The code to check inobt records against inode clusters is a mess of
poorly named variables and unnecessary parameters. Clean the
unnecessary inode number parameters out of _check_cluster_freemask in
favor of computing them inside the function instead of making the caller
do it. In xchk_iallocbt_check_cluster, rename the variables to make it
more obvious just what chunk_ino and cluster_ino represent.
Add a tracepoint to make it easier to track each inode cluster as we
scrub it.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Hoist the inode cluster checks out of the inobt record check loop into
a separate function in preparation for refactoring of that loop. No
functional changes here; that's in the next patch.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
On a big block filesystem, there may be multiple inobt records covering
a single inode cluster. These records obviously won't be aligned to
cluster alignment rules, and they must cover the entire cluster. Teach
scrub to check for these things.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
In xchk_iallocbt_rec, check the alignment of ir_startino by converting
the inode cluster block alignment into units of inodes instead of the
other way around (converting ir_startino to blocks). This prevents us
from tripping over off-by-one errors in ir_startino which are obscured
by the inode -> block conversion.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Make sure we never check more than XFS_INODES_PER_CHUNK inodes for any
given inobt record since there can be more than one inobt record mapped
to an inode cluster.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
In xrep_findroot_block, we work out the btree type and correctness of a
given block by calling different btree verifiers on root block
candidates. However, we leave the NULL b_ops while ->verify_read
validates the block, which means that if the verifier calls
xfs_buf_verifier_error it'll crash on the null b_ops. Fix it to set
b_ops before calling the verifier and unsetting it if the verifier
fails.
Furthermore, improve the documentation around xfs_buf_ensure_ops, which
is the function that is responsible for cleaning up the b_ops state of
buffers that go through xrep_findroot_block but don't match anything.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
As of commit e339dd8d8b ("xfs: use sync buffer I/O for sync delwri
queue submission"), the delwri submission code uses sync buffer I/O
for sync delwri I/O. Instead of waiting on async I/O to unlock the
buffer, it uses the underlying sync I/O completion mechanism.
If delwri buffer submission fails due to a shutdown scenario, an
error is set on the buffer and buffer completion never occurs. This
can cause xfs_buf_delwri_submit() to deadlock waiting on a
completion event.
We could check the error state before waiting on such buffers, but
that doesn't serialize against the case of an error set via a racing
I/O completion. Instead, invoke I/O completion in the shutdown case
regardless of buffer I/O type.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
The cached writeback mapping is EOF trimmed to try and avoid races
between post-eof block management and writeback that result in
sending cached data to a stale location. The cached mapping is
currently trimmed on the validation check, which leaves a race
window between the time the mapping is cached and when it is trimmed
against the current inode size.
For example, if a new mapping is cached by delalloc conversion on a
blocksize == page size fs, we could cycle various locks, perform
memory allocations, etc. in the writeback codepath before the
associated mapping is eventually trimmed to i_size. This leaves
enough time for a post-eof truncate and file append before the
cached mapping is trimmed. The former event essentially invalidates
a range of the cached mapping and the latter bumps the inode size
such the trim on the next writepage event won't trim all of the
invalid blocks. fstest generic/464 reproduces this scenario
occasionally and causes a lost writeback and stale delalloc blocks
warning on inode inactivation.
To work around this problem, trim the cached writeback mapping as
soon as it is cached in addition to on subsequent validation checks.
This is a minor tweak to tighten the race window as much as possible
until a proper invalidation mechanism is available.
Fixes: 40214d128e ("xfs: trim writepage mapping to within eof")
Cc: <stable@vger.kernel.org> # v4.14+
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Allison Henderson <allison.henderson@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Drop LIST_HEAD where the variable it declares is never used.
Commit 0410c3bb2b ("xfs: factor ag btree root block
initialisation") stopped using buffer_list and started using a
buffer list in an aghdr_init_data structure, but the declaration
of buffer_list was not removed.
The semantic patch that fixes this problem is as follows:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
identifier x;
@@
- LIST_HEAD(x);
... when != x
// </smpl>
Fixes: 0410c3bb2b ("xfs: factor ag btree root block initialisation")
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Drop LIST_HEAD where the variable it declares has never
been used.
