfs_alloc_btree.h
@@ -61,5 +61,7 @@ extern struct xfs_btree_cur *xfs_allocbt_init_cursor(struct xfs_mount *,
 		struct xfs_trans *, struct xfs_buf *,
 		xfs_agnumber_t, xfs_btnum_t);
 extern int xfs_allocbt_maxrecs(struct xfs_mount *, int, int);
+extern xfs_extlen_t xfs_allocbt_calc_size(struct xfs_mount *mp,
+		unsigned long long len);
 
 #endif	/* __XFS_ALLOC_BTREE_H__ */
diff --git a/fs/xfs/libxfs/xfs_bmap_btree.c b/fs/xfs/libxfs/xfs_bmap_btree.c
index fde4a29..ed431a5 100644
--- a/fs/xfs/libxfs/xfs_bmap_btree.c
+++ b/fs/xfs/libxfs/xfs_bmap_btree.c
@@ -912,3 +912,12 @@ xfs_bmbt_change_owner(
 	xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
 	return error;
 }
+
+/* Calculate the bmap btree size for some records. */
+unsigned long long
+xfs_bmbt_calc_size(
+	struct xfs_mount	*mp,
+	unsigned long long	len)
+{
+	return xfs_btree_calc_size(mp, mp->m_bmap_dmnr, len);
+}
diff --git a/fs/xfs/libxfs/xfs_bmap_btree.h b/fs/xfs/libxfs/xfs_bmap_btree.h
index 819a8a4..835f0a3 100644
--- a/fs/xfs/libxfs/xfs_bmap_btree.h
+++ b/fs/xfs/libxfs/xfs_bmap_btree.h
@@ -140,4 +140,7 @@ extern int xfs_bmbt_change_owner(struct xfs_trans *tp, struct xfs_inode *ip,
 extern struct xfs_btree_cur *xfs_bmbt_init_cursor(struct xfs_mount *,
 		struct xfs_trans *, struct xfs_inode *, int);
 
+extern unsigned long long xfs_bmbt_calc_size(struct xfs_mount *mp,
+		unsigned long long len);
+
 #endif	/* __XFS_BMAP_BTREE_H__ */
diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c
index fecfd45..1f03736 100644
--- a/fs/xfs/libxfs/xfs_btree.c
+++ b/fs/xfs/libxfs/xfs_btree.c
@@ -4842,7 +4842,7 @@ xfs_btree_query_all(
  * Calculate the number of blocks needed to store a given number of records
  * in a short-format (per-AG metadata) btree.
  */
-xfs_extlen_t
+unsigned long long
 xfs_btree_calc_size(
 	struct xfs_mount	*mp,
 	uint			*limits,
@@ -4850,7 +4850,7 @@ xfs_btree_calc_size(
 {
 	int			level;
 	int			maxrecs;
-	xfs_extlen_t		rval;
+	unsigned long long	rval;
 
 	maxrecs = limits[0];
 	for (level = 0, rval = 0; len > 1; level++) {
diff --git a/fs/xfs/libxfs/xfs_btree.h b/fs/xfs/libxfs/xfs_btree.h
index f1b8241..6092e2b 100644
--- a/fs/xfs/libxfs/xfs_btree.h
+++ b/fs/xfs/libxfs/xfs_btree.h
@@ -480,7 +480,7 @@ bool xfs_btree_sblock_v5hdr_verify(struct xfs_buf *bp);
 bool xfs_btree_sblock_verify(struct xfs_buf *bp, unsigned int max_recs);
 uint xfs_btree_compute_maxlevels(struct xfs_mount *mp, uint *limits,
 				 unsigned long len);
-xfs_extlen_t xfs_btree_calc_size(struct xfs_mount *mp, uint *limits,
+unsigned long long xfs_btree_calc_size(struct xfs_mount *mp, uint *limits,
 		unsigned long long len);
 
 /* return codes */
diff --git a/fs/xfs/libxfs/xfs_ialloc_btree.c b/fs/xfs/libxfs/xfs_ialloc_btree.c
index 09d8cb0..ec784c3 100644
--- a/fs/xfs/libxfs/xfs_ialloc_btree.c
+++ b/fs/xfs/libxfs/xfs_ialloc_btree.c
@@ -500,3 +500,12 @@ xfs_inobt_rec_check_count(
 	return 0;
 }
 #endif	/* DEBUG */
+
+/* Calculate the inobt btree size for some records. */
+xfs_extlen_t
+xfs_iallocbt_calc_size(
+	struct xfs_mount	*mp,
+	unsigned long long	len)
+{
+	return xfs_btree_calc_size(mp, mp->m_inobt_mnr, len);
+}
diff --git a/fs/xfs/libxfs/xfs_ialloc_btree.h b/fs/xfs/libxfs/xfs_ialloc_btree.h
index bd88453..3046c11 100644
--- a/fs/xfs/libxfs/xfs_ialloc_btree.h
+++ b/fs/xfs/libxfs/xfs_ialloc_btree.h
@@ -72,4 +72,7 @@ int xfs_inobt_rec_check_count(struct xfs_mount *,
 #define xfs_inobt_rec_check_count(mp, rec)	0
 #endif	/* DEBUG */
 
+extern xfs_extlen_t xfs_iallocbt_calc_size(struct xfs_mount *mp,
+		unsigned long long len);
+
 #endif	/* __XFS_IALLOC_BTREE_H__ */
diff --git a/fs/xfs/libxfs/xfs_rmap.c b/fs/xfs/libxfs/xfs_rmap.c
index cce51cb..e61d816 100644
--- a/fs/xfs/libxfs/xfs_rmap.c
+++ b/fs/xfs/libxfs/xfs_rmap.c
@@ -2364,3 +2364,54 @@ xfs_rmap_record_exists(
 		     irec.rm_startblock + irec.rm_blockcount >= bno + len);
 	return 0;
 }
+
+struct xfs_rmap_has_other_keys {
+	uint64_t			owner;
+	uint64_t			offset;
+	bool				*has_rmap;
+	unsigned int			flags;
+};
+
+/* For each rmap given, figure out if it doesn't match the key we want. */
+STATIC int
+xfs_rmap_has_other_keys_helper(
+	struct xfs_btree_cur		*cur,
+	struct xfs_rmap_irec		*rec,
+	void				*priv)
+{
+	struct xfs_rmap_has_other_keys	*rhok = priv;
+
+	if (rhok->owner == rec->rm_owner && rhok->offset == rec->rm_offset &&
+	    ((rhok->flags & rec->rm_flags) & XFS_RMAP_KEY_FLAGS) == rhok->flags)
+		return 0;
+	*rhok->has_rmap = true;
+	return XFS_BTREE_QUERY_RANGE_ABORT;
+}
+
+/*
+ * Given an extent and some owner info, can we find records overlapping
+ * the extent whose owner info does not match the given owner?
+ */
+int
+xfs_rmap_has_other_keys(
+	struct xfs_btree_cur		*cur,
+	xfs_fsblock_t			bno,
+	xfs_filblks_t			len,
+	struct xfs_owner_info		*oinfo,
+	bool				*has_rmap)
+{
+	struct xfs_rmap_irec		low = {0};
+	struct xfs_rmap_irec		high;
+	struct xfs_rmap_has_other_keys	rhok;
+
+	xfs_owner_info_unpack(oinfo, &rhok.owner, &rhok.offset, &rhok.flags);
+	*has_rmap = false;
+	rhok.has_rmap = has_rmap;
+
+	low.rm_startblock = bno;
+	memset(&high, 0xFF, sizeof(high));
+	high.rm_startblock = bno + len - 1;
+
+	return xfs_rmap_query_range(cur, &low, &high,
+			xfs_rmap_has_other_keys_helper, &rhok);
+}
diff --git a/fs/xfs/libxfs/xfs_rmap.h b/fs/xfs/libxfs/xfs_rmap.h
index ea359ab..606efe3 100644
--- a/fs/xfs/libxfs/xfs_rmap.h
+++ b/fs/xfs/libxfs/xfs_rmap.h
@@ -222,5 +222,8 @@ int xfs_rmap_has_record(struct xfs_btree_cur *cur, xfs_fsblock_t bno,
 int xfs_rmap_record_exists(struct xfs_btree_cur *cur, xfs_fsblock_t bno,
 		xfs_filblks_t len, struct xfs_owner_info *oinfo,
 		bool *has_rmap);
+int xfs_rmap_has_other_keys(struct xfs_btree_cur *cur, xfs_fsblock_t bno,
+		xfs_filblks_t len, struct xfs_owner_info *oinfo,
+		bool *has_rmap);
 
 #endif	/* __XFS_RMAP_H__ */
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 8b6384f..9de7502 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -641,6 +641,8 @@ xfs_scrub_teardown(
 	struct xfs_inode		*ip_in,
 	int				error)
 {
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+
 	xfs_scrub_ag_free(&sc->sa);
 	if (sc->ag_lock.agmask != sc->ag_lock.__agmask)
 		kmem_free(sc->ag_lock.agmask);
@@ -662,6 +664,8 @@ xfs_scrub_teardown(
 		kmem_free(sc->buf);
 		sc->buf = NULL;
 	}
+	if (sc->reset_counters && !error)
+		error = xfs_repair_reset_counters(mp);
 	return error;
 }
 
@@ -674,11 +678,13 @@ xfs_scrub_setup(
 	bool				retry_deadlocked)
 {
 	struct xfs_mount		*mp = ip->i_mount;
+	xfs_extlen_t			resblks;
 
 	memset(sc, 0, sizeof(*sc));
 	sc->sm = sm;
+	resblks = xfs_repair_calc_ag_resblks(sc, ip, sm);
 	return xfs_scrub_trans_alloc(sm, mp, &M_RES(mp)->tr_itruncate,
-			0, 0, 0, &sc->tp);
+			resblks, 0, 0, &sc->tp);
 }
 
 /*
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 11c04d5..992d21b 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -61,6 +61,7 @@ struct xfs_scrub_context {
 	struct xfs_inode		*ip;
 	void				*buf;
 	bool				retry;
+	bool				reset_counters;
 
 	/* State tracking for multi-AG operations. */
 	struct xfs_scrub_ag_lock	ag_lock;
@@ -97,6 +98,9 @@ xfs_scrub_trans_alloc(
 	uint				flags,
 	struct xfs_trans		**tpp)
 {
+	if (sm->sm_flags & XFS_SCRUB_FLAG_REPAIR)
+		return xfs_trans_alloc(mp, resp, blocks, rtextents, flags, tpp);
+
 	return xfs_trans_alloc_empty(mp, tpp);
 }
 
@@ -322,4 +326,50 @@ int xfs_scrub_symlink(struct xfs_scrub_context *sc);
 int xfs_scrub_rtbitmap(struct xfs_scrub_context *sc);
 int xfs_scrub_rtsummary(struct xfs_scrub_context *sc);
 
+/* Repair helpers */
+
+struct xfs_repair_find_ag_btree {
+	uint64_t			rmap_owner;
+	const struct xfs_buf_ops	*buf_ops;
+	uint32_t			magic;
+	xfs_agblock_t			root;
+	unsigned int			level;
+};
+
+struct xfs_repair_btree_extent {
+	struct list_head		list;
+	xfs_fsblock_t			fsbno;
+	xfs_extlen_t			len;
+};
+
+int xfs_repair_roll_ag_trans(struct xfs_scrub_context *sc);
+bool xfs_repair_ag_has_space(struct xfs_perag *pag, xfs_extlen_t nr_blocks,
+			     enum xfs_ag_resv_type type);
+int xfs_repair_alloc_ag_block(struct xfs_scrub_context *sc,
+			      struct xfs_owner_info *oinfo,
+			      xfs_fsblock_t *fsbno, enum xfs_ag_resv_type resv);
+int xfs_repair_init_btblock(struct xfs_scrub_context *sc, xfs_fsblock_t fsb,
+			    struct xfs_buf **bpp, __u32 magic,
+			    const struct xfs_buf_ops *ops);
+int xfs_repair_fix_freelist(struct xfs_scrub_context *sc, bool can_shrink);
+int xfs_repair_put_freelist(struct xfs_scrub_context *sc, xfs_agblock_t agbno);
+int xfs_repair_collect_btree_extent(struct xfs_mount *mp,
+				    struct list_head *btlist,
+				    xfs_fsblock_t fsbno, xfs_extlen_t len);
+int xfs_repair_reap_btree_extents(struct xfs_scrub_context *sc,
+				  struct list_head *btlist,
+				  struct xfs_owner_info *oinfo,
+				  enum xfs_ag_resv_type type);
+void xfs_repair_cancel_btree_extents(struct xfs_scrub_context *sc,
+				     struct list_head *btlist);
+int xfs_repair_subtract_extents(struct xfs_mount *mp, struct list_head *exlist,
+				struct list_head *sublist);
+int xfs_repair_find_ag_btree_roots(struct xfs_scrub_context *sc,
+				   struct xfs_buf *agf_bp,
+				   struct xfs_repair_find_ag_btree *btree_info);
+int xfs_repair_reset_counters(struct xfs_mount	*mp);
+xfs_extlen_t xfs_repair_calc_ag_resblks(struct xfs_scrub_context *sc,
+					struct xfs_inode *ip,
+					struct xfs_scrub_metadata *sm);
+
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/repair.c b/fs/xfs/scrub/repair.c
new file mode 100644
index 0000000..896e42e3
--- /dev/null
+++ b/fs/xfs/scrub/repair.c
@@ -0,0 +1,763 @@
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_btree.h"
+#include "xfs_bit.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_trace.h"
+#include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_alloc.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_ialloc.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_rmap.h"
+#include "xfs_rmap_btree.h"
+#include "xfs_refcount.h"
+#include "xfs_refcount_btree.h"
+#include "xfs_extent_busy.h"
+#include "xfs_ag_resv.h"
+#include "xfs_trans_space.h"
+#include "scrub/common.h"
+
+/*
+ * Roll a transaction, keeping the AG headers locked and reinitializing
+ * the btree cursors.
+ */
+int
+xfs_repair_roll_ag_trans(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_trans		*tp;
+	int				error;
+
+	/* Keep the AG header buffers locked so we can keep going. */
+	xfs_trans_bhold(sc->tp, sc->sa.agi_bp);
+	xfs_trans_bhold(sc->tp, sc->sa.agf_bp);
+	xfs_trans_bhold(sc->tp, sc->sa.agfl_bp);
+
+	/* Roll the transaction. */
+	tp = sc->tp;
+	error = xfs_trans_roll(&sc->tp, NULL);
+	if (error)
+		return error;
+
+	/* Join the buffer to the new transaction or release the hold. */
+	if (sc->tp != tp) {
+		xfs_trans_bjoin(sc->tp, sc->sa.agi_bp);
+		xfs_trans_bjoin(sc->tp, sc->sa.agf_bp);
+		xfs_trans_bjoin(sc->tp, sc->sa.agfl_bp);
+	} else {
+		xfs_trans_bhold_release(sc->tp, sc->sa.agi_bp);
+		xfs_trans_bhold_release(sc->tp, sc->sa.agf_bp);
+		xfs_trans_bhold_release(sc->tp, sc->sa.agfl_bp);
+	}
+
+	return error;
+}
+
+/*
+ * Does the given AG have enough space to rebuild a btree?  Neither AG
+ * reservation can be critical, and we must have enough space (factoring
+ * in AG reservations) to construct a whole btree.
+ */
+bool
+xfs_repair_ag_has_space(
+	struct xfs_perag		*pag,
+	xfs_extlen_t			nr_blocks,
+	enum xfs_ag_resv_type		type)
+{
+	return  !xfs_ag_resv_critical(pag, XFS_AG_RESV_AGFL) &&
+		!xfs_ag_resv_critical(pag, XFS_AG_RESV_METADATA) &&
+		pag->pagf_freeblks - xfs_ag_resv_needed(pag, type) > nr_blocks;
+}
+
+/* Allocate a block in an AG. */
+int
+xfs_repair_alloc_ag_block(
+	struct xfs_scrub_context	*sc,
+	struct xfs_owner_info		*oinfo,
+	xfs_fsblock_t			*fsbno,
+	enum xfs_ag_resv_type		resv)
+{
+	struct xfs_alloc_arg		args = {0};
+	xfs_agblock_t			bno;
+	int				error;
+
+	if (resv == XFS_AG_RESV_AGFL) {
+		error = xfs_alloc_get_freelist(sc->tp, sc->sa.agf_bp, &bno, 1);
+		if (error)
+			return error;
+		xfs_extent_busy_reuse(sc->tp->t_mountp, sc->sa.agno, bno,
+				1, false);
+		*fsbno = XFS_AGB_TO_FSB(sc->tp->t_mountp, sc->sa.agno, bno);
+		return 0;
+	}
+
+	args.tp = sc->tp;
+	args.mp = sc->tp->t_mountp;
+	args.oinfo = *oinfo;
+	args.fsbno = XFS_AGB_TO_FSB(args.mp, sc->sa.agno, 0);
+	args.minlen = 1;
+	args.maxlen = 1;
+	args.prod = 1;
+	args.type = XFS_ALLOCTYPE_NEAR_BNO;
+	args.resv = resv;
+
+	error = xfs_alloc_vextent(&args);
+	if (error)
+		return error;
+	if (args.fsbno == NULLFSBLOCK)
+		return -ENOSPC;
+	ASSERT(args.len == 1);
+	*fsbno = args.fsbno;
+
+	return 0;
+}
+
+/* Initialize an AG block to a zeroed out btree header. */
+int
+xfs_repair_init_btblock(
+	struct xfs_scrub_context	*sc,
+	xfs_fsblock_t			fsb,
+	struct xfs_buf			**bpp,
+	__u32				magic,
+	const struct xfs_buf_ops	*ops)
+{
+	struct xfs_trans		*tp = sc->tp;
+	struct xfs_mount		*mp = tp->t_mountp;
+	struct xfs_buf			*bp;
+
+	trace_xfs_repair_init_btblock(mp, XFS_FSB_TO_AGNO(mp, fsb),
+			XFS_FSB_TO_AGBNO(mp, fsb), magic);
+
+	ASSERT(XFS_FSB_TO_AGNO(mp, fsb) == sc->sa.agno);
+	bp = xfs_trans_get_buf(tp, mp->m_ddev_targp, XFS_FSB_TO_DADDR(mp, fsb),
+			XFS_FSB_TO_BB(mp, 1), 0);
+	xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
+	xfs_btree_init_block(mp, bp, magic, 0, 0, sc->sa.agno,
+			XFS_BTREE_CRC_BLOCKS);
+	xfs_trans_buf_set_type(tp, bp, XFS_BLFT_BTREE_BUF);
+	xfs_trans_log_buf(tp, bp, 0, bp->b_length);
+	bp->b_ops = ops;
+	*bpp = bp;
+
+	return 0;
+}
+
+/* Ensure the freelist is full. */
+int
+xfs_repair_fix_freelist(
+	struct xfs_scrub_context	*sc,
+	bool				can_shrink)
+{
+	struct xfs_alloc_arg		args = {0};
+	int				error;
+
+	args.mp = sc->tp->t_mountp;
+	args.tp = sc->tp;
+	args.agno = sc->sa.agno;
+	args.alignment = 1;
+	args.pag = xfs_perag_get(args.mp, sc->sa.agno);
+
+	error = xfs_alloc_fix_freelist(&args,
+			can_shrink ? 0 : XFS_ALLOC_FLAG_NOSHRINK);
+	xfs_perag_put(args.pag);
+
+	return error;
+}
+
+/* Put a block back on the AGFL. */
+int
+xfs_repair_put_freelist(
+	struct xfs_scrub_context	*sc,
+	xfs_agblock_t			agbno)
+{
+	struct xfs_owner_info		oinfo;
+	int				error;
+
+	/*
+	 * Since we're "freeing" a lost block onto the AGFL, we have to
+	 * create an rmap for the block prior to merging it or else other
+	 * parts will break.
+	 */
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_AG);
+	error = xfs_rmap_alloc(sc->tp, sc->sa.agf_bp, sc->sa.agno, agbno, 1,
+			&oinfo);
+	if (error)
+		return error;
+
+	/* Put the block on the AGFL. */
+	error = xfs_alloc_put_freelist(sc->tp, sc->sa.agf_bp, sc->sa.agfl_bp,
+			agbno, 0);
+	if (error)
+		return error;
+	xfs_extent_busy_insert(sc->tp, sc->sa.agno, agbno, 1,
+			XFS_EXTENT_BUSY_SKIP_DISCARD);
+
+	/* Make sure the AGFL doesn't overfill. */
+	return xfs_repair_fix_freelist(sc, true);
+}
+
+/*
+ * For a given metadata extent and owner, delete the associated rmap.
+ * If the block has no other owners, free it.
+ */
+STATIC int
+xfs_repair_free_or_unmap_extent(
+	struct xfs_scrub_context	*sc,
+	xfs_fsblock_t			fsbno,
+	xfs_extlen_t			len,
+	struct xfs_owner_info		*oinfo,
+	enum xfs_ag_resv_type		resv)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_btree_cur		*rmap_cur;
+	struct xfs_buf			*agf_bp = NULL;
+	xfs_agnumber_t			agno;
+	xfs_agblock_t			agbno;
+	bool				has_other_rmap;
+	int				error = 0;
+
+	ASSERT(xfs_sb_version_hasrmapbt(&mp->m_sb));
+	agno = XFS_FSB_TO_AGNO(mp, fsbno);
+	agbno = XFS_FSB_TO_AGBNO(mp, fsbno);
+
+	trace_xfs_repair_free_or_unmap_extent(mp, agno, agbno, len);
+
+	for (; len > 0 && !error; len--, agbno++, fsbno++) {
+		ASSERT(sc->ip != NULL || agno == sc->sa.agno);
+
+		/* Can we find any other rmappings? */
+		if (sc->ip) {
+			error = xfs_alloc_read_agf(mp, sc->tp, agno, 0,
+					&agf_bp);
+			if (error)
+				break;
+		}
+		rmap_cur = xfs_rmapbt_init_cursor(mp, sc->tp,
+				agf_bp ? agf_bp : sc->sa.agf_bp, agno);
+		error = xfs_rmap_has_other_keys(rmap_cur, agbno, 1, oinfo,
+				&has_other_rmap);
+		if (error)
+			goto out_cur;
+		xfs_btree_del_cursor(rmap_cur, XFS_BTREE_NOERROR);
+		if (agf_bp)
+			xfs_trans_brelse(sc->tp, agf_bp);
+
+		/*
+		 * If there are other rmappings, this block is cross
+		 * linked and must not be freed.  Remove the reverse
+		 * mapping and move on.  Otherwise, we were the only
+		 * owner of the block, so free the extent, which will
+		 * also remove the rmap.
+		 */
+		if (has_other_rmap)
+			error = xfs_rmap_free(sc->tp, agf_bp, agno, agbno, 1,
+					oinfo);
+		else if (resv == XFS_AG_RESV_AGFL)
+			error = xfs_repair_put_freelist(sc, agbno);
+		else
+			error = xfs_free_extent(sc->tp, fsbno, 1, oinfo, resv);
+		if (error)
+			break;
+
+		if (sc->ip)
+			error = xfs_trans_roll(&sc->tp, sc->ip);
+		else
+			error = xfs_repair_roll_ag_trans(sc);
+	}
+
+	return error;
+out_cur:
+	xfs_btree_del_cursor(rmap_cur, XFS_BTREE_ERROR);
+	if (agf_bp)
+		xfs_trans_brelse(sc->tp, agf_bp);
+	return error;
+}
+
+/* Collect a dead btree extent for later disposal. */
+int
+xfs_repair_collect_btree_extent(
+	struct xfs_mount		*mp,
+	struct list_head		*btlist,
+	xfs_fsblock_t			fsbno,
+	xfs_extlen_t			len)
+{
+	struct xfs_repair_btree_extent	*rbe;
+
+	trace_xfs_repair_collect_btree_extent(mp, XFS_FSB_TO_AGNO(mp, fsbno),
+			XFS_FSB_TO_AGBNO(mp, fsbno), len);
+
+	rbe = kmem_alloc(sizeof(*rbe), KM_NOFS);
+	if (!rbe)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&rbe->list);
+	rbe->fsbno = fsbno;
+	rbe->len = len;
+	list_add_tail(&rbe->list, btlist);
+
+	return 0;
+}
+
+/* Dispose of dead btree extents.  If oinfo is NULL, just delete the list. */
+int
+xfs_repair_reap_btree_extents(
+	struct xfs_scrub_context	*sc,
+	struct list_head		*btlist,
+	struct xfs_owner_info		*oinfo,
+	enum xfs_ag_resv_type		type)
+{
+	struct xfs_repair_btree_extent	*rbe;
+	struct xfs_repair_btree_extent	*n;
+	int				error = 0;
+
+	list_for_each_entry_safe(rbe, n, btlist, list) {
+		if (oinfo) {
+			error = xfs_repair_free_or_unmap_extent(sc, rbe->fsbno,
+					rbe->len, oinfo, type);
+			if (error)
+				oinfo = NULL;
+		}
+		list_del(&rbe->list);
+		kmem_free(rbe);
+	}
+
+	return error;
+}
+
+/* Errors happened, just delete the dead btree extent list. */
+void
+xfs_repair_cancel_btree_extents(
+	struct xfs_scrub_context	*sc,
+	struct list_head		*btlist)
+{
+	xfs_repair_reap_btree_extents(sc, btlist, NULL, XFS_AG_RESV_NONE);
+}
+
+/* Compare two btree extents. */
+static int
+xfs_repair_btree_extent_cmp(
+	void				*priv,
+	struct list_head		*a,
+	struct list_head		*b)
+{
+	struct xfs_repair_btree_extent	*ap;
+	struct xfs_repair_btree_extent	*bp;
+
+	ap = container_of(a, struct xfs_repair_btree_extent, list);
+	bp = container_of(b, struct xfs_repair_btree_extent, list);
+
+	if (ap->fsbno > bp->fsbno)
+		return 1;
+	else if (ap->fsbno < bp->fsbno)
+		return -1;
+	return 0;
+}
+
+/* Remove all the blocks in sublist from exlist. */
+int
+xfs_repair_subtract_extents(
+	struct xfs_mount		*mp,
+	struct list_head		*exlist,
+	struct list_head		*sublist)
+{
+	struct xfs_repair_btree_extent	*newrbe;
+	struct xfs_repair_btree_extent	*rbe;
+	struct xfs_repair_btree_extent	*n;
+	struct xfs_repair_btree_extent	*subp;
+	struct xfs_repair_btree_extent	sub;
+	xfs_fsblock_t			fsb;
+	xfs_fsblock_t			newfsb;
+	xfs_extlen_t			newlen;
+
+	list_sort(NULL, exlist, xfs_repair_btree_extent_cmp);
+	list_sort(NULL, sublist, xfs_repair_btree_extent_cmp);
+
+	subp = list_first_entry(sublist, struct xfs_repair_btree_extent, list);
+	if (subp == NULL)
+		return 0;
+
+	sub = *subp;
+	/* For every block mentioned in exlist... */
+	list_for_each_entry_safe(rbe, n, exlist, list) {
+		newfsb = NULLFSBLOCK;
+		newlen = 0;
+		for (fsb = rbe->fsbno; fsb < rbe->fsbno + rbe->len; fsb++) {
+			/*
+			 * If the current location of the extent list is
+			 * beyond the subtract list, move the subtract list
+			 * forward by one block or by one record.
+			 */
+			while (fsb > sub.fsbno || sub.len == 0) {
+				if (sub.len) {
+					sub.len--;
+					sub.fsbno++;
+				} else {
+					/*
+					 * Get the next subtract extent.  If
+					 * there isn't one, make the current
+					 * extent match the unprocessed part of
+					 * that extent, and jump out.
+					 */
+					if (subp->list.next == sublist ||
+					    subp->list.next == NULL) {
+						rbe->len -= fsb - rbe->fsbno;
+						rbe->fsbno = fsb;
+						subp = NULL;
+						rbe = NULL;
+						goto out_frag;
+					}
+					subp = list_next_entry(subp, list);
+					sub = *subp;
+				}
+			}
+
+			if (fsb != sub.fsbno) {
+				/*
+				 * Block not in the subtract list; stash
+				 * it for later reinsertion in the list.
+				 */
+				if (newfsb == NULLFSBLOCK) {
+					newfsb = fsb;
+					newlen = 1;
+				} else
+					newlen++;
+			} else {
+				/* Match! */
+				if (newfsb != NULLFSBLOCK) {
+					/*
+					 * Last block of the extent and we have
+					 * a saved extent.  Store the saved
+					 * extent in this extent.
+					 */
+					if (fsb == rbe->fsbno + rbe->len - 1) {
+						rbe->fsbno = newfsb;
+						rbe->len = newlen;
+						newfsb = NULLFSBLOCK;
+						rbe = NULL;
+						goto out_frag;
+					}
+					/* Stash the new extent in the list. */
+					newrbe = kmem_alloc(sizeof(*newrbe),
+							KM_NOFS);
+					if (!newrbe)
+						return -ENOMEM;
+					INIT_LIST_HEAD(&newrbe->list);
+					newrbe->fsbno = newfsb;
+					newrbe->len = newlen;
+					list_add_tail(&newrbe->list,
+							&rbe->list);
+				}
+
+				newfsb = NULLFSBLOCK;
+				newlen = 0;
+			}
+		} /* end for loop */
+
+out_frag:
+		/* If we have an extent to add back, do that now. */
+		if (newfsb != NULLFSBLOCK) {
+			if (rbe) {
+				newrbe = rbe;
+				rbe = NULL;
+			} else {
+				newrbe = kmem_alloc(sizeof(*newrbe), KM_NOFS);
+				if (!newrbe)
+					return -ENOMEM;
+				INIT_LIST_HEAD(&newrbe->list);
+				list_add_tail(&newrbe->list, &rbe->list);
+			}
+			newrbe->fsbno = newfsb;
+			newrbe->len = newlen;
+		}
+		if (rbe) {
+			list_del(&rbe->list);
+			kmem_free(rbe);
+		}
+		if (subp == NULL)
+			break;
+	}
+
+	return 0;
+}
+
+/* Find btree roots from the AGF. */
+STATIC int
+xfs_repair_find_ag_btree_roots_helper(
+	struct xfs_btree_cur		*cur,
+	struct xfs_rmap_irec		*rec,
+	void				*priv)
+{
+	struct xfs_mount		*mp = cur->bc_mp;
+	struct xfs_repair_find_ag_btree	*btree_info = priv;
+	struct xfs_repair_find_ag_btree	*fab;
+	struct xfs_buf			*bp;
+	struct xfs_btree_block		*btblock;
+	xfs_daddr_t			daddr;
+	xfs_agblock_t			agbno;
+	int				error;
+
+	if (!XFS_RMAP_NON_INODE_OWNER(rec->rm_owner))
+		return 0;
+
+	for (agbno = 0; agbno < rec->rm_blockcount; agbno++) {
+		daddr = XFS_AGB_TO_DADDR(mp, cur->bc_private.a.agno,
+				rec->rm_startblock + agbno);
+		for (fab = btree_info; fab->buf_ops; fab++) {
+			if (rec->rm_owner != fab->rmap_owner)
+				continue;
+
+			error = xfs_trans_read_buf(mp, cur->bc_tp,
+					mp->m_ddev_targp, daddr, mp->m_bsize,
+					0, &bp, NULL);
+			if (error)
+				return error;
+
+			btblock = XFS_BUF_TO_BLOCK(bp);
+			if (be32_to_cpu(btblock->bb_magic) != fab->magic)
+				goto next_fab;
+			if (fab->root != NULLAGBLOCK &&
+			    xfs_btree_get_level(btblock) <= fab->level)
+				goto next_fab;
+
+			bp->b_ops = fab->buf_ops;
+			bp->b_ops->verify_read(bp);
+			if (bp->b_error)
+				goto next_fab;
+			fab->root = rec->rm_startblock + agbno;
+			fab->level = xfs_btree_get_level(btblock);
+
+			trace_xfs_repair_find_ag_btree_roots_helper(mp,
+					cur->bc_private.a.agno,
+					rec->rm_startblock + agbno,
+					be32_to_cpu(btblock->bb_magic),
+					fab->level);
+next_fab:
+			xfs_trans_brelse(cur->bc_tp, bp);
+			if (be32_to_cpu(btblock->bb_magic) == fab->magic)
+				break;
+		}
+	}
+
+	return error;
+}
+
+/* Find the roots of the given btrees from the rmap info. */
+int
+xfs_repair_find_ag_btree_roots(
+	struct xfs_scrub_context	*sc,
+	struct xfs_buf			*agf_bp,
+	struct xfs_repair_find_ag_btree	*btree_info)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_repair_find_ag_btree	*fab;
+	struct xfs_btree_cur		*cur;
+	int				error;
+
+	for (fab = btree_info; fab->buf_ops; fab++) {
+		fab->root = NULLAGBLOCK;
+		fab->level = 0;
+	}
+
+	cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno);
+	error = xfs_rmap_query_all(cur, xfs_repair_find_ag_btree_roots_helper,
+			btree_info);
+	xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
+
+	for (fab = btree_info; !error && fab->buf_ops; fab++)
+		if (fab->root != NULLAGBLOCK)
+			fab->level++;
+
+	return error;
+}
+
+/* Reset the superblock counters from the AGF/AGI. */
+int
+xfs_repair_reset_counters(
+	struct xfs_mount	*mp)
+{
+	struct xfs_trans	*tp;
+	struct xfs_buf		*agi_bp;
+	struct xfs_buf		*agf_bp;
+	struct xfs_agi		*agi;
+	struct xfs_agf		*agf;
+	xfs_agnumber_t		agno;
+	xfs_ino_t		icount = 0;
+	xfs_ino_t		ifree = 0;
+	xfs_filblks_t		fdblocks = 0;
+	int64_t			delta_icount;
+	int64_t			delta_ifree;
+	int64_t			delta_fdblocks;
+	int			error;
+
+	trace_xfs_repair_reset_counters(mp);
+
+	error = xfs_trans_alloc_empty(mp, &tp);
+	if (error)
+		return error;
+
+	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
+		/* Count all the inodes... */
+		error = xfs_ialloc_read_agi(mp, tp, agno, &agi_bp);
+		if (error)
+			goto out;
+		agi = XFS_BUF_TO_AGI(agi_bp);
+		icount += be32_to_cpu(agi->agi_count);
+		ifree += be32_to_cpu(agi->agi_freecount);
+
+		/* Add up the free/freelist/bnobt/cntbt blocks... */
+		error = xfs_alloc_read_agf(mp, tp, agno, 0, &agf_bp);
+		if (error)
+			goto out;
+		agf = XFS_BUF_TO_AGF(agf_bp);
+		fdblocks += be32_to_cpu(agf->agf_freeblks);
+		fdblocks += be32_to_cpu(agf->agf_flcount);
+		fdblocks += be32_to_cpu(agf->agf_btreeblks);
+	}
+
+	/*
+	 * Reinitialize the counters.  The on-disk and in-core counters
+	 * differ by the number of inodes/blocks reserved by the admin,
+	 * the per-AG reservation, and any transactions in progress, so
+	 * we have to account for that.
+	 */
+	spin_lock(&mp->m_sb_lock);
+	delta_icount = (int64_t)mp->m_sb.sb_icount - icount;
+	delta_ifree = (int64_t)mp->m_sb.sb_ifree - ifree;
+	delta_fdblocks = (int64_t)mp->m_sb.sb_fdblocks - fdblocks;
+	mp->m_sb.sb_icount = icount;
+	mp->m_sb.sb_ifree = ifree;
+	mp->m_sb.sb_fdblocks = fdblocks;
+	spin_unlock(&mp->m_sb_lock);
+
+	if (delta_icount) {
+		error = xfs_mod_icount(mp, delta_icount);
+		if (error)
+			goto out;
+	}
+	if (delta_ifree) {
+		error = xfs_mod_ifree(mp, delta_ifree);
+		if (error)
+			goto out;
+	}
+	if (delta_fdblocks) {
+		error = xfs_mod_fdblocks(mp, delta_fdblocks, false);
+		if (error)
+			goto out;
+	}
+
+out:
+	xfs_trans_cancel(tp);
+	return error;
+}
+
+/* Figure out how many blocks to reserve for an AG repair. */
+xfs_extlen_t
+xfs_repair_calc_ag_resblks(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+	struct xfs_agi			*agi;
+	struct xfs_agf			*agf;
+	struct xfs_buf			*bp;
+	xfs_agino_t			icount;
+	xfs_extlen_t			aglen;
+	xfs_extlen_t			usedlen;
+	xfs_extlen_t			freelen;
+	xfs_extlen_t			bnobt_sz;
+	xfs_extlen_t			inobt_sz;
+	xfs_extlen_t			rmapbt_sz;
+	xfs_extlen_t			refcbt_sz;
+	int				error;
+
+	if (!(sm->sm_flags & XFS_SCRUB_FLAG_REPAIR))
+		return 0;
+
+	if (sm->sm_agno >= mp->m_sb.sb_agcount)
+		return -EINVAL;
+
+	/*
+	 * Try to get the actual counters from disk; if not, make
+	 * some worst case assumptions.
+	 */
+	error = xfs_read_agi(mp, NULL, sm->sm_agno, &bp);
+	if (!error) {
+		agi = XFS_BUF_TO_AGI(bp);
+		icount = be32_to_cpu(agi->agi_count);
+		xfs_trans_brelse(NULL, bp);
+	} else
+		icount = mp->m_sb.sb_agblocks / mp->m_sb.sb_inopblock;
+
+	error = xfs_alloc_read_agf(mp, NULL, sm->sm_agno, 0, &bp);
+	if (!error) {
+		agf = XFS_BUF_TO_AGF(bp);
+		aglen = be32_to_cpu(agf->agf_length);
+		freelen = be32_to_cpu(agf->agf_freeblks);
+		usedlen = aglen - freelen;
+		xfs_trans_brelse(NULL, bp);
+	} else {
+		aglen = mp->m_sb.sb_agblocks;
+		freelen = aglen;
+		usedlen = aglen;
+	}
+
+	trace_xfs_repair_calc_ag_resblks(mp, sm->sm_agno, icount, aglen,
+			freelen, usedlen);
+
+	/*
+	 * Figure out how many blocks we'd need worst case to rebuild
+	 * each type of btree.  Note that we can only rebuild the
+	 * bnobt/cntbt or inobt/finobt as pairs.
+	 */
+	bnobt_sz = 2 * xfs_allocbt_calc_size(mp, freelen);
+	if (xfs_sb_version_hassparseinodes(&mp->m_sb))
+		inobt_sz = xfs_iallocbt_calc_size(mp, icount /
+				XFS_INODES_PER_HOLEMASK_BIT);
+	else
+		inobt_sz = xfs_iallocbt_calc_size(mp, icount /
+				XFS_INODES_PER_CHUNK);
+	if (xfs_sb_version_hasfinobt(&mp->m_sb))
+		inobt_sz *= 2;
+	if (xfs_sb_version_hasreflink(&mp->m_sb)) {
+		rmapbt_sz = xfs_rmapbt_calc_size(mp, aglen);
+		refcbt_sz = xfs_refcountbt_calc_size(mp, usedlen);
+	} else {
+		rmapbt_sz = xfs_rmapbt_calc_size(mp, usedlen);
+		refcbt_sz = 0;
+	}
+	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+		rmapbt_sz = 0;
+
+	trace_xfs_repair_calc_ag_resblks_btsize(mp, sm->sm_agno, bnobt_sz,
+			inobt_sz, rmapbt_sz, refcbt_sz);
+
+	return max(max(bnobt_sz, inobt_sz), max(rmapbt_sz, refcbt_sz));
+}

