* [PATCH 00/25, V3] xfs: metadata buffer verifiers
@ 2012-10-25 6:33 Dave Chinner
2012-10-25 6:33 ` [PATCH 01/25] xfs: growfs: don't read garbage for new secondary superblocks Dave Chinner
` (24 more replies)
0 siblings, 25 replies; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
Hi folks,
Third version of the buffer verifier series. The read verifier
infrastructure is described here:
http://oss.sgi.com/archives/xfs/2012-10/msg00146.html
This version converts to a buffer operations structure rather than
specific write/iodone callback installations and adds log recovery
write verifiers. There are also several bugs fixed and review
comments address.
This series is essentially now functionally complete, so there is
nothing really left to add to this except for addressing review
comments and bug fixing. Comments welcome. ;)
FYI, I do have more changes lined up for the 3.8 window, but I will
be posting them as separate patches on top of this series and not as
part of it.
Cheers,
Dave.
--
Changes in version 3:
- update agfl verfier commit to mention debug checks are being done
unconditionally now.
- fixed agfl verifier null point crash when invalid block numbers
are found
- ifdef'd out agfl verifier checks as they are not reliable because
mkfs does not initialise the full AGFL to known values.
- fixed quiet mount flag handling for superblock verification.
- directorry -> directory
- convert to struct buf_ops method of attaching verifiers to the
buffer. This provides a much cleaner abstraction and simpler
future expansion of operations on the buffer. It removes a great
deal of code that is repeated through all the verifiers, too, by
separating them from buffer IO completion processing.
- add initial support for log write verifiers
Log write verifiers are, in general, identical to the existing
verifiers. There are only a small number of modifications
necessary, mainly due to log recovery occurring before certain
in-memory structures are initialised (e.g. the struct xfs_perag).
Write verifiers that need different checks during recovery do so
via detection of the XLOG_ACTIVE_RECOVERY flag on the log.
Log recovery does not do read verification of the buffers at this
point in time, mainly due to the fact we don't know what the
contents of the buffer is before we read it - the buffer logging
is generic and content unaware. However, almost all metadata has
magic numbers in it, so after the changes have been replayed into
the buffer we can snoop the magic number out of the buffer and
attach the appropriate verifier before it is written back. Hence
we should catch gross corruptions introduced by recovery errors.
Changes in Version 2:
- fixed use of xfs_dir2_db_t instead of xfs_dablk_t in directory and
attr read functions (found when testing xfstests --large-fs on a
500TB fs and attribute block numbers went beyond 32 bits). This
mistake was copy-n-pasted several times.
- fixed use of "int map_type" instead of "xfs_daddr_t mappedbno" in
directory and attr read functions.
- fixed incorrect logic in xfs_dir2_block_verify where a failed
block check would not clear the block_ok flag correctly
- invalidate allocbt->freelist buffers so they don't get written
after being freed and while still on the freelist
- added initial suppor for write verifiers.
Write verifiers are similar to read verifiers, the are simply
called just prior to issuing the IO on the buffer. The buffer is
locked at this point, so we are guaranteed an unchanging buffer
to work from.
The initial write verifiers are simply the same as the read
verifiers, except they don't have the ioend processing in them. A
failure of the write verifier will cause the filesystem to shut
down as writing invalid metadata to disk is a bad thing. The write
verifier for the alloc btree blocks was what discovered the
writing of freed allocbt blocks to disk from the free list.
Eventually, the metadata CRC will be calculated in the write
verifier after validating that the buffer contents are valid.
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* [PATCH 01/25] xfs: growfs: don't read garbage for new secondary superblocks
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
@ 2012-10-25 6:33 ` Dave Chinner
2012-10-30 0:17 ` Phil White
2012-10-25 6:33 ` [PATCH 02/25] xfs: invalidate allocbt blocks moved to the free list Dave Chinner
` (23 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
When updating new secondary superblocks in a growfs operation, the
sueprblock buffer is read from the newly grown region of the
underlying device. This is not guaranteed to be zero, so violates
the underlying assumption that the unused parts of superblocks are
zero filled. Get a new buffer for these secondary superblocks to
ensure that the unused regions are zero filled correctly.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
fs/xfs/xfs_fsops.c | 21 +++++++++++++++++++--
1 file changed, 19 insertions(+), 2 deletions(-)
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index c25b094..4beaede 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -399,9 +399,26 @@ xfs_growfs_data_private(
/* update secondary superblocks. */
for (agno = 1; agno < nagcount; agno++) {
- error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
+ error = 0;
+ /*
+ * new secondary superblocks need to be zeroed, not read from
+ * disk as the contents of the new area we are growing into is
+ * completely unknown.
+ */
+ if (agno < oagcount) {
+ error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
XFS_FSS_TO_BB(mp, 1), 0, &bp);
+ } else {
+ bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
+ XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
+ XFS_FSS_TO_BB(mp, 1), 0);
+ if (bp)
+ xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
+ else
+ error = ENOMEM;
+ }
+
if (error) {
xfs_warn(mp,
"error %d reading secondary superblock for ag %d",
@@ -423,7 +440,7 @@ xfs_growfs_data_private(
break; /* no point in continuing */
}
}
- return 0;
+ return error;
error0:
xfs_trans_cancel(tp, XFS_TRANS_ABORT);
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 02/25] xfs: invalidate allocbt blocks moved to the free list
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
2012-10-25 6:33 ` [PATCH 01/25] xfs: growfs: don't read garbage for new secondary superblocks Dave Chinner
@ 2012-10-25 6:33 ` Dave Chinner
2012-10-26 8:47 ` Christoph Hellwig
2012-10-30 0:22 ` Phil White
2012-10-25 6:33 ` [PATCH 03/25] xfs: make buffer read verication an IO completion function Dave Chinner
` (22 subsequent siblings)
24 siblings, 2 replies; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
When we free a block from the alloc btree tree, we move it to the
freelist held in the AGFL and mark it busy in the busy extent tree.
This typically happens when we merge btree blocks.
Once the transaction is committed and checkpointed, the block can
remain on the free list for an indefinite amount of time. Now, this
isn't the end of the world at this point - if the free list is
shortened, the buffer is invalidated in the transaction that moves
it back to free space. If the buffer is allocated as metadata from
the free list, then all the modifications getted logged, and we have
no issues, either. And if it gets allocated as userdata direct from
the freelist, it gets invalidated and so will never get written.
However, during the time it sits on the free list, pressure on the
log can cause the AIL to be pushed and the buffer that covers the
block gets pushed for write. IOWs, we end up writing a freed
metadata block to disk. Again, this isn't the end of the world
because we know from the above we are only writing to free space.
The problem, however, is for validation callbacks. If the block was
on old btree root block, then the level of the block is going to be
higher than the current tree root, and so will fail validation.
There may be other inconsistencies in the block as well, and
currently we don't care because the block is in free space. Shutting
down the filesystem because a freed block doesn't pass write
validation, OTOH, is rather unfriendly.
So, make sure we always invalidate buffers as they move from the
free space trees to the free list so that we guarantee they never
get written to disk while on the free list.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_alloc_btree.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/fs/xfs/xfs_alloc_btree.c b/fs/xfs/xfs_alloc_btree.c
index f1647ca..f7876c6 100644
--- a/fs/xfs/xfs_alloc_btree.c
+++ b/fs/xfs/xfs_alloc_btree.c
@@ -121,6 +121,8 @@ xfs_allocbt_free_block(
xfs_extent_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1,
XFS_EXTENT_BUSY_SKIP_DISCARD);
xfs_trans_agbtree_delta(cur->bc_tp, -1);
+
+ xfs_trans_binval(cur->bc_tp, bp);
return 0;
}
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 03/25] xfs: make buffer read verication an IO completion function
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
2012-10-25 6:33 ` [PATCH 01/25] xfs: growfs: don't read garbage for new secondary superblocks Dave Chinner
2012-10-25 6:33 ` [PATCH 02/25] xfs: invalidate allocbt blocks moved to the free list Dave Chinner
@ 2012-10-25 6:33 ` Dave Chinner
2012-10-30 0:29 ` Phil White
2012-10-25 6:33 ` [PATCH 04/25] xfs: uncached buffer reads need to return an error Dave Chinner
` (21 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Add a verifier function callback capability to the buffer read
interfaces. This will be used by the callers to supply a function
that verifies the contents of the buffer when it is read from disk.
This patch does not provide callback functions, but simply modifies
the interfaces to allow them to be called.
The reason for adding this to the read interfaces is that it is very
difficult to tell fom the outside is a buffer was just read from
disk or whether we just pulled it out of cache. Supplying a callbck
allows the buffer cache to use it's internal knowledge of the buffer
to execute it only when the buffer is read from disk.
It is intended that the verifier functions will mark the buffer with
an EFSCORRUPTED error when verification fails. This allows the
reading context to distinguish a verification error from an IO
error, and potentially take further actions on the buffer (e.g.
attempt repair) based on the error reported.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
fs/xfs/xfs_alloc.c | 4 ++--
fs/xfs/xfs_attr.c | 2 +-
fs/xfs/xfs_btree.c | 21 ++++++++++++---------
fs/xfs/xfs_buf.c | 13 +++++++++----
fs/xfs/xfs_buf.h | 20 ++++++++++++--------
fs/xfs/xfs_da_btree.c | 4 ++--
fs/xfs/xfs_dir2_leaf.c | 2 +-
fs/xfs/xfs_dquot.c | 4 ++--
fs/xfs/xfs_fsops.c | 4 ++--
fs/xfs/xfs_ialloc.c | 2 +-
fs/xfs/xfs_inode.c | 2 +-
fs/xfs/xfs_log.c | 3 +--
fs/xfs/xfs_log_recover.c | 8 +++++---
fs/xfs/xfs_mount.c | 6 +++---
fs/xfs/xfs_qm.c | 5 +++--
fs/xfs/xfs_rtalloc.c | 6 +++---
fs/xfs/xfs_trans.h | 19 ++++++++-----------
fs/xfs/xfs_trans_buf.c | 9 ++++++---
fs/xfs/xfs_vnodeops.c | 2 +-
19 files changed, 75 insertions(+), 61 deletions(-)
diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
index 335206a..21c3db0 100644
--- a/fs/xfs/xfs_alloc.c
+++ b/fs/xfs/xfs_alloc.c
@@ -447,7 +447,7 @@ xfs_alloc_read_agfl(
error = xfs_trans_read_buf(
mp, tp, mp->m_ddev_targp,
XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
- XFS_FSS_TO_BB(mp, 1), 0, &bp);
+ XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
if (error)
return error;
ASSERT(!xfs_buf_geterror(bp));
@@ -2110,7 +2110,7 @@ xfs_read_agf(
error = xfs_trans_read_buf(
mp, tp, mp->m_ddev_targp,
XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
- XFS_FSS_TO_BB(mp, 1), flags, bpp);
+ XFS_FSS_TO_BB(mp, 1), flags, bpp, NULL);
if (error)
return error;
if (!*bpp)
diff --git a/fs/xfs/xfs_attr.c b/fs/xfs/xfs_attr.c
index 0ca1f0b..ebacb8d 100644
--- a/fs/xfs/xfs_attr.c
+++ b/fs/xfs/xfs_attr.c
@@ -1980,7 +1980,7 @@ xfs_attr_rmtval_get(xfs_da_args_t *args)
dblkno = XFS_FSB_TO_DADDR(mp, map[i].br_startblock);
blkcnt = XFS_FSB_TO_BB(mp, map[i].br_blockcount);
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
- dblkno, blkcnt, 0, &bp);
+ dblkno, blkcnt, 0, &bp, NULL);
if (error)
return(error);
diff --git a/fs/xfs/xfs_btree.c b/fs/xfs/xfs_btree.c
index e53e317..1937c9b 100644
--- a/fs/xfs/xfs_btree.c
+++ b/fs/xfs/xfs_btree.c
@@ -266,9 +266,12 @@ xfs_btree_dup_cursor(
for (i = 0; i < new->bc_nlevels; i++) {
new->bc_ptrs[i] = cur->bc_ptrs[i];
new->bc_ra[i] = cur->bc_ra[i];
- if ((bp = cur->bc_bufs[i])) {
- if ((error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
- XFS_BUF_ADDR(bp), mp->m_bsize, 0, &bp))) {
+ bp = cur->bc_bufs[i];
+ if (bp) {
+ error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
+ XFS_BUF_ADDR(bp), mp->m_bsize,
+ 0, &bp, NULL);
+ if (error) {
xfs_btree_del_cursor(new, error);
*ncur = NULL;
return error;
@@ -624,10 +627,10 @@ xfs_btree_read_bufl(
ASSERT(fsbno != NULLFSBLOCK);
d = XFS_FSB_TO_DADDR(mp, fsbno);
- if ((error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, d,
- mp->m_bsize, lock, &bp))) {
+ error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, d,
+ mp->m_bsize, lock, &bp, NULL);
+ if (error)
return error;
- }
ASSERT(!xfs_buf_geterror(bp));
if (bp)
xfs_buf_set_ref(bp, refval);
@@ -650,7 +653,7 @@ xfs_btree_reada_bufl(
ASSERT(fsbno != NULLFSBLOCK);
d = XFS_FSB_TO_DADDR(mp, fsbno);
- xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count);
+ xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, NULL);
}
/*
@@ -670,7 +673,7 @@ xfs_btree_reada_bufs(
ASSERT(agno != NULLAGNUMBER);
ASSERT(agbno != NULLAGBLOCK);
d = XFS_AGB_TO_DADDR(mp, agno, agbno);
- xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count);
+ xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, NULL);
}
STATIC int
@@ -998,7 +1001,7 @@ xfs_btree_read_buf_block(
d = xfs_btree_ptr_to_daddr(cur, ptr);
error = xfs_trans_read_buf(mp, cur->bc_tp, mp->m_ddev_targp, d,
- mp->m_bsize, flags, bpp);
+ mp->m_bsize, flags, bpp, NULL);
if (error)
return error;
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index 933b793..7cab1b3 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -654,7 +654,8 @@ xfs_buf_read_map(
struct xfs_buftarg *target,
struct xfs_buf_map *map,
int nmaps,
- xfs_buf_flags_t flags)
+ xfs_buf_flags_t flags,
+ xfs_buf_iodone_t verify)
{
struct xfs_buf *bp;
@@ -666,6 +667,7 @@ xfs_buf_read_map(
if (!XFS_BUF_ISDONE(bp)) {
XFS_STATS_INC(xb_get_read);
+ bp->b_iodone = verify;
_xfs_buf_read(bp, flags);
} else if (flags & XBF_ASYNC) {
/*
@@ -691,13 +693,14 @@ void
xfs_buf_readahead_map(
struct xfs_buftarg *target,
struct xfs_buf_map *map,
- int nmaps)
+ int nmaps,
+ xfs_buf_iodone_t verify)
{
if (bdi_read_congested(target->bt_bdi))
return;
xfs_buf_read_map(target, map, nmaps,
- XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD);
+ XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD, verify);
}
/*
@@ -709,7 +712,8 @@ xfs_buf_read_uncached(
struct xfs_buftarg *target,
xfs_daddr_t daddr,
size_t numblks,
- int flags)
+ int flags,
+ xfs_buf_iodone_t verify)
{
xfs_buf_t *bp;
int error;
@@ -723,6 +727,7 @@ xfs_buf_read_uncached(
bp->b_bn = daddr;
bp->b_maps[0].bm_bn = daddr;
bp->b_flags |= XBF_READ;
+ bp->b_iodone = verify;
xfsbdstrat(target->bt_mount, bp);
error = xfs_buf_iowait(bp);
diff --git a/fs/xfs/xfs_buf.h b/fs/xfs/xfs_buf.h
index 7c0b6a0..677b1dc 100644
--- a/fs/xfs/xfs_buf.h
+++ b/fs/xfs/xfs_buf.h
@@ -100,6 +100,7 @@ typedef struct xfs_buftarg {
struct xfs_buf;
typedef void (*xfs_buf_iodone_t)(struct xfs_buf *);
+
#define XB_PAGES 2
struct xfs_buf_map {
@@ -159,7 +160,6 @@ typedef struct xfs_buf {
#endif
} xfs_buf_t;
-
/* Finding and Reading Buffers */
struct xfs_buf *_xfs_buf_find(struct xfs_buftarg *target,
struct xfs_buf_map *map, int nmaps,
@@ -196,9 +196,10 @@ struct xfs_buf *xfs_buf_get_map(struct xfs_buftarg *target,
xfs_buf_flags_t flags);
struct xfs_buf *xfs_buf_read_map(struct xfs_buftarg *target,
struct xfs_buf_map *map, int nmaps,
- xfs_buf_flags_t flags);
+ xfs_buf_flags_t flags, xfs_buf_iodone_t verify);
void xfs_buf_readahead_map(struct xfs_buftarg *target,
- struct xfs_buf_map *map, int nmaps);
+ struct xfs_buf_map *map, int nmaps,
+ xfs_buf_iodone_t verify);
static inline struct xfs_buf *
xfs_buf_get(
@@ -216,20 +217,22 @@ xfs_buf_read(
struct xfs_buftarg *target,
xfs_daddr_t blkno,
size_t numblks,
- xfs_buf_flags_t flags)
+ xfs_buf_flags_t flags,
+ xfs_buf_iodone_t verify)
{
DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
- return xfs_buf_read_map(target, &map, 1, flags);
+ return xfs_buf_read_map(target, &map, 1, flags, verify);
}
static inline void
xfs_buf_readahead(
struct xfs_buftarg *target,
xfs_daddr_t blkno,
- size_t numblks)
+ size_t numblks,
+ xfs_buf_iodone_t verify)
{
DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
- return xfs_buf_readahead_map(target, &map, 1);
+ return xfs_buf_readahead_map(target, &map, 1, verify);
}
struct xfs_buf *xfs_buf_get_empty(struct xfs_buftarg *target, size_t numblks);
@@ -239,7 +242,8 @@ int xfs_buf_associate_memory(struct xfs_buf *bp, void *mem, size_t length);
struct xfs_buf *xfs_buf_get_uncached(struct xfs_buftarg *target, size_t numblks,
int flags);
struct xfs_buf *xfs_buf_read_uncached(struct xfs_buftarg *target,
- xfs_daddr_t daddr, size_t numblks, int flags);
+ xfs_daddr_t daddr, size_t numblks, int flags,
+ xfs_buf_iodone_t verify);
void xfs_buf_hold(struct xfs_buf *bp);
/* Releasing Buffers */
diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
index 7bfb7dd..41d8764 100644
--- a/fs/xfs/xfs_da_btree.c
+++ b/fs/xfs/xfs_da_btree.c
@@ -2155,7 +2155,7 @@ xfs_da_read_buf(
error = xfs_trans_read_buf_map(dp->i_mount, trans,
dp->i_mount->m_ddev_targp,
- mapp, nmap, 0, &bp);
+ mapp, nmap, 0, &bp, NULL);
if (error)
goto out_free;
@@ -2231,7 +2231,7 @@ xfs_da_reada_buf(
}
mappedbno = mapp[0].bm_bn;
- xfs_buf_readahead_map(dp->i_mount->m_ddev_targp, mapp, nmap);
+ xfs_buf_readahead_map(dp->i_mount->m_ddev_targp, mapp, nmap, NULL);
out_free:
if (mapp != &map)
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 0b29625..bac8698 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -926,7 +926,7 @@ xfs_dir2_leaf_readbuf(
XFS_FSB_TO_DADDR(mp,
map[mip->ra_index].br_startblock +
mip->ra_offset),
- (int)BTOBB(mp->m_dirblksize));
+ (int)BTOBB(mp->m_dirblksize), NULL);
mip->ra_current = i;
}
diff --git a/fs/xfs/xfs_dquot.c b/fs/xfs/xfs_dquot.c
index bf27fcc..e95f800 100644
--- a/fs/xfs/xfs_dquot.c
+++ b/fs/xfs/xfs_dquot.c
@@ -439,7 +439,7 @@ xfs_qm_dqtobp(
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
dqp->q_blkno,
mp->m_quotainfo->qi_dqchunklen,
- 0, &bp);
+ 0, &bp, NULL);
if (error || !bp)
return XFS_ERROR(error);
}
@@ -920,7 +920,7 @@ xfs_qm_dqflush(
* Get the buffer containing the on-disk dquot
*/
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno,
- mp->m_quotainfo->qi_dqchunklen, 0, &bp);
+ mp->m_quotainfo->qi_dqchunklen, 0, &bp, NULL);
if (error)
goto out_unlock;
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index 4beaede..917e121 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -146,7 +146,7 @@ xfs_growfs_data_private(
dpct = pct - mp->m_sb.sb_imax_pct;
bp = xfs_buf_read_uncached(mp->m_ddev_targp,
XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
- XFS_FSS_TO_BB(mp, 1), 0);
+ XFS_FSS_TO_BB(mp, 1), 0, NULL);
if (!bp)
return EIO;
xfs_buf_relse(bp);
@@ -408,7 +408,7 @@ xfs_growfs_data_private(
if (agno < oagcount) {
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
- XFS_FSS_TO_BB(mp, 1), 0, &bp);
+ XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
} else {
bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
diff --git a/fs/xfs/xfs_ialloc.c b/fs/xfs/xfs_ialloc.c
index c5c4ef4..7c944e1 100644
--- a/fs/xfs/xfs_ialloc.c
+++ b/fs/xfs/xfs_ialloc.c
@@ -1490,7 +1490,7 @@ xfs_read_agi(
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
- XFS_FSS_TO_BB(mp, 1), 0, bpp);
+ XFS_FSS_TO_BB(mp, 1), 0, bpp, NULL);
if (error)
return error;
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index bba8f37..0b03578 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -408,7 +408,7 @@ xfs_imap_to_bp(
buf_flags |= XBF_UNMAPPED;
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
- (int)imap->im_len, buf_flags, &bp);
+ (int)imap->im_len, buf_flags, &bp, NULL);
if (error) {
if (error != EAGAIN) {
xfs_warn(mp,
diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c
index 46b6986..1d6d2ee 100644
--- a/fs/xfs/xfs_log.c
+++ b/fs/xfs/xfs_log.c
@@ -1129,8 +1129,7 @@ xlog_iodone(xfs_buf_t *bp)
* with it being freed after writing the unmount record to the
* log.
*/
-
-} /* xlog_iodone */
+}
/*
* Return size of each in-core log record buffer.
diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
index 651c988..757688a 100644
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -2144,7 +2144,7 @@ xlog_recover_buffer_pass2(
buf_flags |= XBF_UNMAPPED;
bp = xfs_buf_read(mp->m_ddev_targp, buf_f->blf_blkno, buf_f->blf_len,
- buf_flags);
+ buf_flags, NULL);
if (!bp)
return XFS_ERROR(ENOMEM);
error = bp->b_error;
@@ -2237,7 +2237,8 @@ xlog_recover_inode_pass2(
}
trace_xfs_log_recover_inode_recover(log, in_f);
- bp = xfs_buf_read(mp->m_ddev_targp, in_f->ilf_blkno, in_f->ilf_len, 0);
+ bp = xfs_buf_read(mp->m_ddev_targp, in_f->ilf_blkno, in_f->ilf_len, 0,
+ NULL);
if (!bp) {
error = ENOMEM;
goto error;
@@ -2548,7 +2549,8 @@ xlog_recover_dquot_pass2(
ASSERT(dq_f->qlf_len == 1);
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dq_f->qlf_blkno,
- XFS_FSB_TO_BB(mp, dq_f->qlf_len), 0, &bp);
+ XFS_FSB_TO_BB(mp, dq_f->qlf_len), 0, &bp,
+ NULL);
if (error)
return error;
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index 6f1c997..d39ad72 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -652,7 +652,7 @@ xfs_readsb(xfs_mount_t *mp, int flags)
reread:
bp = xfs_buf_read_uncached(mp->m_ddev_targp, XFS_SB_DADDR,
- BTOBB(sector_size), 0);
+ BTOBB(sector_size), 0, NULL);
if (!bp) {
if (loud)
xfs_warn(mp, "SB buffer read failed");
@@ -1002,7 +1002,7 @@ xfs_check_sizes(xfs_mount_t *mp)
}
bp = xfs_buf_read_uncached(mp->m_ddev_targp,
d - XFS_FSS_TO_BB(mp, 1),
- XFS_FSS_TO_BB(mp, 1), 0);
+ XFS_FSS_TO_BB(mp, 1), 0, NULL);
if (!bp) {
xfs_warn(mp, "last sector read failed");
return EIO;
@@ -1017,7 +1017,7 @@ xfs_check_sizes(xfs_mount_t *mp)
}
bp = xfs_buf_read_uncached(mp->m_logdev_targp,
d - XFS_FSB_TO_BB(mp, 1),
- XFS_FSB_TO_BB(mp, 1), 0);
+ XFS_FSB_TO_BB(mp, 1), 0, NULL);
if (!bp) {
xfs_warn(mp, "log device read failed");
return EIO;
diff --git a/fs/xfs/xfs_qm.c b/fs/xfs/xfs_qm.c
index 48c750b..688f608 100644
--- a/fs/xfs/xfs_qm.c
+++ b/fs/xfs/xfs_qm.c
@@ -892,7 +892,7 @@ xfs_qm_dqiter_bufs(
while (blkcnt--) {
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
XFS_FSB_TO_DADDR(mp, bno),
- mp->m_quotainfo->qi_dqchunklen, 0, &bp);
+ mp->m_quotainfo->qi_dqchunklen, 0, &bp, NULL);
if (error)
break;
@@ -979,7 +979,8 @@ xfs_qm_dqiterate(
while (rablkcnt--) {
xfs_buf_readahead(mp->m_ddev_targp,
XFS_FSB_TO_DADDR(mp, rablkno),
- mp->m_quotainfo->qi_dqchunklen);
+ mp->m_quotainfo->qi_dqchunklen,
+ NULL);
rablkno++;
}
}
diff --git a/fs/xfs/xfs_rtalloc.c b/fs/xfs/xfs_rtalloc.c
index a69e0b4..b271ed9 100644
--- a/fs/xfs/xfs_rtalloc.c
+++ b/fs/xfs/xfs_rtalloc.c
@@ -870,7 +870,7 @@ xfs_rtbuf_get(
ASSERT(map.br_startblock != NULLFSBLOCK);
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
XFS_FSB_TO_DADDR(mp, map.br_startblock),
- mp->m_bsize, 0, &bp);
+ mp->m_bsize, 0, &bp, NULL);
if (error)
return error;
ASSERT(!xfs_buf_geterror(bp));
@@ -1873,7 +1873,7 @@ xfs_growfs_rt(
*/
bp = xfs_buf_read_uncached(mp->m_rtdev_targp,
XFS_FSB_TO_BB(mp, nrblocks - 1),
- XFS_FSB_TO_BB(mp, 1), 0);
+ XFS_FSB_TO_BB(mp, 1), 0, NULL);
if (!bp)
return EIO;
xfs_buf_relse(bp);
@@ -2220,7 +2220,7 @@ xfs_rtmount_init(
}
bp = xfs_buf_read_uncached(mp->m_rtdev_targp,
d - XFS_FSB_TO_BB(mp, 1),
- XFS_FSB_TO_BB(mp, 1), 0);
+ XFS_FSB_TO_BB(mp, 1), 0, NULL);
if (!bp) {
xfs_warn(mp, "realtime device size check failed");
return EIO;
diff --git a/fs/xfs/xfs_trans.h b/fs/xfs/xfs_trans.h
index db05654..f02d402 100644
--- a/fs/xfs/xfs_trans.h
+++ b/fs/xfs/xfs_trans.h
@@ -464,10 +464,7 @@ xfs_trans_get_buf(
int numblks,
uint flags)
{
- struct xfs_buf_map map = {
- .bm_bn = blkno,
- .bm_len = numblks,
- };
+ DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
return xfs_trans_get_buf_map(tp, target, &map, 1, flags);
}
@@ -476,7 +473,8 @@ int xfs_trans_read_buf_map(struct xfs_mount *mp,
struct xfs_buftarg *target,
struct xfs_buf_map *map, int nmaps,
xfs_buf_flags_t flags,
- struct xfs_buf **bpp);
+ struct xfs_buf **bpp,
+ xfs_buf_iodone_t verify);
static inline int
xfs_trans_read_buf(
@@ -486,13 +484,12 @@ xfs_trans_read_buf(
xfs_daddr_t blkno,
int numblks,
xfs_buf_flags_t flags,
- struct xfs_buf **bpp)
+ struct xfs_buf **bpp,
+ xfs_buf_iodone_t verify)
{
- struct xfs_buf_map map = {
- .bm_bn = blkno,
- .bm_len = numblks,
- };
- return xfs_trans_read_buf_map(mp, tp, target, &map, 1, flags, bpp);
+ DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
+ return xfs_trans_read_buf_map(mp, tp, target, &map, 1,
+ flags, bpp, verify);
}
struct xfs_buf *xfs_trans_getsb(xfs_trans_t *, struct xfs_mount *, int);
diff --git a/fs/xfs/xfs_trans_buf.c b/fs/xfs/xfs_trans_buf.c
index 6311b99..9776282 100644
--- a/fs/xfs/xfs_trans_buf.c
+++ b/fs/xfs/xfs_trans_buf.c
@@ -257,7 +257,8 @@ xfs_trans_read_buf_map(
struct xfs_buf_map *map,
int nmaps,
xfs_buf_flags_t flags,
- struct xfs_buf **bpp)
+ struct xfs_buf **bpp,
+ xfs_buf_iodone_t verify)
{
xfs_buf_t *bp;
xfs_buf_log_item_t *bip;
@@ -265,7 +266,7 @@ xfs_trans_read_buf_map(
*bpp = NULL;
if (!tp) {
- bp = xfs_buf_read_map(target, map, nmaps, flags);
+ bp = xfs_buf_read_map(target, map, nmaps, flags, verify);
if (!bp)
return (flags & XBF_TRYLOCK) ?
EAGAIN : XFS_ERROR(ENOMEM);
@@ -312,7 +313,9 @@ xfs_trans_read_buf_map(
if (!(XFS_BUF_ISDONE(bp))) {
trace_xfs_trans_read_buf_io(bp, _RET_IP_);
ASSERT(!XFS_BUF_ISASYNC(bp));
+ ASSERT(bp->b_iodone == NULL);
XFS_BUF_READ(bp);
+ bp->b_iodone = verify;
xfsbdstrat(tp->t_mountp, bp);
error = xfs_buf_iowait(bp);
if (error) {
@@ -349,7 +352,7 @@ xfs_trans_read_buf_map(
return 0;
}
- bp = xfs_buf_read_map(target, map, nmaps, flags);
+ bp = xfs_buf_read_map(target, map, nmaps, flags, verify);
if (bp == NULL) {
*bpp = NULL;
return (flags & XBF_TRYLOCK) ?
diff --git a/fs/xfs/xfs_vnodeops.c b/fs/xfs/xfs_vnodeops.c
index 2ee1f49..f409fda 100644
--- a/fs/xfs/xfs_vnodeops.c
+++ b/fs/xfs/xfs_vnodeops.c
@@ -80,7 +80,7 @@ xfs_readlink_bmap(
d = XFS_FSB_TO_DADDR(mp, mval[n].br_startblock);
byte_cnt = XFS_FSB_TO_B(mp, mval[n].br_blockcount);
- bp = xfs_buf_read(mp->m_ddev_targp, d, BTOBB(byte_cnt), 0);
+ bp = xfs_buf_read(mp->m_ddev_targp, d, BTOBB(byte_cnt), 0, NULL);
if (!bp)
return XFS_ERROR(ENOMEM);
error = bp->b_error;
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 04/25] xfs: uncached buffer reads need to return an error
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (2 preceding siblings ...)
2012-10-25 6:33 ` [PATCH 03/25] xfs: make buffer read verication an IO completion function Dave Chinner
@ 2012-10-25 6:33 ` Dave Chinner
2012-10-26 8:48 ` Christoph Hellwig
2012-10-30 0:36 ` Phil White
2012-10-25 6:33 ` [PATCH 05/25] xfs: verify superblocks as they are read from disk Dave Chinner
` (20 subsequent siblings)
24 siblings, 2 replies; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
With verification being done as an IO completion callback, different
errors can be returned from a read. Uncached reads only return a
buffer or NULL on failure, which means the verification error cannot
be returned to the caller.
Split the error handling for these reads into two - a failure to get
a buffer will still return NULL, but a read error will return a
referenced buffer with b_error set rather than NULL. The caller is
responsible for checking the error state of the buffer returned.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_buf.c | 9 ++-------
fs/xfs/xfs_fsops.c | 5 +++++
fs/xfs/xfs_mount.c | 6 ++++++
fs/xfs/xfs_rtalloc.c | 9 ++++++++-
4 files changed, 21 insertions(+), 8 deletions(-)
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index 7cab1b3..62b7e89 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -715,8 +715,7 @@ xfs_buf_read_uncached(
int flags,
xfs_buf_iodone_t verify)
{
- xfs_buf_t *bp;
- int error;
+ struct xfs_buf *bp;
bp = xfs_buf_get_uncached(target, numblks, flags);
if (!bp)
@@ -730,11 +729,7 @@ xfs_buf_read_uncached(
bp->b_iodone = verify;
xfsbdstrat(target->bt_mount, bp);
- error = xfs_buf_iowait(bp);
- if (error) {
- xfs_buf_relse(bp);
- return NULL;
- }
+ xfs_buf_iowait(bp);
return bp;
}
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index 917e121..dee14eb 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -149,6 +149,11 @@ xfs_growfs_data_private(
XFS_FSS_TO_BB(mp, 1), 0, NULL);
if (!bp)
return EIO;
+ if (bp->b_error) {
+ int error = bp->b_error;
+ xfs_buf_relse(bp);
+ return error;
+ }
xfs_buf_relse(bp);
new = nb; /* use new as a temporary here */
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index d39ad72..dc51e32 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -658,6 +658,12 @@ reread:
xfs_warn(mp, "SB buffer read failed");
return EIO;
}
+ if (bp->b_error) {
+ error = bp->b_error;
+ if (loud)
+ xfs_warn(mp, "SB validate failed");
+ goto release_buf;
+ }
/*
* Initialize the mount structure from the superblock.
diff --git a/fs/xfs/xfs_rtalloc.c b/fs/xfs/xfs_rtalloc.c
index b271ed9..98dc670 100644
--- a/fs/xfs/xfs_rtalloc.c
+++ b/fs/xfs/xfs_rtalloc.c
@@ -1876,6 +1876,11 @@ xfs_growfs_rt(
XFS_FSB_TO_BB(mp, 1), 0, NULL);
if (!bp)
return EIO;
+ if (bp->b_error) {
+ error = bp->b_error;
+ xfs_buf_relse(bp);
+ return error;
+ }
xfs_buf_relse(bp);
/*
@@ -2221,8 +2226,10 @@ xfs_rtmount_init(
bp = xfs_buf_read_uncached(mp->m_rtdev_targp,
d - XFS_FSB_TO_BB(mp, 1),
XFS_FSB_TO_BB(mp, 1), 0, NULL);
- if (!bp) {
+ if (!bp || bp->b_error) {
xfs_warn(mp, "realtime device size check failed");
+ if (bp)
+ xfs_buf_relse(bp);
return EIO;
}
xfs_buf_relse(bp);
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 05/25] xfs: verify superblocks as they are read from disk
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (3 preceding siblings ...)
2012-10-25 6:33 ` [PATCH 04/25] xfs: uncached buffer reads need to return an error Dave Chinner
@ 2012-10-25 6:33 ` Dave Chinner
2012-10-30 0:48 ` Phil White
2012-10-25 6:33 ` [PATCH 06/25] xfs: verify AGF blocks " Dave Chinner
` (19 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Add a superblock verify callback function and pass it into the
buffer read functions. Remove the now redundant verification code
that is currently in use.
Adding verification shows that secondary superblocks never have
their "sb_inprogress" flag cleared by mkfs.xfs, so when validating
the secondary superblocks during a grow operation we have to avoid
checking this field. Even if we fix mkfs, we will still have to
ignore this field for verification purposes unless a version of mkfs
that does not have this bug was used.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_fsops.c | 4 +-
fs/xfs/xfs_log_recover.c | 5 ++-
fs/xfs/xfs_mount.c | 98 +++++++++++++++++++++++++++++-----------------
fs/xfs/xfs_mount.h | 3 +-
4 files changed, 69 insertions(+), 41 deletions(-)
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index dee14eb..302b99c 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -413,7 +413,8 @@ xfs_growfs_data_private(
if (agno < oagcount) {
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
- XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
+ XFS_FSS_TO_BB(mp, 1), 0, &bp,
+ xfs_sb_read_verify);
} else {
bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
@@ -431,6 +432,7 @@ xfs_growfs_data_private(
break;
}
xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
+
/*
* If we get an error writing out the alternate superblocks,
* just issue a warning and continue. The real work is
diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
index 757688a..4cf7ae8 100644
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -3692,13 +3692,14 @@ xlog_do_recover(
/*
* Now that we've finished replaying all buffer and inode
- * updates, re-read in the superblock.
+ * updates, re-read in the superblock and reverify it.
*/
bp = xfs_getsb(log->l_mp, 0);
XFS_BUF_UNDONE(bp);
ASSERT(!(XFS_BUF_ISWRITE(bp)));
XFS_BUF_READ(bp);
XFS_BUF_UNASYNC(bp);
+ bp->b_iodone = xfs_sb_read_verify;
xfsbdstrat(log->l_mp, bp);
error = xfs_buf_iowait(bp);
if (error) {
@@ -3710,7 +3711,7 @@ xlog_do_recover(
/* Convert superblock from on-disk format */
sbp = &log->l_mp->m_sb;
- xfs_sb_from_disk(log->l_mp, XFS_BUF_TO_SBP(bp));
+ xfs_sb_from_disk(sbp, XFS_BUF_TO_SBP(bp));
ASSERT(sbp->sb_magicnum == XFS_SB_MAGIC);
ASSERT(xfs_sb_good_version(sbp));
xfs_buf_relse(bp);
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index dc51e32..8699e5e 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -304,9 +304,8 @@ STATIC int
xfs_mount_validate_sb(
xfs_mount_t *mp,
xfs_sb_t *sbp,
- int flags)
+ bool check_inprogress)
{
- int loud = !(flags & XFS_MFSI_QUIET);
/*
* If the log device and data device have the
@@ -316,21 +315,18 @@ xfs_mount_validate_sb(
* a volume filesystem in a non-volume manner.
*/
if (sbp->sb_magicnum != XFS_SB_MAGIC) {
- if (loud)
- xfs_warn(mp, "bad magic number");
+ xfs_warn(mp, "bad magic number");
return XFS_ERROR(EWRONGFS);
}
if (!xfs_sb_good_version(sbp)) {
- if (loud)
- xfs_warn(mp, "bad version");
+ xfs_warn(mp, "bad version");
return XFS_ERROR(EWRONGFS);
}
if (unlikely(
sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
- if (loud)
- xfs_warn(mp,
+ xfs_warn(mp,
"filesystem is marked as having an external log; "
"specify logdev on the mount command line.");
return XFS_ERROR(EINVAL);
@@ -338,8 +334,7 @@ xfs_mount_validate_sb(
if (unlikely(
sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
- if (loud)
- xfs_warn(mp,
+ xfs_warn(mp,
"filesystem is marked as having an internal log; "
"do not specify logdev on the mount command line.");
return XFS_ERROR(EINVAL);
@@ -373,8 +368,7 @@ xfs_mount_validate_sb(
sbp->sb_dblocks == 0 ||
sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) ||
sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp))) {
- if (loud)
- XFS_CORRUPTION_ERROR("SB sanity check failed",
+ XFS_CORRUPTION_ERROR("SB sanity check failed",
XFS_ERRLEVEL_LOW, mp, sbp);
return XFS_ERROR(EFSCORRUPTED);
}
@@ -383,12 +377,10 @@ xfs_mount_validate_sb(
* Until this is fixed only page-sized or smaller data blocks work.
*/
if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
- if (loud) {
- xfs_warn(mp,
+ xfs_warn(mp,
"File system with blocksize %d bytes. "
"Only pagesize (%ld) or less will currently work.",
sbp->sb_blocksize, PAGE_SIZE);
- }
return XFS_ERROR(ENOSYS);
}
@@ -402,23 +394,20 @@ xfs_mount_validate_sb(
case 2048:
break;
default:
- if (loud)
- xfs_warn(mp, "inode size of %d bytes not supported",
+ xfs_warn(mp, "inode size of %d bytes not supported",
sbp->sb_inodesize);
return XFS_ERROR(ENOSYS);
}
if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
- if (loud)
- xfs_warn(mp,
+ xfs_warn(mp,
"file system too large to be mounted on this system.");
return XFS_ERROR(EFBIG);
}
- if (unlikely(sbp->sb_inprogress)) {
- if (loud)
- xfs_warn(mp, "file system busy");
+ if (check_inprogress && sbp->sb_inprogress) {
+ xfs_warn(mp, "Offline file system operation in progress!");
return XFS_ERROR(EFSCORRUPTED);
}
@@ -426,9 +415,7 @@ xfs_mount_validate_sb(
* Version 1 directory format has never worked on Linux.
*/
if (unlikely(!xfs_sb_version_hasdirv2(sbp))) {
- if (loud)
- xfs_warn(mp,
- "file system using version 1 directory format");
+ xfs_warn(mp, "file system using version 1 directory format");
return XFS_ERROR(ENOSYS);
}
@@ -521,11 +508,9 @@ out_unwind:
void
xfs_sb_from_disk(
- struct xfs_mount *mp,
+ struct xfs_sb *to,
xfs_dsb_t *from)
{
- struct xfs_sb *to = &mp->m_sb;
-
to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
@@ -627,6 +612,50 @@ xfs_sb_to_disk(
}
}
+void
+xfs_sb_read_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_sb sb;
+ int error;
+
+ xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp));
+
+ /*
+ * Only check the in progress field for the primary superblock as
+ * mkfs.xfs doesn't clear it from secondary superblocks.
+ */
+ error = xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR);
+ if (error)
+ xfs_buf_ioerror(bp, error);
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
+/*
+ * We may be probed for a filesystem match, so we may not want to emit
+ * messages when the superblock buffer is not actually an XFS superblock.
+ * If we find an XFS superblock, the run a normal, noisy mount because we are
+ * really going to mount it and want to know about errors.
+ */
+void
+xfs_sb_quiet_read_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_sb sb;
+
+ xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp));
+
+ if (sb.sb_magicnum == XFS_SB_MAGIC) {
+ /* XFS filesystem, verify noisily! */
+ xfs_sb_read_verify(bp);
+ return;
+ }
+ /* quietly fail */
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+}
+
/*
* xfs_readsb
*
@@ -652,7 +681,9 @@ xfs_readsb(xfs_mount_t *mp, int flags)
reread:
bp = xfs_buf_read_uncached(mp->m_ddev_targp, XFS_SB_DADDR,
- BTOBB(sector_size), 0, NULL);
+ BTOBB(sector_size), 0,
+ loud ? xfs_sb_read_verify
+ : xfs_sb_quiet_read_verify);
if (!bp) {
if (loud)
xfs_warn(mp, "SB buffer read failed");
@@ -667,15 +698,8 @@ reread:
/*
* Initialize the mount structure from the superblock.
- * But first do some basic consistency checking.
*/
- xfs_sb_from_disk(mp, XFS_BUF_TO_SBP(bp));
- error = xfs_mount_validate_sb(mp, &(mp->m_sb), flags);
- if (error) {
- if (loud)
- xfs_warn(mp, "SB validate failed");
- goto release_buf;
- }
+ xfs_sb_from_disk(&mp->m_sb, XFS_BUF_TO_SBP(bp));
/*
* We must be able to do sector-sized and sector-aligned IO.
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index a631ca3..82b8fda 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -382,10 +382,11 @@ extern void xfs_set_low_space_thresholds(struct xfs_mount *);
#endif /* __KERNEL__ */
+extern void xfs_sb_read_verify(struct xfs_buf *);
extern void xfs_mod_sb(struct xfs_trans *, __int64_t);
extern int xfs_initialize_perag(struct xfs_mount *, xfs_agnumber_t,
xfs_agnumber_t *);
-extern void xfs_sb_from_disk(struct xfs_mount *, struct xfs_dsb *);
+extern void xfs_sb_from_disk(struct xfs_sb *, struct xfs_dsb *);
extern void xfs_sb_to_disk(struct xfs_dsb *, struct xfs_sb *, __int64_t);
#endif /* __XFS_MOUNT_H__ */
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 06/25] xfs: verify AGF blocks as they are read from disk
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (4 preceding siblings ...)
2012-10-25 6:33 ` [PATCH 05/25] xfs: verify superblocks as they are read from disk Dave Chinner
@ 2012-10-25 6:33 ` Dave Chinner
2012-10-30 0:51 ` Phil White
2012-10-25 6:33 ` [PATCH 07/25] xfs: verify AGI " Dave Chinner
` (18 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Add an AGF block verify callback function and pass it into the
buffer read functions. This replaces the existing verification that
is done after the read completes.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
fs/xfs/xfs_alloc.c | 60 +++++++++++++++++++++++++++++-----------------------
1 file changed, 34 insertions(+), 26 deletions(-)
diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
index 21c3db0..bd565a2 100644
--- a/fs/xfs/xfs_alloc.c
+++ b/fs/xfs/xfs_alloc.c
@@ -2091,6 +2091,39 @@ xfs_alloc_put_freelist(
return 0;
}
+static void
+xfs_agf_read_verify(
+ struct xfs_buf *bp)
+ {
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_agf *agf;
+ int agf_ok;
+
+ agf = XFS_BUF_TO_AGF(bp);
+
+ agf_ok = agf->agf_magicnum == cpu_to_be32(XFS_AGF_MAGIC) &&
+ XFS_AGF_GOOD_VERSION(be32_to_cpu(agf->agf_versionnum)) &&
+ be32_to_cpu(agf->agf_freeblks) <= be32_to_cpu(agf->agf_length) &&
+ be32_to_cpu(agf->agf_flfirst) < XFS_AGFL_SIZE(mp) &&
+ be32_to_cpu(agf->agf_fllast) < XFS_AGFL_SIZE(mp) &&
+ be32_to_cpu(agf->agf_flcount) <= XFS_AGFL_SIZE(mp) &&
+ be32_to_cpu(agf->agf_seqno) == bp->b_pag->pag_agno;
+
+ if (xfs_sb_version_haslazysbcount(&mp->m_sb))
+ agf_ok = agf_ok && be32_to_cpu(agf->agf_btreeblks) <=
+ be32_to_cpu(agf->agf_length);
+
+ if (unlikely(XFS_TEST_ERROR(!agf_ok, mp, XFS_ERRTAG_ALLOC_READ_AGF,
+ XFS_RANDOM_ALLOC_READ_AGF))) {
+ XFS_CORRUPTION_ERROR("xfs_alloc_read_agf",
+ XFS_ERRLEVEL_LOW, mp, agf);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
/*
* Read in the allocation group header (free/alloc section).
*/
@@ -2102,44 +2135,19 @@ xfs_read_agf(
int flags, /* XFS_BUF_ */
struct xfs_buf **bpp) /* buffer for the ag freelist header */
{
- struct xfs_agf *agf; /* ag freelist header */
- int agf_ok; /* set if agf is consistent */
int error;
ASSERT(agno != NULLAGNUMBER);
error = xfs_trans_read_buf(
mp, tp, mp->m_ddev_targp,
XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
- XFS_FSS_TO_BB(mp, 1), flags, bpp, NULL);
+ XFS_FSS_TO_BB(mp, 1), flags, bpp, xfs_agf_read_verify);
if (error)
return error;
if (!*bpp)
return 0;
ASSERT(!(*bpp)->b_error);
- agf = XFS_BUF_TO_AGF(*bpp);
-
- /*
- * Validate the magic number of the agf block.
- */
- agf_ok =
- agf->agf_magicnum == cpu_to_be32(XFS_AGF_MAGIC) &&
- XFS_AGF_GOOD_VERSION(be32_to_cpu(agf->agf_versionnum)) &&
- be32_to_cpu(agf->agf_freeblks) <= be32_to_cpu(agf->agf_length) &&
- be32_to_cpu(agf->agf_flfirst) < XFS_AGFL_SIZE(mp) &&
- be32_to_cpu(agf->agf_fllast) < XFS_AGFL_SIZE(mp) &&
- be32_to_cpu(agf->agf_flcount) <= XFS_AGFL_SIZE(mp) &&
- be32_to_cpu(agf->agf_seqno) == agno;
- if (xfs_sb_version_haslazysbcount(&mp->m_sb))
- agf_ok = agf_ok && be32_to_cpu(agf->agf_btreeblks) <=
- be32_to_cpu(agf->agf_length);
- if (unlikely(XFS_TEST_ERROR(!agf_ok, mp, XFS_ERRTAG_ALLOC_READ_AGF,
- XFS_RANDOM_ALLOC_READ_AGF))) {
- XFS_CORRUPTION_ERROR("xfs_alloc_read_agf",
- XFS_ERRLEVEL_LOW, mp, agf);
- xfs_trans_brelse(tp, *bpp);
- return XFS_ERROR(EFSCORRUPTED);
- }
xfs_buf_set_ref(*bpp, XFS_AGF_REF);
return 0;
}
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 07/25] xfs: verify AGI blocks as they are read from disk
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (5 preceding siblings ...)
2012-10-25 6:33 ` [PATCH 06/25] xfs: verify AGF blocks " Dave Chinner
@ 2012-10-25 6:33 ` Dave Chinner
2012-10-30 0:53 ` Phil White
2012-10-25 6:33 ` [PATCH 08/25] xfs: verify AGFL " Dave Chinner
` (17 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Add an AGI block verify callback function and pass it into the
buffer read functions. Remove the now redundant verification code
that is currently in use.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
fs/xfs/xfs_ialloc.c | 47 ++++++++++++++++++++++++++---------------------
1 file changed, 26 insertions(+), 21 deletions(-)
diff --git a/fs/xfs/xfs_ialloc.c b/fs/xfs/xfs_ialloc.c
index 7c944e1..9311ae5 100644
--- a/fs/xfs/xfs_ialloc.c
+++ b/fs/xfs/xfs_ialloc.c
@@ -1472,6 +1472,31 @@ xfs_check_agi_unlinked(
#define xfs_check_agi_unlinked(agi)
#endif
+static void
+xfs_agi_read_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_agi *agi = XFS_BUF_TO_AGI(bp);
+ int agi_ok;
+
+ /*
+ * Validate the magic number of the agi block.
+ */
+ agi_ok = agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC) &&
+ XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum)) &&
+ be32_to_cpu(agi->agi_seqno) == bp->b_pag->pag_agno;
+ if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI,
+ XFS_RANDOM_IALLOC_READ_AGI))) {
+ XFS_CORRUPTION_ERROR("xfs_read_agi", XFS_ERRLEVEL_LOW,
+ mp, agi);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+ xfs_check_agi_unlinked(agi);
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
/*
* Read in the allocation group header (inode allocation section)
*/
@@ -1482,38 +1507,18 @@ xfs_read_agi(
xfs_agnumber_t agno, /* allocation group number */
struct xfs_buf **bpp) /* allocation group hdr buf */
{
- struct xfs_agi *agi; /* allocation group header */
- int agi_ok; /* agi is consistent */
int error;
ASSERT(agno != NULLAGNUMBER);
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
- XFS_FSS_TO_BB(mp, 1), 0, bpp, NULL);
+ XFS_FSS_TO_BB(mp, 1), 0, bpp, xfs_agi_read_verify);
if (error)
return error;
ASSERT(!xfs_buf_geterror(*bpp));
- agi = XFS_BUF_TO_AGI(*bpp);
-
- /*
- * Validate the magic number of the agi block.
- */
- agi_ok = agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC) &&
- XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum)) &&
- be32_to_cpu(agi->agi_seqno) == agno;
- if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI,
- XFS_RANDOM_IALLOC_READ_AGI))) {
- XFS_CORRUPTION_ERROR("xfs_read_agi", XFS_ERRLEVEL_LOW,
- mp, agi);
- xfs_trans_brelse(tp, *bpp);
- return XFS_ERROR(EFSCORRUPTED);
- }
-
xfs_buf_set_ref(*bpp, XFS_AGI_REF);
-
- xfs_check_agi_unlinked(agi);
return 0;
}
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 08/25] xfs: verify AGFL blocks as they are read from disk
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (6 preceding siblings ...)
2012-10-25 6:33 ` [PATCH 07/25] xfs: verify AGI " Dave Chinner
@ 2012-10-25 6:33 ` Dave Chinner
2012-10-30 1:00 ` Phil White
2012-10-25 6:33 ` [PATCH 09/25] xfs: verify inode buffers " Dave Chinner
` (16 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Add an AGFL block verify callback function and pass it into the
buffer read functions.
While this commit adds verification code to the AGFL, it cannot be
used reliably until the CRC format change comes along as mkfs does
not initialise the full AGFL. Hence it can be full of garbage at the
first mount and will fail verification right now. CRC enabled
filesystems won't have this problem, so leave the code that has
already been written ifdef'd out until the proper time.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_alloc.c | 40 +++++++++++++++++++++++++++++++++++++++-
1 file changed, 39 insertions(+), 1 deletion(-)
diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
index bd565a2..0fa37a7 100644
--- a/fs/xfs/xfs_alloc.c
+++ b/fs/xfs/xfs_alloc.c
@@ -430,6 +430,44 @@ xfs_alloc_fixup_trees(
return 0;
}
+void
+xfs_agfl_read_verify(
+ struct xfs_buf *bp)
+{
+#ifdef WHEN_CRCS_COME_ALONG
+ /*
+ * we cannot actually do any verification of the AGFL because mkfs does
+ * not initialise the AGFL to zero or NULL. Hence the only valid part of
+ * the AGFL is what the AGF says is active. We can't get to the AGF, so
+ * we can't verify just those entries are valid.
+ *
+ * This problem goes away when the CRC format change comes along as that
+ * requires the AGFL to be initialised by mkfs. At that point, we can
+ * verify the blocks in the agfl -active or not- lie within the bounds
+ * of the AG. Until then, just leave this check ifdef'd out.
+ */
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_agfl *agfl = XFS_BUF_TO_AGFL(bp);
+ int agfl_ok = 1;
+
+ int i;
+
+ for (i = 0; i < XFS_AGFL_SIZE(mp); i++) {
+ if (be32_to_cpu(agfl->agfl_bno[i]) == NULLAGBLOCK ||
+ be32_to_cpu(agfl->agfl_bno[i]) >= mp->m_sb.sb_agblocks)
+ agfl_ok = 0;
+ }
+
+ if (!agfl_ok) {
+ XFS_CORRUPTION_ERROR("xfs_agfl_read_verify",
+ XFS_ERRLEVEL_LOW, mp, agfl);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+#endif
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
/*
* Read in the allocation group free block array.
*/
@@ -447,7 +485,7 @@ xfs_alloc_read_agfl(
error = xfs_trans_read_buf(
mp, tp, mp->m_ddev_targp,
XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
- XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
+ XFS_FSS_TO_BB(mp, 1), 0, &bp, xfs_agfl_read_verify);
if (error)
return error;
ASSERT(!xfs_buf_geterror(bp));
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 09/25] xfs: verify inode buffers as they are read from disk
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (7 preceding siblings ...)