The semantic patch that fixes this problem is as follows:
(http://coccinelle.lip6.fr/)
// <smpl>
@@
identifier x;
@@
- LIST_HEAD(x);
... when != x
// </smpl>
Fixes: 26f1fe858f ("xfs: reduce lock hold times in buffer writeback")
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
At mount time, we allocate m_rsum_cache with the number of realtime
bitmap blocks. However, xfs_growfs_rt() can increase the number of
realtime bitmap blocks. Using the cache after this happens may access
out of the bounds of the cache. Fix it by reallocating the cache in this
case.
Fixes: 355e353213 ("xfs: cache minimum realtime summary level")
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Use __print_symbolic to print the scrub type in ftrace output.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Use __print_symbolic to print the btree type in ftrace output.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Move XFS_INODE_FORMAT_STR to libxfs so that we don't forget to keep it
updated, and add necessary TRACE_DEFINE_ENUM.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Move XFS_AG_BTREE_CMP_FORMAT_STR to libxfs so that we don't forget to
keep it updated, and TRACE_DEFINE_ENUM the values while we're at it.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
ftrace's __print_symbolic() has a (very poorly documented) requirement
that any enum values used in the symbol to string translation table be
wrapped in a TRACE_DEFINE_ENUM so that the enum value can be encoded in
the ftrace ring buffer. Fix this unsatisfied requirement.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Use %pS instead of %pF in ftrace strings so that we record the actual
function address instead of the function descriptor.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Several ioctl structs change size between native 32-bit (ia32) and x32
applications, because x32 follows the native 64-bit (amd64) integer
alignment rules and uses 64-bit time_t. In these instances, the ioctl
number changes so userspace simply gets -ENOTTY. This scenario can be
handled by simply adding more cases.
Looking at the different ioctls implemented here:
- All the ones marked 'No size or alignment issue on any arch' should
presumably all be fine.
- All the ones under BROKEN_X86_ALIGNMENT are different under integer
alignment rules. Since x32 matches amd64 here, we just need both
sets of cases handled.
- XFS_IOC_SWAPEXT has both integer alignment differences and time_t
differences. Since x32 matches amd64 here, we need to add a case
which calls the native implementation.
- The remaining ioctls have neither 64-bit integers nor time_t, so
x32 matches ia32 here and no change is required at this level. The
bulkstat ioctl implementations have some pointer chasing which is
handled separately.
Signed-off-by: Nick Bowler <nbowler@draconx.ca>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
The bulkstat family of ioctls are problematic on x32, because there is
a mixup of native 32-bit and 64-bit conventions. The xfs_fsop_bulkreq
struct contains pointers and 32-bit integers so that matches the native
32-bit layout, and that means the ioctl implementation goes into the
regular compat path on x32.
However, the 'ubuffer' member of that struct in turn refers to either
struct xfs_inogrp or xfs_bstat (or an array of these). On x32, those
structures match the native 64-bit layout. The compat implementation
writes out the 32-bit version of these structures. This is not the
expected format for x32 userspace, causing problems.
Fortunately the functions which actually output these xfs_inogrp and
xfs_bstat structures have an easy way to select which output format
is required, so we just need a little tweak to select the right format
on x32.
Signed-off-by: Nick Bowler <nbowler@draconx.ca>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
While inspecting the ioctl implementations, I noticed that the compat
implementation of XFS_IOC_ATTRLIST_BY_HANDLE does not do exactly the
same thing as the native implementation. Specifically, the "cursor"
does not appear to be written out to userspace on the compat path,
like it is on the native path.
This adjusts the compat implementation to copy out the cursor just
like the native implementation does. The attrlist cursor does not
require any special compat handling. This fixes xfstests xfs/269
on both IA-32 and x32 userspace, when running on an amd64 kernel.