[PATCH 45/55] xfs: repair superblocks

[Darrick J. Wong]

If one of the backup superblocks is found to differ seriously from
superblock 0, write out a fresh copy from the in-core sb.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/agheader.c |   35 +++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.c   |    2 +-
 fs/xfs/scrub/common.h   |    4 ++++
 3 files changed, 40 insertions(+), 1 deletion(-)


diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index d94129c..df1305e 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -353,6 +353,41 @@ xfs_scrub_superblock(
 #undef XFS_SCRUB_SB_OP_ERROR_GOTO
 #undef XFS_SCRUB_SB_CHECK
 
+/* Repair the superblock. */
+int
+xfs_repair_superblock(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_buf			*bp;
+	struct xfs_dsb			*sbp;
+	xfs_agnumber_t			agno;
+	int				error;
+
+	/* Don't try to repair AG 0's sb; let xfs_repair deal with it. */
+	agno = sc->sm->sm_agno;
+	if (agno == 0)
+		return -EOPNOTSUPP;
+
+	error = xfs_trans_read_buf(mp, sc->tp, mp->m_ddev_targp,
+		  XFS_AG_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
+		  XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
+	if (error)
+		return error;
+	bp->b_ops = &xfs_sb_buf_ops;
+
+	/* Copy AG 0's superblock to this one. */
+	sbp = XFS_BUF_TO_SBP(bp);
+	memset(sbp, 0, mp->m_sb.sb_sectsize);
+	xfs_sb_to_disk(sbp, &mp->m_sb);
+	sbp->sb_bad_features2 = sbp->sb_features2;
+
+	/* Write this to disk. */
+	xfs_trans_buf_set_type(sc->tp, bp, XFS_BLFT_SB_BUF);
+	xfs_trans_log_buf(sc->tp, bp, 0, mp->m_sb.sb_sectsize - 1);
+	return error;
+}
+
 /* AGF */
 
 /* Tally freespace record lengths. */
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 9de7502..9264ce3 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -761,7 +761,7 @@ struct xfs_scrub_meta_fns {
 
 static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup, xfs_scrub_dummy, NULL, NULL},
-	{xfs_scrub_setup_ag, xfs_scrub_superblock, NULL, NULL},
+	{xfs_scrub_setup_ag, xfs_scrub_superblock, xfs_repair_superblock, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agf, NULL, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agfl, NULL, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agi, NULL, NULL},
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 992d21b..92d294c 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -372,4 +372,8 @@ xfs_extlen_t xfs_repair_calc_ag_resblks(struct xfs_scrub_context *sc,
 					struct xfs_inode *ip,
 					struct xfs_scrub_metadata *sm);
 
+/* Metadata repairers */
+
+int xfs_repair_superblock(struct xfs_scrub_context *sc);
+
 #endif	/* __XFS_REPAIR_COMMON_H__ */

[PATCH 46/55] xfs: repair the AGF and AGFL

[Darrick J. Wong]

Regenerate the AGF and AGFL from the rmap data.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/agheader.c |  392 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.c   |    4 
 fs/xfs/scrub/common.h   |    2 
 3 files changed, 396 insertions(+), 2 deletions(-)


diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index df1305e..6916933 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -32,9 +32,13 @@
 #include "xfs_sb.h"
 #include "xfs_inode.h"
 #include "xfs_alloc.h"
+#include "xfs_alloc_btree.h"
 #include "xfs_ialloc.h"
+#include "xfs_ialloc_btree.h"
 #include "xfs_rmap.h"
+#include "xfs_rmap_btree.h"
 #include "xfs_refcount.h"
+#include "xfs_refcount_btree.h"
 #include "scrub/common.h"
 
 /* Set us up to check an AG header. */
@@ -621,6 +625,182 @@ xfs_scrub_agf(
 #undef XFS_SCRUB_AGF_OP_ERROR_GOTO
 #undef XFS_SCRUB_AGF_CHECK
 
+struct xfs_repair_agf_allocbt {
+	xfs_agblock_t			freeblks;
+	xfs_agblock_t			longest;
+};
+
+/* Record free space shape information. */
+STATIC int
+xfs_repair_agf_walk_allocbt(
+	struct xfs_btree_cur		*cur,
+	struct xfs_alloc_rec_incore	*rec,
+	void				*priv)
+{
+	struct xfs_repair_agf_allocbt	*raa = priv;
+	int				error = 0;
+
+	if (xfs_scrub_should_terminate(&error))
+		return error;
+
+	raa->freeblks += rec->ar_blockcount;
+	if (rec->ar_blockcount > raa->longest)
+		raa->longest = rec->ar_blockcount;
+	return error;
+}
+
+/* Repair the AGF. */
+int
+xfs_repair_agf(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_repair_find_ag_btree	fab[] = {
+		{XFS_RMAP_OWN_AG, &xfs_allocbt_buf_ops, XFS_ABTB_CRC_MAGIC, 0, 0},
+		{XFS_RMAP_OWN_AG, &xfs_allocbt_buf_ops, XFS_ABTC_CRC_MAGIC, 0, 0},
+		{XFS_RMAP_OWN_AG, &xfs_rmapbt_buf_ops, XFS_RMAP_CRC_MAGIC, 0, 0},
+		{XFS_RMAP_OWN_REFC, &xfs_refcountbt_buf_ops, XFS_REFC_CRC_MAGIC, 0, 0},
+		{0, NULL, 0, 0, 0},
+	};
+	struct xfs_repair_agf_allocbt	raa = {0};
+	struct xfs_agf			old_agf;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_buf			*agf_bp;
+	struct xfs_agf			*agf;
+	struct xfs_btree_cur		*cur = NULL;
+	struct xfs_perag		*pag;
+	xfs_agblock_t			blocks;
+	xfs_agblock_t			freesp_blocks;
+	int				error;
+
+	/* We require the rmapbt to rebuild anything. */
+	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+		return -EOPNOTSUPP;
+
+	error = xfs_trans_read_buf(mp, sc->tp, mp->m_ddev_targp,
+			XFS_AG_DADDR(mp, sc->sa.agno, XFS_AGF_DADDR(mp)),
+			XFS_FSS_TO_BB(mp, 1), 0, &agf_bp, NULL);
+	if (error)
+		return error;
+	agf_bp->b_ops = &xfs_agf_buf_ops;
+
+	/* Find the btree roots. */
+	error = xfs_repair_find_ag_btree_roots(sc, agf_bp, fab);
+	if (error)
+		return error;
+	if (fab[0].root == NULLAGBLOCK || fab[0].level > XFS_BTREE_MAXLEVELS ||
+	    fab[1].root == NULLAGBLOCK || fab[1].level > XFS_BTREE_MAXLEVELS ||
+	    fab[2].root == NULLAGBLOCK || fab[2].level > XFS_BTREE_MAXLEVELS)
+		return -EFSCORRUPTED;
+
+	/* Start rewriting the header. */
+	agf = XFS_BUF_TO_AGF(agf_bp);
+	old_agf = *agf;
+	memset(agf, 0, mp->m_sb.sb_sectsize);
+	agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
+	agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
+	agf->agf_seqno = cpu_to_be32(sc->sa.agno);
+	agf->agf_length = cpu_to_be32(xfs_scrub_ag_blocks(mp, sc->sa.agno));
+	agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(fab[0].root);
+	agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(fab[1].root);
+	agf->agf_roots[XFS_BTNUM_RMAPi] = cpu_to_be32(fab[2].root);
+	agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(fab[0].level);
+	agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(fab[1].level);
+	agf->agf_levels[XFS_BTNUM_RMAPi] = cpu_to_be32(fab[2].level);
+	agf->agf_flfirst = old_agf.agf_flfirst;
+	agf->agf_fllast = old_agf.agf_fllast;
+	agf->agf_flcount = old_agf.agf_flcount;
+	if (xfs_sb_version_hascrc(&mp->m_sb))
+		uuid_copy(&agf->agf_uuid, &mp->m_sb.sb_meta_uuid);
+	if (xfs_sb_version_hasreflink(&mp->m_sb)) {
+		agf->agf_refcount_root = cpu_to_be32(fab[3].root);
+		agf->agf_refcount_level = cpu_to_be32(fab[3].level);
+	}
+
+	/* Update the AGF counters from the bnobt. */
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno,
+			XFS_BTNUM_BNO);
+	error = xfs_alloc_query_all(cur, xfs_repair_agf_walk_allocbt, &raa);
+	if (error)
+		goto err;
+	error = xfs_btree_count_blocks(cur, &blocks);
+	if (error)
+		goto err;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	freesp_blocks = blocks - 1;
+	agf->agf_freeblks = cpu_to_be32(raa.freeblks);
+	agf->agf_longest = cpu_to_be32(raa.longest);
+
+	/* Update the AGF counters from the cntbt. */
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno,
+			XFS_BTNUM_CNT);
+	error = xfs_btree_count_blocks(cur, &blocks);
+	if (error)
+		goto err;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	freesp_blocks += blocks - 1;
+
+	/* Update the AGF counters from the rmapbt. */
+	cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno);
+	error = xfs_btree_count_blocks(cur, &blocks);
+	if (error)
+		goto err;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	agf->agf_rmap_blocks = cpu_to_be32(blocks);
+	freesp_blocks += blocks - 1;
+
+	/* Update the AGF counters from the refcountbt. */
+	if (xfs_sb_version_hasreflink(&mp->m_sb)) {
+		cur = xfs_refcountbt_init_cursor(mp, sc->tp, agf_bp,
+				sc->sa.agno, NULL);
+		error = xfs_btree_count_blocks(cur, &blocks);
+		if (error)
+			goto err;
+		xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+		agf->agf_refcount_blocks = cpu_to_be32(blocks);
+	}
+	agf->agf_btreeblks = cpu_to_be32(freesp_blocks);
+	cur = NULL;
+
+	/* Trigger reinitialization of the in-core data. */
+	if (raa.freeblks != be32_to_cpu(old_agf.agf_freeblks) ||
+	    freesp_blocks != be32_to_cpu(old_agf.agf_btreeblks) ||
+	    raa.longest != be32_to_cpu(old_agf.agf_longest) ||
+	    fab[0].level != be32_to_cpu(old_agf.agf_levels[XFS_BTNUM_BNOi]) ||
+	    fab[1].level != be32_to_cpu(old_agf.agf_levels[XFS_BTNUM_CNTi]) ||
+	    fab[2].level != be32_to_cpu(old_agf.agf_levels[XFS_BTNUM_RMAPi]) ||
+	    fab[3].level != be32_to_cpu(old_agf.agf_refcount_level)) {
+		pag = xfs_perag_get(mp, sc->sa.agno);
+		if (pag->pagf_init) {
+			pag->pagf_freeblks = be32_to_cpu(agf->agf_freeblks);
+			pag->pagf_btreeblks = be32_to_cpu(agf->agf_btreeblks);
+			pag->pagf_flcount = be32_to_cpu(agf->agf_flcount);
+			pag->pagf_longest = be32_to_cpu(agf->agf_longest);
+			pag->pagf_levels[XFS_BTNUM_BNOi] =
+				be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNOi]);
+			pag->pagf_levels[XFS_BTNUM_CNTi] =
+				be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]);
+			pag->pagf_levels[XFS_BTNUM_RMAPi] =
+				be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAPi]);
+			pag->pagf_refcount_level =
+				be32_to_cpu(agf->agf_refcount_level);
+		}
+		xfs_perag_put(pag);
+		sc->reset_counters = true;
+	}
+
+	/* Write this to disk. */
+	xfs_trans_buf_set_type(sc->tp, agf_bp, XFS_BLFT_AGF_BUF);
+	xfs_trans_log_buf(sc->tp, agf_bp, 0, mp->m_sb.sb_sectsize - 1);
+	return error;
+
+err:
+	if (cur)
+		xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR :
+				XFS_BTREE_NOERROR);
+	*agf = old_agf;
+	return error;
+}
+
 /* AGFL */
 
 #define XFS_SCRUB_AGFL_CHECK(fs_ok) \
@@ -779,6 +959,218 @@ xfs_scrub_agfl(
 #undef XFS_SCRUB_AGFL_XCHECK
 #undef XFS_SCRUB_AGFL_CHECK
 
+/* AGFL repair. */
+
+struct xfs_repair_agfl {
+	struct list_head		freesp_list;
+	struct list_head		agmeta_list;
+};
+
+/* Record all freespace information. */
+STATIC int
+xfs_repair_agfl_rmap_fn(
+	struct xfs_btree_cur		*cur,
+	struct xfs_rmap_irec		*rec,
+	void				*priv)
+{
+	struct xfs_repair_agfl		*ra = priv;
+	struct xfs_buf			*bp;
+	xfs_fsblock_t			fsb;
+	int				i;
+	int				error = 0;
+
+	if (xfs_scrub_should_terminate(&error))
+		return error;
+
+	/* Record all the OWN_AG blocks... */
+	if (rec->rm_owner == XFS_RMAP_OWN_AG) {
+		fsb = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_private.a.agno,
+				rec->rm_startblock);
+		error = xfs_repair_collect_btree_extent(cur->bc_mp,
+				&ra->freesp_list, fsb, rec->rm_blockcount);
+		if (error)
+			return error;
+	}
+
+	/* ...and all the rmapbt blocks... */
+	for (i = 0; i < cur->bc_nlevels && cur->bc_ptrs[i] == 1; i++) {
+		xfs_btree_get_block(cur, i, &bp);
+		if (!bp)
+			continue;
+		fsb = XFS_DADDR_TO_FSB(cur->bc_mp, bp->b_bn);
+		error = xfs_repair_collect_btree_extent(cur->bc_mp,
+				&ra->agmeta_list, fsb, 1);
+		if (error)
+			return error;
+	}
+
+	return 0;
+}
+
+/* Add a btree block to the agmeta list. */
+STATIC int
+xfs_repair_agfl_visit_btblock(
+	struct xfs_btree_cur		*cur,
+	int				level,
+	void				*priv)
+{
+	struct xfs_repair_agfl		*ra = priv;
+	struct xfs_buf			*bp;
+	xfs_fsblock_t			fsb;
+	int				error = 0;
+
+	if (xfs_scrub_should_terminate(&error))
+		return error;
+
+	xfs_btree_get_block(cur, level, &bp);
+	if (!bp)
+		return 0;
+
+	fsb = XFS_DADDR_TO_FSB(cur->bc_mp, bp->b_bn);
+	return xfs_repair_collect_btree_extent(cur->bc_mp, &ra->agmeta_list,
+			fsb, 1);
+}
+
+/* Repair the AGFL. */
+int
+xfs_repair_agfl(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_repair_agfl		ra;
+	struct xfs_owner_info		oinfo;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_buf			*agf_bp;
+	struct xfs_buf			*agfl_bp;
+	struct xfs_agf			*agf;
+	struct xfs_agfl			*agfl;
+	struct xfs_btree_cur		*cur = NULL;
+	struct xfs_perag		*pag;
+	__be32				*agfl_bno;
+	struct xfs_repair_btree_extent	*rbe;
+	struct xfs_repair_btree_extent	*n;
+	xfs_agblock_t			flcount;
+	xfs_agblock_t			agbno;
+	xfs_agblock_t			bno;
+	xfs_agblock_t			old_flcount;
+	int				error;
+
+	/* We require the rmapbt to rebuild anything. */
+	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+		return -EOPNOTSUPP;
+
+	INIT_LIST_HEAD(&ra.freesp_list);
+	INIT_LIST_HEAD(&ra.agmeta_list);
+	error = xfs_alloc_read_agf(mp, sc->tp, sc->sa.agno, 0, &agf_bp);
+	if (error)
+		return error;
+
+	error = xfs_trans_read_buf(mp, sc->tp, mp->m_ddev_targp,
+			XFS_AG_DADDR(mp, sc->sa.agno, XFS_AGFL_DADDR(mp)),
+			XFS_FSS_TO_BB(mp, 1), 0, &agfl_bp, NULL);
+	if (error)
+		return error;
+	agfl_bp->b_ops = &xfs_agfl_buf_ops;
+
+	/* Find all space used by the free space btrees & rmapbt. */
+	cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno);
+	error = xfs_rmap_query_all(cur, xfs_repair_agfl_rmap_fn, &ra);
+	if (error)
+		goto err;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+
+	/* Find all space used by bnobt. */
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno,
+			XFS_BTNUM_BNO);
+	error = xfs_btree_visit_blocks(cur, xfs_repair_agfl_visit_btblock,
+			&ra);
+	if (error)
+		goto err;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+
+	/* Find all space used by cntbt. */
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.agno,
+			XFS_BTNUM_CNT);
+	error = xfs_btree_visit_blocks(cur, xfs_repair_agfl_visit_btblock,
+			&ra);
+	if (error)
+		goto err;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	cur = NULL;
+
+	/*
+	 * Drop the freesp meta blocks that are in use by btrees.
+	 * The remaining blocks /should/ be AGFL blocks.
+	 */
+	error = xfs_repair_subtract_extents(mp, &ra.freesp_list,
+			&ra.agmeta_list);
+	if (error)
+		goto err;
+	xfs_repair_cancel_btree_extents(sc, &ra.agmeta_list);
+
+	/* Start rewriting the header. */
+	agfl = XFS_BUF_TO_AGFL(agfl_bp);
+	memset(agfl, 0xFF, mp->m_sb.sb_sectsize);
+	agfl->agfl_magicnum = cpu_to_be32(XFS_AGFL_MAGIC);
+	agfl->agfl_seqno = cpu_to_be32(sc->sa.agno);
+	uuid_copy(&agfl->agfl_uuid, &mp->m_sb.sb_meta_uuid);
+
+	/* Fill the AGFL with the remaining blocks. */
+	flcount = 0;
+	agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, agfl_bp);
+	list_for_each_entry_safe(rbe, n, &ra.freesp_list, list) {
+		agbno = XFS_FSB_TO_AGBNO(mp, rbe->fsbno);
+
+		trace_xfs_repair_agfl_insert(mp, sc->sa.agno, agbno, rbe->len);
+
+		for (bno = 0; bno < rbe->len; bno++) {
+			if (flcount >= XFS_AGFL_SIZE(mp))
+				break;
+			agfl_bno[flcount] = cpu_to_be32(agbno + bno);
+			flcount++;
+		}
+		rbe->fsbno += bno;
+		rbe->len -= bno;
+		if (rbe->len)
+			break;
+		list_del(&rbe->list);
+		kmem_free(rbe);
+	}
+
+	/* Update the AGF counters. */
+	agf = XFS_BUF_TO_AGF(agf_bp);
+	old_flcount = be32_to_cpu(agf->agf_flcount);
+	agf->agf_flfirst = 0;
+	agf->agf_flcount = cpu_to_be32(flcount);
+	agf->agf_fllast = cpu_to_be32(flcount - 1);
+
+	/* Trigger reinitialization of the in-core data. */
+	if (flcount != old_flcount) {
+		pag = xfs_perag_get(mp, sc->sa.agno);
+		if (pag->pagf_init)
+			pag->pagf_flcount = flcount;
+		xfs_perag_put(pag);
+		sc->reset_counters = true;
+	}
+
+	/* Write AGF and AGFL to disk. */
+	xfs_alloc_log_agf(sc->tp, agf_bp,
+			XFS_AGF_FLFIRST | XFS_AGF_FLLAST | XFS_AGF_FLCOUNT);
+	xfs_trans_buf_set_type(sc->tp, agfl_bp, XFS_BLFT_AGFL_BUF);
+	xfs_trans_log_buf(sc->tp, agfl_bp, 0, mp->m_sb.sb_sectsize - 1);
+
+	/* Dump any AGFL overflow. */
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_AG);
+	return xfs_repair_reap_btree_extents(sc, &ra.freesp_list, &oinfo,
+			XFS_AG_RESV_AGFL);
+err:
+	xfs_repair_cancel_btree_extents(sc, &ra.agmeta_list);
+	xfs_repair_cancel_btree_extents(sc, &ra.freesp_list);
+	if (cur)
+		xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR :
+				XFS_BTREE_NOERROR);
+	return error;
+}
+
 /* AGI */
 
 #define XFS_SCRUB_AGI_CHECK(fs_ok) \
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 9264ce3..4abb504 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -762,8 +762,8 @@ struct xfs_scrub_meta_fns {
 static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup, xfs_scrub_dummy, NULL, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_superblock, xfs_repair_superblock, NULL},
-	{xfs_scrub_setup_ag, xfs_scrub_agf, NULL, NULL},
-	{xfs_scrub_setup_ag, xfs_scrub_agfl, NULL, NULL},
+	{xfs_scrub_setup_ag, xfs_scrub_agf, xfs_repair_agf, NULL},
+	{xfs_scrub_setup_ag, xfs_scrub_agfl, xfs_repair_agfl, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agi, NULL, NULL},
 	{xfs_scrub_setup_ag_header, xfs_scrub_bnobt, NULL, NULL},
 	{xfs_scrub_setup_ag_header, xfs_scrub_cntbt, NULL, NULL},
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 92d294c..aa7049b 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -375,5 +375,7 @@ xfs_extlen_t xfs_repair_calc_ag_resblks(struct xfs_scrub_context *sc,
 /* Metadata repairers */
 
 int xfs_repair_superblock(struct xfs_scrub_context *sc);
+int xfs_repair_agf(struct xfs_scrub_context *sc);
+int xfs_repair_agfl(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */

[PATCH 47/55] xfs: rebuild the AGI

[Darrick J. Wong]