2012-10-25 6:33 ` [PATCH 08/25] xfs: verify AGFL " Dave Chinner
@ 2012-10-25 6:33 ` Dave Chinner
2012-10-30 1:06 ` Phil White
2012-10-25 6:33 ` [PATCH 10/25] xfs: verify btree blocks " Dave Chinner
` (15 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Add an inode buffer verify callback function and pass it into the
buffer read functions. Inodes are special in that the verbose checks
will be done when reading the inode, but we still need to sanity
check the buffer when that is first read. Always verify the magic
numbers in all inodes in the buffer, rather than jus ton debug
kernels.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_inode.c | 100 +++++++++++++++++++++++++++-------------------------
1 file changed, 51 insertions(+), 49 deletions(-)
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 0b03578..5baf6cb 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -382,6 +382,46 @@ xfs_inobp_check(
}
#endif
+static void
+xfs_inode_buf_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ int i;
+ int ni;
+
+ /*
+ * Validate the magic number and version of every inode in the buffer
+ */
+ ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
+ for (i = 0; i < ni; i++) {
+ int di_ok;
+ xfs_dinode_t *dip;
+
+ dip = (struct xfs_dinode *)xfs_buf_offset(bp,
+ (i << mp->m_sb.sb_inodelog));
+ di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
+ XFS_DINODE_GOOD_VERSION(dip->di_version);
+ if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
+ XFS_ERRTAG_ITOBP_INOTOBP,
+ XFS_RANDOM_ITOBP_INOTOBP))) {
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_HIGH,
+ mp, dip);
+#ifdef DEBUG
+ xfs_emerg(mp,
+ "bad inode magic/vsn daddr %lld #%d (magic=%x)",
+ (unsigned long long)bp->b_bn, i,
+ be16_to_cpu(dip->di_magic));
+ ASSERT(0);
+#endif
+ }
+ }
+ xfs_inobp_check(mp, bp);
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
/*
* This routine is called to map an inode to the buffer containing the on-disk
* version of the inode. It returns a pointer to the buffer containing the
@@ -396,71 +436,33 @@ xfs_imap_to_bp(
struct xfs_mount *mp,
struct xfs_trans *tp,
struct xfs_imap *imap,
- struct xfs_dinode **dipp,
+ struct xfs_dinode **dipp,
struct xfs_buf **bpp,
uint buf_flags,
uint iget_flags)
{
struct xfs_buf *bp;
int error;
- int i;
- int ni;
buf_flags |= XBF_UNMAPPED;
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
- (int)imap->im_len, buf_flags, &bp, NULL);
+ (int)imap->im_len, buf_flags, &bp,
+ xfs_inode_buf_verify);
if (error) {
- if (error != EAGAIN) {
- xfs_warn(mp,
- "%s: xfs_trans_read_buf() returned error %d.",
- __func__, error);
- } else {
+ if (error == EAGAIN) {
ASSERT(buf_flags & XBF_TRYLOCK);
+ return error;
}
- return error;
- }
- /*
- * Validate the magic number and version of every inode in the buffer
- * (if DEBUG kernel) or the first inode in the buffer, otherwise.
- */
-#ifdef DEBUG
- ni = BBTOB(imap->im_len) >> mp->m_sb.sb_inodelog;
-#else /* usual case */
- ni = 1;
-#endif
+ if (error == EFSCORRUPTED &&
+ (iget_flags & XFS_IGET_UNTRUSTED))
+ return XFS_ERROR(EINVAL);
- for (i = 0; i < ni; i++) {
- int di_ok;
- xfs_dinode_t *dip;
-
- dip = (xfs_dinode_t *)xfs_buf_offset(bp,
- (i << mp->m_sb.sb_inodelog));
- di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
- XFS_DINODE_GOOD_VERSION(dip->di_version);
- if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
- XFS_ERRTAG_ITOBP_INOTOBP,
- XFS_RANDOM_ITOBP_INOTOBP))) {
- if (iget_flags & XFS_IGET_UNTRUSTED) {
- xfs_trans_brelse(tp, bp);
- return XFS_ERROR(EINVAL);
- }
- XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_HIGH,
- mp, dip);
-#ifdef DEBUG
- xfs_emerg(mp,
- "bad inode magic/vsn daddr %lld #%d (magic=%x)",
- (unsigned long long)imap->im_blkno, i,
- be16_to_cpu(dip->di_magic));
- ASSERT(0);
-#endif
- xfs_trans_brelse(tp, bp);
- return XFS_ERROR(EFSCORRUPTED);
- }
+ xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
+ __func__, error);
+ return error;
}
- xfs_inobp_check(mp, bp);
-
*bpp = bp;
*dipp = (struct xfs_dinode *)xfs_buf_offset(bp, imap->im_boffset);
return 0;
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 10/25] xfs: verify btree blocks as they are read from disk
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (8 preceding siblings ...)
2012-10-25 6:33 ` [PATCH 09/25] xfs: verify inode buffers " Dave Chinner
@ 2012-10-25 6:33 ` Dave Chinner
2012-10-30 1:14 ` Phil White
2012-10-25 6:34 ` [PATCH 11/25] xfs: verify dquot " Dave Chinner
` (14 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:33 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Add an btree block verify callback function and pass it into the
buffer read functions. Because each different btree block type
requires different verification, add a function to the ops structure
that is called from the generic code.
Also, propagate the verification callback functions through the
readahead functions, and into the external bmap and bulkstat inode
readahead code that uses the generic btree buffer read functions.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_alloc_btree.c | 49 +++++++++++++++++++++++++++++++++
fs/xfs/xfs_bmap.c | 60 ++++++++++++++++++++++++-----------------
fs/xfs/xfs_bmap_btree.c | 47 ++++++++++++++++++++++++++++++++
fs/xfs/xfs_bmap_btree.h | 1 +
fs/xfs/xfs_btree.c | 66 +++++++++++++++++++++++----------------------
fs/xfs/xfs_btree.h | 10 ++++---
fs/xfs/xfs_ialloc_btree.c | 40 +++++++++++++++++++++++++++
fs/xfs/xfs_inode.c | 2 +-
fs/xfs/xfs_inode.h | 1 +
fs/xfs/xfs_itable.c | 3 ++-
10 files changed, 218 insertions(+), 61 deletions(-)
diff --git a/fs/xfs/xfs_alloc_btree.c b/fs/xfs/xfs_alloc_btree.c
index f7876c6..4167e72 100644
--- a/fs/xfs/xfs_alloc_btree.c
+++ b/fs/xfs/xfs_alloc_btree.c
@@ -272,6 +272,54 @@ xfs_allocbt_key_diff(
return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
}
+void
+xfs_allocbt_read_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
+ struct xfs_perag *pag = bp->b_pag;
+ unsigned int level;
+ int sblock_ok; /* block passes checks */
+
+ /* magic number and level verification */
+ level = be16_to_cpu(block->bb_level);
+ switch (block->bb_magic) {
+ case cpu_to_be32(XFS_ABTB_MAGIC):
+ sblock_ok = level < pag->pagf_levels[XFS_BTNUM_BNOi];
+ break;
+ case cpu_to_be32(XFS_ABTC_MAGIC):
+ sblock_ok = level < pag->pagf_levels[XFS_BTNUM_CNTi];
+ break;
+ default:
+ sblock_ok = 0;
+ break;
+ }
+
+ /* numrecs verification */
+ sblock_ok = sblock_ok &&
+ be16_to_cpu(block->bb_numrecs) <= mp->m_alloc_mxr[level != 0];
+
+ /* sibling pointer verification */
+ sblock_ok = sblock_ok &&
+ (block->bb_u.s.bb_leftsib == cpu_to_be32(NULLAGBLOCK) ||
+ be32_to_cpu(block->bb_u.s.bb_leftsib) < mp->m_sb.sb_agblocks) &&
+ block->bb_u.s.bb_leftsib &&
+ (block->bb_u.s.bb_rightsib == cpu_to_be32(NULLAGBLOCK) ||
+ be32_to_cpu(block->bb_u.s.bb_rightsib) < mp->m_sb.sb_agblocks) &&
+ block->bb_u.s.bb_rightsib;
+
+ if (!sblock_ok) {
+ trace_xfs_btree_corrupt(bp, _RET_IP_);
+ XFS_CORRUPTION_ERROR("xfs_allocbt_read_verify",
+ XFS_ERRLEVEL_LOW, mp, block);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
#ifdef DEBUG
STATIC int
xfs_allocbt_keys_inorder(
@@ -327,6 +375,7 @@ static const struct xfs_btree_ops xfs_allocbt_ops = {
.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,
+ .read_verify = xfs_allocbt_read_verify,
#ifdef DEBUG
.keys_inorder = xfs_allocbt_keys_inorder,
.recs_inorder = xfs_allocbt_recs_inorder,
diff --git a/fs/xfs/xfs_bmap.c b/fs/xfs/xfs_bmap.c
index 83d0cf3..8e944bb 100644
--- a/fs/xfs/xfs_bmap.c
+++ b/fs/xfs/xfs_bmap.c
@@ -2662,8 +2662,9 @@ xfs_bmap_btree_to_extents(
if ((error = xfs_btree_check_lptr(cur, cbno, 1)))
return error;
#endif
- if ((error = xfs_btree_read_bufl(mp, tp, cbno, 0, &cbp,
- XFS_BMAP_BTREE_REF)))
+ error = xfs_btree_read_bufl(mp, tp, cbno, 0, &cbp, XFS_BMAP_BTREE_REF,
+ xfs_bmbt_read_verify);
+ if (error)
return error;
cblock = XFS_BUF_TO_BLOCK(cbp);
if ((error = xfs_btree_check_block(cur, cblock, 0, cbp)))
@@ -4078,8 +4079,9 @@ xfs_bmap_read_extents(
* pointer (leftmost) at each level.
*/
while (level-- > 0) {
- if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
- XFS_BMAP_BTREE_REF)))
+ error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
+ XFS_BMAP_BTREE_REF, xfs_bmbt_read_verify);
+ if (error)
return error;
block = XFS_BUF_TO_BLOCK(bp);
XFS_WANT_CORRUPTED_GOTO(
@@ -4124,7 +4126,8 @@ xfs_bmap_read_extents(
*/
nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
if (nextbno != NULLFSBLOCK)
- xfs_btree_reada_bufl(mp, nextbno, 1);
+ xfs_btree_reada_bufl(mp, nextbno, 1,
+ xfs_bmbt_read_verify);
/*
* Copy records into the extent records.
*/
@@ -4156,8 +4159,9 @@ xfs_bmap_read_extents(
*/
if (bno == NULLFSBLOCK)
break;
- if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
- XFS_BMAP_BTREE_REF)))
+ error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
+ XFS_BMAP_BTREE_REF, xfs_bmbt_read_verify);
+ if (error)
return error;
block = XFS_BUF_TO_BLOCK(bp);
}
@@ -5868,15 +5872,16 @@ xfs_bmap_check_leaf_extents(
*/
while (level-- > 0) {
/* See if buf is in cur first */
+ bp_release = 0;
bp = xfs_bmap_get_bp(cur, XFS_FSB_TO_DADDR(mp, bno));
- if (bp) {
- bp_release = 0;
- } else {
+ if (!bp) {
bp_release = 1;
+ error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
+ XFS_BMAP_BTREE_REF,
+ xfs_bmbt_read_verify);
+ if (error)
+ goto error_norelse;
}
- if (!bp && (error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
- XFS_BMAP_BTREE_REF)))
- goto error_norelse;
block = XFS_BUF_TO_BLOCK(bp);
XFS_WANT_CORRUPTED_GOTO(
xfs_bmap_sanity_check(mp, bp, level),
@@ -5953,15 +5958,16 @@ xfs_bmap_check_leaf_extents(
if (bno == NULLFSBLOCK)
break;
+ bp_release = 0;
bp = xfs_bmap_get_bp(cur, XFS_FSB_TO_DADDR(mp, bno));
- if (bp) {
- bp_release = 0;
- } else {
+ if (!bp) {
bp_release = 1;
+ error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
+ XFS_BMAP_BTREE_REF,
+ xfs_bmbt_read_verify);
+ if (error)
+ goto error_norelse;
}
- if (!bp && (error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
- XFS_BMAP_BTREE_REF)))
- goto error_norelse;
block = XFS_BUF_TO_BLOCK(bp);
}
if (bp_release) {
@@ -6052,7 +6058,9 @@ xfs_bmap_count_tree(
struct xfs_btree_block *block, *nextblock;
int numrecs;
- if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp, XFS_BMAP_BTREE_REF)))
+ error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp, XFS_BMAP_BTREE_REF,
+ xfs_bmbt_read_verify);
+ if (error)
return error;
*count += 1;
block = XFS_BUF_TO_BLOCK(bp);
@@ -6061,8 +6069,10 @@ xfs_bmap_count_tree(
/* Not at node above leaves, count this level of nodes */
nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
while (nextbno != NULLFSBLOCK) {
- if ((error = xfs_btree_read_bufl(mp, tp, nextbno,
- 0, &nbp, XFS_BMAP_BTREE_REF)))
+ error = xfs_btree_read_bufl(mp, tp, nextbno, 0, &nbp,
+ XFS_BMAP_BTREE_REF,
+ xfs_bmbt_read_verify);
+ if (error)
return error;
*count += 1;
nextblock = XFS_BUF_TO_BLOCK(nbp);
@@ -6091,8 +6101,10 @@ xfs_bmap_count_tree(
if (nextbno == NULLFSBLOCK)
break;
bno = nextbno;
- if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
- XFS_BMAP_BTREE_REF)))
+ error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
+ XFS_BMAP_BTREE_REF,
+ xfs_bmbt_read_verify);
+ if (error)
return error;
*count += 1;
block = XFS_BUF_TO_BLOCK(bp);
diff --git a/fs/xfs/xfs_bmap_btree.c b/fs/xfs/xfs_bmap_btree.c
index 862084a..bddca9b 100644
--- a/fs/xfs/xfs_bmap_btree.c
+++ b/fs/xfs/xfs_bmap_btree.c
@@ -36,6 +36,7 @@
#include "xfs_bmap.h"
#include "xfs_error.h"
#include "xfs_quota.h"
+#include "xfs_trace.h"
/*
* Determine the extent state.
@@ -707,6 +708,51 @@ xfs_bmbt_key_diff(
cur->bc_rec.b.br_startoff;
}
+void
+xfs_bmbt_read_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
+ unsigned int level;
+ int lblock_ok; /* block passes checks */
+
+ /* magic number and level verification.
+ *
+ * We don't know waht fork we belong to, so just verify that the level
+ * is less than the maximum of the two. Later checks will be more
+ * precise.
+ */
+ level = be16_to_cpu(block->bb_level);
+ lblock_ok = block->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC) &&
+ level < max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]);
+
+ /* numrecs verification */
+ lblock_ok = lblock_ok &&
+ be16_to_cpu(block->bb_numrecs) <= mp->m_bmap_dmxr[level != 0];
+
+ /* sibling pointer verification */
+ lblock_ok = lblock_ok &&
+ block->bb_u.l.bb_leftsib &&
+ (block->bb_u.l.bb_leftsib == cpu_to_be64(NULLDFSBNO) ||
+ XFS_FSB_SANITY_CHECK(mp,
+ be64_to_cpu(block->bb_u.l.bb_leftsib))) &&
+ block->bb_u.l.bb_rightsib &&
+ (block->bb_u.l.bb_rightsib == cpu_to_be64(NULLDFSBNO) ||
+ XFS_FSB_SANITY_CHECK(mp,
+ be64_to_cpu(block->bb_u.l.bb_rightsib)));
+
+ if (!lblock_ok) {
+ trace_xfs_btree_corrupt(bp, _RET_IP_);
+ XFS_CORRUPTION_ERROR("xfs_bmbt_read_verify",
+ XFS_ERRLEVEL_LOW, mp, block);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
#ifdef DEBUG
STATIC int
xfs_bmbt_keys_inorder(
@@ -746,6 +792,7 @@ static const struct xfs_btree_ops xfs_bmbt_ops = {
.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,
+ .read_verify = xfs_bmbt_read_verify,
#ifdef DEBUG
.keys_inorder = xfs_bmbt_keys_inorder,
.recs_inorder = xfs_bmbt_recs_inorder,
diff --git a/fs/xfs/xfs_bmap_btree.h b/fs/xfs/xfs_bmap_btree.h
index 0e66c4e..1d00fbe 100644
--- a/fs/xfs/xfs_bmap_btree.h
+++ b/fs/xfs/xfs_bmap_btree.h
@@ -232,6 +232,7 @@ extern void xfs_bmbt_to_bmdr(struct xfs_mount *, struct xfs_btree_block *, int,
extern int xfs_bmbt_get_maxrecs(struct xfs_btree_cur *, int level);
extern int xfs_bmdr_maxrecs(struct xfs_mount *, int blocklen, int leaf);
extern int xfs_bmbt_maxrecs(struct xfs_mount *, int blocklen, int leaf);
+extern void xfs_bmbt_read_verify(struct xfs_buf *bp);
extern struct xfs_btree_cur *xfs_bmbt_init_cursor(struct xfs_mount *,
struct xfs_trans *, struct xfs_inode *, int);
diff --git a/fs/xfs/xfs_btree.c b/fs/xfs/xfs_btree.c
index 1937c9b..b680949 100644
--- a/fs/xfs/xfs_btree.c
+++ b/fs/xfs/xfs_btree.c
@@ -270,7 +270,8 @@ xfs_btree_dup_cursor(
if (bp) {
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
XFS_BUF_ADDR(bp), mp->m_bsize,
- 0, &bp, NULL);
+ 0, &bp,
+ cur->bc_ops->read_verify);
if (error) {
xfs_btree_del_cursor(new, error);
*ncur = NULL;
@@ -612,23 +613,24 @@ xfs_btree_offsets(
* Get a buffer for the block, return it read in.
* Long-form addressing.
*/
-int /* error */
+int
xfs_btree_read_bufl(
- xfs_mount_t *mp, /* file system mount point */
- xfs_trans_t *tp, /* transaction pointer */
- xfs_fsblock_t fsbno, /* file system block number */
- uint lock, /* lock flags for read_buf */
- xfs_buf_t **bpp, /* buffer for fsbno */
- int refval) /* ref count value for buffer */
-{
- xfs_buf_t *bp; /* return value */
+ struct xfs_mount *mp, /* file system mount point */
+ struct xfs_trans *tp, /* transaction pointer */
+ xfs_fsblock_t fsbno, /* file system block number */
+ uint lock, /* lock flags for read_buf */
+ struct xfs_buf **bpp, /* buffer for fsbno */
+ int refval, /* ref count value for buffer */
+ xfs_buf_iodone_t verify)
+{
+ struct xfs_buf *bp; /* return value */
xfs_daddr_t d; /* real disk block address */
- int error;
+ int error;
ASSERT(fsbno != NULLFSBLOCK);
d = XFS_FSB_TO_DADDR(mp, fsbno);
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, d,
- mp->m_bsize, lock, &bp, NULL);
+ mp->m_bsize, lock, &bp, verify);
if (error)
return error;
ASSERT(!xfs_buf_geterror(bp));
@@ -645,15 +647,16 @@ xfs_btree_read_bufl(
/* ARGSUSED */
void
xfs_btree_reada_bufl(
- xfs_mount_t *mp, /* file system mount point */
- xfs_fsblock_t fsbno, /* file system block number */
- xfs_extlen_t count) /* count of filesystem blocks */
+ struct xfs_mount *mp, /* file system mount point */
+ xfs_fsblock_t fsbno, /* file system block number */
+ xfs_extlen_t count, /* count of filesystem blocks */
+ xfs_buf_iodone_t verify)
{
xfs_daddr_t d;
ASSERT(fsbno != NULLFSBLOCK);
d = XFS_FSB_TO_DADDR(mp, fsbno);
- xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, NULL);
+ xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, verify);
}
/*
@@ -663,17 +666,18 @@ xfs_btree_reada_bufl(
/* ARGSUSED */
void
xfs_btree_reada_bufs(
- xfs_mount_t *mp, /* file system mount point */
- xfs_agnumber_t agno, /* allocation group number */
- xfs_agblock_t agbno, /* allocation group block number */
- xfs_extlen_t count) /* count of filesystem blocks */
+ struct xfs_mount *mp, /* file system mount point */
+ xfs_agnumber_t agno, /* allocation group number */
+ xfs_agblock_t agbno, /* allocation group block number */
+ xfs_extlen_t count, /* count of filesystem blocks */
+ xfs_buf_iodone_t verify)
{
xfs_daddr_t d;
ASSERT(agno != NULLAGNUMBER);
ASSERT(agbno != NULLAGBLOCK);
d = XFS_AGB_TO_DADDR(mp, agno, agbno);
- xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, NULL);
+ xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, verify);
}
STATIC int
@@ -687,12 +691,14 @@ xfs_btree_readahead_lblock(
xfs_dfsbno_t right = be64_to_cpu(block->bb_u.l.bb_rightsib);
if ((lr & XFS_BTCUR_LEFTRA) && left != NULLDFSBNO) {
- xfs_btree_reada_bufl(cur->bc_mp, left, 1);
+ xfs_btree_reada_bufl(cur->bc_mp, left, 1,
+ cur->bc_ops->read_verify);
rval++;
}
if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLDFSBNO) {
- xfs_btree_reada_bufl(cur->bc_mp, right, 1);
+ xfs_btree_reada_bufl(cur->bc_mp, right, 1,
+ cur->bc_ops->read_verify);
rval++;
}
@@ -712,13 +718,13 @@ xfs_btree_readahead_sblock(
if ((lr & XFS_BTCUR_LEFTRA) && left != NULLAGBLOCK) {
xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
- left, 1);
+ left, 1, cur->bc_ops->read_verify);
rval++;
}
if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLAGBLOCK) {
xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
- right, 1);
+ right, 1, cur->bc_ops->read_verify);
rval++;
}
@@ -1001,19 +1007,15 @@ xfs_btree_read_buf_block(
d = xfs_btree_ptr_to_daddr(cur, ptr);
error = xfs_trans_read_buf(mp, cur->bc_tp, mp->m_ddev_targp, d,
- mp->m_bsize, flags, bpp, NULL);
+ mp->m_bsize, flags, bpp,
+ cur->bc_ops->read_verify);
if (error)
return error;
ASSERT(!xfs_buf_geterror(*bpp));
-
xfs_btree_set_refs(cur, *bpp);
*block = XFS_BUF_TO_BLOCK(*bpp);
-
- error = xfs_btree_check_block(cur, *block, level, *bpp);
- if (error)
- xfs_trans_brelse(cur->bc_tp, *bpp);
- return error;
+ return 0;
}
/*
diff --git a/fs/xfs/xfs_btree.h b/fs/xfs/xfs_btree.h
index 5b240de..7c3cb8d 100644
--- a/fs/xfs/xfs_btree.h
+++ b/fs/xfs/xfs_btree.h
@@ -188,6 +188,7 @@ struct xfs_btree_ops {
__int64_t (*key_diff)(struct xfs_btree_cur *cur,
union xfs_btree_key *key);
+ void (*read_verify)(struct xfs_buf *bp);
#ifdef DEBUG
/* check that k1 is lower than k2 */
int (*keys_inorder)(struct xfs_btree_cur *cur,
@@ -355,7 +356,8 @@ xfs_btree_read_bufl(
xfs_fsblock_t fsbno, /* file system block number */
uint lock, /* lock flags for read_buf */
struct xfs_buf **bpp, /* buffer for fsbno */
- int refval);/* ref count value for buffer */
+ int refval, /* ref count value for buffer */
+ xfs_buf_iodone_t verify);
/*
* Read-ahead the block, don't wait for it, don't return a buffer.
@@ -365,7 +367,8 @@ void /* error */
xfs_btree_reada_bufl(
struct xfs_mount *mp, /* file system mount point */
xfs_fsblock_t fsbno, /* file system block number */
- xfs_extlen_t count); /* count of filesystem blocks */
+ xfs_extlen_t count, /* count of filesystem blocks */
+ xfs_buf_iodone_t verify);
/*
* Read-ahead the block, don't wait for it, don't return a buffer.
@@ -376,7 +379,8 @@ xfs_btree_reada_bufs(
struct xfs_mount *mp, /* file system mount point */
xfs_agnumber_t agno, /* allocation group number */
xfs_agblock_t agbno, /* allocation group block number */
- xfs_extlen_t count); /* count of filesystem blocks */
+ xfs_extlen_t count, /* count of filesystem blocks */
+ xfs_buf_iodone_t verify);
/*
diff --git a/fs/xfs/xfs_ialloc_btree.c b/fs/xfs/xfs_ialloc_btree.c
index 2b8b7a3..11306c6 100644
--- a/fs/xfs/xfs_ialloc_btree.c
+++ b/fs/xfs/xfs_ialloc_btree.c
@@ -33,6 +33,7 @@
#include "xfs_ialloc.h"
#include "xfs_alloc.h"
#include "xfs_error.h"
+#include "xfs_trace.h"
STATIC int
@@ -181,6 +182,44 @@ xfs_inobt_key_diff(
cur->bc_rec.i.ir_startino;
}
+void
+xfs_inobt_read_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
+ unsigned int level;
+ int sblock_ok; /* block passes checks */
+
+ /* magic number and level verification */
+ level = be16_to_cpu(block->bb_level);
+ sblock_ok = block->bb_magic == cpu_to_be32(XFS_IBT_MAGIC) &&
+ level < mp->m_in_maxlevels;
+
+ /* numrecs verification */
+ sblock_ok = sblock_ok &&
+ be16_to_cpu(block->bb_numrecs) <= mp->m_inobt_mxr[level != 0];
+
+ /* sibling pointer verification */
+ sblock_ok = sblock_ok &&
+ (block->bb_u.s.bb_leftsib == cpu_to_be32(NULLAGBLOCK) ||
+ be32_to_cpu(block->bb_u.s.bb_leftsib) < mp->m_sb.sb_agblocks) &&
+ block->bb_u.s.bb_leftsib &&
+ (block->bb_u.s.bb_rightsib == cpu_to_be32(NULLAGBLOCK) ||
+ be32_to_cpu(block->bb_u.s.bb_rightsib) < mp->m_sb.sb_agblocks) &&
+ block->bb_u.s.bb_rightsib;
+
+ if (!sblock_ok) {
+ trace_xfs_btree_corrupt(bp, _RET_IP_);
+ XFS_CORRUPTION_ERROR("xfs_inobt_read_verify",
+ XFS_ERRLEVEL_LOW, mp, block);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
#ifdef DEBUG
STATIC int
xfs_inobt_keys_inorder(
@@ -218,6 +257,7 @@ static const struct xfs_btree_ops xfs_inobt_ops = {
.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,
+ .read_verify = xfs_inobt_read_verify,
#ifdef DEBUG
.keys_inorder = xfs_inobt_keys_inorder,
.recs_inorder = xfs_inobt_recs_inorder,
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 5baf6cb..0905e72 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -382,7 +382,7 @@ xfs_inobp_check(
}
#endif
-static void
+void
xfs_inode_buf_verify(
struct xfs_buf *bp)
{
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index 1fc2065..3c1d831 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -554,6 +554,7 @@ int xfs_imap_to_bp(struct xfs_mount *, struct xfs_trans *,
struct xfs_buf **, uint, uint);
int xfs_iread(struct xfs_mount *, struct xfs_trans *,
struct xfs_inode *, uint);
+void xfs_inode_buf_verify(struct xfs_buf *);
void xfs_dinode_to_disk(struct xfs_dinode *,
struct xfs_icdinode *);
void xfs_idestroy_fork(struct xfs_inode *, int);
diff --git a/fs/xfs/xfs_itable.c b/fs/xfs/xfs_itable.c
index 3998fd2..0f18d41 100644
--- a/fs/xfs/xfs_itable.c
+++ b/fs/xfs/xfs_itable.c
@@ -396,7 +396,8 @@ xfs_bulkstat(
if (xfs_inobt_maskn(chunkidx, nicluster)
& ~r.ir_free)
xfs_btree_reada_bufs(mp, agno,
- agbno, nbcluster);
+ agbno, nbcluster,
+ xfs_inode_buf_verify);
}
irbp->ir_startino = r.ir_startino;
irbp->ir_freecount = r.ir_freecount;
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 11/25] xfs: verify dquot blocks as they are read from disk
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (9 preceding siblings ...)
2012-10-25 6:33 ` [PATCH 10/25] xfs: verify btree blocks " Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 1:36 ` Phil White
2012-10-25 6:34 ` [PATCH 12/25] xfs: add verifier callback to directory read code Dave Chinner
` (13 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Add a dquot buffer verify callback function and pass it into the
buffer read functions. This checks all the dquots in a buffer, but
cannot completely verify the dquot ids are correct. Also, errors
cannot be repaired, so an additional function is added to repair bad
dquots in the buffer if such an error is detected in a context where
repair is allowed.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_dquot.c | 117 ++++++++++++++++++++++++++++++++++++++++++----------
1 file changed, 95 insertions(+), 22 deletions(-)
diff --git a/fs/xfs/xfs_dquot.c b/fs/xfs/xfs_dquot.c
index e95f800..2e18382 100644
--- a/fs/xfs/xfs_dquot.c
+++ b/fs/xfs/xfs_dquot.c
@@ -360,6 +360,89 @@ xfs_qm_dqalloc(
return (error);
}
+STATIC void
+xfs_dquot_read_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_dqblk *d = (struct xfs_dqblk *)bp->b_addr;
+ struct xfs_disk_dquot *ddq;
+ xfs_dqid_t id = 0;
+ int i;
+
+ /*
+ * On the first read of the buffer, verify that each dquot is valid.
+ * We don't know what the id of the dquot is supposed to be, just that
+ * they should be increasing monotonically within the buffer. If the
+ * first id is corrupt, then it will fail on the second dquot in the
+ * buffer so corruptions could point to the wrong dquot in this case.
+ */
+ for (i = 0; i < mp->m_quotainfo->qi_dqperchunk; i++) {
+ int error;
+
+ ddq = &d[i].dd_diskdq;
+
+ if (i == 0)
+ id = be32_to_cpu(ddq->d_id);
+
+ error = xfs_qm_dqcheck(mp, ddq, id + i, 0, XFS_QMOPT_DOWARN,
+ "xfs_dquot_read_verify");
+ if (error) {
+ XFS_CORRUPTION_ERROR("xfs_dquot_read_verify",
+ XFS_ERRLEVEL_LOW, mp, d);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ break;
+ }
+ }
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
+STATIC int
+xfs_qm_dqrepair(
+ struct xfs_mount *mp,
+ struct xfs_trans *tp,
+ struct xfs_dquot *dqp,
+ xfs_dqid_t firstid,
+ struct xfs_buf **bpp)
+{
+ int error;
+ struct xfs_disk_dquot *ddq;
+ struct xfs_dqblk *d;
+ int i;
+
+ /*
+ * Read the buffer without verification so we get the corrupted
+ * buffer returned to us.
+ */
+ error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, dqp->q_blkno,
+ mp->m_quotainfo->qi_dqchunklen,
+ 0, bpp, NULL);
+
+ if (error) {
+ ASSERT(*bpp == NULL);
+ return XFS_ERROR(error);
+ }
+
+ ASSERT(xfs_buf_islocked(*bpp));
+ d = (struct xfs_dqblk *)(*bpp)->b_addr;
+
+ /* Do the actual repair of dquots in this buffer */
+ for (i = 0; i < mp->m_quotainfo->qi_dqperchunk; i++) {
+ ddq = &d[i].dd_diskdq;
+ error = xfs_qm_dqcheck(mp, ddq, firstid + i,
+ dqp->dq_flags & XFS_DQ_ALLTYPES,
+ XFS_QMOPT_DQREPAIR, "xfs_qm_dqrepair");
+ if (error) {
+ /* repair failed, we're screwed */
+ xfs_trans_brelse(tp, *bpp);
+ return XFS_ERROR(EIO);
+ }
+ }
+
+ return 0;
+}
+
/*
* Maps a dquot to the buffer containing its on-disk version.
* This returns a ptr to the buffer containing the on-disk dquot
@@ -378,7 +461,6 @@ xfs_qm_dqtobp(
xfs_buf_t *bp;
xfs_inode_t *quotip = XFS_DQ_TO_QIP(dqp);
xfs_mount_t *mp = dqp->q_mount;
- xfs_disk_dquot_t *ddq;
xfs_dqid_t id = be32_to_cpu(dqp->q_core.d_id);
xfs_trans_t *tp = (tpp ? *tpp : NULL);
@@ -439,33 +521,24 @@ xfs_qm_dqtobp(
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
dqp->q_blkno,
mp->m_quotainfo->qi_dqchunklen,
- 0, &bp, NULL);
- if (error || !bp)
- return XFS_ERROR(error);
- }
+ 0, &bp, xfs_dquot_read_verify);
- ASSERT(xfs_buf_islocked(bp));
-
- /*
- * calculate the location of the dquot inside the buffer.
- */
- ddq = bp->b_addr + dqp->q_bufoffset;
+ if (error == EFSCORRUPTED && (flags & XFS_QMOPT_DQREPAIR)) {
+ xfs_dqid_t firstid = (xfs_dqid_t)map.br_startoff *
+ mp->m_quotainfo->qi_dqperchunk;
+ ASSERT(bp == NULL);
+ error = xfs_qm_dqrepair(mp, tp, dqp, firstid, &bp);
+ }
- /*
- * A simple sanity check in case we got a corrupted dquot...
- */
- error = xfs_qm_dqcheck(mp, ddq, id, dqp->dq_flags & XFS_DQ_ALLTYPES,
- flags & (XFS_QMOPT_DQREPAIR|XFS_QMOPT_DOWARN),
- "dqtobp");
- if (error) {
- if (!(flags & XFS_QMOPT_DQREPAIR)) {
- xfs_trans_brelse(tp, bp);
- return XFS_ERROR(EIO);
+ if (error) {
+ ASSERT(bp == NULL);
+ return XFS_ERROR(error);
}
}
+ ASSERT(xfs_buf_islocked(bp));
*O_bpp = bp;
- *O_ddpp = ddq;
+ *O_ddpp = bp->b_addr + dqp->q_bufoffset;
return (0);
}
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 12/25] xfs: add verifier callback to directory read code
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (10 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 11/25] xfs: verify dquot " Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 3:15 ` Phil White
2012-10-25 6:34 ` [PATCH 13/25] xfs: factor dir2 block read operations Dave Chinner
` (12 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
---
fs/xfs/xfs_attr.c | 23 ++++++++++++-----------
fs/xfs/xfs_attr_leaf.c | 18 +++++++++---------
fs/xfs/xfs_da_btree.c | 44 ++++++++++++++++++++++++++++----------------
fs/xfs/xfs_da_btree.h | 7 ++++---
fs/xfs/xfs_dir2_block.c | 23 ++++++++++++-----------
fs/xfs/xfs_dir2_leaf.c | 33 ++++++++++++++++-----------------
fs/xfs/xfs_dir2_node.c | 43 ++++++++++++++++++++-----------------------
fs/xfs/xfs_file.c | 2 +-
8 files changed, 102 insertions(+), 91 deletions(-)
diff --git a/fs/xfs/xfs_attr.c b/fs/xfs/xfs_attr.c
index ebacb8d..956c2ba 100644
--- a/fs/xfs/xfs_attr.c
+++ b/fs/xfs/xfs_attr.c
@@ -904,7 +904,7 @@ xfs_attr_leaf_addname(xfs_da_args_t *args)
dp = args->dp;
args->blkno = 0;
error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
- XFS_ATTR_FORK);
+ XFS_ATTR_FORK, NULL);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1032,7 +1032,7 @@ xfs_attr_leaf_addname(xfs_da_args_t *args)
* remove the "old" attr from that block (neat, huh!)
*/
error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1,
- &bp, XFS_ATTR_FORK);
+ &bp, XFS_ATTR_FORK, NULL);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1101,7 +1101,7 @@ xfs_attr_leaf_removename(xfs_da_args_t *args)
dp = args->dp;
args->blkno = 0;
error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
- XFS_ATTR_FORK);
+ XFS_ATTR_FORK, NULL);
if (error) {
return(error);
}
@@ -1157,7 +1157,7 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
args->blkno = 0;
error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
- XFS_ATTR_FORK);
+ XFS_ATTR_FORK, NULL);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1186,7 +1186,8 @@ xfs_attr_leaf_list(xfs_attr_list_context_t *context)
struct xfs_buf *bp;
context->cursor->blkno = 0;
- error = xfs_da_read_buf(NULL, context->dp, 0, -1, &bp, XFS_ATTR_FORK);
+ error = xfs_da_read_buf(NULL, context->dp, 0, -1, &bp, XFS_ATTR_FORK,
+ NULL);
if (error)
return XFS_ERROR(error);
ASSERT(bp != NULL);
@@ -1601,7 +1602,7 @@ xfs_attr_node_removename(xfs_da_args_t *args)
state->path.blk[0].bp = NULL;
error = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp,
- XFS_ATTR_FORK);
+ XFS_ATTR_FORK, NULL);
if (error)
goto out;
ASSERT((((xfs_attr_leafblock_t *)bp->b_addr)->hdr.info.magic) ==
@@ -1710,7 +1711,7 @@ xfs_attr_refillstate(xfs_da_state_t *state)
error = xfs_da_read_buf(state->args->trans,
state->args->dp,
blk->blkno, blk->disk_blkno,
- &blk->bp, XFS_ATTR_FORK);
+ &blk->bp, XFS_ATTR_FORK, NULL);
if (error)
return(error);
} else {
@@ -1729,7 +1730,7 @@ xfs_attr_refillstate(xfs_da_state_t *state)
error = xfs_da_read_buf(state->args->trans,
state->args->dp,
blk->blkno, blk->disk_blkno,
- &blk->bp, XFS_ATTR_FORK);
+ &blk->bp, XFS_ATTR_FORK, NULL);
if (error)
return(error);
} else {
@@ -1815,7 +1816,7 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
bp = NULL;
if (cursor->blkno > 0) {
error = xfs_da_read_buf(NULL, context->dp, cursor->blkno, -1,
- &bp, XFS_ATTR_FORK);
+ &bp, XFS_ATTR_FORK, NULL);
if ((error != 0) && (error != EFSCORRUPTED))
return(error);
if (bp) {
@@ -1858,7 +1859,7 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
for (;;) {
error = xfs_da_read_buf(NULL, context->dp,
cursor->blkno, -1, &bp,
- XFS_ATTR_FORK);
+ XFS_ATTR_FORK, NULL);
if (error)
return(error);
if (unlikely(bp == NULL)) {
@@ -1925,7 +1926,7 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
cursor->blkno = be32_to_cpu(leaf->hdr.info.forw);
xfs_trans_brelse(NULL, bp);
error = xfs_da_read_buf(NULL, context->dp, cursor->blkno, -1,
- &bp, XFS_ATTR_FORK);
+ &bp, XFS_ATTR_FORK, NULL);
if (error)
return(error);
if (unlikely((bp == NULL))) {
diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c
index d330111..f2b698e 100644
--- a/fs/xfs/xfs_attr_leaf.c
+++ b/fs/xfs/xfs_attr_leaf.c
@@ -870,7 +870,7 @@ xfs_attr_leaf_to_node(xfs_da_args_t *args)
if (error)
goto out;
error = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp1,
- XFS_ATTR_FORK);
+ XFS_ATTR_FORK, NULL);
if (error)
goto out;
ASSERT(bp1 != NULL);
@@ -1621,7 +1621,7 @@ xfs_attr_leaf_toosmall(xfs_da_state_t *state, int *action)
if (blkno == 0)
continue;
error = xfs_da_read_buf(state->args->trans, state->args->dp,
- blkno, -1, &bp, XFS_ATTR_FORK);
+ blkno, -1, &bp, XFS_ATTR_FORK, NULL);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -2496,7 +2496,7 @@ xfs_attr_leaf_clearflag(xfs_da_args_t *args)
* Set up the operation.
*/
error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
- XFS_ATTR_FORK);
+ XFS_ATTR_FORK, NULL);
if (error) {
return(error);
}
@@ -2561,7 +2561,7 @@ xfs_attr_leaf_setflag(xfs_da_args_t *args)
* Set up the operation.
*/
error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
- XFS_ATTR_FORK);
+ XFS_ATTR_FORK, NULL);
if (error) {
return(error);
}
@@ -2618,7 +2618,7 @@ xfs_attr_leaf_flipflags(xfs_da_args_t *args)
* Read the block containing the "old" attr
*/
error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp1,
- XFS_ATTR_FORK);
+ XFS_ATTR_FORK, NULL);
if (error) {
return(error);
}
@@ -2629,7 +2629,7 @@ xfs_attr_leaf_flipflags(xfs_da_args_t *args)
*/
if (args->blkno2 != args->blkno) {
error = xfs_da_read_buf(args->trans, args->dp, args->blkno2,
- -1, &bp2, XFS_ATTR_FORK);
+ -1, &bp2, XFS_ATTR_FORK, NULL);
if (error) {
return(error);
}
@@ -2730,7 +2730,7 @@ xfs_attr_root_inactive(xfs_trans_t **trans, xfs_inode_t *dp)
* the extents in reverse order the extent containing
* block 0 must still be there.
*/
- error = xfs_da_read_buf(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK);
+ error = xfs_da_read_buf(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK, NULL);
if (error)
return(error);
blkno = XFS_BUF_ADDR(bp);
@@ -2816,7 +2816,7 @@ xfs_attr_node_inactive(
* before we come back to this one.
*/
error = xfs_da_read_buf(*trans, dp, child_fsb, -2, &child_bp,
- XFS_ATTR_FORK);
+ XFS_ATTR_FORK, NULL);
if (error)
return(error);
if (child_bp) {
@@ -2857,7 +2857,7 @@ xfs_attr_node_inactive(
*/
if ((i+1) < count) {
error = xfs_da_read_buf(*trans, dp, 0, parent_blkno,
- &bp, XFS_ATTR_FORK);
+ &bp, XFS_ATTR_FORK, NULL);
if (error)
return(error);
child_fsb = be32_to_cpu(node->btree[i+1].before);
diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
index 41d8764..a46035b 100644
--- a/fs/xfs/xfs_da_btree.c
+++ b/fs/xfs/xfs_da_btree.c
@@ -747,7 +747,7 @@ xfs_da_root_join(xfs_da_state_t *state, xfs_da_state_blk_t *root_blk)
child = be32_to_cpu(oldroot->btree[0].before);
ASSERT(child != 0);
error = xfs_da_read_buf(args->trans, args->dp, child, -1, &bp,
- args->whichfork);
+ args->whichfork, NULL);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -836,7 +836,8 @@ xfs_da_node_toosmall(xfs_da_state_t *state, int *action)
if (blkno == 0)
continue;
error = xfs_da_read_buf(state->args->trans, state->args->dp,
- blkno, -1, &bp, state->args->whichfork);
+ blkno, -1, &bp, state->args->whichfork,
+ NULL);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1080,7 +1081,7 @@ xfs_da_node_lookup_int(xfs_da_state_t *state, int *result)
*/
blk->blkno = blkno;
error = xfs_da_read_buf(args->trans, args->dp, blkno,
- -1, &blk->bp, args->whichfork);
+ -1, &blk->bp, args->whichfork, NULL);
if (error) {
blk->blkno = 0;
state->path.active--;
@@ -1243,7 +1244,7 @@ xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
if (old_info->back) {
error = xfs_da_read_buf(args->trans, args->dp,
be32_to_cpu(old_info->back),
- -1, &bp, args->whichfork);
+ -1, &bp, args->whichfork, NULL);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1264,7 +1265,7 @@ xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
if (old_info->forw) {
error = xfs_da_read_buf(args->trans, args->dp,
be32_to_cpu(old_info->forw),
- -1, &bp, args->whichfork);
+ -1, &bp, args->whichfork, NULL);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1364,7 +1365,7 @@ xfs_da_blk_unlink(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
if (drop_info->back) {
error = xfs_da_read_buf(args->trans, args->dp,
be32_to_cpu(drop_info->back),
- -1, &bp, args->whichfork);
+ -1, &bp, args->whichfork, NULL);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1381,7 +1382,7 @@ xfs_da_blk_unlink(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
if (drop_info->forw) {
error = xfs_da_read_buf(args->trans, args->dp,
be32_to_cpu(drop_info->forw),
- -1, &bp, args->whichfork);
+ -1, &bp, args->whichfork, NULL);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1464,7 +1465,7 @@ xfs_da_path_shift(xfs_da_state_t *state, xfs_da_state_path_t *path,
*/
blk->blkno = blkno;
error = xfs_da_read_buf(args->trans, args->dp, blkno, -1,
- &blk->bp, args->whichfork);
+ &blk->bp, args->whichfork, NULL);
if (error)
return(error);
ASSERT(blk->bp != NULL);
@@ -1727,7 +1728,8 @@ xfs_da_swap_lastblock(
* Read the last block in the btree space.
*/
last_blkno = (xfs_dablk_t)lastoff - mp->m_dirblkfsbs;
- if ((error = xfs_da_read_buf(tp, ip, last_blkno, -1, &last_buf, w)))
+ error = xfs_da_read_buf(tp, ip, last_blkno, -1, &last_buf, w, NULL);
+ if (error)
return error;
/*
* Copy the last block into the dead buffer and log it.
@@ -1753,7 +1755,9 @@ xfs_da_swap_lastblock(
* If the moved block has a left sibling, fix up the pointers.
*/
if ((sib_blkno = be32_to_cpu(dead_info->back))) {
- if ((error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w)))
+ error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w,
+ NULL);
+ if (error)
goto done;
sib_info = sib_buf->b_addr;
if (unlikely(
@@ -1774,7 +1778,9 @@ xfs_da_swap_lastblock(
* If the moved block has a right sibling, fix up the pointers.
*/
if ((sib_blkno = be32_to_cpu(dead_info->forw))) {
- if ((error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w)))
+ error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w,
+ NULL);
+ if (error)
goto done;
sib_info = sib_buf->b_addr;
if (unlikely(
@@ -1797,7 +1803,9 @@ xfs_da_swap_lastblock(
* Walk down the tree looking for the parent of the moved block.
*/
for (;;) {
- if ((error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w)))
+ error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w,
+ NULL);
+ if (error)
goto done;
par_node = par_buf->b_addr;
if (unlikely(par_node->hdr.info.magic !=
@@ -1847,7 +1855,9 @@ xfs_da_swap_lastblock(
error = XFS_ERROR(EFSCORRUPTED);
goto done;
}
- if ((error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w)))
+ error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w,
+ NULL);
+ if (error)
goto done;
par_node = par_buf->b_addr;
if (unlikely(
@@ -2133,7 +2143,8 @@ xfs_da_read_buf(
xfs_dablk_t bno,
xfs_daddr_t mappedbno,
struct xfs_buf **bpp,
- int whichfork)
+ int whichfork,
+ xfs_buf_iodone_t verifier)
{
struct xfs_buf *bp;
struct xfs_buf_map map;
@@ -2155,7 +2166,7 @@ xfs_da_read_buf(
error = xfs_trans_read_buf_map(dp->i_mount, trans,
dp->i_mount->m_ddev_targp,
- mapp, nmap, 0, &bp, NULL);
+ mapp, nmap, 0, &bp, verifier);
if (error)
goto out_free;
@@ -2211,7 +2222,8 @@ xfs_da_reada_buf(
struct xfs_trans *trans,
struct xfs_inode *dp,
xfs_dablk_t bno,
- int whichfork)
+ int whichfork,
+ xfs_buf_iodone_t verifier)
{
xfs_daddr_t mappedbno = -1;
struct xfs_buf_map map;
diff --git a/fs/xfs/xfs_da_btree.h b/fs/xfs/xfs_da_btree.h
index 132adaf..bf8bfaa 100644
--- a/fs/xfs/xfs_da_btree.h
+++ b/fs/xfs/xfs_da_btree.h
@@ -18,7 +18,6 @@
#ifndef __XFS_DA_BTREE_H__
#define __XFS_DA_BTREE_H__
-struct xfs_buf;
struct xfs_bmap_free;
struct xfs_inode;
struct xfs_mount;
@@ -226,9 +225,11 @@ int xfs_da_get_buf(struct xfs_trans *trans, struct xfs_inode *dp,
struct xfs_buf **bp, int whichfork);
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);
+ struct xfs_buf **bpp, int whichfork,
+ xfs_buf_iodone_t verifier);
xfs_daddr_t xfs_da_reada_buf(struct xfs_trans *trans, struct xfs_inode *dp,
- xfs_dablk_t bno, int whichfork);
+ xfs_dablk_t bno, int whichfork,
+ xfs_buf_iodone_t verifier);
int xfs_da_shrink_inode(xfs_da_args_t *args, xfs_dablk_t dead_blkno,
struct xfs_buf *dead_buf);
diff --git a/fs/xfs/xfs_dir2_block.c b/fs/xfs/xfs_dir2_block.c
index e93ca8f..53666ca 100644
--- a/fs/xfs/xfs_dir2_block.c
+++ b/fs/xfs/xfs_dir2_block.c
@@ -97,10 +97,10 @@ xfs_dir2_block_addname(
/*
* Read the (one and only) directory block into dabuf bp.
*/
- if ((error =
- xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &bp, XFS_DATA_FORK))) {
+ error = xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &bp,
+ XFS_DATA_FORK, NULL);
+ if (error)
return error;
- }
ASSERT(bp != NULL);
hdr = bp->b_addr;
/*
@@ -457,7 +457,7 @@ xfs_dir2_block_getdents(
* Can't read the block, give up, else get dabuf in bp.
*/
error = xfs_da_read_buf(NULL, dp, mp->m_dirdatablk, -1,
- &bp, XFS_DATA_FORK);
+ &bp, XFS_DATA_FORK, NULL);
if (error)
return error;
@@ -640,10 +640,10 @@ xfs_dir2_block_lookup_int(
/*
* Read the buffer, return error if we can't get it.
*/
- if ((error =
- xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &bp, XFS_DATA_FORK))) {
+ error = xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &bp,
+ XFS_DATA_FORK, NULL);
+ if (error)
return error;
- }
ASSERT(bp != NULL);
hdr = bp->b_addr;
xfs_dir2_data_check(dp, bp);
@@ -917,10 +917,11 @@ xfs_dir2_leaf_to_block(
/*
* Read the data block if we don't already have it, give up if it fails.
*/
- if (dbp == NULL &&
- (error = xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &dbp,
- XFS_DATA_FORK))) {
- return error;
+ if (!dbp) {
+ error = xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &dbp,
+ XFS_DATA_FORK, NULL);
+ if (error)
+ return error;
}
hdr = dbp->b_addr;
ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC));
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index bac8698..86e3dc1 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -315,10 +315,9 @@ xfs_dir2_leaf_addname(
* Read the leaf block.
*/
error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
- XFS_DATA_FORK);
- if (error) {
+ XFS_DATA_FORK, NULL);
+ if (error)
return error;
- }
ASSERT(lbp != NULL);
/*
* Look up the entry by hash value and name.
@@ -500,9 +499,9 @@ xfs_dir2_leaf_addname(
* Just read that one in.
*/
else {
- if ((error =
- xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, use_block),
- -1, &dbp, XFS_DATA_FORK))) {
+ error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, use_block),
+ -1, &dbp, XFS_DATA_FORK, NULL);
+ if (error) {
xfs_trans_brelse(tp, lbp);
return error;
}
@@ -895,7 +894,7 @@ xfs_dir2_leaf_readbuf(
error = xfs_da_read_buf(NULL, dp, map->br_startoff,
map->br_blockcount >= mp->m_dirblkfsbs ?
XFS_FSB_TO_DADDR(mp, map->br_startblock) : -1,
- &bp, XFS_DATA_FORK);
+ &bp, XFS_DATA_FORK, NULL);
/*
* Should just skip over the data block instead of giving up.
@@ -938,7 +937,7 @@ xfs_dir2_leaf_readbuf(
xfs_da_reada_buf(NULL, dp,
map[mip->ra_index].br_startoff +
mip->ra_offset,
- XFS_DATA_FORK);
+ XFS_DATA_FORK, NULL);
mip->ra_current = i;
}
@@ -1376,7 +1375,7 @@ xfs_dir2_leaf_lookup_int(
* Read the leaf block into the buffer.
*/
error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
- XFS_DATA_FORK);
+ XFS_DATA_FORK, NULL);
if (error)
return error;
*lbpp = lbp;
@@ -1411,7 +1410,7 @@ xfs_dir2_leaf_lookup_int(
xfs_trans_brelse(tp, dbp);
error = xfs_da_read_buf(tp, dp,
xfs_dir2_db_to_da(mp, newdb),
- -1, &dbp, XFS_DATA_FORK);
+ -1, &dbp, XFS_DATA_FORK, NULL);
if (error) {
xfs_trans_brelse(tp, lbp);
return error;
@@ -1453,7 +1452,7 @@ xfs_dir2_leaf_lookup_int(
xfs_trans_brelse(tp, dbp);
error = xfs_da_read_buf(tp, dp,
xfs_dir2_db_to_da(mp, cidb),
- -1, &dbp, XFS_DATA_FORK);
+ -1, &dbp, XFS_DATA_FORK, NULL);
if (error) {
xfs_trans_brelse(tp, lbp);
return error;
@@ -1738,10 +1737,10 @@ xfs_dir2_leaf_trim_data(
/*
* Read the offending data block. We need its buffer.
*/
- if ((error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, db), -1, &dbp,
- XFS_DATA_FORK))) {
+ error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, db), -1, &dbp,
+ XFS_DATA_FORK, NULL);
+ if (error)
return error;
- }
leaf = lbp->b_addr;
ltp = xfs_dir2_leaf_tail_p(mp, leaf);
@@ -1864,10 +1863,10 @@ xfs_dir2_node_to_leaf(
/*
* Read the freespace block.
*/
- if ((error = xfs_da_read_buf(tp, dp, mp->m_dirfreeblk, -1, &fbp,
- XFS_DATA_FORK))) {
+ error = xfs_da_read_buf(tp, dp, mp->m_dirfreeblk, -1, &fbp,
+ XFS_DATA_FORK, NULL);
+ if (error)
return error;
- }
free = fbp->b_addr;
ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC));
ASSERT(!free->hdr.firstdb);
diff --git a/fs/xfs/xfs_dir2_node.c b/fs/xfs/xfs_dir2_node.c
index 6c70524..290c2b1 100644
--- a/fs/xfs/xfs_dir2_node.c
+++ b/fs/xfs/xfs_dir2_node.c
@@ -399,7 +399,7 @@ xfs_dir2_leafn_lookup_for_addname(
*/
error = xfs_da_read_buf(tp, dp,
xfs_dir2_db_to_da(mp, newfdb),
- -1, &curbp, XFS_DATA_FORK);
+ -1, &curbp, XFS_DATA_FORK, NULL);
if (error)
return error;
free = curbp->b_addr;
@@ -536,7 +536,7 @@ xfs_dir2_leafn_lookup_for_entry(
} else {
error = xfs_da_read_buf(tp, dp,
xfs_dir2_db_to_da(mp, newdb),
- -1, &curbp, XFS_DATA_FORK);
+ -1, &curbp, XFS_DATA_FORK, NULL);
if (error)
return error;
}
@@ -915,10 +915,10 @@ xfs_dir2_leafn_remove(
* read in the free block.
*/
fdb = xfs_dir2_db_to_fdb(mp, db);
- if ((error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, fdb),
- -1, &fbp, XFS_DATA_FORK))) {
+ error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, fdb),
+ -1, &fbp, XFS_DATA_FORK, NULL);
+ if (error)
return error;
- }
free = fbp->b_addr;
ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC));
ASSERT(be32_to_cpu(free->hdr.firstdb) ==
@@ -1169,11 +1169,10 @@ xfs_dir2_leafn_toosmall(
/*
* Read the sibling leaf block.
*/
- if ((error =
- xfs_da_read_buf(state->args->trans, state->args->dp, blkno,
- -1, &bp, XFS_DATA_FORK))) {
+ error = xfs_da_read_buf(state->args->trans, state->args->dp,
+ blkno, -1, &bp, XFS_DATA_FORK, NULL);
+ if (error)
return error;
- }
ASSERT(bp != NULL);
/*
* Count bytes in the two blocks combined.
@@ -1454,14 +1453,13 @@ xfs_dir2_node_addname_int(
* This should be really rare, so there's no reason
* to avoid it.
*/
- if ((error = xfs_da_read_buf(tp, dp,
- xfs_dir2_db_to_da(mp, fbno), -2, &fbp,
- XFS_DATA_FORK))) {
+ error = xfs_da_read_buf(tp, dp,
+ xfs_dir2_db_to_da(mp, fbno), -2,
+ &fbp, XFS_DATA_FORK, NULL);
+ if (error)
return error;
- }
- if (unlikely(fbp == NULL)) {
+ if (!fbp)
continue;
- }
free = fbp->b_addr;
ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC));
findex = 0;
@@ -1520,9 +1518,9 @@ xfs_dir2_node_addname_int(
* that was just allocated.
*/
fbno = xfs_dir2_db_to_fdb(mp, dbno);
- if (unlikely(error = xfs_da_read_buf(tp, dp,
- xfs_dir2_db_to_da(mp, fbno), -2, &fbp,
- XFS_DATA_FORK)))
+ error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, fbno), -2,
+ &fbp, XFS_DATA_FORK, NULL);
+ if (error)
return error;
/*
@@ -1631,7 +1629,7 @@ xfs_dir2_node_addname_int(
* Read the data block in.
*/
error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, dbno),
- -1, &dbp, XFS_DATA_FORK);
+ -1, &dbp, XFS_DATA_FORK, NULL);
if (error)
return error;
hdr = dbp->b_addr;
@@ -1917,11 +1915,10 @@ xfs_dir2_node_trim_free(
/*
* Read the freespace block.