Signed-off-by: Nick Bowler <nbowler@draconx.ca>
Fixes: 0facef7fb0 ("xfs: in _attrlist_by_handle, copy the cursor back to userspace")
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Since mkfs always formats the filesystem with the realtime bitmap and
summary inodes immediately after the root directory, we should expect
that both of them are present and loadable, even if there isn't a
realtime volume attached. There's no reason to skip this if rbmino ==
NULLFSINO; in fact, this causes an immediate crash if the there /is/ a
realtime volume and someone writes to it.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Bill O'Donnell <billodo@redhat.com>
The realtime summary is a two-dimensional array on disk, effectively:
u32 rsum[log2(number of realtime extents) + 1][number of blocks in the bitmap]
rsum[log][bbno] is the number of extents of size 2**log which start in
bitmap block bbno.
xfs_rtallocate_extent_near() uses xfs_rtany_summary() to check whether
rsum[log][bbno] != 0 for any log level. However, the summary array is
stored in row-major order (i.e., like an array in C), so all of these
entries are not adjacent, but rather spread across the entire summary
file. In the worst case (a full bitmap block), xfs_rtany_summary() has
to check every level.
This means that on a moderately-used realtime device, an allocation will
waste a lot of time finding, reading, and releasing buffers for the
realtime summary. In particular, one of our storage services (which runs
on servers with 8 very slow CPUs and 15 8 TB XFS realtime filesystems)
spends almost 5% of its CPU cycles in xfs_rtbuf_get() and
xfs_trans_brelse() called from xfs_rtany_summary().
One solution would be to also store the summary with the dimensions
swapped. However, this would require a disk format change to a very old
component of XFS.
Instead, we can cache the minimum size which contains any extents. We do
so lazily; rather than guaranteeing that the cache contains the precise
minimum, it always contains a loose lower bound which we tighten when we
read or update a summary block. This only uses a few kilobytes of memory
and is already serialized via the realtime bitmap and summary inode
locks, so the cost is minimal. With this change, the same workload only
spends 0.2% of its CPU cycles in the realtime allocator.
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
A big block filesystem might require more than one inobt record to cover
all the inodes in the block. In these cases it is not correct to round
the irec count up to the nearest block because this causes us to
overestimate the number of inode blocks we expect to find.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Store the inode cluster alignment information in units of inodes and
blocks in the mount data so that we don't have to keep recalculating
them.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Store the number of inodes and blocks per inode cluster in the mount
data so that we don't have to keep recalculating them.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Add new helpers to convert units of fs blocks into inodes, and AG blocks
into AG inodes, respectively. Convert all the open-coded conversions
and XFS_OFFBNO_TO_AGINO(, , 0) calls to use them, as appropriate. The
OFFBNO_TO_AGINO macro is retained for xfs_repair.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Owner information for static fs metadata can be defined readonly at
build time because it never changes across filesystems. This enables us
to reduce stack usage (particularly in scrub) because we can use the
statically defined oinfo structures.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Only certain functions actually change the contents of an
xfs_owner_info; the rest can accept a const struct pointer. This will
enable us to save stack space by hoisting static owner info types to
be const global variables.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
There's no need to bundle a pointer to the defer op type into the defer
op control structure. Instead, store the defer op type enum, which
enables us to shorten some of the lines.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Recently, we forgot to port a new defer op type to xfsprogs, which
caused us some userspace pain. Reorganize the way we make libxfs
clients supply defer op type information so that all type information
has to be provided at build time instead of risky runtime dynamic
configuration.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
A log recovery failure has been reproduced where a symlink inode has
a zero length in extent form. It was caused by a shutdown during a
combined fstress+fsmark workload.
The underlying problem is the issue in xfs_inactive_symlink(): the
inode is unlocked between the symlink inactivation/truncation and
the inode being freed. This opens a window for the inode to be
written to disk before it xfs_ifree() removes it from the unlinked
list, marks it free in the inobt and zeros the mode.
For shortform inodes, the fix is simple. xfs_ifree() clears the data
fork state, so there's no need to do it in xfs_inactive_symlink().
This means the shortform fork verifier will not see a zero length
data fork as it mirrors the inode size through to xfs_ifree()), and
hence if the inode gets written back and the fork verifiers are run
they will still see a fork that matches the on-disk inode size.