Rebuild the AGI header items with some help from the rmapbt.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/agheader.c |   98 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.c   |    2 -
 fs/xfs/scrub/common.h   |    1 
 3 files changed, 100 insertions(+), 1 deletion(-)


diff --git a/fs/xfs/scrub/agheader.c b/fs/xfs/scrub/agheader.c
index 6916933..f3ae4e0 100644
--- a/fs/xfs/scrub/agheader.c
+++ b/fs/xfs/scrub/agheader.c
@@ -1337,3 +1337,101 @@ xfs_scrub_agi(
 #undef XFS_SCRUB_AGI_XCHECK
 #undef XFS_SCRUB_AGI_CHECK
 #undef XFS_SCRUB_AGI_OP_ERROR_GOTO
+
+/* Repair the AGI. */
+int
+xfs_repair_agi(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_repair_find_ag_btree	fab[] = {
+		{XFS_RMAP_OWN_INOBT, &xfs_inobt_buf_ops, XFS_IBT_CRC_MAGIC, 0, 0},
+		{XFS_RMAP_OWN_INOBT, &xfs_inobt_buf_ops, XFS_FIBT_CRC_MAGIC, 0, 0},
+		{0, NULL, 0, 0, 0},
+	};
+	struct xfs_agi			old_agi;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_buf			*agi_bp;
+	struct xfs_buf			*agf_bp;
+	struct xfs_agi			*agi;
+	struct xfs_btree_cur		*cur;
+	struct xfs_perag		*pag;
+	xfs_agino_t			old_count;
+	xfs_agino_t			old_freecount;
+	xfs_agino_t			count;
+	xfs_agino_t			freecount;
+	int				bucket;
+	int				error;
+
+	/* We require the rmapbt to rebuild anything. */
+	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+		return -EOPNOTSUPP;
+
+	error = xfs_trans_read_buf(mp, sc->tp, mp->m_ddev_targp,
+			XFS_AG_DADDR(mp, sc->sa.agno, XFS_AGI_DADDR(mp)),
+			XFS_FSS_TO_BB(mp, 1), 0, &agi_bp, NULL);
+	if (error)
+		return error;
+	agi_bp->b_ops = &xfs_agi_buf_ops;
+
+	error = xfs_alloc_read_agf(mp, sc->tp, sc->sa.agno, 0, &agf_bp);
+	if (error)
+		return error;
+
+	/* Find the btree roots. */
+	error = xfs_repair_find_ag_btree_roots(sc, agf_bp, fab);
+	if (error)
+		return error;
+	if (fab[0].root == NULLAGBLOCK || fab[0].level > XFS_BTREE_MAXLEVELS)
+		return -EFSCORRUPTED;
+	if (xfs_sb_version_hasfinobt(&mp->m_sb) &&
+	    (fab[1].root == NULLAGBLOCK || fab[1].level > XFS_BTREE_MAXLEVELS))
+		return -EFSCORRUPTED;
+
+	/* Start rewriting the header. */
+	agi = XFS_BUF_TO_AGI(agi_bp);
+	old_agi = *agi;
+	old_count = be32_to_cpu(old_agi.agi_count);
+	old_freecount = be32_to_cpu(old_agi.agi_freecount);
+	memset(agi, 0, mp->m_sb.sb_sectsize);
+	agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
+	agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
+	agi->agi_seqno = cpu_to_be32(sc->sa.agno);
+	agi->agi_length = cpu_to_be32(xfs_scrub_ag_blocks(mp, sc->sa.agno));
+	agi->agi_newino = cpu_to_be32(NULLAGINO);
+	agi->agi_dirino = cpu_to_be32(NULLAGINO);
+	if (xfs_sb_version_hascrc(&mp->m_sb))
+		uuid_copy(&agi->agi_uuid, &mp->m_sb.sb_meta_uuid);
+	for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
+		agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
+	agi->agi_root = cpu_to_be32(fab[0].root);
+	agi->agi_level = cpu_to_be32(fab[0].level);
+	if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
+		agi->agi_free_root = cpu_to_be32(fab[1].root);
+		agi->agi_free_level = cpu_to_be32(fab[1].level);
+	}
+
+	/* Update the AGI counters. */
+	cur = xfs_inobt_init_cursor(mp, sc->tp, agi_bp, sc->sa.agno,
+			XFS_BTNUM_INO);
+	error = xfs_ialloc_count_inodes(cur, &count, &freecount);
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	if (error)
+		goto err;
+	agi->agi_count = cpu_to_be32(count);
+	agi->agi_freecount = cpu_to_be32(freecount);
+	if (old_count != count || old_freecount != freecount) {
+		pag = xfs_perag_get(mp, sc->sa.agno);
+		pag->pagi_init = 0;
+		xfs_perag_put(pag);
+		sc->reset_counters = true;
+	}
+
+	/* Write this to disk. */
+	xfs_trans_buf_set_type(sc->tp, agi_bp, XFS_BLFT_AGI_BUF);
+	xfs_trans_log_buf(sc->tp, agi_bp, 0, mp->m_sb.sb_sectsize - 1);
+	return error;
+
+err:
+	*agi = old_agi;
+	return error;
+}
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 4abb504..ceaa528 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -764,7 +764,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag, xfs_scrub_superblock, xfs_repair_superblock, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agf, xfs_repair_agf, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agfl, xfs_repair_agfl, NULL},
-	{xfs_scrub_setup_ag, xfs_scrub_agi, NULL, NULL},
+	{xfs_scrub_setup_ag, xfs_scrub_agi, xfs_repair_agi, NULL},
 	{xfs_scrub_setup_ag_header, xfs_scrub_bnobt, NULL, NULL},
 	{xfs_scrub_setup_ag_header, xfs_scrub_cntbt, NULL, NULL},
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_inobt, NULL, NULL},
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index aa7049b..e23a22a 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -377,5 +377,6 @@ xfs_extlen_t xfs_repair_calc_ag_resblks(struct xfs_scrub_context *sc,
 int xfs_repair_superblock(struct xfs_scrub_context *sc);
 int xfs_repair_agf(struct xfs_scrub_context *sc);
 int xfs_repair_agfl(struct xfs_scrub_context *sc);
+int xfs_repair_agi(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */

[PATCH 48/55] xfs: repair free space btrees

[Darrick J. Wong]

Rebuild the free space btrees from the gaps in the rmap btree.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/alloc.c  |  414 +++++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/scrub/common.c |    4 
 fs/xfs/scrub/common.h |    5 +
 3 files changed, 421 insertions(+), 2 deletions(-)


diff --git a/fs/xfs/scrub/alloc.c b/fs/xfs/scrub/alloc.c
index 8d81875..da7fbce 100644
--- a/fs/xfs/scrub/alloc.c
+++ b/fs/xfs/scrub/alloc.c
@@ -30,9 +30,14 @@
 #include "xfs_trans.h"
 #include "xfs_trace.h"
 #include "xfs_sb.h"
+#include "xfs_inode.h"
 #include "xfs_rmap.h"
+#include "xfs_log.h"
+#include "xfs_trans_priv.h"
 #include "xfs_alloc.h"
+#include "xfs_alloc_btree.h"
 #include "xfs_ialloc.h"
+#include "xfs_rmap_btree.h"
 #include "xfs_refcount.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
@@ -58,6 +63,39 @@ xfs_scrub_setup_ag_header(
 	return error;
 }
 
+/*
+ * Set us up with AG headers and btree cursors.  Push everything out
+ * of the log so that the busy extent list is empty.
+ */
+int
+xfs_scrub_setup_ag_allocbt(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+	int				error;
+
+	/*
+	 * Push everything out of the log onto disk prior to repairing.
+	 * The busy extent list must be cleared before we can do anything.
+	 */
+	if (sm->sm_flags & XFS_SCRUB_FLAG_REPAIR) {
+		error = _xfs_log_force(mp, XFS_LOG_SYNC, NULL);
+		if (error)
+			goto out;
+		xfs_ail_push_all_sync(mp->m_ail);
+	}
+
+	error = xfs_scrub_setup_ag_header(sc, ip, sm, retry_deadlocked);
+	if (error)
+		goto out;
+	sc->retry = retry_deadlocked;
+out:
+	return error;
+}
+
 /* Free space btree scrubber. */
 
 /* Scrub a bnobt/cntbt record. */
@@ -184,3 +222,379 @@ xfs_scrub_cntbt(
 {
 	return xfs_scrub_allocbt(sc, XFS_BTNUM_CNT);
 }
+
+/* Free space btree repair. */
+
+struct xfs_repair_alloc_extent {
+	struct list_head		list;
+	xfs_agblock_t			bno;
+	xfs_extlen_t			len;
+};
+
+struct xfs_repair_alloc {
+	struct list_head		extlist;
+	struct list_head		btlist;	  /* OWN_AG blocks */
+	struct list_head		nobtlist; /* rmapbt/agfl blocks */
+	xfs_agblock_t			next_bno;
+	uint64_t			nr_records;
+};
+
+/* Record extents that aren't in use from gaps in the rmap records. */
+STATIC int
+xfs_repair_alloc_extent_fn(
+	struct xfs_btree_cur		*cur,
+	struct xfs_rmap_irec		*rec,
+	void				*priv)
+{
+	struct xfs_repair_alloc		*ra = priv;
+	struct xfs_repair_alloc_extent	*rae;
+	struct xfs_buf			*bp;
+	struct xfs_mount		*mp = cur->bc_mp;
+	xfs_fsblock_t			fsb;
+	int				i;
+	int				error;
+
+	/* Record all the OWN_AG blocks... */
+	if (rec->rm_owner == XFS_RMAP_OWN_AG) {
+		fsb = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_private.a.agno,
+				rec->rm_startblock);
+		error = xfs_repair_collect_btree_extent(cur->bc_mp,
+				&ra->btlist, fsb, rec->rm_blockcount);
+		if (error)
+			return error;
+	}
+
+	/* ...and all the rmapbt blocks... */
+	for (i = 0; i < cur->bc_nlevels && cur->bc_ptrs[i] == 1; i++) {
+		xfs_btree_get_block(cur, i, &bp);
+		if (!bp)
+			continue;
+		fsb = XFS_DADDR_TO_FSB(cur->bc_mp, bp->b_bn);
+		error = xfs_repair_collect_btree_extent(cur->bc_mp,
+				&ra->nobtlist, fsb, 1);
+		if (error)
+			return error;
+	}
+
+	/* ...and all the free space. */
+	if (rec->rm_startblock > ra->next_bno) {
+		trace_xfs_repair_alloc_extent_fn(mp, cur->bc_private.a.agno,
+				rec->rm_startblock, rec->rm_blockcount,
+				rec->rm_owner, rec->rm_offset, rec->rm_flags);
+
+		rae = kmem_alloc(sizeof(*rae), KM_NOFS);
+		if (!rae)
+			return -ENOMEM;
+		INIT_LIST_HEAD(&rae->list);
+		rae->bno = ra->next_bno;
+		rae->len = rec->rm_startblock - ra->next_bno;
+		list_add_tail(&rae->list, &ra->extlist);
+		ra->nr_records++;
+	}
+	ra->next_bno = max_t(xfs_agblock_t, ra->next_bno,
+			rec->rm_startblock + rec->rm_blockcount);
+	return 0;
+}
+
+/* Find the longest free extent in the list. */
+static struct xfs_repair_alloc_extent *
+xfs_repair_allocbt_get_longest(
+	struct xfs_repair_alloc		*ra)
+{
+	struct xfs_repair_alloc_extent	*rae;
+	struct xfs_repair_alloc_extent	*longest = NULL;
+
+	list_for_each_entry(rae, &ra->extlist, list)
+		if (!longest || rae->len > longest->len)
+			longest = rae;
+	return longest;
+}
+
+/* Collect an AGFL block for the not-to-release list. */
+static int
+xfs_repair_collect_agfl_block(
+	struct xfs_scrub_context	*sc,
+	xfs_agblock_t			bno,
+	void				*data)
+{
+	struct xfs_repair_alloc		*ra = data;
+	xfs_fsblock_t			fsb;
+
+	fsb = XFS_AGB_TO_FSB(sc->tp->t_mountp, sc->sa.agno, bno);
+	return xfs_repair_collect_btree_extent(sc->tp->t_mountp,
+			&ra->nobtlist, fsb, 1);
+}
+
+/* Compare two btree extents. */
+static int
+xfs_repair_allocbt_extent_cmp(
+	void				*priv,
+	struct list_head		*a,
+	struct list_head		*b)
+{
+	struct xfs_repair_alloc_extent	*ap;
+	struct xfs_repair_alloc_extent	*bp;
+
+	ap = container_of(a, struct xfs_repair_alloc_extent, list);
+	bp = container_of(b, struct xfs_repair_alloc_extent, list);
+
+	if (ap->bno > bp->bno)
+		return 1;
+	else if (ap->bno < bp->bno)
+		return -1;
+	return 0;
+}
+
+/* Put an extent onto the free list. */
+STATIC int
+xfs_repair_allocbt_free_extent(
+	struct xfs_scrub_context	*sc,
+	xfs_fsblock_t			fsbno,
+	xfs_extlen_t			len,
+	struct xfs_owner_info		*oinfo)
+{
+	int				error;
+
+	error = xfs_free_extent(sc->tp, fsbno, len, oinfo, 0);
+	if (error)
+		return error;
+	error = xfs_repair_roll_ag_trans(sc);
+	if (error)
+		return error;
+	return xfs_mod_fdblocks(sc->tp->t_mountp, -(int64_t)len, false);
+}
+
+/* Repair the freespace btrees for some AG. */
+int
+xfs_repair_allocbt(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_repair_alloc		ra;
+	struct xfs_owner_info		oinfo;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_btree_cur		*cur = NULL;
+	struct xfs_repair_alloc_extent	*longest;
+	struct xfs_repair_alloc_extent	*rae;
+	struct xfs_repair_alloc_extent	*n;
+	struct xfs_perag		*pag;
+	struct xfs_agf			*agf;
+	struct xfs_buf			*bp;
+	xfs_fsblock_t			bnofsb;
+	xfs_fsblock_t			cntfsb;
+	xfs_extlen_t			oldf;
+	xfs_extlen_t			nr_blocks;
+	xfs_agblock_t			agend;
+	int				error;
+
+	/* We require the rmapbt to rebuild anything. */
+	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+		return -EOPNOTSUPP;
+
+	/*
+	 * Make sure the busy extent list is clear because we can't put
+	 * extents on there twice.
+	 */
+	pag = xfs_perag_get(mp, sc->sa.agno);
+	spin_lock(&pag->pagb_lock);
+	if (pag->pagb_tree.rb_node) {
+		spin_unlock(&pag->pagb_lock);
+		xfs_perag_put(pag);
+		return -EDEADLOCK;
+	}
+	spin_unlock(&pag->pagb_lock);
+	xfs_perag_put(pag);
+
+	/*
+	 * Collect all reverse mappings for free extents, and the rmapbt
+	 * blocks.  We can discover the rmapbt blocks completely from a
+	 * query_all handler because there are always rmapbt entries.
+	 * (One cannot use on query_all to visit all of a btree's blocks
+	 * unless that btree is guaranteed to have at least one entry.)
+	 */
+	INIT_LIST_HEAD(&ra.extlist);
+	INIT_LIST_HEAD(&ra.btlist);
+	INIT_LIST_HEAD(&ra.nobtlist);
+	ra.next_bno = 0;
+	ra.nr_records = 0;
+	cur = xfs_rmapbt_init_cursor(mp, sc->tp, sc->sa.agf_bp, sc->sa.agno);
+	error = xfs_rmap_query_all(cur, xfs_repair_alloc_extent_fn, &ra);
+	if (error)
+		goto out;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	cur = NULL;
+
+	/* Insert a record for space between the last rmap and EOAG. */
+	agf = XFS_BUF_TO_AGF(sc->sa.agf_bp);
+	agend = be32_to_cpu(agf->agf_length);
+	if (ra.next_bno < agend) {
+		rae = kmem_alloc(sizeof(*rae), KM_NOFS);
+		if (!rae) {
+			error = -ENOMEM;
+			goto out;
+		}
+		INIT_LIST_HEAD(&rae->list);
+		rae->bno = ra.next_bno;
+		rae->len = agend - ra.next_bno;
+		list_add_tail(&rae->list, &ra.extlist);
+		ra.nr_records++;
+	}
+
+	/* Collect all the AGFL blocks. */
+	error = xfs_scrub_walk_agfl(sc, xfs_repair_collect_agfl_block, &ra);
+	if (error)
+		goto out;
+
+	/* Do we actually have enough space to do this? */
+	pag = xfs_perag_get(mp, sc->sa.agno);
+	nr_blocks = 2 * xfs_allocbt_calc_size(mp, ra.nr_records);
+	if (!xfs_repair_ag_has_space(pag, nr_blocks, XFS_AG_RESV_NONE)) {
+		xfs_perag_put(pag);
+		error = -ENOSPC;
+		goto out;
+	}
+	xfs_perag_put(pag);
+
+	/* Allocate new bnobt root. */
+	longest = xfs_repair_allocbt_get_longest(&ra);
+	if (longest == NULL) {
+		error = -ENOSPC;
+		goto out;
+	}
+	bnofsb = XFS_AGB_TO_FSB(mp, sc->sa.agno, longest->bno);
+	longest->bno++;
+	longest->len--;
+
+	/* Allocate new cntbt root. */
+	if (longest->len == 0) {
+		list_del(&longest->list);
+		kmem_free(longest);
+		longest = xfs_repair_allocbt_get_longest(&ra);
+		if (longest == NULL) {
+			error = -ENOSPC;
+			goto out;
+		}
+	}
+	cntfsb = XFS_AGB_TO_FSB(mp, sc->sa.agno, longest->bno);
+	longest->bno++;
+	longest->len--;
+	if (longest->len == 0) {
+		list_del(&longest->list);
+		kmem_free(longest);
+		longest = xfs_repair_allocbt_get_longest(&ra);
+	}
+
+	agf = XFS_BUF_TO_AGF(sc->sa.agf_bp);
+	/* Initialize new bnobt root. */
+	error = xfs_repair_init_btblock(sc, bnofsb, &bp, XFS_ABTB_CRC_MAGIC,
+			&xfs_allocbt_buf_ops);
+	if (error)
+		goto out;
+	agf->agf_roots[XFS_BTNUM_BNOi] =
+			cpu_to_be32(XFS_FSB_TO_AGBNO(mp, bnofsb));
+	agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
+
+	/* Initialize new cntbt root. */
+	error = xfs_repair_init_btblock(sc, cntfsb, &bp, XFS_ABTC_CRC_MAGIC,
+			&xfs_allocbt_buf_ops);
+	if (error)
+		goto out;
+	agf->agf_roots[XFS_BTNUM_CNTi] =
+			cpu_to_be32(XFS_FSB_TO_AGBNO(mp, cntfsb));
+	agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
+
+	/*
+	 * Since we're abandoning the old bnobt/cntbt, we have to
+	 * decrease fdblocks by the # of blocks in those trees.
+	 * btreeblks counts the non-root blocks of the free space
+	 * and rmap btrees.  Do this before resetting the AGF counters.
+	 */
+	pag = xfs_perag_get(mp, sc->sa.agno);
+	oldf = pag->pagf_btreeblks + 2;
+	oldf -= (be32_to_cpu(agf->agf_rmap_blocks) - 1);
+	error = xfs_mod_fdblocks(mp, -(int64_t)oldf, false);
+	if (error) {
+		xfs_perag_put(pag);
+		goto out;
+	}
+
+	/* Reset the perag info. */
+	pag->pagf_btreeblks = be32_to_cpu(agf->agf_rmap_blocks) - 1;
+	pag->pagf_freeblks = 0;
+	pag->pagf_longest = 0;
+	pag->pagf_levels[XFS_BTNUM_BNOi] =
+			be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNOi]);
+	pag->pagf_levels[XFS_BTNUM_CNTi] =
+			be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNTi]);
+
+	/* Now reset the AGF counters. */
+	agf->agf_btreeblks = cpu_to_be32(pag->pagf_btreeblks);
+	agf->agf_freeblks = cpu_to_be32(pag->pagf_freeblks);
+	agf->agf_longest = cpu_to_be32(pag->pagf_longest);
+	xfs_perag_put(pag);
+	xfs_alloc_log_agf(sc->tp, sc->sa.agf_bp,
+			XFS_AGF_ROOTS | XFS_AGF_LEVELS | XFS_AGF_BTREEBLKS |
+			XFS_AGF_LONGEST | XFS_AGF_FREEBLKS);
+	error = xfs_repair_roll_ag_trans(sc);
+	if (error)
+		goto out;
+
+	/*
+	 * Insert the longest free extent in case it's necessary to
+	 * refresh the AGFL with multiple blocks.
+	 */
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_UNKNOWN);
+	if (longest && longest->len == 0) {
+		error = xfs_repair_allocbt_free_extent(sc,
+				XFS_AGB_TO_FSB(mp, sc->sa.agno, longest->bno),
+				longest->len, &oinfo);
+		if (error)
+			goto out;
+		list_del(&longest->list);
+		kmem_free(longest);
+	}
+
+	/* Insert records into the new btrees. */
+	list_sort(NULL, &ra.extlist, xfs_repair_allocbt_extent_cmp);
+	list_for_each_entry_safe(rae, n, &ra.extlist, list) {
+		error = xfs_repair_allocbt_free_extent(sc,
+				XFS_AGB_TO_FSB(mp, sc->sa.agno, rae->bno),
+				rae->len, &oinfo);
+		if (error)
+			goto out;
+		list_del(&rae->list);
+		kmem_free(rae);
+	}
+
+	/* Add rmap records for the btree roots */
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_AG);
+	error = xfs_rmap_alloc(sc->tp, sc->sa.agf_bp, sc->sa.agno,
+			XFS_FSB_TO_AGBNO(mp, bnofsb), 1, &oinfo);
+	if (error)
+		goto out;
+	error = xfs_rmap_alloc(sc->tp, sc->sa.agf_bp, sc->sa.agno,
+			XFS_FSB_TO_AGBNO(mp, cntfsb), 1, &oinfo);
+	if (error)
+		goto out;
+
+	/* Free all the OWN_AG blocks that are not in the rmapbt/agfl. */
+	error = xfs_repair_subtract_extents(mp, &ra.btlist, &ra.nobtlist);
+	if (error)
+		goto out;
+	xfs_repair_cancel_btree_extents(sc, &ra.nobtlist);
+	error = xfs_repair_reap_btree_extents(sc, &ra.btlist, &oinfo,
+			XFS_AG_RESV_NONE);
+	if (error)
+		goto out;
+
+	return 0;
+out:
+	xfs_repair_cancel_btree_extents(sc, &ra.btlist);
+	xfs_repair_cancel_btree_extents(sc, &ra.nobtlist);
+	if (cur)
+		xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
+	list_for_each_entry_safe(rae, n, &ra.extlist, list) {
+		list_del(&rae->list);
+		kmem_free(rae);
+	}
+	return error;
+}
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index ceaa528..d975be9 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -765,8 +765,8 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag, xfs_scrub_agf, xfs_repair_agf, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agfl, xfs_repair_agfl, NULL},
 	{xfs_scrub_setup_ag, xfs_scrub_agi, xfs_repair_agi, NULL},
-	{xfs_scrub_setup_ag_header, xfs_scrub_bnobt, NULL, NULL},
-	{xfs_scrub_setup_ag_header, xfs_scrub_cntbt, NULL, NULL},
+	{xfs_scrub_setup_ag_allocbt, xfs_scrub_bnobt, xfs_repair_allocbt, NULL},
+	{xfs_scrub_setup_ag_allocbt, xfs_scrub_cntbt, xfs_repair_allocbt, NULL},
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_inobt, NULL, NULL},
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_finobt, NULL, xfs_sb_version_hasfinobt},
 	{xfs_scrub_setup_ag_header, xfs_scrub_rmapbt, NULL, xfs_sb_version_hasrmapbt},
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index e23a22a..4dc5890 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -275,6 +275,10 @@ int xfs_scrub_setup_ag_header(struct xfs_scrub_context *sc,
 			      struct xfs_inode *ip,
 			      struct xfs_scrub_metadata *sm,
 			      bool retry_deadlocked);
+int xfs_scrub_setup_ag_allocbt(struct xfs_scrub_context *sc,
+			       struct xfs_inode *ip,
+			       struct xfs_scrub_metadata *sm,
+			       bool retry_deadlocked);
 int xfs_scrub_setup_ag_iallocbt(struct xfs_scrub_context *sc,
 				struct xfs_inode *ip,
 				struct xfs_scrub_metadata *sm,
@@ -378,5 +382,6 @@ int xfs_repair_superblock(struct xfs_scrub_context *sc);
 int xfs_repair_agf(struct xfs_scrub_context *sc);
 int xfs_repair_agfl(struct xfs_scrub_context *sc);
 int xfs_repair_agi(struct xfs_scrub_context *sc);
+int xfs_repair_allocbt(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */

[PATCH 49/55] xfs: repair inode btrees

[Darrick J. Wong]

Use the rmapbt to find inode chunks, query the chunks to compute
hole and free masks, and with that information rebuild the inobt
and finobt.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_ialloc.c |    2 
 fs/xfs/libxfs/xfs_ialloc.h |    3 
 fs/xfs/scrub/common.c      |    4 
 fs/xfs/scrub/common.h      |    1 
 fs/xfs/scrub/ialloc.c      |  380 ++++++++++++++++++++++++++++++++++++++++++++
 5 files changed, 387 insertions(+), 3 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_ialloc.c b/fs/xfs/libxfs/xfs_ialloc.c
index 0fb7ba0..e34053d 100644
--- a/fs/xfs/libxfs/xfs_ialloc.c
+++ b/fs/xfs/libxfs/xfs_ialloc.c
@@ -147,7 +147,7 @@ xfs_inobt_get_rec(
 /*
  * Insert a single inobt record. Cursor must already point to desired location.
  */
-STATIC int
+int
 xfs_inobt_insert_rec(
 	struct xfs_btree_cur	*cur,
 	__uint16_t		holemask,
diff --git a/fs/xfs/libxfs/xfs_ialloc.h b/fs/xfs/libxfs/xfs_ialloc.h
index 17f0f1b..95216bb 100644
--- a/fs/xfs/libxfs/xfs_ialloc.h
+++ b/fs/xfs/libxfs/xfs_ialloc.h
@@ -177,5 +177,8 @@ int xfs_ialloc_has_inode_record(struct xfs_btree_cur *cur, xfs_agino_t low,
 		xfs_agino_t high, bool *exists);
 int xfs_ialloc_count_inodes(struct xfs_btree_cur *cur, xfs_agino_t *count,
 		xfs_agino_t *freecount);
+int xfs_inobt_insert_rec(struct xfs_btree_cur *cur, __uint16_t holemask,
+		__uint8_t count, __int32_t freecount, xfs_inofree_t free,
+		int *stat);
 