*/
- if (unlikely(error = xfs_da_read_buf(tp, dp, (xfs_dablk_t)fo, -2, &bp,
- XFS_DATA_FORK))) {
+ error = xfs_da_read_buf(tp, dp, (xfs_dablk_t)fo, -2, &bp,
+ XFS_DATA_FORK, NULL);
+ if (error)
return error;
- }
-
/*
* There can be holes in freespace. If fo is a hole, there's
* nothing to do.
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index daf4066..d949bad 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -890,7 +890,7 @@ xfs_dir_open(
*/
mode = xfs_ilock_map_shared(ip);
if (ip->i_d.di_nextents > 0)
- xfs_da_reada_buf(NULL, ip, 0, XFS_DATA_FORK);
+ xfs_da_reada_buf(NULL, ip, 0, XFS_DATA_FORK, NULL);
xfs_iunlock(ip, mode);
return 0;
}
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 13/25] xfs: factor dir2 block read operations
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (11 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 12/25] xfs: add verifier callback to directory read code Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 3:23 ` Phil White
2012-10-25 6:34 ` [PATCH 14/25] xfs: verify dir2 block format buffers Dave Chinner
` (11 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
In preparation for verifying dir2 block format buffers, factor
the read operations out of the block operations (lookup, addname,
getdents) and some of the additional logic to make it easier to
understand an dmodify the code.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_dir2_block.c | 386 +++++++++++++++++++++++++----------------------
1 file changed, 209 insertions(+), 177 deletions(-)
diff --git a/fs/xfs/xfs_dir2_block.c b/fs/xfs/xfs_dir2_block.c
index 53666ca..25ce409 100644
--- a/fs/xfs/xfs_dir2_block.c
+++ b/fs/xfs/xfs_dir2_block.c
@@ -56,6 +56,178 @@ xfs_dir_startup(void)
xfs_dir_hash_dotdot = xfs_da_hashname((unsigned char *)"..", 2);
}
+static int
+xfs_dir2_block_read(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ struct xfs_buf **bpp)
+{
+ struct xfs_mount *mp = dp->i_mount;
+
+ return xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, bpp,
+ XFS_DATA_FORK, NULL);
+}
+
+static void
+xfs_dir2_block_need_space(
+ struct xfs_dir2_data_hdr *hdr,
+ struct xfs_dir2_block_tail *btp,
+ struct xfs_dir2_leaf_entry *blp,
+ __be16 **tagpp,
+ struct xfs_dir2_data_unused **dupp,
+ struct xfs_dir2_data_unused **enddupp,
+ int *compact,
+ int len)
+{
+ struct xfs_dir2_data_free *bf;
+ __be16 *tagp = NULL;
+ struct xfs_dir2_data_unused *dup = NULL;
+ struct xfs_dir2_data_unused *enddup = NULL;
+
+ *compact = 0;
+ bf = hdr->bestfree;
+
+ /*
+ * If there are stale entries we'll use one for the leaf.
+ */
+ if (btp->stale) {
+ if (be16_to_cpu(bf[0].length) >= len) {
+ /*
+ * The biggest entry enough to avoid compaction.
+ */
+ dup = (xfs_dir2_data_unused_t *)
+ ((char *)hdr + be16_to_cpu(bf[0].offset));
+ goto out;
+ }
+
+ /*
+ * Will need to compact to make this work.
+ * Tag just before the first leaf entry.
+ */
+ *compact = 1;
+ tagp = (__be16 *)blp - 1;
+
+ /* Data object just before the first leaf entry. */
+ dup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
+
+ /*
+ * If it's not free then the data will go where the
+ * leaf data starts now, if it works at all.
+ */
+ if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
+ if (be16_to_cpu(dup->length) + (be32_to_cpu(btp->stale) - 1) *
+ (uint)sizeof(*blp) < len)
+ dup = NULL;
+ } else if ((be32_to_cpu(btp->stale) - 1) * (uint)sizeof(*blp) < len)
+ dup = NULL;
+ else
+ dup = (xfs_dir2_data_unused_t *)blp;
+ goto out;
+ }
+
+ /*
+ * no stale entries, so just use free space.
+ * Tag just before the first leaf entry.
+ */
+ tagp = (__be16 *)blp - 1;
+
+ /* Data object just before the first leaf entry. */
+ enddup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
+
+ /*
+ * If it's not free then can't do this add without cleaning up:
+ * the space before the first leaf entry needs to be free so it
+ * can be expanded to hold the pointer to the new entry.
+ */
+ if (be16_to_cpu(enddup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
+ /*
+ * Check out the biggest freespace and see if it's the same one.
+ */
+ dup = (xfs_dir2_data_unused_t *)
+ ((char *)hdr + be16_to_cpu(bf[0].offset));
+ if (dup != enddup) {
+ /*
+ * Not the same free entry, just check its length.
+ */
+ if (be16_to_cpu(dup->length) < len)
+ dup = NULL;
+ goto out;
+ }
+
+ /*
+ * It is the biggest freespace, can it hold the leaf too?
+ */
+ if (be16_to_cpu(dup->length) < len + (uint)sizeof(*blp)) {
+ /*
+ * Yes, use the second-largest entry instead if it works.
+ */
+ if (be16_to_cpu(bf[1].length) >= len)
+ dup = (xfs_dir2_data_unused_t *)
+ ((char *)hdr + be16_to_cpu(bf[1].offset));
+ else
+ dup = NULL;
+ }
+ }
+out:
+ *tagpp = tagp;
+ *dupp = dup;
+ *enddupp = enddup;
+}
+
+/*
+ * compact the leaf entries.
+ * Leave the highest-numbered stale entry stale.
+ * XXX should be the one closest to mid but mid is not yet computed.
+ */
+static void
+xfs_dir2_block_compact(
+ struct xfs_trans *tp,
+ struct xfs_buf *bp,
+ struct xfs_dir2_data_hdr *hdr,
+ struct xfs_dir2_block_tail *btp,
+ struct xfs_dir2_leaf_entry *blp,
+ int *needlog,
+ int *lfloghigh,
+ int *lfloglow)
+{
+ int fromidx; /* source leaf index */
+ int toidx; /* target leaf index */
+ int needscan = 0;
+ int highstale; /* high stale index */
+
+ fromidx = toidx = be32_to_cpu(btp->count) - 1;
+ highstale = *lfloghigh = -1;
+ for (; fromidx >= 0; fromidx--) {
+ if (blp[fromidx].address == cpu_to_be32(XFS_DIR2_NULL_DATAPTR)) {
+ if (highstale == -1)
+ highstale = toidx;
+ else {
+ if (*lfloghigh == -1)
+ *lfloghigh = toidx;
+ continue;
+ }
+ }
+ if (fromidx < toidx)
+ blp[toidx] = blp[fromidx];
+ toidx--;
+ }
+ *lfloglow = toidx + 1 - (be32_to_cpu(btp->stale) - 1);
+ *lfloghigh -= be32_to_cpu(btp->stale) - 1;
+ be32_add_cpu(&btp->count, -(be32_to_cpu(btp->stale) - 1));
+ xfs_dir2_data_make_free(tp, bp,
+ (xfs_dir2_data_aoff_t)((char *)blp - (char *)hdr),
+ (xfs_dir2_data_aoff_t)((be32_to_cpu(btp->stale) - 1) * sizeof(*blp)),
+ needlog, &needscan);
+ blp += be32_to_cpu(btp->stale) - 1;
+ btp->stale = cpu_to_be32(1);
+ /*
+ * If we now need to rebuild the bestfree map, do so.
+ * This needs to happen before the next call to use_free.
+ */
+ if (needscan)
+ xfs_dir2_data_freescan(tp->t_mountp, hdr, needlog);
+}
+
/*
* Add an entry to a block directory.
*/
@@ -63,7 +235,6 @@ int /* error */
xfs_dir2_block_addname(
xfs_da_args_t *args) /* directory op arguments */
{
- xfs_dir2_data_free_t *bf; /* bestfree table in block */
xfs_dir2_data_hdr_t *hdr; /* block header */
xfs_dir2_leaf_entry_t *blp; /* block leaf entries */
struct xfs_buf *bp; /* buffer for block */
@@ -94,134 +265,44 @@ xfs_dir2_block_addname(
dp = args->dp;
tp = args->trans;
mp = dp->i_mount;
- /*
- * Read the (one and only) directory block into dabuf bp.
- */
- error = xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &bp,
- XFS_DATA_FORK, NULL);
+
+ /* Read the (one and only) directory block into bp. */
+ error = xfs_dir2_block_read(tp, dp, &bp);
if (error)
return error;
- ASSERT(bp != NULL);
- hdr = bp->b_addr;
- /*
- * Check the magic number, corrupted if wrong.
- */
- if (unlikely(hdr->magic != cpu_to_be32(XFS_DIR2_BLOCK_MAGIC))) {
- XFS_CORRUPTION_ERROR("xfs_dir2_block_addname",
- XFS_ERRLEVEL_LOW, mp, hdr);
- xfs_trans_brelse(tp, bp);
- return XFS_ERROR(EFSCORRUPTED);
- }
+
len = xfs_dir2_data_entsize(args->namelen);
+
/*
* Set up pointers to parts of the block.
*/
- bf = hdr->bestfree;
+ hdr = bp->b_addr;
btp = xfs_dir2_block_tail_p(mp, hdr);
blp = xfs_dir2_block_leaf_p(btp);
+
/*
- * No stale entries? Need space for entry and new leaf.
- */
- if (!btp->stale) {
- /*
- * Tag just before the first leaf entry.
- */
- tagp = (__be16 *)blp - 1;
- /*
- * Data object just before the first leaf entry.
- */
- enddup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
- /*
- * If it's not free then can't do this add without cleaning up:
- * the space before the first leaf entry needs to be free so it
- * can be expanded to hold the pointer to the new entry.
- */
- if (be16_to_cpu(enddup->freetag) != XFS_DIR2_DATA_FREE_TAG)
- dup = enddup = NULL;
- /*
- * Check out the biggest freespace and see if it's the same one.
- */
- else {
- dup = (xfs_dir2_data_unused_t *)
- ((char *)hdr + be16_to_cpu(bf[0].offset));
- if (dup == enddup) {
- /*
- * It is the biggest freespace, is it too small
- * to hold the new leaf too?
- */
- if (be16_to_cpu(dup->length) < len + (uint)sizeof(*blp)) {
- /*
- * Yes, we use the second-largest
- * entry instead if it works.
- */
- if (be16_to_cpu(bf[1].length) >= len)
- dup = (xfs_dir2_data_unused_t *)
- ((char *)hdr +
- be16_to_cpu(bf[1].offset));
- else
- dup = NULL;
- }
- } else {
- /*
- * Not the same free entry,
- * just check its length.
- */
- if (be16_to_cpu(dup->length) < len) {
- dup = NULL;
- }
- }
- }
- compact = 0;
- }
- /*
- * If there are stale entries we'll use one for the leaf.
- * Is the biggest entry enough to avoid compaction?
- */
- else if (be16_to_cpu(bf[0].length) >= len) {
- dup = (xfs_dir2_data_unused_t *)
- ((char *)hdr + be16_to_cpu(bf[0].offset));
- compact = 0;
- }
- /*
- * Will need to compact to make this work.
+ * Find out if we can reuse stale entries or whether we need extra
+ * space for entry and new leaf.
*/
- else {
- /*
- * Tag just before the first leaf entry.
- */
- tagp = (__be16 *)blp - 1;
- /*
- * Data object just before the first leaf entry.
- */
- dup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
- /*
- * If it's not free then the data will go where the
- * leaf data starts now, if it works at all.
- */
- if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
- if (be16_to_cpu(dup->length) + (be32_to_cpu(btp->stale) - 1) *
- (uint)sizeof(*blp) < len)
- dup = NULL;
- } else if ((be32_to_cpu(btp->stale) - 1) * (uint)sizeof(*blp) < len)
- dup = NULL;
- else
- dup = (xfs_dir2_data_unused_t *)blp;
- compact = 1;
- }
+ xfs_dir2_block_need_space(hdr, btp, blp, &tagp, &dup,
+ &enddup, &compact, len);
+
/*
- * If this isn't a real add, we're done with the buffer.
+ * Done everything we need for a space check now.
*/
- if (args->op_flags & XFS_DA_OP_JUSTCHECK)
+ if (args->op_flags & XFS_DA_OP_JUSTCHECK) {
xfs_trans_brelse(tp, bp);
+ if (!dup)
+ return XFS_ERROR(ENOSPC);
+ return 0;
+ }
+
/*
* If we don't have space for the new entry & leaf ...
*/
if (!dup) {
- /*
- * Not trying to actually do anything, or don't have
- * a space reservation: return no-space.
- */
- if ((args->op_flags & XFS_DA_OP_JUSTCHECK) || args->total == 0)
+ /* Don't have a space reservation: return no-space. */
+ if (args->total == 0)
return XFS_ERROR(ENOSPC);
/*
* Convert to the next larger format.
@@ -232,65 +313,24 @@ xfs_dir2_block_addname(
return error;
return xfs_dir2_leaf_addname(args);
}
- /*
- * Just checking, and it would work, so say so.
- */
- if (args->op_flags & XFS_DA_OP_JUSTCHECK)
- return 0;
+
needlog = needscan = 0;
+
/*
* If need to compact the leaf entries, do it now.
- * Leave the highest-numbered stale entry stale.
- * XXX should be the one closest to mid but mid is not yet computed.
- */
- if (compact) {
- int fromidx; /* source leaf index */
- int toidx; /* target leaf index */
-
- for (fromidx = toidx = be32_to_cpu(btp->count) - 1,
- highstale = lfloghigh = -1;
- fromidx >= 0;
- fromidx--) {
- if (blp[fromidx].address ==
- cpu_to_be32(XFS_DIR2_NULL_DATAPTR)) {
- if (highstale == -1)
- highstale = toidx;
- else {
- if (lfloghigh == -1)
- lfloghigh = toidx;
- continue;
- }
- }
- if (fromidx < toidx)
- blp[toidx] = blp[fromidx];
- toidx--;
- }
- lfloglow = toidx + 1 - (be32_to_cpu(btp->stale) - 1);
- lfloghigh -= be32_to_cpu(btp->stale) - 1;
- be32_add_cpu(&btp->count, -(be32_to_cpu(btp->stale) - 1));
- xfs_dir2_data_make_free(tp, bp,
- (xfs_dir2_data_aoff_t)((char *)blp - (char *)hdr),
- (xfs_dir2_data_aoff_t)((be32_to_cpu(btp->stale) - 1) * sizeof(*blp)),
- &needlog, &needscan);
- blp += be32_to_cpu(btp->stale) - 1;
- btp->stale = cpu_to_be32(1);
- /*
- * If we now need to rebuild the bestfree map, do so.
- * This needs to happen before the next call to use_free.
- */
- if (needscan) {
- xfs_dir2_data_freescan(mp, hdr, &needlog);
- needscan = 0;
- }
- }
- /*
- * Set leaf logging boundaries to impossible state.
- * For the no-stale case they're set explicitly.
*/
+ if (compact)
+ xfs_dir2_block_compact(tp, bp, hdr, btp, blp, &needlog,
+ &lfloghigh, &lfloglow);
else if (btp->stale) {
+ /*
+ * Set leaf logging boundaries to impossible state.
+ * For the no-stale case they're set explicitly.
+ */
lfloglow = be32_to_cpu(btp->count);
lfloghigh = -1;
}
+
/*
* Find the slot that's first lower than our hash value, -1 if none.
*/
@@ -450,18 +490,13 @@ xfs_dir2_block_getdents(
/*
* If the block number in the offset is out of range, we're done.
*/
- if (xfs_dir2_dataptr_to_db(mp, *offset) > mp->m_dirdatablk) {
+ if (xfs_dir2_dataptr_to_db(mp, *offset) > mp->m_dirdatablk)
return 0;
- }
- /*
- * Can't read the block, give up, else get dabuf in bp.
- */
- error = xfs_da_read_buf(NULL, dp, mp->m_dirdatablk, -1,
- &bp, XFS_DATA_FORK, NULL);
+
+ error = xfs_dir2_block_read(NULL, dp, &bp);
if (error)
return error;
- ASSERT(bp != NULL);
/*
* Extract the byte offset we start at from the seek pointer.
* We'll skip entries before this.
@@ -637,14 +672,11 @@ xfs_dir2_block_lookup_int(
dp = args->dp;
tp = args->trans;
mp = dp->i_mount;
- /*
- * Read the buffer, return error if we can't get it.
- */
- error = xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &bp,
- XFS_DATA_FORK, NULL);
+
+ error = xfs_dir2_block_read(tp, dp, &bp);
if (error)
return error;
- ASSERT(bp != NULL);
+
hdr = bp->b_addr;
xfs_dir2_data_check(dp, bp);
btp = xfs_dir2_block_tail_p(mp, hdr);
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 14/25] xfs: verify dir2 block format buffers
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (12 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 13/25] xfs: factor dir2 block read operations Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 3:26 ` Phil White
2012-10-25 6:34 ` [PATCH 15/25] xfs: factor dir2 free block reading Dave Chinner
` (10 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Add a dir2 block format read verifier. To fully verify every block
when read, call xfs_dir2_data_check() on them. Change
xfs_dir2_data_check() to do runtime checking, convert ASSERT()
checks to XFS_WANT_CORRUPTED_RETURN(), which will trigger an ASSERT
failure on debug kernels, but on production kernels will dump an
error to dmesg and return EFSCORRUPTED to the caller.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_dir2_block.c | 22 +++++++++++++-
fs/xfs/xfs_dir2_data.c | 73 ++++++++++++++++++++++++++++-------------------
fs/xfs/xfs_dir2_priv.h | 4 ++-
3 files changed, 68 insertions(+), 31 deletions(-)
diff --git a/fs/xfs/xfs_dir2_block.c b/fs/xfs/xfs_dir2_block.c
index 25ce409..57351b8 100644
--- a/fs/xfs/xfs_dir2_block.c
+++ b/fs/xfs/xfs_dir2_block.c
@@ -56,6 +56,26 @@ xfs_dir_startup(void)
xfs_dir_hash_dotdot = xfs_da_hashname((unsigned char *)"..", 2);
}
+static void
+xfs_dir2_block_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_dir2_data_hdr *hdr = bp->b_addr;
+ int block_ok = 0;
+
+ block_ok = hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC);
+ block_ok = block_ok && __xfs_dir2_data_check(NULL, bp) == 0;
+
+ if (!block_ok) {
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
static int
xfs_dir2_block_read(
struct xfs_trans *tp,
@@ -65,7 +85,7 @@ xfs_dir2_block_read(
struct xfs_mount *mp = dp->i_mount;
return xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, bpp,
- XFS_DATA_FORK, NULL);
+ XFS_DATA_FORK, xfs_dir2_block_verify);
}
static void
diff --git a/fs/xfs/xfs_dir2_data.c b/fs/xfs/xfs_dir2_data.c
index 44ffd4d..c45107d 100644
--- a/fs/xfs/xfs_dir2_data.c
+++ b/fs/xfs/xfs_dir2_data.c
@@ -34,14 +34,13 @@
STATIC xfs_dir2_data_free_t *
xfs_dir2_data_freefind(xfs_dir2_data_hdr_t *hdr, xfs_dir2_data_unused_t *dup);
-#ifdef DEBUG
/*
* Check the consistency of the data block.
* The input can also be a block-format directory.
- * Pop an assert if we find anything bad.
+ * Return 0 is the buffer is good, otherwise an error.
*/
-void
-xfs_dir2_data_check(
+bool
+__xfs_dir2_data_check(
struct xfs_inode *dp, /* incore inode pointer */
struct xfs_buf *bp) /* data block's buffer */
{
@@ -64,18 +63,23 @@ xfs_dir2_data_check(
int stale; /* count of stale leaves */
struct xfs_name name;
- mp = dp->i_mount;
+ mp = bp->b_target->bt_mount;
hdr = bp->b_addr;
bf = hdr->bestfree;
p = (char *)(hdr + 1);
- if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC)) {
+ switch (hdr->magic) {
+ case cpu_to_be32(XFS_DIR2_BLOCK_MAGIC):
btp = xfs_dir2_block_tail_p(mp, hdr);
lep = xfs_dir2_block_leaf_p(btp);
endp = (char *)lep;
- } else {
- ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC));
+ break;
+ case cpu_to_be32(XFS_DIR2_DATA_MAGIC):
endp = (char *)hdr + mp->m_dirblksize;
+ break;
+ default:
+ XFS_ERROR_REPORT("Bad Magic", XFS_ERRLEVEL_LOW, mp);
+ return EFSCORRUPTED;
}
count = lastfree = freeseen = 0;
@@ -83,19 +87,22 @@ xfs_dir2_data_check(
* Account for zero bestfree entries.
*/
if (!bf[0].length) {
- ASSERT(!bf[0].offset);
+ XFS_WANT_CORRUPTED_RETURN(!bf[0].offset);
freeseen |= 1 << 0;
}
if (!bf[1].length) {
- ASSERT(!bf[1].offset);
+ XFS_WANT_CORRUPTED_RETURN(!bf[1].offset);
freeseen |= 1 << 1;
}
if (!bf[2].length) {
- ASSERT(!bf[2].offset);
+ XFS_WANT_CORRUPTED_RETURN(!bf[2].offset);
freeseen |= 1 << 2;
}
- ASSERT(be16_to_cpu(bf[0].length) >= be16_to_cpu(bf[1].length));
- ASSERT(be16_to_cpu(bf[1].length) >= be16_to_cpu(bf[2].length));
+
+ XFS_WANT_CORRUPTED_RETURN(be16_to_cpu(bf[0].length) >=
+ be16_to_cpu(bf[1].length));
+ XFS_WANT_CORRUPTED_RETURN(be16_to_cpu(bf[1].length) >=
+ be16_to_cpu(bf[2].length));
/*
* Loop over the data/unused entries.
*/
@@ -107,17 +114,20 @@ xfs_dir2_data_check(
* doesn't need to be there.
*/
if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
- ASSERT(lastfree == 0);
- ASSERT(be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup)) ==
- (char *)dup - (char *)hdr);
+ XFS_WANT_CORRUPTED_RETURN(lastfree == 0);
+ XFS_WANT_CORRUPTED_RETURN(
+ be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup)) ==
+ (char *)dup - (char *)hdr);
dfp = xfs_dir2_data_freefind(hdr, dup);
if (dfp) {
i = (int)(dfp - bf);
- ASSERT((freeseen & (1 << i)) == 0);
+ XFS_WANT_CORRUPTED_RETURN(
+ (freeseen & (1 << i)) == 0);
freeseen |= 1 << i;
} else {
- ASSERT(be16_to_cpu(dup->length) <=
- be16_to_cpu(bf[2].length));
+ XFS_WANT_CORRUPTED_RETURN(
+ be16_to_cpu(dup->length) <=
+ be16_to_cpu(bf[2].length));
}
p += be16_to_cpu(dup->length);
lastfree = 1;
@@ -130,10 +140,12 @@ xfs_dir2_data_check(
* The linear search is crude but this is DEBUG code.
*/
dep = (xfs_dir2_data_entry_t *)p;
- ASSERT(dep->namelen != 0);
- ASSERT(xfs_dir_ino_validate(mp, be64_to_cpu(dep->inumber)) == 0);
- ASSERT(be16_to_cpu(*xfs_dir2_data_entry_tag_p(dep)) ==
- (char *)dep - (char *)hdr);
+ XFS_WANT_CORRUPTED_RETURN(dep->namelen != 0);
+ XFS_WANT_CORRUPTED_RETURN(
+ !xfs_dir_ino_validate(mp, be64_to_cpu(dep->inumber)));
+ XFS_WANT_CORRUPTED_RETURN(
+ be16_to_cpu(*xfs_dir2_data_entry_tag_p(dep)) ==
+ (char *)dep - (char *)hdr);
count++;
lastfree = 0;
if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC)) {
@@ -148,27 +160,30 @@ xfs_dir2_data_check(
be32_to_cpu(lep[i].hashval) == hash)
break;
}
- ASSERT(i < be32_to_cpu(btp->count));
+ XFS_WANT_CORRUPTED_RETURN(i < be32_to_cpu(btp->count));
}
p += xfs_dir2_data_entsize(dep->namelen);
}
/*
* Need to have seen all the entries and all the bestfree slots.
*/
- ASSERT(freeseen == 7);
+ XFS_WANT_CORRUPTED_RETURN(freeseen == 7);
if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC)) {
for (i = stale = 0; i < be32_to_cpu(btp->count); i++) {
if (lep[i].address ==
cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
stale++;
if (i > 0)
- ASSERT(be32_to_cpu(lep[i].hashval) >= be32_to_cpu(lep[i - 1].hashval));
+ XFS_WANT_CORRUPTED_RETURN(
+ be32_to_cpu(lep[i].hashval) >=
+ be32_to_cpu(lep[i - 1].hashval));
}
- ASSERT(count == be32_to_cpu(btp->count) - be32_to_cpu(btp->stale));
- ASSERT(stale == be32_to_cpu(btp->stale));
+ XFS_WANT_CORRUPTED_RETURN(count ==
+ be32_to_cpu(btp->count) - be32_to_cpu(btp->stale));
+ XFS_WANT_CORRUPTED_RETURN(stale == be32_to_cpu(btp->stale));
}
+ return 0;
}
-#endif
/*
* Given a data block and an unused entry from that block,
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index 3523d3e..e1c02ca 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -41,10 +41,12 @@ extern int xfs_dir2_leaf_to_block(struct xfs_da_args *args,
/* xfs_dir2_data.c */
#ifdef DEBUG
-extern void xfs_dir2_data_check(struct xfs_inode *dp, struct xfs_buf *bp);
+#define xfs_dir2_data_check(dp,bp) __xfs_dir2_data_check(dp, bp);
#else
#define xfs_dir2_data_check(dp,bp)
#endif
+extern bool __xfs_dir2_data_check(struct xfs_inode *dp, struct xfs_buf *bp);
+
extern struct xfs_dir2_data_free *
xfs_dir2_data_freeinsert(struct xfs_dir2_data_hdr *hdr,
struct xfs_dir2_data_unused *dup, int *loghead);
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 15/25] xfs: factor dir2 free block reading
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (13 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 14/25] xfs: verify dir2 block format buffers Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 13:14 ` Phil White
2012-10-25 6:34 ` [PATCH 16/25] xfs: factor out dir2 data " Dave Chinner
` (9 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Also factor out the updating of the free block when removing entries
from leaf blocks, and add a verifier callback for reads.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_dir2_leaf.c | 3 +-
fs/xfs/xfs_dir2_node.c | 218 +++++++++++++++++++++++++++++++-----------------
fs/xfs/xfs_dir2_priv.h | 2 +
3 files changed, 143 insertions(+), 80 deletions(-)
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 86e3dc1..6c1359d 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -1863,8 +1863,7 @@ xfs_dir2_node_to_leaf(
/*
* Read the freespace block.
*/
- error = xfs_da_read_buf(tp, dp, mp->m_dirfreeblk, -1, &fbp,
- XFS_DATA_FORK, NULL);
+ error = xfs_dir2_free_read(tp, dp, mp->m_dirfreeblk, &fbp);
if (error)
return error;
free = fbp->b_addr;
diff --git a/fs/xfs/xfs_dir2_node.c b/fs/xfs/xfs_dir2_node.c
index 290c2b1..d7f899d 100644
--- a/fs/xfs/xfs_dir2_node.c
+++ b/fs/xfs/xfs_dir2_node.c
@@ -55,6 +55,57 @@ static int xfs_dir2_leafn_remove(xfs_da_args_t *args, struct xfs_buf *bp,
static int xfs_dir2_node_addname_int(xfs_da_args_t *args,
xfs_da_state_blk_t *fblk);
+static void
+xfs_dir2_free_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_dir2_free_hdr *hdr = bp->b_addr;
+ int block_ok = 0;
+
+ block_ok = hdr->magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC);
+ if (!block_ok) {
+ XFS_CORRUPTION_ERROR("xfs_dir2_free_verify magic",
+ XFS_ERRLEVEL_LOW, mp, hdr);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
+static int
+__xfs_dir2_free_read(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ xfs_dablk_t fbno,
+ xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp)
+{
+ return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
+ XFS_DATA_FORK, xfs_dir2_free_verify);
+}
+
+int
+xfs_dir2_free_read(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ xfs_dablk_t fbno,
+ struct xfs_buf **bpp)
+{
+ return __xfs_dir2_free_read(tp, dp, fbno, -1, bpp);
+}
+
+static int
+xfs_dir2_free_try_read(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ xfs_dablk_t fbno,
+ struct xfs_buf **bpp)
+{
+ return __xfs_dir2_free_read(tp, dp, fbno, -2, bpp);
+}
+
/*
* Log entries from a freespace block.
*/
@@ -394,12 +445,10 @@ xfs_dir2_leafn_lookup_for_addname(
*/
if (curbp)
xfs_trans_brelse(tp, curbp);
- /*
- * Read the free block.
- */
- error = xfs_da_read_buf(tp, dp,
+
+ error = xfs_dir2_free_read(tp, dp,
xfs_dir2_db_to_da(mp, newfdb),
- -1, &curbp, XFS_DATA_FORK, NULL);
+ &curbp);
if (error)
return error;
free = curbp->b_addr;
@@ -825,6 +874,77 @@ xfs_dir2_leafn_rebalance(
}
}
+static int
+xfs_dir2_data_block_free(
+ xfs_da_args_t *args,
+ struct xfs_dir2_data_hdr *hdr,
+ struct xfs_dir2_free *free,
+ xfs_dir2_db_t fdb,
+ int findex,
+ struct xfs_buf *fbp,
+ int longest)
+{
+ struct xfs_trans *tp = args->trans;
+ int logfree = 0;
+
+ if (!hdr) {
+ /* One less used entry in the free table. */
+ be32_add_cpu(&free->hdr.nused, -1);
+ xfs_dir2_free_log_header(tp, fbp);
+
+ /*
+ * If this was the last entry in the table, we can trim the
+ * table size back. There might be other entries at the end
+ * referring to non-existent data blocks, get those too.
+ */
+ if (findex == be32_to_cpu(free->hdr.nvalid) - 1) {
+ int i; /* free entry index */
+
+ for (i = findex - 1; i >= 0; i--) {
+ if (free->bests[i] != cpu_to_be16(NULLDATAOFF))
+ break;
+ }
+ free->hdr.nvalid = cpu_to_be32(i + 1);
+ logfree = 0;
+ } else {
+ /* Not the last entry, just punch it out. */
+ free->bests[findex] = cpu_to_be16(NULLDATAOFF);
+ logfree = 1;
+ }
+ /*
+ * If there are no useful entries left in the block,
+ * get rid of the block if we can.
+ */
+ if (!free->hdr.nused) {
+ int error;
+
+ error = xfs_dir2_shrink_inode(args, fdb, fbp);
+ if (error == 0) {
+ fbp = NULL;
+ logfree = 0;
+ } else if (error != ENOSPC || args->total != 0)
+ return error;
+ /*
+ * It's possible to get ENOSPC if there is no
+ * space reservation. In this case some one
+ * else will eventually get rid of this block.
+ */
+ }
+ } else {
+ /*
+ * Data block is not empty, just set the free entry to the new
+ * value.
+ */
+ free->bests[findex] = cpu_to_be16(longest);
+ logfree = 1;
+ }
+
+ /* Log the free entry that changed, unless we got rid of it. */
+ if (logfree)
+ xfs_dir2_free_log_bests(tp, fbp, findex, findex);
+ return 0;
+}
+
/*
* Remove an entry from a node directory.
* This removes the leaf entry and the data entry,
@@ -908,15 +1028,14 @@ xfs_dir2_leafn_remove(
xfs_dir2_db_t fdb; /* freeblock block number */
int findex; /* index in freeblock entries */
xfs_dir2_free_t *free; /* freeblock structure */
- int logfree; /* need to log free entry */
/*
* Convert the data block number to a free block,
* read in the free block.
*/
fdb = xfs_dir2_db_to_fdb(mp, db);
- error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, fdb),
- -1, &fbp, XFS_DATA_FORK, NULL);
+ error = xfs_dir2_free_read(tp, dp, xfs_dir2_db_to_da(mp, fdb),
+ &fbp);
if (error)
return error;
free = fbp->b_addr;
@@ -954,68 +1073,12 @@ xfs_dir2_leafn_remove(
* If we got rid of the data block, we can eliminate that entry
* in the free block.
*/
- if (hdr == NULL) {
- /*
- * One less used entry in the free table.
- */
- be32_add_cpu(&free->hdr.nused, -1);
- xfs_dir2_free_log_header(tp, fbp);
- /*
- * If this was the last entry in the table, we can
- * trim the table size back. There might be other
- * entries at the end referring to non-existent
- * data blocks, get those too.
- */
- if (findex == be32_to_cpu(free->hdr.nvalid) - 1) {
- int i; /* free entry index */
-
- for (i = findex - 1;
- i >= 0 &&
- free->bests[i] == cpu_to_be16(NULLDATAOFF);
- i--)
- continue;
- free->hdr.nvalid = cpu_to_be32(i + 1);
- logfree = 0;
- }
- /*
- * Not the last entry, just punch it out.
- */
- else {
- free->bests[findex] = cpu_to_be16(NULLDATAOFF);
- logfree = 1;
- }
- /*
- * If there are no useful entries left in the block,
- * get rid of the block if we can.
- */
- if (!free->hdr.nused) {
- error = xfs_dir2_shrink_inode(args, fdb, fbp);
- if (error == 0) {
- fbp = NULL;
- logfree = 0;
- } else if (error != ENOSPC || args->total != 0)
- return error;
- /*
- * It's possible to get ENOSPC if there is no
- * space reservation. In this case some one
- * else will eventually get rid of this block.
- */
- }
- }
- /*
- * Data block is not empty, just set the free entry to
- * the new value.
- */
- else {
- free->bests[findex] = cpu_to_be16(longest);
- logfree = 1;
- }
- /*
- * Log the free entry that changed, unless we got rid of it.
- */
- if (logfree)
- xfs_dir2_free_log_bests(tp, fbp, findex, findex);
+ error = xfs_dir2_data_block_free(args, hdr, free,
+ fdb, findex, fbp, longest);
+ if (error)
+ return error;
}
+
xfs_dir2_leafn_check(dp, bp);
/*
* Return indication of whether this leaf block is empty enough
@@ -1453,9 +1516,9 @@ xfs_dir2_node_addname_int(
* This should be really rare, so there's no reason
* to avoid it.
*/
- error = xfs_da_read_buf(tp, dp,
- xfs_dir2_db_to_da(mp, fbno), -2,
- &fbp, XFS_DATA_FORK, NULL);
+ error = xfs_dir2_free_try_read(tp, dp,
+ xfs_dir2_db_to_da(mp, fbno),
+ &fbp);
if (error)
return error;
if (!fbp)
@@ -1518,8 +1581,9 @@ xfs_dir2_node_addname_int(
* that was just allocated.
*/
fbno = xfs_dir2_db_to_fdb(mp, dbno);
- error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, fbno), -2,
- &fbp, XFS_DATA_FORK, NULL);
+ error = xfs_dir2_free_try_read(tp, dp,
+ xfs_dir2_db_to_da(mp, fbno),
+ &fbp);
if (error)
return error;
@@ -1915,17 +1979,15 @@ xfs_dir2_node_trim_free(
/*
* Read the freespace block.
*/
- error = xfs_da_read_buf(tp, dp, (xfs_dablk_t)fo, -2, &bp,
- XFS_DATA_FORK, NULL);
+ error = xfs_dir2_free_try_read(tp, dp, fo, &bp);
if (error)
return error;
/*
* There can be holes in freespace. If fo is a hole, there's
* nothing to do.
*/
- if (bp == NULL) {
+ if (!bp)
return 0;
- }
free = bp->b_addr;
ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC));
/*
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index e1c02ca..91d936b 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -117,6 +117,8 @@ extern int xfs_dir2_node_removename(struct xfs_da_args *args);
extern int xfs_dir2_node_replace(struct xfs_da_args *args);
extern int xfs_dir2_node_trim_free(struct xfs_da_args *args, xfs_fileoff_t fo,
int *rvalp);
+extern int xfs_dir2_free_read(struct xfs_trans *tp, struct xfs_inode *dp,
+ xfs_dablk_t fbno, struct xfs_buf **bpp);
/* xfs_dir2_sf.c */
extern xfs_ino_t xfs_dir2_sf_get_parent_ino(struct xfs_dir2_sf_hdr *sfp);
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 16/25] xfs: factor out dir2 data block reading
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (14 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 15/25] xfs: factor dir2 free block reading Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 13:21 ` Phil White
2012-10-25 6:34 ` [PATCH 17/25] xfs: factor dir2 leaf read Dave Chinner
` (8 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
And add a verifier callback function while there.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_dir2_block.c | 3 +--
fs/xfs/xfs_dir2_data.c | 32 ++++++++++++++++++++++++++++++++
fs/xfs/xfs_dir2_leaf.c | 38 +++++++++++++++++---------------------
fs/xfs/xfs_dir2_node.c | 8 ++++----
fs/xfs/xfs_dir2_priv.h | 2 ++
5 files changed, 56 insertions(+), 27 deletions(-)
diff --git a/fs/xfs/xfs_dir2_block.c b/fs/xfs/xfs_dir2_block.c
index 57351b8..ca03b10 100644
--- a/fs/xfs/xfs_dir2_block.c
+++ b/fs/xfs/xfs_dir2_block.c
@@ -970,8 +970,7 @@ xfs_dir2_leaf_to_block(
* Read the data block if we don't already have it, give up if it fails.
*/
if (!dbp) {
- error = xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &dbp,
- XFS_DATA_FORK, NULL);
+ error = xfs_dir2_data_read(tp, dp, mp->m_dirdatablk, -1, &dbp);
if (error)
return error;
}
diff --git a/fs/xfs/xfs_dir2_data.c b/fs/xfs/xfs_dir2_data.c
index c45107d..43c8426 100644
--- a/fs/xfs/xfs_dir2_data.c
+++ b/fs/xfs/xfs_dir2_data.c
@@ -185,6 +185,38 @@ __xfs_dir2_data_check(
return 0;
}
+static void
+xfs_dir2_data_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_dir2_data_hdr *hdr = bp->b_addr;
+ int block_ok = 0;
+
+ block_ok = hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC);
+ block_ok = block_ok && __xfs_dir2_data_check(NULL, bp) == 0;
+
+ if (!block_ok) {
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
+int
+xfs_dir2_data_read(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ xfs_dablk_t bno,
+ xfs_daddr_t mapped_bno,
+ struct xfs_buf **bpp)
+{
+ return xfs_da_read_buf(tp, dp, bno, mapped_bno, bpp,
+ XFS_DATA_FORK, xfs_dir2_data_verify);
+}
+
/*
* Given a data block and an unused entry from that block,
* return the bestfree entry if any that corresponds to it.
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 6c1359d..0fdf765 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -493,14 +493,14 @@ xfs_dir2_leaf_addname(
hdr = dbp->b_addr;
bestsp[use_block] = hdr->bestfree[0].length;
grown = 1;
- }
- /*
- * Already had space in some data block.
- * Just read that one in.
- */
- else {
- error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, use_block),
- -1, &dbp, XFS_DATA_FORK, NULL);
+ } else {
+ /*
+ * Already had space in some data block.
+ * Just read that one in.
+ */
+ error = xfs_dir2_data_read(tp, dp,
+ xfs_dir2_db_to_da(mp, use_block),
+ -1, &dbp);
if (error) {
xfs_trans_brelse(tp, lbp);
return error;
@@ -508,7 +508,6 @@ xfs_dir2_leaf_addname(
hdr = dbp->b_addr;
grown = 0;
}
- xfs_dir2_data_check(dp, dbp);
/*
* Point to the biggest freespace in our data block.
*/
@@ -891,10 +890,9 @@ xfs_dir2_leaf_readbuf(
* Read the directory block starting at the first mapping.
*/
mip->curdb = xfs_dir2_da_to_db(mp, map->br_startoff);
- error = xfs_da_read_buf(NULL, dp, map->br_startoff,
+ error = xfs_dir2_data_read(NULL, dp, map->br_startoff,
map->br_blockcount >= mp->m_dirblkfsbs ?
- XFS_FSB_TO_DADDR(mp, map->br_startblock) : -1,
- &bp, XFS_DATA_FORK, NULL);
+ XFS_FSB_TO_DADDR(mp, map->br_startblock) : -1, &bp);
/*
* Should just skip over the data block instead of giving up.
@@ -1408,14 +1406,13 @@ xfs_dir2_leaf_lookup_int(
if (newdb != curdb) {
if (dbp)
xfs_trans_brelse(tp, dbp);
- error = xfs_da_read_buf(tp, dp,
- xfs_dir2_db_to_da(mp, newdb),
- -1, &dbp, XFS_DATA_FORK, NULL);
+ error = xfs_dir2_data_read(tp, dp,
+ xfs_dir2_db_to_da(mp, newdb),
+ -1, &dbp);
if (error) {
xfs_trans_brelse(tp, lbp);
return error;
}
- xfs_dir2_data_check(dp, dbp);
curdb = newdb;
}
/*
@@ -1450,9 +1447,9 @@ xfs_dir2_leaf_lookup_int(
ASSERT(cidb != -1);
if (cidb != curdb) {
xfs_trans_brelse(tp, dbp);
- error = xfs_da_read_buf(tp, dp,
- xfs_dir2_db_to_da(mp, cidb),
- -1, &dbp, XFS_DATA_FORK, NULL);
+ error = xfs_dir2_data_read(tp, dp,
+ xfs_dir2_db_to_da(mp, cidb),
+ -1, &dbp);
if (error) {
xfs_trans_brelse(tp, lbp);
return error;
@@ -1737,8 +1734,7 @@ xfs_dir2_leaf_trim_data(
/*
* Read the offending data block. We need its buffer.
*/
- error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, db), -1, &dbp,
- XFS_DATA_FORK, NULL);
+ error = xfs_dir2_data_read(tp, dp, xfs_dir2_db_to_da(mp, db), -1, &dbp);
if (error)
return error;
diff --git a/fs/xfs/xfs_dir2_node.c b/fs/xfs/xfs_dir2_node.c
index d7f899d..67b811c 100644
--- a/fs/xfs/xfs_dir2_node.c
+++ b/fs/xfs/xfs_dir2_node.c
@@ -583,9 +583,9 @@ xfs_dir2_leafn_lookup_for_entry(
ASSERT(state->extravalid);
curbp = state->extrablk.bp;
} else {
- error = xfs_da_read_buf(tp, dp,
+ error = xfs_dir2_data_read(tp, dp,
xfs_dir2_db_to_da(mp, newdb),
- -1, &curbp, XFS_DATA_FORK, NULL);
+ -1, &curbp);
if (error)
return error;
}
@@ -1692,8 +1692,8 @@ xfs_dir2_node_addname_int(
/*
* Read the data block in.
*/
- error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, dbno),
- -1, &dbp, XFS_DATA_FORK, NULL);
+ error = xfs_dir2_data_read(tp, dp, xfs_dir2_db_to_da(mp, dbno),
+ -1, &dbp);
if (error)
return error;
hdr = dbp->b_addr;
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index 91d936b..7d8c302a 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -46,6 +46,8 @@ extern int xfs_dir2_leaf_to_block(struct xfs_da_args *args,
#define xfs_dir2_data_check(dp,bp)
#endif
extern bool __xfs_dir2_data_check(struct xfs_inode *dp, struct xfs_buf *bp);
+extern int xfs_dir2_data_read(struct xfs_trans *tp, struct xfs_inode *dp,
+ xfs_dablk_t bno, xfs_daddr_t mapped_bno, struct xfs_buf **bpp);
extern struct xfs_dir2_data_free *
xfs_dir2_data_freeinsert(struct xfs_dir2_data_hdr *hdr,
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 17/25] xfs: factor dir2 leaf read
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (15 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 16/25] xfs: factor out dir2 data " Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 13:22 ` Phil White
2012-10-25 6:34 ` [PATCH 18/25] xfs: factor and verify attr leaf reads Dave Chinner
` (7 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_dir2_leaf.c | 73 ++++++++++++++++++++++++++++++++++++++++--------
fs/xfs/xfs_dir2_node.c | 6 ++--
fs/xfs/xfs_dir2_priv.h | 2 ++
3 files changed, 67 insertions(+), 14 deletions(-)
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 0fdf765..97408e3 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -48,6 +48,62 @@ static void xfs_dir2_leaf_log_bests(struct xfs_trans *tp, struct xfs_buf *bp,
int first, int last);
static void xfs_dir2_leaf_log_tail(struct xfs_trans *tp, struct xfs_buf *bp);
+static void
+xfs_dir2_leaf_verify(
+ struct xfs_buf *bp,
+ __be16 magic)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_dir2_leaf_hdr *hdr = bp->b_addr;
+ int block_ok = 0;
+
+ block_ok = hdr->info.magic == magic;
+ if (!block_ok) {
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
+static void
+xfs_dir2_leaf1_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
+}
+
+static void
+xfs_dir2_leafn_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
+}
+
+static int
+xfs_dir2_leaf_read(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ xfs_dablk_t fbno,
+ xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp)
+{
+ return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
+ XFS_DATA_FORK, xfs_dir2_leaf1_verify);
+}
+
+int
+xfs_dir2_leafn_read(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ xfs_dablk_t fbno,
+ xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp)
+{
+ return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
+ XFS_DATA_FORK, xfs_dir2_leafn_verify);
+}
/*
* Convert a block form directory to a leaf form directory.
@@ -311,14 +367,11 @@ xfs_dir2_leaf_addname(
dp = args->dp;
tp = args->trans;
mp = dp->i_mount;
- /*
- * Read the leaf block.
- */
- error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
- XFS_DATA_FORK, NULL);
+
+ error = xfs_dir2_leaf_read(tp, dp, mp->m_dirleafblk, -1, &lbp);
if (error)
return error;
- ASSERT(lbp != NULL);
+
/*
* Look up the entry by hash value and name.
* We know it's not there, our caller has already done a lookup.
@@ -1369,13 +1422,11 @@ xfs_dir2_leaf_lookup_int(
dp = args->dp;
tp = args->trans;
mp = dp->i_mount;
- /*
- * Read the leaf block into the buffer.
- */
- error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
- XFS_DATA_FORK, NULL);
+
+ error = xfs_dir2_leaf_read(tp, dp, mp->m_dirleafblk, -1, &lbp);
if (error)
return error;
+
*lbpp = lbp;
leaf = lbp->b_addr;
xfs_dir2_leaf_check(dp, lbp);
diff --git a/fs/xfs/xfs_dir2_node.c b/fs/xfs/xfs_dir2_node.c
index 67b811c..7c6f956 100644
--- a/fs/xfs/xfs_dir2_node.c
+++ b/fs/xfs/xfs_dir2_node.c
@@ -1232,11 +1232,11 @@ xfs_dir2_leafn_toosmall(
/*
* Read the sibling leaf block.
*/
- error = xfs_da_read_buf(state->args->trans, state->args->dp,
- blkno, -1, &bp, XFS_DATA_FORK, NULL);
+ error = xfs_dir2_leafn_read(state->args->trans, state->args->dp,
+ blkno, -1, &bp);
if (error)
return error;
- ASSERT(bp != NULL);
+
/*
* Count bytes in the two blocks combined.
*/
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index 7d8c302a..ecf75d9 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -70,6 +70,8 @@ extern void xfs_dir2_data_use_free(struct xfs_trans *tp, struct xfs_buf *bp,
xfs_dir2_data_aoff_t len, int *needlogp, int *needscanp);
/* xfs_dir2_leaf.c */
+extern int xfs_dir2_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,
struct xfs_buf *dbp);
extern int xfs_dir2_leaf_addname(struct xfs_da_args *args);
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 18/25] xfs: factor and verify attr leaf reads
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (16 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 17/25] xfs: factor dir2 leaf read Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 13:26 ` Phil White
2012-10-25 6:34 ` [PATCH 19/25] xfs: add xfs_da_node verification Dave Chinner
` (6 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Some reads are not converted yet because it isn't obvious ahead of
time what the format of the block is going to be. Need to determine
how to tell if the first block in the tree is a node or leaf format
block. That will be done in later patches.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_attr.c | 70 +++++++++++--------------------------------
fs/xfs/xfs_attr_leaf.c | 78 ++++++++++++++++++++++++++++--------------------
fs/xfs/xfs_attr_leaf.h | 3 ++
3 files changed, 66 insertions(+), 85 deletions(-)
diff --git a/fs/xfs/xfs_attr.c b/fs/xfs/xfs_attr.c
index 956c2ba..548e910 100644
--- a/fs/xfs/xfs_attr.c
+++ b/fs/xfs/xfs_attr.c
@@ -903,11 +903,9 @@ xfs_attr_leaf_addname(xfs_da_args_t *args)
*/
dp = args->dp;
args->blkno = 0;
- error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
- XFS_ATTR_FORK, NULL);
+ error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp);
if (error)
- return(error);
- ASSERT(bp != NULL);
+ return error;
/*
* Look up the given attribute in the leaf block. Figure out if
@@ -1031,12 +1029,12 @@ xfs_attr_leaf_addname(xfs_da_args_t *args)
* Read in the block containing the "old" attr, then
* remove the "old" attr from that block (neat, huh!)
*/
- error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1,
- &bp, XFS_ATTR_FORK, NULL);
+ error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno,
+ -1, &bp);
if (error)
- return(error);
- ASSERT(bp != NULL);
- (void)xfs_attr_leaf_remove(bp, args);
+ return error;
+
+ xfs_attr_leaf_remove(bp, args);
/*
* If the result is small enough, shrink it all into the inode.
@@ -1100,20 +1098,17 @@ xfs_attr_leaf_removename(xfs_da_args_t *args)
*/
dp = args->dp;
args->blkno = 0;
- error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
- XFS_ATTR_FORK, NULL);
- if (error) {
- return(error);
- }
+ error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp);
+ if (error)
+ return error;
- ASSERT(bp != NULL);
error = xfs_attr_leaf_lookup_int(bp, args);
if (error == ENOATTR) {
xfs_trans_brelse(args->trans, bp);
return(error);
}
- (void)xfs_attr_leaf_remove(bp, args);
+ xfs_attr_leaf_remove(bp, args);
/*
* If the result is small enough, shrink it all into the inode.
@@ -1156,11 +1151,9 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
int error;
args->blkno = 0;
- error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
- XFS_ATTR_FORK, NULL);
+ error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp);
if (error)
- return(error);
- ASSERT(bp != NULL);
+ return error;
error = xfs_attr_leaf_lookup_int(bp, args);
if (error != EEXIST) {
@@ -1181,23 +1174,13 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
STATIC int
xfs_attr_leaf_list(xfs_attr_list_context_t *context)
{
- xfs_attr_leafblock_t *leaf;
int error;
struct xfs_buf *bp;
context->cursor->blkno = 0;
- error = xfs_da_read_buf(NULL, context->dp, 0, -1, &bp, XFS_ATTR_FORK,
- NULL);
+ error = xfs_attr_leaf_read(NULL, context->dp, 0, -1, &bp);
if (error)
return XFS_ERROR(error);
- ASSERT(bp != NULL);
- leaf = bp->b_addr;
- if (unlikely(leaf->hdr.info.magic != cpu_to_be16(XFS_ATTR_LEAF_MAGIC))) {
- XFS_CORRUPTION_ERROR("xfs_attr_leaf_list", XFS_ERRLEVEL_LOW,
- context->dp->i_mount, leaf);
- xfs_trans_brelse(NULL, bp);
- return XFS_ERROR(EFSCORRUPTED);
- }
error = xfs_attr_leaf_list_int(bp, context);
xfs_trans_brelse(NULL, bp);
@@ -1601,12 +1584,9 @@ xfs_attr_node_removename(xfs_da_args_t *args)
ASSERT(state->path.blk[0].bp);
state->path.blk[0].bp = NULL;
- error = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp,
- XFS_ATTR_FORK, NULL);
+ error = xfs_attr_leaf_read(args->trans, args->dp, 0, -1, &bp);
if (error)
goto out;
- ASSERT((((xfs_attr_leafblock_t *)bp->b_addr)->hdr.info.magic) ==
- cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
if ((forkoff = xfs_attr_shortform_allfit(bp, dp))) {
xfs_bmap_init(args->flist, args->firstblock);
@@ -1908,14 +1888,6 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
*/
for (;;) {
leaf = bp->b_addr;
- if (unlikely(leaf->hdr.info.magic !=
- cpu_to_be16(XFS_ATTR_LEAF_MAGIC))) {
- XFS_CORRUPTION_ERROR("xfs_attr_node_list(4)",
- XFS_ERRLEVEL_LOW,
- context->dp->i_mount, leaf);
- xfs_trans_brelse(NULL, bp);
- return(XFS_ERROR(EFSCORRUPTED));
- }
error = xfs_attr_leaf_list_int(bp, context);
if (error) {
xfs_trans_brelse(NULL, bp);
@@ -1925,16 +1897,10 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
break;
cursor->blkno = be32_to_cpu(leaf->hdr.info.forw);
xfs_trans_brelse(NULL, bp);
- error = xfs_da_read_buf(NULL, context->dp, cursor->blkno, -1,
- &bp, XFS_ATTR_FORK, NULL);
+ error = xfs_attr_leaf_read(NULL, context->dp, cursor->blkno, -1,
+ &bp);
if (error)
- return(error);
- if (unlikely((bp == NULL))) {
- XFS_ERROR_REPORT("xfs_attr_node_list(5)",
- XFS_ERRLEVEL_LOW,
- context->dp->i_mount);
- return(XFS_ERROR(EFSCORRUPTED));
- }
+ return error;
}
xfs_trans_brelse(NULL, bp);
return(0);
diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c
index f2b698e..7891d06 100644
--- a/fs/xfs/xfs_attr_leaf.c
+++ b/fs/xfs/xfs_attr_leaf.c
@@ -87,6 +87,36 @@ STATIC void xfs_attr_leaf_moveents(xfs_attr_leafblock_t *src_leaf,
xfs_mount_t *mp);
STATIC int xfs_attr_leaf_entsize(xfs_attr_leafblock_t *leaf, int index);
+static void
+xfs_attr_leaf_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_attr_leaf_hdr *hdr = bp->b_addr;
+ int block_ok = 0;
+
+ block_ok = hdr->info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC);
+ if (!block_ok) {
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
+int
+xfs_attr_leaf_read(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ xfs_dablk_t bno,
+ xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp)
+{
+ return xfs_da_read_buf(tp, dp, bno, mappedbno, bpp,
+ XFS_ATTR_FORK, xfs_attr_leaf_verify);
+}
+
/*========================================================================
* Namespace helper routines
*========================================================================*/
@@ -869,11 +899,10 @@ xfs_attr_leaf_to_node(xfs_da_args_t *args)
error = xfs_da_grow_inode(args, &blkno);
if (error)
goto out;
- error = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp1,
- XFS_ATTR_FORK, NULL);
+ error = xfs_attr_leaf_read(args->trans, args->dp, 0, -1, &bp1);
if (error)
goto out;
- ASSERT(bp1 != NULL);
+
bp2 = NULL;
error = xfs_da_get_buf(args->trans, args->dp, blkno, -1, &bp2,
XFS_ATTR_FORK);
@@ -1620,18 +1649,16 @@ xfs_attr_leaf_toosmall(xfs_da_state_t *state, int *action)
blkno = be32_to_cpu(info->back);
if (blkno == 0)
continue;
- error = xfs_da_read_buf(state->args->trans, state->args->dp,
- blkno, -1, &bp, XFS_ATTR_FORK, NULL);
+ error = xfs_attr_leaf_read(state->args->trans, state->args->dp,
+ blkno, -1, &bp);
if (error)
return(error);
- ASSERT(bp != NULL);
leaf = (xfs_attr_leafblock_t *)info;
count = be16_to_cpu(leaf->hdr.count);
bytes = state->blocksize - (state->blocksize>>2);
bytes -= be16_to_cpu(leaf->hdr.usedbytes);
leaf = bp->b_addr;
- ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
count += be16_to_cpu(leaf->hdr.count);
bytes -= be16_to_cpu(leaf->hdr.usedbytes);
bytes -= count * sizeof(xfs_attr_leaf_entry_t);
@@ -2495,15 +2522,11 @@ xfs_attr_leaf_clearflag(xfs_da_args_t *args)
/*
* Set up the operation.