For extent form (remote) symlinks, it is a little more tricky. Here
we explicitly set the inode size to zero, so the above race can lead
to zero length symlinks on disk. Because the inode is unlinked at
this point (i.e. on the unlinked list) and unreferenced, it can
never be seen again by a user. Hence when we set the inode size to
zeor, also change the type to S_IFREG. xfs_ifree() expects S_IFREG
inodes to be of zero length, and so this avoids all the problems of
zero length symlinks ever hitting the disk. It also avoids the
problem of needing to handle zero length symlink inodes in log
recovery to replay the extent free intents and the remaining
deferops to free the extents the symlink used.
Also add a couple of asserts to warn us if zero length symlinks end
up in either the symlink create or inactivation paths.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
There is a statement that has an unwanted space in the indentation.
Remove it.
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
The function xfs_alloc_get_freelist calls xfs_perag_put to drop the
reference. However, pag->pagf_btreeblks is read and written after the
put operation. This patch moves the put operation later.
Signed-off-by: Pan Bian <bianpan2016@163.com>
Reviewed-by: Carlos Maiolino <cmaiolino@redhat.com>
[darrick: minor changelog edits]
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
In xfs_reflink_end_cow, we allocate a single transaction for the entire
end_cow operation and then loop the CoW fork mappings to move them to
the data fork. This design fails on a heavily fragmented filesystem
where an inode's data fork has exactly one more extent than would fit in
an extents-format fork, because the unmap can collapse the data fork
into extents format (freeing the bmbt block) but the remap can expand
the data fork back into a (newly allocated) bmbt block. If the number
of extents we end up remapping is large, we can overflow the block
reservation because we reserved blocks assuming that we were adding
mappings into an already-cleared area of the data fork.
Let's say we have 8 extents in the data fork, 8 extents in the CoW fork,
and the data fork can hold at most 7 extents before needing to convert
to btree format; and that blocks A-P are discontiguous single-block
extents:
0......7
D: ABCDEFGH
C: IJKLMNOP
When a write to file blocks 0-7 completes, we must remap I-P into the
data fork. We start by removing H from the btree-format data fork. Now
we have 7 extents, so we convert the fork to extents format, freeing the
bmbt block. We then move P into the data fork and it now has 8 extents
again. We must convert the data fork back to btree format, requiring a
block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0,
we'll need a total of 8 block allocations to remap all 8 blocks. We
reserved only enough blocks to handle one btree split (5 blocks on a 4k
block filesystem), which means we overflow the block reservation.
To fix this issue, create a separate helper function to remap a single
extent, and change _reflink_end_cow to call it in a tight loop over the
entire range we're completing. As a side effect this also removes the
size restrictions on how many extents we can end_cow at a time, though
nobody ever hit that. It is not reasonable to reserve N blocks to remap
N blocks.
Note that this can be reproduced after ~320 million fsx ops while
running generic/938 (long soak directio fsx exerciser):
XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116
<machine registers snipped>
Call Trace:
xfs_trans_dup+0x211/0x250 [xfs]
xfs_trans_roll+0x6d/0x180 [xfs]
xfs_defer_trans_roll+0x10c/0x3b0 [xfs]
xfs_defer_finish_noroll+0xdf/0x740 [xfs]
xfs_defer_finish+0x13/0x70 [xfs]
xfs_reflink_end_cow+0x2c6/0x680 [xfs]
xfs_dio_write_end_io+0x115/0x220 [xfs]
iomap_dio_complete+0x3f/0x130
iomap_dio_rw+0x3c3/0x420
xfs_file_dio_aio_write+0x132/0x3c0 [xfs]
xfs_file_write_iter+0x8b/0xc0 [xfs]
__vfs_write+0x193/0x1f0
vfs_write+0xba/0x1c0
ksys_write+0x52/0xc0
do_syscall_64+0x50/0x160
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
xfs_btree_sblock_verify_crc is a bool so should not be returning
a failaddr_t; worse, if xfs_log_check_lsn fails it returns
__this_address which looks like a boolean true (i.e. success)
to the caller.
(interestingly xfs_btree_lblock_verify_crc doesn't have the issue)
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
In commit e53c4b598, I *tried* to teach xfs to force writeback when we
fzero/fpunch right up to EOF so that if EOF is in the middle of a page,
the post-EOF part of the page gets zeroed before we return to userspace.