 #endif	/* __XFS_IALLOC_H__ */
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index d975be9..4b77550 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -767,8 +767,8 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag, xfs_scrub_agi, xfs_repair_agi, NULL},
 	{xfs_scrub_setup_ag_allocbt, xfs_scrub_bnobt, xfs_repair_allocbt, NULL},
 	{xfs_scrub_setup_ag_allocbt, xfs_scrub_cntbt, xfs_repair_allocbt, NULL},
-	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_inobt, NULL, NULL},
-	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_finobt, NULL, xfs_sb_version_hasfinobt},
+	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_inobt, xfs_repair_iallocbt, NULL},
+	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_finobt, xfs_repair_iallocbt, xfs_sb_version_hasfinobt},
 	{xfs_scrub_setup_ag_header, xfs_scrub_rmapbt, NULL, xfs_sb_version_hasrmapbt},
 	{xfs_scrub_setup_ag_header, xfs_scrub_refcountbt, NULL, xfs_sb_version_hasreflink},
 	{xfs_scrub_setup_inode_raw, xfs_scrub_inode, NULL, NULL},
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 4dc5890..641e0c2 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -383,5 +383,6 @@ int xfs_repair_agf(struct xfs_scrub_context *sc);
 int xfs_repair_agfl(struct xfs_scrub_context *sc);
 int xfs_repair_agi(struct xfs_scrub_context *sc);
 int xfs_repair_allocbt(struct xfs_scrub_context *sc);
+int xfs_repair_iallocbt(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/ialloc.c b/fs/xfs/scrub/ialloc.c
index 7fa4ebf..217e9db 100644
--- a/fs/xfs/scrub/ialloc.c
+++ b/fs/xfs/scrub/ialloc.c
@@ -38,7 +38,9 @@
 #include "xfs_log.h"
 #include "xfs_trans_priv.h"
 #include "xfs_alloc.h"
+#include "xfs_rmap_btree.h"
 #include "xfs_refcount.h"
+#include "xfs_error.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -430,3 +432,381 @@ xfs_scrub_finobt(
 {
 	return xfs_scrub_iallocbt(sc, XFS_BTNUM_FINO);
 }
+
+/* Inode btree repair. */
+
+struct xfs_repair_ialloc_extent {
+	struct list_head		list;
+	xfs_inofree_t			freemask;
+	xfs_agino_t			startino;
+	unsigned int			count;
+	unsigned int			usedcount;
+	__uint16_t			holemask;
+};
+
+struct xfs_repair_ialloc {
+	struct list_head		extlist;
+	struct list_head		btlist;
+	uint64_t			nr_records;
+};
+
+/* Set usedmask if the inode is in use. */
+STATIC int
+xfs_repair_ialloc_check_free(
+	struct xfs_trans	*tp,
+	struct xfs_buf		*bp,
+	xfs_ino_t		fsino,
+	xfs_agino_t		chunkino,
+	xfs_agino_t		clusterino,
+	xfs_inofree_t		*usedmask,
+	int			*usedcount)
+{
+	struct xfs_mount	*mp = tp->t_mountp;
+	struct xfs_inode	*ip;
+	struct xfs_dinode	*dip;
+	int			error;
+
+	error = xfs_iget(mp, tp, fsino + clusterino, XFS_IGET_HITONLY, 0, &ip);
+	if (error == -ENOENT) {
+		return 0;
+	} else if (!error && ip) {
+		if (VFS_I(ip)->i_mode) {
+			*usedmask |= 1ULL << (chunkino + clusterino);
+			(*usedcount)++;
+		}
+		IRELE(ip);
+		return 0;
+	}
+
+	dip = xfs_buf_offset(bp, clusterino * mp->m_sb.sb_inodesize);
+	if (be16_to_cpu(dip->di_magic) != XFS_DINODE_MAGIC)
+		return -EFSCORRUPTED;
+
+	if (dip->di_version >= 3 &&
+	    be64_to_cpu(dip->di_ino) != fsino + clusterino)
+		return -EFSCORRUPTED;
+
+	if (dip->di_mode) {
+		*usedmask |= 1ULL << (chunkino + clusterino);
+		(*usedcount)++;
+	}
+
+	return 0;
+}
+
+/* Record extents that belong to inode btrees. */
+STATIC int
+xfs_repair_ialloc_extent_fn(
+	struct xfs_btree_cur		*cur,
+	struct xfs_rmap_irec		*rec,
+	void				*priv)
+{
+	struct xfs_imap			imap;
+	struct xfs_repair_ialloc	*ri = priv;
+	struct xfs_repair_ialloc_extent	*rie;
+	struct xfs_dinode		*dip;
+	struct xfs_buf			*bp;
+	struct xfs_mount		*mp = cur->bc_mp;
+	xfs_ino_t			fsino;
+	xfs_inofree_t			usedmask;
+	xfs_fsblock_t			fsbno;
+	xfs_agnumber_t			agno;
+	xfs_agblock_t			agbno;
+	xfs_agino_t			agino;
+	xfs_agino_t			startino;
+	xfs_agino_t			chunkino;
+	xfs_agino_t			nr_inodes;
+	xfs_agino_t			i;
+	__uint16_t			fillmask;
+	int				blks_per_cluster;
+	int				usedcount;
+	int				error = 0;
+
+	if (xfs_scrub_should_terminate(&error))
+		return error;
+
+	/* Fragment of the old btrees; dispose of them later. */
+	if (rec->rm_owner == XFS_RMAP_OWN_INOBT) {
+		fsbno = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_private.a.agno,
+				rec->rm_startblock);
+		return xfs_repair_collect_btree_extent(mp, &ri->btlist,
+				fsbno, rec->rm_blockcount);
+	}
+
+	/* Skip extents which are not owned by this inode and fork. */
+	if (rec->rm_owner != XFS_RMAP_OWN_INODES)
+		return 0;
+
+	agno = cur->bc_private.a.agno;
+	blks_per_cluster = xfs_icluster_size_fsb(mp);
+	nr_inodes = XFS_OFFBNO_TO_AGINO(mp, blks_per_cluster, 0);
+
+	if (rec->rm_startblock % blks_per_cluster != 0)
+		return -EFSCORRUPTED;
+
+	trace_xfs_repair_ialloc_extent_fn(mp, cur->bc_private.a.agno,
+			rec->rm_startblock, rec->rm_blockcount, rec->rm_owner,
+			rec->rm_offset, rec->rm_flags);
+
+	for (agbno = rec->rm_startblock;
+	     agbno < rec->rm_startblock + rec->rm_blockcount;
+	     agbno += blks_per_cluster) {
+		agino = XFS_OFFBNO_TO_AGINO(mp, agbno, 0);
+		fsino = XFS_AGINO_TO_INO(mp, agno, agino);
+		chunkino = agino & (XFS_INODES_PER_CHUNK - 1);
+		startino = agino & ~(XFS_INODES_PER_CHUNK - 1);
+
+		/* Which inodes are not holes? */
+		fillmask = xfs_inobt_maskn(
+				chunkino / XFS_INODES_PER_HOLEMASK_BIT,
+				nr_inodes / XFS_INODES_PER_HOLEMASK_BIT);
+
+		/* Grab the inode cluster buffer. */
+		imap.im_blkno = XFS_AGB_TO_DADDR(mp, agno, agbno);
+		imap.im_len = XFS_FSB_TO_BB(mp, blks_per_cluster);
+		imap.im_boffset = 0;
+
+		error = xfs_imap_to_bp(mp, cur->bc_tp, &imap,
+				&dip, &bp, 0, XFS_IGET_UNTRUSTED);
+		if (error)
+			return error;
+
+		/* Which inodes are free? */
+		for (usedmask = 0, usedcount = 0, i = 0; i < nr_inodes; i++) {
+			error = xfs_repair_ialloc_check_free(cur->bc_tp, bp,
+					fsino, chunkino, i, &usedmask,
+					&usedcount);
+			if (error) {
+				xfs_trans_brelse(cur->bc_tp, bp);
+				return error;
+			}
+		}
+		xfs_trans_brelse(cur->bc_tp, bp);
+
+		/*
+		 * If the last item in the list is our chunk record,
+		 * update that.
+		 */
+		if (!list_empty(&ri->extlist)) {
+			rie = list_last_entry(&ri->extlist,
+					struct xfs_repair_ialloc_extent, list);
+			if (rie->startino == startino) {
+				rie->freemask &= ~usedmask;
+				rie->holemask &= ~fillmask;
+				rie->count += nr_inodes;
+				rie->usedcount += usedcount;
+				continue;
+			}
+		}
+
+		/* New inode chunk; add to the list. */
+		rie = kmem_alloc(sizeof(*rie), KM_NOFS);
+		if (!rie)
+			return -ENOMEM;
+
+		INIT_LIST_HEAD(&rie->list);
+		rie->startino = startino;
+		rie->freemask = XFS_INOBT_ALL_FREE & ~usedmask;
+		rie->holemask = XFS_INOBT_ALL_FREE & ~fillmask;
+		rie->count = nr_inodes;
+		rie->usedcount = usedcount;
+		list_add_tail(&rie->list, &ri->extlist);
+		ri->nr_records++;
+	}
+
+	return 0;
+}
+
+/* Compare two ialloc extents. */
+static int
+xfs_repair_ialloc_extent_cmp(
+	void				*priv,
+	struct list_head		*a,
+	struct list_head		*b)
+{
+	struct xfs_repair_ialloc_extent	*ap;
+	struct xfs_repair_ialloc_extent	*bp;
+
+	ap = container_of(a, struct xfs_repair_ialloc_extent, list);
+	bp = container_of(b, struct xfs_repair_ialloc_extent, list);
+
+	if (ap->startino > bp->startino)
+		return 1;
+	else if (ap->startino < bp->startino)
+		return -1;
+	return 0;
+}
+
+/* Repair both inode btrees. */
+int
+xfs_repair_iallocbt(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_repair_ialloc	ri;
+	struct xfs_owner_info		oinfo;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_buf			*bp;
+	struct xfs_repair_ialloc_extent	*rie;
+	struct xfs_repair_ialloc_extent	*n;
+	struct xfs_agi			*agi;
+	struct xfs_btree_cur		*cur = NULL;
+	struct xfs_perag		*pag;
+	xfs_fsblock_t			inofsb;
+	xfs_fsblock_t			finofsb;
+	xfs_extlen_t			nr_blocks;
+	unsigned int			count;
+	unsigned int			usedcount;
+	int				stat;
+	int				logflags;
+	int				error = 0;
+
+	/* We require the rmapbt to rebuild anything. */
+	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+		return -EOPNOTSUPP;
+
+	/* Collect all reverse mappings for inode blocks. */
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_INOBT);
+	INIT_LIST_HEAD(&ri.extlist);
+	INIT_LIST_HEAD(&ri.btlist);
+	ri.nr_records = 0;
+	cur = xfs_rmapbt_init_cursor(mp, sc->tp, sc->sa.agf_bp, sc->sa.agno);
+	error = xfs_rmap_query_all(cur, xfs_repair_ialloc_extent_fn, &ri);
+	if (error)
+		goto out;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	cur = NULL;
+
+	/* Do we actually have enough space to do this? */
+	pag = xfs_perag_get(mp, sc->sa.agno);
+	nr_blocks = xfs_iallocbt_calc_size(mp, ri.nr_records);
+	if (xfs_sb_version_hasfinobt(&mp->m_sb))
+		nr_blocks *= 2;
+	if (!xfs_repair_ag_has_space(pag, nr_blocks, XFS_AG_RESV_NONE)) {
+		xfs_perag_put(pag);
+		error = -ENOSPC;
+		goto out;
+	}
+	xfs_perag_put(pag);
+
+	agi = XFS_BUF_TO_AGI(sc->sa.agi_bp);
+	/* Initialize new btree roots. */
+	error = xfs_repair_alloc_ag_block(sc, &oinfo, &inofsb,
+			XFS_AG_RESV_NONE);
+	if (error)
+		goto out;
+	error = xfs_repair_init_btblock(sc, inofsb, &bp, XFS_IBT_CRC_MAGIC,
+			&xfs_inobt_buf_ops);
+	if (error)
+		goto out;
+	agi->agi_root = cpu_to_be32(XFS_FSB_TO_AGBNO(mp, inofsb));
+	agi->agi_level = cpu_to_be32(1);
+	logflags = XFS_AGI_ROOT | XFS_AGI_LEVEL;
+
+	if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
+		error = xfs_repair_alloc_ag_block(sc, &oinfo, &finofsb,
+				XFS_AG_RESV_NONE);
+		if (error)
+			goto out;
+		error = xfs_repair_init_btblock(sc, finofsb, &bp,
+				XFS_FIBT_CRC_MAGIC, &xfs_inobt_buf_ops);
+		if (error)
+			goto out;
+		agi->agi_free_root = cpu_to_be32(XFS_FSB_TO_AGBNO(mp, finofsb));
+		agi->agi_free_level = cpu_to_be32(1);
+		logflags |= XFS_AGI_FREE_ROOT | XFS_AGI_FREE_LEVEL;
+	}
+
+	xfs_ialloc_log_agi(sc->tp, sc->sa.agi_bp, logflags);
+	error = xfs_repair_roll_ag_trans(sc);
+	if (error)
+		goto out;
+
+	/* Insert records into the new btrees. */
+	count = 0;
+	usedcount = 0;
+	list_sort(NULL, &ri.extlist, xfs_repair_ialloc_extent_cmp);
+	list_for_each_entry_safe(rie, n, &ri.extlist, list) {
+		count += rie->count;
+		usedcount += rie->usedcount;
+
+		trace_xfs_repair_ialloc_insert(mp, sc->sa.agno, rie->startino,
+				rie->holemask, rie->count,
+				rie->count - rie->usedcount, rie->freemask);
+
+		/* Insert into the inobt. */
+		cur = xfs_inobt_init_cursor(mp, sc->tp, sc->sa.agi_bp,
+				sc->sa.agno, XFS_BTNUM_INO);
+		error = xfs_inobt_lookup(cur, rie->startino, XFS_LOOKUP_EQ,
+				&stat);
+		if (error)
+			goto out;
+		XFS_WANT_CORRUPTED_GOTO(mp, stat == 0, out);
+		error = xfs_inobt_insert_rec(cur, rie->holemask, rie->count,
+				rie->count - rie->usedcount, rie->freemask,
+				&stat);
+		if (error)
+			goto out;
+		XFS_WANT_CORRUPTED_GOTO(mp, stat == 1, out);
+		xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+		cur = NULL;
+
+		/* Insert into the finobt. */
+		if (rie->count != rie->usedcount &&
+		    xfs_sb_version_hasfinobt(&mp->m_sb)) {
+			cur = xfs_inobt_init_cursor(mp, sc->tp, sc->sa.agi_bp,
+					sc->sa.agno, XFS_BTNUM_FINO);
+			error = xfs_inobt_lookup(cur, rie->startino,
+					XFS_LOOKUP_EQ, &stat);
+			if (error)
+				goto out;
+			XFS_WANT_CORRUPTED_GOTO(mp, stat == 0, out);
+			error = xfs_inobt_insert_rec(cur, rie->holemask,
+					rie->count, rie->count - rie->usedcount,
+					rie->freemask, &stat);
+			if (error)
+				goto out;
+			XFS_WANT_CORRUPTED_GOTO(mp, stat == 1, out);
+			xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+			cur = NULL;
+		}
+
+		error = xfs_repair_roll_ag_trans(sc);
+		if (error)
+			goto out;
+
+		list_del(&rie->list);
+		kmem_free(rie);
+	}
+
+	/* Update the AGI counters. */
+	agi = XFS_BUF_TO_AGI(sc->sa.agi_bp);
+	if (be32_to_cpu(agi->agi_count) != count ||
+	    be32_to_cpu(agi->agi_freecount) != count - usedcount) {
+		pag = xfs_perag_get(mp, sc->sa.agno);
+		pag->pagi_init = 0;
+		xfs_perag_put(pag);
+
+		agi->agi_count = cpu_to_be32(count);
+		agi->agi_freecount = cpu_to_be32(count - usedcount);
+		xfs_ialloc_log_agi(sc->tp, sc->sa.agi_bp,
+				XFS_AGI_COUNT | XFS_AGI_FREECOUNT);
+		sc->reset_counters = true;
+	}
+
+	/* Free the old inode btree blocks if they're not in use. */
+	error = xfs_repair_reap_btree_extents(sc, &ri.btlist, &oinfo,
+			XFS_AG_RESV_NONE);
+	if (error)
+		goto out;
+
+	return error;
+out:
+	if (cur)
+		xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
+	xfs_repair_cancel_btree_extents(sc, &ri.btlist);
+	list_for_each_entry_safe(rie, n, &ri.extlist, list) {
+		list_del(&rie->list);
+		kmem_free(rie);
+	}
+	return error;
+}

[PATCH 50/55] xfs: rebuild the rmapbt

[Darrick J. Wong]

Rebuild the reverse mapping btree from all primary metadata.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_refcount.c |    2 
 fs/xfs/libxfs/xfs_refcount.h |    3 
 fs/xfs/libxfs/xfs_rmap.c     |   28 +
 fs/xfs/libxfs/xfs_rmap.h     |    1 
 fs/xfs/scrub/common.c        |   10 -
 fs/xfs/scrub/common.h        |    8 
 fs/xfs/scrub/rmap.c          |  793 ++++++++++++++++++++++++++++++++++++++++++
 7 files changed, 843 insertions(+), 2 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_refcount.c b/fs/xfs/libxfs/xfs_refcount.c
index c6c875d..f63cfdb 100644
--- a/fs/xfs/libxfs/xfs_refcount.c
+++ b/fs/xfs/libxfs/xfs_refcount.c
@@ -88,7 +88,7 @@ xfs_refcount_lookup_ge(
 }
 
 /* Convert on-disk record to in-core format. */
-static inline void
+void
 xfs_refcount_btrec_to_irec(
 	union xfs_btree_rec		*rec,
 	struct xfs_refcount_irec	*irec)
diff --git a/fs/xfs/libxfs/xfs_refcount.h b/fs/xfs/libxfs/xfs_refcount.h
index 78cb142..5973c56 100644
--- a/fs/xfs/libxfs/xfs_refcount.h
+++ b/fs/xfs/libxfs/xfs_refcount.h
@@ -69,5 +69,8 @@ extern int xfs_refcount_recover_cow_leftovers(struct xfs_mount *mp,
 
 extern int xfs_refcount_has_record(struct xfs_btree_cur *cur,
 		xfs_agblock_t bno, xfs_extlen_t len, bool *exists);
+union xfs_btree_rec;
+extern void xfs_refcount_btrec_to_irec(union xfs_btree_rec *rec,
+		struct xfs_refcount_irec *irec);
 
 #endif	/* __XFS_REFCOUNT_H__ */
diff --git a/fs/xfs/libxfs/xfs_rmap.c b/fs/xfs/libxfs/xfs_rmap.c
index e61d816..8531cbc 100644
--- a/fs/xfs/libxfs/xfs_rmap.c
+++ b/fs/xfs/libxfs/xfs_rmap.c
@@ -1977,6 +1977,34 @@ xfs_rmap_map_shared(
 	return error;
 }
 
+/* Insert a raw rmap into the rmapbt. */
+int
+xfs_rmap_map_raw(
+	struct xfs_btree_cur	*cur,
+	struct xfs_rmap_irec	*rmap)
+{
+	struct xfs_owner_info	oinfo;
+
+	oinfo.oi_owner = rmap->rm_owner;
+	oinfo.oi_offset = rmap->rm_offset;
+	oinfo.oi_flags = 0;
+	if (rmap->rm_flags & XFS_RMAP_ATTR_FORK)
+		oinfo.oi_flags |= XFS_OWNER_INFO_ATTR_FORK;
+	if (rmap->rm_flags & XFS_RMAP_BMBT_BLOCK)
+		oinfo.oi_flags |= XFS_OWNER_INFO_BMBT_BLOCK;
+
+	if (rmap->rm_flags || XFS_RMAP_NON_INODE_OWNER(rmap->rm_owner))
+		return xfs_rmap_map(cur, rmap->rm_startblock,
+				rmap->rm_blockcount,
+				rmap->rm_flags & XFS_RMAP_UNWRITTEN,
+				&oinfo);
+
+	return xfs_rmap_map_shared(cur, rmap->rm_startblock,
+			rmap->rm_blockcount,
+			rmap->rm_flags & XFS_RMAP_UNWRITTEN,
+			&oinfo);
+}
+
 struct xfs_rmap_query_range_info {
 	xfs_rmap_query_range_fn	fn;
 	void				*priv;
diff --git a/fs/xfs/libxfs/xfs_rmap.h b/fs/xfs/libxfs/xfs_rmap.h
index 606efe3..eac90d7 100644
--- a/fs/xfs/libxfs/xfs_rmap.h
+++ b/fs/xfs/libxfs/xfs_rmap.h
@@ -225,5 +225,6 @@ int xfs_rmap_record_exists(struct xfs_btree_cur *cur, xfs_fsblock_t bno,
 int xfs_rmap_has_other_keys(struct xfs_btree_cur *cur, xfs_fsblock_t bno,
 		xfs_filblks_t len, struct xfs_owner_info *oinfo,
 		bool *has_rmap);
+int xfs_rmap_map_raw(struct xfs_btree_cur *cur, struct xfs_rmap_irec *rmap);
 
 #endif	/* __XFS_RMAP_H__ */
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 4b77550..d962395 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -642,6 +642,7 @@ xfs_scrub_teardown(
 	int				error)
 {
 	struct xfs_mount		*mp = sc->tp->t_mountp;
+	int				err2;
 
 	xfs_scrub_ag_free(&sc->sa);
 	if (sc->ag_lock.agmask != sc->ag_lock.__agmask)
@@ -652,6 +653,13 @@ xfs_scrub_teardown(
 	else
 		xfs_trans_cancel(sc->tp);
 	sc->tp = NULL;
+
+	if (sc->teardown) {
+		err2 = sc->teardown(sc, ip_in, error);
+		if (!error && err2)
+			error = err2;
+	}
+
 	if (sc->ip != NULL) {
 		xfs_iunlock(sc->ip, XFS_ILOCK_EXCL);
 		xfs_iunlock(sc->ip, XFS_IOLOCK_EXCL);
@@ -769,7 +777,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag_allocbt, xfs_scrub_cntbt, xfs_repair_allocbt, NULL},
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_inobt, xfs_repair_iallocbt, NULL},
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_finobt, xfs_repair_iallocbt, xfs_sb_version_hasfinobt},
-	{xfs_scrub_setup_ag_header, xfs_scrub_rmapbt, NULL, xfs_sb_version_hasrmapbt},
+	{xfs_scrub_setup_ag_header_freeze, xfs_scrub_rmapbt, xfs_repair_rmapbt, xfs_sb_version_hasrmapbt},
 	{xfs_scrub_setup_ag_header, xfs_scrub_refcountbt, NULL, xfs_sb_version_hasreflink},
 	{xfs_scrub_setup_inode_raw, xfs_scrub_inode, NULL, NULL},
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_data, NULL, NULL},
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 641e0c2..c198ef3 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -68,6 +68,9 @@ struct xfs_scrub_context {
 
 	/* State tracking for single-AG operations. */
 	struct xfs_scrub_ag		sa;
+
+	int				(*teardown)(struct xfs_scrub_context *,
+						    struct xfs_inode *, int);
 };
 
 /* Should we end the scrub early? */
@@ -275,6 +278,10 @@ int xfs_scrub_setup_ag_header(struct xfs_scrub_context *sc,
 			      struct xfs_inode *ip,
 			      struct xfs_scrub_metadata *sm,
 			      bool retry_deadlocked);
+int xfs_scrub_setup_ag_header_freeze(struct xfs_scrub_context *sc,
+				     struct xfs_inode *ip,
+				     struct xfs_scrub_metadata *sm,
+				     bool retry_deadlocked);
 int xfs_scrub_setup_ag_allocbt(struct xfs_scrub_context *sc,
 			       struct xfs_inode *ip,
 			       struct xfs_scrub_metadata *sm,
@@ -384,5 +391,6 @@ int xfs_repair_agfl(struct xfs_scrub_context *sc);
 int xfs_repair_agi(struct xfs_scrub_context *sc);
 int xfs_repair_allocbt(struct xfs_scrub_context *sc);
 int xfs_repair_iallocbt(struct xfs_scrub_context *sc);
+int xfs_repair_rmapbt(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/rmap.c b/fs/xfs/scrub/rmap.c
index 9b7ed8b..26b75d5 100644
--- a/fs/xfs/scrub/rmap.c
+++ b/fs/xfs/scrub/rmap.c
@@ -30,13 +30,93 @@
 #include "xfs_trans.h"
 #include "xfs_trace.h"
 #include "xfs_sb.h"
+#include "xfs_inode.h"
+#include "xfs_icache.h"
 #include "xfs_rmap.h"
+#include "xfs_rmap_btree.h"
 #include "xfs_alloc.h"
+#include "xfs_alloc_btree.h"
 #include "xfs_ialloc.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_bmap.h"
+#include "xfs_bmap_btree.h"
 #include "xfs_refcount.h"
+#include "xfs_refcount_btree.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
+/* Unfreeze the FS. */
+STATIC int
+xfs_scrub_teardown_thaw(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	int				error)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+	struct super_block		*sb = mp->m_super;
+	int				err2;
+
+	/* Re-freeze the last level of filesystem. */
+	down_write(&sb->s_umount);
+	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_PAGEFAULT);
+	sb->s_writers.frozen = SB_FREEZE_COMPLETE;
+	up_write(&sb->s_umount);
+	err2 = thaw_super(sb);
+	if (!error && err2)
+		error = err2;
+
+	return error;
+}
+
+/* Set us up with AG headers and btree cursors, and freeze the FS. */
+int
+xfs_scrub_setup_ag_header_freeze(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	struct xfs_mount		*mp = ip->i_mount;
+	struct super_block		*sb = mp->m_super;
+	int				error;
+
+	if (!(sm->sm_flags & XFS_SCRUB_FLAG_REPAIR))
+		return xfs_scrub_setup_ag_header(sc, ip, sm, retry_deadlocked);
+
+	/* Freeze out any further writes or page faults. */
+	error = freeze_super(sb);
+	if (error)
+		return error;
+
+	/* Thaw it to the point that we can make transactions. */
+	down_write(&sb->s_umount);
+	percpu_up_write(sb->s_writers.rw_sem + SB_FREEZE_PAGEFAULT);
+	sb->s_writers.frozen = SB_FREEZE_FS;
+	up_write(&sb->s_umount);
+
+	/* Check the AG number and set up the scrub context. */
+	error = xfs_scrub_setup_ag(sc, ip, sm, retry_deadlocked);
+	if (error)
+		return xfs_scrub_teardown_thaw(sc, ip, error);
+
+	/* Lock all the AG header buffers. */
+	sc->teardown = xfs_scrub_teardown_thaw;
+	xfs_scrub_ag_lock_init(mp, &sc->ag_lock);
+	error = xfs_scrub_ag_lock_all(sc);
+	if (error)
+		return error;
+
+	/* Now grab the headers of the AGF we want. */
+	sc->sa.agno = sm->sm_agno;
+	error = xfs_scrub_ag_read_headers(sc, sm->sm_agno, &sc->sa.agi_bp,
+			&sc->sa.agf_bp, &sc->sa.agfl_bp);
+	if (error)
+		return error;
+
+	/* ...and initialize the btree cursors for xref. */
+	return xfs_scrub_ag_btcur_init(sc, &sc->sa);
+}
+
 /* Reverse-mapping scrubber. */
 