*/
- error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
- XFS_ATTR_FORK, NULL);
- if (error) {
+ error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp);
+ if (error)
return(error);
- }
- ASSERT(bp != NULL);
leaf = bp->b_addr;
- ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
ASSERT(args->index < be16_to_cpu(leaf->hdr.count));
ASSERT(args->index >= 0);
entry = &leaf->entries[ args->index ];
@@ -2560,15 +2583,11 @@ xfs_attr_leaf_setflag(xfs_da_args_t *args)
/*
* Set up the operation.
*/
- error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
- XFS_ATTR_FORK, NULL);
- if (error) {
+ error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp);
+ if (error)
return(error);
- }
- ASSERT(bp != NULL);
leaf = bp->b_addr;
- ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
ASSERT(args->index < be16_to_cpu(leaf->hdr.count));
ASSERT(args->index >= 0);
entry = &leaf->entries[ args->index ];
@@ -2617,35 +2636,28 @@ xfs_attr_leaf_flipflags(xfs_da_args_t *args)
/*
* Read the block containing the "old" attr
*/
- error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp1,
- XFS_ATTR_FORK, NULL);
- if (error) {
- return(error);
- }
- ASSERT(bp1 != NULL);
+ error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp1);
+ if (error)
+ return error;
/*
* Read the block containing the "new" attr, if it is different
*/
if (args->blkno2 != args->blkno) {
- error = xfs_da_read_buf(args->trans, args->dp, args->blkno2,
- -1, &bp2, XFS_ATTR_FORK, NULL);
- if (error) {
- return(error);
- }
- ASSERT(bp2 != NULL);
+ error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno2,
+ -1, &bp2);
+ if (error)
+ return error;
} else {
bp2 = bp1;
}
leaf1 = bp1->b_addr;
- ASSERT(leaf1->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
ASSERT(args->index < be16_to_cpu(leaf1->hdr.count));
ASSERT(args->index >= 0);
entry1 = &leaf1->entries[ args->index ];
leaf2 = bp2->b_addr;
- ASSERT(leaf2->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
ASSERT(args->index2 < be16_to_cpu(leaf2->hdr.count));
ASSERT(args->index2 >= 0);
entry2 = &leaf2->entries[ args->index2 ];
diff --git a/fs/xfs/xfs_attr_leaf.h b/fs/xfs/xfs_attr_leaf.h
index dea1772..8f7ab98 100644
--- a/fs/xfs/xfs_attr_leaf.h
+++ b/fs/xfs/xfs_attr_leaf.h
@@ -227,6 +227,9 @@ int xfs_attr_leaf_to_shortform(struct xfs_buf *bp,
int xfs_attr_leaf_clearflag(struct xfs_da_args *args);
int xfs_attr_leaf_setflag(struct xfs_da_args *args);
int xfs_attr_leaf_flipflags(xfs_da_args_t *args);
+int xfs_attr_leaf_read(struct xfs_trans *tp, struct xfs_inode *dp,
+ xfs_dablk_t bno, xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp);
/*
* Routines used for growing the Btree.
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 19/25] xfs: add xfs_da_node verification
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (17 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 18/25] xfs: factor and verify attr leaf reads Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 13:30 ` Phil White
2012-10-25 6:34 ` [PATCH 20/25] xfs: Add verifiers to dir2 data readahead Dave Chinner
` (5 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_attr.c | 22 ++++------
fs/xfs/xfs_attr_leaf.c | 12 +++---
fs/xfs/xfs_attr_leaf.h | 8 ++--
fs/xfs/xfs_da_btree.c | 108 ++++++++++++++++++++++++++++++++++++------------
fs/xfs/xfs_da_btree.h | 3 ++
fs/xfs/xfs_dir2_leaf.c | 2 +-
fs/xfs/xfs_dir2_priv.h | 1 +
7 files changed, 106 insertions(+), 50 deletions(-)
diff --git a/fs/xfs/xfs_attr.c b/fs/xfs/xfs_attr.c
index 548e910..4b862ed 100644
--- a/fs/xfs/xfs_attr.c
+++ b/fs/xfs/xfs_attr.c
@@ -1688,10 +1688,10 @@ xfs_attr_refillstate(xfs_da_state_t *state)
ASSERT((path->active >= 0) && (path->active < XFS_DA_NODE_MAXDEPTH));
for (blk = path->blk, level = 0; level < path->active; blk++, level++) {
if (blk->disk_blkno) {
- error = xfs_da_read_buf(state->args->trans,
+ error = xfs_da_node_read(state->args->trans,
state->args->dp,
blk->blkno, blk->disk_blkno,
- &blk->bp, XFS_ATTR_FORK, NULL);
+ &blk->bp, XFS_ATTR_FORK);
if (error)
return(error);
} else {
@@ -1707,10 +1707,10 @@ xfs_attr_refillstate(xfs_da_state_t *state)
ASSERT((path->active >= 0) && (path->active < XFS_DA_NODE_MAXDEPTH));
for (blk = path->blk, level = 0; level < path->active; blk++, level++) {
if (blk->disk_blkno) {
- error = xfs_da_read_buf(state->args->trans,
+ error = xfs_da_node_read(state->args->trans,
state->args->dp,
blk->blkno, blk->disk_blkno,
- &blk->bp, XFS_ATTR_FORK, NULL);
+ &blk->bp, XFS_ATTR_FORK);
if (error)
return(error);
} else {
@@ -1795,8 +1795,8 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
*/
bp = NULL;
if (cursor->blkno > 0) {
- error = xfs_da_read_buf(NULL, context->dp, cursor->blkno, -1,
- &bp, XFS_ATTR_FORK, NULL);
+ error = xfs_da_node_read(NULL, context->dp, cursor->blkno, -1,
+ &bp, XFS_ATTR_FORK);
if ((error != 0) && (error != EFSCORRUPTED))
return(error);
if (bp) {
@@ -1837,17 +1837,11 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
if (bp == NULL) {
cursor->blkno = 0;
for (;;) {
- error = xfs_da_read_buf(NULL, context->dp,
+ error = xfs_da_node_read(NULL, context->dp,
cursor->blkno, -1, &bp,
- XFS_ATTR_FORK, NULL);
+ XFS_ATTR_FORK);
if (error)
return(error);
- if (unlikely(bp == NULL)) {
- XFS_ERROR_REPORT("xfs_attr_node_list(2)",
- XFS_ERRLEVEL_LOW,
- context->dp->i_mount);
- return(XFS_ERROR(EFSCORRUPTED));
- }
node = bp->b_addr;
if (node->hdr.info.magic ==
cpu_to_be16(XFS_ATTR_LEAF_MAGIC))
diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c
index 7891d06..5ba92eb 100644
--- a/fs/xfs/xfs_attr_leaf.c
+++ b/fs/xfs/xfs_attr_leaf.c
@@ -87,7 +87,7 @@ STATIC void xfs_attr_leaf_moveents(xfs_attr_leafblock_t *src_leaf,
xfs_mount_t *mp);
STATIC int xfs_attr_leaf_entsize(xfs_attr_leafblock_t *leaf, int index);
-static void
+void
xfs_attr_leaf_verify(
struct xfs_buf *bp)
{
@@ -2742,7 +2742,7 @@ xfs_attr_root_inactive(xfs_trans_t **trans, xfs_inode_t *dp)
* the extents in reverse order the extent containing
* block 0 must still be there.
*/
- error = xfs_da_read_buf(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK, NULL);
+ error = xfs_da_node_read(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK);
if (error)
return(error);
blkno = XFS_BUF_ADDR(bp);
@@ -2827,8 +2827,8 @@ xfs_attr_node_inactive(
* traversal of the tree so we may deal with many blocks
* before we come back to this one.
*/
- error = xfs_da_read_buf(*trans, dp, child_fsb, -2, &child_bp,
- XFS_ATTR_FORK, NULL);
+ error = xfs_da_node_read(*trans, dp, child_fsb, -2, &child_bp,
+ XFS_ATTR_FORK);
if (error)
return(error);
if (child_bp) {
@@ -2868,8 +2868,8 @@ xfs_attr_node_inactive(
* child block number.
*/
if ((i+1) < count) {
- error = xfs_da_read_buf(*trans, dp, 0, parent_blkno,
- &bp, XFS_ATTR_FORK, NULL);
+ error = xfs_da_node_read(*trans, dp, 0, parent_blkno,
+ &bp, XFS_ATTR_FORK);
if (error)
return(error);
child_fsb = be32_to_cpu(node->btree[i+1].before);
diff --git a/fs/xfs/xfs_attr_leaf.h b/fs/xfs/xfs_attr_leaf.h
index 8f7ab98..098e9a5 100644
--- a/fs/xfs/xfs_attr_leaf.h
+++ b/fs/xfs/xfs_attr_leaf.h
@@ -227,9 +227,6 @@ int xfs_attr_leaf_to_shortform(struct xfs_buf *bp,
int xfs_attr_leaf_clearflag(struct xfs_da_args *args);
int xfs_attr_leaf_setflag(struct xfs_da_args *args);
int xfs_attr_leaf_flipflags(xfs_da_args_t *args);
-int xfs_attr_leaf_read(struct xfs_trans *tp, struct xfs_inode *dp,
- xfs_dablk_t bno, xfs_daddr_t mappedbno,
- struct xfs_buf **bpp);
/*
* Routines used for growing the Btree.
@@ -264,4 +261,9 @@ int xfs_attr_leaf_order(struct xfs_buf *leaf1_bp,
struct xfs_buf *leaf2_bp);
int xfs_attr_leaf_newentsize(int namelen, int valuelen, int blocksize,
int *local);
+int xfs_attr_leaf_read(struct xfs_trans *tp, struct xfs_inode *dp,
+ xfs_dablk_t bno, xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp);
+void xfs_attr_leaf_verify(struct xfs_buf *bp);
+
#endif /* __XFS_ATTR_LEAF_H__ */
diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
index a46035b..e950192 100644
--- a/fs/xfs/xfs_da_btree.c
+++ b/fs/xfs/xfs_da_btree.c
@@ -91,6 +91,67 @@ STATIC int xfs_da_blk_unlink(xfs_da_state_t *state,
xfs_da_state_blk_t *save_blk);
STATIC void xfs_da_state_kill_altpath(xfs_da_state_t *state);
+static void
+__xfs_da_node_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_da_node_hdr *hdr = bp->b_addr;
+ int block_ok = 0;
+
+ block_ok = hdr->info.magic == cpu_to_be16(XFS_DA_NODE_MAGIC);
+ block_ok |= hdr->level > 0;
+ block_ok |= hdr->count > 0;
+ if (!block_ok) {
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
+static void
+xfs_da_node_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_da_blkinfo *info = bp->b_addr;
+
+ switch (be16_to_cpu(info->magic)) {
+ case XFS_DA_NODE_MAGIC:
+ __xfs_da_node_verify(bp);
+ return;
+ case XFS_ATTR_LEAF_MAGIC:
+ xfs_attr_leaf_verify(bp);
+ return;
+ case XFS_DIR2_LEAFN_MAGIC:
+ xfs_dir2_leafn_verify(bp);
+ return;
+ default:
+ break;
+ }
+
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, info);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
+int
+xfs_da_node_read(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ xfs_dablk_t bno,
+ xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp,
+ int which_fork)
+{
+ return xfs_da_read_buf(tp, dp, bno, mappedbno, bpp,
+ which_fork, xfs_da_node_verify);
+}
+
/*========================================================================
* Routines used for growing the Btree.
*========================================================================*/
@@ -746,8 +807,8 @@ xfs_da_root_join(xfs_da_state_t *state, xfs_da_state_blk_t *root_blk)
*/
child = be32_to_cpu(oldroot->btree[0].before);
ASSERT(child != 0);
- error = xfs_da_read_buf(args->trans, args->dp, child, -1, &bp,
- args->whichfork, NULL);
+ error = xfs_da_node_read(args->trans, args->dp, child, -1, &bp,
+ args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -835,9 +896,8 @@ xfs_da_node_toosmall(xfs_da_state_t *state, int *action)
blkno = be32_to_cpu(info->back);
if (blkno == 0)
continue;
- error = xfs_da_read_buf(state->args->trans, state->args->dp,
- blkno, -1, &bp, state->args->whichfork,
- NULL);
+ error = xfs_da_node_read(state->args->trans, state->args->dp,
+ blkno, -1, &bp, state->args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1080,8 +1140,8 @@ xfs_da_node_lookup_int(xfs_da_state_t *state, int *result)
* Read the next node down in the tree.
*/
blk->blkno = blkno;
- error = xfs_da_read_buf(args->trans, args->dp, blkno,
- -1, &blk->bp, args->whichfork, NULL);
+ error = xfs_da_node_read(args->trans, args->dp, blkno,
+ -1, &blk->bp, args->whichfork);
if (error) {
blk->blkno = 0;
state->path.active--;
@@ -1242,9 +1302,9 @@ xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
new_info->forw = cpu_to_be32(old_blk->blkno);
new_info->back = old_info->back;
if (old_info->back) {
- error = xfs_da_read_buf(args->trans, args->dp,
+ error = xfs_da_node_read(args->trans, args->dp,
be32_to_cpu(old_info->back),
- -1, &bp, args->whichfork, NULL);
+ -1, &bp, args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1263,9 +1323,9 @@ xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
new_info->forw = old_info->forw;
new_info->back = cpu_to_be32(old_blk->blkno);
if (old_info->forw) {
- error = xfs_da_read_buf(args->trans, args->dp,
+ error = xfs_da_node_read(args->trans, args->dp,
be32_to_cpu(old_info->forw),
- -1, &bp, args->whichfork, NULL);
+ -1, &bp, args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1363,9 +1423,9 @@ xfs_da_blk_unlink(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
trace_xfs_da_unlink_back(args);
save_info->back = drop_info->back;
if (drop_info->back) {
- error = xfs_da_read_buf(args->trans, args->dp,
+ error = xfs_da_node_read(args->trans, args->dp,
be32_to_cpu(drop_info->back),
- -1, &bp, args->whichfork, NULL);
+ -1, &bp, args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1380,9 +1440,9 @@ xfs_da_blk_unlink(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
trace_xfs_da_unlink_forward(args);
save_info->forw = drop_info->forw;
if (drop_info->forw) {
- error = xfs_da_read_buf(args->trans, args->dp,
+ error = xfs_da_node_read(args->trans, args->dp,
be32_to_cpu(drop_info->forw),
- -1, &bp, args->whichfork, NULL);
+ -1, &bp, args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1464,8 +1524,8 @@ xfs_da_path_shift(xfs_da_state_t *state, xfs_da_state_path_t *path,
* Read the next child block.
*/
blk->blkno = blkno;
- error = xfs_da_read_buf(args->trans, args->dp, blkno, -1,
- &blk->bp, args->whichfork, NULL);
+ error = xfs_da_node_read(args->trans, args->dp, blkno, -1,
+ &blk->bp, args->whichfork);
if (error)
return(error);
ASSERT(blk->bp != NULL);
@@ -1728,7 +1788,7 @@ xfs_da_swap_lastblock(
* Read the last block in the btree space.
*/
last_blkno = (xfs_dablk_t)lastoff - mp->m_dirblkfsbs;
- error = xfs_da_read_buf(tp, ip, last_blkno, -1, &last_buf, w, NULL);
+ error = xfs_da_node_read(tp, ip, last_blkno, -1, &last_buf, w);
if (error)
return error;
/*
@@ -1755,8 +1815,7 @@ xfs_da_swap_lastblock(
* If the moved block has a left sibling, fix up the pointers.
*/
if ((sib_blkno = be32_to_cpu(dead_info->back))) {
- error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w,
- NULL);
+ error = xfs_da_node_read(tp, ip, sib_blkno, -1, &sib_buf, w);
if (error)
goto done;
sib_info = sib_buf->b_addr;
@@ -1778,8 +1837,7 @@ xfs_da_swap_lastblock(
* If the moved block has a right sibling, fix up the pointers.
*/
if ((sib_blkno = be32_to_cpu(dead_info->forw))) {
- error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w,
- NULL);
+ error = xfs_da_node_read(tp, ip, sib_blkno, -1, &sib_buf, w);
if (error)
goto done;
sib_info = sib_buf->b_addr;
@@ -1803,8 +1861,7 @@ xfs_da_swap_lastblock(
* Walk down the tree looking for the parent of the moved block.
*/
for (;;) {
- error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w,
- NULL);
+ error = xfs_da_node_read(tp, ip, par_blkno, -1, &par_buf, w);
if (error)
goto done;
par_node = par_buf->b_addr;
@@ -1855,8 +1912,7 @@ xfs_da_swap_lastblock(
error = XFS_ERROR(EFSCORRUPTED);
goto done;
}
- error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w,
- NULL);
+ error = xfs_da_node_read(tp, ip, par_blkno, -1, &par_buf, w);
if (error)
goto done;
par_node = par_buf->b_addr;
diff --git a/fs/xfs/xfs_da_btree.h b/fs/xfs/xfs_da_btree.h
index bf8bfaa..2d1bec4 100644
--- a/fs/xfs/xfs_da_btree.h
+++ b/fs/xfs/xfs_da_btree.h
@@ -213,6 +213,9 @@ int xfs_da_path_shift(xfs_da_state_t *state, xfs_da_state_path_t *path,
*/
int xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
xfs_da_state_blk_t *new_blk);
+int xfs_da_node_read(struct xfs_trans *tp, struct xfs_inode *dp,
+ xfs_dablk_t bno, xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp, int which_fork);
/*
* Utility routines.
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 97408e3..67cc21c 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -74,7 +74,7 @@ xfs_dir2_leaf1_verify(
xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
}
-static void
+void
xfs_dir2_leafn_verify(
struct xfs_buf *bp)
{
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index ecf75d9..1f42e81 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -70,6 +70,7 @@ extern void xfs_dir2_data_use_free(struct xfs_trans *tp, struct xfs_buf *bp,
xfs_dir2_data_aoff_t len, int *needlogp, int *needscanp);
/* xfs_dir2_leaf.c */
+extern void xfs_dir2_leafn_verify(struct xfs_buf *bp);
extern int xfs_dir2_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,
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 20/25] xfs: Add verifiers to dir2 data readahead.
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (18 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 19/25] xfs: add xfs_da_node verification Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 13:31 ` Phil White
2012-10-25 6:34 ` [PATCH 21/25] xfs: add buffer pre-write callback Dave Chinner
` (4 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_da_btree.c | 4 ++--
fs/xfs/xfs_da_btree.h | 4 ++--
fs/xfs/xfs_dir2_data.c | 13 ++++++++++++-
fs/xfs/xfs_dir2_leaf.c | 11 +++++------
fs/xfs/xfs_dir2_priv.h | 2 ++
fs/xfs/xfs_file.c | 4 +++-
6 files changed, 26 insertions(+), 12 deletions(-)
diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
index e950192..7656c14 100644
--- a/fs/xfs/xfs_da_btree.c
+++ b/fs/xfs/xfs_da_btree.c
@@ -2278,10 +2278,10 @@ xfs_da_reada_buf(
struct xfs_trans *trans,
struct xfs_inode *dp,
xfs_dablk_t bno,
+ xfs_daddr_t mappedbno,
int whichfork,
xfs_buf_iodone_t verifier)
{
- xfs_daddr_t mappedbno = -1;
struct xfs_buf_map map;
struct xfs_buf_map *mapp;
int nmap;
@@ -2289,7 +2289,7 @@ xfs_da_reada_buf(
mapp = ↦
nmap = 1;
- error = xfs_dabuf_map(trans, dp, bno, -1, whichfork,
+ error = xfs_dabuf_map(trans, dp, bno, mappedbno, whichfork,
&mapp, &nmap);
if (error) {
/* mapping a hole is not an error, but we don't continue */
diff --git a/fs/xfs/xfs_da_btree.h b/fs/xfs/xfs_da_btree.h
index 2d1bec4..521b008 100644
--- a/fs/xfs/xfs_da_btree.h
+++ b/fs/xfs/xfs_da_btree.h
@@ -231,8 +231,8 @@ int xfs_da_read_buf(struct xfs_trans *trans, struct xfs_inode *dp,
struct xfs_buf **bpp, int whichfork,
xfs_buf_iodone_t verifier);
xfs_daddr_t xfs_da_reada_buf(struct xfs_trans *trans, struct xfs_inode *dp,
- xfs_dablk_t bno, int whichfork,
- xfs_buf_iodone_t verifier);
+ xfs_dablk_t bno, xfs_daddr_t mapped_bno,
+ int whichfork, xfs_buf_iodone_t verifier);
int xfs_da_shrink_inode(xfs_da_args_t *args, xfs_dablk_t dead_blkno,
struct xfs_buf *dead_buf);
diff --git a/fs/xfs/xfs_dir2_data.c b/fs/xfs/xfs_dir2_data.c
index 43c8426..795cfdd 100644
--- a/fs/xfs/xfs_dir2_data.c
+++ b/fs/xfs/xfs_dir2_data.c
@@ -185,7 +185,7 @@ __xfs_dir2_data_check(
return 0;
}
-static void
+void
xfs_dir2_data_verify(
struct xfs_buf *bp)
{
@@ -217,6 +217,17 @@ xfs_dir2_data_read(
XFS_DATA_FORK, xfs_dir2_data_verify);
}
+int
+xfs_dir2_data_readahead(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ xfs_dablk_t bno,
+ xfs_daddr_t mapped_bno)
+{
+ return xfs_da_reada_buf(tp, dp, bno, mapped_bno,
+ XFS_DATA_FORK, xfs_dir2_data_verify);
+}
+
/*
* Given a data block and an unused entry from that block,
* return the bestfree entry if any that corresponds to it.
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 67cc21c..8a95547 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -972,11 +972,11 @@ xfs_dir2_leaf_readbuf(
*/
if (i > mip->ra_current &&
map[mip->ra_index].br_blockcount >= mp->m_dirblkfsbs) {
- xfs_buf_readahead(mp->m_ddev_targp,
+ xfs_dir2_data_readahead(NULL, dp,
+ map[mip->ra_index].br_startoff + mip->ra_offset,
XFS_FSB_TO_DADDR(mp,
map[mip->ra_index].br_startblock +
- mip->ra_offset),
- (int)BTOBB(mp->m_dirblksize), NULL);
+ mip->ra_offset));
mip->ra_current = i;
}
@@ -985,10 +985,9 @@ xfs_dir2_leaf_readbuf(
* use our mapping, but this is a very rare case.
*/
else if (i > mip->ra_current) {
- xfs_da_reada_buf(NULL, dp,
+ xfs_dir2_data_readahead(NULL, dp,
map[mip->ra_index].br_startoff +
- mip->ra_offset,
- XFS_DATA_FORK, NULL);
+ mip->ra_offset, -1);
mip->ra_current = i;
}
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index 1f42e81..aa06174 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -48,6 +48,8 @@ extern int xfs_dir2_leaf_to_block(struct xfs_da_args *args,
extern bool __xfs_dir2_data_check(struct xfs_inode *dp, struct xfs_buf *bp);
extern int xfs_dir2_data_read(struct xfs_trans *tp, struct xfs_inode *dp,
xfs_dablk_t bno, xfs_daddr_t mapped_bno, struct xfs_buf **bpp);
+extern int xfs_dir2_data_readahead(struct xfs_trans *tp, struct xfs_inode *dp,
+ xfs_dablk_t bno, xfs_daddr_t mapped_bno);
extern struct xfs_dir2_data_free *
xfs_dir2_data_freeinsert(struct xfs_dir2_data_hdr *hdr,
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index d949bad..2cc2361 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -31,6 +31,8 @@
#include "xfs_error.h"
#include "xfs_vnodeops.h"
#include "xfs_da_btree.h"
+#include "xfs_dir2_format.h"
+#include "xfs_dir2_priv.h"
#include "xfs_ioctl.h"
#include "xfs_trace.h"
@@ -890,7 +892,7 @@ xfs_dir_open(
*/
mode = xfs_ilock_map_shared(ip);
if (ip->i_d.di_nextents > 0)
- xfs_da_reada_buf(NULL, ip, 0, XFS_DATA_FORK, NULL);
+ xfs_dir2_data_readahead(NULL, ip, 0, -1);
xfs_iunlock(ip, mode);
return 0;
}
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 21/25] xfs: add buffer pre-write callback
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (19 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 20/25] xfs: Add verifiers to dir2 data readahead Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-26 8:50 ` Christoph Hellwig
2012-10-30 13:32 ` Phil White
2012-10-25 6:34 ` [PATCH 22/25] xfs: add pre-write metadata buffer verifier callbacks Dave Chinner
` (3 subsequent siblings)
24 siblings, 2 replies; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Add a callback to the buffer write path to enable verification of
the buffer and CRC calculation prior to issuing the write to the
underlying storage.
If the callback function detects some kind of failure or error
condition, it must mark the buffer with an error so that the caller
can take appropriate action. In the case of xfs_buf_ioapply(), a
corrupt metadta buffer willt rigger a shutdown of the filesystem,
because something is clearly wrong and we can't allow corrupt
metadata to be written to disk.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_buf.c | 16 ++++++++++++++++
fs/xfs/xfs_buf.h | 3 +++
2 files changed, 19 insertions(+)
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index 62b7e89..c073236 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -569,7 +569,9 @@ found:
*/
if (bp->b_flags & XBF_STALE) {
ASSERT((bp->b_flags & _XBF_DELWRI_Q) == 0);
+ ASSERT(bp->b_iodone == NULL);
bp->b_flags &= _XBF_KMEM | _XBF_PAGES;
+ bp->b_pre_io = NULL;
}
trace_xfs_buf_find(bp, flags, _RET_IP_);
@@ -1314,6 +1316,20 @@ _xfs_buf_ioapply(
rw |= REQ_META;
/*
+ * run the pre-io callback function if it exists. If this function
+ * fails it will mark the buffer with an error and the IO should
+ * not be dispatched.
+ */
+ if (bp->b_pre_io) {
+ bp->b_pre_io(bp);
+ if (bp->b_error) {
+ xfs_force_shutdown(bp->b_target->bt_mount,
+ SHUTDOWN_CORRUPT_INCORE);
+ return;
+ }
+ }
+
+ /*
* Walk all the vectors issuing IO on them. Set up the initial offset
* into the buffer and the desired IO size before we start -
* _xfs_buf_ioapply_vec() will modify them appropriately for each
diff --git a/fs/xfs/xfs_buf.h b/fs/xfs/xfs_buf.h
index 677b1dc..51bc16a 100644
--- a/fs/xfs/xfs_buf.h
+++ b/fs/xfs/xfs_buf.h
@@ -155,6 +155,9 @@ typedef struct xfs_buf {
unsigned int b_offset; /* page offset in first page */
unsigned short b_error; /* error code on I/O */
+ void (*b_pre_io)(struct xfs_buf *);
+ /* pre-io callback function */
+
#ifdef XFS_BUF_LOCK_TRACKING
int b_last_holder;
#endif
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 22/25] xfs: add pre-write metadata buffer verifier callbacks
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (20 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 21/25] xfs: add buffer pre-write callback Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 13:34 ` Phil White
2012-10-25 6:34 ` [PATCH 23/25] xfs: connect up write verifiers to new buffers Dave Chinner
` (2 subsequent siblings)
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
These verifiers are essentially the same code as the read verifiers,
but do not require ioend processing. Hence factor the read verifier
functions and add a new write verifier wrapper that is used as the
callback.
This is done as one large patch for all verifiers rather than one
patch per verifier as the change is largely mechanical. This
includes hooking up the write verifier via the read verifier
function.
Hooking up the write verifier for buffers obtained via
xfs_trans_get_buf() will be done in a separate patch as that touches
code in many different places rather than just the verifier
functions.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_alloc.c | 38 +++++++++++++++++++++++++++++++++-----
fs/xfs/xfs_alloc_btree.c | 21 +++++++++++++++++----
fs/xfs/xfs_attr_leaf.c | 19 +++++++++++++++++--
fs/xfs/xfs_attr_leaf.h | 2 +-
fs/xfs/xfs_bmap_btree.c | 21 +++++++++++++++++----
fs/xfs/xfs_da_btree.c | 30 ++++++++++++++++++------------
fs/xfs/xfs_dir2_block.c | 16 +++++++++++++++-
fs/xfs/xfs_dir2_data.c | 19 +++++++++++++++++--
fs/xfs/xfs_dir2_leaf.c | 31 ++++++++++++++++++++++++-------
fs/xfs/xfs_dir2_node.c | 17 ++++++++++++++++-
fs/xfs/xfs_dir2_priv.h | 2 +-
fs/xfs/xfs_dquot.c | 22 ++++++++++++++++++----
fs/xfs/xfs_ialloc.c | 17 ++++++++++++++++-
fs/xfs/xfs_ialloc_btree.c | 19 ++++++++++++++++---
fs/xfs/xfs_inode.c | 19 +++++++++++++++++--
fs/xfs/xfs_inode.h | 2 +-
fs/xfs/xfs_itable.c | 2 +-
fs/xfs/xfs_mount.c | 19 +++++++++++++++++--
18 files changed, 262 insertions(+), 54 deletions(-)
diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
index 0fa37a7..5f42e53 100644
--- a/fs/xfs/xfs_alloc.c
+++ b/fs/xfs/xfs_alloc.c
@@ -430,8 +430,8 @@ xfs_alloc_fixup_trees(
return 0;
}
-void
-xfs_agfl_read_verify(
+static void
+xfs_agfl_verify(
struct xfs_buf *bp)
{
#ifdef WHEN_CRCS_COME_ALONG
@@ -459,11 +459,25 @@ xfs_agfl_read_verify(
}
if (!agfl_ok) {
- XFS_CORRUPTION_ERROR("xfs_agfl_read_verify",
- XFS_ERRLEVEL_LOW, mp, agfl);
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, agfl);
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
#endif
+}
+
+static void
+xfs_agfl_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_agfl_verify(bp);
+}
+
+void
+xfs_agfl_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_agfl_verify(bp);
+ bp->b_pre_io = xfs_agfl_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
@@ -2130,7 +2144,7 @@ xfs_alloc_put_freelist(
}
static void
-xfs_agf_read_verify(
+xfs_agf_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
@@ -2157,7 +2171,21 @@ xfs_agf_read_verify(
XFS_ERRLEVEL_LOW, mp, agf);
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
+}
+
+static void
+xfs_agf_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_agf_verify(bp);
+}
+void
+xfs_agf_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_agf_verify(bp);
+ bp->b_pre_io = xfs_agf_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
diff --git a/fs/xfs/xfs_alloc_btree.c b/fs/xfs/xfs_alloc_btree.c
index 4167e72..6fc432b 100644
--- a/fs/xfs/xfs_alloc_btree.c
+++ b/fs/xfs/xfs_alloc_btree.c
@@ -272,8 +272,8 @@ xfs_allocbt_key_diff(
return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
}
-void
-xfs_allocbt_read_verify(
+static void
+xfs_allocbt_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
@@ -311,11 +311,24 @@ xfs_allocbt_read_verify(
if (!sblock_ok) {
trace_xfs_btree_corrupt(bp, _RET_IP_);
- XFS_CORRUPTION_ERROR("xfs_allocbt_read_verify",
- XFS_ERRLEVEL_LOW, mp, block);
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, block);
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
+}
+static void
+xfs_allocbt_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_allocbt_verify(bp);
+}
+
+void
+xfs_allocbt_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_allocbt_verify(bp);
+ bp->b_pre_io = xfs_allocbt_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c
index 5ba92eb..bb96c55 100644
--- a/fs/xfs/xfs_attr_leaf.c
+++ b/fs/xfs/xfs_attr_leaf.c
@@ -87,7 +87,7 @@ STATIC void xfs_attr_leaf_moveents(xfs_attr_leafblock_t *src_leaf,
xfs_mount_t *mp);
STATIC int xfs_attr_leaf_entsize(xfs_attr_leafblock_t *leaf, int index);
-void
+static void
xfs_attr_leaf_verify(
struct xfs_buf *bp)
{
@@ -100,11 +100,26 @@ xfs_attr_leaf_verify(
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
+}
+
+static void
+xfs_attr_leaf_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_attr_leaf_verify(bp);
+}
+void
+xfs_attr_leaf_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_attr_leaf_verify(bp);
+ bp->b_pre_io = xfs_attr_leaf_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
+
int
xfs_attr_leaf_read(
struct xfs_trans *tp,
@@ -114,7 +129,7 @@ xfs_attr_leaf_read(
struct xfs_buf **bpp)
{
return xfs_da_read_buf(tp, dp, bno, mappedbno, bpp,
- XFS_ATTR_FORK, xfs_attr_leaf_verify);
+ XFS_ATTR_FORK, xfs_attr_leaf_read_verify);
}
/*========================================================================
diff --git a/fs/xfs/xfs_attr_leaf.h b/fs/xfs/xfs_attr_leaf.h
index 098e9a5..3bbf627 100644
--- a/fs/xfs/xfs_attr_leaf.h
+++ b/fs/xfs/xfs_attr_leaf.h
@@ -264,6 +264,6 @@ int xfs_attr_leaf_newentsize(int namelen, int valuelen, int blocksize,
int xfs_attr_leaf_read(struct xfs_trans *tp, struct xfs_inode *dp,
xfs_dablk_t bno, xfs_daddr_t mappedbno,
struct xfs_buf **bpp);
-void xfs_attr_leaf_verify(struct xfs_buf *bp);
+void xfs_attr_leaf_read_verify(struct xfs_buf *bp);
#endif /* __XFS_ATTR_LEAF_H__ */
diff --git a/fs/xfs/xfs_bmap_btree.c b/fs/xfs/xfs_bmap_btree.c
index bddca9b..17d7423 100644
--- a/fs/xfs/xfs_bmap_btree.c
+++ b/fs/xfs/xfs_bmap_btree.c
@@ -708,8 +708,8 @@ xfs_bmbt_key_diff(
cur->bc_rec.b.br_startoff;
}
-void
-xfs_bmbt_read_verify(
+static void
+xfs_bmbt_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
@@ -744,11 +744,24 @@ xfs_bmbt_read_verify(
if (!lblock_ok) {
trace_xfs_btree_corrupt(bp, _RET_IP_);
- XFS_CORRUPTION_ERROR("xfs_bmbt_read_verify",
- XFS_ERRLEVEL_LOW, mp, block);
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, block);
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
+}
+static void
+xfs_bmbt_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_bmbt_verify(bp);
+}
+
+void
+xfs_bmbt_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_bmbt_verify(bp);
+ bp->b_pre_io = xfs_bmbt_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
index 7656c14..179173e 100644
--- a/fs/xfs/xfs_da_btree.c
+++ b/fs/xfs/xfs_da_btree.c
@@ -92,7 +92,7 @@ STATIC int xfs_da_blk_unlink(xfs_da_state_t *state,
STATIC void xfs_da_state_kill_altpath(xfs_da_state_t *state);
static void
-__xfs_da_node_verify(
+xfs_da_node_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
@@ -107,12 +107,17 @@ __xfs_da_node_verify(
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
static void
-xfs_da_node_verify(
+xfs_da_node_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_da_node_verify(bp);
+}
+
+static void
+xfs_da_node_read_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
@@ -120,21 +125,22 @@ xfs_da_node_verify(
switch (be16_to_cpu(info->magic)) {
case XFS_DA_NODE_MAGIC:
- __xfs_da_node_verify(bp);
- return;
+ xfs_da_node_verify(bp);
+ break;
case XFS_ATTR_LEAF_MAGIC:
- xfs_attr_leaf_verify(bp);
+ xfs_attr_leaf_read_verify(bp);
return;
case XFS_DIR2_LEAFN_MAGIC:
- xfs_dir2_leafn_verify(bp);
+ xfs_dir2_leafn_read_verify(bp);
return;
default:
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW,
+ mp, info);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
break;
}
- XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, info);
- xfs_buf_ioerror(bp, EFSCORRUPTED);
-
+ bp->b_pre_io = xfs_da_node_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
@@ -149,7 +155,7 @@ xfs_da_node_read(
int which_fork)
{
return xfs_da_read_buf(tp, dp, bno, mappedbno, bpp,
- which_fork, xfs_da_node_verify);
+ which_fork, xfs_da_node_read_verify);
}
/*========================================================================
diff --git a/fs/xfs/xfs_dir2_block.c b/fs/xfs/xfs_dir2_block.c
index ca03b10..0f8793c 100644
--- a/fs/xfs/xfs_dir2_block.c
+++ b/fs/xfs/xfs_dir2_block.c
@@ -71,7 +71,21 @@ xfs_dir2_block_verify(
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
+}
+static void
+xfs_dir2_block_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dir2_block_verify(bp);
+}
+
+void
+xfs_dir2_block_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dir2_block_verify(bp);
+ bp->b_pre_io = xfs_dir2_block_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
@@ -85,7 +99,7 @@ xfs_dir2_block_read(
struct xfs_mount *mp = dp->i_mount;
return xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, bpp,
- XFS_DATA_FORK, xfs_dir2_block_verify);
+ XFS_DATA_FORK, xfs_dir2_block_read_verify);
}
static void
diff --git a/fs/xfs/xfs_dir2_data.c b/fs/xfs/xfs_dir2_data.c
index 795cfdd..0af533b 100644
--- a/fs/xfs/xfs_dir2_data.c
+++ b/fs/xfs/xfs_dir2_data.c
@@ -200,11 +200,26 @@ xfs_dir2_data_verify(
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
+}
+
+static void
+xfs_dir2_data_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dir2_data_verify(bp);
+}
+void
+xfs_dir2_data_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dir2_data_verify(bp);
+ bp->b_pre_io = xfs_dir2_data_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
+
int
xfs_dir2_data_read(
struct xfs_trans *tp,
@@ -214,7 +229,7 @@ xfs_dir2_data_read(
struct xfs_buf **bpp)
{
return xfs_da_read_buf(tp, dp, bno, mapped_bno, bpp,
- XFS_DATA_FORK, xfs_dir2_data_verify);
+ XFS_DATA_FORK, xfs_dir2_data_read_verify);
}
int
@@ -225,7 +240,7 @@ xfs_dir2_data_readahead(
xfs_daddr_t mapped_bno)
{
return xfs_da_reada_buf(tp, dp, bno, mapped_bno,
- XFS_DATA_FORK, xfs_dir2_data_verify);
+ XFS_DATA_FORK, xfs_dir2_data_read_verify);
}
/*
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 8a95547..5b3bcab 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -62,23 +62,40 @@ xfs_dir2_leaf_verify(
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
+}
+
+static void
+xfs_dir2_leaf1_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
+}
+static void
+xfs_dir2_leaf1_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
+ bp->b_pre_io = xfs_dir2_leaf1_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
static void
-xfs_dir2_leaf1_verify(
- struct xfs_buf *bp)
+xfs_dir2_leafn_write_verify(
+ struct xfs_buf *bp)
{
- xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
+ xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
}
void
-xfs_dir2_leafn_verify(
- struct xfs_buf *bp)
+xfs_dir2_leafn_read_verify(
+ struct xfs_buf *bp)
{
xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
+ bp->b_pre_io = xfs_dir2_leafn_write_verify;
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
}
static int
@@ -90,7 +107,7 @@ xfs_dir2_leaf_read(
struct xfs_buf **bpp)
{
return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
- XFS_DATA_FORK, xfs_dir2_leaf1_verify);
+ XFS_DATA_FORK, xfs_dir2_leaf1_read_verify);
}
int
@@ -102,7 +119,7 @@ xfs_dir2_leafn_read(
struct xfs_buf **bpp)
{
return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
- XFS_DATA_FORK, xfs_dir2_leafn_verify);
+ XFS_DATA_FORK, xfs_dir2_leafn_read_verify);
}
/*
diff --git a/fs/xfs/xfs_dir2_node.c b/fs/xfs/xfs_dir2_node.c
index 7c6f956..a58abe1 100644
--- a/fs/xfs/xfs_dir2_node.c
+++ b/fs/xfs/xfs_dir2_node.c
@@ -69,11 +69,26 @@ xfs_dir2_free_verify(
XFS_ERRLEVEL_LOW, mp, hdr);
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
+}
+
+static void
+xfs_dir2_free_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dir2_free_verify(bp);
+}
+void
+xfs_dir2_free_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dir2_free_verify(bp);
+ bp->b_pre_io = xfs_dir2_free_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
+
static int
__xfs_dir2_free_read(
struct xfs_trans *tp,
@@ -83,7 +98,7 @@ __xfs_dir2_free_read(
struct xfs_buf **bpp)
{
return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
- XFS_DATA_FORK, xfs_dir2_free_verify);
+ XFS_DATA_FORK, xfs_dir2_free_read_verify);
}
int
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index aa06174..4aeef62 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -72,7 +72,7 @@ extern void xfs_dir2_data_use_free(struct xfs_trans *tp, struct xfs_buf *bp,
xfs_dir2_data_aoff_t len, int *needlogp, int *needscanp);
/* xfs_dir2_leaf.c */
-extern void xfs_dir2_leafn_verify(struct xfs_buf *bp);
+extern void xfs_dir2_leafn_read_verify(struct xfs_buf *bp);
extern int xfs_dir2_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/xfs_dquot.c b/fs/xfs/xfs_dquot.c
index 2e18382..eff7586 100644
--- a/fs/xfs/xfs_dquot.c
+++ b/fs/xfs/xfs_dquot.c
@@ -361,7 +361,7 @@ xfs_qm_dqalloc(
}
STATIC void
-xfs_dquot_read_verify(
+xfs_dquot_buf_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
@@ -388,12 +388,26 @@ xfs_dquot_read_verify(
error = xfs_qm_dqcheck(mp, ddq, id + i, 0, XFS_QMOPT_DOWARN,
"xfs_dquot_read_verify");
if (error) {
- XFS_CORRUPTION_ERROR("xfs_dquot_read_verify",
- XFS_ERRLEVEL_LOW, mp, d);
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, d);
xfs_buf_ioerror(bp, EFSCORRUPTED);
break;
}
}
+}
+
+static void
+xfs_dquot_buf_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dquot_buf_verify(bp);
+}
+
+static void
+xfs_dquot_buf_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dquot_buf_verify(bp);
+ bp->b_pre_io = xfs_dquot_buf_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
@@ -521,7 +535,7 @@ xfs_qm_dqtobp(
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
dqp->q_blkno,
mp->m_quotainfo->qi_dqchunklen,
- 0, &bp, xfs_dquot_read_verify);
+ 0, &bp, xfs_dquot_buf_read_verify);
if (error == EFSCORRUPTED && (flags & XFS_QMOPT_DQREPAIR)) {
xfs_dqid_t firstid = (xfs_dqid_t)map.br_startoff *
diff --git a/fs/xfs/xfs_ialloc.c b/fs/xfs/xfs_ialloc.c
index 9311ae5..f260a36 100644
--- a/fs/xfs/xfs_ialloc.c
+++ b/fs/xfs/xfs_ialloc.c
@@ -1473,7 +1473,7 @@ xfs_check_agi_unlinked(
#endif
static void
-xfs_agi_read_verify(
+xfs_agi_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
@@ -1493,6 +1493,21 @@ xfs_agi_read_verify(
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
xfs_check_agi_unlinked(agi);
+}
+
+static void
+xfs_agi_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_agi_verify(bp);
+}
+
+void
+xfs_agi_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_agi_verify(bp);
+ bp->b_pre_io = xfs_agi_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
diff --git a/fs/xfs/xfs_ialloc_btree.c b/fs/xfs/xfs_ialloc_btree.c
index 11306c6..15a79f8 100644
--- a/fs/xfs/xfs_ialloc_btree.c
+++ b/fs/xfs/xfs_ialloc_btree.c
@@ -183,7 +183,7 @@ xfs_inobt_key_diff(
}
void
-xfs_inobt_read_verify(
+xfs_inobt_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
@@ -211,11 +211,24 @@ xfs_inobt_read_verify(
if (!sblock_ok) {
trace_xfs_btree_corrupt(bp, _RET_IP_);
- XFS_CORRUPTION_ERROR("xfs_inobt_read_verify",
- XFS_ERRLEVEL_LOW, mp, block);
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, block);
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
+}
+
+static void
+xfs_inobt_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_inobt_verify(bp);
+}
+void
+xfs_inobt_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_inobt_verify(bp);
+ bp->b_pre_io = xfs_inobt_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 0905e72..875ceb2 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -382,7 +382,7 @@ xfs_inobp_check(
}
#endif
-void
+static void
xfs_inode_buf_verify(
struct xfs_buf *bp)
{
@@ -418,6 +418,21 @@ xfs_inode_buf_verify(
}
}
xfs_inobp_check(mp, bp);
+}
+
+static void
+xfs_inode_buf_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_inode_buf_verify(bp);
+}
+
+void
+xfs_inode_buf_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_inode_buf_verify(bp);
+ bp->b_pre_io = xfs_inode_buf_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
@@ -447,7 +462,7 @@ xfs_imap_to_bp(
buf_flags |= XBF_UNMAPPED;
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
(int)imap->im_len, buf_flags, &bp,
- xfs_inode_buf_verify);
+ xfs_inode_buf_read_verify);
if (error) {
if (error == EAGAIN) {
ASSERT(buf_flags & XBF_TRYLOCK);
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index 3c1d831..32b6d70 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -554,7 +554,7 @@ int xfs_imap_to_bp(struct xfs_mount *, struct xfs_trans *,
struct xfs_buf **, uint, uint);
int xfs_iread(struct xfs_mount *, struct xfs_trans *,
struct xfs_inode *, uint);
-void xfs_inode_buf_verify(struct xfs_buf *);
+void xfs_inode_buf_read_verify(struct xfs_buf *);
void xfs_dinode_to_disk(struct xfs_dinode *,
struct xfs_icdinode *);
void xfs_idestroy_fork(struct xfs_inode *, int);
diff --git a/fs/xfs/xfs_itable.c b/fs/xfs/xfs_itable.c
index 0f18d41..7f86fda 100644
--- a/fs/xfs/xfs_itable.c
+++ b/fs/xfs/xfs_itable.c
@@ -397,7 +397,7 @@ xfs_bulkstat(
& ~r.ir_free)
xfs_btree_reada_bufs(mp, agno,
agbno, nbcluster,
- xfs_inode_buf_verify);
+ xfs_inode_buf_read_verify);
}
irbp->ir_startino = r.ir_startino;
irbp->ir_freecount = r.ir_freecount;
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index 8699e5e..a622f6f 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -612,8 +612,8 @@ xfs_sb_to_disk(
}
}
-void
-xfs_sb_read_verify(
+static void
+xfs_sb_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
@@ -629,6 +629,21 @@ xfs_sb_read_verify(
error = xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR);
if (error)
xfs_buf_ioerror(bp, error);
+}
+
+static void
+xfs_sb_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_sb_verify(bp);
+}
+
+void
+xfs_sb_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_sb_verify(bp);
+ bp->b_pre_io = xfs_sb_write_verify;
bp->b_iodone = NULL;
xfs_buf_ioend(bp, 0);
}
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 23/25] xfs: connect up write verifiers to new buffers
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (21 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 22/25] xfs: add pre-write metadata buffer verifier callbacks Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 13:39 ` Phil White
2012-10-25 6:34 ` [PATCH 24/25] xfs: convert buffer verifiers to an ops structure Dave Chinner
2012-10-25 6:34 ` [PATCH 25/25] xfs: add write verifiers to log recovery Dave Chinner
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Metadata buffers that are read from disk have write verifiers
already attached to them, but newly allocated buffers do not. Add
appropriate write verifiers to all new metadata buffers.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_alloc.c | 6 +--
fs/xfs/xfs_alloc.h | 2 +
fs/xfs/xfs_alloc_btree.c | 1 +
fs/xfs/xfs_attr_leaf.c | 4 +-
fs/xfs/xfs_bmap.c | 2 +
fs/xfs/xfs_bmap_btree.c | 3 +-
fs/xfs/xfs_bmap_btree.h | 1 +
fs/xfs/xfs_btree.c | 1 +
fs/xfs/xfs_btree.h | 2 +
fs/xfs/xfs_da_btree.c | 3 ++
fs/xfs/xfs_dir2_block.c | 2 +
fs/xfs/xfs_dir2_data.c | 11 +++--
fs/xfs/xfs_dir2_leaf.c | 19 ++++++---
fs/xfs/xfs_dir2_node.c | 24 +++++++----
fs/xfs/xfs_dir2_priv.h | 2 +
fs/xfs/xfs_dquot.c | 104 ++++++++++++++++++++++-----------------------
fs/xfs/xfs_fsops.c | 7 ++-
fs/xfs/xfs_ialloc.c | 5 ++-
fs/xfs/xfs_ialloc.h | 4 +-
fs/xfs/xfs_ialloc_btree.c | 1 +
fs/xfs/xfs_inode.c | 14 +++++-
fs/xfs/xfs_inode.h | 1 +
fs/xfs/xfs_mount.c | 2 +-
fs/xfs/xfs_mount.h | 1 +
24 files changed, 135 insertions(+), 87 deletions(-)
diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
index 5f42e53..578d2e8 100644
--- a/fs/xfs/xfs_alloc.c
+++ b/fs/xfs/xfs_alloc.c
@@ -472,7 +472,7 @@ xfs_agfl_write_verify(
xfs_agfl_verify(bp);
}
-void
+static void
xfs_agfl_read_verify(
struct xfs_buf *bp)
{
@@ -2173,14 +2173,14 @@ xfs_agf_verify(
}
}
-static void
+void
xfs_agf_write_verify(
struct xfs_buf *bp)
{
xfs_agf_verify(bp);
}
-void
+static void
xfs_agf_read_verify(
struct xfs_buf *bp)
{
diff --git a/fs/xfs/xfs_alloc.h b/fs/xfs/xfs_alloc.h
index feacb06..4b6c2c4 100644
--- a/fs/xfs/xfs_alloc.h
+++ b/fs/xfs/xfs_alloc.h
@@ -231,4 +231,6 @@ xfs_alloc_get_rec(
xfs_extlen_t *len, /* output: length of extent */
int *stat); /* output: success/failure */
+void xfs_agf_write_verify(struct xfs_buf *bp);
+
#endif /* __XFS_ALLOC_H__ */
diff --git a/fs/xfs/xfs_alloc_btree.c b/fs/xfs/xfs_alloc_btree.c
index 6fc432b..7f8e704 100644
--- a/fs/xfs/xfs_alloc_btree.c
+++ b/fs/xfs/xfs_alloc_btree.c
@@ -389,6 +389,7 @@ static const struct xfs_btree_ops xfs_allocbt_ops = {
.init_ptr_from_cur = xfs_allocbt_init_ptr_from_cur,
.key_diff = xfs_allocbt_key_diff,
.read_verify = xfs_allocbt_read_verify,
+ .write_verify = xfs_allocbt_write_verify,
#ifdef DEBUG
.keys_inorder = xfs_allocbt_keys_inorder,
.recs_inorder = xfs_allocbt_recs_inorder,
diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c
index bb96c55..5d56886 100644
--- a/fs/xfs/xfs_attr_leaf.c
+++ b/fs/xfs/xfs_attr_leaf.c
@@ -923,7 +923,7 @@ xfs_attr_leaf_to_node(xfs_da_args_t *args)
XFS_ATTR_FORK);
if (error)
goto out;
- ASSERT(bp2 != NULL);
+ bp2->b_pre_io = bp1->b_pre_io;
memcpy(bp2->b_addr, bp1->b_addr, XFS_LBSIZE(dp->i_mount));
bp1 = NULL;
xfs_trans_log_buf(args->trans, bp2, 0, XFS_LBSIZE(dp->i_mount) - 1);
@@ -977,7 +977,7 @@ xfs_attr_leaf_create(
XFS_ATTR_FORK);
if (error)
return(error);
- ASSERT(bp != NULL);
+ bp->b_pre_io = xfs_attr_leaf_write_verify;
leaf = bp->b_addr;
memset((char *)leaf, 0, XFS_LBSIZE(dp->i_mount));
hdr = &leaf->hdr;
diff --git a/fs/xfs/xfs_bmap.c b/fs/xfs/xfs_bmap.c
index 8e944bb..fe5438b 100644
--- a/fs/xfs/xfs_bmap.c
+++ b/fs/xfs/xfs_bmap.c
@@ -3124,6 +3124,7 @@ xfs_bmap_extents_to_btree(
/*
* Fill in the child block.
*/
+ abp->b_pre_io = xfs_bmbt_write_verify;
ablock = XFS_BUF_TO_BLOCK(abp);
ablock->bb_magic = cpu_to_be32(XFS_BMAP_MAGIC);
ablock->bb_level = 0;
@@ -3270,6 +3271,7 @@ xfs_bmap_local_to_extents(
ASSERT(args.len == 1);
*firstblock = args.fsbno;
bp = xfs_btree_get_bufl(args.mp, tp, args.fsbno, 0);
+ bp->b_pre_io = xfs_bmbt_write_verify;
memcpy(bp->b_addr, ifp->if_u1.if_data, ifp->if_bytes);
xfs_trans_log_buf(tp, bp, 0, ifp->if_bytes - 1);
xfs_bmap_forkoff_reset(args.mp, ip, whichfork);
diff --git a/fs/xfs/xfs_bmap_btree.c b/fs/xfs/xfs_bmap_btree.c
index 17d7423..79758e1 100644
--- a/fs/xfs/xfs_bmap_btree.c
+++ b/fs/xfs/xfs_bmap_btree.c
@@ -749,7 +749,7 @@ xfs_bmbt_verify(
}
}
-static void
+void
xfs_bmbt_write_verify(
struct xfs_buf *bp)
{
@@ -806,6 +806,7 @@ static const struct xfs_btree_ops xfs_bmbt_ops = {
.init_ptr_from_cur = xfs_bmbt_init_ptr_from_cur,
.key_diff = xfs_bmbt_key_diff,
.read_verify = xfs_bmbt_read_verify,
+ .write_verify = xfs_bmbt_write_verify,
#ifdef DEBUG
.keys_inorder = xfs_bmbt_keys_inorder,
.recs_inorder = xfs_bmbt_recs_inorder,
diff --git a/fs/xfs/xfs_bmap_btree.h b/fs/xfs/xfs_bmap_btree.h
index 1d00fbe..938c859 100644
--- a/fs/xfs/xfs_bmap_btree.h
+++ b/fs/xfs/xfs_bmap_btree.h
@@ -233,6 +233,7 @@ extern int xfs_bmbt_get_maxrecs(struct xfs_btree_cur *, int level);
extern int xfs_bmdr_maxrecs(struct xfs_mount *, int blocklen, int leaf);
extern int xfs_bmbt_maxrecs(struct xfs_mount *, int blocklen, int leaf);
extern void xfs_bmbt_read_verify(struct xfs_buf *bp);
+extern void xfs_bmbt_write_verify(struct xfs_buf *bp);
extern struct xfs_btree_cur *xfs_bmbt_init_cursor(struct xfs_mount *,
struct xfs_trans *, struct xfs_inode *, int);
diff --git a/fs/xfs/xfs_btree.c b/fs/xfs/xfs_btree.c
index b680949..1ccd42b 100644
--- a/fs/xfs/xfs_btree.c
+++ b/fs/xfs/xfs_btree.c
@@ -981,6 +981,7 @@ xfs_btree_get_buf_block(
if (!*bpp)
return ENOMEM;
+ (*bpp)->b_pre_io = cur->bc_ops->write_verify;
*block = XFS_BUF_TO_BLOCK(*bpp);
return 0;
}
diff --git a/fs/xfs/xfs_btree.h b/fs/xfs/xfs_btree.h
index 7c3cb8d..2ac403a 100644
--- a/fs/xfs/xfs_btree.h
+++ b/fs/xfs/xfs_btree.h
@@ -189,6 +189,8 @@ struct xfs_btree_ops {
union xfs_btree_key *key);
void (*read_verify)(struct xfs_buf *bp);
+ void (*write_verify)(struct xfs_buf *bp);
+
#ifdef DEBUG
/* check that k1 is lower than k2 */
int (*keys_inorder)(struct xfs_btree_cur *cur,
diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
index 179173e..eff604f 100644
--- a/fs/xfs/xfs_da_btree.c
+++ b/fs/xfs/xfs_da_btree.c
@@ -192,6 +192,7 @@ xfs_da_node_create(xfs_da_args_t *args, xfs_dablk_t blkno, int level,
xfs_trans_log_buf(tp, bp,
XFS_DA_LOGRANGE(node, &node->hdr, sizeof(node->hdr)));
+ bp->b_pre_io = xfs_da_node_write_verify;
*bpp = bp;
return(0);
}
@@ -391,6 +392,8 @@ xfs_da_root_split(xfs_da_state_t *state, xfs_da_state_blk_t *blk1,
}
memcpy(node, oldroot, size);
xfs_trans_log_buf(tp, bp, 0, size - 1);
+
+ bp->b_pre_io = blk1->bp->b_pre_io;
blk1->bp = bp;
blk1->blkno = blkno;
diff --git a/fs/xfs/xfs_dir2_block.c b/fs/xfs/xfs_dir2_block.c
index 0f8793c..e2fdc6f 100644
--- a/fs/xfs/xfs_dir2_block.c
+++ b/fs/xfs/xfs_dir2_block.c
@@ -1010,6 +1010,7 @@ xfs_dir2_leaf_to_block(
/*
* Start converting it to block form.