Unfortunately, I missed the part where PAGE_MASK is ~(PAGE_SIZE - 1),
which means that we totally fail to zero if we're fpunching and EOF is
within the first page. Worse yet, the same PAGE_MASK thinko plagues the
filemap_write_and_wait_range call, so we'd initiate writeback of the
entire file, which (mostly) masked the thinko.
Drop the tricky PAGE_MASK and replace it with correct usage of PAGE_SIZE
and the proper rounding macros.
Fixes: e53c4b598 ("xfs: ensure post-EOF zeroing happens after zeroing part of a file")
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
When project is set, we should use inode limit minus the used count
Signed-off-by: Ye Yin <dbyin@tencent.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Long saga. There have been days spent following this through dead end
after dead end in multi-GB event traces. This morning, after writing
a trace-cmd wrapper that enabled me to be more selective about XFS
trace points, I discovered that I could get just enough essential
tracepoints enabled that there was a 50:50 chance the fsx config
would fail at ~115k ops. If it didn't fail at op 115547, I stopped
fsx at op 115548 anyway.
That gave me two traces - one where the problem manifested, and one
where it didn't. After refining the traces to have the necessary
information, I found that in the failing case there was a real
extent in the COW fork compared to an unwritten extent in the
working case.
Walking back through the two traces to the point where the CWO fork
extents actually diverged, I found that the bad case had an extra
unwritten extent in it. This is likely because the bug it led me to
had triggered multiple times in those 115k ops, leaving stray
COW extents around. What I saw was a COW delalloc conversion to an
unwritten extent (as they should always be through
xfs_iomap_write_allocate()) resulted in a /written extent/:
xfs_writepage: dev 259:0 ino 0x83 pgoff 0x17000 size 0x79a00 offset 0 length 0
xfs_iext_remove: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/2 offset 32 block 152 count 20 flag 1 caller xfs_bmap_add_extent_delay_real
xfs_bmap_pre_update: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/1 offset 1 block 4503599627239429 count 31 flag 0 caller xfs_bmap_add_extent_delay_real
xfs_bmap_post_update: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/1 offset 1 block 121 count 51 flag 0 caller xfs_bmap_add_ex
Basically, Cow fork before:
0 1 32 52
+H+DDDDDDDDDDDD+UUUUUUUUUUU+
PREV RIGHT
COW delalloc conversion allocates:
1 32
+uuuuuuuuuuuu+
NEW
And the result according to the xfs_bmap_post_update trace was:
0 1 32 52
+H+wwwwwwwwwwwwwwwwwwwwwwww+
PREV
Which is clearly wrong - it should be a merged unwritten extent,
not an unwritten extent.
That lead me to look at the LEFT_FILLING|RIGHT_FILLING|RIGHT_CONTIG
case in xfs_bmap_add_extent_delay_real(), and sure enough, there's
the bug.
It takes the old delalloc extent (PREV) and adds the length of the
RIGHT extent to it, takes the start block from NEW, removes the
RIGHT extent and then updates PREV with the new extent.
What it fails to do is update PREV.br_state. For delalloc, this is
always XFS_EXT_NORM, while in this case we are converting the
delayed allocation to unwritten, so it needs to be updated to
XFS_EXT_UNWRITTEN. This LF|RF|RC case does not do this, and so
the resultant extent is always written.
And that's the bug I've been chasing for a week - a bmap btree bug,
not a reflink/dedupe/copy_file_range bug, but a BMBT bug introduced
with the recent in core extent tree scalability enhancements.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
On a sub-page block size filesystem, fsx is failing with a data
corruption after a series of operations involving copying a file
with the destination offset beyond EOF of the destination of the file:
8093(157 mod 256): TRUNCATE DOWN from 0x7a120 to 0x50000 ******WWWW
8094(158 mod 256): INSERT 0x25000 thru 0x25fff (0x1000 bytes)
8095(159 mod 256): COPY 0x18000 thru 0x1afff (0x3000 bytes) to 0x2f400
8096(160 mod 256): WRITE 0x5da00 thru 0x651ff (0x7800 bytes) HOLE
8097(161 mod 256): COPY 0x2000 thru 0x5fff (0x4000 bytes) to 0x6fc00
The second copy here is beyond EOF, and it is to sub-page (4k) but
block aligned (1k) offset. The clone runs the EOF zeroing, landing
in a pre-existing post-eof delalloc extent. This zeroes the post-eof
extents in the page cache just fine, dirtying the pages correctly.