 /* Scrub an rmapbt record. */
@@ -219,3 +299,716 @@ xfs_scrub_rmapbt(
 	return xfs_scrub_btree(sc, sc->sa.rmap_cur, xfs_scrub_rmapbt_helper,
 			&oinfo, NULL);
 }
+
+/* Reverse-mapping repair. */
+
+struct xfs_repair_rmapbt_extent {
+	struct list_head		list;
+	struct xfs_rmap_irec		rmap;
+};
+
+struct xfs_repair_rmapbt {
+	struct list_head		rmaplist;
+	struct list_head		rmap_freelist;
+	struct list_head		bno_freelist;
+	struct xfs_scrub_context	*sc;
+	uint64_t			owner;
+	xfs_extlen_t			btblocks;
+	xfs_agblock_t			next_bno;
+	uint64_t			nr_records;
+};
+
+/* Initialize an rmap. */
+static inline int
+xfs_repair_rmapbt_new_rmap(
+	struct xfs_repair_rmapbt	*rr,
+	xfs_agblock_t			startblock,
+	xfs_extlen_t			blockcount,
+	__uint64_t			owner,
+	__uint64_t			offset,
+	unsigned int			flags)
+{
+	struct xfs_repair_rmapbt_extent	*rre;
+	int				error = 0;
+
+	trace_xfs_repair_rmap_extent_fn(rr->sc->tp->t_mountp, rr->sc->sa.agno,
+			startblock, blockcount, owner, offset, flags);
+
+	if (xfs_scrub_should_terminate(&error))
+		return error;
+
+	rre = kmem_alloc(sizeof(*rre), KM_NOFS);
+	if (!rre)
+		return -ENOMEM;
+	INIT_LIST_HEAD(&rre->list);
+	rre->rmap.rm_startblock = startblock;
+	rre->rmap.rm_blockcount = blockcount;
+	rre->rmap.rm_owner = owner;
+	rre->rmap.rm_offset = offset;
+	rre->rmap.rm_flags = flags;
+	list_add_tail(&rre->list, &rr->rmaplist);
+	rr->nr_records++;
+
+	return 0;
+}
+
+/* Add an AGFL block to the rmap list. */
+STATIC int
+xfs_repair_rmapbt_walk_agfl(
+	struct xfs_scrub_context	*sc,
+	xfs_agblock_t			bno,
+	void				*priv)
+{
+	struct xfs_repair_rmapbt	*rr = priv;
+
+	return xfs_repair_rmapbt_new_rmap(rr, bno, 1, XFS_RMAP_OWN_AG, 0, 0);
+}
+
+/* Add a btree block to the rmap list. */
+STATIC int
+xfs_repair_rmapbt_visit_btblock(
+	struct xfs_btree_cur		*cur,
+	int				level,
+	void				*priv)
+{
+	struct xfs_repair_rmapbt	*rr = priv;
+	struct xfs_buf			*bp;
+	xfs_fsblock_t			fsb;
+
+	xfs_btree_get_block(cur, level, &bp);
+	if (!bp)
+		return 0;
+
+	rr->btblocks++;
+	fsb = XFS_DADDR_TO_FSB(cur->bc_mp, bp->b_bn);
+	return xfs_repair_rmapbt_new_rmap(rr, XFS_FSB_TO_AGBNO(cur->bc_mp, fsb),
+			1, rr->owner, 0, 0);
+}
+
+/* Record inode btree rmaps. */
+STATIC int
+xfs_repair_rmapbt_inodes(
+	struct xfs_btree_cur		*cur,
+	union xfs_btree_rec		*rec,
+	void				*priv)
+{
+	struct xfs_inobt_rec_incore	irec;
+	struct xfs_repair_rmapbt	*rr = priv;
+	struct xfs_mount		*mp = cur->bc_mp;
+	struct xfs_buf			*bp;
+	xfs_fsblock_t			fsb;
+	xfs_agino_t			agino;
+	xfs_agino_t			iperhole;
+	unsigned int			i;
+	int				error;
+
+	/* Record the inobt blocks */
+	for (i = 0; i < cur->bc_nlevels && cur->bc_ptrs[i] == 1; i++) {
+		xfs_btree_get_block(cur, i, &bp);
+		if (!bp)
+			continue;
+		fsb = XFS_DADDR_TO_FSB(mp, bp->b_bn);
+		error = xfs_repair_rmapbt_new_rmap(rr,
+				XFS_FSB_TO_AGBNO(mp, fsb), 1,
+				XFS_RMAP_OWN_INOBT, 0, 0);
+		if (error)
+			return error;
+	}
+
+	xfs_inobt_btrec_to_irec(mp, rec, &irec);
+
+	/* Record a non-sparse inode chunk. */
+	if (irec.ir_holemask == XFS_INOBT_HOLEMASK_FULL)
+		return xfs_repair_rmapbt_new_rmap(rr,
+				XFS_AGINO_TO_AGBNO(mp, irec.ir_startino),
+				XFS_INODES_PER_CHUNK / mp->m_sb.sb_inopblock,
+				XFS_RMAP_OWN_INODES, 0, 0);
+
+	/* Iterate each chunk. */
+	iperhole = max_t(xfs_agino_t, mp->m_sb.sb_inopblock,
+			XFS_INODES_PER_HOLEMASK_BIT);
+	for (i = 0, agino = irec.ir_startino;
+	     i < XFS_INOBT_HOLEMASK_BITS;
+	     i += iperhole / XFS_INODES_PER_HOLEMASK_BIT, agino += iperhole) {
+		/* Skip holes. */
+		if (irec.ir_holemask & (1 << i))
+			continue;
+
+		/* Record the inode chunk otherwise. */
+		error = xfs_repair_rmapbt_new_rmap(rr,
+				XFS_AGINO_TO_AGBNO(mp, agino),
+				iperhole / mp->m_sb.sb_inopblock,
+				XFS_RMAP_OWN_INODES, 0, 0);
+		if (error)
+			return error;
+	}
+
+	return 0;
+}
+
+/* Record a CoW staging extent. */
+STATIC int
+xfs_repair_rmapbt_refcount(
+	struct xfs_btree_cur		*cur,
+	union xfs_btree_rec		*rec,
+	void				*priv)
+{
+	struct xfs_repair_rmapbt	*rr = priv;
+	struct xfs_refcount_irec	refc;
+
+	xfs_refcount_btrec_to_irec(rec, &refc);
+	if (refc.rc_refcount != 1)
+		return -EFSCORRUPTED;
+
+	return xfs_repair_rmapbt_new_rmap(rr,
+			refc.rc_startblock - XFS_REFC_COW_START,
+			refc.rc_blockcount, XFS_RMAP_OWN_COW, 0, 0);
+}
+
+/* Add a bmbt block to the rmap list. */
+STATIC int
+xfs_repair_rmapbt_visit_bmbt(
+	struct xfs_btree_cur		*cur,
+	int				level,
+	void				*priv)
+{
+	struct xfs_repair_rmapbt	*rr = priv;
+	struct xfs_buf			*bp;
+	xfs_fsblock_t			fsb;
+	unsigned int			flags = XFS_RMAP_BMBT_BLOCK;
+
+	xfs_btree_get_block(cur, level, &bp);
+	if (!bp)
+		return 0;
+
+	fsb = XFS_DADDR_TO_FSB(cur->bc_mp, bp->b_bn);
+	if (XFS_FSB_TO_AGNO(cur->bc_mp, fsb) != rr->sc->sa.agno)
+		return 0;
+
+	if (cur->bc_private.b.whichfork == XFS_ATTR_FORK)
+		flags |= XFS_RMAP_ATTR_FORK;
+	return xfs_repair_rmapbt_new_rmap(rr,
+			XFS_FSB_TO_AGBNO(cur->bc_mp, fsb), 1,
+			cur->bc_private.b.ip->i_ino, 0, flags);
+}
+
+/* Determine rmap flags from fork and bmbt state. */
+static inline unsigned int
+xfs_repair_rmapbt_bmap_flags(
+	int			whichfork,
+	xfs_exntst_t		state)
+{
+	return  (whichfork == XFS_ATTR_FORK ? XFS_RMAP_ATTR_FORK : 0) |
+		(state == XFS_EXT_UNWRITTEN ? XFS_RMAP_UNWRITTEN : 0);
+}
+
+/* Find all the extents from a given AG in an inode fork. */
+STATIC int
+xfs_repair_rmapbt_scan_ifork(
+	struct xfs_repair_rmapbt	*rr,
+	struct xfs_inode		*ip,
+	int				whichfork)
+{
+	struct xfs_bmbt_irec		rec;
+	struct xfs_mount		*mp = rr->sc->tp->t_mountp;
+	struct xfs_btree_cur		*cur = NULL;
+	xfs_fileoff_t			off;
+	xfs_fileoff_t			endoff;
+	unsigned int			bflags;
+	unsigned int			rflags;
+	int				nmaps;
+	int				fmt;
+	int				error;
+
+	/* Do we even have data mapping extents? */
+	fmt = XFS_IFORK_FORMAT(ip, whichfork);
+	switch (fmt) {
+	case XFS_DINODE_FMT_BTREE:
+	case XFS_DINODE_FMT_EXTENTS:
+		break;
+	default:
+		return 0;
+	}
+	if (!XFS_IFORK_PTR(ip, whichfork))
+		return 0;
+
+	/* Find all the BMBT blocks in the AG. */
+	if (fmt == XFS_DINODE_FMT_BTREE) {
+		cur = xfs_bmbt_init_cursor(mp, rr->sc->tp, ip, whichfork);
+		error = xfs_btree_visit_blocks(cur,
+				xfs_repair_rmapbt_visit_bmbt, rr);
+		if (error)
+			goto out;
+		xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+		cur = NULL;
+	}
+
+	/* We're done if this is an rt inode's data fork. */
+	if (whichfork == XFS_DATA_FORK && XFS_IS_REALTIME_INODE(ip))
+		return 0;
+
+	/* Find the offset of the last extent in the mapping. */
+	error = xfs_bmap_last_offset(ip, &endoff, whichfork);
+	if (error)
+		goto out;
+
+	/* Find all the extents in the AG. */
+	bflags = whichfork == XFS_ATTR_FORK ? XFS_BMAPI_ATTRFORK : 0;
+	off = 0;
+	while (true) {
+		nmaps = 1;
+		error = xfs_bmapi_read(ip, off, endoff - off, &rec,
+				&nmaps, bflags);
+		if (error || nmaps == 0)
+			break;
+		/* Stash non-hole extent. */
+		if (rec.br_startblock != HOLESTARTBLOCK &&
+		    rec.br_startblock != DELAYSTARTBLOCK &&
+		    XFS_FSB_TO_AGNO(mp, rec.br_startblock) == rr->sc->sa.agno) {
+			rflags = xfs_repair_rmapbt_bmap_flags(whichfork,
+					rec.br_state);
+			error = xfs_repair_rmapbt_new_rmap(rr,
+					XFS_FSB_TO_AGBNO(mp, rec.br_startblock),
+					rec.br_blockcount, ip->i_ino,
+					rec.br_startoff, rflags);
+			if (error)
+				goto out;
+		}
+
+		off += rec.br_blockcount;
+	}
+out:
+	if (cur)
+		xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
+	return error;
+}
+
+/* Iterate all the inodes in an AG group. */
+STATIC int
+xfs_repair_rmapbt_scan_inobt(
+	struct xfs_btree_cur		*cur,
+	union xfs_btree_rec		*rec,
+	void				*priv)
+{
+	struct xfs_inobt_rec_incore	irec;
+	struct xfs_mount		*mp = cur->bc_mp;
+	struct xfs_inode		*ip = NULL;
+	xfs_ino_t			ino;
+	xfs_agino_t			agino;
+	int				chunkidx;
+	int				error;
+
+	xfs_inobt_btrec_to_irec(mp, rec, &irec);
+
+	for (chunkidx = 0, agino = irec.ir_startino;
+	     chunkidx < XFS_INODES_PER_CHUNK;
+	     chunkidx++, agino++) {
+		/* Skip if this inode is free */
+		if (XFS_INOBT_MASK(chunkidx) & irec.ir_free)
+			continue;
+		ino = XFS_AGINO_TO_INO(mp, cur->bc_private.a.agno, agino);
+		error = xfs_iget(mp, cur->bc_tp, ino, 0, XFS_ILOCK_EXCL, &ip);
+		if (error)
+			break;
+
+		/* Check the data fork. */
+		error = xfs_repair_rmapbt_scan_ifork(priv, ip, XFS_DATA_FORK);
+		if (error)
+			break;
+
+		/* Check the attr fork. */
+		error = xfs_repair_rmapbt_scan_ifork(priv, ip, XFS_ATTR_FORK);
+		if (error)
+			break;
+
+		xfs_iunlock(ip, XFS_ILOCK_EXCL);
+		IRELE(ip);
+		ip = NULL;
+	}
+
+	if (ip) {
+		xfs_iunlock(ip, XFS_ILOCK_EXCL);
+		IRELE(ip);
+	}
+	return error;
+}
+
+/* Record extents that aren't in use from gaps in the rmap records. */
+STATIC int
+xfs_repair_rmapbt_record_rmap_freesp(
+	struct xfs_btree_cur		*cur,
+	struct xfs_rmap_irec		*rec,
+	void				*priv)
+{
+	struct xfs_repair_rmapbt	*rr = priv;
+	xfs_fsblock_t			fsb;
+	int				error;
+
+	/* Record the free space we find. */
+	if (rec->rm_startblock > rr->next_bno) {
+		fsb = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_private.a.agno,
+				rr->next_bno);
+		error = xfs_repair_collect_btree_extent(cur->bc_mp,
+				&rr->rmap_freelist, fsb,
+				rec->rm_startblock - rr->next_bno);
+		if (error)
+			return error;
+	}
+	rr->next_bno = max_t(xfs_agblock_t, rr->next_bno,
+			rec->rm_startblock + rec->rm_blockcount);
+	return 0;
+}
+
+/* Record extents that aren't in use from the bnobt records. */
+STATIC int
+xfs_repair_rmapbt_record_bno_freesp(
+	struct xfs_btree_cur		*cur,
+	struct xfs_alloc_rec_incore	*rec,
+	void				*priv)
+{
+	struct xfs_repair_rmapbt	*rr = priv;
+	xfs_fsblock_t			fsb;
+
+	/* Record the free space we find. */
+	fsb = XFS_AGB_TO_FSB(cur->bc_mp, cur->bc_private.a.agno,
+			rec->ar_startblock);
+	return xfs_repair_collect_btree_extent(cur->bc_mp, &rr->bno_freelist,
+			fsb, rec->ar_blockcount);
+}
+
+/* Compare two rmapbt extents. */
+static int
+xfs_repair_rmapbt_extent_cmp(
+	void				*priv,
+	struct list_head		*a,
+	struct list_head		*b)
+{
+	struct xfs_repair_rmapbt_extent	*ap;
+	struct xfs_repair_rmapbt_extent	*bp;
+	__u64				oa;
+	__u64				ob;
+
+	ap = container_of(a, struct xfs_repair_rmapbt_extent, list);
+	bp = container_of(b, struct xfs_repair_rmapbt_extent, list);
+	oa = xfs_rmap_irec_offset_pack(&ap->rmap);
+	ob = xfs_rmap_irec_offset_pack(&bp->rmap);
+
+	if (ap->rmap.rm_startblock > bp->rmap.rm_startblock)
+		return 1;
+	else if (ap->rmap.rm_startblock < bp->rmap.rm_startblock)
+		return -1;
+	else if (ap->rmap.rm_owner > bp->rmap.rm_owner)
+		return 1;
+	else if (ap->rmap.rm_owner < bp->rmap.rm_owner)
+		return -1;
+	else if (oa > ob)
+		return 1;
+	else if (oa < ob)
+		return -1;
+	return 0;
+}
+
+#define RMAP(type, startblock, blockcount) xfs_repair_rmapbt_new_rmap( \
+		&rr, (startblock), (blockcount), \
+		XFS_RMAP_OWN_##type, 0, 0)
+/* Repair the rmap btree for some AG. */
+int
+xfs_repair_rmapbt(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_repair_rmapbt	rr;
+	struct xfs_owner_info		oinfo;
+	struct xfs_repair_rmapbt_extent	*rre;
+	struct xfs_repair_rmapbt_extent	*n;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_btree_cur		*cur = NULL;
+	struct xfs_buf			*bp = NULL;
+	struct xfs_agf			*agf;
+	struct xfs_agi			*agi;
+	struct xfs_perag		*pag;
+	xfs_fsblock_t			btfsb;
+	xfs_agnumber_t			ag;
+	xfs_agblock_t			agend;
+	xfs_extlen_t			freesp_btblocks;
+	int				error;
+
+	INIT_LIST_HEAD(&rr.rmaplist);
+	INIT_LIST_HEAD(&rr.rmap_freelist);
+	INIT_LIST_HEAD(&rr.bno_freelist);
+	rr.sc = sc;
+	rr.nr_records = 0;
+
+	/* Collect rmaps for all AG headers. */
+	error = RMAP(FS, XFS_SB_BLOCK(mp), 1);
+	if (error)
+		goto out;
+	rre = list_last_entry(&rr.rmaplist, struct xfs_repair_rmapbt_extent,
+			list);
+
+	if (rre->rmap.rm_startblock != XFS_AGF_BLOCK(mp)) {
+		error = RMAP(FS, XFS_AGF_BLOCK(mp), 1);
+		if (error)
+			goto out;
+		rre = list_last_entry(&rr.rmaplist,
+				struct xfs_repair_rmapbt_extent, list);
+	}
+
+	if (rre->rmap.rm_startblock != XFS_AGI_BLOCK(mp)) {
+		error = RMAP(FS, XFS_AGI_BLOCK(mp), 1);
+		if (error)
+			goto out;
+		rre = list_last_entry(&rr.rmaplist,
+				struct xfs_repair_rmapbt_extent, list);
+	}
+
+	if (rre->rmap.rm_startblock != XFS_AGFL_BLOCK(mp)) {
+		error = RMAP(FS, XFS_AGFL_BLOCK(mp), 1);
+		if (error)
+			goto out;
+	}
+
+	error = xfs_scrub_walk_agfl(sc, xfs_repair_rmapbt_walk_agfl, &rr);
+	if (error)
+		goto out;
+
+	/* Collect rmap for the log if it's in this AG. */
+	if (mp->m_sb.sb_logstart &&
+	    XFS_FSB_TO_AGNO(mp, mp->m_sb.sb_logstart) == sc->sa.agno) {
+		error = RMAP(LOG, XFS_FSB_TO_AGBNO(mp, mp->m_sb.sb_logstart),
+				mp->m_sb.sb_logblocks);
+		if (error)
+			goto out;
+	}
+
+	/* Collect rmaps for the free space btrees. */
+	rr.owner = XFS_RMAP_OWN_AG;
+	rr.btblocks = 0;
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, sc->sa.agf_bp, sc->sa.agno,
+			XFS_BTNUM_BNO);
+	error = xfs_btree_visit_blocks(cur, xfs_repair_rmapbt_visit_btblock,
+			&rr);
+	if (error)
+		goto out;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	cur = NULL;
+
+	/* Collect rmaps for the cntbt. */
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, sc->sa.agf_bp, sc->sa.agno,
+			XFS_BTNUM_CNT);
+	error = xfs_btree_visit_blocks(cur, xfs_repair_rmapbt_visit_btblock,
+			&rr);
+	if (error)
+		goto out;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	cur = NULL;
+	freesp_btblocks = rr.btblocks;
+
+	/* Collect rmaps for the inode btree. */
+	cur = xfs_inobt_init_cursor(mp, sc->tp, sc->sa.agi_bp, sc->sa.agno,
+			XFS_BTNUM_INO);
+	error = xfs_btree_query_all(cur, xfs_repair_rmapbt_inodes, &rr);
+	if (error)
+		goto out;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+
+	/* If there are no inodes, we have to include the inobt root. */
+	agi = XFS_BUF_TO_AGI(sc->sa.agi_bp);
+	if (agi->agi_count == cpu_to_be32(0)) {
+		error = xfs_repair_rmapbt_new_rmap(&rr,
+				be32_to_cpu(agi->agi_root), 1,
+				XFS_RMAP_OWN_INOBT, 0, 0);
+		if (error)
+			goto out;
+	}
+
+	/* Collect rmaps for the free inode btree. */
+	if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
+		rr.owner = XFS_RMAP_OWN_INOBT;
+		cur = xfs_inobt_init_cursor(mp, sc->tp, sc->sa.agi_bp,
+				sc->sa.agno, XFS_BTNUM_FINO);
+		error = xfs_btree_visit_blocks(cur,
+				xfs_repair_rmapbt_visit_btblock, &rr);
+		if (error)
+			goto out;
+		xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+		cur = NULL;
+	}
+
+	/* Collect rmaps for the refcount btree. */
+	if (xfs_sb_version_hasreflink(&mp->m_sb)) {
+		union xfs_btree_irec		low;
+		union xfs_btree_irec		high;
+
+		rr.owner = XFS_RMAP_OWN_REFC;
+		cur = xfs_refcountbt_init_cursor(mp, sc->tp, sc->sa.agf_bp,
+				sc->sa.agno, NULL);
+		error = xfs_btree_visit_blocks(cur,
+				xfs_repair_rmapbt_visit_btblock, &rr);
+		if (error)
+			goto out;
+
+		/* Collect rmaps for CoW staging extents. */
+		memset(&low, 0, sizeof(low));
+		low.rc.rc_startblock = XFS_REFC_COW_START;
+		memset(&high, 0xFF, sizeof(high));
+		error = xfs_btree_query_range(cur, &low, &high,
+				xfs_repair_rmapbt_refcount, &rr);
+		if (error)
+			goto out;
+		xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+		cur = NULL;
+	}
+
+	/* Iterate all AGs for inodes. */
+	for (ag = 0; ag < mp->m_sb.sb_agcount; ag++) {
+		ASSERT(xfs_scrub_ag_can_lock(sc, ag));
+		error = xfs_ialloc_read_agi(mp, sc->tp, ag, &bp);
+		if (error)
+			goto out;
+		cur = xfs_inobt_init_cursor(mp, sc->tp, bp, ag, XFS_BTNUM_INO);
+		error = xfs_btree_query_all(cur, xfs_repair_rmapbt_scan_inobt,
+				&rr);
+		if (error)
+			goto out;
+		xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+		cur = NULL;
+		xfs_trans_brelse(sc->tp, bp);
+		bp = NULL;
+	}
+
+	/* Do we actually have enough space to do this? */
+	pag = xfs_perag_get(mp, sc->sa.agno);
+	if (!xfs_repair_ag_has_space(pag,
+			xfs_rmapbt_calc_size(mp, rr.nr_records),
+			XFS_AG_RESV_AGFL)) {
+		xfs_perag_put(pag);
+		error = -ENOSPC;
+		goto out;
+	}
+
+	/* Initialize a new rmapbt root. */
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_UNKNOWN);
+	agf = XFS_BUF_TO_AGF(sc->sa.agf_bp);
+	error = xfs_repair_alloc_ag_block(sc, &oinfo, &btfsb, XFS_AG_RESV_AGFL);
+	if (error) {
+		xfs_perag_put(pag);
+		goto out;
+	}
+	error = xfs_repair_init_btblock(sc, btfsb, &bp, XFS_RMAP_CRC_MAGIC,
+			&xfs_rmapbt_buf_ops);
+	if (error) {
+		xfs_perag_put(pag);
+		goto out;
+	}
+	agf->agf_roots[XFS_BTNUM_RMAPi] = cpu_to_be32(XFS_FSB_TO_AGBNO(mp,
+			btfsb));
+	agf->agf_levels[XFS_BTNUM_RMAPi] = cpu_to_be32(1);
+	agf->agf_rmap_blocks = cpu_to_be32(1);
+
+	/* Reset the perag info. */
+	pag->pagf_btreeblks = freesp_btblocks - 2;
+	pag->pagf_levels[XFS_BTNUM_RMAPi] =
+			be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAPi]);
+
+	/* Now reset the AGF counters. */
+	agf->agf_btreeblks = cpu_to_be32(pag->pagf_btreeblks);
+	xfs_perag_put(pag);
+	xfs_alloc_log_agf(sc->tp, sc->sa.agf_bp, XFS_AGF_ROOTS |
+			XFS_AGF_LEVELS | XFS_AGF_RMAP_BLOCKS |
+			XFS_AGF_BTREEBLKS);
+	bp = NULL;
+	error = xfs_repair_roll_ag_trans(sc);
+	if (error)
+		goto out;
+
+	/* Insert all the metadata rmaps. */
+	list_sort(NULL, &rr.rmaplist, xfs_repair_rmapbt_extent_cmp);
+	list_for_each_entry_safe(rre, n, &rr.rmaplist, list) {
+		/*
+		 * Ensure the freelist is full, but don't let it shrink.
+		 * The rmapbt isn't fully set up yet, which means that
+		 * the current AGFL blocks might not be reflected in the
+		 * rmapbt, which is a problem if we want to unmap blocks
+		 * from the AGFL.
+		 */
+		error = xfs_repair_fix_freelist(sc, false);
+		if (error)
+			goto out;
+
+		/* Add the rmap. */
+		cur = xfs_rmapbt_init_cursor(mp, sc->tp, sc->sa.agf_bp,
+				sc->sa.agno);
+		error = xfs_rmap_map_raw(cur, &rre->rmap);
+		if (error)
+			goto out;
+		xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+		cur = NULL;
+
+		error = xfs_repair_roll_ag_trans(sc);
+		if (error)
+			goto out;
+
+		list_del(&rre->list);
+		kmem_free(rre);
+	}
+
+	/* Compute free space from the new rmapbt. */
+	rr.next_bno = 0;
+	cur = xfs_rmapbt_init_cursor(mp, sc->tp, sc->sa.agf_bp, sc->sa.agno);
+	error = xfs_rmap_query_all(cur, xfs_repair_rmapbt_record_rmap_freesp,
+			&rr);
+	if (error)
+		goto out;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	cur = NULL;
+
+	/* Insert a record for space between the last rmap and EOAG. */
+	agf = XFS_BUF_TO_AGF(sc->sa.agf_bp);
+	agend = be32_to_cpu(agf->agf_length);
+	if (rr.next_bno < agend) {
+		btfsb = XFS_AGB_TO_FSB(mp, sc->sa.agno, rr.next_bno);
+		error = xfs_repair_collect_btree_extent(mp, &rr.rmap_freelist,
+				btfsb, agend - rr.next_bno);
+		if (error)
+			goto out;
+	}
+
+	/* Compute free space from the existing bnobt. */
+	cur = xfs_allocbt_init_cursor(mp, sc->tp, sc->sa.agf_bp, sc->sa.agno,
+			XFS_BTNUM_BNO);
+	error = xfs_alloc_query_all(cur, xfs_repair_rmapbt_record_bno_freesp,
+			&rr);
+	if (error)
+		goto out;
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	cur = NULL;
+
+	/*
+	 * Free the "free" blocks that the new rmapbt knows about but
+	 * the old bnobt doesn't.  These are the old rmapbt blocks.
+	 */
+	error = xfs_repair_subtract_extents(mp, &rr.rmap_freelist,
+			&rr.bno_freelist);
+	if (error)
+		goto out;
+	xfs_repair_cancel_btree_extents(sc, &rr.bno_freelist);
+	error = xfs_repair_reap_btree_extents(sc, &rr.rmap_freelist, &oinfo,
+			XFS_AG_RESV_AGFL);
+	if (error)
+		goto out;
+
+	return 0;
+out:
+	if (cur)
+		xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
+	if (bp)
+		xfs_trans_brelse(sc->tp, bp);
+	xfs_repair_cancel_btree_extents(sc, &rr.bno_freelist);
+	xfs_repair_cancel_btree_extents(sc, &rr.rmap_freelist);
+	list_for_each_entry_safe(rre, n, &rr.rmaplist, list) {
+		list_del(&rre->list);
+		kmem_free(rre);
+	}
+	return error;
+}
+#undef RMAP

[PATCH 51/55] xfs: repair refcount btrees

[Darrick J. Wong]

Reconstruct the refcount data from the rmap btree.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_btree.c    |   21 ++
 fs/xfs/libxfs/xfs_btree.h    |    1 
 fs/xfs/libxfs/xfs_refcount.c |   19 ++
 fs/xfs/libxfs/xfs_refcount.h |    4 
 fs/xfs/scrub/common.c        |    2 
 fs/xfs/scrub/common.h        |    1 
 fs/xfs/scrub/refcount.c      |  467 ++++++++++++++++++++++++++++++++++++++++++
 7 files changed, 513 insertions(+), 2 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c
index 1f03736..b81e5e2 100644
--- a/fs/xfs/libxfs/xfs_btree.c
+++ b/fs/xfs/libxfs/xfs_btree.c
@@ -4913,3 +4913,24 @@ xfs_btree_has_record(
 
 	return 0;
 }
+
+/* Are there more records in this btree? */
+bool
+xfs_btree_has_more_records(
+	struct xfs_btree_cur	*cur)
+{
+	struct xfs_btree_block	*block;
+	struct xfs_buf		*bp;
+
+	block = xfs_btree_get_block(cur, 0, &bp);
+
+	/* There are still records in this block. */
+	if (cur->bc_ptrs[0] < xfs_btree_get_numrecs(block))
+		return true;
+
+	/* There are more record blocks. */
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
+		return block->bb_u.l.bb_rightsib != cpu_to_be64(NULLFSBLOCK);
+	else
+		return block->bb_u.s.bb_rightsib != cpu_to_be32(NULLAGBLOCK);
+}
diff --git a/fs/xfs/libxfs/xfs_btree.h b/fs/xfs/libxfs/xfs_btree.h
index 6092e2b..a033dff 100644
--- a/fs/xfs/libxfs/xfs_btree.h
+++ b/fs/xfs/libxfs/xfs_btree.h
@@ -516,5 +516,6 @@ struct xfs_btree_block *xfs_btree_get_block(struct xfs_btree_cur *cur,
 		int level, struct xfs_buf **bpp);
 int xfs_btree_has_record(struct xfs_btree_cur *cur, union xfs_btree_irec *low,
 		union xfs_btree_irec *high, bool *exists);
+bool xfs_btree_has_more_records(struct xfs_btree_cur *);
 
 #endif	/* __XFS_BTREE_H__ */
diff --git a/fs/xfs/libxfs/xfs_refcount.c b/fs/xfs/libxfs/xfs_refcount.c
index f63cfdb..1c47671 100644
--- a/fs/xfs/libxfs/xfs_refcount.c
+++ b/fs/xfs/libxfs/xfs_refcount.c
@@ -87,6 +87,23 @@ xfs_refcount_lookup_ge(
 	return xfs_btree_lookup(cur, XFS_LOOKUP_GE, stat);
 }
 
+/*
+ * Look up the first record equal to [bno, len] in the btree
+ * given by cur.
+ */
+int
+xfs_refcount_lookup_eq(
+	struct xfs_btree_cur	*cur,
+	xfs_agblock_t		bno,
+	int			*stat)
+{
+	trace_xfs_refcount_lookup(cur->bc_mp, cur->bc_private.a.agno, bno,
+			XFS_LOOKUP_LE);
+	cur->bc_rec.rc.rc_startblock = bno;
+	cur->bc_rec.rc.rc_blockcount = 0;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_EQ, stat);
+}
+
 /* Convert on-disk record to in-core format. */
 void
 xfs_refcount_btrec_to_irec(
@@ -148,7 +165,7 @@ xfs_refcount_update(
  * by [bno, len, refcount].
  * This either works (return 0) or gets an EFSCORRUPTED error.
  */
-STATIC int
+int
 xfs_refcount_insert(
 	struct xfs_btree_cur		*cur,
 	struct xfs_refcount_irec	*irec,
diff --git a/fs/xfs/libxfs/xfs_refcount.h b/fs/xfs/libxfs/xfs_refcount.h
index 5973c56..cad61de 100644
--- a/fs/xfs/libxfs/xfs_refcount.h
+++ b/fs/xfs/libxfs/xfs_refcount.h
@@ -24,6 +24,8 @@ extern int xfs_refcount_lookup_le(struct xfs_btree_cur *cur,
 		xfs_agblock_t bno, int *stat);
 extern int xfs_refcount_lookup_ge(struct xfs_btree_cur *cur,
 		xfs_agblock_t bno, int *stat);
+extern int xfs_refcount_lookup_eq(struct xfs_btree_cur *cur,
+		xfs_agblock_t bno, int *stat);
 extern int xfs_refcount_get_rec(struct xfs_btree_cur *cur,
 		struct xfs_refcount_irec *irec, int *stat);
 
@@ -72,5 +74,7 @@ extern int xfs_refcount_has_record(struct xfs_btree_cur *cur,
 union xfs_btree_rec;
 extern void xfs_refcount_btrec_to_irec(union xfs_btree_rec *rec,
 		struct xfs_refcount_irec *irec);
+extern int xfs_refcount_insert(struct xfs_btree_cur *cur,
+		struct xfs_refcount_irec *irec, int *stat);
 
 #endif	/* __XFS_REFCOUNT_H__ */
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index d962395..5c61985 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -778,7 +778,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_inobt, xfs_repair_iallocbt, NULL},
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_finobt, xfs_repair_iallocbt, xfs_sb_version_hasfinobt},
 	{xfs_scrub_setup_ag_header_freeze, xfs_scrub_rmapbt, xfs_repair_rmapbt, xfs_sb_version_hasrmapbt},
-	{xfs_scrub_setup_ag_header, xfs_scrub_refcountbt, NULL, xfs_sb_version_hasreflink},
+	{xfs_scrub_setup_ag_header, xfs_scrub_refcountbt, xfs_repair_refcountbt, xfs_sb_version_hasreflink},
 	{xfs_scrub_setup_inode_raw, xfs_scrub_inode, NULL, NULL},
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_data, NULL, NULL},
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_attr, NULL, NULL},
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index c198ef3..896d6c0 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -392,5 +392,6 @@ int xfs_repair_agi(struct xfs_scrub_context *sc);
 int xfs_repair_allocbt(struct xfs_scrub_context *sc);
 int xfs_repair_iallocbt(struct xfs_scrub_context *sc);
 int xfs_repair_rmapbt(struct xfs_scrub_context *sc);
+int xfs_repair_refcountbt(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/refcount.c b/fs/xfs/scrub/refcount.c
index 2194be9..b5b17d8 100644
--- a/fs/xfs/scrub/refcount.c
+++ b/fs/xfs/scrub/refcount.c
@@ -30,9 +30,14 @@
 #include "xfs_trans.h"
 #include "xfs_trace.h"
 #include "xfs_sb.h"
+#include "xfs_itable.h"
 #include "xfs_rmap.h"
+#include "xfs_rmap_btree.h"
+#include "xfs_refcount.h"
+#include "xfs_refcount_btree.h"
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
+#include "xfs_error.h"
 #include "scrub/common.h"
 #include "scrub/btree.h"
 