*/
+ dbp->b_pre_io = xfs_dir2_block_write_verify;
hdr->magic = cpu_to_be32(XFS_DIR2_BLOCK_MAGIC);
needlog = 1;
needscan = 0;
@@ -1139,6 +1140,7 @@ xfs_dir2_sf_to_block(
kmem_free(sfp);
return error;
}
+ bp->b_pre_io = xfs_dir2_block_write_verify;
hdr = bp->b_addr;
hdr->magic = cpu_to_be32(XFS_DIR2_BLOCK_MAGIC);
/*
diff --git a/fs/xfs/xfs_dir2_data.c b/fs/xfs/xfs_dir2_data.c
index 0af533b..c759c7b 100644
--- a/fs/xfs/xfs_dir2_data.c
+++ b/fs/xfs/xfs_dir2_data.c
@@ -185,7 +185,7 @@ __xfs_dir2_data_check(
return 0;
}
-void
+static void
xfs_dir2_data_verify(
struct xfs_buf *bp)
{
@@ -202,14 +202,14 @@ xfs_dir2_data_verify(
}
}
-static void
+void
xfs_dir2_data_write_verify(
struct xfs_buf *bp)
{
xfs_dir2_data_verify(bp);
}
-void
+static void
xfs_dir2_data_read_verify(
struct xfs_buf *bp)
{
@@ -482,10 +482,9 @@ xfs_dir2_data_init(
*/
error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(mp, blkno), -1, &bp,
XFS_DATA_FORK);
- if (error) {
+ if (error)
return error;
- }
- ASSERT(bp != NULL);
+ bp->b_pre_io = xfs_dir2_data_write_verify;
/*
* Initialize the header.
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 5b3bcab..3002ab7 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -81,7 +81,7 @@ xfs_dir2_leaf1_read_verify(
xfs_buf_ioend(bp, 0);
}
-static void
+void
xfs_dir2_leafn_write_verify(
struct xfs_buf *bp)
{
@@ -198,6 +198,7 @@ xfs_dir2_block_to_leaf(
/*
* Fix up the block header, make it a data block.
*/
+ dbp->b_pre_io = xfs_dir2_data_write_verify;
hdr->magic = cpu_to_be32(XFS_DIR2_DATA_MAGIC);
if (needscan)
xfs_dir2_data_freescan(mp, hdr, &needlog);
@@ -1243,15 +1244,14 @@ xfs_dir2_leaf_init(
* Get the buffer for the block.
*/
error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(mp, bno), -1, &bp,
- XFS_DATA_FORK);
- if (error) {
+ XFS_DATA_FORK);
+ if (error)
return error;
- }
- ASSERT(bp != NULL);
- leaf = bp->b_addr;
+
/*
* Initialize the header.
*/
+ leaf = bp->b_addr;
leaf->hdr.info.magic = cpu_to_be16(magic);
leaf->hdr.info.forw = 0;
leaf->hdr.info.back = 0;
@@ -1264,10 +1264,12 @@ xfs_dir2_leaf_init(
* the block.
*/
if (magic == XFS_DIR2_LEAF1_MAGIC) {
+ bp->b_pre_io = xfs_dir2_leaf1_write_verify;
ltp = xfs_dir2_leaf_tail_p(mp, leaf);
ltp->bestcount = 0;
xfs_dir2_leaf_log_tail(tp, bp);
- }
+ } else
+ bp->b_pre_io = xfs_dir2_leafn_write_verify;
*bpp = bp;
return 0;
}
@@ -1951,7 +1953,10 @@ xfs_dir2_node_to_leaf(
xfs_dir2_leaf_compact(args, lbp);
else
xfs_dir2_leaf_log_header(tp, lbp);
+
+ lbp->b_pre_io = xfs_dir2_leaf1_write_verify;
leaf->hdr.info.magic = cpu_to_be16(XFS_DIR2_LEAF1_MAGIC);
+
/*
* Set up the leaf tail from the freespace block.
*/
diff --git a/fs/xfs/xfs_dir2_node.c b/fs/xfs/xfs_dir2_node.c
index a58abe1..da90a91 100644
--- a/fs/xfs/xfs_dir2_node.c
+++ b/fs/xfs/xfs_dir2_node.c
@@ -197,11 +197,12 @@ xfs_dir2_leaf_to_node(
/*
* Get the buffer for the new freespace block.
*/
- if ((error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(mp, fdb), -1, &fbp,
- XFS_DATA_FORK))) {
+ error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(mp, fdb), -1, &fbp,
+ XFS_DATA_FORK);
+ if (error)
return error;
- }
- ASSERT(fbp != NULL);
+ fbp->b_pre_io = xfs_dir2_free_write_verify;
+
free = fbp->b_addr;
leaf = lbp->b_addr;
ltp = xfs_dir2_leaf_tail_p(mp, leaf);
@@ -223,7 +224,10 @@ xfs_dir2_leaf_to_node(
*to = cpu_to_be16(off);
}
free->hdr.nused = cpu_to_be32(n);
+
+ lbp->b_pre_io = xfs_dir2_leafn_write_verify;
leaf->hdr.info.magic = cpu_to_be16(XFS_DIR2_LEAFN_MAGIC);
+
/*
* Log everything.
*/
@@ -632,6 +636,7 @@ xfs_dir2_leafn_lookup_for_entry(
state->extrablk.index = (int)((char *)dep -
(char *)curbp->b_addr);
state->extrablk.magic = XFS_DIR2_DATA_MAGIC;
+ curbp->b_pre_io = xfs_dir2_data_write_verify;
if (cmp == XFS_CMP_EXACT)
return XFS_ERROR(EEXIST);
}
@@ -646,6 +651,7 @@ xfs_dir2_leafn_lookup_for_entry(
state->extrablk.index = -1;
state->extrablk.blkno = curdb;
state->extrablk.magic = XFS_DIR2_DATA_MAGIC;
+ curbp->b_pre_io = xfs_dir2_data_write_verify;
} else {
/* If the curbp is not the CI match block, drop it */
if (state->extrablk.bp != curbp)
@@ -1638,12 +1644,12 @@ xfs_dir2_node_addname_int(
/*
* Get a buffer for the new block.
*/
- if ((error = xfs_da_get_buf(tp, dp,
- xfs_dir2_db_to_da(mp, fbno),
- -1, &fbp, XFS_DATA_FORK))) {
+ error = xfs_da_get_buf(tp, dp,
+ xfs_dir2_db_to_da(mp, fbno),
+ -1, &fbp, XFS_DATA_FORK);
+ if (error)
return error;
- }
- ASSERT(fbp != NULL);
+ fbp->b_pre_io = xfs_dir2_free_write_verify;
/*
* Initialize the new block to be empty, and remember
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index 4aeef62..ac49e12 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -45,6 +45,7 @@ extern int xfs_dir2_leaf_to_block(struct xfs_da_args *args,
#else
#define xfs_dir2_data_check(dp,bp)
#endif
+extern void xfs_dir2_data_write_verify(struct xfs_buf *bp);
extern bool __xfs_dir2_data_check(struct xfs_inode *dp, struct xfs_buf *bp);
extern int xfs_dir2_data_read(struct xfs_trans *tp, struct xfs_inode *dp,
xfs_dablk_t bno, xfs_daddr_t mapped_bno, struct xfs_buf **bpp);
@@ -73,6 +74,7 @@ extern void xfs_dir2_data_use_free(struct xfs_trans *tp, struct xfs_buf *bp,
/* xfs_dir2_leaf.c */
extern void xfs_dir2_leafn_read_verify(struct xfs_buf *bp);
+extern void xfs_dir2_leafn_write_verify(struct xfs_buf *bp);
extern int xfs_dir2_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/xfs_dquot.c b/fs/xfs/xfs_dquot.c
index eff7586..d6d4d6b 100644
--- a/fs/xfs/xfs_dquot.c
+++ b/fs/xfs/xfs_dquot.c
@@ -249,6 +249,57 @@ xfs_qm_init_dquot_blk(
}
+STATIC void
+xfs_dquot_buf_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_dqblk *d = (struct xfs_dqblk *)bp->b_addr;
+ struct xfs_disk_dquot *ddq;
+ xfs_dqid_t id = 0;
+ int i;
+
+ /*
+ * On the first read of the buffer, verify that each dquot is valid.
+ * We don't know what the id of the dquot is supposed to be, just that
+ * they should be increasing monotonically within the buffer. If the
+ * first id is corrupt, then it will fail on the second dquot in the
+ * buffer so corruptions could point to the wrong dquot in this case.
+ */
+ for (i = 0; i < mp->m_quotainfo->qi_dqperchunk; i++) {
+ int error;
+
+ ddq = &d[i].dd_diskdq;
+
+ if (i == 0)
+ id = be32_to_cpu(ddq->d_id);
+
+ error = xfs_qm_dqcheck(mp, ddq, id + i, 0, XFS_QMOPT_DOWARN,
+ "xfs_dquot_read_verify");
+ if (error) {
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, d);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ break;
+ }
+ }
+}
+
+static void
+xfs_dquot_buf_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dquot_buf_verify(bp);
+}
+
+static void
+xfs_dquot_buf_read_verify(
+ struct xfs_buf *bp)
+{
+ xfs_dquot_buf_verify(bp);
+ bp->b_pre_io = xfs_dquot_buf_write_verify;
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
/*
* Allocate a block and fill it with dquots.
@@ -315,6 +366,7 @@ xfs_qm_dqalloc(
error = xfs_buf_geterror(bp);
if (error)
goto error1;
+ bp->b_pre_io = xfs_dquot_buf_write_verify;
/*
* Make a chunk of dquots out of this buffer and log
@@ -360,58 +412,6 @@ xfs_qm_dqalloc(
return (error);
}
-STATIC void
-xfs_dquot_buf_verify(
- struct xfs_buf *bp)
-{
- struct xfs_mount *mp = bp->b_target->bt_mount;
- struct xfs_dqblk *d = (struct xfs_dqblk *)bp->b_addr;
- struct xfs_disk_dquot *ddq;
- xfs_dqid_t id = 0;
- int i;
-
- /*
- * On the first read of the buffer, verify that each dquot is valid.
- * We don't know what the id of the dquot is supposed to be, just that
- * they should be increasing monotonically within the buffer. If the
- * first id is corrupt, then it will fail on the second dquot in the
- * buffer so corruptions could point to the wrong dquot in this case.
- */
- for (i = 0; i < mp->m_quotainfo->qi_dqperchunk; i++) {
- int error;
-
- ddq = &d[i].dd_diskdq;
-
- if (i == 0)
- id = be32_to_cpu(ddq->d_id);
-
- error = xfs_qm_dqcheck(mp, ddq, id + i, 0, XFS_QMOPT_DOWARN,
- "xfs_dquot_read_verify");
- if (error) {
- XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, d);
- xfs_buf_ioerror(bp, EFSCORRUPTED);
- break;
- }
- }
-}
-
-static void
-xfs_dquot_buf_write_verify(
- struct xfs_buf *bp)
-{
- xfs_dquot_buf_verify(bp);
-}
-
-static void
-xfs_dquot_buf_read_verify(
- struct xfs_buf *bp)
-{
- xfs_dquot_buf_verify(bp);
- bp->b_pre_io = xfs_dquot_buf_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
-}
-
STATIC int
xfs_qm_dqrepair(
struct xfs_mount *mp,
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index 302b99c..8ebc457 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -200,6 +200,7 @@ xfs_growfs_data_private(
error = ENOMEM;
goto error0;
}
+ bp->b_pre_io = xfs_agf_write_verify;
agf = XFS_BUF_TO_AGF(bp);
memset(agf, 0, mp->m_sb.sb_sectsize);
agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
@@ -237,6 +238,7 @@ xfs_growfs_data_private(
error = ENOMEM;
goto error0;
}
+ bp->b_pre_io = xfs_agi_write_verify;
agi = XFS_BUF_TO_AGI(bp);
memset(agi, 0, mp->m_sb.sb_sectsize);
agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
@@ -419,9 +421,10 @@ xfs_growfs_data_private(
bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
XFS_FSS_TO_BB(mp, 1), 0);
- if (bp)
+ if (bp) {
xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
- else
+ bp->b_pre_io = xfs_sb_write_verify;
+ } else
error = ENOMEM;
}
diff --git a/fs/xfs/xfs_ialloc.c b/fs/xfs/xfs_ialloc.c
index f260a36..12d1c94 100644
--- a/fs/xfs/xfs_ialloc.c
+++ b/fs/xfs/xfs_ialloc.c
@@ -210,6 +210,7 @@ xfs_ialloc_inode_init(
* to log a whole cluster of inodes instead of all the
* individual transactions causing a lot of log traffic.
*/
+ fbuf->b_pre_io = xfs_inode_buf_write_verify;
xfs_buf_zero(fbuf, 0, ninodes << mp->m_sb.sb_inodelog);
for (i = 0; i < ninodes; i++) {
int ioffset = i << mp->m_sb.sb_inodelog;
@@ -1495,14 +1496,14 @@ xfs_agi_verify(
xfs_check_agi_unlinked(agi);
}
-static void
+void
xfs_agi_write_verify(
struct xfs_buf *bp)
{
xfs_agi_verify(bp);
}
-void
+static void
xfs_agi_read_verify(
struct xfs_buf *bp)
{
diff --git a/fs/xfs/xfs_ialloc.h b/fs/xfs/xfs_ialloc.h
index 1fd6ea4..7a169e3 100644
--- a/fs/xfs/xfs_ialloc.h
+++ b/fs/xfs/xfs_ialloc.h
@@ -147,7 +147,9 @@ int xfs_inobt_lookup(struct xfs_btree_cur *cur, xfs_agino_t ino,
/*
* Get the data from the pointed-to record.
*/
-extern int xfs_inobt_get_rec(struct xfs_btree_cur *cur,
+int xfs_inobt_get_rec(struct xfs_btree_cur *cur,
xfs_inobt_rec_incore_t *rec, int *stat);
+void xfs_agi_write_verify(struct xfs_buf *bp);
+
#endif /* __XFS_IALLOC_H__ */
diff --git a/fs/xfs/xfs_ialloc_btree.c b/fs/xfs/xfs_ialloc_btree.c
index 15a79f8..7761e1e 100644
--- a/fs/xfs/xfs_ialloc_btree.c
+++ b/fs/xfs/xfs_ialloc_btree.c
@@ -271,6 +271,7 @@ static const struct xfs_btree_ops xfs_inobt_ops = {
.init_ptr_from_cur = xfs_inobt_init_ptr_from_cur,
.key_diff = xfs_inobt_key_diff,
.read_verify = xfs_inobt_read_verify,
+ .write_verify = xfs_inobt_write_verify,
#ifdef DEBUG
.keys_inorder = xfs_inobt_keys_inorder,
.recs_inorder = xfs_inobt_recs_inorder,
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 875ceb2..4f2e99a 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -420,7 +420,7 @@ xfs_inode_buf_verify(
xfs_inobp_check(mp, bp);
}
-static void
+void
xfs_inode_buf_write_verify(
struct xfs_buf *bp)
{
@@ -1781,6 +1781,18 @@ xfs_ifree_cluster(
if (!bp)
return ENOMEM;
+
+ /*
+ * This buffer may not have been correctly initialised as we
+ * didn't read it from disk. That's not important because we are
+ * only using to mark the buffer as stale in the log, and to
+ * attach stale cached inodes on it. That means it will never be
+ * dispatched for IO. If it is, we want to know about it, and we
+ * want it to fail. We can acheive this by adding a write
+ * verifier to the buffer.
+ */
+ bp->b_pre_io = xfs_inode_buf_write_verify;
+
/*
* Walk the inodes already attached to the buffer and mark them
* stale. These will all have the flush locks held, so an
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index 32b6d70..a866d28 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -555,6 +555,7 @@ int xfs_imap_to_bp(struct xfs_mount *, struct xfs_trans *,
int xfs_iread(struct xfs_mount *, struct xfs_trans *,
struct xfs_inode *, uint);
void xfs_inode_buf_read_verify(struct xfs_buf *);
+void xfs_inode_buf_write_verify(struct xfs_buf *);
void xfs_dinode_to_disk(struct xfs_dinode *,
struct xfs_icdinode *);
void xfs_idestroy_fork(struct xfs_inode *, int);
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index a622f6f..18b853e 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -631,7 +631,7 @@ xfs_sb_verify(
xfs_buf_ioerror(bp, error);
}
-static void
+void
xfs_sb_write_verify(
struct xfs_buf *bp)
{
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 82b8fda..69e4082 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -383,6 +383,7 @@ extern void xfs_set_low_space_thresholds(struct xfs_mount *);
#endif /* __KERNEL__ */
extern void xfs_sb_read_verify(struct xfs_buf *);
+extern void xfs_sb_write_verify(struct xfs_buf *bp);
extern void xfs_mod_sb(struct xfs_trans *, __int64_t);
extern int xfs_initialize_perag(struct xfs_mount *, xfs_agnumber_t,
xfs_agnumber_t *);
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 24/25] xfs: convert buffer verifiers to an ops structure.
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (22 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 23/25] xfs: connect up write verifiers to new buffers Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-30 13:41 ` Phil White
2012-10-25 6:34 ` [PATCH 25/25] xfs: add write verifiers to log recovery Dave Chinner
24 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
To separate the verifiers from iodone functions and associate read
and write verifiers at the same time, introduce a buffer verifier
operations structure to the xfs_buf.
This avoids the need for assigning the write verifier, clearing the
iodone function and re-running ioend processing in the read
verifier, and gets rid of the nasty "b_pre_io" name for the write
verifier function pointer. If we ever need to, it will also be
easier to add further content specific callbacks to a buffer with an
ops structure in place.
We also avoid needing to export verifier functions, instead we
can simply export the ops structures for those that are needed
outside the function they are defined in.
This patch also fixes a directory block readahead verifier issue
it exposed.
This patch also adds ops callbacks to the inode/alloc btree blocks
initialised by growfs. These will need more work before they will
work with CRCs.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_ag.h | 4 +++
fs/xfs/xfs_alloc.c | 26 +++++++++++--------
fs/xfs/xfs_alloc.h | 2 +-
fs/xfs/xfs_alloc_btree.c | 18 +++++++------
fs/xfs/xfs_alloc_btree.h | 2 ++
fs/xfs/xfs_attr_leaf.c | 19 +++++++-------
fs/xfs/xfs_attr_leaf.h | 3 ++-
fs/xfs/xfs_bmap.c | 22 ++++++++--------
fs/xfs/xfs_bmap_btree.c | 20 +++++++-------
fs/xfs/xfs_bmap_btree.h | 3 +--
fs/xfs/xfs_btree.c | 26 +++++++++----------
fs/xfs/xfs_btree.h | 9 +++----
fs/xfs/xfs_buf.c | 63 ++++++++++++++++++++++++++-------------------
fs/xfs/xfs_buf.h | 24 ++++++++++-------
fs/xfs/xfs_da_btree.c | 28 ++++++++++----------
fs/xfs/xfs_da_btree.h | 4 +--
fs/xfs/xfs_dir2_block.c | 20 +++++++-------
fs/xfs/xfs_dir2_data.c | 52 ++++++++++++++++++++++++++++++-------
fs/xfs/xfs_dir2_leaf.c | 36 ++++++++++++++------------
fs/xfs/xfs_dir2_node.c | 26 ++++++++++---------
fs/xfs/xfs_dir2_priv.h | 10 ++++---
fs/xfs/xfs_dquot.c | 16 +++++++-----
fs/xfs/xfs_fsops.c | 26 ++++++++++++++++---
fs/xfs/xfs_ialloc.c | 18 +++++++------
fs/xfs/xfs_ialloc.h | 2 +-
fs/xfs/xfs_ialloc_btree.c | 17 ++++++------
fs/xfs/xfs_ialloc_btree.h | 2 ++
fs/xfs/xfs_inode.c | 22 +++++++++-------
fs/xfs/xfs_inode.h | 3 +--
fs/xfs/xfs_itable.c | 2 +-
fs/xfs/xfs_log_recover.c | 2 +-
fs/xfs/xfs_mount.c | 35 +++++++++++++++----------
fs/xfs/xfs_mount.h | 4 +--
fs/xfs/xfs_trans.h | 6 ++---
fs/xfs/xfs_trans_buf.c | 8 +++---
35 files changed, 346 insertions(+), 234 deletions(-)
diff --git a/fs/xfs/xfs_ag.h b/fs/xfs/xfs_ag.h
index 44d65c1..48c873a 100644
--- a/fs/xfs/xfs_ag.h
+++ b/fs/xfs/xfs_ag.h
@@ -108,6 +108,8 @@ typedef struct xfs_agf {
extern int xfs_read_agf(struct xfs_mount *mp, struct xfs_trans *tp,
xfs_agnumber_t agno, int flags, struct xfs_buf **bpp);
+extern const struct xfs_buf_ops xfs_agf_buf_ops;
+
/*
* Size of the unlinked inode hash table in the agi.
*/
@@ -161,6 +163,8 @@ typedef struct xfs_agi {
extern int xfs_read_agi(struct xfs_mount *mp, struct xfs_trans *tp,
xfs_agnumber_t agno, struct xfs_buf **bpp);
+extern const struct xfs_buf_ops xfs_agi_buf_ops;
+
/*
* The third a.g. block contains the a.g. freelist, an array
* of block pointers to blocks owned by the allocation btree code.
diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
index 578d2e8..f9231b2 100644
--- a/fs/xfs/xfs_alloc.c
+++ b/fs/xfs/xfs_alloc.c
@@ -477,11 +477,13 @@ xfs_agfl_read_verify(
struct xfs_buf *bp)
{
xfs_agfl_verify(bp);
- bp->b_pre_io = xfs_agfl_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+static const struct xfs_buf_ops xfs_agfl_buf_ops = {
+ .verify_read = xfs_agfl_read_verify,
+ .verify_write = xfs_agfl_write_verify,
+};
+
/*
* Read in the allocation group free block array.
*/
@@ -499,7 +501,7 @@ xfs_alloc_read_agfl(
error = xfs_trans_read_buf(
mp, tp, mp->m_ddev_targp,
XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
- XFS_FSS_TO_BB(mp, 1), 0, &bp, xfs_agfl_read_verify);
+ XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_agfl_buf_ops);
if (error)
return error;
ASSERT(!xfs_buf_geterror(bp));
@@ -2173,23 +2175,25 @@ xfs_agf_verify(
}
}
-void
-xfs_agf_write_verify(
+static void
+xfs_agf_read_verify(
struct xfs_buf *bp)
{
xfs_agf_verify(bp);
}
static void
-xfs_agf_read_verify(
+xfs_agf_write_verify(
struct xfs_buf *bp)
{
xfs_agf_verify(bp);
- bp->b_pre_io = xfs_agf_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+const struct xfs_buf_ops xfs_agf_buf_ops = {
+ .verify_read = xfs_agf_read_verify,
+ .verify_write = xfs_agf_write_verify,
+};
+
/*
* Read in the allocation group header (free/alloc section).
*/
@@ -2207,7 +2211,7 @@ xfs_read_agf(
error = xfs_trans_read_buf(
mp, tp, mp->m_ddev_targp,
XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
- XFS_FSS_TO_BB(mp, 1), flags, bpp, xfs_agf_read_verify);
+ XFS_FSS_TO_BB(mp, 1), flags, bpp, &xfs_agf_buf_ops);
if (error)
return error;
if (!*bpp)
diff --git a/fs/xfs/xfs_alloc.h b/fs/xfs/xfs_alloc.h
index 4b6c2c4..aaf7ff1 100644
--- a/fs/xfs/xfs_alloc.h
+++ b/fs/xfs/xfs_alloc.h
@@ -231,6 +231,6 @@ xfs_alloc_get_rec(
xfs_extlen_t *len, /* output: length of extent */
int *stat); /* output: success/failure */
-void xfs_agf_write_verify(struct xfs_buf *bp);
+extern const struct xfs_buf_ops xfs_agf_buf_ops;
#endif /* __XFS_ALLOC_H__ */
diff --git a/fs/xfs/xfs_alloc_btree.c b/fs/xfs/xfs_alloc_btree.c
index 7f8e704..b14ff21 100644
--- a/fs/xfs/xfs_alloc_btree.c
+++ b/fs/xfs/xfs_alloc_btree.c
@@ -317,22 +317,25 @@ xfs_allocbt_verify(
}
static void
-xfs_allocbt_write_verify(
+xfs_allocbt_read_verify(
struct xfs_buf *bp)
{
xfs_allocbt_verify(bp);
}
-void
-xfs_allocbt_read_verify(
+static void
+xfs_allocbt_write_verify(
struct xfs_buf *bp)
{
xfs_allocbt_verify(bp);
- bp->b_pre_io = xfs_allocbt_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+const struct xfs_buf_ops xfs_allocbt_buf_ops = {
+ .verify_read = xfs_allocbt_read_verify,
+ .verify_write = xfs_allocbt_write_verify,
+};
+
+
#ifdef DEBUG
STATIC int
xfs_allocbt_keys_inorder(
@@ -388,8 +391,7 @@ static const struct xfs_btree_ops xfs_allocbt_ops = {
.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,
- .read_verify = xfs_allocbt_read_verify,
- .write_verify = xfs_allocbt_write_verify,
+ .buf_ops = &xfs_allocbt_buf_ops,
#ifdef DEBUG
.keys_inorder = xfs_allocbt_keys_inorder,
.recs_inorder = xfs_allocbt_recs_inorder,
diff --git a/fs/xfs/xfs_alloc_btree.h b/fs/xfs/xfs_alloc_btree.h
index 359fb86..7e89a2b 100644
--- a/fs/xfs/xfs_alloc_btree.h
+++ b/fs/xfs/xfs_alloc_btree.h
@@ -93,4 +93,6 @@ extern struct xfs_btree_cur *xfs_allocbt_init_cursor(struct xfs_mount *,
xfs_agnumber_t, xfs_btnum_t);
extern int xfs_allocbt_maxrecs(struct xfs_mount *, int, int);
+extern const struct xfs_buf_ops xfs_allocbt_buf_ops;
+
#endif /* __XFS_ALLOC_BTREE_H__ */
diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c
index 5d56886..d790860 100644
--- a/fs/xfs/xfs_attr_leaf.c
+++ b/fs/xfs/xfs_attr_leaf.c
@@ -103,22 +103,23 @@ xfs_attr_leaf_verify(
}
static void
-xfs_attr_leaf_write_verify(
+xfs_attr_leaf_read_verify(
struct xfs_buf *bp)
{
xfs_attr_leaf_verify(bp);
}
-void
-xfs_attr_leaf_read_verify(
+static void
+xfs_attr_leaf_write_verify(
struct xfs_buf *bp)
{
xfs_attr_leaf_verify(bp);
- bp->b_pre_io = xfs_attr_leaf_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+const struct xfs_buf_ops xfs_attr_leaf_buf_ops = {
+ .verify_read = xfs_attr_leaf_read_verify,
+ .verify_write = xfs_attr_leaf_write_verify,
+};
int
xfs_attr_leaf_read(
@@ -129,7 +130,7 @@ xfs_attr_leaf_read(
struct xfs_buf **bpp)
{
return xfs_da_read_buf(tp, dp, bno, mappedbno, bpp,
- XFS_ATTR_FORK, xfs_attr_leaf_read_verify);
+ XFS_ATTR_FORK, &xfs_attr_leaf_buf_ops);
}
/*========================================================================
@@ -923,7 +924,7 @@ xfs_attr_leaf_to_node(xfs_da_args_t *args)
XFS_ATTR_FORK);
if (error)
goto out;
- bp2->b_pre_io = bp1->b_pre_io;
+ bp2->b_ops = bp1->b_ops;
memcpy(bp2->b_addr, bp1->b_addr, XFS_LBSIZE(dp->i_mount));
bp1 = NULL;
xfs_trans_log_buf(args->trans, bp2, 0, XFS_LBSIZE(dp->i_mount) - 1);
@@ -977,7 +978,7 @@ xfs_attr_leaf_create(
XFS_ATTR_FORK);
if (error)
return(error);
- bp->b_pre_io = xfs_attr_leaf_write_verify;
+ bp->b_ops = &xfs_attr_leaf_buf_ops;
leaf = bp->b_addr;
memset((char *)leaf, 0, XFS_LBSIZE(dp->i_mount));
hdr = &leaf->hdr;
diff --git a/fs/xfs/xfs_attr_leaf.h b/fs/xfs/xfs_attr_leaf.h
index 3bbf627..77de139 100644
--- a/fs/xfs/xfs_attr_leaf.h
+++ b/fs/xfs/xfs_attr_leaf.h
@@ -264,6 +264,7 @@ int xfs_attr_leaf_newentsize(int namelen, int valuelen, int blocksize,
int xfs_attr_leaf_read(struct xfs_trans *tp, struct xfs_inode *dp,
xfs_dablk_t bno, xfs_daddr_t mappedbno,
struct xfs_buf **bpp);
-void xfs_attr_leaf_read_verify(struct xfs_buf *bp);
+
+extern const struct xfs_buf_ops xfs_attr_leaf_buf_ops;
#endif /* __XFS_ATTR_LEAF_H__ */
diff --git a/fs/xfs/xfs_bmap.c b/fs/xfs/xfs_bmap.c
index fe5438b..4fb8276 100644
--- a/fs/xfs/xfs_bmap.c
+++ b/fs/xfs/xfs_bmap.c
@@ -2663,7 +2663,7 @@ xfs_bmap_btree_to_extents(
return error;
#endif
error = xfs_btree_read_bufl(mp, tp, cbno, 0, &cbp, XFS_BMAP_BTREE_REF,
- xfs_bmbt_read_verify);
+ &xfs_bmbt_buf_ops);
if (error)
return error;
cblock = XFS_BUF_TO_BLOCK(cbp);
@@ -3124,7 +3124,7 @@ xfs_bmap_extents_to_btree(
/*
* Fill in the child block.
*/
- abp->b_pre_io = xfs_bmbt_write_verify;
+ abp->b_ops = &xfs_bmbt_buf_ops;
ablock = XFS_BUF_TO_BLOCK(abp);
ablock->bb_magic = cpu_to_be32(XFS_BMAP_MAGIC);
ablock->bb_level = 0;
@@ -3271,7 +3271,7 @@ xfs_bmap_local_to_extents(
ASSERT(args.len == 1);
*firstblock = args.fsbno;
bp = xfs_btree_get_bufl(args.mp, tp, args.fsbno, 0);
- bp->b_pre_io = xfs_bmbt_write_verify;
+ bp->b_ops = &xfs_bmbt_buf_ops;
memcpy(bp->b_addr, ifp->if_u1.if_data, ifp->if_bytes);
xfs_trans_log_buf(tp, bp, 0, ifp->if_bytes - 1);
xfs_bmap_forkoff_reset(args.mp, ip, whichfork);
@@ -4082,7 +4082,7 @@ xfs_bmap_read_extents(
*/
while (level-- > 0) {
error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
- XFS_BMAP_BTREE_REF, xfs_bmbt_read_verify);
+ XFS_BMAP_BTREE_REF, &xfs_bmbt_buf_ops);
if (error)
return error;
block = XFS_BUF_TO_BLOCK(bp);
@@ -4129,7 +4129,7 @@ xfs_bmap_read_extents(
nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
if (nextbno != NULLFSBLOCK)
xfs_btree_reada_bufl(mp, nextbno, 1,
- xfs_bmbt_read_verify);
+ &xfs_bmbt_buf_ops);
/*
* Copy records into the extent records.
*/
@@ -4162,7 +4162,7 @@ xfs_bmap_read_extents(
if (bno == NULLFSBLOCK)
break;
error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
- XFS_BMAP_BTREE_REF, xfs_bmbt_read_verify);
+ XFS_BMAP_BTREE_REF, &xfs_bmbt_buf_ops);
if (error)
return error;
block = XFS_BUF_TO_BLOCK(bp);
@@ -5880,7 +5880,7 @@ xfs_bmap_check_leaf_extents(
bp_release = 1;
error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
XFS_BMAP_BTREE_REF,
- xfs_bmbt_read_verify);
+ &xfs_bmbt_buf_ops);
if (error)
goto error_norelse;
}
@@ -5966,7 +5966,7 @@ xfs_bmap_check_leaf_extents(
bp_release = 1;
error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
XFS_BMAP_BTREE_REF,
- xfs_bmbt_read_verify);
+ &xfs_bmbt_buf_ops);
if (error)
goto error_norelse;
}
@@ -6061,7 +6061,7 @@ xfs_bmap_count_tree(
int numrecs;
error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp, XFS_BMAP_BTREE_REF,
- xfs_bmbt_read_verify);
+ &xfs_bmbt_buf_ops);
if (error)
return error;
*count += 1;
@@ -6073,7 +6073,7 @@ xfs_bmap_count_tree(
while (nextbno != NULLFSBLOCK) {
error = xfs_btree_read_bufl(mp, tp, nextbno, 0, &nbp,
XFS_BMAP_BTREE_REF,
- xfs_bmbt_read_verify);
+ &xfs_bmbt_buf_ops);
if (error)
return error;
*count += 1;
@@ -6105,7 +6105,7 @@ xfs_bmap_count_tree(
bno = nextbno;
error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
XFS_BMAP_BTREE_REF,
- xfs_bmbt_read_verify);
+ &xfs_bmbt_buf_ops);
if (error)
return error;
*count += 1;
diff --git a/fs/xfs/xfs_bmap_btree.c b/fs/xfs/xfs_bmap_btree.c
index 79758e1..061b45c 100644
--- a/fs/xfs/xfs_bmap_btree.c
+++ b/fs/xfs/xfs_bmap_btree.c
@@ -749,23 +749,26 @@ xfs_bmbt_verify(
}
}
-void
-xfs_bmbt_write_verify(
+static void
+xfs_bmbt_read_verify(
struct xfs_buf *bp)
{
xfs_bmbt_verify(bp);
}
-void
-xfs_bmbt_read_verify(
+static void
+xfs_bmbt_write_verify(
struct xfs_buf *bp)
{
xfs_bmbt_verify(bp);
- bp->b_pre_io = xfs_bmbt_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+const struct xfs_buf_ops xfs_bmbt_buf_ops = {
+ .verify_read = xfs_bmbt_read_verify,
+ .verify_write = xfs_bmbt_write_verify,
+};
+
+
#ifdef DEBUG
STATIC int
xfs_bmbt_keys_inorder(
@@ -805,8 +808,7 @@ static const struct xfs_btree_ops xfs_bmbt_ops = {
.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,
- .read_verify = xfs_bmbt_read_verify,
- .write_verify = xfs_bmbt_write_verify,
+ .buf_ops = &xfs_bmbt_buf_ops,
#ifdef DEBUG
.keys_inorder = xfs_bmbt_keys_inorder,
.recs_inorder = xfs_bmbt_recs_inorder,
diff --git a/fs/xfs/xfs_bmap_btree.h b/fs/xfs/xfs_bmap_btree.h
index 938c859..88469ca 100644
--- a/fs/xfs/xfs_bmap_btree.h
+++ b/fs/xfs/xfs_bmap_btree.h
@@ -232,11 +232,10 @@ extern void xfs_bmbt_to_bmdr(struct xfs_mount *, struct xfs_btree_block *, int,
extern int xfs_bmbt_get_maxrecs(struct xfs_btree_cur *, int level);
extern int xfs_bmdr_maxrecs(struct xfs_mount *, int blocklen, int leaf);
extern int xfs_bmbt_maxrecs(struct xfs_mount *, int blocklen, int leaf);
-extern void xfs_bmbt_read_verify(struct xfs_buf *bp);
-extern void xfs_bmbt_write_verify(struct xfs_buf *bp);
extern struct xfs_btree_cur *xfs_bmbt_init_cursor(struct xfs_mount *,
struct xfs_trans *, struct xfs_inode *, int);
+extern const struct xfs_buf_ops xfs_bmbt_buf_ops;
#endif /* __XFS_BMAP_BTREE_H__ */
diff --git a/fs/xfs/xfs_btree.c b/fs/xfs/xfs_btree.c
index 1ccd42b..79039fa 100644
--- a/fs/xfs/xfs_btree.c
+++ b/fs/xfs/xfs_btree.c
@@ -271,7 +271,7 @@ xfs_btree_dup_cursor(
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
XFS_BUF_ADDR(bp), mp->m_bsize,
0, &bp,
- cur->bc_ops->read_verify);
+ cur->bc_ops->buf_ops);
if (error) {
xfs_btree_del_cursor(new, error);
*ncur = NULL;
@@ -621,7 +621,7 @@ xfs_btree_read_bufl(
uint lock, /* lock flags for read_buf */
struct xfs_buf **bpp, /* buffer for fsbno */
int refval, /* ref count value for buffer */
- xfs_buf_iodone_t verify)
+ const struct xfs_buf_ops *ops)
{
struct xfs_buf *bp; /* return value */
xfs_daddr_t d; /* real disk block address */
@@ -630,7 +630,7 @@ xfs_btree_read_bufl(
ASSERT(fsbno != NULLFSBLOCK);
d = XFS_FSB_TO_DADDR(mp, fsbno);
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, d,
- mp->m_bsize, lock, &bp, verify);
+ mp->m_bsize, lock, &bp, ops);
if (error)
return error;
ASSERT(!xfs_buf_geterror(bp));
@@ -650,13 +650,13 @@ xfs_btree_reada_bufl(
struct xfs_mount *mp, /* file system mount point */
xfs_fsblock_t fsbno, /* file system block number */
xfs_extlen_t count, /* count of filesystem blocks */
- xfs_buf_iodone_t verify)
+ const struct xfs_buf_ops *ops)
{
xfs_daddr_t d;
ASSERT(fsbno != NULLFSBLOCK);
d = XFS_FSB_TO_DADDR(mp, fsbno);
- xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, verify);
+ xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, ops);
}
/*
@@ -670,14 +670,14 @@ xfs_btree_reada_bufs(
xfs_agnumber_t agno, /* allocation group number */
xfs_agblock_t agbno, /* allocation group block number */
xfs_extlen_t count, /* count of filesystem blocks */
- xfs_buf_iodone_t verify)
+ const struct xfs_buf_ops *ops)
{
xfs_daddr_t d;
ASSERT(agno != NULLAGNUMBER);
ASSERT(agbno != NULLAGBLOCK);
d = XFS_AGB_TO_DADDR(mp, agno, agbno);
- xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, verify);
+ xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, ops);
}
STATIC int
@@ -692,13 +692,13 @@ xfs_btree_readahead_lblock(
if ((lr & XFS_BTCUR_LEFTRA) && left != NULLDFSBNO) {
xfs_btree_reada_bufl(cur->bc_mp, left, 1,
- cur->bc_ops->read_verify);
+ cur->bc_ops->buf_ops);
rval++;
}
if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLDFSBNO) {
xfs_btree_reada_bufl(cur->bc_mp, right, 1,
- cur->bc_ops->read_verify);
+ cur->bc_ops->buf_ops);
rval++;
}
@@ -718,13 +718,13 @@ xfs_btree_readahead_sblock(
if ((lr & XFS_BTCUR_LEFTRA) && left != NULLAGBLOCK) {
xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
- left, 1, cur->bc_ops->read_verify);
+ left, 1, cur->bc_ops->buf_ops);
rval++;
}
if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLAGBLOCK) {
xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
- right, 1, cur->bc_ops->read_verify);
+ right, 1, cur->bc_ops->buf_ops);
rval++;
}
@@ -981,7 +981,7 @@ xfs_btree_get_buf_block(
if (!*bpp)
return ENOMEM;
- (*bpp)->b_pre_io = cur->bc_ops->write_verify;
+ (*bpp)->b_ops = cur->bc_ops->buf_ops;
*block = XFS_BUF_TO_BLOCK(*bpp);
return 0;
}
@@ -1009,7 +1009,7 @@ xfs_btree_read_buf_block(
d = xfs_btree_ptr_to_daddr(cur, ptr);
error = xfs_trans_read_buf(mp, cur->bc_tp, mp->m_ddev_targp, d,
mp->m_bsize, flags, bpp,
- cur->bc_ops->read_verify);
+ cur->bc_ops->buf_ops);
if (error)
return error;
diff --git a/fs/xfs/xfs_btree.h b/fs/xfs/xfs_btree.h
index 2ac403a..c7f187a 100644
--- a/fs/xfs/xfs_btree.h
+++ b/fs/xfs/xfs_btree.h
@@ -188,8 +188,7 @@ struct xfs_btree_ops {
__int64_t (*key_diff)(struct xfs_btree_cur *cur,
union xfs_btree_key *key);
- void (*read_verify)(struct xfs_buf *bp);
- void (*write_verify)(struct xfs_buf *bp);
+ const struct xfs_buf_ops *buf_ops;
#ifdef DEBUG
/* check that k1 is lower than k2 */
@@ -359,7 +358,7 @@ xfs_btree_read_bufl(
uint lock, /* lock flags for read_buf */
struct xfs_buf **bpp, /* buffer for fsbno */
int refval, /* ref count value for buffer */
- xfs_buf_iodone_t verify);
+ const struct xfs_buf_ops *ops);
/*
* Read-ahead the block, don't wait for it, don't return a buffer.
@@ -370,7 +369,7 @@ xfs_btree_reada_bufl(
struct xfs_mount *mp, /* file system mount point */
xfs_fsblock_t fsbno, /* file system block number */
xfs_extlen_t count, /* count of filesystem blocks */
- xfs_buf_iodone_t verify);
+ const struct xfs_buf_ops *ops);
/*
* Read-ahead the block, don't wait for it, don't return a buffer.
@@ -382,7 +381,7 @@ xfs_btree_reada_bufs(
xfs_agnumber_t agno, /* allocation group number */
xfs_agblock_t agbno, /* allocation group block number */
xfs_extlen_t count, /* count of filesystem blocks */
- xfs_buf_iodone_t verify);
+ const struct xfs_buf_ops *ops);
/*
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index c073236..aacd017 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -571,7 +571,7 @@ found:
ASSERT((bp->b_flags & _XBF_DELWRI_Q) == 0);
ASSERT(bp->b_iodone == NULL);
bp->b_flags &= _XBF_KMEM | _XBF_PAGES;
- bp->b_pre_io = NULL;
+ bp->b_ops = NULL;
}
trace_xfs_buf_find(bp, flags, _RET_IP_);
@@ -657,7 +657,7 @@ xfs_buf_read_map(
struct xfs_buf_map *map,
int nmaps,
xfs_buf_flags_t flags,
- xfs_buf_iodone_t verify)
+ const struct xfs_buf_ops *ops)
{
struct xfs_buf *bp;
@@ -669,7 +669,7 @@ xfs_buf_read_map(
if (!XFS_BUF_ISDONE(bp)) {
XFS_STATS_INC(xb_get_read);
- bp->b_iodone = verify;
+ bp->b_ops = ops;
_xfs_buf_read(bp, flags);
} else if (flags & XBF_ASYNC) {
/*
@@ -696,13 +696,13 @@ xfs_buf_readahead_map(
struct xfs_buftarg *target,
struct xfs_buf_map *map,
int nmaps,
- xfs_buf_iodone_t verify)
+ const struct xfs_buf_ops *ops)
{
if (bdi_read_congested(target->bt_bdi))
return;
xfs_buf_read_map(target, map, nmaps,
- XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD, verify);
+ XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD, ops);
}
/*
@@ -715,7 +715,7 @@ xfs_buf_read_uncached(
xfs_daddr_t daddr,
size_t numblks,
int flags,
- xfs_buf_iodone_t verify)
+ const struct xfs_buf_ops *ops)
{
struct xfs_buf *bp;
@@ -728,7 +728,7 @@ xfs_buf_read_uncached(
bp->b_bn = daddr;
bp->b_maps[0].bm_bn = daddr;
bp->b_flags |= XBF_READ;
- bp->b_iodone = verify;
+ bp->b_ops = ops;
xfsbdstrat(target->bt_mount, bp);
xfs_buf_iowait(bp);
@@ -1001,27 +1001,37 @@ STATIC void
xfs_buf_iodone_work(
struct work_struct *work)
{
- xfs_buf_t *bp =
+ struct xfs_buf *bp =
container_of(work, xfs_buf_t, b_iodone_work);
+ bool read = !!(bp->b_flags & XBF_READ);
+
+ bp->b_flags &= ~(XBF_READ | XBF_WRITE | XBF_READ_AHEAD);
+ if (read && bp->b_ops)
+ bp->b_ops->verify_read(bp);
if (bp->b_iodone)
(*(bp->b_iodone))(bp);
else if (bp->b_flags & XBF_ASYNC)
xfs_buf_relse(bp);
+ else {
+ ASSERT(read && bp->b_ops);
+ complete(&bp->b_iowait);
+ }
}
void
xfs_buf_ioend(
- xfs_buf_t *bp,
- int schedule)
+ struct xfs_buf *bp,
+ int schedule)
{
+ bool read = !!(bp->b_flags & XBF_READ);
+
trace_xfs_buf_iodone(bp, _RET_IP_);
- bp->b_flags &= ~(XBF_READ | XBF_WRITE | XBF_READ_AHEAD);
if (bp->b_error == 0)
bp->b_flags |= XBF_DONE;
- if ((bp->b_iodone) || (bp->b_flags & XBF_ASYNC)) {
+ if (bp->b_iodone || (read && bp->b_ops) || (bp->b_flags & XBF_ASYNC)) {
if (schedule) {
INIT_WORK(&bp->b_iodone_work, xfs_buf_iodone_work);
queue_work(xfslogd_workqueue, &bp->b_iodone_work);
@@ -1029,6 +1039,7 @@ xfs_buf_ioend(
xfs_buf_iodone_work(&bp->b_iodone_work);
}
} else {
+ bp->b_flags &= ~(XBF_READ | XBF_WRITE | XBF_READ_AHEAD);
complete(&bp->b_iowait);
}
}
@@ -1306,6 +1317,20 @@ _xfs_buf_ioapply(
rw |= REQ_FUA;
if (bp->b_flags & XBF_FLUSH)
rw |= REQ_FLUSH;
+
+ /*
+ * Run the write verifier callback function if it exists. If
+ * this function fails it will mark the buffer with an error and
+ * the IO should not be dispatched.
+ */
+ if (bp->b_ops) {
+ bp->b_ops->verify_write(bp);
+ if (bp->b_error) {
+ xfs_force_shutdown(bp->b_target->bt_mount,
+ SHUTDOWN_CORRUPT_INCORE);
+ return;
+ }
+ }
} else if (bp->b_flags & XBF_READ_AHEAD) {
rw = READA;
} else {
@@ -1316,20 +1341,6 @@ _xfs_buf_ioapply(
rw |= REQ_META;
/*
- * run the pre-io callback function if it exists. If this function
- * fails it will mark the buffer with an error and the IO should
- * not be dispatched.
- */
- if (bp->b_pre_io) {
- bp->b_pre_io(bp);
- if (bp->b_error) {
- xfs_force_shutdown(bp->b_target->bt_mount,
- SHUTDOWN_CORRUPT_INCORE);
- return;
- }
- }
-
- /*
* Walk all the vectors issuing IO on them. Set up the initial offset
* into the buffer and the desired IO size before we start -
* _xfs_buf_ioapply_vec() will modify them appropriately for each
diff --git a/fs/xfs/xfs_buf.h b/fs/xfs/xfs_buf.h
index 51bc16a..23f5642 100644
--- a/fs/xfs/xfs_buf.h
+++ b/fs/xfs/xfs_buf.h
@@ -111,6 +111,11 @@ struct xfs_buf_map {
#define DEFINE_SINGLE_BUF_MAP(map, blkno, numblk) \
struct xfs_buf_map (map) = { .bm_bn = (blkno), .bm_len = (numblk) };
+struct xfs_buf_ops {
+ void (*verify_read)(struct xfs_buf *);
+ void (*verify_write)(struct xfs_buf *);
+};
+
typedef struct xfs_buf {
/*
* first cacheline holds all the fields needed for an uncontended cache
@@ -154,9 +159,7 @@ typedef struct xfs_buf {
unsigned int b_page_count; /* size of page array */
unsigned int b_offset; /* page offset in first page */
unsigned short b_error; /* error code on I/O */
-
- void (*b_pre_io)(struct xfs_buf *);
- /* pre-io callback function */
+ const struct xfs_buf_ops *b_ops;
#ifdef XFS_BUF_LOCK_TRACKING
int b_last_holder;
@@ -199,10 +202,11 @@ struct xfs_buf *xfs_buf_get_map(struct xfs_buftarg *target,
xfs_buf_flags_t flags);
struct xfs_buf *xfs_buf_read_map(struct xfs_buftarg *target,
struct xfs_buf_map *map, int nmaps,
- xfs_buf_flags_t flags, xfs_buf_iodone_t verify);
+ xfs_buf_flags_t flags,
+ const struct xfs_buf_ops *ops);
void xfs_buf_readahead_map(struct xfs_buftarg *target,
struct xfs_buf_map *map, int nmaps,
- xfs_buf_iodone_t verify);
+ const struct xfs_buf_ops *ops);
static inline struct xfs_buf *
xfs_buf_get(
@@ -221,10 +225,10 @@ xfs_buf_read(
xfs_daddr_t blkno,
size_t numblks,
xfs_buf_flags_t flags,
- xfs_buf_iodone_t verify)
+ const struct xfs_buf_ops *ops)
{
DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
- return xfs_buf_read_map(target, &map, 1, flags, verify);
+ return xfs_buf_read_map(target, &map, 1, flags, ops);
}
static inline void
@@ -232,10 +236,10 @@ xfs_buf_readahead(
struct xfs_buftarg *target,
xfs_daddr_t blkno,
size_t numblks,
- xfs_buf_iodone_t verify)
+ const struct xfs_buf_ops *ops)
{
DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
- return xfs_buf_readahead_map(target, &map, 1, verify);
+ return xfs_buf_readahead_map(target, &map, 1, ops);
}
struct xfs_buf *xfs_buf_get_empty(struct xfs_buftarg *target, size_t numblks);
@@ -246,7 +250,7 @@ struct xfs_buf *xfs_buf_get_uncached(struct xfs_buftarg *target, size_t numblks,
int flags);
struct xfs_buf *xfs_buf_read_uncached(struct xfs_buftarg *target,
xfs_daddr_t daddr, size_t numblks, int flags,
- xfs_buf_iodone_t verify);
+ const struct xfs_buf_ops *ops);
void xfs_buf_hold(struct xfs_buf *bp);
/* Releasing Buffers */
diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
index eff604f..816cade 100644
--- a/fs/xfs/xfs_da_btree.c
+++ b/fs/xfs/xfs_da_btree.c
@@ -128,10 +128,10 @@ xfs_da_node_read_verify(
xfs_da_node_verify(bp);
break;
case XFS_ATTR_LEAF_MAGIC:
- xfs_attr_leaf_read_verify(bp);
+ xfs_attr_leaf_buf_ops.verify_read(bp);
return;
case XFS_DIR2_LEAFN_MAGIC:
- xfs_dir2_leafn_read_verify(bp);
+ xfs_dir2_leafn_buf_ops.verify_read(bp);
return;
default:
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW,
@@ -139,12 +139,14 @@ xfs_da_node_read_verify(
xfs_buf_ioerror(bp, EFSCORRUPTED);
break;
}
-
- bp->b_pre_io = xfs_da_node_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+const struct xfs_buf_ops xfs_da_node_buf_ops = {
+ .verify_read = xfs_da_node_read_verify,
+ .verify_write = xfs_da_node_write_verify,
+};
+
+
int
xfs_da_node_read(
struct xfs_trans *tp,
@@ -155,7 +157,7 @@ xfs_da_node_read(
int which_fork)
{
return xfs_da_read_buf(tp, dp, bno, mappedbno, bpp,
- which_fork, xfs_da_node_read_verify);
+ which_fork, &xfs_da_node_buf_ops);
}
/*========================================================================
@@ -192,7 +194,7 @@ xfs_da_node_create(xfs_da_args_t *args, xfs_dablk_t blkno, int level,
xfs_trans_log_buf(tp, bp,
XFS_DA_LOGRANGE(node, &node->hdr, sizeof(node->hdr)));
- bp->b_pre_io = xfs_da_node_write_verify;
+ bp->b_ops = &xfs_da_node_buf_ops;
*bpp = bp;
return(0);
}
@@ -393,7 +395,7 @@ xfs_da_root_split(xfs_da_state_t *state, xfs_da_state_blk_t *blk1,
memcpy(node, oldroot, size);
xfs_trans_log_buf(tp, bp, 0, size - 1);
- bp->b_pre_io = blk1->bp->b_pre_io;
+ bp->b_ops = blk1->bp->b_ops;
blk1->bp = bp;
blk1->blkno = blkno;
@@ -2209,7 +2211,7 @@ xfs_da_read_buf(
xfs_daddr_t mappedbno,
struct xfs_buf **bpp,
int whichfork,
- xfs_buf_iodone_t verifier)
+ const struct xfs_buf_ops *ops)
{
struct xfs_buf *bp;
struct xfs_buf_map map;
@@ -2231,7 +2233,7 @@ xfs_da_read_buf(
error = xfs_trans_read_buf_map(dp->i_mount, trans,
dp->i_mount->m_ddev_targp,
- mapp, nmap, 0, &bp, verifier);
+ mapp, nmap, 0, &bp, ops);
if (error)
goto out_free;
@@ -2289,7 +2291,7 @@ xfs_da_reada_buf(
xfs_dablk_t bno,
xfs_daddr_t mappedbno,
int whichfork,
- xfs_buf_iodone_t verifier)
+ const struct xfs_buf_ops *ops)
{
struct xfs_buf_map map;
struct xfs_buf_map *mapp;
@@ -2308,7 +2310,7 @@ xfs_da_reada_buf(
}
mappedbno = mapp[0].bm_bn;
- xfs_buf_readahead_map(dp->i_mount->m_ddev_targp, mapp, nmap, NULL);
+ xfs_buf_readahead_map(dp->i_mount->m_ddev_targp, mapp, nmap, ops);
out_free:
if (mapp != &map)
diff --git a/fs/xfs/xfs_da_btree.h b/fs/xfs/xfs_da_btree.h
index 521b008..ee5170c 100644
--- a/fs/xfs/xfs_da_btree.h
+++ b/fs/xfs/xfs_da_btree.h
@@ -229,10 +229,10 @@ int xfs_da_get_buf(struct xfs_trans *trans, struct xfs_inode *dp,
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,
- xfs_buf_iodone_t verifier);
+ const struct xfs_buf_ops *ops);
xfs_daddr_t xfs_da_reada_buf(struct xfs_trans *trans, struct xfs_inode *dp,
xfs_dablk_t bno, xfs_daddr_t mapped_bno,
- int whichfork, xfs_buf_iodone_t verifier);
+ int whichfork, const struct xfs_buf_ops *ops);
int xfs_da_shrink_inode(xfs_da_args_t *args, xfs_dablk_t dead_blkno,
struct xfs_buf *dead_buf);
diff --git a/fs/xfs/xfs_dir2_block.c b/fs/xfs/xfs_dir2_block.c
index e2fdc6f..7536faa 100644
--- a/fs/xfs/xfs_dir2_block.c
+++ b/fs/xfs/xfs_dir2_block.c
@@ -74,22 +74,24 @@ xfs_dir2_block_verify(
}
static void
-xfs_dir2_block_write_verify(
+xfs_dir2_block_read_verify(
struct xfs_buf *bp)
{
xfs_dir2_block_verify(bp);
}
-void
-xfs_dir2_block_read_verify(
+static void
+xfs_dir2_block_write_verify(
struct xfs_buf *bp)
{
xfs_dir2_block_verify(bp);
- bp->b_pre_io = xfs_dir2_block_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+const struct xfs_buf_ops xfs_dir2_block_buf_ops = {
+ .verify_read = xfs_dir2_block_read_verify,
+ .verify_write = xfs_dir2_block_write_verify,
+};
+
static int
xfs_dir2_block_read(
struct xfs_trans *tp,
@@ -99,7 +101,7 @@ xfs_dir2_block_read(
struct xfs_mount *mp = dp->i_mount;
return xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, bpp,
- XFS_DATA_FORK, xfs_dir2_block_read_verify);
+ XFS_DATA_FORK, &xfs_dir2_block_buf_ops);
}
static void
@@ -1010,7 +1012,7 @@ xfs_dir2_leaf_to_block(
/*
* Start converting it to block form.