The problem is that xfs_reflink_remap_prep() now truncates the page
cache over the range that it is copying it to, and rounds that down
to cover the entire start page. This removes the dirty page over the
delalloc extent from the page cache without having written it back.
Hence later, when the page cache is flushed, the page at offset
0x6f000 has not been written back and hence exposes stale data,
which fsx trips over less than 10 operations later.
Fix this by changing xfs_reflink_remap_prep() to use
xfs_flush_unmap_range().
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
The extent shifting code uses a flush and invalidate mechainsm prior
to shifting extents around. This is similar to what
xfs_free_file_space() does, but it doesn't take into account things
like page cache vs block size differences, and it will fail if there
is a page that it currently busy.
xfs_flush_unmap_range() handles all of these cases, so just convert
xfs_prepare_shift() to us that mechanism rather than having it's own
special sauce.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
The last AG may be very small comapred to all other AGs, and hence
AG reservations based on the superblock AG size may actually consume
more space than the AG actually has. This results on assert failures
like:
XFS: Assertion failed: xfs_perag_resv(pag, XFS_AG_RESV_METADATA)->ar_reserved + xfs_perag_resv(pag, XFS_AG_RESV_RMAPBT)->ar_reserved <= pag->pagf_freeblks + pag->pagf_flcount, file: fs/xfs/libxfs/xfs_ag_resv.c, line: 319
[ 48.932891] xfs_ag_resv_init+0x1bd/0x1d0
[ 48.933853] xfs_fs_reserve_ag_blocks+0x37/0xb0
[ 48.934939] xfs_mountfs+0x5b3/0x920
[ 48.935804] xfs_fs_fill_super+0x462/0x640
[ 48.936784] ? xfs_test_remount_options+0x60/0x60
[ 48.937908] mount_bdev+0x178/0x1b0
[ 48.938751] mount_fs+0x36/0x170
[ 48.939533] vfs_kern_mount.part.43+0x54/0x130
[ 48.940596] do_mount+0x20e/0xcb0
[ 48.941396] ? memdup_user+0x3e/0x70
[ 48.942249] ksys_mount+0xba/0xd0
[ 48.943046] __x64_sys_mount+0x21/0x30
[ 48.943953] do_syscall_64+0x54/0x170
[ 48.944835] entry_SYSCALL_64_after_hwframe+0x49/0xbe
Hence we need to ensure the finobt per-ag space reservations take
into account the size of the last AG rather than treat it like all
the other full size AGs.
Note that both refcountbt and rmapbt already take the size of the AG
into account via reading the AGF length directly.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
When retrying a failed inode or dquot buffer,
xfs_buf_resubmit_failed_buffers() clears all the failed flags from
the inde/dquot log items. In doing so, it also drops all the
reference counts on the buffer that the failed log items hold. This
means it can drop all the active references on the buffer and hence
free the buffer before it queues it for write again.
Putting the buffer on the delwri queue takes a reference to the
buffer (so that it hangs around until it has been written and
completed), but this goes bang if the buffer has already been freed.
Hence we need to add the buffer to the delwri queue before we remove
the failed flags from the log items attached to the buffer to ensure
it always remains referenced during the resubmit process.
Reported-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
xfs_file_remap_range() is only used in fs/xfs/xfs_file.c, so make it
static.
This addresses a gcc warning when -Wmissing-prototypes is enabled.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Page writeback indirectly handles shared extents via the existence
of overlapping COW fork blocks. If COW fork blocks exist, writeback
always performs the associated copy-on-write regardless if the
underlying blocks are actually shared. If the blocks are shared,
then overlapping COW fork blocks must always exist.
fstests shared/010 reproduces a case where a buffered write occurs
over a shared block without performing the requisite COW fork
reservation. This ultimately causes writeback to the shared extent
and data corruption that is detected across md5 checks of the
filesystem across a mount cycle.
The problem occurs when a buffered write lands over a shared extent
that crosses an extent size hint boundary and that also happens to
have a partial COW reservation that doesn't cover the start and end
blocks of the data fork extent.