@@ -312,3 +317,465 @@ xfs_scrub_refcountbt(
 	return xfs_scrub_btree(sc, sc->sa.refc_cur, xfs_scrub_refcountbt_helper,
 			&oinfo, NULL);
 }
+
+/*
+ * Rebuilding the Reference Count Btree
+ *
+ * This algorithm is "borrowed" from xfs_repair.  Imagine the rmap
+ * entries as rectangles representing extents of physical blocks, and
+ * that the rectangles can be laid down to allow them to overlap each
+ * other; then we know that we must emit a refcnt btree entry wherever
+ * the amount of overlap changes, i.e. the emission stimulus is
+ * level-triggered:
+ *
+ *                 -    ---
+ *       --      ----- ----   ---        ------
+ * --   ----     ----------- ----     ---------
+ * -------------------------------- -----------
+ * ^ ^  ^^ ^^    ^ ^^ ^^^  ^^^^  ^ ^^ ^  ^     ^
+ * 2 1  23 21    3 43 234  2123  1 01 2  3     0
+ *
+ * For our purposes, a rmap is a tuple (startblock, len, fileoff, owner).
+ *
+ * Note that in the actual refcnt btree we don't store the refcount < 2
+ * cases because the bnobt tells us which blocks are free; single-use
+ * blocks aren't recorded in the bnobt or the refcntbt.  If the rmapbt
+ * supports storing multiple entries covering a given block we could
+ * theoretically dispense with the refcntbt and simply count rmaps, but
+ * that's inefficient in the (hot) write path, so we'll take the cost of
+ * the extra tree to save time.  Also there's no guarantee that rmap
+ * will be enabled.
+ *
+ * Given an array of rmaps sorted by physical block number, a starting
+ * physical block (sp), a bag to hold rmaps that cover sp, and the next
+ * physical block where the level changes (np), we can reconstruct the
+ * refcount btree as follows:
+ *
+ * While there are still unprocessed rmaps in the array,
+ *  - Set sp to the physical block (pblk) of the next unprocessed rmap.
+ *  - Add to the bag all rmaps in the array where startblock == sp.
+ *  - Set np to the physical block where the bag size will change.  This
+ *    is the minimum of (the pblk of the next unprocessed rmap) and
+ *    (startblock + len of each rmap in the bag).
+ *  - Record the bag size as old_bag_size.
+ *
+ *  - While the bag isn't empty,
+ *     - Remove from the bag all rmaps where startblock + len == np.
+ *     - Add to the bag all rmaps in the array where startblock == np.
+ *     - If the bag size isn't old_bag_size, store the refcount entry
+ *       (sp, np - sp, bag_size) in the refcnt btree.
+ *     - If the bag is empty, break out of the inner loop.
+ *     - Set old_bag_size to the bag size
+ *     - Set sp = np.
+ *     - Set np to the physical block where the bag size will change.
+ *       This is the minimum of (the pblk of the next unprocessed rmap)
+ *       and (startblock + len of each rmap in the bag).
+ *
+ * Like all the other repairers, we make a list of all the refcount
+ * records we need, then reinitialize the refcount btree root and
+ * insert all the records.
+ */
+
+struct xfs_repair_refc_rmap {
+	struct list_head		list;
+	struct xfs_rmap_irec		rmap;
+};
+
+struct xfs_repair_refc_extent {
+	struct list_head		list;
+	struct xfs_refcount_irec	refc;
+};
+
+struct xfs_repair_refc {
+	struct list_head		rmap_bag;  /* rmaps we're tracking */
+	struct list_head		rmap_idle; /* idle rmaps */
+	struct list_head		extlist;   /* refcount extents */
+	struct list_head		btlist;    /* old refcountbt blocks */
+	xfs_extlen_t			btblocks;  /* # of refcountbt blocks */
+};
+
+/* Grab the next record from the rmapbt. */
+STATIC int
+xfs_repair_refcountbt_next_rmap(
+	struct xfs_btree_cur		*cur,
+	struct xfs_repair_refc		*rr,
+	struct xfs_rmap_irec		*rec,
+	bool				*have_rec)
+{
+	struct xfs_rmap_irec		rmap;
+	struct xfs_mount		*mp = cur->bc_mp;
+	struct xfs_repair_refc_extent	*rre;
+	xfs_fsblock_t			fsbno;
+	int				have_gt;
+	int				error = 0;
+
+	*have_rec = false;
+	/*
+	 * Loop through the remaining rmaps.  Remember CoW staging
+	 * extents and the refcountbt blocks from the old tree for later
+	 * disposal.  We can only share written data fork extents, so
+	 * keep looping until we find an rmap for one.
+	 */
+	do {
+		if (xfs_scrub_should_terminate(&error))
+			goto out_error;
+
+		error = xfs_btree_increment(cur, 0, &have_gt);
+		if (error)
+			goto out_error;
+		if (!have_gt)
+			return 0;
+
+		error = xfs_rmap_get_rec(cur, &rmap, &have_gt);
+		if (error)
+			goto out_error;
+		XFS_WANT_CORRUPTED_GOTO(mp, have_gt == 1, out_error);
+
+		if (rmap.rm_owner == XFS_RMAP_OWN_COW) {
+			/* Pass CoW staging extents right through. */
+			rre = kmem_alloc(sizeof(*rre), KM_NOFS);
+			if (!rre)
+				goto out_error;
+
+			INIT_LIST_HEAD(&rre->list);
+			rre->refc.rc_startblock = rmap.rm_startblock +
+					XFS_REFC_COW_START;
+			rre->refc.rc_blockcount = rmap.rm_blockcount;
+			rre->refc.rc_refcount = 1;
+			list_add_tail(&rre->list, &rr->extlist);
+		} else if (rmap.rm_owner == XFS_RMAP_OWN_REFC) {
+			/* refcountbt block, dump it when we're done. */
+			rr->btblocks += rmap.rm_blockcount;
+			fsbno = XFS_AGB_TO_FSB(cur->bc_mp,
+					cur->bc_private.a.agno,
+					rmap.rm_startblock);
+			error = xfs_repair_collect_btree_extent(mp, &rr->btlist,
+					fsbno, rmap.rm_blockcount);
+			if (error)
+				goto out_error;
+		}
+	} while (XFS_RMAP_NON_INODE_OWNER(rmap.rm_owner) ||
+		 xfs_internal_inum(mp, rmap.rm_owner) ||
+		 (rmap.rm_flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK |
+				   XFS_RMAP_UNWRITTEN)));
+
+	*rec = rmap;
+	*have_rec = true;
+	return 0;
+
+out_error:
+	return error;
+}
+
+/* Recycle an idle rmap or allocate a new one. */
+static struct xfs_repair_refc_rmap *
+xfs_repair_refcountbt_get_rmap(
+	struct xfs_repair_refc		*rr)
+{
+	struct xfs_repair_refc_rmap	*rrm;
+
+	if (list_empty(&rr->rmap_idle)) {
+		rrm = kmem_alloc(sizeof(*rrm), KM_NOFS);
+		if (!rrm)
+			return NULL;
+		INIT_LIST_HEAD(&rrm->list);
+		return rrm;
+	}
+
+	rrm = list_first_entry(&rr->rmap_idle, struct xfs_repair_refc_rmap,
+			list);
+	list_del_init(&rrm->list);
+	return rrm;
+}
+
+/* Compare two btree extents. */
+static int
+xfs_repair_refcount_extent_cmp(
+	void				*priv,
+	struct list_head		*a,
+	struct list_head		*b)
+{
+	struct xfs_repair_refc_extent	*ap;
+	struct xfs_repair_refc_extent	*bp;
+
+	ap = container_of(a, struct xfs_repair_refc_extent, list);
+	bp = container_of(b, struct xfs_repair_refc_extent, list);
+
+	if (ap->refc.rc_startblock > bp->refc.rc_startblock)
+		return 1;
+	else if (ap->refc.rc_startblock < bp->refc.rc_startblock)
+		return -1;
+	return 0;
+}
+
+/* Record a reference count extent. */
+STATIC int
+xfs_repair_refcountbt_new_refc(
+	struct xfs_scrub_context	*sc,
+	struct xfs_repair_refc		*rr,
+	xfs_agblock_t			agbno,
+	xfs_extlen_t			len,
+	xfs_nlink_t			refcount)
+{
+	struct xfs_repair_refc_extent	*rre;
+	struct xfs_refcount_irec	irec;
+
+	irec.rc_startblock = agbno;
+	irec.rc_blockcount = len;
+	irec.rc_refcount = refcount;
+
+	trace_xfs_repair_refcount_extent_fn(sc->tp->t_mountp, sc->sa.agno,
+			&irec);
+
+	rre = kmem_alloc(sizeof(*rre), KM_NOFS);
+	if (!rre)
+		return -ENOMEM;
+	INIT_LIST_HEAD(&rre->list);
+	rre->refc = irec;
+	list_add_tail(&rre->list, &rr->extlist);
+
+	return 0;
+}
+
+/* Rebuild the refcount btree. */
+#define RMAP_END(r)	((r).rm_startblock + (r).rm_blockcount)
+int
+xfs_repair_refcountbt(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_repair_refc		rr;
+	struct xfs_rmap_irec		rmap;
+	struct xfs_owner_info		oinfo;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_repair_refc_rmap	*rrm;
+	struct xfs_repair_refc_rmap	*n;
+	struct xfs_repair_refc_extent	*rre;
+	struct xfs_repair_refc_extent	*o;
+	struct xfs_buf			*bp = NULL;
+	struct xfs_agf			*agf;
+	struct xfs_btree_cur		*cur;
+	struct xfs_perag		*pag;
+	uint64_t			nr_records;
+	xfs_fsblock_t			btfsb;
+	size_t				old_stack_sz;
+	size_t				stack_sz = 0;
+	xfs_agblock_t			sbno;
+	xfs_agblock_t			cbno;
+	xfs_agblock_t			nbno;
+	bool				have;
+	int				have_gt;
+	int				error = 0;
+
+	/* We require the rmapbt to rebuild anything. */
+	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+		return -EOPNOTSUPP;
+
+	INIT_LIST_HEAD(&rr.rmap_bag);
+	INIT_LIST_HEAD(&rr.rmap_idle);
+	INIT_LIST_HEAD(&rr.extlist);
+	INIT_LIST_HEAD(&rr.btlist);
+	rr.btblocks = 0;
+	nr_records = 0;
+	xfs_rmap_ag_owner(&oinfo, XFS_RMAP_OWN_REFC);
+
+	/* Start the rmapbt cursor to the left of all records. */
+	cur = xfs_rmapbt_init_cursor(mp, sc->tp, sc->sa.agf_bp, sc->sa.agno);
+	error = xfs_rmap_lookup_le(cur, 0, 0, 0, 0, 0, &have_gt);
+	if (error)
+		return error;
+	ASSERT(have_gt == 0);
+
+	/* Process reverse mappings into refcount data. */
+	while (xfs_btree_has_more_records(cur)) {
+		/* Push all rmaps with pblk == sbno onto the stack */
+		error = xfs_repair_refcountbt_next_rmap(cur, &rr, &rmap, &have);
+		if (error)
+			goto out;
+		if (!have)
+			break;
+		sbno = cbno = rmap.rm_startblock;
+		while (have && rmap.rm_startblock == sbno) {
+			rrm = xfs_repair_refcountbt_get_rmap(&rr);
+			if (!rrm)
+				goto out;
+			rrm->rmap = rmap;
+			list_add_tail(&rrm->list, &rr.rmap_bag);
+			stack_sz++;
+			error = xfs_repair_refcountbt_next_rmap(cur, &rr, &rmap,
+					&have);
+			if (error)
+				goto out;
+		}
+		error = xfs_btree_decrement(cur, 0, &have_gt);
+		if (error)
+			goto out;
+		XFS_WANT_CORRUPTED_GOTO(mp, have_gt, out);
+
+		/* Set nbno to the bno of the next refcount change */
+		nbno = have ? rmap.rm_startblock : NULLAGBLOCK;
+		list_for_each_entry(rrm, &rr.rmap_bag, list)
+			nbno = min_t(xfs_agblock_t, nbno, RMAP_END(rrm->rmap));
+
+		ASSERT(nbno > sbno);
+		old_stack_sz = stack_sz;
+
+		/* While stack isn't empty... */
+		while (stack_sz) {
+			/* Pop all rmaps that end at nbno */
+			list_for_each_entry_safe(rrm, n, &rr.rmap_bag, list) {
+				if (RMAP_END(rrm->rmap) != nbno)
+					continue;
+				stack_sz--;
+				list_del_init(&rrm->list);
+				list_add(&rrm->list, &rr.rmap_idle);
+			}
+
+			/* Push array items that start at nbno */
+			error = xfs_repair_refcountbt_next_rmap(cur, &rr, &rmap,
+					&have);
+			if (error)
+				goto out;
+			while (have && rmap.rm_startblock == nbno) {
+				rrm = xfs_repair_refcountbt_get_rmap(&rr);
+				if (!rrm)
+					goto out;
+				rrm->rmap = rmap;
+				list_add_tail(&rrm->list, &rr.rmap_bag);
+				stack_sz++;
+				error = xfs_repair_refcountbt_next_rmap(cur,
+						&rr, &rmap, &have);
+				if (error)
+					goto out;
+			}
+			error = xfs_btree_decrement(cur, 0, &have_gt);
+			if (error)
+				goto out;
+			XFS_WANT_CORRUPTED_GOTO(mp, have_gt, out);
+
+			/* Emit refcount if necessary */
+			ASSERT(nbno > cbno);
+			if (stack_sz != old_stack_sz) {
+				if (old_stack_sz > 1) {
+					error = xfs_repair_refcountbt_new_refc(
+							sc, &rr, cbno,
+							nbno - cbno,
+							old_stack_sz);
+					if (error)
+						goto out;
+					nr_records++;
+				}
+				cbno = nbno;
+			}
+
+			/* Stack empty, go find the next rmap */
+			if (stack_sz == 0)
+				break;
+			old_stack_sz = stack_sz;
+			sbno = nbno;
+
+			/* Set nbno to the bno of the next refcount change */
+			nbno = have ? rmap.rm_startblock : NULLAGBLOCK;
+			list_for_each_entry(rrm, &rr.rmap_bag, list)
+				nbno = min_t(xfs_agblock_t, nbno,
+						RMAP_END(rrm->rmap));
+
+			/* Emit reverse mappings, if needed */
+			ASSERT(nbno > sbno);
+		}
+	}
+	ASSERT(list_empty(&rr.rmap_bag));
+	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+	cur = NULL;
+
+	/* Free all the rmap records. */
+	list_for_each_entry_safe(rrm, n, &rr.rmap_idle, list) {
+		list_del(&rrm->list);
+		kmem_free(rrm);
+	}
+	list_for_each_entry_safe(rrm, n, &rr.rmap_bag, list) {
+		list_del(&rrm->list);
+		kmem_free(rrm);
+	}
+
+	/* Do we actually have enough space to do this? */
+	pag = xfs_perag_get(mp, sc->sa.agno);
+	if (!xfs_repair_ag_has_space(pag,
+			xfs_refcountbt_calc_size(mp, nr_records),
+			XFS_AG_RESV_METADATA)) {
+		xfs_perag_put(pag);
+		error = -ENOSPC;
+		goto out;
+	}
+	xfs_perag_put(pag);
+
+	agf = XFS_BUF_TO_AGF(sc->sa.agf_bp);
+	/* Initialize a new btree root. */
+	error = xfs_repair_alloc_ag_block(sc, &oinfo, &btfsb,
+			XFS_AG_RESV_METADATA);
+	if (error)
+		goto out;
+	error = xfs_repair_init_btblock(sc, btfsb, &bp, XFS_REFC_CRC_MAGIC,
+			&xfs_refcountbt_buf_ops);
+	if (error)
+		goto out;
+	agf->agf_refcount_root = cpu_to_be32(XFS_FSB_TO_AGBNO(mp, btfsb));
+	agf->agf_refcount_level = cpu_to_be32(1);
+	agf->agf_refcount_blocks = cpu_to_be32(1);
+	xfs_alloc_log_agf(sc->tp, sc->sa.agf_bp, XFS_AGF_REFCOUNT_BLOCKS |
+			XFS_AGF_REFCOUNT_ROOT | XFS_AGF_REFCOUNT_LEVEL);
+	error = xfs_repair_roll_ag_trans(sc);
+	if (error)
+		goto out;
+
+	/* Insert records into the new btree. */
+	list_sort(NULL, &rr.extlist, xfs_repair_refcount_extent_cmp);
+	list_for_each_entry_safe(rre, o, &rr.extlist, list) {
+		/* Insert into the refcountbt. */
+		cur = xfs_refcountbt_init_cursor(mp, sc->tp, sc->sa.agf_bp,
+				sc->sa.agno, NULL);
+		error = xfs_refcount_lookup_eq(cur, rre->refc.rc_startblock,
+				&have_gt);
+		if (error)
+			goto out;
+		XFS_WANT_CORRUPTED_GOTO(mp, have_gt == 0, out);
+		error = xfs_refcount_insert(cur, &rre->refc, &have_gt);
+		if (error)
+			goto out;
+		XFS_WANT_CORRUPTED_GOTO(mp, have_gt == 1, out);
+		xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
+		cur = NULL;
+
+		error = xfs_repair_roll_ag_trans(sc);
+		if (error)
+			goto out;
+
+		list_del(&rre->list);
+		kmem_free(rre);
+	}
+
+	/* Free the old refcountbt blocks if they're not in use. */
+	error = xfs_repair_reap_btree_extents(sc, &rr.btlist, &oinfo,
+			XFS_AG_RESV_METADATA);
+	if (error)
+		goto out;
+
+	return error;
+
+out:
+	if (cur)
+		xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
+	xfs_repair_cancel_btree_extents(sc, &rr.btlist);
+	list_for_each_entry_safe(rrm, n, &rr.rmap_idle, list) {
+		list_del(&rrm->list);
+		kmem_free(rrm);
+	}
+	list_for_each_entry_safe(rrm, n, &rr.rmap_bag, list) {
+		list_del(&rrm->list);
+		kmem_free(rrm);
+	}
+	list_for_each_entry_safe(rre, o, &rr.extlist, list) {
+		list_del(&rre->list);
+		kmem_free(rre);
+	}
+	return error;
+}

[PATCH 52/55] xfs: online repair of inodes

[Darrick J. Wong]

Try to reinitialize corrupt inodes, or clear the reflink flag
if it's not needed.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/common.c  |    2 -
 fs/xfs/scrub/common.h  |    1 
 fs/xfs/scrub/inode.c   |  153 ++++++++++++++++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_bmap_util.c |    4 +
 fs/xfs/xfs_reflink.c   |   15 +++--
 fs/xfs/xfs_reflink.h   |    6 +-
 6 files changed, 168 insertions(+), 13 deletions(-)


diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 5c61985..51ae56d 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -779,7 +779,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag_iallocbt, xfs_scrub_finobt, xfs_repair_iallocbt, xfs_sb_version_hasfinobt},
 	{xfs_scrub_setup_ag_header_freeze, xfs_scrub_rmapbt, xfs_repair_rmapbt, xfs_sb_version_hasrmapbt},
 	{xfs_scrub_setup_ag_header, xfs_scrub_refcountbt, xfs_repair_refcountbt, xfs_sb_version_hasreflink},
-	{xfs_scrub_setup_inode_raw, xfs_scrub_inode, NULL, NULL},
+	{xfs_scrub_setup_inode_raw, xfs_scrub_inode, xfs_repair_inode, NULL},
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_data, NULL, NULL},
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_attr, NULL, NULL},
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_cow, NULL, NULL},
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 896d6c0..137092f 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -393,5 +393,6 @@ int xfs_repair_allocbt(struct xfs_scrub_context *sc);
 int xfs_repair_iallocbt(struct xfs_scrub_context *sc);
 int xfs_repair_rmapbt(struct xfs_scrub_context *sc);
 int xfs_repair_refcountbt(struct xfs_scrub_context *sc);
+int xfs_repair_inode(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/inode.c b/fs/xfs/scrub/inode.c
index e79fe5b..0bec017 100644
--- a/fs/xfs/scrub/inode.c
+++ b/fs/xfs/scrub/inode.c
@@ -41,6 +41,7 @@
 #include "xfs_rmap.h"
 #include "xfs_bmap.h"
 #include "xfs_bmap_util.h"
+#include "xfs_reflink.h"
 #include "scrub/common.h"
 
 /*
@@ -450,3 +451,155 @@ xfs_scrub_inode(
 #undef XFS_SCRUB_INODE_OP_ERROR_GOTO
 #undef XFS_SCRUB_INODE_GOTO
 #undef XFS_SCRUB_INODE_CHECK
+
+/* Repair an inode's fields. */
+int
+xfs_repair_inode(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_imap			imap;
+	struct xfs_mount		*mp = sc->tp->t_mountp;
+	struct xfs_buf			*bp;
+	struct xfs_dinode		*dip;
+	struct xfs_inode		*ip;
+	xfs_ino_t			ino;
+	unsigned long long		count;
+	uint64_t			flags2;
+	uint32_t			nextents;
+	uint16_t			flags;
+	int				error = 0;
+
+	if (!xfs_sb_version_hascrc(&mp->m_sb))
+		return -EOPNOTSUPP;
+
+	/* Are we fixing this thing manually? */
+	if (!sc->ip) {
+		/* Map & read inode. */
+		ino = sc->sm->sm_ino;
+		error = xfs_imap(mp, sc->tp, ino, &imap, XFS_IGET_UNTRUSTED);
+		if (error)
+			goto out;
+
+		error = xfs_trans_read_buf(mp, sc->tp, mp->m_ddev_targp,
+				imap.im_blkno, imap.im_len, XBF_UNMAPPED, &bp,
+				NULL);
+		if (error)
+			goto out;
+
+		/* Fix everything the verifier will complain about. */
+		bp->b_ops = &xfs_inode_buf_ops;
+		dip = xfs_buf_offset(bp, imap.im_boffset);
+		dip->di_magic = cpu_to_be16(XFS_DINODE_MAGIC);
+		if (!xfs_dinode_good_version(mp, dip->di_version))
+			dip->di_version = 3;
+		dip->di_ino = cpu_to_be64(ino);
+		uuid_copy(&dip->di_uuid, &mp->m_sb.sb_meta_uuid);
+		flags = be16_to_cpu(dip->di_flags);
+		flags2 = be64_to_cpu(dip->di_flags2);
+		if (xfs_sb_version_hasreflink(&mp->m_sb))
+			flags2 |= XFS_DIFLAG2_REFLINK;
+		else
+			flags2 &= ~(XFS_DIFLAG2_REFLINK |
+				    XFS_DIFLAG2_COWEXTSIZE);
+		if (flags & XFS_DIFLAG_REALTIME)
+			flags2 &= ~XFS_DIFLAG2_REFLINK;
+		if (flags2 & XFS_DIFLAG2_REFLINK)
+			flags2 &= ~XFS_DIFLAG2_DAX;
+		dip->di_flags = cpu_to_be16(flags);
+		dip->di_flags2 = cpu_to_be64(flags2);
+		dip->di_gen = cpu_to_be32(sc->sm->sm_gen);
+		if (be64_to_cpu(dip->di_size) & (1ULL << 63))
+			dip->di_size = cpu_to_be64((1ULL << 63) - 1);
+
+		/* Write out the inode... */
+		xfs_dinode_calc_crc(mp, dip);
+		xfs_trans_buf_set_type(sc->tp, bp, XFS_BLFT_DINO_BUF);
+		xfs_trans_log_buf(sc->tp, bp, imap.im_boffset,
+				imap.im_boffset + mp->m_sb.sb_inodesize - 1);
+		error = xfs_trans_roll(&sc->tp, NULL);
+		if (error)
+			goto out;
+
+		/* ...and reload it? */
+		error = xfs_iget(mp, sc->tp, ino,
+				XFS_IGET_UNTRUSTED | XFS_IGET_DONTCACHE,
+				0, &sc->ip);
+		if (error)
+			goto out;
+		xfs_ilock(sc->ip, XFS_MMAPLOCK_EXCL);
+		xfs_ilock(sc->ip, XFS_IOLOCK_EXCL);
+		xfs_ilock(sc->ip, XFS_ILOCK_EXCL);
+	}
+
+	ip = sc->ip;
+	xfs_trans_ijoin(sc->tp, ip, 0);
+
+	/* di_size */
+	if (!S_ISDIR(VFS_I(ip)->i_mode) && !S_ISREG(VFS_I(ip)->i_mode) &&
+	    !S_ISLNK(VFS_I(ip)->i_mode)) {
+		i_size_write(VFS_I(ip), 0);
+		ip->i_d.di_size = 0;
+	}
+
+	/* di_flags */
+	flags = ip->i_d.di_flags;
+	if ((flags & XFS_DIFLAG_IMMUTABLE) && (flags & XFS_DIFLAG_APPEND))
+		flags &= ~XFS_DIFLAG_APPEND;
+
+	if ((flags & XFS_DIFLAG_FILESTREAM) && (flags & XFS_DIFLAG_REALTIME))
+		flags &= ~XFS_DIFLAG_FILESTREAM;
+	ip->i_d.di_flags = flags;
+
+	/* di_nblocks/di_nextents/di_anextents */
+	count = 0;
+	error = xfs_bmap_count_blocks(sc->tp, sc->ip, XFS_DATA_FORK,
+			&nextents, &count);
+	if (error)
+		goto out;
+	ip->i_d.di_nextents = nextents;
+
+	error = xfs_bmap_count_blocks(sc->tp, sc->ip, XFS_ATTR_FORK,
+			&nextents, &count);
+	if (error)
+		goto out;
+	ip->i_d.di_anextents = nextents;
+	ip->i_d.di_nblocks = count;
+	if (ip->i_d.di_anextents != 0 && ip->i_d.di_forkoff == 0)
+		ip->i_d.di_anextents = 0;
+
+	/* Do we have prealloc blocks? */
+	if (S_ISREG(VFS_I(ip)->i_mode) && !(flags & XFS_DIFLAG_PREALLOC) &&
+	    (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS ||
+	     ip->i_d.di_format == XFS_DINODE_FMT_BTREE)) {
+		struct xfs_bmbt_irec		got;
+		struct xfs_ifork		*ifp;
+		xfs_fileoff_t			lblk;
+		xfs_extnum_t			idx;
+		bool				found;
+
+		lblk = XFS_B_TO_FSB(mp, i_size_read(VFS_I(sc->ip)));
+		ifp = XFS_IFORK_PTR(sc->ip, XFS_DATA_FORK);
+		found = xfs_iext_lookup_extent(sc->ip, ifp, lblk, &idx, &got);
+		while (found) {
+			if (got.br_startoff >= lblk &&
+			    got.br_state == XFS_EXT_NORM) {
+				ip->i_d.di_flags |= XFS_DIFLAG_PREALLOC;
+				break;
+			}
+			lblk = got.br_startoff + got.br_blockcount;
+			found = xfs_iext_get_extent(ifp, ++idx, &got);
+		}
+	}
+
+	/* Commit inode core changes. */
+	xfs_trans_log_inode(sc->tp, ip, XFS_ILOG_CORE);
+	error = xfs_trans_roll(&sc->tp, ip);
+	if (error)
+		goto out;
+
+	if (xfs_is_reflink_inode(sc->ip))
+		return xfs_reflink_clear_inode_flag(sc->ip, &sc->tp);
+
+out:
+	return error;
+}
diff --git a/fs/xfs/xfs_bmap_util.c b/fs/xfs/xfs_bmap_util.c
index d983f28..cfba408 100644
--- a/fs/xfs/xfs_bmap_util.c
+++ b/fs/xfs/xfs_bmap_util.c
@@ -481,8 +481,8 @@ xfs_getbmap_adjust_shared(
 
 	agno = XFS_FSB_TO_AGNO(mp, map->br_startblock);
 	agbno = XFS_FSB_TO_AGBNO(mp, map->br_startblock);
-	error = xfs_reflink_find_shared(mp, agno, agbno, map->br_blockcount,
-			&ebno, &elen, true);
+	error = xfs_reflink_find_shared(mp, NULL, agno, agbno,
+			map->br_blockcount, &ebno, &elen, true);
 	if (error)
 		return error;
 
diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c
index 07593a3..bdecdb8 100644
--- a/fs/xfs/xfs_reflink.c
+++ b/fs/xfs/xfs_reflink.c
@@ -128,6 +128,7 @@
 int
 xfs_reflink_find_shared(
 	struct xfs_mount	*mp,
+	struct xfs_trans	*tp,
 	xfs_agnumber_t		agno,
 	xfs_agblock_t		agbno,
 	xfs_extlen_t		aglen,
@@ -139,18 +140,18 @@ xfs_reflink_find_shared(
 	struct xfs_btree_cur	*cur;
 	int			error;
 
-	error = xfs_alloc_read_agf(mp, NULL, agno, 0, &agbp);
+	error = xfs_alloc_read_agf(mp, tp, agno, 0, &agbp);
 	if (error)
 		return error;
 
-	cur = xfs_refcountbt_init_cursor(mp, NULL, agbp, agno, NULL);
+	cur = xfs_refcountbt_init_cursor(mp, tp, agbp, agno, NULL);
 
 	error = xfs_refcount_find_shared(cur, agbno, aglen, fbno, flen,
 			find_end_of_shared);
 
 	xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
 
-	xfs_buf_relse(agbp);
+	xfs_trans_brelse(tp, agbp);
 	return error;
 }
 
@@ -194,7 +195,7 @@ xfs_reflink_trim_around_shared(
 	agbno = XFS_FSB_TO_AGBNO(ip->i_mount, irec->br_startblock);
 	aglen = irec->br_blockcount;
 
-	error = xfs_reflink_find_shared(ip->i_mount, agno, agbno,
+	error = xfs_reflink_find_shared(ip->i_mount, NULL, agno, agbno,
 			aglen, &fbno, &flen, true);
 	if (error)
 		return error;
@@ -1256,8 +1257,8 @@ xfs_reflink_dirty_extents(
 			agbno = XFS_FSB_TO_AGBNO(mp, map[1].br_startblock);
 			aglen = map[1].br_blockcount;
 
-			error = xfs_reflink_find_shared(mp, agno, agbno, aglen,
-					&rbno, &rlen, true);
+			error = xfs_reflink_find_shared(mp, NULL, agno, agbno,
+					aglen, &rbno, &rlen, true);
 			if (error)
 				goto out;
 			if (rbno == NULLAGBLOCK)
@@ -1330,7 +1331,7 @@ xfs_reflink_clear_inode_flag(
 		agbno = XFS_FSB_TO_AGBNO(mp, map.br_startblock);
 		aglen = map.br_blockcount;
 
-		error = xfs_reflink_find_shared(mp, agno, agbno, aglen,
+		error = xfs_reflink_find_shared(mp, *tpp, agno, agbno, aglen,
 				&rbno, &rlen, false);
 		if (error)
 			return error;
diff --git a/fs/xfs/xfs_reflink.h b/fs/xfs/xfs_reflink.h
index aa6a4d6..2a18e4d 100644
--- a/fs/xfs/xfs_reflink.h
+++ b/fs/xfs/xfs_reflink.h
@@ -20,9 +20,9 @@
 #ifndef __XFS_REFLINK_H
 #define __XFS_REFLINK_H 1
 
-extern int xfs_reflink_find_shared(struct xfs_mount *mp, xfs_agnumber_t agno,
-		xfs_agblock_t agbno, xfs_extlen_t aglen, xfs_agblock_t *fbno,
-		xfs_extlen_t *flen, bool find_maximal);
+extern int xfs_reflink_find_shared(struct xfs_mount *mp, struct xfs_trans *tp,
+		xfs_agnumber_t agno, xfs_agblock_t agbno, xfs_extlen_t aglen,
+		xfs_agblock_t *fbno, xfs_extlen_t *flen, bool find_maximal);
 extern int xfs_reflink_trim_around_shared(struct xfs_inode *ip,
 		struct xfs_bmbt_irec *irec, bool *shared, bool *trimmed);
 

[PATCH 53/55] xfs: repair inode block maps

[Darrick J. Wong]

Use the reverse-mapping btree information to rebuild an inode fork.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_bmap.c |   20 ++-
 fs/xfs/libxfs/xfs_bmap.h |    6 +
 fs/xfs/scrub/bmap.c      |  315 +++++++++++++++++++++++++++++++++++++++++++++-
 fs/xfs/scrub/common.c    |    4 -
 fs/xfs/scrub/common.h    |   10 +
 fs/xfs/scrub/inode.c     |   45 ++++++-
 6 files changed, 380 insertions(+), 20 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c
index 3ac13be..d4532c7 100644
--- a/fs/xfs/libxfs/xfs_bmap.c
+++ b/fs/xfs/libxfs/xfs_bmap.c
@@ -2151,9 +2151,12 @@ xfs_bmap_add_extent_delay_real(
 	}
 
 	/* add reverse mapping */
-	error = xfs_rmap_map_extent(mp, bma->dfops, bma->ip, whichfork, new);
-	if (error)
-		goto done;
+	if (!(bma->flags & XFS_BMAPI_NORMAP)) {
+		error = xfs_rmap_map_extent(mp, bma->dfops, bma->ip,
+				whichfork, new);
+		if (error)
+			goto done;
+	}
 
 	/* convert to a btree if necessary */
 	if (xfs_bmap_needs_btree(bma->ip, whichfork)) {
@@ -3094,9 +3097,12 @@ xfs_bmap_add_extent_hole_real(
 	}
 
 	/* add reverse mapping */
-	error = xfs_rmap_map_extent(mp, bma->dfops, bma->ip, whichfork, new);
-	if (error)
-		goto done;
+	if (!(bma->flags & XFS_BMAPI_NORMAP)) {
+		error = xfs_rmap_map_extent(mp, bma->dfops, bma->ip,
+				whichfork, new);
+		if (error)
+			goto done;
+	}
 
 	/* convert to a btree if necessary */
 	if (xfs_bmap_needs_btree(bma->ip, whichfork)) {
@@ -4537,8 +4543,6 @@ xfs_bmapi_write(
 	ASSERT(len > 0);
 	ASSERT(XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_LOCAL);
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
-	ASSERT(!(flags & XFS_BMAPI_REMAP) || whichfork == XFS_DATA_FORK);
-	ASSERT(!(flags & XFS_BMAPI_PREALLOC) || !(flags & XFS_BMAPI_REMAP));
 	ASSERT(!(flags & XFS_BMAPI_CONVERT) || !(flags & XFS_BMAPI_REMAP));
 	ASSERT(!(flags & XFS_BMAPI_PREALLOC) || whichfork != XFS_COW_FORK);
 	ASSERT(!(flags & XFS_BMAPI_CONVERT) || whichfork != XFS_COW_FORK);
diff --git a/fs/xfs/libxfs/xfs_bmap.h b/fs/xfs/libxfs/xfs_bmap.h
index cecd094..15454749 100644
--- a/fs/xfs/libxfs/xfs_bmap.h
+++ b/fs/xfs/libxfs/xfs_bmap.h
@@ -110,6 +110,9 @@ struct xfs_extent_free_item
 /* Map something in the CoW fork. */
 #define XFS_BMAPI_COWFORK	0x200
 
+/* Don't update the rmap btree. */
+#define XFS_BMAPI_NORMAP	0x400
+
 #define XFS_BMAPI_FLAGS \
 	{ XFS_BMAPI_ENTIRE,	"ENTIRE" }, \
 	{ XFS_BMAPI_METADATA,	"METADATA" }, \
@@ -120,7 +123,8 @@ struct xfs_extent_free_item
 	{ XFS_BMAPI_CONVERT,	"CONVERT" }, \
 	{ XFS_BMAPI_ZERO,	"ZERO" }, \
 	{ XFS_BMAPI_REMAP,	"REMAP" }, \
-	{ XFS_BMAPI_COWFORK,	"COWFORK" }
+	{ XFS_BMAPI_COWFORK,	"COWFORK" }, \
+	{ XFS_BMAPI_NORMAP,	"NORMAP" }
 
 
 static inline int xfs_bmapi_aflag(int w)
diff --git a/fs/xfs/scrub/bmap.c b/fs/xfs/scrub/bmap.c
index 96dbe66..7b1742b 100644
--- a/fs/xfs/scrub/bmap.c
+++ b/fs/xfs/scrub/bmap.c
@@ -36,6 +36,7 @@
 #include "xfs_bmap_util.h"
 #include "xfs_bmap_btree.h"
 #include "xfs_rmap.h"
+#include "xfs_rmap_btree.h"
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
 #include "xfs_refcount.h"
@@ -44,17 +45,21 @@
 #include "scrub/btree.h"
 
 /* Set us up with an inode and AG headers, if needed. */
-int
-xfs_scrub_setup_inode_bmap(
+STATIC int
+__xfs_scrub_setup_inode_bmap(
 	struct xfs_scrub_context	*sc,
 	struct xfs_inode		*ip,
 	struct xfs_scrub_metadata	*sm,
-	bool				retry_deadlocked)
+	bool				retry_deadlocked,
+	bool				data)
 {
+	bool				is_repair;
 	int				error;
 
-	error = xfs_scrub_setup_inode(sc, ip, sm, retry_deadlocked);
-	if (error || !retry_deadlocked)
+	is_repair = (sm->sm_flags & XFS_SCRUB_FLAG_REPAIR);
+	error = __xfs_scrub_setup_inode(sc, ip, sm, retry_deadlocked,
+			data && is_repair);
+	if (error || (!retry_deadlocked && !is_repair))
 		return error;
 
 	error = xfs_scrub_ag_lock_all(sc);
@@ -66,6 +71,28 @@ xfs_scrub_setup_inode_bmap(
 	return xfs_scrub_teardown(sc, ip, error);
 }
 