*/
- dbp->b_pre_io = xfs_dir2_block_write_verify;
+ dbp->b_ops = &xfs_dir2_block_buf_ops;
hdr->magic = cpu_to_be32(XFS_DIR2_BLOCK_MAGIC);
needlog = 1;
needscan = 0;
@@ -1140,7 +1142,7 @@ xfs_dir2_sf_to_block(
kmem_free(sfp);
return error;
}
- bp->b_pre_io = xfs_dir2_block_write_verify;
+ bp->b_ops = &xfs_dir2_block_buf_ops;
hdr = bp->b_addr;
hdr->magic = cpu_to_be32(XFS_DIR2_BLOCK_MAGIC);
/*
diff --git a/fs/xfs/xfs_dir2_data.c b/fs/xfs/xfs_dir2_data.c
index c759c7b..5ea838b 100644
--- a/fs/xfs/xfs_dir2_data.c
+++ b/fs/xfs/xfs_dir2_data.c
@@ -202,23 +202,57 @@ xfs_dir2_data_verify(
}
}
-void
-xfs_dir2_data_write_verify(
+/*
+ * Readahead of the first block of the directory when it is opened is completely
+ * oblivious to the format of the directory. Hence we can either get a block
+ * format buffer or a data format buffer on readahead.
+ */
+static void
+xfs_dir2_data_reada_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_dir2_data_hdr *hdr = bp->b_addr;
+
+ switch (hdr->magic) {
+ case cpu_to_be32(XFS_DIR2_BLOCK_MAGIC):
+ bp->b_ops = &xfs_dir2_block_buf_ops;
+ bp->b_ops->verify_read(bp);
+ return;
+ case cpu_to_be32(XFS_DIR2_DATA_MAGIC):
+ xfs_dir2_data_verify(bp);
+ return;
+ default:
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ break;
+ }
+}
+
+static void
+xfs_dir2_data_read_verify(
struct xfs_buf *bp)
{
xfs_dir2_data_verify(bp);
}
static void
-xfs_dir2_data_read_verify(
+xfs_dir2_data_write_verify(
struct xfs_buf *bp)
{
xfs_dir2_data_verify(bp);
- bp->b_pre_io = xfs_dir2_data_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+const struct xfs_buf_ops xfs_dir2_data_buf_ops = {
+ .verify_read = xfs_dir2_data_read_verify,
+ .verify_write = xfs_dir2_data_write_verify,
+};
+
+static const struct xfs_buf_ops xfs_dir2_data_reada_buf_ops = {
+ .verify_read = xfs_dir2_data_reada_verify,
+ .verify_write = xfs_dir2_data_write_verify,
+};
+
int
xfs_dir2_data_read(
@@ -229,7 +263,7 @@ xfs_dir2_data_read(
struct xfs_buf **bpp)
{
return xfs_da_read_buf(tp, dp, bno, mapped_bno, bpp,
- XFS_DATA_FORK, xfs_dir2_data_read_verify);
+ XFS_DATA_FORK, &xfs_dir2_data_buf_ops);
}
int
@@ -240,7 +274,7 @@ xfs_dir2_data_readahead(
xfs_daddr_t mapped_bno)
{
return xfs_da_reada_buf(tp, dp, bno, mapped_bno,
- XFS_DATA_FORK, xfs_dir2_data_read_verify);
+ XFS_DATA_FORK, &xfs_dir2_data_reada_buf_ops);
}
/*
@@ -484,7 +518,7 @@ xfs_dir2_data_init(
XFS_DATA_FORK);
if (error)
return error;
- bp->b_pre_io = xfs_dir2_data_write_verify;
+ bp->b_ops = &xfs_dir2_data_buf_ops;
/*
* Initialize the header.
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 3002ab7..60cd2fa 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -65,39 +65,43 @@ xfs_dir2_leaf_verify(
}
static void
-xfs_dir2_leaf1_write_verify(
+xfs_dir2_leaf1_read_verify(
struct xfs_buf *bp)
{
xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
}
static void
-xfs_dir2_leaf1_read_verify(
+xfs_dir2_leaf1_write_verify(
struct xfs_buf *bp)
{
xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
- bp->b_pre_io = xfs_dir2_leaf1_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
void
-xfs_dir2_leafn_write_verify(
+xfs_dir2_leafn_read_verify(
struct xfs_buf *bp)
{
xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
}
void
-xfs_dir2_leafn_read_verify(
+xfs_dir2_leafn_write_verify(
struct xfs_buf *bp)
{
xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
- bp->b_pre_io = xfs_dir2_leafn_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+static const struct xfs_buf_ops xfs_dir2_leaf1_buf_ops = {
+ .verify_read = xfs_dir2_leaf1_read_verify,
+ .verify_write = xfs_dir2_leaf1_write_verify,
+};
+
+const struct xfs_buf_ops xfs_dir2_leafn_buf_ops = {
+ .verify_read = xfs_dir2_leafn_read_verify,
+ .verify_write = xfs_dir2_leafn_write_verify,
+};
+
static int
xfs_dir2_leaf_read(
struct xfs_trans *tp,
@@ -107,7 +111,7 @@ xfs_dir2_leaf_read(
struct xfs_buf **bpp)
{
return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
- XFS_DATA_FORK, xfs_dir2_leaf1_read_verify);
+ XFS_DATA_FORK, &xfs_dir2_leaf1_buf_ops);
}
int
@@ -119,7 +123,7 @@ xfs_dir2_leafn_read(
struct xfs_buf **bpp)
{
return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
- XFS_DATA_FORK, xfs_dir2_leafn_read_verify);
+ XFS_DATA_FORK, &xfs_dir2_leafn_buf_ops);
}
/*
@@ -198,7 +202,7 @@ xfs_dir2_block_to_leaf(
/*
* Fix up the block header, make it a data block.
*/
- dbp->b_pre_io = xfs_dir2_data_write_verify;
+ dbp->b_ops = &xfs_dir2_data_buf_ops;
hdr->magic = cpu_to_be32(XFS_DIR2_DATA_MAGIC);
if (needscan)
xfs_dir2_data_freescan(mp, hdr, &needlog);
@@ -1264,12 +1268,12 @@ xfs_dir2_leaf_init(
* the block.
*/
if (magic == XFS_DIR2_LEAF1_MAGIC) {
- bp->b_pre_io = xfs_dir2_leaf1_write_verify;
+ bp->b_ops = &xfs_dir2_leaf1_buf_ops;
ltp = xfs_dir2_leaf_tail_p(mp, leaf);
ltp->bestcount = 0;
xfs_dir2_leaf_log_tail(tp, bp);
} else
- bp->b_pre_io = xfs_dir2_leafn_write_verify;
+ bp->b_ops = &xfs_dir2_leafn_buf_ops;
*bpp = bp;
return 0;
}
@@ -1954,7 +1958,7 @@ xfs_dir2_node_to_leaf(
else
xfs_dir2_leaf_log_header(tp, lbp);
- lbp->b_pre_io = xfs_dir2_leaf1_write_verify;
+ lbp->b_ops = &xfs_dir2_leaf1_buf_ops;
leaf->hdr.info.magic = cpu_to_be16(XFS_DIR2_LEAF1_MAGIC);
/*
diff --git a/fs/xfs/xfs_dir2_node.c b/fs/xfs/xfs_dir2_node.c
index da90a91..5980f9b 100644
--- a/fs/xfs/xfs_dir2_node.c
+++ b/fs/xfs/xfs_dir2_node.c
@@ -72,22 +72,24 @@ xfs_dir2_free_verify(
}
static void
-xfs_dir2_free_write_verify(
+xfs_dir2_free_read_verify(
struct xfs_buf *bp)
{
xfs_dir2_free_verify(bp);
}
-void
-xfs_dir2_free_read_verify(
+static void
+xfs_dir2_free_write_verify(
struct xfs_buf *bp)
{
xfs_dir2_free_verify(bp);
- bp->b_pre_io = xfs_dir2_free_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+static const struct xfs_buf_ops xfs_dir2_free_buf_ops = {
+ .verify_read = xfs_dir2_free_read_verify,
+ .verify_write = xfs_dir2_free_write_verify,
+};
+
static int
__xfs_dir2_free_read(
@@ -98,7 +100,7 @@ __xfs_dir2_free_read(
struct xfs_buf **bpp)
{
return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
- XFS_DATA_FORK, xfs_dir2_free_read_verify);
+ XFS_DATA_FORK, &xfs_dir2_free_buf_ops);
}
int
@@ -201,7 +203,7 @@ xfs_dir2_leaf_to_node(
XFS_DATA_FORK);
if (error)
return error;
- fbp->b_pre_io = xfs_dir2_free_write_verify;
+ fbp->b_ops = &xfs_dir2_free_buf_ops;
free = fbp->b_addr;
leaf = lbp->b_addr;
@@ -225,7 +227,7 @@ xfs_dir2_leaf_to_node(
}
free->hdr.nused = cpu_to_be32(n);
- lbp->b_pre_io = xfs_dir2_leafn_write_verify;
+ lbp->b_ops = &xfs_dir2_leafn_buf_ops;
leaf->hdr.info.magic = cpu_to_be16(XFS_DIR2_LEAFN_MAGIC);
/*
@@ -636,7 +638,7 @@ xfs_dir2_leafn_lookup_for_entry(
state->extrablk.index = (int)((char *)dep -
(char *)curbp->b_addr);
state->extrablk.magic = XFS_DIR2_DATA_MAGIC;
- curbp->b_pre_io = xfs_dir2_data_write_verify;
+ curbp->b_ops = &xfs_dir2_data_buf_ops;
if (cmp == XFS_CMP_EXACT)
return XFS_ERROR(EEXIST);
}
@@ -651,7 +653,7 @@ xfs_dir2_leafn_lookup_for_entry(
state->extrablk.index = -1;
state->extrablk.blkno = curdb;
state->extrablk.magic = XFS_DIR2_DATA_MAGIC;
- curbp->b_pre_io = xfs_dir2_data_write_verify;
+ curbp->b_ops = &xfs_dir2_data_buf_ops;
} else {
/* If the curbp is not the CI match block, drop it */
if (state->extrablk.bp != curbp)
@@ -1649,7 +1651,7 @@ xfs_dir2_node_addname_int(
-1, &fbp, XFS_DATA_FORK);
if (error)
return error;
- fbp->b_pre_io = xfs_dir2_free_write_verify;
+ fbp->b_ops = &xfs_dir2_free_buf_ops;
/*
* Initialize the new block to be empty, and remember
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index ac49e12..b9a033b 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -30,6 +30,8 @@ extern int xfs_dir_cilookup_result(struct xfs_da_args *args,
const unsigned char *name, int len);
/* xfs_dir2_block.c */
+extern const struct xfs_buf_ops xfs_dir2_block_buf_ops;
+
extern int xfs_dir2_block_addname(struct xfs_da_args *args);
extern int xfs_dir2_block_getdents(struct xfs_inode *dp, void *dirent,
xfs_off_t *offset, filldir_t filldir);
@@ -45,7 +47,9 @@ extern int xfs_dir2_leaf_to_block(struct xfs_da_args *args,
#else
#define xfs_dir2_data_check(dp,bp)
#endif
-extern void xfs_dir2_data_write_verify(struct xfs_buf *bp);
+
+extern const struct xfs_buf_ops xfs_dir2_data_buf_ops;
+
extern bool __xfs_dir2_data_check(struct xfs_inode *dp, struct xfs_buf *bp);
extern int xfs_dir2_data_read(struct xfs_trans *tp, struct xfs_inode *dp,
xfs_dablk_t bno, xfs_daddr_t mapped_bno, struct xfs_buf **bpp);
@@ -73,8 +77,8 @@ extern void xfs_dir2_data_use_free(struct xfs_trans *tp, struct xfs_buf *bp,
xfs_dir2_data_aoff_t len, int *needlogp, int *needscanp);
/* xfs_dir2_leaf.c */
-extern void xfs_dir2_leafn_read_verify(struct xfs_buf *bp);
-extern void xfs_dir2_leafn_write_verify(struct xfs_buf *bp);
+extern const struct xfs_buf_ops xfs_dir2_leafn_buf_ops;
+
extern int xfs_dir2_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/xfs_dquot.c b/fs/xfs/xfs_dquot.c
index d6d4d6b..14d4088 100644
--- a/fs/xfs/xfs_dquot.c
+++ b/fs/xfs/xfs_dquot.c
@@ -285,22 +285,24 @@ xfs_dquot_buf_verify(
}
static void
-xfs_dquot_buf_write_verify(
+xfs_dquot_buf_read_verify(
struct xfs_buf *bp)
{
xfs_dquot_buf_verify(bp);
}
static void
-xfs_dquot_buf_read_verify(
+xfs_dquot_buf_write_verify(
struct xfs_buf *bp)
{
xfs_dquot_buf_verify(bp);
- bp->b_pre_io = xfs_dquot_buf_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+static const struct xfs_buf_ops xfs_dquot_buf_ops = {
+ .verify_read = xfs_dquot_buf_read_verify,
+ .verify_write = xfs_dquot_buf_write_verify,
+};
+
/*
* Allocate a block and fill it with dquots.
* This is called when the bmapi finds a hole.
@@ -366,7 +368,7 @@ xfs_qm_dqalloc(
error = xfs_buf_geterror(bp);
if (error)
goto error1;
- bp->b_pre_io = xfs_dquot_buf_write_verify;
+ bp->b_ops = &xfs_dquot_buf_ops;
/*
* Make a chunk of dquots out of this buffer and log
@@ -535,7 +537,7 @@ xfs_qm_dqtobp(
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
dqp->q_blkno,
mp->m_quotainfo->qi_dqchunklen,
- 0, &bp, xfs_dquot_buf_read_verify);
+ 0, &bp, &xfs_dquot_buf_ops);
if (error == EFSCORRUPTED && (flags & XFS_QMOPT_DQREPAIR)) {
xfs_dqid_t firstid = (xfs_dqid_t)map.br_startoff *
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index 8ebc457..f477723 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -200,7 +200,7 @@ xfs_growfs_data_private(
error = ENOMEM;
goto error0;
}
- bp->b_pre_io = xfs_agf_write_verify;
+ bp->b_ops = &xfs_agf_buf_ops;
agf = XFS_BUF_TO_AGF(bp);
memset(agf, 0, mp->m_sb.sb_sectsize);
agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
@@ -238,7 +238,7 @@ xfs_growfs_data_private(
error = ENOMEM;
goto error0;
}
- bp->b_pre_io = xfs_agi_write_verify;
+ bp->b_ops = &xfs_agi_buf_ops;
agi = XFS_BUF_TO_AGI(bp);
memset(agi, 0, mp->m_sb.sb_sectsize);
agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
@@ -260,6 +260,11 @@ xfs_growfs_data_private(
/*
* BNO btree root block
+ *
+ * XXX: we attach the buf ops after writing the buffer becaus
+ * the perag is not yet initialised fully and hence the buffer
+ * will fail write verification. Attach it after writing. This
+ * needs fixing before CRC protection will work.
*/
bp = xfs_buf_get(mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
@@ -280,12 +285,18 @@ xfs_growfs_data_private(
arec->ar_blockcount = cpu_to_be32(
agsize - be32_to_cpu(arec->ar_startblock));
error = xfs_bwrite(bp);
+ bp->b_ops = &xfs_allocbt_buf_ops;
xfs_buf_relse(bp);
if (error)
goto error0;
/*
* CNT btree root block
+ *
+ * XXX: we attach the buf ops after writing the buffer becaus
+ * the perag is not yet initialised fully and hence the buffer
+ * will fail write verification. Attach it after writing. This
+ * needs fixing before CRC protection will work.
*/
bp = xfs_buf_get(mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
@@ -307,12 +318,18 @@ xfs_growfs_data_private(
agsize - be32_to_cpu(arec->ar_startblock));
nfree += be32_to_cpu(arec->ar_blockcount);
error = xfs_bwrite(bp);
+ bp->b_ops = &xfs_allocbt_buf_ops;
xfs_buf_relse(bp);
if (error)
goto error0;
/*
* INO btree root block
+ *
+ * XXX: we attach the buf ops after writing the buffer becaus
+ * the perag is not yet initialised fully and hence the buffer
+ * will fail write verification. Attach it after writing. This
+ * needs fixing before CRC protection will work.
*/
bp = xfs_buf_get(mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
@@ -329,6 +346,7 @@ xfs_growfs_data_private(
block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
error = xfs_bwrite(bp);
+ bp->b_ops = &xfs_inobt_buf_ops;
xfs_buf_relse(bp);
if (error)
goto error0;
@@ -416,14 +434,14 @@ xfs_growfs_data_private(
error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
XFS_FSS_TO_BB(mp, 1), 0, &bp,
- xfs_sb_read_verify);
+ &xfs_sb_buf_ops);
} else {
bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
XFS_FSS_TO_BB(mp, 1), 0);
if (bp) {
+ bp->b_ops = &xfs_sb_buf_ops;
xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
- bp->b_pre_io = xfs_sb_write_verify;
} else
error = ENOMEM;
}
diff --git a/fs/xfs/xfs_ialloc.c b/fs/xfs/xfs_ialloc.c
index 12d1c94..878efd7 100644
--- a/fs/xfs/xfs_ialloc.c
+++ b/fs/xfs/xfs_ialloc.c
@@ -210,7 +210,7 @@ xfs_ialloc_inode_init(
* to log a whole cluster of inodes instead of all the
* individual transactions causing a lot of log traffic.
*/
- fbuf->b_pre_io = xfs_inode_buf_write_verify;
+ fbuf->b_ops = &xfs_inode_buf_ops;
xfs_buf_zero(fbuf, 0, ninodes << mp->m_sb.sb_inodelog);
for (i = 0; i < ninodes; i++) {
int ioffset = i << mp->m_sb.sb_inodelog;
@@ -1496,23 +1496,25 @@ xfs_agi_verify(
xfs_check_agi_unlinked(agi);
}
-void
-xfs_agi_write_verify(
+static void
+xfs_agi_read_verify(
struct xfs_buf *bp)
{
xfs_agi_verify(bp);
}
static void
-xfs_agi_read_verify(
+xfs_agi_write_verify(
struct xfs_buf *bp)
{
xfs_agi_verify(bp);
- bp->b_pre_io = xfs_agi_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+const struct xfs_buf_ops xfs_agi_buf_ops = {
+ .verify_read = xfs_agi_read_verify,
+ .verify_write = xfs_agi_write_verify,
+};
+
/*
* Read in the allocation group header (inode allocation section)
*/
@@ -1529,7 +1531,7 @@ xfs_read_agi(
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
- XFS_FSS_TO_BB(mp, 1), 0, bpp, xfs_agi_read_verify);
+ XFS_FSS_TO_BB(mp, 1), 0, bpp, &xfs_agi_buf_ops);
if (error)
return error;
diff --git a/fs/xfs/xfs_ialloc.h b/fs/xfs/xfs_ialloc.h
index 7a169e3..c8da3df 100644
--- a/fs/xfs/xfs_ialloc.h
+++ b/fs/xfs/xfs_ialloc.h
@@ -150,6 +150,6 @@ int xfs_inobt_lookup(struct xfs_btree_cur *cur, xfs_agino_t ino,
int xfs_inobt_get_rec(struct xfs_btree_cur *cur,
xfs_inobt_rec_incore_t *rec, int *stat);
-void xfs_agi_write_verify(struct xfs_buf *bp);
+extern const struct xfs_buf_ops xfs_agi_buf_ops;
#endif /* __XFS_IALLOC_H__ */
diff --git a/fs/xfs/xfs_ialloc_btree.c b/fs/xfs/xfs_ialloc_btree.c
index 7761e1e..bec344b 100644
--- a/fs/xfs/xfs_ialloc_btree.c
+++ b/fs/xfs/xfs_ialloc_btree.c
@@ -217,22 +217,24 @@ xfs_inobt_verify(
}
static void
-xfs_inobt_write_verify(
+xfs_inobt_read_verify(
struct xfs_buf *bp)
{
xfs_inobt_verify(bp);
}
-void
-xfs_inobt_read_verify(
+static void
+xfs_inobt_write_verify(
struct xfs_buf *bp)
{
xfs_inobt_verify(bp);
- bp->b_pre_io = xfs_inobt_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+const struct xfs_buf_ops xfs_inobt_buf_ops = {
+ .verify_read = xfs_inobt_read_verify,
+ .verify_write = xfs_inobt_write_verify,
+};
+
#ifdef DEBUG
STATIC int
xfs_inobt_keys_inorder(
@@ -270,8 +272,7 @@ static const struct xfs_btree_ops xfs_inobt_ops = {
.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,
- .read_verify = xfs_inobt_read_verify,
- .write_verify = xfs_inobt_write_verify,
+ .buf_ops = &xfs_inobt_buf_ops,
#ifdef DEBUG
.keys_inorder = xfs_inobt_keys_inorder,
.recs_inorder = xfs_inobt_recs_inorder,
diff --git a/fs/xfs/xfs_ialloc_btree.h b/fs/xfs/xfs_ialloc_btree.h
index f782ad0..25c0239 100644
--- a/fs/xfs/xfs_ialloc_btree.h
+++ b/fs/xfs/xfs_ialloc_btree.h
@@ -109,4 +109,6 @@ extern struct xfs_btree_cur *xfs_inobt_init_cursor(struct xfs_mount *,
struct xfs_trans *, struct xfs_buf *, xfs_agnumber_t);
extern int xfs_inobt_maxrecs(struct xfs_mount *, int, int);
+extern const struct xfs_buf_ops xfs_inobt_buf_ops;
+
#endif /* __XFS_IALLOC_BTREE_H__ */
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 4f2e99a..c4add46 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -420,23 +420,27 @@ xfs_inode_buf_verify(
xfs_inobp_check(mp, bp);
}
-void
-xfs_inode_buf_write_verify(
+
+static void
+xfs_inode_buf_read_verify(
struct xfs_buf *bp)
{
xfs_inode_buf_verify(bp);
}
-void
-xfs_inode_buf_read_verify(
+static void
+xfs_inode_buf_write_verify(
struct xfs_buf *bp)
{
xfs_inode_buf_verify(bp);
- bp->b_pre_io = xfs_inode_buf_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
+const struct xfs_buf_ops xfs_inode_buf_ops = {
+ .verify_read = xfs_inode_buf_read_verify,
+ .verify_write = xfs_inode_buf_write_verify,
+};
+
+
/*
* This routine is called to map an inode to the buffer containing the on-disk
* version of the inode. It returns a pointer to the buffer containing the
@@ -462,7 +466,7 @@ xfs_imap_to_bp(
buf_flags |= XBF_UNMAPPED;
error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
(int)imap->im_len, buf_flags, &bp,
- xfs_inode_buf_read_verify);
+ &xfs_inode_buf_ops);
if (error) {
if (error == EAGAIN) {
ASSERT(buf_flags & XBF_TRYLOCK);
@@ -1791,7 +1795,7 @@ xfs_ifree_cluster(
* want it to fail. We can acheive this by adding a write
* verifier to the buffer.
*/
- bp->b_pre_io = xfs_inode_buf_write_verify;
+ bp->b_ops = &xfs_inode_buf_ops;
/*
* Walk the inodes already attached to the buffer and mark them
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index a866d28..0a5d3c0 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -554,8 +554,6 @@ int xfs_imap_to_bp(struct xfs_mount *, struct xfs_trans *,
struct xfs_buf **, uint, uint);
int xfs_iread(struct xfs_mount *, struct xfs_trans *,
struct xfs_inode *, uint);
-void xfs_inode_buf_read_verify(struct xfs_buf *);
-void xfs_inode_buf_write_verify(struct xfs_buf *);
void xfs_dinode_to_disk(struct xfs_dinode *,
struct xfs_icdinode *);
void xfs_idestroy_fork(struct xfs_inode *, int);
@@ -599,5 +597,6 @@ void xfs_inobp_check(struct xfs_mount *, struct xfs_buf *);
extern struct kmem_zone *xfs_ifork_zone;
extern struct kmem_zone *xfs_inode_zone;
extern struct kmem_zone *xfs_ili_zone;
+extern const struct xfs_buf_ops xfs_inode_buf_ops;
#endif /* __XFS_INODE_H__ */
diff --git a/fs/xfs/xfs_itable.c b/fs/xfs/xfs_itable.c
index 7f86fda..2ea7d40 100644
--- a/fs/xfs/xfs_itable.c
+++ b/fs/xfs/xfs_itable.c
@@ -397,7 +397,7 @@ xfs_bulkstat(
& ~r.ir_free)
xfs_btree_reada_bufs(mp, agno,
agbno, nbcluster,
- xfs_inode_buf_read_verify);
+ &xfs_inode_buf_ops);
}
irbp->ir_startino = r.ir_startino;
irbp->ir_freecount = r.ir_freecount;
diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
index 4cf7ae8..d63d0ca 100644
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -3699,7 +3699,7 @@ xlog_do_recover(
ASSERT(!(XFS_BUF_ISWRITE(bp)));
XFS_BUF_READ(bp);
XFS_BUF_UNASYNC(bp);
- bp->b_iodone = xfs_sb_read_verify;
+ bp->b_ops = &xfs_sb_buf_ops;
xfsbdstrat(log->l_mp, bp);
error = xfs_buf_iowait(bp);
if (error) {
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index 18b853e..247c93d 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -631,21 +631,11 @@ xfs_sb_verify(
xfs_buf_ioerror(bp, error);
}
-void
-xfs_sb_write_verify(
- struct xfs_buf *bp)
-{
- xfs_sb_verify(bp);
-}
-
-void
+static void
xfs_sb_read_verify(
struct xfs_buf *bp)
{
xfs_sb_verify(bp);
- bp->b_pre_io = xfs_sb_write_verify;
- bp->b_iodone = NULL;
- xfs_buf_ioend(bp, 0);
}
/*
@@ -654,7 +644,7 @@ xfs_sb_read_verify(
* If we find an XFS superblock, the run a normal, noisy mount because we are
* really going to mount it and want to know about errors.
*/
-void
+static void
xfs_sb_quiet_read_verify(
struct xfs_buf *bp)
{
@@ -671,6 +661,23 @@ xfs_sb_quiet_read_verify(
xfs_buf_ioerror(bp, EFSCORRUPTED);
}
+static void
+xfs_sb_write_verify(
+ struct xfs_buf *bp)
+{
+ xfs_sb_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_sb_buf_ops = {
+ .verify_read = xfs_sb_read_verify,
+ .verify_write = xfs_sb_write_verify,
+};
+
+static const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
+ .verify_read = xfs_sb_quiet_read_verify,
+ .verify_write = xfs_sb_write_verify,
+};
+
/*
* xfs_readsb
*
@@ -697,8 +704,8 @@ xfs_readsb(xfs_mount_t *mp, int flags)
reread:
bp = xfs_buf_read_uncached(mp->m_ddev_targp, XFS_SB_DADDR,
BTOBB(sector_size), 0,
- loud ? xfs_sb_read_verify
- : xfs_sb_quiet_read_verify);
+ loud ? &xfs_sb_buf_ops
+ : &xfs_sb_quiet_buf_ops);
if (!bp) {
if (loud)
xfs_warn(mp, "SB buffer read failed");
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 69e4082..e4d1510 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -382,12 +382,12 @@ extern void xfs_set_low_space_thresholds(struct xfs_mount *);
#endif /* __KERNEL__ */
-extern void xfs_sb_read_verify(struct xfs_buf *);
-extern void xfs_sb_write_verify(struct xfs_buf *bp);
extern void xfs_mod_sb(struct xfs_trans *, __int64_t);
extern int xfs_initialize_perag(struct xfs_mount *, xfs_agnumber_t,
xfs_agnumber_t *);
extern void xfs_sb_from_disk(struct xfs_sb *, struct xfs_dsb *);
extern void xfs_sb_to_disk(struct xfs_dsb *, struct xfs_sb *, __int64_t);
+extern const struct xfs_buf_ops xfs_sb_buf_ops;
+
#endif /* __XFS_MOUNT_H__ */
diff --git a/fs/xfs/xfs_trans.h b/fs/xfs/xfs_trans.h
index f02d402..c6c0601 100644
--- a/fs/xfs/xfs_trans.h
+++ b/fs/xfs/xfs_trans.h
@@ -474,7 +474,7 @@ int xfs_trans_read_buf_map(struct xfs_mount *mp,
struct xfs_buf_map *map, int nmaps,
xfs_buf_flags_t flags,
struct xfs_buf **bpp,
- xfs_buf_iodone_t verify);
+ const struct xfs_buf_ops *ops);
static inline int
xfs_trans_read_buf(
@@ -485,11 +485,11 @@ xfs_trans_read_buf(
int numblks,
xfs_buf_flags_t flags,
struct xfs_buf **bpp,
- xfs_buf_iodone_t verify)
+ const struct xfs_buf_ops *ops)
{
DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
return xfs_trans_read_buf_map(mp, tp, target, &map, 1,
- flags, bpp, verify);
+ flags, bpp, ops);
}
struct xfs_buf *xfs_trans_getsb(xfs_trans_t *, struct xfs_mount *, int);
diff --git a/fs/xfs/xfs_trans_buf.c b/fs/xfs/xfs_trans_buf.c
index 9776282..4fc17d4 100644
--- a/fs/xfs/xfs_trans_buf.c
+++ b/fs/xfs/xfs_trans_buf.c
@@ -258,7 +258,7 @@ xfs_trans_read_buf_map(
int nmaps,
xfs_buf_flags_t flags,
struct xfs_buf **bpp,
- xfs_buf_iodone_t verify)
+ const struct xfs_buf_ops *ops)
{
xfs_buf_t *bp;
xfs_buf_log_item_t *bip;
@@ -266,7 +266,7 @@ xfs_trans_read_buf_map(
*bpp = NULL;
if (!tp) {
- bp = xfs_buf_read_map(target, map, nmaps, flags, verify);
+ bp = xfs_buf_read_map(target, map, nmaps, flags, ops);
if (!bp)
return (flags & XBF_TRYLOCK) ?
EAGAIN : XFS_ERROR(ENOMEM);
@@ -315,7 +315,7 @@ xfs_trans_read_buf_map(
ASSERT(!XFS_BUF_ISASYNC(bp));
ASSERT(bp->b_iodone == NULL);
XFS_BUF_READ(bp);
- bp->b_iodone = verify;
+ bp->b_ops = ops;
xfsbdstrat(tp->t_mountp, bp);
error = xfs_buf_iowait(bp);
if (error) {
@@ -352,7 +352,7 @@ xfs_trans_read_buf_map(
return 0;
}
- bp = xfs_buf_read_map(target, map, nmaps, flags, verify);
+ bp = xfs_buf_read_map(target, map, nmaps, flags, ops);
if (bp == NULL) {
*bpp = NULL;
return (flags & XBF_TRYLOCK) ?
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* [PATCH 25/25] xfs: add write verifiers to log recovery
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
` (23 preceding siblings ...)
2012-10-25 6:34 ` [PATCH 24/25] xfs: convert buffer verifiers to an ops structure Dave Chinner
@ 2012-10-25 6:34 ` Dave Chinner
2012-10-26 8:54 ` Christoph Hellwig
2012-10-30 13:44 ` Phil White
24 siblings, 2 replies; 69+ messages in thread
From: Dave Chinner @ 2012-10-25 6:34 UTC (permalink / raw)
To: xfs
From: Dave Chinner <dchinner@redhat.com>
Log recovery reads metadata, modifies it and rewrites it to disk.
It is only practical to add write verifiers to metadata buffers
because we do not know the type of the buffer prior to reading it
from disk. Further, if it is an new bufer, the contents might not
contain anything we can verify. Hence we only attempt to verify
after the buffer changes have been replayed and we can peek at the
buffer to find out what it contains to attached the correct
verifier. This ensures that we don't introduce gross corruptions as
a result of replaying transactions in the log.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_alloc.c | 2 +-
fs/xfs/xfs_alloc.h | 1 +
fs/xfs/xfs_alloc_btree.c | 15 ++++---
fs/xfs/xfs_da_btree.h | 1 +
fs/xfs/xfs_dir2_leaf.c | 2 +-
fs/xfs/xfs_dir2_node.c | 2 +-
fs/xfs/xfs_dir2_priv.h | 3 ++
fs/xfs/xfs_dquot.c | 17 +++++++-
fs/xfs/xfs_dquot.h | 2 +
fs/xfs/xfs_log_recover.c | 104 +++++++++++++++++++++++++++++++++++++++++++++-
10 files changed, 138 insertions(+), 11 deletions(-)
diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
index f9231b2..9e30796 100644
--- a/fs/xfs/xfs_alloc.c
+++ b/fs/xfs/xfs_alloc.c
@@ -479,7 +479,7 @@ xfs_agfl_read_verify(
xfs_agfl_verify(bp);
}
-static const struct xfs_buf_ops xfs_agfl_buf_ops = {
+const struct xfs_buf_ops xfs_agfl_buf_ops = {
.verify_read = xfs_agfl_read_verify,
.verify_write = xfs_agfl_write_verify,
};
diff --git a/fs/xfs/xfs_alloc.h b/fs/xfs/xfs_alloc.h
index aaf7ff1..99d0a61 100644
--- a/fs/xfs/xfs_alloc.h
+++ b/fs/xfs/xfs_alloc.h
@@ -232,5 +232,6 @@ xfs_alloc_get_rec(
int *stat); /* output: success/failure */
extern const struct xfs_buf_ops xfs_agf_buf_ops;
+extern const struct xfs_buf_ops xfs_agfl_buf_ops;
#endif /* __XFS_ALLOC_H__ */
diff --git a/fs/xfs/xfs_alloc_btree.c b/fs/xfs/xfs_alloc_btree.c
index b14ff21..5e12e7b 100644
--- a/fs/xfs/xfs_alloc_btree.c
+++ b/fs/xfs/xfs_alloc_btree.c
@@ -33,6 +33,7 @@
#include "xfs_extent_busy.h"
#include "xfs_error.h"
#include "xfs_trace.h"
+#include "xfs_log_priv.h"
STATIC struct xfs_btree_cur *
@@ -279,17 +280,22 @@ xfs_allocbt_verify(
struct xfs_mount *mp = bp->b_target->bt_mount;
struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
struct xfs_perag *pag = bp->b_pag;
- unsigned int level;
+ unsigned int level = 0;
int sblock_ok; /* block passes checks */
- /* magic number and level verification */
+ /*
+ * magic number and level verification. For recovery, the pag has not
+ * been initialised fully yet, so the pagf_level checks cannot be done.
+ */
level = be16_to_cpu(block->bb_level);
switch (block->bb_magic) {
case cpu_to_be32(XFS_ABTB_MAGIC):
- sblock_ok = level < pag->pagf_levels[XFS_BTNUM_BNOi];
+ sblock_ok = (mp->m_log->l_flags & XLOG_ACTIVE_RECOVERY) ||
+ level < pag->pagf_levels[XFS_BTNUM_BNOi];
break;
case cpu_to_be32(XFS_ABTC_MAGIC):
- sblock_ok = level < pag->pagf_levels[XFS_BTNUM_CNTi];
+ sblock_ok = (mp->m_log->l_flags & XLOG_ACTIVE_RECOVERY) ||
+ level < pag->pagf_levels[XFS_BTNUM_CNTi];
break;
default:
sblock_ok = 0;
@@ -335,7 +341,6 @@ const struct xfs_buf_ops xfs_allocbt_buf_ops = {
.verify_write = xfs_allocbt_write_verify,
};
-
#ifdef DEBUG
STATIC int
xfs_allocbt_keys_inorder(
diff --git a/fs/xfs/xfs_da_btree.h b/fs/xfs/xfs_da_btree.h
index ee5170c..eae66b0 100644
--- a/fs/xfs/xfs_da_btree.h
+++ b/fs/xfs/xfs_da_btree.h
@@ -246,5 +246,6 @@ void xfs_da_state_free(xfs_da_state_t *state);
extern struct kmem_zone *xfs_da_state_zone;
extern const struct xfs_nameops xfs_default_nameops;
+extern const struct xfs_buf_ops xfs_da_node_buf_ops;
#endif /* __XFS_DA_BTREE_H__ */
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 60cd2fa..88a27a1 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -92,7 +92,7 @@ xfs_dir2_leafn_write_verify(
xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
}
-static const struct xfs_buf_ops xfs_dir2_leaf1_buf_ops = {
+const struct xfs_buf_ops xfs_dir2_leaf1_buf_ops = {
.verify_read = xfs_dir2_leaf1_read_verify,
.verify_write = xfs_dir2_leaf1_write_verify,
};
diff --git a/fs/xfs/xfs_dir2_node.c b/fs/xfs/xfs_dir2_node.c
index 5980f9b..90d71d2 100644
--- a/fs/xfs/xfs_dir2_node.c
+++ b/fs/xfs/xfs_dir2_node.c
@@ -85,7 +85,7 @@ xfs_dir2_free_write_verify(
xfs_dir2_free_verify(bp);
}
-static const struct xfs_buf_ops xfs_dir2_free_buf_ops = {
+const struct xfs_buf_ops xfs_dir2_free_buf_ops = {
.verify_read = xfs_dir2_free_read_verify,
.verify_write = xfs_dir2_free_write_verify,
};
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index b9a033b..40ff241 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -77,6 +77,7 @@ extern void xfs_dir2_data_use_free(struct xfs_trans *tp, struct xfs_buf *bp,
xfs_dir2_data_aoff_t len, int *needlogp, int *needscanp);
/* xfs_dir2_leaf.c */
+extern const struct xfs_buf_ops xfs_dir2_leaf1_buf_ops;
extern const struct xfs_buf_ops xfs_dir2_leafn_buf_ops;
extern int xfs_dir2_leafn_read(struct xfs_trans *tp, struct xfs_inode *dp,
@@ -110,6 +111,8 @@ xfs_dir2_leaf_find_entry(struct xfs_dir2_leaf *leaf, int index, int compact,
extern int xfs_dir2_node_to_leaf(struct xfs_da_state *state);
/* xfs_dir2_node.c */
+extern const struct xfs_buf_ops xfs_dir2_free_buf_ops;
+
extern int xfs_dir2_leaf_to_node(struct xfs_da_args *args,
struct xfs_buf *lbp);
extern xfs_dahash_t xfs_dir2_leafn_lasthash(struct xfs_buf *bp, int *count);
diff --git a/fs/xfs/xfs_dquot.c b/fs/xfs/xfs_dquot.c
index 14d4088..0b690a2 100644
--- a/fs/xfs/xfs_dquot.c
+++ b/fs/xfs/xfs_dquot.c
@@ -37,6 +37,7 @@
#include "xfs_trans_priv.h"
#include "xfs_qm.h"
#include "xfs_trace.h"
+#include "xfs_log_priv.h"
/*
* Lock order:
@@ -257,16 +258,28 @@ xfs_dquot_buf_verify(
struct xfs_dqblk *d = (struct xfs_dqblk *)bp->b_addr;
struct xfs_disk_dquot *ddq;
xfs_dqid_t id = 0;
+ int dquots_per_buf;
int i;
/*
+ * during log recovery, we don't have a quotainfo structure to
+ * pull the number of dquots per buffer out of, so we have to calculate
+ * it directly.
+ */
+ if (mp->m_log->l_flags & XLOG_ACTIVE_RECOVERY) {
+ dquots_per_buf = BBTOB(bp->b_length);
+ do_div(dquots_per_buf, sizeof(xfs_dqblk_t));
+ } else
+ dquots_per_buf = mp->m_quotainfo->qi_dqperchunk;
+
+ /*
* On the first read of the buffer, verify that each dquot is valid.
* We don't know what the id of the dquot is supposed to be, just that
* they should be increasing monotonically within the buffer. If the
* first id is corrupt, then it will fail on the second dquot in the
* buffer so corruptions could point to the wrong dquot in this case.
*/
- for (i = 0; i < mp->m_quotainfo->qi_dqperchunk; i++) {
+ for (i = 0; i < dquots_per_buf; i++) {
int error;
ddq = &d[i].dd_diskdq;
@@ -298,7 +311,7 @@ xfs_dquot_buf_write_verify(
xfs_dquot_buf_verify(bp);
}
-static const struct xfs_buf_ops xfs_dquot_buf_ops = {
+const struct xfs_buf_ops xfs_dquot_buf_ops = {
.verify_read = xfs_dquot_buf_read_verify,
.verify_write = xfs_dquot_buf_write_verify,
};
diff --git a/fs/xfs/xfs_dquot.h b/fs/xfs/xfs_dquot.h
index 7d20af2..c694a84 100644
--- a/fs/xfs/xfs_dquot.h
+++ b/fs/xfs/xfs_dquot.h
@@ -161,4 +161,6 @@ static inline struct xfs_dquot *xfs_qm_dqhold(struct xfs_dquot *dqp)
return dqp;
}
+extern const struct xfs_buf_ops xfs_dquot_buf_ops;
+
#endif /* __XFS_DQUOT_H__ */
diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
index d63d0ca..e445550 100644
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -43,6 +43,12 @@
#include "xfs_utils.h"
#include "xfs_trace.h"
#include "xfs_icache.h"
+#include "xfs_da_btree.h"
+#include "xfs_dir2_format.h"
+#include "xfs_dir2_priv.h"
+#include "xfs_attr_leaf.h"
+#include "xfs_dquot_item.h"
+#include "xfs_dquot.h"
STATIC int
xlog_find_zeroed(
@@ -1786,6 +1792,8 @@ xlog_recover_do_inode_buffer(
trace_xfs_log_recover_buf_inode_buf(mp->m_log, buf_f);
+ bp->b_ops = &xfs_inode_buf_ops;
+
inodes_per_buf = BBTOB(bp->b_io_length) >> mp->m_sb.sb_inodelog;
for (i = 0; i < inodes_per_buf; i++) {
next_unlinked_offset = (i * mp->m_sb.sb_inodesize) +
@@ -1856,6 +1864,97 @@ xlog_recover_do_inode_buffer(
return 0;
}
+
+/*
+ * If we don't know what the type of buffer is, work it out now
+ * and attached the appropriate write verifier. This is needed to ensure
+ * recovery hasn't corrupted the contents of the buffer, and to
+ * calculate CRC so that the buffer is correct on disk after recovery.
+ *
+ * There is no easy way to do this except for trying a bunch of magic
+ * number matches....
+ */
+static void
+xlog_buf_attach_ops(
+ struct xfs_buf *bp)
+{
+ struct xfs_da_blkinfo *dablk;
+ struct xfs_mount *mp;
+ xfs_agnumber_t agno;
+ __be32 *magic32;
+
+ /*
+ * dquot buffers are already marked here, and inode buffers never get to
+ * this function, so we can ignore them too.
+ */
+ if (bp->b_ops)
+ return;
+
+ /* try all the buffers that have a magic number in the first 32 bits */
+ magic32 = bp->b_addr;
+ switch (be32_to_cpu(*magic32)) {
+ case XFS_SB_MAGIC:
+ bp->b_ops = &xfs_sb_buf_ops;
+ return;
+ case XFS_AGF_MAGIC:
+ bp->b_ops = &xfs_agf_buf_ops;
+ return;
+ case XFS_AGI_MAGIC:
+ bp->b_ops = &xfs_agi_buf_ops;
+ return;
+ case XFS_ABTB_MAGIC:
+ case XFS_ABTC_MAGIC:
+ bp->b_ops = &xfs_allocbt_buf_ops;
+ return;
+ case XFS_BMAP_MAGIC:
+ bp->b_ops = &xfs_bmbt_buf_ops;
+ return;
+ case XFS_IBT_MAGIC:
+ bp->b_ops = &xfs_inobt_buf_ops;
+ return;
+ case XFS_DIR2_BLOCK_MAGIC:
+ bp->b_ops = &xfs_dir2_block_buf_ops;
+ return;
+ case XFS_DIR2_DATA_MAGIC:
+ bp->b_ops = &xfs_dir2_data_buf_ops;
+ return;
+ case XFS_DIR2_FREE_MAGIC:
+ bp->b_ops = &xfs_dir2_free_buf_ops;
+ return;
+ default:
+ break;
+ }
+
+ /* Now check for dablk types with 16 bit magic numbers */
+ dablk = bp->b_addr;
+ switch (be16_to_cpu(dablk->magic)) {
+ case XFS_DA_NODE_MAGIC:
+ bp->b_ops = &xfs_da_node_buf_ops;
+ return;
+ case XFS_ATTR_LEAF_MAGIC:
+ bp->b_ops = &xfs_attr_leaf_buf_ops;
+ return;
+ case XFS_DIR2_LEAF1_MAGIC:
+ bp->b_ops = &xfs_dir2_leaf1_buf_ops;
+ return;
+ case XFS_DIR2_LEAFN_MAGIC:
+ bp->b_ops = &xfs_dir2_leafn_buf_ops;
+ return;
+ default:
+ break;
+ }
+
+ /*
+ * AGFL has no magic number. Detect by finding the AG daddr of the
+ * buffer and matching it to the XFS_AGFL_DADDR.
+ */
+ mp = bp->b_target->bt_mount;
+ agno = xfs_daddr_to_agno(mp, bp->b_bn);
+ if (bp->b_bn == XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)))
+ bp->b_ops = &xfs_agfl_buf_ops;
+
+}
+
/*
* Perform a 'normal' buffer recovery. Each logged region of the
* buffer should be copied over the corresponding region in the
@@ -1928,6 +2027,8 @@ xlog_recover_do_reg_buffer(
/* Shouldn't be any more regions */
ASSERT(i == item->ri_total);
+
+ xlog_buf_attach_ops(bp);
}
/*
@@ -2089,6 +2190,7 @@ xlog_recover_do_dquot_buffer(
if (log->l_quotaoffs_flag & type)
return;
+ bp->b_ops = &xfs_dquot_buf_ops;
xlog_recover_do_reg_buffer(mp, item, bp, buf_f);
}
@@ -2238,7 +2340,7 @@ xlog_recover_inode_pass2(
trace_xfs_log_recover_inode_recover(log, in_f);
bp = xfs_buf_read(mp->m_ddev_targp, in_f->ilf_blkno, in_f->ilf_len, 0,
- NULL);
+ &xfs_inode_buf_ops);
if (!bp) {
error = ENOMEM;
goto error;
--
1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* Re: [PATCH 02/25] xfs: invalidate allocbt blocks moved to the free list
2012-10-25 6:33 ` [PATCH 02/25] xfs: invalidate allocbt blocks moved to the free list Dave Chinner
@ 2012-10-26 8:47 ` Christoph Hellwig
2012-10-30 0:22 ` Phil White
1 sibling, 0 replies; 69+ messages in thread
From: Christoph Hellwig @ 2012-10-26 8:47 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
> index f1647ca..f7876c6 100644
> --- a/fs/xfs/xfs_alloc_btree.c
> +++ b/fs/xfs/xfs_alloc_btree.c
> @@ -121,6 +121,8 @@ xfs_allocbt_free_block(
> xfs_extent_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1,
> XFS_EXTENT_BUSY_SKIP_DISCARD);
> xfs_trans_agbtree_delta(cur->bc_tp, -1);
> +
> + xfs_trans_binval(cur->bc_tp, bp);
> return 0;
I'd be almost tempted to move this into the caller, as the bmap and
ialloc btrees do it and I always wondered how the alloc btree got away
without it.
Otherwise looks ok,
Reviewed-by: Christoph Hellwig <hch@lst.de>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 04/25] xfs: uncached buffer reads need to return an error
2012-10-25 6:33 ` [PATCH 04/25] xfs: uncached buffer reads need to return an error Dave Chinner
@ 2012-10-26 8:48 ` Christoph Hellwig
2012-10-30 0:36 ` Phil White
1 sibling, 0 replies; 69+ messages in thread
From: Christoph Hellwig @ 2012-10-26 8:48 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:33:53PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> With verification being done as an IO completion callback, different
> errors can be returned from a read. Uncached reads only return a
> buffer or NULL on failure, which means the verification error cannot
> be returned to the caller.
>
> Split the error handling for these reads into two - a failure to get
> a buffer will still return NULL, but a read error will return a
> referenced buffer with b_error set rather than NULL. The caller is
> responsible for checking the error state of the buffer returned.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
I'd still prefer the error as return value, but in case you don't want
that yet:
Reviewed-by: Christoph Hellwig <hch@lst.de>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 21/25] xfs: add buffer pre-write callback
2012-10-25 6:34 ` [PATCH 21/25] xfs: add buffer pre-write callback Dave Chinner
@ 2012-10-26 8:50 ` Christoph Hellwig
2012-10-30 22:30 ` Dave Chinner
2012-10-30 13:32 ` Phil White
1 sibling, 1 reply; 69+ messages in thread
From: Christoph Hellwig @ 2012-10-26 8:50 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
> /*
> + * run the pre-io callback function if it exists. If this function
> + * fails it will mark the buffer with an error and the IO should
> + * not be dispatched.
> + */
> + if (bp->b_pre_io) {
> + bp->b_pre_io(bp);
> + if (bp->b_error) {
Wouldn't it be a cleaner calling convention to return the erro from the
callback?
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 25/25] xfs: add write verifiers to log recovery
2012-10-25 6:34 ` [PATCH 25/25] xfs: add write verifiers to log recovery Dave Chinner
@ 2012-10-26 8:54 ` Christoph Hellwig
2012-10-26 20:31 ` Dave Chinner
2012-10-30 13:44 ` Phil White
1 sibling, 1 reply; 69+ messages in thread
From: Christoph Hellwig @ 2012-10-26 8:54 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
> + * during log recovery, we don't have a quotainfo structure to
> + * pull the number of dquots per buffer out of, so we have to calculate
> + * it directly.
> + */
> + if (mp->m_log->l_flags & XLOG_ACTIVE_RECOVERY) {
> + dquots_per_buf = BBTOB(bp->b_length);
> + do_div(dquots_per_buf, sizeof(xfs_dqblk_t));
No need for do_div when dividing a 32-bit value by a constant.
I'd be almost tempted to do the calculation unconditionally to make the
code cleaner, too.
> + * There is no easy way to do this except for trying a bunch of magic
> + * number matches....
How do we make sure buffers used for the symlink or attr payload don't
match this?
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 25/25] xfs: add write verifiers to log recovery
2012-10-26 8:54 ` Christoph Hellwig
@ 2012-10-26 20:31 ` Dave Chinner
2012-10-30 12:23 ` Christoph Hellwig
0 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-26 20:31 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: xfs
On Fri, Oct 26, 2012 at 04:54:48AM -0400, Christoph Hellwig wrote:
> > + * during log recovery, we don't have a quotainfo structure to
> > + * pull the number of dquots per buffer out of, so we have to calculate
> > + * it directly.
> > + */
> > + if (mp->m_log->l_flags & XLOG_ACTIVE_RECOVERY) {
> > + dquots_per_buf = BBTOB(bp->b_length);
> > + do_div(dquots_per_buf, sizeof(xfs_dqblk_t));
>
> No need for do_div when dividing a 32-bit value by a constant.
>
> I'd be almost tempted to do the calculation unconditionally to make the
> code cleaner, too.
Ok.
> > + * There is no easy way to do this except for trying a bunch of magic
> > + * number matches....
>
> How do we make sure buffers used for the symlink or attr payload don't
> match this?
Remote attr buffers aren't logged - they are written sycnhronously
during the transaction - so won't get found by this. As for remote
symlink buffers, yeah, that might be a problem. Ultimately, both of
these buffer types are going to grow headers for CRCs, so this
problem will go away. I'm not sure how to address this problem
in the mean time short of putting the buffer content type into all
the buf_log_format headers. Do you have any better ideas?
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 01/25] xfs: growfs: don't read garbage for new secondary superblocks
2012-10-25 6:33 ` [PATCH 01/25] xfs: growfs: don't read garbage for new secondary superblocks Dave Chinner
@ 2012-10-30 0:17 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 0:17 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:33:50PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> When updating new secondary superblocks in a growfs operation, the
> sueprblock buffer is read from the newly grown region of the
> underlying device. This is not guaranteed to be zero, so violates
> the underlying assumption that the unused parts of superblocks are
> zero filled. Get a new buffer for these secondary superblocks to
> ensure that the unused regions are zero filled correctly.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> Reviewed-by: Christoph Hellwig <hch@lst.de>
> ---
> [ snip ]
Looks good to me too.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 02/25] xfs: invalidate allocbt blocks moved to the free list
2012-10-25 6:33 ` [PATCH 02/25] xfs: invalidate allocbt blocks moved to the free list Dave Chinner
2012-10-26 8:47 ` Christoph Hellwig
@ 2012-10-30 0:22 ` Phil White
1 sibling, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 0:22 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
This looks OK by me.
Reviewed-by: Phil White <pwhite@sgi.com>
On Thu, Oct 25, 2012 at 05:33:51PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> When we free a block from the alloc btree tree, we move it to the
> freelist held in the AGFL and mark it busy in the busy extent tree.
> This typically happens when we merge btree blocks.
>
> Once the transaction is committed and checkpointed, the block can
> remain on the free list for an indefinite amount of time. Now, this
> isn't the end of the world at this point - if the free list is
> shortened, the buffer is invalidated in the transaction that moves
> it back to free space. If the buffer is allocated as metadata from
> the free list, then all the modifications getted logged, and we have
> no issues, either. And if it gets allocated as userdata direct from
> the freelist, it gets invalidated and so will never get written.
>
> However, during the time it sits on the free list, pressure on the
> log can cause the AIL to be pushed and the buffer that covers the
> block gets pushed for write. IOWs, we end up writing a freed
> metadata block to disk. Again, this isn't the end of the world
> because we know from the above we are only writing to free space.
>
> The problem, however, is for validation callbacks. If the block was
> on old btree root block, then the level of the block is going to be
> higher than the current tree root, and so will fail validation.
> There may be other inconsistencies in the block as well, and
> currently we don't care because the block is in free space. Shutting
> down the filesystem because a freed block doesn't pass write
> validation, OTOH, is rather unfriendly.
>
> So, make sure we always invalidate buffers as they move from the
> free space trees to the free list so that we guarantee they never
> get written to disk while on the free list.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_alloc_btree.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/fs/xfs/xfs_alloc_btree.c b/fs/xfs/xfs_alloc_btree.c
> index f1647ca..f7876c6 100644
> --- a/fs/xfs/xfs_alloc_btree.c
> +++ b/fs/xfs/xfs_alloc_btree.c
> @@ -121,6 +121,8 @@ xfs_allocbt_free_block(
> xfs_extent_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1,
> XFS_EXTENT_BUSY_SKIP_DISCARD);
> xfs_trans_agbtree_delta(cur->bc_tp, -1);
> +
> + xfs_trans_binval(cur->bc_tp, bp);
> return 0;
> }
>
> --
> 1.7.10
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 03/25] xfs: make buffer read verication an IO completion function
2012-10-25 6:33 ` [PATCH 03/25] xfs: make buffer read verication an IO completion function Dave Chinner
@ 2012-10-30 0:29 ` Phil White
2012-10-30 0:45 ` Dave Chinner
0 siblings, 1 reply; 69+ messages in thread
From: Phil White @ 2012-10-30 0:29 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:33:52PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Add a verifier function callback capability to the buffer read
> interfaces. This will be used by the callers to supply a function
> that verifies the contents of the buffer when it is read from disk.
> This patch does not provide callback functions, but simply modifies
> the interfaces to allow them to be called.
>
> The reason for adding this to the read interfaces is that it is very
> difficult to tell fom the outside is a buffer was just read from
> disk or whether we just pulled it out of cache. Supplying a callbck
> allows the buffer cache to use it's internal knowledge of the buffer
> to execute it only when the buffer is read from disk.
>
> It is intended that the verifier functions will mark the buffer with
> an EFSCORRUPTED error when verification fails. This allows the
> reading context to distinguish a verification error from an IO
> error, and potentially take further actions on the buffer (e.g.