For example, a buffered write occurs across the file offset (in FSB
units) range of [29, 57]. A shared extent exists at blocks [29, 35]
and COW reservation already exists at blocks [32, 34]. After
accommodating a COW extent size hint of 32 blocks and the existing
reservation at offset 32, xfs_reflink_reserve_cow() allocates 32
blocks of reservation at offset 0 and returns with COW reservation
across the range of [0, 34]. The associated data fork extent is
still [29, 35], however, which isn't fully covered by the COW
reservation.
This leads to a buffered write at file offset 35 over a shared
extent without associated COW reservation. Writeback eventually
kicks in, performs an overwrite of the underlying shared block and
causes the associated data corruption.
Update xfs_reflink_reserve_cow() to accommodate the fact that a
delalloc allocation request may not fully cover the extent in the
data fork. Trim the data fork extent appropriately, just as is done
for shared extent boundaries and/or existing COW reservations that
happen to overlap the start of the data fork extent. This prevents
shared/010 failures due to data corruption on reflink enabled
filesystems.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
generic/070 on 64k block size filesystems is failing with a verifier
corruption on writeback or an attribute leaf block:
[ 94.973083] XFS (pmem0): Metadata corruption detected at xfs_attr3_leaf_verify+0x246/0x260, xfs_attr3_leaf block 0x811480
[ 94.975623] XFS (pmem0): Unmount and run xfs_repair
[ 94.976720] XFS (pmem0): First 128 bytes of corrupted metadata buffer:
[ 94.978270] 000000004b2e7b45: 00 00 00 00 00 00 00 00 3b ee 00 00 00 00 00 00 ........;.......
[ 94.980268] 000000006b1db90b: 00 00 00 00 00 81 14 80 00 00 00 00 00 00 00 00 ................
[ 94.982251] 00000000433f2407: 22 7b 5c 82 2d 5c 47 4c bb 31 1c 37 fa a9 ce d6 "{\.-\GL.1.7....
[ 94.984157] 0000000010dc7dfb: 00 00 00 00 00 81 04 8a 00 0a 18 e8 dd 94 01 00 ................
[ 94.986215] 00000000d5a19229: 00 a0 dc f4 fe 98 01 68 f0 d8 07 e0 00 00 00 00 .......h........
[ 94.988171] 00000000521df36c: 0c 2d 32 e2 fe 20 01 00 0c 2d 58 65 fe 0c 01 00 .-2.. ...-Xe....
[ 94.990162] 000000008477ae06: 0c 2d 5b 66 fe 8c 01 00 0c 2d 71 35 fe 7c 01 00 .-[f.....-q5.|..
[ 94.992139] 00000000a4a6bca6: 0c 2d 72 37 fc d4 01 00 0c 2d d8 b8 f0 90 01 00 .-r7.....-......
[ 94.994789] XFS (pmem0): xfs_do_force_shutdown(0x8) called from line 1453 of file fs/xfs/xfs_buf.c. Return address = ffffffff815365f3
This is failing this check:
end = ichdr.freemap[i].base + ichdr.freemap[i].size;
if (end < ichdr.freemap[i].base)
>>>>> return __this_address;
if (end > mp->m_attr_geo->blksize)
return __this_address;
And from the buffer output above, the freemap array is:
freemap[0].base = 0x00a0
freemap[0].size = 0xdcf4 end = 0xdd94
freemap[1].base = 0xfe98
freemap[1].size = 0x0168 end = 0x10000
freemap[2].base = 0xf0d8
freemap[2].size = 0x07e0 end = 0xf8b8
These all look valid - the block size is 0x10000 and so from the
last check in the above verifier fragment we know that the end
of freemap[1] is valid. The problem is that end is declared as:
uint16_t end;
And (uint16_t)0x10000 = 0. So we have a verifier bug here, not a
corruption. Fix the verifier to use uint32_t types for the check and
hence avoid the overflow.
Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=201577
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Use DUMP_PREFIX_OFFSET when printing hex dumps of corrupt buffers
because modern Linux now prints a 32-bit hash of our 64-bit pointer when
using DUMP_PREFIX_ADDRESS:
00000000b4bb4297: 00 00 00 00 00 00 00 00 3b ee 00 00 00 00 00 00 ........;.......