+/* Set us up with an inode and AG headers, if needed. */
+int
+xfs_scrub_setup_inode_bmap(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				deadlocked)
+{
+	return __xfs_scrub_setup_inode_bmap(sc, ip, sm, deadlocked, false);
+}
+
+/* Set us up with an inode and AG headers, if needed. */
+int
+xfs_scrub_setup_inode_bmap_data(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				deadlocked)
+{
+	return __xfs_scrub_setup_inode_bmap(sc, ip, sm, deadlocked, true);
+}
+
 /*
  * Inode fork block mapping (BMBT) scrubber.
  * More complex than the others because we have to scrub
@@ -552,3 +579,281 @@ xfs_scrub_bmap_cow(
 
 	return xfs_scrub_bmap(sc, XFS_COW_FORK);
 }
+
+/* Inode fork block mapping (BMBT) repair. */
+
+struct xfs_repair_bmap_extent {
+	struct list_head		list;
+	struct xfs_rmap_irec		rmap;
+	xfs_agnumber_t			agno;
+};
+
+struct xfs_repair_bmap {
+	struct list_head		extlist;
+	struct list_head		btlist;
+	xfs_ino_t			ino;
+	xfs_rfsblock_t			bmbt_blocks;
+	int				whichfork;
+};
+
+/* Record extents that belong to this inode's fork. */
+STATIC int
+xfs_repair_bmap_extent_fn(
+	struct xfs_btree_cur		*cur,
+	struct xfs_rmap_irec		*rec,
+	void				*priv)
+{
+	struct xfs_repair_bmap		*rb = priv;
+	struct xfs_repair_bmap_extent	*rbe;
+	struct xfs_mount		*mp = cur->bc_mp;
+	xfs_fsblock_t			fsbno;
+	int				error = 0;
+
+	if (xfs_scrub_should_terminate(&error))
+		return error;
+
+	/* Skip extents which are not owned by this inode and fork. */
+	if (rec->rm_owner != rb->ino)
+		return 0;
+	else if (rb->whichfork == XFS_DATA_FORK &&
+		 (rec->rm_flags & XFS_RMAP_ATTR_FORK))
+		return 0;
+	else if (rb->whichfork == XFS_ATTR_FORK &&
+		 !(rec->rm_flags & XFS_RMAP_ATTR_FORK))
+		return 0;
+
+	/* Delete the old bmbt blocks later. */
+	if (rec->rm_flags & XFS_RMAP_BMBT_BLOCK) {
+		fsbno = XFS_AGB_TO_FSB(mp, cur->bc_private.a.agno,
+				rec->rm_startblock);
+		rb->bmbt_blocks += rec->rm_blockcount;
+		return xfs_repair_collect_btree_extent(mp, &rb->btlist,
+				fsbno, rec->rm_blockcount);
+	}
+
+	/* Remember this rmap. */
+	trace_xfs_repair_bmap_extent_fn(mp, cur->bc_private.a.agno,
+			rec->rm_startblock, rec->rm_blockcount, rec->rm_owner,
+			rec->rm_offset, rec->rm_flags);
+
+	rbe = kmem_alloc(sizeof(*rbe), KM_NOFS);
+	if (!rbe)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&rbe->list);
+	rbe->rmap = *rec;
+	rbe->agno = cur->bc_private.a.agno;
+	list_add_tail(&rbe->list, &rb->extlist);
+
+	return 0;
+}
+
+/* Compare two bmap extents. */
+static int
+xfs_repair_bmap_extent_cmp(
+	void				*priv,
+	struct list_head		*a,
+	struct list_head		*b)
+{
+	struct xfs_repair_bmap_extent	*ap;
+	struct xfs_repair_bmap_extent	*bp;
+
+	ap = container_of(a, struct xfs_repair_bmap_extent, list);
+	bp = container_of(b, struct xfs_repair_bmap_extent, list);
+
+	if (ap->rmap.rm_offset > bp->rmap.rm_offset)
+		return 1;
+	else if (ap->rmap.rm_offset < bp->rmap.rm_offset)
+		return -1;
+	return 0;
+}
+
+/* Repair an inode fork. */
+STATIC int
+xfs_repair_bmap(
+	struct xfs_scrub_context	*sc,
+	int				whichfork)
+{
+	struct xfs_repair_bmap		rb = {0};
+	struct xfs_bmbt_irec		bmap;
+	struct xfs_defer_ops		dfops;
+	struct xfs_owner_info		oinfo;
+	struct xfs_inode		*ip = sc->ip;
+	struct xfs_mount		*mp = ip->i_mount;
+	struct xfs_buf			*agf_bp = NULL;
+	struct xfs_repair_bmap_extent	*rbe;
+	struct xfs_repair_bmap_extent	*n;
+	struct xfs_btree_cur		*cur;
+	xfs_fsblock_t			firstfsb;
+	xfs_agnumber_t			agno;
+	xfs_extlen_t			extlen;
+	int				baseflags;
+	int				flags;
+	int				nimaps;
+	int				error = 0;
+
+	ASSERT(whichfork == XFS_DATA_FORK || whichfork == XFS_ATTR_FORK);
+
+	/* Don't know how to repair the other fork formats. */
+	if (XFS_IFORK_FORMAT(sc->ip, whichfork) != XFS_DINODE_FMT_EXTENTS &&
+	    XFS_IFORK_FORMAT(sc->ip, whichfork) != XFS_DINODE_FMT_BTREE)
+		return -EOPNOTSUPP;
+
+	/* Only files, symlinks, and directories get to have data forks. */
+	if (whichfork == XFS_DATA_FORK && !S_ISREG(VFS_I(ip)->i_mode) &&
+	    !S_ISDIR(VFS_I(ip)->i_mode) && !S_ISLNK(VFS_I(ip)->i_mode))
+		return -EINVAL;
+
+	/* If we somehow have delalloc extents, forget it. */
+	if (whichfork == XFS_DATA_FORK && ip->i_delayed_blks)
+		return -EBUSY;
+
+	/* We require the rmapbt to rebuild anything. */
+	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+		return -EOPNOTSUPP;
+
+	/* Don't know how to rebuild realtime data forks. */
+	if (XFS_IS_REALTIME_INODE(ip) && whichfork == XFS_DATA_FORK)
+		return -EOPNOTSUPP;
+
+	/*
+	 * If this is a file data fork, wait for all pending directio to
+	 * complete, then tear everything out of the page cache.
+	 */
+	if (S_ISREG(VFS_I(ip)->i_mode) && whichfork == XFS_DATA_FORK) {
+		inode_dio_wait(VFS_I(ip));
+		truncate_inode_pages(VFS_I(ip)->i_mapping, 0);
+	}
+
+	/* Collect all reverse mappings for this fork's extents. */
+	INIT_LIST_HEAD(&rb.extlist);
+	INIT_LIST_HEAD(&rb.btlist);
+	rb.ino = ip->i_ino;
+	rb.whichfork = whichfork;
+	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
+		ASSERT(xfs_scrub_ag_can_lock(sc, agno));
+		error = xfs_alloc_read_agf(mp, sc->tp, agno, 0, &agf_bp);
+		if (error)
+			goto out;
+		cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, agno);
+		error = xfs_rmap_query_all(cur, xfs_repair_bmap_extent_fn, &rb);
+		xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR :
+				XFS_BTREE_NOERROR);
+		if (error)
+			goto out;
+	}
+
+	/* Blow out the in-core fork and zero the on-disk fork. */
+	xfs_trans_ijoin(sc->tp, sc->ip, 0);
+	if (XFS_IFORK_PTR(ip, whichfork) != NULL)
+		xfs_idestroy_fork(sc->ip, whichfork);
+	XFS_IFORK_FMT_SET(sc->ip, whichfork, XFS_DINODE_FMT_EXTENTS);
+	XFS_IFORK_NEXT_SET(sc->ip, whichfork, 0);
+
+	/* Reinitialize the on-disk fork. */
+	if (whichfork == XFS_DATA_FORK) {
+		memset(&ip->i_df, 0, sizeof(struct xfs_ifork));
+		ip->i_df.if_flags |= XFS_IFEXTENTS;
+	} else if (whichfork == XFS_ATTR_FORK) {
+		if (list_empty(&rb.extlist))
+			ip->i_afp = NULL;
+		else {
+			ip->i_afp = kmem_zone_zalloc(xfs_ifork_zone, KM_NOFS);
+			ip->i_afp->if_flags |= XFS_IFEXTENTS;
+		}
+	}
+	xfs_trans_log_inode(sc->tp, sc->ip, XFS_ILOG_CORE);
+	error = xfs_trans_roll(&sc->tp, sc->ip);
+	if (error)
+		goto out;
+
+	baseflags = XFS_BMAPI_REMAP | XFS_BMAPI_NORMAP;
+	if (whichfork == XFS_ATTR_FORK)
+		baseflags |= XFS_BMAPI_ATTRFORK;
+
+	/* "Remap" the extents into the fork. */
+	list_sort(NULL, &rb.extlist, xfs_repair_bmap_extent_cmp);
+	list_for_each_entry_safe(rbe, n, &rb.extlist, list) {
+		/* Form the "new" mapping... */
+		bmap.br_startblock = XFS_AGB_TO_FSB(mp, rbe->agno,
+				rbe->rmap.rm_startblock);
+		bmap.br_startoff = rbe->rmap.rm_offset;
+		flags = 0;
+		if (rbe->rmap.rm_flags & XFS_RMAP_UNWRITTEN)
+			flags = XFS_BMAPI_PREALLOC;
+		while (rbe->rmap.rm_blockcount > 0) {
+			xfs_defer_init(&dfops, &firstfsb);
+			extlen = min_t(xfs_extlen_t, rbe->rmap.rm_blockcount,
+					MAXEXTLEN);
+			bmap.br_blockcount = extlen;
+
+			/* Drop the block counter... */
+			sc->ip->i_d.di_nblocks -= extlen;
+
+			/* Re-add the extent to the fork. */
+			nimaps = 1;
+			firstfsb = bmap.br_startblock;
+			error = xfs_bmapi_write(sc->tp, sc->ip,
+					bmap.br_startoff,
+					extlen, baseflags | flags, &firstfsb,
+					extlen, &bmap, &nimaps,
+					&dfops);
+			if (error)
+				goto out;
+
+			bmap.br_startblock += extlen;
+			bmap.br_startoff += extlen;
+			rbe->rmap.rm_blockcount -= extlen;
+			error = xfs_defer_finish(&sc->tp, &dfops, sc->ip);
+			if (error)
+				goto out;
+			/* Make sure we roll the transaction. */
+			error = xfs_trans_roll(&sc->tp, sc->ip);
+			if (error)
+				goto out;
+		}
+		list_del(&rbe->list);
+		kmem_free(rbe);
+	}
+
+	/* Decrease nblocks to reflect the freed bmbt blocks. */
+	if (rb.bmbt_blocks) {
+		sc->ip->i_d.di_nblocks -= rb.bmbt_blocks;
+		xfs_trans_log_inode(sc->tp, sc->ip, XFS_ILOG_CORE);
+		error = xfs_trans_roll(&sc->tp, sc->ip);
+		if (error)
+			goto out;
+	}
+
+	/* Dispose of all the old bmbt blocks. */
+	xfs_rmap_ino_bmbt_owner(&oinfo, sc->ip->i_ino, whichfork);
+	error = xfs_repair_reap_btree_extents(sc, &rb.btlist, &oinfo,
+			XFS_AG_RESV_NONE);
+	if (error)
+		goto out;
+
+	return error;
+out:
+	xfs_repair_cancel_btree_extents(sc, &rb.btlist);
+	list_for_each_entry_safe(rbe, n, &rb.extlist, list) {
+		list_del(&rbe->list);
+		kmem_free(rbe);
+	}
+	return error;
+}
+
+/* Repair an inode's data fork. */
+int
+xfs_repair_bmap_data(
+	struct xfs_scrub_context	*sc)
+{
+	return xfs_repair_bmap(sc, XFS_DATA_FORK);
+}
+
+/* Repair an inode's attr fork. */
+int
+xfs_repair_bmap_attr(
+	struct xfs_scrub_context	*sc)
+{
+	return xfs_repair_bmap(sc, XFS_ATTR_FORK);
+}
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 51ae56d..2460a66 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -780,8 +780,8 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_ag_header_freeze, xfs_scrub_rmapbt, xfs_repair_rmapbt, xfs_sb_version_hasrmapbt},
 	{xfs_scrub_setup_ag_header, xfs_scrub_refcountbt, xfs_repair_refcountbt, xfs_sb_version_hasreflink},
 	{xfs_scrub_setup_inode_raw, xfs_scrub_inode, xfs_repair_inode, NULL},
-	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_data, NULL, NULL},
-	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_attr, NULL, NULL},
+	{xfs_scrub_setup_inode_bmap_data, xfs_scrub_bmap_data, xfs_repair_bmap_data, NULL},
+	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_attr, xfs_repair_bmap_attr, NULL},
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_cow, NULL, NULL},
 	{xfs_scrub_setup_inode, xfs_scrub_directory, NULL, NULL},
 	{xfs_scrub_setup_inode_xattr, xfs_scrub_xattr, NULL, NULL},
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 137092f..19b307a 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -290,6 +290,10 @@ int xfs_scrub_setup_ag_iallocbt(struct xfs_scrub_context *sc,
 				struct xfs_inode *ip,
 				struct xfs_scrub_metadata *sm,
 				bool retry_deadlocked);
+int __xfs_scrub_setup_inode(struct xfs_scrub_context *sc,
+			  struct xfs_inode *ip,
+			  struct xfs_scrub_metadata *sm,
+			  bool retry_deadlocked, bool flush_data);
 int xfs_scrub_setup_inode(struct xfs_scrub_context *sc,
 			  struct xfs_inode *ip,
 			  struct xfs_scrub_metadata *sm,
@@ -302,6 +306,10 @@ int xfs_scrub_setup_inode_bmap(struct xfs_scrub_context *sc,
 			       struct xfs_inode *ip,
 			       struct xfs_scrub_metadata *sm,
 			       bool retry_deadlocked);
+int xfs_scrub_setup_inode_bmap_data(struct xfs_scrub_context *sc,
+				    struct xfs_inode *ip,
+				    struct xfs_scrub_metadata *sm,
+				    bool retry_deadlocked);
 int xfs_scrub_setup_inode_xattr(struct xfs_scrub_context *sc,
 				struct xfs_inode *ip,
 				struct xfs_scrub_metadata *sm,
@@ -394,5 +402,7 @@ int xfs_repair_iallocbt(struct xfs_scrub_context *sc);
 int xfs_repair_rmapbt(struct xfs_scrub_context *sc);
 int xfs_repair_refcountbt(struct xfs_scrub_context *sc);
 int xfs_repair_inode(struct xfs_scrub_context *sc);
+int xfs_repair_bmap_data(struct xfs_scrub_context *sc);
+int xfs_repair_bmap_attr(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/inode.c b/fs/xfs/scrub/inode.c
index 0bec017..e92ca7b 100644
--- a/fs/xfs/scrub/inode.c
+++ b/fs/xfs/scrub/inode.c
@@ -42,6 +42,7 @@
 #include "xfs_bmap.h"
 #include "xfs_bmap_util.h"
 #include "xfs_reflink.h"
+#include "xfs_bmap_btree.h"
 #include "scrub/common.h"
 
 /*
@@ -88,13 +89,15 @@ xfs_scrub_get_inode(
 
 /* Set us up with an inode. */
 int
-xfs_scrub_setup_inode(
+__xfs_scrub_setup_inode(
 	struct xfs_scrub_context	*sc,
 	struct xfs_inode		*ip,
 	struct xfs_scrub_metadata	*sm,
-	bool				retry_deadlocked)
+	bool				retry_deadlocked,
+	bool				flush_data)
 {
 	struct xfs_mount		*mp = ip->i_mount;
+	unsigned long long		resblks;
 	int				error;
 
 	memset(sc, 0, sizeof(*sc));
@@ -107,8 +110,31 @@ xfs_scrub_setup_inode(
 
 	xfs_ilock(sc->ip, XFS_IOLOCK_EXCL);
 	xfs_ilock(sc->ip, XFS_MMAPLOCK_EXCL);
+
+	/*
+	 * We don't want any ephemeral data fork updates sitting around
+	 * while we inspect block mappings, so wait for directio to finish
+	 * and flush dirty data if we have delalloc reservations.
+	 */
+	if (flush_data) {
+		inode_dio_wait(VFS_I(sc->ip));
+		error = filemap_write_and_wait(VFS_I(sc->ip)->i_mapping);
+		if (error)
+			goto out_unlock;
+	}
+
+	/*
+	 * Guess how many blocks we're going to need to rebuild an
+	 * entire bmap.  We don't actually know which fork, so err
+	 * on the side of asking for more blocks than we might
+	 * actually need.  Since we're reloading the btree sequentially
+	 * there should be fewer splits.
+	 */
+	resblks = xfs_bmbt_calc_size(mp,
+			max_t(xfs_extnum_t, sc->ip->i_d.di_nextents,
+				sc->ip->i_d.di_anextents));
 	error = xfs_scrub_trans_alloc(sm, mp, &M_RES(mp)->tr_itruncate,
-			0, 0, 0, &sc->tp);
+			resblks, 0, 0, &sc->tp);
 	if (error)
 		goto out_unlock;
 	xfs_ilock(sc->ip, XFS_ILOCK_EXCL);
@@ -137,7 +163,7 @@ xfs_scrub_setup_inode_raw(
 	if (sm->sm_ino && xfs_internal_inum(mp, sm->sm_ino))
 		return -ENOENT;
 
-	error = xfs_scrub_setup_inode(sc, ip, sm, retry_deadlocked);
+	error = __xfs_scrub_setup_inode(sc, ip, sm, retry_deadlocked, false);
 	if (error) {
 		memset(sc, 0, sizeof(*sc));
 		sc->ip = NULL;
@@ -155,6 +181,17 @@ xfs_scrub_setup_inode_raw(
 	return 0;
 }
 
+/* Set us up with an inode. */
+int
+xfs_scrub_setup_inode(
+	struct xfs_scrub_context	*sc,
+	struct xfs_inode		*ip,
+	struct xfs_scrub_metadata	*sm,
+	bool				retry_deadlocked)
+{
+	return __xfs_scrub_setup_inode(sc, ip, sm, retry_deadlocked, false);
+}
+
 /* Inode core */
 
 #define XFS_SCRUB_INODE_CHECK(fs_ok) \

[PATCH 54/55] xfs: repair damaged symlinks

[Darrick J. Wong]

Repair inconsistent symbolic link data.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/scrub/common.c  |    2 
 fs/xfs/scrub/common.h  |    1 
 fs/xfs/scrub/inode.c   |    1 
 fs/xfs/scrub/symlink.c |  235 ++++++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 238 insertions(+), 1 deletion(-)


diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 2460a66..ac2c492 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -785,7 +785,7 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
 	{xfs_scrub_setup_inode_bmap, xfs_scrub_bmap_cow, NULL, NULL},
 	{xfs_scrub_setup_inode, xfs_scrub_directory, NULL, NULL},
 	{xfs_scrub_setup_inode_xattr, xfs_scrub_xattr, NULL, NULL},
-	{xfs_scrub_setup_inode_symlink, xfs_scrub_symlink, NULL, NULL},
+	{xfs_scrub_setup_inode_symlink, xfs_scrub_symlink, xfs_repair_symlink, NULL},
 #ifdef CONFIG_XFS_RT
 	{xfs_scrub_setup_rt, xfs_scrub_rtbitmap, NULL, xfs_sb_version_hasrealtime},
 	{xfs_scrub_setup_rt, xfs_scrub_rtsummary, NULL, xfs_sb_version_hasrealtime},
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index 19b307a..accb2ac 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -404,5 +404,6 @@ int xfs_repair_refcountbt(struct xfs_scrub_context *sc);
 int xfs_repair_inode(struct xfs_scrub_context *sc);
 int xfs_repair_bmap_data(struct xfs_scrub_context *sc);
 int xfs_repair_bmap_attr(struct xfs_scrub_context *sc);
+int xfs_repair_symlink(struct xfs_scrub_context *sc);
 
 #endif	/* __XFS_REPAIR_COMMON_H__ */
diff --git a/fs/xfs/scrub/inode.c b/fs/xfs/scrub/inode.c
index e92ca7b..36bccf0 100644
--- a/fs/xfs/scrub/inode.c
+++ b/fs/xfs/scrub/inode.c
@@ -133,6 +133,7 @@ __xfs_scrub_setup_inode(
 	resblks = xfs_bmbt_calc_size(mp,
 			max_t(xfs_extnum_t, sc->ip->i_d.di_nextents,
 				sc->ip->i_d.di_anextents));
+	resblks = max_t(unsigned long long, resblks, XFS_SYMLINK_MAPS);
 	error = xfs_scrub_trans_alloc(sm, mp, &M_RES(mp)->tr_itruncate,
 			resblks, 0, 0, &sc->tp);
 	if (error)
diff --git a/fs/xfs/scrub/symlink.c b/fs/xfs/scrub/symlink.c
index 80a1e70..58784d1 100644
--- a/fs/xfs/scrub/symlink.c
+++ b/fs/xfs/scrub/symlink.c
@@ -33,6 +33,8 @@
 #include "xfs_inode.h"
 #include "xfs_inode_fork.h"
 #include "xfs_symlink.h"
+#include "xfs_bmap.h"
+#include "xfs_quota.h"
 #include "scrub/common.h"
 
 /* Set us up with an inode and a buffer for reading symlink targets. */
@@ -105,3 +107,236 @@ xfs_scrub_symlink(
 }
 #undef XFS_SCRUB_SYMLINK_GOTO
 #undef XFS_SCRUB_SYMLINK_CHECK
+
+/* Blow out the whole symlink; replace contents. */
+STATIC int
+xfs_repair_symlink_rewrite(
+	struct xfs_trans	**tpp,
+	struct xfs_inode	*ip,
+	const char		*target_path,
+	int			pathlen)
+{
+	struct xfs_defer_ops	dfops;
+	struct xfs_bmbt_irec	mval[XFS_SYMLINK_MAPS];
+	struct xfs_ifork	*ifp;
+	const char		*cur_chunk;
+	struct xfs_mount	*mp = (*tpp)->t_mountp;
+	struct xfs_buf		*bp;
+	xfs_fsblock_t		first_block;
+	xfs_fileoff_t		first_fsb;
+	xfs_filblks_t		fs_blocks;
+	xfs_daddr_t		d;
+	uint			resblks;
+	int			byte_cnt;
+	int			n;
+	int			nmaps;
+	int			offset;
+	int			error = 0;
+
+	ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
+
+	/* Truncate the whole data fork if it wasn't inline. */
+	if (!(ifp->if_flags & XFS_IFINLINE)) {
+		error = xfs_itruncate_extents(tpp, ip, XFS_DATA_FORK, 0);
+		if (error)
+			goto out;
+	}
+
+	/* Blow out the in-core fork and zero the on-disk fork. */
+	xfs_idestroy_fork(ip, XFS_DATA_FORK);
+	ip->i_d.di_format = XFS_DINODE_FMT_EXTENTS;
+	ip->i_d.di_nextents = 0;
+	memset(&ip->i_df, 0, sizeof(struct xfs_ifork));
+	ip->i_df.if_flags |= XFS_IFEXTENTS;
+
+	/* Rewrite an inline symlink. */
+	if (pathlen <= XFS_IFORK_DSIZE(ip)) {
+		xfs_init_local_fork(ip, XFS_DATA_FORK, target_path, pathlen);
+
+		i_size_write(VFS_I(ip), pathlen);
+		ip->i_d.di_size = pathlen;
+		ip->i_d.di_format = XFS_DINODE_FMT_LOCAL;
+		xfs_trans_log_inode(*tpp, ip, XFS_ILOG_DDATA | XFS_ILOG_CORE);
+		goto out;
+
+	}
+
+	/* Rewrite a remote symlink. */
+	fs_blocks = xfs_symlink_blocks(mp, pathlen);
+	first_fsb = 0;
+	nmaps = XFS_SYMLINK_MAPS;
+
+	/* Reserve quota for new blocks. */
+	error = xfs_trans_reserve_quota_nblks(*tpp, ip, fs_blocks, 0,
+			XFS_QMOPT_RES_REGBLKS);
+	if (error)
+		goto out;
+
+	/* Map blocks, write symlink target. */
+	xfs_defer_init(&dfops, &first_block);
+
+	error = xfs_bmapi_write(*tpp, ip, first_fsb, fs_blocks,
+			  XFS_BMAPI_METADATA, &first_block, fs_blocks,
+			  mval, &nmaps, &dfops);
+	if (error)
+		goto out_bmap_cancel;
+
+	if (resblks)
+		resblks -= fs_blocks;
+	ip->i_d.di_size = pathlen;
+	i_size_write(VFS_I(ip), pathlen);
+	xfs_trans_log_inode(*tpp, ip, XFS_ILOG_CORE);
+
+	cur_chunk = target_path;
+	offset = 0;
+	for (n = 0; n < nmaps; n++) {
+		char	*buf;
+
+		d = XFS_FSB_TO_DADDR(mp, mval[n].br_startblock);
+		byte_cnt = XFS_FSB_TO_B(mp, mval[n].br_blockcount);
+		bp = xfs_trans_get_buf(*tpp, mp->m_ddev_targp, d,
+				       BTOBB(byte_cnt), 0);
+		if (!bp) {
+			error = -ENOMEM;
+			goto out_bmap_cancel;
+		}
+		bp->b_ops = &xfs_symlink_buf_ops;
+
+		byte_cnt = XFS_SYMLINK_BUF_SPACE(mp, byte_cnt);
+		byte_cnt = min(byte_cnt, pathlen);
+
+		buf = bp->b_addr;
+		buf += xfs_symlink_hdr_set(mp, ip->i_ino, offset,
+					   byte_cnt, bp);
+
+		memcpy(buf, cur_chunk, byte_cnt);
+
+		cur_chunk += byte_cnt;
+		pathlen -= byte_cnt;
+		offset += byte_cnt;
+
+		xfs_trans_buf_set_type(*tpp, bp, XFS_BLFT_SYMLINK_BUF);
+		xfs_trans_log_buf(*tpp, bp, 0, (buf + byte_cnt - 1) -
+						(char *)bp->b_addr);
+	}
+	ASSERT(pathlen == 0);
+
+	error = xfs_defer_finish(tpp, &dfops, NULL);
+	if (error)
+		goto out_bmap_cancel;
+
+	return 0;
+
+out_bmap_cancel:
+	xfs_defer_cancel(&dfops);
+out:
+	return error;
+}
+
+int
+xfs_repair_symlink(
+	struct xfs_scrub_context	*sc)
+{
+	struct xfs_bmbt_irec		mval[XFS_SYMLINK_MAPS];
+	struct xfs_inode		*ip = sc->ip;
+	struct xfs_mount		*mp = ip->i_mount;
+	struct xfs_ifork		*ifp;
+	struct xfs_buf			*bp;
+	loff_t				len;
+	size_t				newlen;
+	xfs_daddr_t			d;
+	int				fsblocks;
+	int				nmaps = XFS_SYMLINK_MAPS;
+	int				nr;
+	int				offset;
+	int				n;
+	int				byte_cnt;
+	int				error = 0;
+
+	ifp = XFS_IFORK_PTR(ip, XFS_DATA_FORK);
+	len = i_size_read(VFS_I(ip));
+	xfs_trans_ijoin(sc->tp, ip, 0);
+
+	/* Truncate the inode if there's a zero inside the length. */
+	if (ifp->if_flags & XFS_IFINLINE) {
+		if (ifp->if_u1.if_data)
+			newlen = strnlen(ifp->if_u1.if_data,
+					XFS_IFORK_DSIZE(ip));
+		else {
+			newlen = 1;
+			ifp->if_u1.if_data = ifp->if_u2.if_inline_data;
+			ifp->if_u1.if_data[0] = '/';
+		}
+		if (len > newlen) {
+			i_size_write(VFS_I(ip), newlen);
+			ip->i_d.di_size = newlen;
+			xfs_trans_log_inode(sc->tp, ip, XFS_ILOG_DDATA |
+					XFS_ILOG_CORE);
+		}
+		goto out;
+	}
+
+	fsblocks = xfs_symlink_blocks(mp, len);
+	error = xfs_bmapi_read(ip, 0, fsblocks, mval, &nmaps, 0);
+	if (error)
+		goto out;
+
+	/* Fix everything that fails the verifiers. */
+	offset = 0;
+	for (n = 0; n < nmaps; n++) {
+		d = XFS_FSB_TO_DADDR(mp, mval[n].br_startblock);
+		byte_cnt = XFS_FSB_TO_B(mp, mval[n].br_blockcount);
+
+		error = xfs_trans_read_buf(mp, sc->tp, mp->m_ddev_targp,
+				d, BTOBB(byte_cnt), 0, &bp, NULL);
+		if (error)
+			goto out;
+		bp->b_ops = &xfs_symlink_buf_ops;
+
+		byte_cnt = XFS_SYMLINK_BUF_SPACE(mp, byte_cnt);
+		if (len < byte_cnt)
+			byte_cnt = len;
+
+		nr = xfs_symlink_hdr_set(mp, ip->i_ino, offset, byte_cnt, bp);
+
+		len -= byte_cnt;
+		offset += byte_cnt;
+
+		xfs_trans_buf_set_type(sc->tp, bp, XFS_BLFT_SYMLINK_BUF);
+		xfs_trans_log_buf(sc->tp, bp, 0, nr - 1);
+		xfs_trans_brelse(sc->tp, bp);
+	}
+	if (len != 0) {
+		error = -EFSCORRUPTED;
+		goto out;
+	}
+
+	/* Roll transaction, release buffers. */
+	error = xfs_trans_roll(&sc->tp, ip);
+	if (error)
+		goto out;
+
+	/* Size set correctly? */
+	len = i_size_read(VFS_I(ip));
+	xfs_iunlock(ip, XFS_ILOCK_EXCL);
+	error = xfs_readlink(ip, sc->buf);
+	xfs_ilock(ip, XFS_ILOCK_EXCL);
+	if (error)
+		goto out;
+
+	/*
+	 * Figure out the new target length.  We can't handle zero-length
+	 * symlinks, so make sure that we don't write that out.
+	 */
+	newlen = strnlen(sc->buf, MAXPATHLEN);
+	if (newlen == 0) {
+		*((char *)sc->buf) = '/';
+		newlen = 1;
+	}
+
+	if (len > newlen)
+		error = xfs_repair_symlink_rewrite(&sc->tp, ip, sc->buf,
+				newlen);
+out:
+	return error;
+}

[PATCH 55/55] xfs: avoid mount-time deadlock in CoW extent recovery

[Darrick J. Wong]

If a malicious user corrupts the refcount btree to cause a cycle between
different levels of the tree, the next mount attempt will deadlock in
the CoW recovery routine.  The scrub code uses the ability to re-grab a
buffer that was previous locked to a transaction to avoid deadlocks, so
do that here too.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/libxfs/xfs_refcount.c |   16 ++++++++++++----
 1 file changed, 12 insertions(+), 4 deletions(-)


diff --git a/fs/xfs/libxfs/xfs_refcount.c b/fs/xfs/libxfs/xfs_refcount.c
index 1c47671..c845253 100644
--- a/fs/xfs/libxfs/xfs_refcount.c
+++ b/fs/xfs/libxfs/xfs_refcount.c
@@ -1646,10 +1646,14 @@ xfs_refcount_recover_cow_leftovers(
 	if (mp->m_sb.sb_agblocks >= XFS_REFC_COW_START)
 		return -EOPNOTSUPP;
 
-	error = xfs_alloc_read_agf(mp, NULL, agno, 0, &agbp);
+	error = xfs_trans_alloc_empty(mp, &tp);
 	if (error)
 		return error;
-	cur = xfs_refcountbt_init_cursor(mp, NULL, agbp, agno, NULL);
+
+	error = xfs_alloc_read_agf(mp, tp, agno, 0, &agbp);
+	if (error)
+		goto out_trans;
+	cur = xfs_refcountbt_init_cursor(mp, tp, agbp, agno, NULL);
 
 	/* Find all the leftover CoW staging extents. */
 	INIT_LIST_HEAD(&debris);
@@ -1662,7 +1666,7 @@ xfs_refcount_recover_cow_leftovers(
 	if (error)
 		goto out_cursor;
 	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
-	xfs_buf_relse(agbp);
+	xfs_trans_cancel(tp);
 
 	/* Now iterate the list to free the leftovers */
 	list_for_each_entry(rr, &debris, rr_list) {
@@ -1705,13 +1709,17 @@ xfs_refcount_recover_cow_leftovers(
 
 out_cursor:
 	xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
-	xfs_buf_relse(agbp);
+	xfs_trans_cancel(tp);
 	goto out_free;
 
 out_defer:
 	xfs_defer_cancel(&dfops);
 	xfs_trans_cancel(tp);
 	goto out_free;
+
+out_trans:
+	xfs_trans_cancel(tp);
+	return error;
 }
 
 /* Is there a record covering a given extent? */

Re: [PATCH v5 00/55] xfs: online scrub/repair support

[Brian Foster]

Trimmed CC to XFS.