> attempt repair) based on the error reported.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> Reviewed-by: Christoph Hellwig <hch@lst.de>
> ---
> fs/xfs/xfs_alloc.c | 4 ++--
> fs/xfs/xfs_attr.c | 2 +-
> fs/xfs/xfs_btree.c | 21 ++++++++++++---------
> fs/xfs/xfs_buf.c | 13 +++++++++----
> fs/xfs/xfs_buf.h | 20 ++++++++++++--------
> fs/xfs/xfs_da_btree.c | 4 ++--
> fs/xfs/xfs_dir2_leaf.c | 2 +-
> fs/xfs/xfs_dquot.c | 4 ++--
> fs/xfs/xfs_fsops.c | 4 ++--
> fs/xfs/xfs_ialloc.c | 2 +-
> fs/xfs/xfs_inode.c | 2 +-
> fs/xfs/xfs_log.c | 3 +--
> fs/xfs/xfs_log_recover.c | 8 +++++---
> fs/xfs/xfs_mount.c | 6 +++---
> fs/xfs/xfs_qm.c | 5 +++--
> fs/xfs/xfs_rtalloc.c | 6 +++---
> fs/xfs/xfs_trans.h | 19 ++++++++-----------
> fs/xfs/xfs_trans_buf.c | 9 ++++++---
> fs/xfs/xfs_vnodeops.c | 2 +-
> 19 files changed, 75 insertions(+), 61 deletions(-)
>
> diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
> index 335206a..21c3db0 100644
> --- a/fs/xfs/xfs_alloc.c
> +++ b/fs/xfs/xfs_alloc.c
> @@ -447,7 +447,7 @@ xfs_alloc_read_agfl(
> error = xfs_trans_read_buf(
> mp, tp, mp->m_ddev_targp,
> XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
> - XFS_FSS_TO_BB(mp, 1), 0, &bp);
> + XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
> if (error)
> return error;
> ASSERT(!xfs_buf_geterror(bp));
> @@ -2110,7 +2110,7 @@ xfs_read_agf(
> error = xfs_trans_read_buf(
> mp, tp, mp->m_ddev_targp,
> XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
> - XFS_FSS_TO_BB(mp, 1), flags, bpp);
> + XFS_FSS_TO_BB(mp, 1), flags, bpp, NULL);
> if (error)
> return error;
> if (!*bpp)
> diff --git a/fs/xfs/xfs_attr.c b/fs/xfs/xfs_attr.c
> index 0ca1f0b..ebacb8d 100644
> --- a/fs/xfs/xfs_attr.c
> +++ b/fs/xfs/xfs_attr.c
> @@ -1980,7 +1980,7 @@ xfs_attr_rmtval_get(xfs_da_args_t *args)
> dblkno = XFS_FSB_TO_DADDR(mp, map[i].br_startblock);
> blkcnt = XFS_FSB_TO_BB(mp, map[i].br_blockcount);
> error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
> - dblkno, blkcnt, 0, &bp);
> + dblkno, blkcnt, 0, &bp, NULL);
> if (error)
> return(error);
>
> diff --git a/fs/xfs/xfs_btree.c b/fs/xfs/xfs_btree.c
> index e53e317..1937c9b 100644
> --- a/fs/xfs/xfs_btree.c
> +++ b/fs/xfs/xfs_btree.c
> @@ -266,9 +266,12 @@ xfs_btree_dup_cursor(
> for (i = 0; i < new->bc_nlevels; i++) {
> new->bc_ptrs[i] = cur->bc_ptrs[i];
> new->bc_ra[i] = cur->bc_ra[i];
> - if ((bp = cur->bc_bufs[i])) {
> - if ((error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
> - XFS_BUF_ADDR(bp), mp->m_bsize, 0, &bp))) {
> + bp = cur->bc_bufs[i];
> + if (bp) {
> + error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
> + XFS_BUF_ADDR(bp), mp->m_bsize,
> + 0, &bp, NULL);
> + if (error) {
> xfs_btree_del_cursor(new, error);
> *ncur = NULL;
> return error;
> @@ -624,10 +627,10 @@ xfs_btree_read_bufl(
>
> ASSERT(fsbno != NULLFSBLOCK);
> d = XFS_FSB_TO_DADDR(mp, fsbno);
> - if ((error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, d,
> - mp->m_bsize, lock, &bp))) {
> + error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, d,
> + mp->m_bsize, lock, &bp, NULL);
> + if (error)
> return error;
> - }
> ASSERT(!xfs_buf_geterror(bp));
> if (bp)
> xfs_buf_set_ref(bp, refval);
> @@ -650,7 +653,7 @@ xfs_btree_reada_bufl(
>
> ASSERT(fsbno != NULLFSBLOCK);
> d = XFS_FSB_TO_DADDR(mp, fsbno);
> - xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count);
> + xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, NULL);
> }
>
> /*
> @@ -670,7 +673,7 @@ xfs_btree_reada_bufs(
> ASSERT(agno != NULLAGNUMBER);
> ASSERT(agbno != NULLAGBLOCK);
> d = XFS_AGB_TO_DADDR(mp, agno, agbno);
> - xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count);
> + xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, NULL);
> }
>
> STATIC int
> @@ -998,7 +1001,7 @@ xfs_btree_read_buf_block(
>
> d = xfs_btree_ptr_to_daddr(cur, ptr);
> error = xfs_trans_read_buf(mp, cur->bc_tp, mp->m_ddev_targp, d,
> - mp->m_bsize, flags, bpp);
> + mp->m_bsize, flags, bpp, NULL);
> if (error)
> return error;
>
> diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
> index 933b793..7cab1b3 100644
> --- a/fs/xfs/xfs_buf.c
> +++ b/fs/xfs/xfs_buf.c
> @@ -654,7 +654,8 @@ xfs_buf_read_map(
> struct xfs_buftarg *target,
> struct xfs_buf_map *map,
> int nmaps,
> - xfs_buf_flags_t flags)
> + xfs_buf_flags_t flags,
> + xfs_buf_iodone_t verify)
> {
> struct xfs_buf *bp;
>
> @@ -666,6 +667,7 @@ xfs_buf_read_map(
>
> if (!XFS_BUF_ISDONE(bp)) {
> XFS_STATS_INC(xb_get_read);
> + bp->b_iodone = verify;
> _xfs_buf_read(bp, flags);
> } else if (flags & XBF_ASYNC) {
> /*
> @@ -691,13 +693,14 @@ void
> xfs_buf_readahead_map(
> struct xfs_buftarg *target,
> struct xfs_buf_map *map,
> - int nmaps)
> + int nmaps,
> + xfs_buf_iodone_t verify)
> {
> if (bdi_read_congested(target->bt_bdi))
> return;
>
> xfs_buf_read_map(target, map, nmaps,
> - XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD);
> + XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD, verify);
> }
>
> /*
> @@ -709,7 +712,8 @@ xfs_buf_read_uncached(
> struct xfs_buftarg *target,
> xfs_daddr_t daddr,
> size_t numblks,
> - int flags)
> + int flags,
> + xfs_buf_iodone_t verify)
> {
> xfs_buf_t *bp;
> int error;
> @@ -723,6 +727,7 @@ xfs_buf_read_uncached(
> bp->b_bn = daddr;
> bp->b_maps[0].bm_bn = daddr;
> bp->b_flags |= XBF_READ;
> + bp->b_iodone = verify;
>
> xfsbdstrat(target->bt_mount, bp);
> error = xfs_buf_iowait(bp);
> diff --git a/fs/xfs/xfs_buf.h b/fs/xfs/xfs_buf.h
> index 7c0b6a0..677b1dc 100644
> --- a/fs/xfs/xfs_buf.h
> +++ b/fs/xfs/xfs_buf.h
> @@ -100,6 +100,7 @@ typedef struct xfs_buftarg {
> struct xfs_buf;
> typedef void (*xfs_buf_iodone_t)(struct xfs_buf *);
>
> +
> #define XB_PAGES 2
>
> struct xfs_buf_map {
> @@ -159,7 +160,6 @@ typedef struct xfs_buf {
> #endif
> } xfs_buf_t;
>
> -
> /* Finding and Reading Buffers */
> struct xfs_buf *_xfs_buf_find(struct xfs_buftarg *target,
> struct xfs_buf_map *map, int nmaps,
> @@ -196,9 +196,10 @@ struct xfs_buf *xfs_buf_get_map(struct xfs_buftarg *target,
> xfs_buf_flags_t flags);
> struct xfs_buf *xfs_buf_read_map(struct xfs_buftarg *target,
> struct xfs_buf_map *map, int nmaps,
> - xfs_buf_flags_t flags);
> + xfs_buf_flags_t flags, xfs_buf_iodone_t verify);
> void xfs_buf_readahead_map(struct xfs_buftarg *target,
> - struct xfs_buf_map *map, int nmaps);
> + struct xfs_buf_map *map, int nmaps,
> + xfs_buf_iodone_t verify);
>
> static inline struct xfs_buf *
> xfs_buf_get(
> @@ -216,20 +217,22 @@ xfs_buf_read(
> struct xfs_buftarg *target,
> xfs_daddr_t blkno,
> size_t numblks,
> - xfs_buf_flags_t flags)
> + xfs_buf_flags_t flags,
> + xfs_buf_iodone_t verify)
> {
> DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
> - return xfs_buf_read_map(target, &map, 1, flags);
> + return xfs_buf_read_map(target, &map, 1, flags, verify);
> }
>
> static inline void
> xfs_buf_readahead(
> struct xfs_buftarg *target,
> xfs_daddr_t blkno,
> - size_t numblks)
> + size_t numblks,
> + xfs_buf_iodone_t verify)
> {
> DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
> - return xfs_buf_readahead_map(target, &map, 1);
> + return xfs_buf_readahead_map(target, &map, 1, verify);
> }
>
> struct xfs_buf *xfs_buf_get_empty(struct xfs_buftarg *target, size_t numblks);
> @@ -239,7 +242,8 @@ int xfs_buf_associate_memory(struct xfs_buf *bp, void *mem, size_t length);
> struct xfs_buf *xfs_buf_get_uncached(struct xfs_buftarg *target, size_t numblks,
> int flags);
> struct xfs_buf *xfs_buf_read_uncached(struct xfs_buftarg *target,
> - xfs_daddr_t daddr, size_t numblks, int flags);
> + xfs_daddr_t daddr, size_t numblks, int flags,
> + xfs_buf_iodone_t verify);
> void xfs_buf_hold(struct xfs_buf *bp);
>
> /* Releasing Buffers */
> diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
> index 7bfb7dd..41d8764 100644
> --- a/fs/xfs/xfs_da_btree.c
> +++ b/fs/xfs/xfs_da_btree.c
> @@ -2155,7 +2155,7 @@ xfs_da_read_buf(
>
> error = xfs_trans_read_buf_map(dp->i_mount, trans,
> dp->i_mount->m_ddev_targp,
> - mapp, nmap, 0, &bp);
> + mapp, nmap, 0, &bp, NULL);
> if (error)
> goto out_free;
>
> @@ -2231,7 +2231,7 @@ xfs_da_reada_buf(
> }
>
> mappedbno = mapp[0].bm_bn;
> - xfs_buf_readahead_map(dp->i_mount->m_ddev_targp, mapp, nmap);
> + xfs_buf_readahead_map(dp->i_mount->m_ddev_targp, mapp, nmap, NULL);
>
> out_free:
> if (mapp != &map)
> diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
> index 0b29625..bac8698 100644
> --- a/fs/xfs/xfs_dir2_leaf.c
> +++ b/fs/xfs/xfs_dir2_leaf.c
> @@ -926,7 +926,7 @@ xfs_dir2_leaf_readbuf(
> XFS_FSB_TO_DADDR(mp,
> map[mip->ra_index].br_startblock +
> mip->ra_offset),
> - (int)BTOBB(mp->m_dirblksize));
> + (int)BTOBB(mp->m_dirblksize), NULL);
> mip->ra_current = i;
> }
>
> diff --git a/fs/xfs/xfs_dquot.c b/fs/xfs/xfs_dquot.c
> index bf27fcc..e95f800 100644
> --- a/fs/xfs/xfs_dquot.c
> +++ b/fs/xfs/xfs_dquot.c
> @@ -439,7 +439,7 @@ xfs_qm_dqtobp(
> error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
> dqp->q_blkno,
> mp->m_quotainfo->qi_dqchunklen,
> - 0, &bp);
> + 0, &bp, NULL);
> if (error || !bp)
> return XFS_ERROR(error);
> }
> @@ -920,7 +920,7 @@ xfs_qm_dqflush(
> * Get the buffer containing the on-disk dquot
> */
> error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno,
> - mp->m_quotainfo->qi_dqchunklen, 0, &bp);
> + mp->m_quotainfo->qi_dqchunklen, 0, &bp, NULL);
> if (error)
> goto out_unlock;
>
> diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
> index 4beaede..917e121 100644
> --- a/fs/xfs/xfs_fsops.c
> +++ b/fs/xfs/xfs_fsops.c
> @@ -146,7 +146,7 @@ xfs_growfs_data_private(
> dpct = pct - mp->m_sb.sb_imax_pct;
> bp = xfs_buf_read_uncached(mp->m_ddev_targp,
> XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
> - XFS_FSS_TO_BB(mp, 1), 0);
> + XFS_FSS_TO_BB(mp, 1), 0, NULL);
> if (!bp)
> return EIO;
> xfs_buf_relse(bp);
> @@ -408,7 +408,7 @@ xfs_growfs_data_private(
> if (agno < oagcount) {
> error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
> XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
> - XFS_FSS_TO_BB(mp, 1), 0, &bp);
> + XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
> } else {
> bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
> XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
> diff --git a/fs/xfs/xfs_ialloc.c b/fs/xfs/xfs_ialloc.c
> index c5c4ef4..7c944e1 100644
> --- a/fs/xfs/xfs_ialloc.c
> +++ b/fs/xfs/xfs_ialloc.c
> @@ -1490,7 +1490,7 @@ xfs_read_agi(
>
> error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
> XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
> - XFS_FSS_TO_BB(mp, 1), 0, bpp);
> + XFS_FSS_TO_BB(mp, 1), 0, bpp, NULL);
> if (error)
> return error;
>
> diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
> index bba8f37..0b03578 100644
> --- a/fs/xfs/xfs_inode.c
> +++ b/fs/xfs/xfs_inode.c
> @@ -408,7 +408,7 @@ xfs_imap_to_bp(
>
> buf_flags |= XBF_UNMAPPED;
> error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
> - (int)imap->im_len, buf_flags, &bp);
> + (int)imap->im_len, buf_flags, &bp, NULL);
> if (error) {
> if (error != EAGAIN) {
> xfs_warn(mp,
> diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c
> index 46b6986..1d6d2ee 100644
> --- a/fs/xfs/xfs_log.c
> +++ b/fs/xfs/xfs_log.c
> @@ -1129,8 +1129,7 @@ xlog_iodone(xfs_buf_t *bp)
> * with it being freed after writing the unmount record to the
> * log.
> */
> -
> -} /* xlog_iodone */
> +}
>
> /*
> * Return size of each in-core log record buffer.
> diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
> index 651c988..757688a 100644
> --- a/fs/xfs/xfs_log_recover.c
> +++ b/fs/xfs/xfs_log_recover.c
> @@ -2144,7 +2144,7 @@ xlog_recover_buffer_pass2(
> buf_flags |= XBF_UNMAPPED;
>
> bp = xfs_buf_read(mp->m_ddev_targp, buf_f->blf_blkno, buf_f->blf_len,
> - buf_flags);
> + buf_flags, NULL);
> if (!bp)
> return XFS_ERROR(ENOMEM);
> error = bp->b_error;
> @@ -2237,7 +2237,8 @@ xlog_recover_inode_pass2(
> }
> trace_xfs_log_recover_inode_recover(log, in_f);
>
> - bp = xfs_buf_read(mp->m_ddev_targp, in_f->ilf_blkno, in_f->ilf_len, 0);
> + bp = xfs_buf_read(mp->m_ddev_targp, in_f->ilf_blkno, in_f->ilf_len, 0,
> + NULL);
> if (!bp) {
> error = ENOMEM;
> goto error;
> @@ -2548,7 +2549,8 @@ xlog_recover_dquot_pass2(
> ASSERT(dq_f->qlf_len == 1);
>
> error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dq_f->qlf_blkno,
> - XFS_FSB_TO_BB(mp, dq_f->qlf_len), 0, &bp);
> + XFS_FSB_TO_BB(mp, dq_f->qlf_len), 0, &bp,
> + NULL);
> if (error)
> return error;
>
> diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
> index 6f1c997..d39ad72 100644
> --- a/fs/xfs/xfs_mount.c
> +++ b/fs/xfs/xfs_mount.c
> @@ -652,7 +652,7 @@ xfs_readsb(xfs_mount_t *mp, int flags)
>
> reread:
> bp = xfs_buf_read_uncached(mp->m_ddev_targp, XFS_SB_DADDR,
> - BTOBB(sector_size), 0);
> + BTOBB(sector_size), 0, NULL);
> if (!bp) {
> if (loud)
> xfs_warn(mp, "SB buffer read failed");
> @@ -1002,7 +1002,7 @@ xfs_check_sizes(xfs_mount_t *mp)
> }
> bp = xfs_buf_read_uncached(mp->m_ddev_targp,
> d - XFS_FSS_TO_BB(mp, 1),
> - XFS_FSS_TO_BB(mp, 1), 0);
> + XFS_FSS_TO_BB(mp, 1), 0, NULL);
> if (!bp) {
> xfs_warn(mp, "last sector read failed");
> return EIO;
> @@ -1017,7 +1017,7 @@ xfs_check_sizes(xfs_mount_t *mp)
> }
> bp = xfs_buf_read_uncached(mp->m_logdev_targp,
> d - XFS_FSB_TO_BB(mp, 1),
> - XFS_FSB_TO_BB(mp, 1), 0);
> + XFS_FSB_TO_BB(mp, 1), 0, NULL);
> if (!bp) {
> xfs_warn(mp, "log device read failed");
> return EIO;
> diff --git a/fs/xfs/xfs_qm.c b/fs/xfs/xfs_qm.c
> index 48c750b..688f608 100644
> --- a/fs/xfs/xfs_qm.c
> +++ b/fs/xfs/xfs_qm.c
> @@ -892,7 +892,7 @@ xfs_qm_dqiter_bufs(
> while (blkcnt--) {
> error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
> XFS_FSB_TO_DADDR(mp, bno),
> - mp->m_quotainfo->qi_dqchunklen, 0, &bp);
> + mp->m_quotainfo->qi_dqchunklen, 0, &bp, NULL);
> if (error)
> break;
>
> @@ -979,7 +979,8 @@ xfs_qm_dqiterate(
> while (rablkcnt--) {
> xfs_buf_readahead(mp->m_ddev_targp,
> XFS_FSB_TO_DADDR(mp, rablkno),
> - mp->m_quotainfo->qi_dqchunklen);
> + mp->m_quotainfo->qi_dqchunklen,
> + NULL);
> rablkno++;
> }
> }
> diff --git a/fs/xfs/xfs_rtalloc.c b/fs/xfs/xfs_rtalloc.c
> index a69e0b4..b271ed9 100644
> --- a/fs/xfs/xfs_rtalloc.c
> +++ b/fs/xfs/xfs_rtalloc.c
> @@ -870,7 +870,7 @@ xfs_rtbuf_get(
> ASSERT(map.br_startblock != NULLFSBLOCK);
> error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
> XFS_FSB_TO_DADDR(mp, map.br_startblock),
> - mp->m_bsize, 0, &bp);
> + mp->m_bsize, 0, &bp, NULL);
> if (error)
> return error;
> ASSERT(!xfs_buf_geterror(bp));
> @@ -1873,7 +1873,7 @@ xfs_growfs_rt(
> */
> bp = xfs_buf_read_uncached(mp->m_rtdev_targp,
> XFS_FSB_TO_BB(mp, nrblocks - 1),
> - XFS_FSB_TO_BB(mp, 1), 0);
> + XFS_FSB_TO_BB(mp, 1), 0, NULL);
> if (!bp)
> return EIO;
> xfs_buf_relse(bp);
> @@ -2220,7 +2220,7 @@ xfs_rtmount_init(
> }
> bp = xfs_buf_read_uncached(mp->m_rtdev_targp,
> d - XFS_FSB_TO_BB(mp, 1),
> - XFS_FSB_TO_BB(mp, 1), 0);
> + XFS_FSB_TO_BB(mp, 1), 0, NULL);
> if (!bp) {
> xfs_warn(mp, "realtime device size check failed");
> return EIO;
> diff --git a/fs/xfs/xfs_trans.h b/fs/xfs/xfs_trans.h
> index db05654..f02d402 100644
> --- a/fs/xfs/xfs_trans.h
> +++ b/fs/xfs/xfs_trans.h
> @@ -464,10 +464,7 @@ xfs_trans_get_buf(
> int numblks,
> uint flags)
> {
> - struct xfs_buf_map map = {
> - .bm_bn = blkno,
> - .bm_len = numblks,
> - };
> + DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
> return xfs_trans_get_buf_map(tp, target, &map, 1, flags);
> }
>
> @@ -476,7 +473,8 @@ int xfs_trans_read_buf_map(struct xfs_mount *mp,
> struct xfs_buftarg *target,
> struct xfs_buf_map *map, int nmaps,
> xfs_buf_flags_t flags,
> - struct xfs_buf **bpp);
> + struct xfs_buf **bpp,
> + xfs_buf_iodone_t verify);
>
> static inline int
> xfs_trans_read_buf(
> @@ -486,13 +484,12 @@ xfs_trans_read_buf(
> xfs_daddr_t blkno,
> int numblks,
> xfs_buf_flags_t flags,
> - struct xfs_buf **bpp)
> + struct xfs_buf **bpp,
> + xfs_buf_iodone_t verify)
> {
> - struct xfs_buf_map map = {
> - .bm_bn = blkno,
> - .bm_len = numblks,
> - };
> - return xfs_trans_read_buf_map(mp, tp, target, &map, 1, flags, bpp);
> + DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
> + return xfs_trans_read_buf_map(mp, tp, target, &map, 1,
> + flags, bpp, verify);
> }
>
> struct xfs_buf *xfs_trans_getsb(xfs_trans_t *, struct xfs_mount *, int);
> diff --git a/fs/xfs/xfs_trans_buf.c b/fs/xfs/xfs_trans_buf.c
> index 6311b99..9776282 100644
> --- a/fs/xfs/xfs_trans_buf.c
> +++ b/fs/xfs/xfs_trans_buf.c
> @@ -257,7 +257,8 @@ xfs_trans_read_buf_map(
> struct xfs_buf_map *map,
> int nmaps,
> xfs_buf_flags_t flags,
> - struct xfs_buf **bpp)
> + struct xfs_buf **bpp,
> + xfs_buf_iodone_t verify)
> {
> xfs_buf_t *bp;
> xfs_buf_log_item_t *bip;
> @@ -265,7 +266,7 @@ xfs_trans_read_buf_map(
>
> *bpp = NULL;
> if (!tp) {
> - bp = xfs_buf_read_map(target, map, nmaps, flags);
> + bp = xfs_buf_read_map(target, map, nmaps, flags, verify);
> if (!bp)
> return (flags & XBF_TRYLOCK) ?
> EAGAIN : XFS_ERROR(ENOMEM);
> @@ -312,7 +313,9 @@ xfs_trans_read_buf_map(
> if (!(XFS_BUF_ISDONE(bp))) {
> trace_xfs_trans_read_buf_io(bp, _RET_IP_);
> ASSERT(!XFS_BUF_ISASYNC(bp));
> + ASSERT(bp->b_iodone == NULL);
> XFS_BUF_READ(bp);
> + bp->b_iodone = verify;
> xfsbdstrat(tp->t_mountp, bp);
> error = xfs_buf_iowait(bp);
> if (error) {
> @@ -349,7 +352,7 @@ xfs_trans_read_buf_map(
> return 0;
> }
>
> - bp = xfs_buf_read_map(target, map, nmaps, flags);
> + bp = xfs_buf_read_map(target, map, nmaps, flags, verify);
> if (bp == NULL) {
> *bpp = NULL;
> return (flags & XBF_TRYLOCK) ?
> diff --git a/fs/xfs/xfs_vnodeops.c b/fs/xfs/xfs_vnodeops.c
> index 2ee1f49..f409fda 100644
> --- a/fs/xfs/xfs_vnodeops.c
> +++ b/fs/xfs/xfs_vnodeops.c
> @@ -80,7 +80,7 @@ xfs_readlink_bmap(
> d = XFS_FSB_TO_DADDR(mp, mval[n].br_startblock);
> byte_cnt = XFS_FSB_TO_B(mp, mval[n].br_blockcount);
>
> - bp = xfs_buf_read(mp->m_ddev_targp, d, BTOBB(byte_cnt), 0);
> + bp = xfs_buf_read(mp->m_ddev_targp, d, BTOBB(byte_cnt), 0, NULL);
> if (!bp)
> return XFS_ERROR(ENOMEM);
> error = bp->b_error;
> --
> 1.7.10
This is OK with me so far, but I have comments on some of the callbacks.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 04/25] xfs: uncached buffer reads need to return an error
2012-10-25 6:33 ` [PATCH 04/25] xfs: uncached buffer reads need to return an error Dave Chinner
2012-10-26 8:48 ` Christoph Hellwig
@ 2012-10-30 0:36 ` Phil White
1 sibling, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 0:36 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:33:53PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> With verification being done as an IO completion callback, different
> errors can be returned from a read. Uncached reads only return a
> buffer or NULL on failure, which means the verification error cannot
> be returned to the caller.
>
> Split the error handling for these reads into two - a failure to get
> a buffer will still return NULL, but a read error will return a
> referenced buffer with b_error set rather than NULL. The caller is
> responsible for checking the error state of the buffer returned.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_buf.c | 9 ++-------
> fs/xfs/xfs_fsops.c | 5 +++++
> fs/xfs/xfs_mount.c | 6 ++++++
> fs/xfs/xfs_rtalloc.c | 9 ++++++++-
> 4 files changed, 21 insertions(+), 8 deletions(-)
>
> diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
> index 7cab1b3..62b7e89 100644
> --- a/fs/xfs/xfs_buf.c
> +++ b/fs/xfs/xfs_buf.c
> @@ -715,8 +715,7 @@ xfs_buf_read_uncached(
> int flags,
> xfs_buf_iodone_t verify)
> {
> - xfs_buf_t *bp;
> - int error;
> + struct xfs_buf *bp;
>
> bp = xfs_buf_get_uncached(target, numblks, flags);
> if (!bp)
> @@ -730,11 +729,7 @@ xfs_buf_read_uncached(
> bp->b_iodone = verify;
>
> xfsbdstrat(target->bt_mount, bp);
> - error = xfs_buf_iowait(bp);
> - if (error) {
> - xfs_buf_relse(bp);
> - return NULL;
> - }
> + xfs_buf_iowait(bp);
> return bp;
> }
>
> diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
> index 917e121..dee14eb 100644
> --- a/fs/xfs/xfs_fsops.c
> +++ b/fs/xfs/xfs_fsops.c
> @@ -149,6 +149,11 @@ xfs_growfs_data_private(
> XFS_FSS_TO_BB(mp, 1), 0, NULL);
> if (!bp)
> return EIO;
> + if (bp->b_error) {
> + int error = bp->b_error;
> + xfs_buf_relse(bp);
> + return error;
> + }
> xfs_buf_relse(bp);
>
> new = nb; /* use new as a temporary here */
> diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
> index d39ad72..dc51e32 100644
> --- a/fs/xfs/xfs_mount.c
> +++ b/fs/xfs/xfs_mount.c
> @@ -658,6 +658,12 @@ reread:
> xfs_warn(mp, "SB buffer read failed");
> return EIO;
> }
> + if (bp->b_error) {
> + error = bp->b_error;
> + if (loud)
> + xfs_warn(mp, "SB validate failed");
> + goto release_buf;
> + }
>
> /*
> * Initialize the mount structure from the superblock.
> diff --git a/fs/xfs/xfs_rtalloc.c b/fs/xfs/xfs_rtalloc.c
> index b271ed9..98dc670 100644
> --- a/fs/xfs/xfs_rtalloc.c
> +++ b/fs/xfs/xfs_rtalloc.c
> @@ -1876,6 +1876,11 @@ xfs_growfs_rt(
> XFS_FSB_TO_BB(mp, 1), 0, NULL);
> if (!bp)
> return EIO;
> + if (bp->b_error) {
> + error = bp->b_error;
> + xfs_buf_relse(bp);
> + return error;
> + }
> xfs_buf_relse(bp);
>
> /*
> @@ -2221,8 +2226,10 @@ xfs_rtmount_init(
> bp = xfs_buf_read_uncached(mp->m_rtdev_targp,
> d - XFS_FSB_TO_BB(mp, 1),
> XFS_FSB_TO_BB(mp, 1), 0, NULL);
> - if (!bp) {
> + if (!bp || bp->b_error) {
> xfs_warn(mp, "realtime device size check failed");
> + if (bp)
> + xfs_buf_relse(bp);
> return EIO;
> }
> xfs_buf_relse(bp);
> --
> 1.7.10
>
> _______________________________________________
> xfs mailing list
> xfs@oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs
Looks OK.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 03/25] xfs: make buffer read verication an IO completion function
2012-10-30 0:29 ` Phil White
@ 2012-10-30 0:45 ` Dave Chinner
2012-10-30 0:55 ` Phil White
0 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-30 0:45 UTC (permalink / raw)
To: Phil White; +Cc: xfs
On Mon, Oct 29, 2012 at 05:29:17PM -0700, Phil White wrote:
> On Thu, Oct 25, 2012 at 05:33:52PM +1100, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@redhat.com>
> >
> > Add a verifier function callback capability to the buffer read
> > interfaces. This will be used by the callers to supply a function
> > that verifies the contents of the buffer when it is read from disk.
> > This patch does not provide callback functions, but simply modifies
> > the interfaces to allow them to be called.
> >
> > The reason for adding this to the read interfaces is that it is very
> > difficult to tell fom the outside is a buffer was just read from
> > disk or whether we just pulled it out of cache. Supplying a callbck
> > allows the buffer cache to use it's internal knowledge of the buffer
> > to execute it only when the buffer is read from disk.
> >
> > It is intended that the verifier functions will mark the buffer with
> > an EFSCORRUPTED error when verification fails. This allows the
> > reading context to distinguish a verification error from an IO
> > error, and potentially take further actions on the buffer (e.g.
> > attempt repair) based on the error reported.
> >
> > Signed-off-by: Dave Chinner <dchinner@redhat.com>
> > Reviewed-by: Christoph Hellwig <hch@lst.de>
......
> This is OK with me so far, but I have comments on some of the callbacks.
/me looks around, doesn't find anything...
The comments must be elsewhere. ;)
BTW, Phil, can you trim away all the bits of the patch you aren't
commenting on? Having to scroll through hundreds of lines of quoted
email to find your replies is a little slow, and it's quite easy to
miss comments when they are widely spread apart. Most people just
quote the patch hunk they are making the comment about to avoid this
problem....
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 05/25] xfs: verify superblocks as they are read from disk
2012-10-25 6:33 ` [PATCH 05/25] xfs: verify superblocks as they are read from disk Dave Chinner
@ 2012-10-30 0:48 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 0:48 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:33:54PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Add a superblock verify callback function and pass it into the
> buffer read functions. Remove the now redundant verification code
> that is currently in use.
>
> Adding verification shows that secondary superblocks never have
> their "sb_inprogress" flag cleared by mkfs.xfs, so when validating
> the secondary superblocks during a grow operation we have to avoid
> checking this field. Even if we fix mkfs, we will still have to
> ignore this field for verification purposes unless a version of mkfs
> that does not have this bug was used.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_fsops.c | 4 +-
> fs/xfs/xfs_log_recover.c | 5 ++-
> fs/xfs/xfs_mount.c | 98 +++++++++++++++++++++++++++++-----------------
> fs/xfs/xfs_mount.h | 3 +-
> 4 files changed, 69 insertions(+), 41 deletions(-)
>
> diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
> index dee14eb..302b99c 100644
> --- a/fs/xfs/xfs_fsops.c
> +++ b/fs/xfs/xfs_fsops.c
> @@ -413,7 +413,8 @@ xfs_growfs_data_private(
> if (agno < oagcount) {
> error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
> XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
> - XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
> + XFS_FSS_TO_BB(mp, 1), 0, &bp,
> + xfs_sb_read_verify);
> } else {
> bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
> XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
> @@ -431,6 +432,7 @@ xfs_growfs_data_private(
> break;
> }
> xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
> +
> /*
> * If we get an error writing out the alternate superblocks,
> * just issue a warning and continue. The real work is
> diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
> index 757688a..4cf7ae8 100644
> --- a/fs/xfs/xfs_log_recover.c
> +++ b/fs/xfs/xfs_log_recover.c
> @@ -3692,13 +3692,14 @@ xlog_do_recover(
>
> /*
> * Now that we've finished replaying all buffer and inode
> - * updates, re-read in the superblock.
> + * updates, re-read in the superblock and reverify it.
> */
> bp = xfs_getsb(log->l_mp, 0);
> XFS_BUF_UNDONE(bp);
> ASSERT(!(XFS_BUF_ISWRITE(bp)));
> XFS_BUF_READ(bp);
> XFS_BUF_UNASYNC(bp);
> + bp->b_iodone = xfs_sb_read_verify;
> xfsbdstrat(log->l_mp, bp);
> error = xfs_buf_iowait(bp);
> if (error) {
> @@ -3710,7 +3711,7 @@ xlog_do_recover(
>
> /* Convert superblock from on-disk format */
> sbp = &log->l_mp->m_sb;
> - xfs_sb_from_disk(log->l_mp, XFS_BUF_TO_SBP(bp));
> + xfs_sb_from_disk(sbp, XFS_BUF_TO_SBP(bp));
> ASSERT(sbp->sb_magicnum == XFS_SB_MAGIC);
> ASSERT(xfs_sb_good_version(sbp));
> xfs_buf_relse(bp);
> diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
> index dc51e32..8699e5e 100644
> --- a/fs/xfs/xfs_mount.c
> +++ b/fs/xfs/xfs_mount.c
> @@ -304,9 +304,8 @@ STATIC int
> xfs_mount_validate_sb(
> xfs_mount_t *mp,
> xfs_sb_t *sbp,
> - int flags)
> + bool check_inprogress)
> {
> - int loud = !(flags & XFS_MFSI_QUIET);
>
> /*
> * If the log device and data device have the
> @@ -316,21 +315,18 @@ xfs_mount_validate_sb(
> * a volume filesystem in a non-volume manner.
> */
> if (sbp->sb_magicnum != XFS_SB_MAGIC) {
> - if (loud)
> - xfs_warn(mp, "bad magic number");
> + xfs_warn(mp, "bad magic number");
> return XFS_ERROR(EWRONGFS);
> }
>
> if (!xfs_sb_good_version(sbp)) {
> - if (loud)
> - xfs_warn(mp, "bad version");
> + xfs_warn(mp, "bad version");
> return XFS_ERROR(EWRONGFS);
> }
>
> if (unlikely(
> sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
> - if (loud)
> - xfs_warn(mp,
> + xfs_warn(mp,
> "filesystem is marked as having an external log; "
> "specify logdev on the mount command line.");
> return XFS_ERROR(EINVAL);
> @@ -338,8 +334,7 @@ xfs_mount_validate_sb(
>
> if (unlikely(
> sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
> - if (loud)
> - xfs_warn(mp,
> + xfs_warn(mp,
> "filesystem is marked as having an internal log; "
> "do not specify logdev on the mount command line.");
> return XFS_ERROR(EINVAL);
> @@ -373,8 +368,7 @@ xfs_mount_validate_sb(
> sbp->sb_dblocks == 0 ||
> sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp) ||
> sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp))) {
> - if (loud)
> - XFS_CORRUPTION_ERROR("SB sanity check failed",
> + XFS_CORRUPTION_ERROR("SB sanity check failed",
> XFS_ERRLEVEL_LOW, mp, sbp);
> return XFS_ERROR(EFSCORRUPTED);
> }
> @@ -383,12 +377,10 @@ xfs_mount_validate_sb(
> * Until this is fixed only page-sized or smaller data blocks work.
> */
> if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
> - if (loud) {
> - xfs_warn(mp,
> + xfs_warn(mp,
> "File system with blocksize %d bytes. "
> "Only pagesize (%ld) or less will currently work.",
> sbp->sb_blocksize, PAGE_SIZE);
> - }
> return XFS_ERROR(ENOSYS);
> }
>
> @@ -402,23 +394,20 @@ xfs_mount_validate_sb(
> case 2048:
> break;
> default:
> - if (loud)
> - xfs_warn(mp, "inode size of %d bytes not supported",
> + xfs_warn(mp, "inode size of %d bytes not supported",
> sbp->sb_inodesize);
> return XFS_ERROR(ENOSYS);
> }
>
> if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
> xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
> - if (loud)
> - xfs_warn(mp,
> + xfs_warn(mp,
> "file system too large to be mounted on this system.");
> return XFS_ERROR(EFBIG);
> }
>
> - if (unlikely(sbp->sb_inprogress)) {
> - if (loud)
> - xfs_warn(mp, "file system busy");
> + if (check_inprogress && sbp->sb_inprogress) {
> + xfs_warn(mp, "Offline file system operation in progress!");
> return XFS_ERROR(EFSCORRUPTED);
> }
>
> @@ -426,9 +415,7 @@ xfs_mount_validate_sb(
> * Version 1 directory format has never worked on Linux.
> */
> if (unlikely(!xfs_sb_version_hasdirv2(sbp))) {
> - if (loud)
> - xfs_warn(mp,
> - "file system using version 1 directory format");
> + xfs_warn(mp, "file system using version 1 directory format");
> return XFS_ERROR(ENOSYS);
> }
>
> @@ -521,11 +508,9 @@ out_unwind:
>
> void
> xfs_sb_from_disk(
> - struct xfs_mount *mp,
> + struct xfs_sb *to,
> xfs_dsb_t *from)
> {
> - struct xfs_sb *to = &mp->m_sb;
> -
> to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
> to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
> to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
> @@ -627,6 +612,50 @@ xfs_sb_to_disk(
> }
> }
>
> +void
> +xfs_sb_read_verify(
> + struct xfs_buf *bp)
> +{
> + struct xfs_mount *mp = bp->b_target->bt_mount;
> + struct xfs_sb sb;
> + int error;
> +
> + xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp));
> +
> + /*
> + * Only check the in progress field for the primary superblock as
> + * mkfs.xfs doesn't clear it from secondary superblocks.
> + */
> + error = xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR);
> + if (error)
> + xfs_buf_ioerror(bp, error);
> + bp->b_iodone = NULL;
> + xfs_buf_ioend(bp, 0);
> +}
> +
> +/*
> + * We may be probed for a filesystem match, so we may not want to emit
> + * messages when the superblock buffer is not actually an XFS superblock.
> + * If we find an XFS superblock, the run a normal, noisy mount because we are
> + * really going to mount it and want to know about errors.
> + */
> +void
> +xfs_sb_quiet_read_verify(
> + struct xfs_buf *bp)
> +{
> + struct xfs_sb sb;
> +
> + xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp));
> +
> + if (sb.sb_magicnum == XFS_SB_MAGIC) {
> + /* XFS filesystem, verify noisily! */
> + xfs_sb_read_verify(bp);
> + return;
> + }
> + /* quietly fail */
> + xfs_buf_ioerror(bp, EFSCORRUPTED);
> +}
> +
> /*
> * xfs_readsb
> *
> @@ -652,7 +681,9 @@ xfs_readsb(xfs_mount_t *mp, int flags)
>
> reread:
> bp = xfs_buf_read_uncached(mp->m_ddev_targp, XFS_SB_DADDR,
> - BTOBB(sector_size), 0, NULL);
> + BTOBB(sector_size), 0,
> + loud ? xfs_sb_read_verify
> + : xfs_sb_quiet_read_verify);
> if (!bp) {
> if (loud)
> xfs_warn(mp, "SB buffer read failed");
> @@ -667,15 +698,8 @@ reread:
>
> /*
> * Initialize the mount structure from the superblock.
> - * But first do some basic consistency checking.
> */
> - xfs_sb_from_disk(mp, XFS_BUF_TO_SBP(bp));
> - error = xfs_mount_validate_sb(mp, &(mp->m_sb), flags);
> - if (error) {
> - if (loud)
> - xfs_warn(mp, "SB validate failed");
> - goto release_buf;
> - }
> + xfs_sb_from_disk(&mp->m_sb, XFS_BUF_TO_SBP(bp));
>
> /*
> * We must be able to do sector-sized and sector-aligned IO.
> diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
> index a631ca3..82b8fda 100644
> --- a/fs/xfs/xfs_mount.h
> +++ b/fs/xfs/xfs_mount.h
> @@ -382,10 +382,11 @@ extern void xfs_set_low_space_thresholds(struct xfs_mount *);
>
> #endif /* __KERNEL__ */
>
> +extern void xfs_sb_read_verify(struct xfs_buf *);
> extern void xfs_mod_sb(struct xfs_trans *, __int64_t);
> extern int xfs_initialize_perag(struct xfs_mount *, xfs_agnumber_t,
> xfs_agnumber_t *);
> -extern void xfs_sb_from_disk(struct xfs_mount *, struct xfs_dsb *);
> +extern void xfs_sb_from_disk(struct xfs_sb *, struct xfs_dsb *);
> extern void xfs_sb_to_disk(struct xfs_dsb *, struct xfs_sb *, __int64_t);
>
> #endif /* __XFS_MOUNT_H__ */
> --
> 1.7.10
>
> _______________________________________________
> xfs mailing list
> xfs@oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs
Looks good to me.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 06/25] xfs: verify AGF blocks as they are read from disk
2012-10-25 6:33 ` [PATCH 06/25] xfs: verify AGF blocks " Dave Chinner
@ 2012-10-30 0:51 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 0:51 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:33:55PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Add an AGF block verify callback function and pass it into the
> buffer read functions. This replaces the existing verification that
> is done after the read completes.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> Reviewed-by: Christoph Hellwig <hch@lst.de>
> ---
> fs/xfs/xfs_alloc.c | 60 +++++++++++++++++++++++++++++-----------------------
> 1 file changed, 34 insertions(+), 26 deletions(-)
>
> diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
> index 21c3db0..bd565a2 100644
> --- a/fs/xfs/xfs_alloc.c
> +++ b/fs/xfs/xfs_alloc.c
> @@ -2091,6 +2091,39 @@ xfs_alloc_put_freelist(
> return 0;
> }
>
> +static void
> +xfs_agf_read_verify(
> + struct xfs_buf *bp)
> + {
> + struct xfs_mount *mp = bp->b_target->bt_mount;
> + struct xfs_agf *agf;
> + int agf_ok;
> +
> + agf = XFS_BUF_TO_AGF(bp);
> +
> + agf_ok = agf->agf_magicnum == cpu_to_be32(XFS_AGF_MAGIC) &&
> + XFS_AGF_GOOD_VERSION(be32_to_cpu(agf->agf_versionnum)) &&
> + be32_to_cpu(agf->agf_freeblks) <= be32_to_cpu(agf->agf_length) &&
> + be32_to_cpu(agf->agf_flfirst) < XFS_AGFL_SIZE(mp) &&
> + be32_to_cpu(agf->agf_fllast) < XFS_AGFL_SIZE(mp) &&
> + be32_to_cpu(agf->agf_flcount) <= XFS_AGFL_SIZE(mp) &&
> + be32_to_cpu(agf->agf_seqno) == bp->b_pag->pag_agno;
> +
> + if (xfs_sb_version_haslazysbcount(&mp->m_sb))
> + agf_ok = agf_ok && be32_to_cpu(agf->agf_btreeblks) <=
> + be32_to_cpu(agf->agf_length);
> +
> + if (unlikely(XFS_TEST_ERROR(!agf_ok, mp, XFS_ERRTAG_ALLOC_READ_AGF,
> + XFS_RANDOM_ALLOC_READ_AGF))) {
> + XFS_CORRUPTION_ERROR("xfs_alloc_read_agf",
> + XFS_ERRLEVEL_LOW, mp, agf);
> + xfs_buf_ioerror(bp, EFSCORRUPTED);
> + }
Shouldn't this be XFS_CORRUPTION_ERROR("xfs_agf_read_verify", ...) ?
-Phil
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 07/25] xfs: verify AGI blocks as they are read from disk
2012-10-25 6:33 ` [PATCH 07/25] xfs: verify AGI " Dave Chinner
@ 2012-10-30 0:53 ` Phil White
2012-10-30 22:13 ` Dave Chinner
0 siblings, 1 reply; 69+ messages in thread
From: Phil White @ 2012-10-30 0:53 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:33:56PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Add an AGI block verify callback function and pass it into the
> buffer read functions. Remove the now redundant verification code
> that is currently in use.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> Reviewed-by: Christoph Hellwig <hch@lst.de>
> ---
> fs/xfs/xfs_ialloc.c | 47 ++++++++++++++++++++++++++---------------------
> 1 file changed, 26 insertions(+), 21 deletions(-)
>
> diff --git a/fs/xfs/xfs_ialloc.c b/fs/xfs/xfs_ialloc.c
> index 7c944e1..9311ae5 100644
> --- a/fs/xfs/xfs_ialloc.c
> +++ b/fs/xfs/xfs_ialloc.c
> @@ -1472,6 +1472,31 @@ xfs_check_agi_unlinked(
> #define xfs_check_agi_unlinked(agi)
> #endif
>
> +static void
> +xfs_agi_read_verify(
> + struct xfs_buf *bp)
> +{
> + struct xfs_mount *mp = bp->b_target->bt_mount;
> + struct xfs_agi *agi = XFS_BUF_TO_AGI(bp);
> + int agi_ok;
> +
> + /*
> + * Validate the magic number of the agi block.
> + */
> + agi_ok = agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC) &&
> + XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum)) &&
> + be32_to_cpu(agi->agi_seqno) == bp->b_pag->pag_agno;
> + if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI,
> + XFS_RANDOM_IALLOC_READ_AGI))) {
> + XFS_CORRUPTION_ERROR("xfs_read_agi", XFS_ERRLEVEL_LOW,
> + mp, agi);
> + xfs_buf_ioerror(bp, EFSCORRUPTED);
> + }
> + xfs_check_agi_unlinked(agi);
> + bp->b_iodone = NULL;
> + xfs_buf_ioend(bp, 0);
> +}
> +
In like fashion, shouldn't this be XFS_CORRUPTION_ERROR("xfs_agi_read_verify",
...)? In principle, it might be called from somewhere else in the future.
-Phil
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 03/25] xfs: make buffer read verication an IO completion function
2012-10-30 0:45 ` Dave Chinner
@ 2012-10-30 0:55 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 0:55 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs, Phil White
On Tue, Oct 30, 2012 at 11:45:46AM +1100, Dave Chinner wrote:
> > This is OK with me so far, but I have comments on some of the callbacks.
>
> /me looks around, doesn't find anything...
>
> The comments must be elsewhere. ;)
You surmised correctly.
> BTW, Phil, can you trim away all the bits of the patch you aren't
> commenting on? Having to scroll through hundreds of lines of quoted
> email to find your replies is a little slow, and it's quite easy to
> miss comments when they are widely spread apart. Most people just
> quote the patch hunk they are making the comment about to avoid this
> problem....
Will do. Sorry about that!
-Phil
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 08/25] xfs: verify AGFL blocks as they are read from disk
2012-10-25 6:33 ` [PATCH 08/25] xfs: verify AGFL " Dave Chinner
@ 2012-10-30 1:00 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 1:00 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:33:57PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Add an AGFL block verify callback function and pass it into the
> buffer read functions.
>
> While this commit adds verification code to the AGFL, it cannot be
> used reliably until the CRC format change comes along as mkfs does
> not initialise the full AGFL. Hence it can be full of garbage at the
> first mount and will fail verification right now. CRC enabled
> filesystems won't have this problem, so leave the code that has
> already been written ifdef'd out until the proper time.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_alloc.c | 40 +++++++++++++++++++++++++++++++++++++++-
> 1 file changed, 39 insertions(+), 1 deletion(-)
Okee doke by me.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 09/25] xfs: verify inode buffers as they are read from disk
2012-10-25 6:33 ` [PATCH 09/25] xfs: verify inode buffers " Dave Chinner
@ 2012-10-30 1:06 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 1:06 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:33:58PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Add an inode buffer verify callback function and pass it into the
> buffer read functions. Inodes are special in that the verbose checks
> will be done when reading the inode, but we still need to sanity
> check the buffer when that is first read. Always verify the magic
> numbers in all inodes in the buffer, rather than jus ton debug
> kernels.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_inode.c | 100 +++++++++++++++++++++++++++-------------------------
> 1 file changed, 51 insertions(+), 49 deletions(-)
Hunky dory as it's essentially just lifting the same code into a
function. I wonder a bit if all those checks are strictly necessary vs.
simply being belt-and-suspenders safe, but that thought's not related to
your patch.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 10/25] xfs: verify btree blocks as they are read from disk
2012-10-25 6:33 ` [PATCH 10/25] xfs: verify btree blocks " Dave Chinner
@ 2012-10-30 1:14 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 1:14 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:33:59PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Add an btree block verify callback function and pass it into the
> buffer read functions. Because each different btree block type
> requires different verification, add a function to the ops structure
> that is called from the generic code.
>
> Also, propagate the verification callback functions through the
> readahead functions, and into the external bmap and bulkstat inode
> readahead code that uses the generic btree buffer read functions.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_alloc_btree.c | 49 +++++++++++++++++++++++++++++++++
> fs/xfs/xfs_bmap.c | 60 ++++++++++++++++++++++++-----------------
> fs/xfs/xfs_bmap_btree.c | 47 ++++++++++++++++++++++++++++++++
> fs/xfs/xfs_bmap_btree.h | 1 +
> fs/xfs/xfs_btree.c | 66 +++++++++++++++++++++++----------------------
> fs/xfs/xfs_btree.h | 10 ++++---
> fs/xfs/xfs_ialloc_btree.c | 40 +++++++++++++++++++++++++++
> fs/xfs/xfs_inode.c | 2 +-
> fs/xfs/xfs_inode.h | 1 +
> fs/xfs/xfs_itable.c | 3 ++-
> 10 files changed, 218 insertions(+), 61 deletions(-)
Looks good. I remember there being conversations about whether or not
XFS needed to go on a stack diet for some platforms. But in general,
I think there's merit to providing these functions as callbacks.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 11/25] xfs: verify dquot blocks as they are read from disk
2012-10-25 6:34 ` [PATCH 11/25] xfs: verify dquot " Dave Chinner
@ 2012-10-30 1:36 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 1:36 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:00PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Add a dquot buffer verify callback function and pass it into the
> buffer read functions. This checks all the dquots in a buffer, but
> cannot completely verify the dquot ids are correct. Also, errors
> cannot be repaired, so an additional function is added to repair bad
> dquots in the buffer if such an error is detected in a context where
> repair is allowed.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
Looks good.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 12/25] xfs: add verifier callback to directory read code
2012-10-25 6:34 ` [PATCH 12/25] xfs: add verifier callback to directory read code Dave Chinner
@ 2012-10-30 3:15 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 3:15 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:01PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> Reviewed-by: Christoph Hellwig <hch@lst.de>
> ---
> fs/xfs/xfs_attr.c | 23 ++++++++++++-----------
> fs/xfs/xfs_attr_leaf.c | 18 +++++++++---------
> fs/xfs/xfs_da_btree.c | 44 ++++++++++++++++++++++++++++----------------
> fs/xfs/xfs_da_btree.h | 7 ++++---
> fs/xfs/xfs_dir2_block.c | 23 ++++++++++++-----------
> fs/xfs/xfs_dir2_leaf.c | 33 ++++++++++++++++-----------------
> fs/xfs/xfs_dir2_node.c | 43 ++++++++++++++++++++-----------------------
> fs/xfs/xfs_file.c | 2 +-
> 8 files changed, 102 insertions(+), 91 deletions(-)
More of the same.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 13/25] xfs: factor dir2 block read operations
2012-10-25 6:34 ` [PATCH 13/25] xfs: factor dir2 block read operations Dave Chinner
@ 2012-10-30 3:23 ` Phil White
2012-10-30 22:16 ` Dave Chinner
0 siblings, 1 reply; 69+ messages in thread
From: Phil White @ 2012-10-30 3:23 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:02PM +1100, Dave Chinner wrote:
> +static void
> +xfs_dir2_block_need_space(
> ...
> + /*
> + * If there are stale entries we'll use one for the leaf.
> + */
> + if (btp->stale) {
> + if (be16_to_cpu(bf[0].length) >= len) {
> + /*
> + * The biggest entry enough to avoid compaction.
> + */
> + dup = (xfs_dir2_data_unused_t *)
> + ((char *)hdr + be16_to_cpu(bf[0].offset));
> + goto out;
> + }
> +
> + /*
> + * Will need to compact to make this work.
> + * Tag just before the first leaf entry.
> + */
> + *compact = 1;
> + tagp = (__be16 *)blp - 1;
> +
> + /* Data object just before the first leaf entry. */
> + dup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
> +
> + /*
> + * If it's not free then the data will go where the
> + * leaf data starts now, if it works at all.
> + */
> + if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
> + if (be16_to_cpu(dup->length) + (be32_to_cpu(btp->stale) - 1) *
> + (uint)sizeof(*blp) < len)
> + dup = NULL;
> + } else if ((be32_to_cpu(btp->stale) - 1) * (uint)sizeof(*blp) < len)
> + dup = NULL;
> + else
> + dup = (xfs_dir2_data_unused_t *)blp;
> + goto out;
> + }
> +
> + /*
> + * no stale entries, so just use free space.
> + * Tag just before the first leaf entry.
> + */
> + tagp = (__be16 *)blp - 1;
Shouldn't tagp just be set before this if statement rather than inside of it
and outside of it? In both cases it's equated to blp-1 and the value of blp
doesn't change to the best of my knowledge, so it's confusing to have two
assignments.
-Phil
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 14/25] xfs: verify dir2 block format buffers
2012-10-25 6:34 ` [PATCH 14/25] xfs: verify dir2 block format buffers Dave Chinner
@ 2012-10-30 3:26 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 3:26 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:03PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Add a dir2 block format read verifier. To fully verify every block
> when read, call xfs_dir2_data_check() on them. Change
> xfs_dir2_data_check() to do runtime checking, convert ASSERT()
> checks to XFS_WANT_CORRUPTED_RETURN(), which will trigger an ASSERT
> failure on debug kernels, but on production kernels will dump an
> error to dmesg and return EFSCORRUPTED to the caller.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
Solid.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 25/25] xfs: add write verifiers to log recovery
2012-10-26 20:31 ` Dave Chinner
@ 2012-10-30 12:23 ` Christoph Hellwig
2012-10-30 22:08 ` Dave Chinner
0 siblings, 1 reply; 69+ messages in thread
From: Christoph Hellwig @ 2012-10-30 12:23 UTC (permalink / raw)
To: Dave Chinner; +Cc: Christoph Hellwig, xfs
> Remote attr buffers aren't logged - they are written sycnhronously
> during the transaction - so won't get found by this.
Oh right. Removing the synchronous writes for the remote attrs is
someting we should tackled on day as well.
> As for remote
> symlink buffers, yeah, that might be a problem. Ultimately, both of
> these buffer types are going to grow headers for CRCs, so this
> problem will go away. I'm not sure how to address this problem
> in the mean time short of putting the buffer content type into all
> the buf_log_format headers. Do you have any better ideas?
I can't think of a really good idea. But introducing user exploitable
issues in log recovery is something I'd avoid. I also hate having to
change the log format now if we're going to bump it soon again. Let
me look a bit more at the log recovery code if there's some way to
interfer the buffer type.