00000005ec77e26: 00 00 00 00 02 d0 5a 00 00 00 00 00 00 00 00 00 ......Z.........
000000015938018: 21 98 e8 b4 fd de 4c 07 bc ea 3c e5 ae b4 7c 48 !.....L...<...|H
This is totally worthless for a sequential dump since we probably only
care about tracking the buffer offsets and afaik there's no way to
recover the actual pointer from the hashed value.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
In this function, once 'buf' has been allocated, we unconditionally
return 0.
However, 'error' is set to some error codes in several error handling
paths.
Before commit 232b51948b ("xfs: simplify the xfs_getbmap interface")
this was not an issue because all error paths were returning directly,
but now that some cleanup at the end may be needed, we must propagate the
error code.
Fixes: 232b51948b ("xfs: simplify the xfs_getbmap interface")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Rework the vfs_clone_file_range and vfs_dedupe_file_range infrastructure to use
a common .remap_file_range method and supply generic bounds and sanity checking
functions that are shared with the data write path. The current VFS
infrastructure has problems with rlimit, LFS file sizes, file time stamps,
maximum filesystem file sizes, stripping setuid bits, etc and so they are
addressed in these commits.
We also introduce the ability for the ->remap_file_range methods to return short
clones so that clones for vfs_copy_file_range() don't get rejected if the entire
range can't be cloned. It also allows filesystems to sliently skip deduplication
of partial EOF blocks if they are not capable of doing so without requiring
errors to be thrown to userspace.
All existing filesystems are converted to user the new .remap_file_range method,
and both XFS and ocfs2 are modified to make use of the new generic checking
infrastructure.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=CE/A
-----END PGP SIGNATURE-----
Merge tag 'xfs-4.20-merge-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Pull vfs dedup fixes from Dave Chinner:
"This reworks the vfs data cloning infrastructure.
We discovered many issues with these interfaces late in the 4.19 cycle
- the worst of them (data corruption, setuid stripping) were fixed for
XFS in 4.19-rc8, but a larger rework of the infrastructure fixing all
the problems was needed. That rework is the contents of this pull
request.
Rework the vfs_clone_file_range and vfs_dedupe_file_range
infrastructure to use a common .remap_file_range method and supply
generic bounds and sanity checking functions that are shared with the
data write path. The current VFS infrastructure has problems with
rlimit, LFS file sizes, file time stamps, maximum filesystem file
sizes, stripping setuid bits, etc and so they are addressed in these
commits.
We also introduce the ability for the ->remap_file_range methods to
return short clones so that clones for vfs_copy_file_range() don't get
rejected if the entire range can't be cloned. It also allows
filesystems to sliently skip deduplication of partial EOF blocks if
they are not capable of doing so without requiring errors to be thrown
to userspace.
Existing filesystems are converted to user the new remap_file_range
method, and both XFS and ocfs2 are modified to make use of the new
generic checking infrastructure"
* tag 'xfs-4.20-merge-2' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (28 commits)
xfs: remove [cm]time update from reflink calls
xfs: remove xfs_reflink_remap_range
xfs: remove redundant remap partial EOF block checks
xfs: support returning partial reflink results
xfs: clean up xfs_reflink_remap_blocks call site
xfs: fix pagecache truncation prior to reflink
ocfs2: remove ocfs2_reflink_remap_range
ocfs2: support partial clone range and dedupe range
ocfs2: fix pagecache truncation prior to reflink
ocfs2: truncate page cache for clone destination file before remapping
vfs: clean up generic_remap_file_range_prep return value
vfs: hide file range comparison function
vfs: enable remap callers that can handle short operations
vfs: plumb remap flags through the vfs dedupe functions
vfs: plumb remap flags through the vfs clone functions
vfs: make remap_file_range functions take and return bytes completed
vfs: remap helper should update destination inode metadata
vfs: pass remap flags to generic_remap_checks
vfs: pass remap flags to generic_remap_file_range_prep
vfs: combine the clone and dedupe into a single remap_file_range
...
Now that the vfs remap helper dirties the inode [cm]time for us, xfs no
longer needs to do that on its own.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