On Sat, Jan 21, 2017 at 12:00:15AM -0800, Darrick J. Wong wrote:
> Hi all,
> 
> This is the fifth revision of a patchset that adds to XFS kernel support
> for online metadata scrubbing and repair.  There aren't any on-disk
> format changes.  Changes since v4 include numerous bug fixes, somewhat
> more aggressive log flushing so that on-disk metadata, and the ability
> to distinguish between metadata that's obviously corrupt and metadata
> that merely fails cross-referencing checks in the status that is sent
> back to userspace.  I have also begun using it to check all my
> development workstations, which has been useful for flushing out more
> bugs.
> 

Hi Darrick,

Sorry I haven't got to looking into this yet.. I have kind of a
logistical suggestion if I may...

Can we reduce and repost this in the smallest possible "mergeable
units?" I ask because, at least for me, this kind of huge patchset tends
to continuously get pushed down my todo list because the size of it
suggests I'm going to need to set aside a decent amount of time to grok
the whole thing, test it, etc.

We obviously lose quite a bit of (already limited) review throughput
(and expertise) without Dave around. I think this would be easier for us
to digest from a review perspective if we could do so in smaller chunks.
For example, and just going by some of the patch titles:

- Some of the patches look like they are standalone bugfixes. If so, a
  collection of those could be put into a single series, reviewed and
  merged probably fairly quickly.
- getfsmap looks like a standalone ioctl()..? That seems like something
  that could also be reviewed and merged incrementally.
- Getting into the scrub stuff, could we separate scrubbing and online
  repair into incremental series?

A nice side effect of that is we don't have to repost the entire thing
if we haven't made progress on the next particular dependency. ;)

That aside, in general I think that the whole patchbomb thing kind of
jams up the review process for the associated patches. IMO, better to
feed the list in digestable chunks such that we can try to make
continuous progress (even if that progress is slow) as opposed to trying
to get the whole thing in at once. I think the latter kind of depends on
having somebody like Dave around who can digest and review the whole
thing much more quickly. Just my .02 though.. thoughts?

Brian

> Online scrub/repair support consists of four major pieces -- first, an
> ioctl that maps physical extents to their owners; second, various
> in-kernel metadata scrubbing ioctls to examine metadata records and
> cross-reference them with other filesystem metadata; third, an in-kernel
> mechanism for rebuilding damaged metadata objects and btrees; and
> fourth, a userspace component to initiate kernel scrubbing, walk all
> inodes and the directory tree, scrub data extents, and ask the kernel to
> repair anything that is broken.
> 
> This new utility, xfs_scrub, is separate from the existing offline
> xfs_repair tool.  Scrub has three main modes of operation -- in its most
> powerful mode, it iterates all XFS metadata and asks the kernel to check
> the metadata and repair it if necessary.  The second most powerful mode
> can use certain VFS methods and XFS ioctls (BULKSTAT, GETBMAP, and
> GETFSMAP) to check as much metadata as it reasonably can from userspace.
> It cannot repair anything.  The least powerful mode uses only VFS
> functions to access as much of the directory/file/xattr graph as
> possible.  It has no mechanism to check internal metadata and also
> cannot repair anything.  This is good enough for scrubbing non-XFS
> filesystems, but the primary goal is first-class XFS support.
> 
> As usual, the first patches in this series are bug fixes for problems
> discovered while running the code through rigorous fuzz testing.
> 
> The next few patches in this series implements the GETFSMAP ioctl that
> maps a device number and physical extent either to filesystem metadata
> or to a range of file blocks.  The initial implementation uses the
> reverse-mapping B+tree to supply the mapping information, however a
> fallback implementation based on the free space btrees is also provided.
> The flexibility of having both implementations is important when it
> comes to the userspace tool -- even without the owner/offset data, we
> still have enough information to set up a read verification.  There's
> also a patch to enable xfs_scrub to query the per-AG block reservations
> so that the summary counters can be sanity-checked.
> 
> The next big chunk of patches implement in-kernel scrubbing.  This is
> implemented as a new ioctl.  Pass in a metadata type and control data
> such as an AG number or inode (when applicable); the kernel will examine
> each record in that metadata structure looking for obvious logical
> errors.  External corruption should be discoverable via the checksum
> embedded in each (v5) filesystem metadata block.  When applicable, the
> metadata record will be cross-referenced with the other metadata
> structures to look for discrepancies.  Should any errors be found, an
> error code is returned to userspace, which in the old days would require
> the administrator to take the filesystem offline and repair it.  I've
> hidden the new online scrubber behind CONFIG_XFS_DEBUG to keep it
> disabled by default.
> 
> Last comes the online *repair* functionality, which largely uses the
> redundancy between the new reverse-mapping feature introduced in 4.8 and
> the existing storage space records (bno, cnt, ino, fino, and bmap) to
> reconstruct primary metadata from the secondary, or secondary metadata
> from the primaries.  That's right, we can regrow (some) of the XFS
> metadata even if parts of the filesystem go bad!  Should the kernel
> succeed, it is not necessary to take the filesystem offline for repair.
> 
> Finally, there's a patch that uses one of the new scrub features to
> prevent mount-time deadlocks if the refcountbt is corrupt.
> 
> If you're going to start using this mess, you probably ought to just
> pull from my github trees.  The kernel patches[1] should apply against
> 4.10-rc4.  xfsprogs[2] and xfstests[3] can be found in their usual
> places.
> 
> The patches have survived all of the new tests in [3] that try to fuzz
> every field in every data structure on disk, which has shaken out
> several bugs in the scrubber and in other parts of XFS.
> 
> This is an extraordinary way to eat your data.  Enjoy! 
> Comments and questions are, as always, welcome.
> 
> --D
> 
> [1] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfs-linux.git/log/?h=djwong-devel
> [2] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfsprogs-dev.git/log/?h=djwong-devel
> [3] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfstests-dev.git/log/?h=djwong-devel
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH v5 00/55] xfs: online scrub/repair support

[Darrick J. Wong]

On Tue, Jan 24, 2017 at 12:08:12PM -0500, Brian Foster wrote:
> Trimmed CC to XFS.
> 
> On Sat, Jan 21, 2017 at 12:00:15AM -0800, Darrick J. Wong wrote:
> > Hi all,
> > 
> > This is the fifth revision of a patchset that adds to XFS kernel support
> > for online metadata scrubbing and repair.  There aren't any on-disk
> > format changes.  Changes since v4 include numerous bug fixes, somewhat
> > more aggressive log flushing so that on-disk metadata, and the ability
> > to distinguish between metadata that's obviously corrupt and metadata
> > that merely fails cross-referencing checks in the status that is sent
> > back to userspace.  I have also begun using it to check all my
> > development workstations, which has been useful for flushing out more
> > bugs.
> > 
> 
> Hi Darrick,
> 
> Sorry I haven't got to looking into this yet.. I have kind of a
> logistical suggestion if I may...
> 
> Can we reduce and repost this in the smallest possible "mergeable
> units?" I ask because, at least for me, this kind of huge patchset tends
> to continuously get pushed down my todo list because the size of it
> suggests I'm going to need to set aside a decent amount of time to grok
> the whole thing, test it, etc.
> 
> We obviously lose quite a bit of (already limited) review throughput
> (and expertise) without Dave around. I think this would be easier for us

Yeah.  I've been reviewing my own patches, but when I encounter things
I simply stuff them into the patches directly.  I'm also fairly sure
that R-v-b'ing my own patches doesn't carry much weight. ;)

> to digest from a review perspective if we could do so in smaller chunks.
> For example, and just going by some of the patch titles:
> 
> - Some of the patches look like they are standalone bugfixes. If so, a
>   collection of those could be put into a single series, reviewed and
>   merged probably fairly quickly.
> - getfsmap looks like a standalone ioctl()..? That seems like something
>   that could also be reviewed and merged incrementally.

Originally the only consumer of getfsmap was the scrub tool itself,
though spaceman is now the second (real) user of it.

(The GET_AG_RESBLKS ioctl retrieves the per-ag reservation counters so
that scrub can compare what the fs reports for block/inode counts
against what it scrubbed, for the purpose of evaluating just how much of
the fs it found.)

> - Getting into the scrub stuff, could we separate scrubbing and online
>   repair into incremental series?

Yes, I could split these into (approximately) these kernel series:

1) The usual random fixes (5 patches)
2) GETFSMAP and GET_AG_RESBLKS (8)
3) Basic scrub (19)
4) Scrub cross-references (9)
5) Repair (13)

Beyond that, there's still:

6) Root btrees in inodes (3)
7) rt reverse-mapping (13)

and for xfsprogs, that translates into (excluding libxfs-apply stuff):

8) The usual random fixes (none at the moment)
9) getfsmap & spaceman (8)
10) scrub (9)
11) rtrmapbt (14)

FWIW patches 1-5 are (1) and patches 6-13 are (2) in this patch series,
if anyone /does/ have review bandwidth for 4.11.  The rest I'll defer
to 4.12 or beyond.

> A nice side effect of that is we don't have to repost the entire thing
> if we haven't made progress on the next particular dependency. ;)
> 
> That aside, in general I think that the whole patchbomb thing kind of
> jams up the review process for the associated patches. IMO, better to
> feed the list in digestable chunks such that we can try to make
> continuous progress (even if that progress is slow) as opposed to trying
> to get the whole thing in at once. I think the latter kind of depends on
> having somebody like Dave around who can digest and review the whole
> thing much more quickly. Just my .02 though.. thoughts?

My reasoning for the patchbombs is that I don't like the idea of sending
out an incomplete subset of a feature or features that don't yet have a
downstream consumer.  I've a couple of worries here -- one is that we
review and merge, say (2), but later on we discover while reviewing (5)
something that really should have gotten changed in (2), but now it's
going to be a PITA to change it.  There's less risk of that since the
scrub ioctl will be hidden to non-developers for a while yet, so at
least we don't need to worry about userspace ABI compatibility.  The
other worry of mine is that we partially merge the kernel scrub (say (2)
without (3)) and then xfs_scrub's test cases start exploding because the
kernel scrubber is still half-brained, and in come a flood of bug
reports.

On the other hand, reviewers are critical which means overwhelming them
is also to be avoided.  I recognize that seeing "[PATCH 77/85]" just
makes the whole process seem all the more overwhelming, so perhaps it's
sufficient just to send separate sets of ~15 or so patches?  I'll also
dedicate more time to reviewing (outside the kernel patches) both to
slow myself down and to increase the supply of review time.

In the longer term, AFAICT there are six or so regulars I see on the
list and/or irc.  I think it'll be difficult to do this but frankly I
think we need to find a way to encourage a few more participants.

--D

> 
> Brian
> 
> > Online scrub/repair support consists of four major pieces -- first, an
> > ioctl that maps physical extents to their owners; second, various
> > in-kernel metadata scrubbing ioctls to examine metadata records and
> > cross-reference them with other filesystem metadata; third, an in-kernel
> > mechanism for rebuilding damaged metadata objects and btrees; and
> > fourth, a userspace component to initiate kernel scrubbing, walk all
> > inodes and the directory tree, scrub data extents, and ask the kernel to
> > repair anything that is broken.
> > 
> > This new utility, xfs_scrub, is separate from the existing offline
> > xfs_repair tool.  Scrub has three main modes of operation -- in its most
> > powerful mode, it iterates all XFS metadata and asks the kernel to check
> > the metadata and repair it if necessary.  The second most powerful mode
> > can use certain VFS methods and XFS ioctls (BULKSTAT, GETBMAP, and
> > GETFSMAP) to check as much metadata as it reasonably can from userspace.
> > It cannot repair anything.  The least powerful mode uses only VFS
> > functions to access as much of the directory/file/xattr graph as
> > possible.  It has no mechanism to check internal metadata and also
> > cannot repair anything.  This is good enough for scrubbing non-XFS
> > filesystems, but the primary goal is first-class XFS support.
> > 
> > As usual, the first patches in this series are bug fixes for problems
> > discovered while running the code through rigorous fuzz testing.
> > 
> > The next few patches in this series implements the GETFSMAP ioctl that
> > maps a device number and physical extent either to filesystem metadata
> > or to a range of file blocks.  The initial implementation uses the
> > reverse-mapping B+tree to supply the mapping information, however a
> > fallback implementation based on the free space btrees is also provided.
> > The flexibility of having both implementations is important when it
> > comes to the userspace tool -- even without the owner/offset data, we
> > still have enough information to set up a read verification.  There's
> > also a patch to enable xfs_scrub to query the per-AG block reservations
> > so that the summary counters can be sanity-checked.
> > 
> > The next big chunk of patches implement in-kernel scrubbing.  This is
> > implemented as a new ioctl.  Pass in a metadata type and control data
> > such as an AG number or inode (when applicable); the kernel will examine
> > each record in that metadata structure looking for obvious logical
> > errors.  External corruption should be discoverable via the checksum
> > embedded in each (v5) filesystem metadata block.  When applicable, the
> > metadata record will be cross-referenced with the other metadata
> > structures to look for discrepancies.  Should any errors be found, an
> > error code is returned to userspace, which in the old days would require
> > the administrator to take the filesystem offline and repair it.  I've
> > hidden the new online scrubber behind CONFIG_XFS_DEBUG to keep it
> > disabled by default.
> > 
> > Last comes the online *repair* functionality, which largely uses the
> > redundancy between the new reverse-mapping feature introduced in 4.8 and
> > the existing storage space records (bno, cnt, ino, fino, and bmap) to
> > reconstruct primary metadata from the secondary, or secondary metadata
> > from the primaries.  That's right, we can regrow (some) of the XFS
> > metadata even if parts of the filesystem go bad!  Should the kernel
> > succeed, it is not necessary to take the filesystem offline for repair.
> > 
> > Finally, there's a patch that uses one of the new scrub features to
> > prevent mount-time deadlocks if the refcountbt is corrupt.
> > 
> > If you're going to start using this mess, you probably ought to just
> > pull from my github trees.  The kernel patches[1] should apply against
> > 4.10-rc4.  xfsprogs[2] and xfstests[3] can be found in their usual
> > places.
> > 
> > The patches have survived all of the new tests in [3] that try to fuzz
> > every field in every data structure on disk, which has shaken out
> > several bugs in the scrubber and in other parts of XFS.
> > 
> > This is an extraordinary way to eat your data.  Enjoy! 
> > Comments and questions are, as always, welcome.
> > 
> > --D
> > 
> > [1] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfs-linux.git/log/?h=djwong-devel
> > [2] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfsprogs-dev.git/log/?h=djwong-devel
> > [3] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfstests-dev.git/log/?h=djwong-devel
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH v5 00/55] xfs: online scrub/repair support

[Brian Foster]

On Tue, Jan 24, 2017 at 11:37:19AM -0800, Darrick J. Wong wrote:
> On Tue, Jan 24, 2017 at 12:08:12PM -0500, Brian Foster wrote:
> > Trimmed CC to XFS.
> > 
> > On Sat, Jan 21, 2017 at 12:00:15AM -0800, Darrick J. Wong wrote:
> > > Hi all,
> > > 
> > > This is the fifth revision of a patchset that adds to XFS kernel support
> > > for online metadata scrubbing and repair.  There aren't any on-disk
> > > format changes.  Changes since v4 include numerous bug fixes, somewhat
> > > more aggressive log flushing so that on-disk metadata, and the ability
> > > to distinguish between metadata that's obviously corrupt and metadata
> > > that merely fails cross-referencing checks in the status that is sent
> > > back to userspace.  I have also begun using it to check all my
> > > development workstations, which has been useful for flushing out more
> > > bugs.
> > > 
> > 
> > Hi Darrick,
> > 
> > Sorry I haven't got to looking into this yet.. I have kind of a
> > logistical suggestion if I may...
> > 
> > Can we reduce and repost this in the smallest possible "mergeable
> > units?" I ask because, at least for me, this kind of huge patchset tends
> > to continuously get pushed down my todo list because the size of it
> > suggests I'm going to need to set aside a decent amount of time to grok
> > the whole thing, test it, etc.
> > 
> > We obviously lose quite a bit of (already limited) review throughput
> > (and expertise) without Dave around. I think this would be easier for us
> 
> Yeah.  I've been reviewing my own patches, but when I encounter things
> I simply stuff them into the patches directly.  I'm also fairly sure
> that R-v-b'ing my own patches doesn't carry much weight. ;)
> 

Heh. :P That's better than nothing I suppose, but yeah, a self r-b is
probably to be avoided. At least I don't recall a point where we had to
resort to that in the recent past (a better question for Dave). On the
flip side, I think it has been naturally taking longer to get things
reviewed and merged irrespective of Dave's vacation.

> > to digest from a review perspective if we could do so in smaller chunks.
> > For example, and just going by some of the patch titles:
> > 
> > - Some of the patches look like they are standalone bugfixes. If so, a
> >   collection of those could be put into a single series, reviewed and
> >   merged probably fairly quickly.
> > - getfsmap looks like a standalone ioctl()..? That seems like something
> >   that could also be reviewed and merged incrementally.
> 
> Originally the only consumer of getfsmap was the scrub tool itself,
> though spaceman is now the second (real) user of it.
> 
> (The GET_AG_RESBLKS ioctl retrieves the per-ag reservation counters so
> that scrub can compare what the fs reports for block/inode counts
> against what it scrubbed, for the purpose of evaluating just how much of
> the fs it found.)
> 
> > - Getting into the scrub stuff, could we separate scrubbing and online
> >   repair into incremental series?
> 
> Yes, I could split these into (approximately) these kernel series:
> 
> 1) The usual random fixes (5 patches)
> 2) GETFSMAP and GET_AG_RESBLKS (8)
> 3) Basic scrub (19)
> 4) Scrub cross-references (9)
> 5) Repair (13)
> 

That looks much more approachable. :)

> Beyond that, there's still:
> 
> 6) Root btrees in inodes (3)
> 7) rt reverse-mapping (13)
> 
> and for xfsprogs, that translates into (excluding libxfs-apply stuff):
> 
> 8) The usual random fixes (none at the moment)
> 9) getfsmap & spaceman (8)
> 10) scrub (9)
> 11) rtrmapbt (14)
> 
> FWIW patches 1-5 are (1) and patches 6-13 are (2) in this patch series,
> if anyone /does/ have review bandwidth for 4.11.  The rest I'll defer
> to 4.12 or beyond.
> 
> > A nice side effect of that is we don't have to repost the entire thing
> > if we haven't made progress on the next particular dependency. ;)
> > 
> > That aside, in general I think that the whole patchbomb thing kind of
> > jams up the review process for the associated patches. IMO, better to
> > feed the list in digestable chunks such that we can try to make
> > continuous progress (even if that progress is slow) as opposed to trying
> > to get the whole thing in at once. I think the latter kind of depends on
> > having somebody like Dave around who can digest and review the whole
> > thing much more quickly. Just my .02 though.. thoughts?
> 
> My reasoning for the patchbombs is that I don't like the idea of sending
> out an incomplete subset of a feature or features that don't yet have a
> downstream consumer.  I've a couple of worries here -- one is that we
> review and merge, say (2), but later on we discover while reviewing (5)
> something that really should have gotten changed in (2), but now it's
> going to be a PITA to change it.  There's less risk of that since the
> scrub ioctl will be hidden to non-developers for a while yet, so at
> least we don't need to worry about userspace ABI compatibility.  The
> other worry of mine is that we partially merge the kernel scrub (say (2)
> without (3)) and then xfs_scrub's test cases start exploding because the
> kernel scrubber is still half-brained, and in come a flood of bug
> reports.
> 

Understood, and fair point. For things like the ioctl(), IMO it's good
enough that we have a clear use case and the userspace code available
for reference and/or testing. Whether it's posted in the same series or
not, we have to review it for generality regardless because it is a
userspace interface (I actually think it's better to make it stand on
its own). The difference is just that the consumer or dependent code can
be sent out separately or just made available in a git tree for the
purposes of reviewing/testing that portion of the functionality.

For other "mid-feature" checkpoints, I think it's a matter of logically
considering a feature in its most basic/minimal form and considering the
follow on bits as enhancements, even if that means the v1 that's been
merged is not yet something useful to an end user. I think what you've
laid out above in terms of 1.) basic informative scrub 2.) enhanced
cross checking support (assuming I understand what that means :P) and
3.) online repair sounds like a perfect application of that iterative
model to me. If we have to change something in 1. because of new
requirements introduced in 3., well then, I don't see how that differs
from anything else we do as software developers, really. :P

> On the other hand, reviewers are critical which means overwhelming them
> is also to be avoided.  I recognize that seeing "[PATCH 77/85]" just
> makes the whole process seem all the more overwhelming, so perhaps it's
> sufficient just to send separate sets of ~15 or so patches?  I'll also
> dedicate more time to reviewing (outside the kernel patches) both to
> slow myself down and to increase the supply of review time.
> 

Right, thanks. I think so, as long as the patch series are logical
checkpoints as opposed to arbitrary split-ups (the latter probably being
confusing and ineffective). E.g., I'm more asking for smaller logical
units than patch counts necessarily.

> In the longer term, AFAICT there are six or so regulars I see on the
> list and/or irc.  I think it'll be difficult to do this but frankly I
> think we need to find a way to encourage a few more participants.
> 

Yep. I don't have a good answer for that one, unfortunately. Perhaps
that's something we all should discuss at LSF.

Brian

> --D
> 
> > 
> > Brian
> > 
> > > Online scrub/repair support consists of four major pieces -- first, an
> > > ioctl that maps physical extents to their owners; second, various
> > > in-kernel metadata scrubbing ioctls to examine metadata records and
> > > cross-reference them with other filesystem metadata; third, an in-kernel
> > > mechanism for rebuilding damaged metadata objects and btrees; and
> > > fourth, a userspace component to initiate kernel scrubbing, walk all
> > > inodes and the directory tree, scrub data extents, and ask the kernel to
> > > repair anything that is broken.
> > > 
> > > This new utility, xfs_scrub, is separate from the existing offline
> > > xfs_repair tool.  Scrub has three main modes of operation -- in its most
> > > powerful mode, it iterates all XFS metadata and asks the kernel to check
> > > the metadata and repair it if necessary.  The second most powerful mode
> > > can use certain VFS methods and XFS ioctls (BULKSTAT, GETBMAP, and
> > > GETFSMAP) to check as much metadata as it reasonably can from userspace.
> > > It cannot repair anything.  The least powerful mode uses only VFS
> > > functions to access as much of the directory/file/xattr graph as
> > > possible.  It has no mechanism to check internal metadata and also
> > > cannot repair anything.  This is good enough for scrubbing non-XFS
> > > filesystems, but the primary goal is first-class XFS support.
> > > 
> > > As usual, the first patches in this series are bug fixes for problems
> > > discovered while running the code through rigorous fuzz testing.
> > > 
> > > The next few patches in this series implements the GETFSMAP ioctl that
> > > maps a device number and physical extent either to filesystem metadata
> > > or to a range of file blocks.  The initial implementation uses the
> > > reverse-mapping B+tree to supply the mapping information, however a
> > > fallback implementation based on the free space btrees is also provided.
> > > The flexibility of having both implementations is important when it
> > > comes to the userspace tool -- even without the owner/offset data, we
> > > still have enough information to set up a read verification.  There's
> > > also a patch to enable xfs_scrub to query the per-AG block reservations
> > > so that the summary counters can be sanity-checked.
> > > 
> > > The next big chunk of patches implement in-kernel scrubbing.  This is
> > > implemented as a new ioctl.  Pass in a metadata type and control data
> > > such as an AG number or inode (when applicable); the kernel will examine
> > > each record in that metadata structure looking for obvious logical
> > > errors.  External corruption should be discoverable via the checksum
> > > embedded in each (v5) filesystem metadata block.  When applicable, the
> > > metadata record will be cross-referenced with the other metadata
> > > structures to look for discrepancies.  Should any errors be found, an
> > > error code is returned to userspace, which in the old days would require
> > > the administrator to take the filesystem offline and repair it.  I've
> > > hidden the new online scrubber behind CONFIG_XFS_DEBUG to keep it
> > > disabled by default.
> > > 
> > > Last comes the online *repair* functionality, which largely uses the
> > > redundancy between the new reverse-mapping feature introduced in 4.8 and
> > > the existing storage space records (bno, cnt, ino, fino, and bmap) to
> > > reconstruct primary metadata from the secondary, or secondary metadata
> > > from the primaries.  That's right, we can regrow (some) of the XFS
> > > metadata even if parts of the filesystem go bad!  Should the kernel
> > > succeed, it is not necessary to take the filesystem offline for repair.
> > > 
> > > Finally, there's a patch that uses one of the new scrub features to
> > > prevent mount-time deadlocks if the refcountbt is corrupt.
> > > 
> > > If you're going to start using this mess, you probably ought to just
> > > pull from my github trees.  The kernel patches[1] should apply against
> > > 4.10-rc4.  xfsprogs[2] and xfstests[3] can be found in their usual
> > > places.
> > > 
> > > The patches have survived all of the new tests in [3] that try to fuzz
> > > every field in every data structure on disk, which has shaken out
> > > several bugs in the scrubber and in other parts of XFS.
> > > 
> > > This is an extraordinary way to eat your data.  Enjoy! 
> > > Comments and questions are, as always, welcome.
> > > 
> > > --D
> > > 
> > > [1] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfs-linux.git/log/?h=djwong-devel
> > > [2] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfsprogs-dev.git/log/?h=djwong-devel
> > > [3] https://git.kernel.org/cgit/linux/kernel/git/djwong/xfstests-dev.git/log/?h=djwong-devel
> > > --
> > > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> > > the body of a message to majordomo@vger.kernel.org
> > > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH v5 00/55] xfs: online scrub/repair support

[Dave Chinner]

On Tue, Jan 24, 2017 at 03:50:23PM -0500, Brian Foster wrote:
> On Tue, Jan 24, 2017 at 11:37:19AM -0800, Darrick J. Wong wrote:
> > On Tue, Jan 24, 2017 at 12:08:12PM -0500, Brian Foster wrote:
> > > Trimmed CC to XFS.
> > > 
> > > On Sat, Jan 21, 2017 at 12:00:15AM -0800, Darrick J. Wong wrote:
> > > > Hi all,
> > > > 
> > > > This is the fifth revision of a patchset that adds to XFS kernel support
> > > > for online metadata scrubbing and repair.  There aren't any on-disk
> > > > format changes.  Changes since v4 include numerous bug fixes, somewhat
> > > > more aggressive log flushing so that on-disk metadata, and the ability
> > > > to distinguish between metadata that's obviously corrupt and metadata
> > > > that merely fails cross-referencing checks in the status that is sent
> > > > back to userspace.  I have also begun using it to check all my
> > > > development workstations, which has been useful for flushing out more
> > > > bugs.
> > > > 
> > > 
> > > Hi Darrick,
> > > 
> > > Sorry I haven't got to looking into this yet.. I have kind of a
> > > logistical suggestion if I may...
> > > 
> > > Can we reduce and repost this in the smallest possible "mergeable
> > > units?" I ask because, at least for me, this kind of huge patchset tends
> > > to continuously get pushed down my todo list because the size of it
> > > suggests I'm going to need to set aside a decent amount of time to grok
> > > the whole thing, test it, etc.
> > > 
> > > We obviously lose quite a bit of (already limited) review throughput
> > > (and expertise) without Dave around. I think this would be easier for us
> > 
> > Yeah.  I've been reviewing my own patches, but when I encounter things
> > I simply stuff them into the patches directly.  I'm also fairly sure
> > that R-v-b'ing my own patches doesn't carry much weight. ;)
> > 
> 
> Heh. :P That's better than nothing I suppose, but yeah, a self r-b is
> probably to be avoided. At least I don't recall a point where we had to
> resort to that in the recent past (a better question for Dave). On the
> flip side, I think it has been naturally taking longer to get things
> reviewed and merged irrespective of Dave's vacation.

When it comes to review, the maintainer does not get a free pass. In
fact, it's even more important that the maintainer's code is
reviewed by someone else as it makes it clear that the
maintainer is not "all-powerful" and does not have privileges that
other developers don't have. i.e. there must be extremely compelling
reasons for a maintainer to commit their own code to the kernel tree
without peer review.

> > > to digest from a review perspective if we could do so in smaller chunks.
> > > For example, and just going by some of the patch titles:
> > > 
> > > - Some of the patches look like they are standalone bugfixes. If so, a
> > >   collection of those could be put into a single series, reviewed and
> > >   merged probably fairly quickly.
> > > - getfsmap looks like a standalone ioctl()..? That seems like something
> > >   that could also be reviewed and merged incrementally.
> > 
> > Originally the only consumer of getfsmap was the scrub tool itself,
> > though spaceman is now the second (real) user of it.
> > 
> > (The GET_AG_RESBLKS ioctl retrieves the per-ag reservation counters so
> > that scrub can compare what the fs reports for block/inode counts
> > against what it scrubbed, for the purpose of evaluating just how much of
> > the fs it found.)
> > 
> > > - Getting into the scrub stuff, could we separate scrubbing and online
> > >   repair into incremental series?
> > 
> > Yes, I could split these into (approximately) these kernel series:
> > 
> > 1) The usual random fixes (5 patches)
> > 2) GETFSMAP and GET_AG_RESBLKS (8)
> > 3) Basic scrub (19)
> > 4) Scrub cross-references (9)
> > 5) Repair (13)
> > 
> 
> That looks much more approachable. :)

This is generally how I've approached review of Darrick's patchbombs
- review and merge smaller, self-contained chunks one at a time. I
guess over the years I've gotten used to handling big patch sets
like this, so I have never found it a problem to break them into
manageable chunks myself... :P

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com