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 15/25] xfs: factor dir2 free block reading
2012-10-25 6:34 ` [PATCH 15/25] xfs: factor dir2 free block reading Dave Chinner
@ 2012-10-30 13:14 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 13:14 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:04PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Also factor out the updating of the free block when removing entries
> from leaf blocks, and add a verifier callback for reads.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_dir2_leaf.c | 3 +-
> fs/xfs/xfs_dir2_node.c | 218 +++++++++++++++++++++++++++++++-----------------
> fs/xfs/xfs_dir2_priv.h | 2 +
> 3 files changed, 143 insertions(+), 80 deletions(-)
Looks like good work to me.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 16/25] xfs: factor out dir2 data block reading
2012-10-25 6:34 ` [PATCH 16/25] xfs: factor out dir2 data " Dave Chinner
@ 2012-10-30 13:21 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 13:21 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:05PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> And add a verifier callback function while there.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_dir2_block.c | 3 +--
> fs/xfs/xfs_dir2_data.c | 32 ++++++++++++++++++++++++++++++++
> fs/xfs/xfs_dir2_leaf.c | 38 +++++++++++++++++---------------------
> fs/xfs/xfs_dir2_node.c | 8 ++++----
> fs/xfs/xfs_dir2_priv.h | 2 ++
> 5 files changed, 56 insertions(+), 27 deletions(-)
This one actually made a little light bulb go off. I'd been wondering
why you conjoined logical conditions in an assignment to block_ok
earlier, but split things out here. I see now that it's to explicitly
and unambiguosly separate out the order of evaluation.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 17/25] xfs: factor dir2 leaf read
2012-10-25 6:34 ` [PATCH 17/25] xfs: factor dir2 leaf read Dave Chinner
@ 2012-10-30 13:22 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 13:22 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:06PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_dir2_leaf.c | 73 ++++++++++++++++++++++++++++++++++++++++--------
> fs/xfs/xfs_dir2_node.c | 6 ++--
> fs/xfs/xfs_dir2_priv.h | 2 ++
> 3 files changed, 67 insertions(+), 14 deletions(-)
More of the same, essentially.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 18/25] xfs: factor and verify attr leaf reads
2012-10-25 6:34 ` [PATCH 18/25] xfs: factor and verify attr leaf reads Dave Chinner
@ 2012-10-30 13:26 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 13:26 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:07PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Some reads are not converted yet because it isn't obvious ahead of
> time what the format of the block is going to be. Need to determine
> how to tell if the first block in the tree is a node or leaf format
> block. That will be done in later patches.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
Keep on truckin'
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 19/25] xfs: add xfs_da_node verification
2012-10-25 6:34 ` [PATCH 19/25] xfs: add xfs_da_node verification Dave Chinner
@ 2012-10-30 13:30 ` Phil White
2012-10-30 22:23 ` Dave Chinner
0 siblings, 1 reply; 69+ messages in thread
From: Phil White @ 2012-10-30 13:30 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:08PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_attr.c | 22 ++++------
> fs/xfs/xfs_attr_leaf.c | 12 +++---
> fs/xfs/xfs_attr_leaf.h | 8 ++--
> fs/xfs/xfs_da_btree.c | 108 ++++++++++++++++++++++++++++++++++++------------
> fs/xfs/xfs_da_btree.h | 3 ++
> fs/xfs/xfs_dir2_leaf.c | 2 +-
> fs/xfs/xfs_dir2_priv.h | 1 +
> 7 files changed, 106 insertions(+), 50 deletions(-)
Reviewed-by: Phil White <pwhite@sgi.com>
One minor comment:
> diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
> index a46035b..e950192 100644
> --- a/fs/xfs/xfs_da_btree.c
> +++ b/fs/xfs/xfs_da_btree.c
> @@ -91,6 +91,67 @@ STATIC int xfs_da_blk_unlink(xfs_da_state_t *state,
> xfs_da_state_blk_t *save_blk);
> STATIC void xfs_da_state_kill_altpath(xfs_da_state_t *state);
>
> +static void
> +__xfs_da_node_verify(
> + struct xfs_buf *bp)
> +{
> + struct xfs_mount *mp = bp->b_target->bt_mount;
> + struct xfs_da_node_hdr *hdr = bp->b_addr;
> + int block_ok = 0;
> +
> + block_ok = hdr->info.magic == cpu_to_be16(XFS_DA_NODE_MAGIC);
> + block_ok |= hdr->level > 0;
> + block_ok |= hdr->count > 0;
This particular assignment seemed a little inconsistent, compared to
other usages. Functionally, it's fine though.
-Phil
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 20/25] xfs: Add verifiers to dir2 data readahead.
2012-10-25 6:34 ` [PATCH 20/25] xfs: Add verifiers to dir2 data readahead Dave Chinner
@ 2012-10-30 13:31 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 13:31 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:09PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_da_btree.c | 4 ++--
> fs/xfs/xfs_da_btree.h | 4 ++--
> fs/xfs/xfs_dir2_data.c | 13 ++++++++++++-
> fs/xfs/xfs_dir2_leaf.c | 11 +++++------
> fs/xfs/xfs_dir2_priv.h | 2 ++
> fs/xfs/xfs_file.c | 4 +++-
> 6 files changed, 26 insertions(+), 12 deletions(-)
This one's a slam dunk
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 21/25] xfs: add buffer pre-write callback
2012-10-25 6:34 ` [PATCH 21/25] xfs: add buffer pre-write callback Dave Chinner
2012-10-26 8:50 ` Christoph Hellwig
@ 2012-10-30 13:32 ` Phil White
1 sibling, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 13:32 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:10PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Add a callback to the buffer write path to enable verification of
> the buffer and CRC calculation prior to issuing the write to the
> underlying storage.
>
> If the callback function detects some kind of failure or error
> condition, it must mark the buffer with an error so that the caller
> can take appropriate action. In the case of xfs_buf_ioapply(), a
> corrupt metadta buffer willt rigger a shutdown of the filesystem,
> because something is clearly wrong and we can't allow corrupt
> metadata to be written to disk.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 22/25] xfs: add pre-write metadata buffer verifier callbacks
2012-10-25 6:34 ` [PATCH 22/25] xfs: add pre-write metadata buffer verifier callbacks Dave Chinner
@ 2012-10-30 13:34 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 13:34 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:11PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> These verifiers are essentially the same code as the read verifiers,
> but do not require ioend processing. Hence factor the read verifier
> functions and add a new write verifier wrapper that is used as the
> callback.
>
> This is done as one large patch for all verifiers rather than one
> patch per verifier as the change is largely mechanical. This
> includes hooking up the write verifier via the read verifier
> function.
>
> Hooking up the write verifier for buffers obtained via
> xfs_trans_get_buf() will be done in a separate patch as that touches
> code in many different places rather than just the verifier
> functions.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_alloc.c | 38 +++++++++++++++++++++++++++++++++-----
> fs/xfs/xfs_alloc_btree.c | 21 +++++++++++++++++----
> fs/xfs/xfs_attr_leaf.c | 19 +++++++++++++++++--
> fs/xfs/xfs_attr_leaf.h | 2 +-
> fs/xfs/xfs_bmap_btree.c | 21 +++++++++++++++++----
> fs/xfs/xfs_da_btree.c | 30 ++++++++++++++++++------------
> fs/xfs/xfs_dir2_block.c | 16 +++++++++++++++-
> fs/xfs/xfs_dir2_data.c | 19 +++++++++++++++++--
> fs/xfs/xfs_dir2_leaf.c | 31 ++++++++++++++++++++++++-------
> fs/xfs/xfs_dir2_node.c | 17 ++++++++++++++++-
> fs/xfs/xfs_dir2_priv.h | 2 +-
> fs/xfs/xfs_dquot.c | 22 ++++++++++++++++++----
> fs/xfs/xfs_ialloc.c | 17 ++++++++++++++++-
> fs/xfs/xfs_ialloc_btree.c | 19 ++++++++++++++++---
> fs/xfs/xfs_inode.c | 19 +++++++++++++++++--
> fs/xfs/xfs_inode.h | 2 +-
> fs/xfs/xfs_itable.c | 2 +-
> fs/xfs/xfs_mount.c | 19 +++++++++++++++++--
> 18 files changed, 262 insertions(+), 54 deletions(-)
Looks OK by me.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 23/25] xfs: connect up write verifiers to new buffers
2012-10-25 6:34 ` [PATCH 23/25] xfs: connect up write verifiers to new buffers Dave Chinner
@ 2012-10-30 13:39 ` Phil White
2012-10-30 22:34 ` Dave Chinner
0 siblings, 1 reply; 69+ messages in thread
From: Phil White @ 2012-10-30 13:39 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:12PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Metadata buffers that are read from disk have write verifiers
> already attached to them, but newly allocated buffers do not. Add
> appropriate write verifiers to all new metadata buffers.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_alloc.c | 6 +--
> fs/xfs/xfs_alloc.h | 2 +
> fs/xfs/xfs_alloc_btree.c | 1 +
> fs/xfs/xfs_attr_leaf.c | 4 +-
> fs/xfs/xfs_bmap.c | 2 +
> fs/xfs/xfs_bmap_btree.c | 3 +-
> fs/xfs/xfs_bmap_btree.h | 1 +
> fs/xfs/xfs_btree.c | 1 +
> fs/xfs/xfs_btree.h | 2 +
> fs/xfs/xfs_da_btree.c | 3 ++
> fs/xfs/xfs_dir2_block.c | 2 +
> fs/xfs/xfs_dir2_data.c | 11 +++--
> fs/xfs/xfs_dir2_leaf.c | 19 ++++++---
> fs/xfs/xfs_dir2_node.c | 24 +++++++----
> fs/xfs/xfs_dir2_priv.h | 2 +
> fs/xfs/xfs_dquot.c | 104 ++++++++++++++++++++++-----------------------
> fs/xfs/xfs_fsops.c | 7 ++-
> fs/xfs/xfs_ialloc.c | 5 ++-
> fs/xfs/xfs_ialloc.h | 4 +-
> fs/xfs/xfs_ialloc_btree.c | 1 +
> fs/xfs/xfs_inode.c | 14 +++++-
> fs/xfs/xfs_inode.h | 1 +
> fs/xfs/xfs_mount.c | 2 +-
> fs/xfs/xfs_mount.h | 1 +
> 24 files changed, 135 insertions(+), 87 deletions(-)
>
A few comments:
> diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c
> index bb96c55..5d56886 100644
> --- a/fs/xfs/xfs_attr_leaf.c
> +++ b/fs/xfs/xfs_attr_leaf.c
> @@ -923,7 +923,7 @@ xfs_attr_leaf_to_node(xfs_da_args_t *args)
> XFS_ATTR_FORK);
> if (error)
> goto out;
> - ASSERT(bp2 != NULL);
> + bp2->b_pre_io = bp1->b_pre_io;
> memcpy(bp2->b_addr, bp1->b_addr, XFS_LBSIZE(dp->i_mount));
> bp1 = NULL;
> xfs_trans_log_buf(args->trans, bp2, 0, XFS_LBSIZE(dp->i_mount) - 1);
> @@ -977,7 +977,7 @@ xfs_attr_leaf_create(
> XFS_ATTR_FORK);
> if (error)
> return(error);
> - ASSERT(bp != NULL);
> + bp->b_pre_io = xfs_attr_leaf_write_verify;
> leaf = bp->b_addr;
> memset((char *)leaf, 0, XFS_LBSIZE(dp->i_mount));
> hdr = &leaf->hdr;
I'm unclear as to why you're removing the asserts here. There must be
a reason that you think bp is guaranteed to be safe, but I haven't
grasped it here.
-Phil
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 24/25] xfs: convert buffer verifiers to an ops structure.
2012-10-25 6:34 ` [PATCH 24/25] xfs: convert buffer verifiers to an ops structure Dave Chinner
@ 2012-10-30 13:41 ` Phil White
0 siblings, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 13:41 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:13PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> To separate the verifiers from iodone functions and associate read
> and write verifiers at the same time, introduce a buffer verifier
> operations structure to the xfs_buf.
>
> This avoids the need for assigning the write verifier, clearing the
> iodone function and re-running ioend processing in the read
> verifier, and gets rid of the nasty "b_pre_io" name for the write
> verifier function pointer. If we ever need to, it will also be
> easier to add further content specific callbacks to a buffer with an
> ops structure in place.
>
> We also avoid needing to export verifier functions, instead we
> can simply export the ops structures for those that are needed
> outside the function they are defined in.
>
> This patch also fixes a directory block readahead verifier issue
> it exposed.
>
> This patch also adds ops callbacks to the inode/alloc btree blocks
> initialised by growfs. These will need more work before they will
> work with CRCs.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_ag.h | 4 +++
> fs/xfs/xfs_alloc.c | 26 +++++++++++--------
> fs/xfs/xfs_alloc.h | 2 +-
> fs/xfs/xfs_alloc_btree.c | 18 +++++++------
> fs/xfs/xfs_alloc_btree.h | 2 ++
> fs/xfs/xfs_attr_leaf.c | 19 +++++++-------
> fs/xfs/xfs_attr_leaf.h | 3 ++-
> fs/xfs/xfs_bmap.c | 22 ++++++++--------
> fs/xfs/xfs_bmap_btree.c | 20 +++++++-------
> fs/xfs/xfs_bmap_btree.h | 3 +--
> fs/xfs/xfs_btree.c | 26 +++++++++----------
> fs/xfs/xfs_btree.h | 9 +++----
> fs/xfs/xfs_buf.c | 63 ++++++++++++++++++++++++++-------------------
> fs/xfs/xfs_buf.h | 24 ++++++++++-------
> fs/xfs/xfs_da_btree.c | 28 ++++++++++----------
> fs/xfs/xfs_da_btree.h | 4 +--
> fs/xfs/xfs_dir2_block.c | 20 +++++++-------
> fs/xfs/xfs_dir2_data.c | 52 ++++++++++++++++++++++++++++++-------
> fs/xfs/xfs_dir2_leaf.c | 36 ++++++++++++++------------
> fs/xfs/xfs_dir2_node.c | 26 ++++++++++---------
> fs/xfs/xfs_dir2_priv.h | 10 ++++---
> fs/xfs/xfs_dquot.c | 16 +++++++-----
> fs/xfs/xfs_fsops.c | 26 ++++++++++++++++---
> fs/xfs/xfs_ialloc.c | 18 +++++++------
> fs/xfs/xfs_ialloc.h | 2 +-
> fs/xfs/xfs_ialloc_btree.c | 17 ++++++------
> fs/xfs/xfs_ialloc_btree.h | 2 ++
> fs/xfs/xfs_inode.c | 22 +++++++++-------
> fs/xfs/xfs_inode.h | 3 +--
> fs/xfs/xfs_itable.c | 2 +-
> fs/xfs/xfs_log_recover.c | 2 +-
> fs/xfs/xfs_mount.c | 35 +++++++++++++++----------
> fs/xfs/xfs_mount.h | 4 +--
> fs/xfs/xfs_trans.h | 6 ++---
> fs/xfs/xfs_trans_buf.c | 8 +++---
> 35 files changed, 346 insertions(+), 234 deletions(-)
More mechanical changes. Looks good.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 25/25] xfs: add write verifiers to log recovery
2012-10-25 6:34 ` [PATCH 25/25] xfs: add write verifiers to log recovery Dave Chinner
2012-10-26 8:54 ` Christoph Hellwig
@ 2012-10-30 13:44 ` Phil White
1 sibling, 0 replies; 69+ messages in thread
From: Phil White @ 2012-10-30 13:44 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Thu, Oct 25, 2012 at 05:34:14PM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
>
> Log recovery reads metadata, modifies it and rewrites it to disk.
> It is only practical to add write verifiers to metadata buffers
> because we do not know the type of the buffer prior to reading it
> from disk. Further, if it is an new bufer, the contents might not
> contain anything we can verify. Hence we only attempt to verify
> after the buffer changes have been replayed and we can peek at the
> buffer to find out what it contains to attached the correct
> verifier. This ensures that we don't introduce gross corruptions as
> a result of replaying transactions in the log.
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_alloc.c | 2 +-
> fs/xfs/xfs_alloc.h | 1 +
> fs/xfs/xfs_alloc_btree.c | 15 ++++---
> fs/xfs/xfs_da_btree.h | 1 +
> fs/xfs/xfs_dir2_leaf.c | 2 +-
> fs/xfs/xfs_dir2_node.c | 2 +-
> fs/xfs/xfs_dir2_priv.h | 3 ++
> fs/xfs/xfs_dquot.c | 17 +++++++-
> fs/xfs/xfs_dquot.h | 2 +
> fs/xfs/xfs_log_recover.c | 104 +++++++++++++++++++++++++++++++++++++++++++++-
> 10 files changed, 138 insertions(+), 11 deletions(-)
Generally, a good read. It kept me on the edge of my seat.
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 25/25] xfs: add write verifiers to log recovery
2012-10-30 12:23 ` Christoph Hellwig
@ 2012-10-30 22:08 ` Dave Chinner
2012-10-31 10:19 ` Christoph Hellwig
0 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-30 22:08 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: xfs
On Tue, Oct 30, 2012 at 08:23:31AM -0400, Christoph Hellwig wrote:
> > Remote attr buffers aren't logged - they are written sycnhronously
> > during the transaction - so won't get found by this.
>
> Oh right. Removing the synchronous writes for the remote attrs is
> someting we should tackled on day as well.
>
> > As for remote
> > symlink buffers, yeah, that might be a problem. Ultimately, both of
> > these buffer types are going to grow headers for CRCs, so this
> > problem will go away. I'm not sure how to address this problem
> > in the mean time short of putting the buffer content type into all
> > the buf_log_format headers. Do you have any better ideas?
>
> I can't think of a really good idea. But introducing user exploitable
> issues in log recovery is something I'd avoid. I also hate having to
> change the log format now if we're going to bump it soon again. Let
> me look a bit more at the log recovery code if there's some way to
> interfer the buffer type.
I couldn't find any short of magic number matches. The ones that can
be inferred (inode and dquot buffers) are done that way, but for
everything else they are anonymous buffers being recovered. I might
just drop this patch for now, and only re-introduce it for CRC
enabled filesystems when that is added.
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 07/25] xfs: verify AGI blocks as they are read from disk
2012-10-30 0:53 ` Phil White
@ 2012-10-30 22:13 ` Dave Chinner
0 siblings, 0 replies; 69+ messages in thread
From: Dave Chinner @ 2012-10-30 22:13 UTC (permalink / raw)
To: Phil White; +Cc: xfs
On Mon, Oct 29, 2012 at 05:53:22PM -0700, Phil White wrote:
> On Thu, Oct 25, 2012 at 05:33:56PM +1100, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@redhat.com>
> >
> > Add an AGI block verify callback function and pass it into the
> > buffer read functions. Remove the now redundant verification code
> > that is currently in use.
> >
> > Signed-off-by: Dave Chinner <dchinner@redhat.com>
> > Reviewed-by: Christoph Hellwig <hch@lst.de>
> > ---
> > fs/xfs/xfs_ialloc.c | 47 ++++++++++++++++++++++++++---------------------
> > 1 file changed, 26 insertions(+), 21 deletions(-)
> >
> > diff --git a/fs/xfs/xfs_ialloc.c b/fs/xfs/xfs_ialloc.c
> > index 7c944e1..9311ae5 100644
> > --- a/fs/xfs/xfs_ialloc.c
> > +++ b/fs/xfs/xfs_ialloc.c
> > @@ -1472,6 +1472,31 @@ xfs_check_agi_unlinked(
> > #define xfs_check_agi_unlinked(agi)
> > #endif
> >
> > +static void
> > +xfs_agi_read_verify(
> > + struct xfs_buf *bp)
> > +{
> > + struct xfs_mount *mp = bp->b_target->bt_mount;
> > + struct xfs_agi *agi = XFS_BUF_TO_AGI(bp);
> > + int agi_ok;
> > +
> > + /*
> > + * Validate the magic number of the agi block.
> > + */
> > + agi_ok = agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC) &&
> > + XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum)) &&
> > + be32_to_cpu(agi->agi_seqno) == bp->b_pag->pag_agno;
> > + if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI,
> > + XFS_RANDOM_IALLOC_READ_AGI))) {
> > + XFS_CORRUPTION_ERROR("xfs_read_agi", XFS_ERRLEVEL_LOW,
> > + mp, agi);
> > + xfs_buf_ioerror(bp, EFSCORRUPTED);
> > + }
> > + xfs_check_agi_unlinked(agi);
> > + bp->b_iodone = NULL;
> > + xfs_buf_ioend(bp, 0);
> > +}
> > +
>
> In like fashion, shouldn't this be XFS_CORRUPTION_ERROR("xfs_agi_read_verify",
> ...)? In principle, it might be called from somewhere else in the future.
It should be converted to __func__. I thought I caught most of them,
but I didn't. It's not immediately critical - if I need to repost
the series, I'll fix them.
Most of these messages get revamped when CRC checking is enabled,
anyway, because there are different errors and more useful
information that is worth reporting. hence I haven't been too
concerned about little things like this...
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 13/25] xfs: factor dir2 block read operations
2012-10-30 3:23 ` Phil White
@ 2012-10-30 22:16 ` Dave Chinner
0 siblings, 0 replies; 69+ messages in thread
From: Dave Chinner @ 2012-10-30 22:16 UTC (permalink / raw)
To: Phil White; +Cc: xfs
On Mon, Oct 29, 2012 at 08:23:09PM -0700, Phil White wrote:
> On Thu, Oct 25, 2012 at 05:34:02PM +1100, Dave Chinner wrote:
> > +static void
> > +xfs_dir2_block_need_space(
> > ...
> > + /*
> > + * If there are stale entries we'll use one for the leaf.
> > + */
> > + if (btp->stale) {
> > + if (be16_to_cpu(bf[0].length) >= len) {
> > + /*
> > + * The biggest entry enough to avoid compaction.
> > + */
> > + dup = (xfs_dir2_data_unused_t *)
> > + ((char *)hdr + be16_to_cpu(bf[0].offset));
> > + goto out;
> > + }
> > +
> > + /*
> > + * Will need to compact to make this work.
> > + * Tag just before the first leaf entry.
> > + */
> > + *compact = 1;
> > + tagp = (__be16 *)blp - 1;
> > +
> > + /* Data object just before the first leaf entry. */
> > + dup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
> > +
> > + /*
> > + * If it's not free then the data will go where the
> > + * leaf data starts now, if it works at all.
> > + */
> > + if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
> > + if (be16_to_cpu(dup->length) + (be32_to_cpu(btp->stale) - 1) *
> > + (uint)sizeof(*blp) < len)
> > + dup = NULL;
> > + } else if ((be32_to_cpu(btp->stale) - 1) * (uint)sizeof(*blp) < len)
> > + dup = NULL;
> > + else
> > + dup = (xfs_dir2_data_unused_t *)blp;
> > + goto out;
> > + }
> > +
> > + /*
> > + * no stale entries, so just use free space.
> > + * Tag just before the first leaf entry.
> > + */
> > + tagp = (__be16 *)blp - 1;
>
> Shouldn't tagp just be set before this if statement rather than inside of it
> and outside of it?
No, because there is a case where it isn't set.
In general, when factoring code it's not a good idea to change logic
because that's where bugs most commonly creep in. At some point in
the future this could probably do with a more robust cleanup (rather
than just factoring), but right now that's out-of-scope for what I'm
doing...
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 19/25] xfs: add xfs_da_node verification
2012-10-30 13:30 ` Phil White
@ 2012-10-30 22:23 ` Dave Chinner
2012-10-31 0:23 ` Phil White
0 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-30 22:23 UTC (permalink / raw)
To: Phil White; +Cc: xfs
On Tue, Oct 30, 2012 at 06:30:26AM -0700, Phil White wrote:
> On Thu, Oct 25, 2012 at 05:34:08PM +1100, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@redhat.com>
> >
> > Signed-off-by: Dave Chinner <dchinner@redhat.com>
> > ---
> > fs/xfs/xfs_attr.c | 22 ++++------
> > fs/xfs/xfs_attr_leaf.c | 12 +++---
> > fs/xfs/xfs_attr_leaf.h | 8 ++--
> > fs/xfs/xfs_da_btree.c | 108 ++++++++++++++++++++++++++++++++++++------------
> > fs/xfs/xfs_da_btree.h | 3 ++
> > fs/xfs/xfs_dir2_leaf.c | 2 +-
> > fs/xfs/xfs_dir2_priv.h | 1 +
> > 7 files changed, 106 insertions(+), 50 deletions(-)
>
> Reviewed-by: Phil White <pwhite@sgi.com>
>
> One minor comment:
>
> > diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
> > index a46035b..e950192 100644
> > --- a/fs/xfs/xfs_da_btree.c
> > +++ b/fs/xfs/xfs_da_btree.c
> > @@ -91,6 +91,67 @@ STATIC int xfs_da_blk_unlink(xfs_da_state_t *state,
> > xfs_da_state_blk_t *save_blk);
> > STATIC void xfs_da_state_kill_altpath(xfs_da_state_t *state);
> >
> > +static void
> > +__xfs_da_node_verify(
> > + struct xfs_buf *bp)
> > +{
> > + struct xfs_mount *mp = bp->b_target->bt_mount;
> > + struct xfs_da_node_hdr *hdr = bp->b_addr;
> > + int block_ok = 0;
> > +
> > + block_ok = hdr->info.magic == cpu_to_be16(XFS_DA_NODE_MAGIC);
> > + block_ok |= hdr->level > 0;
> > + block_ok |= hdr->count > 0;
>
> This particular assignment seemed a little inconsistent, compared to
> other usages. Functionally, it's fine though.
Actaully, it's wrong. If level or count are good, that will override
a bad magic number. I thought i caught all those thinko's I copied
around the place. Obviously not.
There's also another problem with this - endian swapping is missing.
Fixed patch is below.
-Dave
--
Dave Chinner
david@fromorbit.com
xfs: add xfs_da_node verification
From: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
---
fs/xfs/xfs_attr.c | 22 ++++------
fs/xfs/xfs_attr_leaf.c | 12 +++---
fs/xfs/xfs_attr_leaf.h | 8 ++--
fs/xfs/xfs_da_btree.c | 109 ++++++++++++++++++++++++++++++++++++------------
fs/xfs/xfs_da_btree.h | 3 ++
fs/xfs/xfs_dir2_leaf.c | 2 +-
fs/xfs/xfs_dir2_priv.h | 1 +
7 files changed, 107 insertions(+), 50 deletions(-)
diff --git a/fs/xfs/xfs_attr.c b/fs/xfs/xfs_attr.c
index 548e910..4b862ed 100644
--- a/fs/xfs/xfs_attr.c
+++ b/fs/xfs/xfs_attr.c
@@ -1688,10 +1688,10 @@ xfs_attr_refillstate(xfs_da_state_t *state)
ASSERT((path->active >= 0) && (path->active < XFS_DA_NODE_MAXDEPTH));
for (blk = path->blk, level = 0; level < path->active; blk++, level++) {
if (blk->disk_blkno) {
- error = xfs_da_read_buf(state->args->trans,
+ error = xfs_da_node_read(state->args->trans,
state->args->dp,
blk->blkno, blk->disk_blkno,
- &blk->bp, XFS_ATTR_FORK, NULL);
+ &blk->bp, XFS_ATTR_FORK);
if (error)
return(error);
} else {
@@ -1707,10 +1707,10 @@ xfs_attr_refillstate(xfs_da_state_t *state)
ASSERT((path->active >= 0) && (path->active < XFS_DA_NODE_MAXDEPTH));
for (blk = path->blk, level = 0; level < path->active; blk++, level++) {
if (blk->disk_blkno) {
- error = xfs_da_read_buf(state->args->trans,
+ error = xfs_da_node_read(state->args->trans,
state->args->dp,
blk->blkno, blk->disk_blkno,
- &blk->bp, XFS_ATTR_FORK, NULL);
+ &blk->bp, XFS_ATTR_FORK);
if (error)
return(error);
} else {
@@ -1795,8 +1795,8 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
*/
bp = NULL;
if (cursor->blkno > 0) {
- error = xfs_da_read_buf(NULL, context->dp, cursor->blkno, -1,
- &bp, XFS_ATTR_FORK, NULL);
+ error = xfs_da_node_read(NULL, context->dp, cursor->blkno, -1,
+ &bp, XFS_ATTR_FORK);
if ((error != 0) && (error != EFSCORRUPTED))
return(error);
if (bp) {
@@ -1837,17 +1837,11 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
if (bp == NULL) {
cursor->blkno = 0;
for (;;) {
- error = xfs_da_read_buf(NULL, context->dp,
+ error = xfs_da_node_read(NULL, context->dp,
cursor->blkno, -1, &bp,
- XFS_ATTR_FORK, NULL);
+ XFS_ATTR_FORK);
if (error)
return(error);
- if (unlikely(bp == NULL)) {
- XFS_ERROR_REPORT("xfs_attr_node_list(2)",
- XFS_ERRLEVEL_LOW,
- context->dp->i_mount);
- return(XFS_ERROR(EFSCORRUPTED));
- }
node = bp->b_addr;
if (node->hdr.info.magic ==
cpu_to_be16(XFS_ATTR_LEAF_MAGIC))
diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c
index 7891d06..5ba92eb 100644
--- a/fs/xfs/xfs_attr_leaf.c
+++ b/fs/xfs/xfs_attr_leaf.c
@@ -87,7 +87,7 @@ STATIC void xfs_attr_leaf_moveents(xfs_attr_leafblock_t *src_leaf,
xfs_mount_t *mp);
STATIC int xfs_attr_leaf_entsize(xfs_attr_leafblock_t *leaf, int index);
-static void
+void
xfs_attr_leaf_verify(
struct xfs_buf *bp)
{
@@ -2742,7 +2742,7 @@ xfs_attr_root_inactive(xfs_trans_t **trans, xfs_inode_t *dp)
* the extents in reverse order the extent containing
* block 0 must still be there.
*/
- error = xfs_da_read_buf(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK, NULL);
+ error = xfs_da_node_read(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK);
if (error)
return(error);
blkno = XFS_BUF_ADDR(bp);
@@ -2827,8 +2827,8 @@ xfs_attr_node_inactive(
* traversal of the tree so we may deal with many blocks
* before we come back to this one.
*/
- error = xfs_da_read_buf(*trans, dp, child_fsb, -2, &child_bp,
- XFS_ATTR_FORK, NULL);
+ error = xfs_da_node_read(*trans, dp, child_fsb, -2, &child_bp,
+ XFS_ATTR_FORK);
if (error)
return(error);
if (child_bp) {
@@ -2868,8 +2868,8 @@ xfs_attr_node_inactive(
* child block number.
*/
if ((i+1) < count) {
- error = xfs_da_read_buf(*trans, dp, 0, parent_blkno,
- &bp, XFS_ATTR_FORK, NULL);
+ error = xfs_da_node_read(*trans, dp, 0, parent_blkno,
+ &bp, XFS_ATTR_FORK);
if (error)
return(error);
child_fsb = be32_to_cpu(node->btree[i+1].before);
diff --git a/fs/xfs/xfs_attr_leaf.h b/fs/xfs/xfs_attr_leaf.h
index 8f7ab98..098e9a5 100644
--- a/fs/xfs/xfs_attr_leaf.h
+++ b/fs/xfs/xfs_attr_leaf.h
@@ -227,9 +227,6 @@ int xfs_attr_leaf_to_shortform(struct xfs_buf *bp,
int xfs_attr_leaf_clearflag(struct xfs_da_args *args);
int xfs_attr_leaf_setflag(struct xfs_da_args *args);
int xfs_attr_leaf_flipflags(xfs_da_args_t *args);
-int xfs_attr_leaf_read(struct xfs_trans *tp, struct xfs_inode *dp,
- xfs_dablk_t bno, xfs_daddr_t mappedbno,
- struct xfs_buf **bpp);
/*
* Routines used for growing the Btree.
@@ -264,4 +261,9 @@ int xfs_attr_leaf_order(struct xfs_buf *leaf1_bp,
struct xfs_buf *leaf2_bp);
int xfs_attr_leaf_newentsize(int namelen, int valuelen, int blocksize,
int *local);
+int xfs_attr_leaf_read(struct xfs_trans *tp, struct xfs_inode *dp,
+ xfs_dablk_t bno, xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp);
+void xfs_attr_leaf_verify(struct xfs_buf *bp);
+
#endif /* __XFS_ATTR_LEAF_H__ */
diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
index a46035b..9895faf 100644
--- a/fs/xfs/xfs_da_btree.c
+++ b/fs/xfs/xfs_da_btree.c
@@ -91,6 +91,68 @@ STATIC int xfs_da_blk_unlink(xfs_da_state_t *state,
xfs_da_state_blk_t *save_blk);
STATIC void xfs_da_state_kill_altpath(xfs_da_state_t *state);
+static void
+__xfs_da_node_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_da_node_hdr *hdr = bp->b_addr;
+ int block_ok = 0;
+
+ block_ok = hdr->info.magic == cpu_to_be16(XFS_DA_NODE_MAGIC);
+ block_ok = block_ok &&
+ be16_to_cpu(hdr->level) > 0 &&
+ be16_to_cpu(hdr->count) > 0 ;
+ if (!block_ok) {
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+ }
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
+static void
+xfs_da_node_verify(
+ struct xfs_buf *bp)
+{
+ struct xfs_mount *mp = bp->b_target->bt_mount;
+ struct xfs_da_blkinfo *info = bp->b_addr;
+
+ switch (be16_to_cpu(info->magic)) {
+ case XFS_DA_NODE_MAGIC:
+ __xfs_da_node_verify(bp);
+ return;
+ case XFS_ATTR_LEAF_MAGIC:
+ xfs_attr_leaf_verify(bp);
+ return;
+ case XFS_DIR2_LEAFN_MAGIC:
+ xfs_dir2_leafn_verify(bp);
+ return;
+ default:
+ break;
+ }
+
+ XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, info);
+ xfs_buf_ioerror(bp, EFSCORRUPTED);
+
+ bp->b_iodone = NULL;
+ xfs_buf_ioend(bp, 0);
+}
+
+int
+xfs_da_node_read(
+ struct xfs_trans *tp,
+ struct xfs_inode *dp,
+ xfs_dablk_t bno,
+ xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp,
+ int which_fork)
+{
+ return xfs_da_read_buf(tp, dp, bno, mappedbno, bpp,
+ which_fork, xfs_da_node_verify);
+}
+
/*========================================================================
* Routines used for growing the Btree.
*========================================================================*/
@@ -746,8 +808,8 @@ xfs_da_root_join(xfs_da_state_t *state, xfs_da_state_blk_t *root_blk)
*/
child = be32_to_cpu(oldroot->btree[0].before);
ASSERT(child != 0);
- error = xfs_da_read_buf(args->trans, args->dp, child, -1, &bp,
- args->whichfork, NULL);
+ error = xfs_da_node_read(args->trans, args->dp, child, -1, &bp,
+ args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -835,9 +897,8 @@ xfs_da_node_toosmall(xfs_da_state_t *state, int *action)
blkno = be32_to_cpu(info->back);
if (blkno == 0)
continue;
- error = xfs_da_read_buf(state->args->trans, state->args->dp,
- blkno, -1, &bp, state->args->whichfork,
- NULL);
+ error = xfs_da_node_read(state->args->trans, state->args->dp,
+ blkno, -1, &bp, state->args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1080,8 +1141,8 @@ xfs_da_node_lookup_int(xfs_da_state_t *state, int *result)
* Read the next node down in the tree.
*/
blk->blkno = blkno;
- error = xfs_da_read_buf(args->trans, args->dp, blkno,
- -1, &blk->bp, args->whichfork, NULL);
+ error = xfs_da_node_read(args->trans, args->dp, blkno,
+ -1, &blk->bp, args->whichfork);
if (error) {
blk->blkno = 0;
state->path.active--;
@@ -1242,9 +1303,9 @@ xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
new_info->forw = cpu_to_be32(old_blk->blkno);
new_info->back = old_info->back;
if (old_info->back) {
- error = xfs_da_read_buf(args->trans, args->dp,
+ error = xfs_da_node_read(args->trans, args->dp,
be32_to_cpu(old_info->back),
- -1, &bp, args->whichfork, NULL);
+ -1, &bp, args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1263,9 +1324,9 @@ xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
new_info->forw = old_info->forw;
new_info->back = cpu_to_be32(old_blk->blkno);
if (old_info->forw) {
- error = xfs_da_read_buf(args->trans, args->dp,
+ error = xfs_da_node_read(args->trans, args->dp,
be32_to_cpu(old_info->forw),
- -1, &bp, args->whichfork, NULL);
+ -1, &bp, args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1363,9 +1424,9 @@ xfs_da_blk_unlink(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
trace_xfs_da_unlink_back(args);
save_info->back = drop_info->back;
if (drop_info->back) {
- error = xfs_da_read_buf(args->trans, args->dp,
+ error = xfs_da_node_read(args->trans, args->dp,
be32_to_cpu(drop_info->back),
- -1, &bp, args->whichfork, NULL);
+ -1, &bp, args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1380,9 +1441,9 @@ xfs_da_blk_unlink(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
trace_xfs_da_unlink_forward(args);
save_info->forw = drop_info->forw;
if (drop_info->forw) {
- error = xfs_da_read_buf(args->trans, args->dp,
+ error = xfs_da_node_read(args->trans, args->dp,
be32_to_cpu(drop_info->forw),
- -1, &bp, args->whichfork, NULL);
+ -1, &bp, args->whichfork);
if (error)
return(error);
ASSERT(bp != NULL);
@@ -1464,8 +1525,8 @@ xfs_da_path_shift(xfs_da_state_t *state, xfs_da_state_path_t *path,
* Read the next child block.
*/
blk->blkno = blkno;
- error = xfs_da_read_buf(args->trans, args->dp, blkno, -1,
- &blk->bp, args->whichfork, NULL);
+ error = xfs_da_node_read(args->trans, args->dp, blkno, -1,
+ &blk->bp, args->whichfork);
if (error)
return(error);
ASSERT(blk->bp != NULL);
@@ -1728,7 +1789,7 @@ xfs_da_swap_lastblock(
* Read the last block in the btree space.
*/
last_blkno = (xfs_dablk_t)lastoff - mp->m_dirblkfsbs;
- error = xfs_da_read_buf(tp, ip, last_blkno, -1, &last_buf, w, NULL);
+ error = xfs_da_node_read(tp, ip, last_blkno, -1, &last_buf, w);
if (error)
return error;
/*
@@ -1755,8 +1816,7 @@ xfs_da_swap_lastblock(
* If the moved block has a left sibling, fix up the pointers.
*/
if ((sib_blkno = be32_to_cpu(dead_info->back))) {
- error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w,
- NULL);
+ error = xfs_da_node_read(tp, ip, sib_blkno, -1, &sib_buf, w);
if (error)
goto done;
sib_info = sib_buf->b_addr;
@@ -1778,8 +1838,7 @@ xfs_da_swap_lastblock(
* If the moved block has a right sibling, fix up the pointers.
*/
if ((sib_blkno = be32_to_cpu(dead_info->forw))) {
- error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w,
- NULL);
+ error = xfs_da_node_read(tp, ip, sib_blkno, -1, &sib_buf, w);
if (error)
goto done;
sib_info = sib_buf->b_addr;
@@ -1803,8 +1862,7 @@ xfs_da_swap_lastblock(
* Walk down the tree looking for the parent of the moved block.
*/
for (;;) {
- error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w,
- NULL);
+ error = xfs_da_node_read(tp, ip, par_blkno, -1, &par_buf, w);
if (error)
goto done;
par_node = par_buf->b_addr;
@@ -1855,8 +1913,7 @@ xfs_da_swap_lastblock(
error = XFS_ERROR(EFSCORRUPTED);
goto done;
}
- error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w,
- NULL);
+ error = xfs_da_node_read(tp, ip, par_blkno, -1, &par_buf, w);
if (error)
goto done;
par_node = par_buf->b_addr;
diff --git a/fs/xfs/xfs_da_btree.h b/fs/xfs/xfs_da_btree.h
index bf8bfaa..2d1bec4 100644
--- a/fs/xfs/xfs_da_btree.h
+++ b/fs/xfs/xfs_da_btree.h
@@ -213,6 +213,9 @@ int xfs_da_path_shift(xfs_da_state_t *state, xfs_da_state_path_t *path,
*/
int xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
xfs_da_state_blk_t *new_blk);
+int xfs_da_node_read(struct xfs_trans *tp, struct xfs_inode *dp,
+ xfs_dablk_t bno, xfs_daddr_t mappedbno,
+ struct xfs_buf **bpp, int which_fork);
/*
* Utility routines.
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 97408e3..67cc21c 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -74,7 +74,7 @@ xfs_dir2_leaf1_verify(
xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
}
-static void
+void
xfs_dir2_leafn_verify(
struct xfs_buf *bp)
{
diff --git a/fs/xfs/xfs_dir2_priv.h b/fs/xfs/xfs_dir2_priv.h
index ecf75d9..1f42e81 100644
--- a/fs/xfs/xfs_dir2_priv.h
+++ b/fs/xfs/xfs_dir2_priv.h
@@ -70,6 +70,7 @@ extern void xfs_dir2_data_use_free(struct xfs_trans *tp, struct xfs_buf *bp,
xfs_dir2_data_aoff_t len, int *needlogp, int *needscanp);
/* xfs_dir2_leaf.c */
+extern void xfs_dir2_leafn_verify(struct xfs_buf *bp);
extern int xfs_dir2_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,
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply related [flat|nested] 69+ messages in thread
* Re: [PATCH 21/25] xfs: add buffer pre-write callback
2012-10-26 8:50 ` Christoph Hellwig
@ 2012-10-30 22:30 ` Dave Chinner
2012-10-31 10:20 ` Christoph Hellwig
0 siblings, 1 reply; 69+ messages in thread
From: Dave Chinner @ 2012-10-30 22:30 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: xfs
On Fri, Oct 26, 2012 at 04:50:14AM -0400, Christoph Hellwig wrote:
> > /*
> > + * run the pre-io callback function if it exists. If this function
> > + * fails it will mark the buffer with an error and the IO should
> > + * not be dispatched.
> > + */
> > + if (bp->b_pre_io) {
> > + bp->b_pre_io(bp);
> > + if (bp->b_error) {
>
> Wouldn't it be a cleaner calling convention to return the erro from the
> callback?
Perhaps. I just wrote it in a manner consistent with the iodone
function where errors are returned in bp->b_error. Other functions
pass buffer errors like this, too - xfs_buf_ioapply_map(),
xfs_buf_read_map(), and _xfs_buf_ioapply() - so it's not unusual,
really..
I can change it, but that involves changing every callback function
as well and I don't see that as really necessary. i.e. they call
xfs_buf_ioerror() already, so do we really need to have them return
bp->b_error as well?
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 23/25] xfs: connect up write verifiers to new buffers
2012-10-30 13:39 ` Phil White
@ 2012-10-30 22:34 ` Dave Chinner
0 siblings, 0 replies; 69+ messages in thread
From: Dave Chinner @ 2012-10-30 22:34 UTC (permalink / raw)
To: Phil White; +Cc: xfs
On Tue, Oct 30, 2012 at 06:39:38AM -0700, Phil White wrote:
> On Thu, Oct 25, 2012 at 05:34:12PM +1100, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@redhat.com>
> >
> > Metadata buffers that are read from disk have write verifiers
> > already attached to them, but newly allocated buffers do not. Add
> > appropriate write verifiers to all new metadata buffers.
> >
> > Signed-off-by: Dave Chinner <dchinner@redhat.com>
> > ---
> > fs/xfs/xfs_alloc.c | 6 +--
> > fs/xfs/xfs_alloc.h | 2 +
> > fs/xfs/xfs_alloc_btree.c | 1 +
> > fs/xfs/xfs_attr_leaf.c | 4 +-
> > fs/xfs/xfs_bmap.c | 2 +
> > fs/xfs/xfs_bmap_btree.c | 3 +-
> > fs/xfs/xfs_bmap_btree.h | 1 +
> > fs/xfs/xfs_btree.c | 1 +
> > fs/xfs/xfs_btree.h | 2 +
> > fs/xfs/xfs_da_btree.c | 3 ++
> > fs/xfs/xfs_dir2_block.c | 2 +
> > fs/xfs/xfs_dir2_data.c | 11 +++--
> > fs/xfs/xfs_dir2_leaf.c | 19 ++++++---
> > fs/xfs/xfs_dir2_node.c | 24 +++++++----
> > fs/xfs/xfs_dir2_priv.h | 2 +
> > fs/xfs/xfs_dquot.c | 104 ++++++++++++++++++++++-----------------------
> > fs/xfs/xfs_fsops.c | 7 ++-
> > fs/xfs/xfs_ialloc.c | 5 ++-
> > fs/xfs/xfs_ialloc.h | 4 +-
> > fs/xfs/xfs_ialloc_btree.c | 1 +
> > fs/xfs/xfs_inode.c | 14 +++++-
> > fs/xfs/xfs_inode.h | 1 +
> > fs/xfs/xfs_mount.c | 2 +-
> > fs/xfs/xfs_mount.h | 1 +
> > 24 files changed, 135 insertions(+), 87 deletions(-)
> >
>
> A few comments:
>
> > diff --git a/fs/xfs/xfs_attr_leaf.c b/fs/xfs/xfs_attr_leaf.c
> > index bb96c55..5d56886 100644
> > --- a/fs/xfs/xfs_attr_leaf.c
> > +++ b/fs/xfs/xfs_attr_leaf.c
> > @@ -923,7 +923,7 @@ xfs_attr_leaf_to_node(xfs_da_args_t *args)
> > XFS_ATTR_FORK);
> > if (error)
> > goto out;
> > - ASSERT(bp2 != NULL);
> > + bp2->b_pre_io = bp1->b_pre_io;
> > memcpy(bp2->b_addr, bp1->b_addr, XFS_LBSIZE(dp->i_mount));
> > bp1 = NULL;
> > xfs_trans_log_buf(args->trans, bp2, 0, XFS_LBSIZE(dp->i_mount) - 1);
> > @@ -977,7 +977,7 @@ xfs_attr_leaf_create(
> > XFS_ATTR_FORK);
> > if (error)
> > return(error);
> > - ASSERT(bp != NULL);
> > + bp->b_pre_io = xfs_attr_leaf_write_verify;
> > leaf = bp->b_addr;
> > memset((char *)leaf, 0, XFS_LBSIZE(dp->i_mount));
> > hdr = &leaf->hdr;
>
> I'm unclear as to why you're removing the asserts here. There must be
> a reason that you think bp is guaranteed to be safe, but I haven't
> grasped it here.
If bp is NULL then the code is going to oops immediately, anyway.
Hence the assert is redundant. Further, the function that returns
bp either gives us a valid buffer or returns an error, so logically
the assert is also redundant from that perspective.
Hence I killed them to clean up the code a little.
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 19/25] xfs: add xfs_da_node verification
2012-10-30 22:23 ` Dave Chinner
@ 2012-10-31 0:23 ` Phil White
2012-10-31 0:50 ` Dave Chinner
0 siblings, 1 reply; 69+ messages in thread
From: Phil White @ 2012-10-31 0:23 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs, Phil White
On Wed, Oct 31, 2012 at 09:23:32AM +1100, Dave Chinner wrote:
> There's also another problem with this - endian swapping is missing.
Endian swapping doesn't matter. include/linux/types.h defines __be16 as
a __u16 and 0 is 0 is 0, no matter which order you put the bytes.
Doesn't hurt to make it clear though.
> xfs: add xfs_da_node verification
>
> From: Dave Chinner <dchinner@redhat.com>
>
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
> fs/xfs/xfs_attr.c | 22 ++++------
> fs/xfs/xfs_attr_leaf.c | 12 +++---
> fs/xfs/xfs_attr_leaf.h | 8 ++--
> fs/xfs/xfs_da_btree.c | 109 ++++++++++++++++++++++++++++++++++++------------
> fs/xfs/xfs_da_btree.h | 3 ++
> fs/xfs/xfs_dir2_leaf.c | 2 +-
> fs/xfs/xfs_dir2_priv.h | 1 +
> 7 files changed, 107 insertions(+), 50 deletions(-)
I'm a little surprised (and dismayed) that it passed xfstests with that.
Presumably, it never ran into a case where level or count were > 0 on
an invalid xfs_da_node.
Anyway...
Reviewed-by: Phil White <pwhite@sgi.com>
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 19/25] xfs: add xfs_da_node verification
2012-10-31 0:23 ` Phil White
@ 2012-10-31 0:50 ` Dave Chinner
0 siblings, 0 replies; 69+ messages in thread
From: Dave Chinner @ 2012-10-31 0:50 UTC (permalink / raw)
To: Phil White; +Cc: xfs
On Tue, Oct 30, 2012 at 05:23:47PM -0700, Phil White wrote:
> On Wed, Oct 31, 2012 at 09:23:32AM +1100, Dave Chinner wrote:
> > There's also another problem with this - endian swapping is missing.
>
> Endian swapping doesn't matter. include/linux/types.h defines __be16 as
> a __u16 and 0 is 0 is 0, no matter which order you put the bytes.
Except that the compiler based endian checks throw a error because
it is wrong ;)
(make C=2 CF="-D__CHECK_ENDIAN__" fs/xfs/xfs.ko)
> Doesn't hurt to make it clear though.
>
> > xfs: add xfs_da_node verification
> >
> > From: Dave Chinner <dchinner@redhat.com>
> >
> > Signed-off-by: Dave Chinner <dchinner@redhat.com>
> > ---
> > fs/xfs/xfs_attr.c | 22 ++++------
> > fs/xfs/xfs_attr_leaf.c | 12 +++---
> > fs/xfs/xfs_attr_leaf.h | 8 ++--
> > fs/xfs/xfs_da_btree.c | 109 ++++++++++++++++++++++++++++++++++++------------
> > fs/xfs/xfs_da_btree.h | 3 ++
> > fs/xfs/xfs_dir2_leaf.c | 2 +-
> > fs/xfs/xfs_dir2_priv.h | 1 +
> > 7 files changed, 107 insertions(+), 50 deletions(-)
>
> I'm a little surprised (and dismayed) that it passed xfstests with that.
> Presumably, it never ran into a case where level or count were > 0 on
> an invalid xfs_da_node.
Sure, xfstests doesn't usually trigger directory corruption, and so
it never would have had a chance to fail the verification
incorrectly. More than anything, what we are testing with xfstests
is that the verification doesn't introduce regressions, not that the
verification is 100% correct. Review is what catches the thinkos and
typos (as it has in this case), so I think the process is working as
it should. ;)
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 25/25] xfs: add write verifiers to log recovery
2012-10-30 22:08 ` Dave Chinner
@ 2012-10-31 10:19 ` Christoph Hellwig
0 siblings, 0 replies; 69+ messages in thread
From: Christoph Hellwig @ 2012-10-31 10:19 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Wed, Oct 31, 2012 at 09:08:55AM +1100, Dave Chinner wrote:
> I couldn't find any short of magic number matches. The ones that can
> be inferred (inode and dquot buffers) are done that way, but for
> everything else they are anonymous buffers being recovered. I might
> just drop this patch for now, and only re-introduce it for CRC
> enabled filesystems when that is added.
I fear that's the only reasonable way to go. Otoh we were planning
to bump the log version at some point anyway, at which point we could
enable the new log formats even without CRCs.
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
* Re: [PATCH 21/25] xfs: add buffer pre-write callback
2012-10-30 22:30 ` Dave Chinner
@ 2012-10-31 10:20 ` Christoph Hellwig
0 siblings, 0 replies; 69+ messages in thread
From: Christoph Hellwig @ 2012-10-31 10:20 UTC (permalink / raw)
To: Dave Chinner; +Cc: xfs
On Wed, Oct 31, 2012 at 09:30:15AM +1100, Dave Chinner wrote:
> Perhaps. I just wrote it in a manner consistent with the iodone
> function where errors are returned in bp->b_error. Other functions
> pass buffer errors like this, too - xfs_buf_ioapply_map(),
> xfs_buf_read_map(), and _xfs_buf_ioapply() - so it's not unusual,
> really..
>
> I can change it, but that involves changing every callback function
> as well and I don't see that as really necessary. i.e. they call
> xfs_buf_ioerror() already, so do we really need to have them return
> bp->b_error as well?
Let's keep it as is for now to make forward progress, we can still
figure out later if doing it differently is cleaner.
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
^ permalink raw reply [flat|nested] 69+ messages in thread
end of thread, other threads:[~2012-10-31 10:18 UTC | newest]
Thread overview: 69+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2012-10-25 6:33 [PATCH 00/25, V3] xfs: metadata buffer verifiers Dave Chinner
2012-10-25 6:33 ` [PATCH 01/25] xfs: growfs: don't read garbage for new secondary superblocks Dave Chinner
2012-10-30 0:17 ` Phil White
2012-10-25 6:33 ` [PATCH 02/25] xfs: invalidate allocbt blocks moved to the free list Dave Chinner
2012-10-26 8:47 ` Christoph Hellwig
2012-10-30 0:22 ` Phil White
2012-10-25 6:33 ` [PATCH 03/25] xfs: make buffer read verication an IO completion function Dave Chinner
2012-10-30 0:29 ` Phil White
2012-10-30 0:45 ` Dave Chinner
2012-10-30 0:55 ` Phil White
2012-10-25 6:33 ` [PATCH 04/25] xfs: uncached buffer reads need to return an error Dave Chinner
2012-10-26 8:48 ` Christoph Hellwig
2012-10-30 0:36 ` Phil White
2012-10-25 6:33 ` [PATCH 05/25] xfs: verify superblocks as they are read from disk Dave Chinner
2012-10-30 0:48 ` Phil White
2012-10-25 6:33 ` [PATCH 06/25] xfs: verify AGF blocks " Dave Chinner
2012-10-30 0:51 ` Phil White
2012-10-25 6:33 ` [PATCH 07/25] xfs: verify AGI " Dave Chinner
2012-10-30 0:53 ` Phil White
2012-10-30 22:13 ` Dave Chinner
2012-10-25 6:33 ` [PATCH 08/25] xfs: verify AGFL " Dave Chinner
2012-10-30 1:00 ` Phil White
2012-10-25 6:33 ` [PATCH 09/25] xfs: verify inode buffers " Dave Chinner
2012-10-30 1:06 ` Phil White
2012-10-25 6:33 ` [PATCH 10/25] xfs: verify btree blocks " Dave Chinner
2012-10-30 1:14 ` Phil White
2012-10-25 6:34 ` [PATCH 11/25] xfs: verify dquot " Dave Chinner
2012-10-30 1:36 ` Phil White
2012-10-25 6:34 ` [PATCH 12/25] xfs: add verifier callback to directory read code Dave Chinner
2012-10-30 3:15 ` Phil White
2012-10-25 6:34 ` [PATCH 13/25] xfs: factor dir2 block read operations Dave Chinner
2012-10-30 3:23 ` Phil White
2012-10-30 22:16 ` Dave Chinner
2012-10-25 6:34 ` [PATCH 14/25] xfs: verify dir2 block format buffers Dave Chinner
2012-10-30 3:26 ` Phil White
2012-10-25 6:34 ` [PATCH 15/25] xfs: factor dir2 free block reading Dave Chinner
2012-10-30 13:14 ` Phil White
2012-10-25 6:34 ` [PATCH 16/25] xfs: factor out dir2 data " Dave Chinner
2012-10-30 13:21 ` Phil White
2012-10-25 6:34 ` [PATCH 17/25] xfs: factor dir2 leaf read Dave Chinner
2012-10-30 13:22 ` Phil White
2012-10-25 6:34 ` [PATCH 18/25] xfs: factor and verify attr leaf reads Dave Chinner
2012-10-30 13:26 ` Phil White
2012-10-25 6:34 ` [PATCH 19/25] xfs: add xfs_da_node verification Dave Chinner
2012-10-30 13:30 ` Phil White
2012-10-30 22:23 ` Dave Chinner
2012-10-31 0:23 ` Phil White
2012-10-31 0:50 ` Dave Chinner
2012-10-25 6:34 ` [PATCH 20/25] xfs: Add verifiers to dir2 data readahead Dave Chinner
2012-10-30 13:31 ` Phil White
2012-10-25 6:34 ` [PATCH 21/25] xfs: add buffer pre-write callback Dave Chinner
2012-10-26 8:50 ` Christoph Hellwig
2012-10-30 22:30 ` Dave Chinner
2012-10-31 10:20 ` Christoph Hellwig
2012-10-30 13:32 ` Phil White
2012-10-25 6:34 ` [PATCH 22/25] xfs: add pre-write metadata buffer verifier callbacks Dave Chinner
2012-10-30 13:34 ` Phil White
2012-10-25 6:34 ` [PATCH 23/25] xfs: connect up write verifiers to new buffers Dave Chinner
2012-10-30 13:39 ` Phil White
2012-10-30 22:34 ` Dave Chinner
2012-10-25 6:34 ` [PATCH 24/25] xfs: convert buffer verifiers to an ops structure Dave Chinner
2012-10-30 13:41 ` Phil White
2012-10-25 6:34 ` [PATCH 25/25] xfs: add write verifiers to log recovery Dave Chinner
2012-10-26 8:54 ` Christoph Hellwig
2012-10-26 20:31 ` Dave Chinner
2012-10-30 12:23 ` Christoph Hellwig
2012-10-30 22:08 ` Dave Chinner
2012-10-31 10:19 ` Christoph Hellwig
2012-10-30 13:44 ` Phil White
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox