kernel/git/stable/linux-stable.git/fs/xfs, branch linux-6.10.y

xfs: xfs_finobt_count_blocks() walks the wrong btree

2024-09-12T09:12:42+00:00

commit 95179935beadccaf0f0bb461adb778731e293da4 upstream. As a result of the factoring in commit 14dd46cf31f4 ("xfs: split xfs_inobt_init_cursor"), mount started taking a long time on a user's filesystem. For Anders, this made mount times regress from under a second to over 15 minutes for a filesystem with only 30 million inodes in it. Anders bisected it down to the above commit, but even then the bug was not obvious. In this commit, over 20 calls to xfs_inobt_init_cursor() were modified, and some we modified to call a new function named xfs_finobt_init_cursor(). If that takes you a moment to reread those function names to see what the rename was, then you have realised why this bug wasn't spotted during review. And it wasn't spotted on inspection even after the bisect pointed at this commit - a single missing "f" isn't the easiest thing for a human eye to notice.... The result is that xfs_finobt_count_blocks() now incorrectly calls xfs_inobt_init_cursor() so it is now walking the inobt instead of the finobt. Hence when there are lots of allocated inodes in a filesystem, mount takes a -long- time run because it now walks a massive allocated inode btrees instead of the small, nearly empty free inode btrees. It also means all the finobt space reservations are wrong, so mount could potentially given ENOSPC on kernel upgrade. In hindsight, commit 14dd46cf31f4 should have been two commits - the first to convert the finobt callers to the new API, the second to modify the xfs_inobt_init_cursor() API for the inobt callers. That would have made the bug very obvious during review. Fixes: 14dd46cf31f4 ("xfs: split xfs_inobt_init_cursor") Reported-by: Anders Blomdell Signed-off-by: Dave Chinner Reviewed-by: Christoph Hellwig Reviewed-by: Darrick J. Wong Signed-off-by: Darrick J. Wong Signed-off-by: Chandan Babu R Signed-off-by: Greg Kroah-Hartman

xfs: honor init_xattrs in xfs_init_new_inode for !ATTR fs

2024-06-26T08:59:25+00:00

xfs_init_new_inode ignores the init_xattrs parameter for filesystems that do not have ATTR enabled. As a result, the first init_xattrs file to be created by the kernel will not have an attr fork created to store acls. Storing that first acl will add ATTR to the superblock flags, so subsequent files will be created with attr forks. The overhead of this is so small that chances are that nobody has noticed this behavior. However, this is disastrous on a filesystem with parent pointers because it requires that a new linkable file /must/ have a pre-existing attr fork, and the parent pointers code uses init_xattrs to create that fork. The preproduction version of mkfs.xfs used to set this, but the V5 sb verifier only requires ATTR2, not ATTR. There is no guard for filesystems with (PARENT && !ATTR). It turns out that I misunderstood the two flags -- ATTR means that we at some point created an attr fork to store xattrs in a file; ATTR2 apparently means only that inodes have dynamic fork offsets or that the filesystem was mounted with the "attr2" option. Fixes: 2442ee15bb1e ("xfs: eager inode attr fork init needs attr feature awareness") Signed-off-by: Darrick J. Wong Reviewed-by: Christoph Hellwig Signed-off-by: Chandan Babu R

xfs: fix direction in XFS_IOC_EXCHANGE_RANGE

2024-06-26T08:59:25+00:00

The kernel reads userspace's buffer but does not write it back. Therefore this is really an _IOW ioctl. Change this before 6.10 final releases. Signed-off-by: Darrick J. Wong Reviewed-by: Christoph Hellwig Signed-off-by: Chandan Babu R

xfs: allow unlinked symlinks and dirs with zero size

2024-06-26T08:59:25+00:00

For a very very long time, inode inactivation has set the inode size to zero before unmapping the extents associated with the data fork. Unfortunately, commit 3c6f46eacd876 changed the inode verifier to prohibit zero-length symlinks and directories. If an inode happens to get logged in this state and the system crashes before freeing the inode, log recovery will also fail on the broken inode. Therefore, allow zero-size symlinks and directories as long as the link count is zero; nobody will be able to open these files by handle so there isn't any risk of data exposure. Fixes: 3c6f46eacd876 ("xfs: sanity check directory inode di_size") Signed-off-by: Darrick J. Wong Reviewed-by: Christoph Hellwig Signed-off-by: Chandan Babu R

xfs: restrict when we try to align cow fork delalloc to cowextsz hints

2024-06-26T08:59:24+00:00

xfs/205 produces the following failure when always_cow is enabled: --- a/tests/xfs/205.out 2024-02-28 16:20:24.437887970 -0800 +++ b/tests/xfs/205.out.bad 2024-06-03 21:13:40.584000000 -0700 @@ -1,4 +1,5 @@ QA output created by 205 *** one file + !!! disk full (expected) *** one file, a few bytes at a time *** done This is the result of overly aggressive attempts to align cow fork delalloc reservations to the CoW extent size hint. Looking at the trace data, we're trying to append a single fsblock to the "fred" file. Trying to create a speculative post-eof reservation fails because there's not enough space. We then set @prealloc_blocks to zero and try again, but the cowextsz alignment code triggers, which expands our request for a 1-fsblock reservation into a 39-block reservation. There's not enough space for that, so the whole write fails with ENOSPC even though there's sufficient space in the filesystem to allocate the single block that we need to land the write. There are two things wrong here -- first, we shouldn't be attempting speculative preallocations beyond what was requested when we're low on space. Second, if we've already computed a posteof preallocation, we shouldn't bother trying to align that to the cowextsize hint. Fix both of these problems by adding a flag that only enables the expansion of the delalloc reservation to the cowextsize if we're doing a non-extending write, and only if we're not doing an ENOSPC retry. This requires us to move the ENOSPC retry logic to xfs_bmapi_reserve_delalloc. I probably should have caught this six years ago when 6ca30729c206d was being reviewed, but oh well. Update the comments to reflect what the code does now. Fixes: 6ca30729c206d ("xfs: bmap code cleanup") Signed-off-by: Darrick J. Wong Reviewed-by: Christoph Hellwig Signed-off-by: Chandan Babu R

xfs: fix freeing speculative preallocations for preallocated files

2024-06-26T08:59:17+00:00

xfs_can_free_eofblocks returns false for files that have persistent preallocations unless the force flag is passed and there are delayed blocks. This means it won't free delalloc reservations for files with persistent preallocations unless the force flag is set, and it will also free the persistent preallocations if the force flag is set and the file happens to have delayed allocations. Both of these are bad, so do away with the force flag and always free only post-EOF delayed allocations for files with the XFS_DIFLAG_PREALLOC or APPEND flags set. Signed-off-by: Christoph Hellwig Reviewed-by: Darrick J. Wong Signed-off-by: Chandan Babu R

xfs: fix unlink vs cluster buffer instantiation race

2024-06-17T05:47:09+00:00

Luis has been reporting an assert failure when freeing an inode cluster during inode inactivation for a while. The assert looks like: XFS: Assertion failed: bp->b_flags & XBF_DONE, file: fs/xfs/xfs_trans_buf.c, line: 241 ------------[ cut here ]------------ kernel BUG at fs/xfs/xfs_message.c:102! Oops: invalid opcode: 0000 [#1] PREEMPT SMP KASAN NOPTI CPU: 4 PID: 73 Comm: kworker/4:1 Not tainted 6.10.0-rc1 #4 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014 Workqueue: xfs-inodegc/loop5 xfs_inodegc_worker [xfs] RIP: 0010:assfail (fs/xfs/xfs_message.c:102) xfs RSP: 0018:ffff88810188f7f0 EFLAGS: 00010202 RAX: 0000000000000000 RBX: ffff88816e748250 RCX: 1ffffffff844b0e7 RDX: 0000000000000004 RSI: ffff88810188f558 RDI: ffffffffc2431fa0 RBP: 1ffff11020311f01 R08: 0000000042431f9f R09: ffffed1020311e9b R10: ffff88810188f4df R11: ffffffffac725d70 R12: ffff88817a3f4000 R13: ffff88812182f000 R14: ffff88810188f998 R15: ffffffffc2423f80 FS: 0000000000000000(0000) GS:ffff8881c8400000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 000055fe9d0f109c CR3: 000000014426c002 CR4: 0000000000770ef0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe07f0 DR7: 0000000000000400 PKRU: 55555554 Call Trace: xfs_trans_read_buf_map (fs/xfs/xfs_trans_buf.c:241 (discriminator 1)) xfs xfs_imap_to_bp (fs/xfs/xfs_trans.h:210 fs/xfs/libxfs/xfs_inode_buf.c:138) xfs xfs_inode_item_precommit (fs/xfs/xfs_inode_item.c:145) xfs xfs_trans_run_precommits (fs/xfs/xfs_trans.c:931) xfs __xfs_trans_commit (fs/xfs/xfs_trans.c:966) xfs xfs_inactive_ifree (fs/xfs/xfs_inode.c:1811) xfs xfs_inactive (fs/xfs/xfs_inode.c:2013) xfs xfs_inodegc_worker (fs/xfs/xfs_icache.c:1841 fs/xfs/xfs_icache.c:1886) xfs process_one_work (kernel/workqueue.c:3231) worker_thread (kernel/workqueue.c:3306 (discriminator 2) kernel/workqueue.c:3393 (discriminator 2)) kthread (kernel/kthread.c:389) ret_from_fork (arch/x86/kernel/process.c:147) ret_from_fork_asm (arch/x86/entry/entry_64.S:257) And occurs when the the inode precommit handlers is attempt to look up the inode cluster buffer to attach the inode for writeback. The trail of logic that I can reconstruct is as follows. 1. the inode is clean when inodegc runs, so it is not attached to a cluster buffer when precommit runs. 2. #1 implies the inode cluster buffer may be clean and not pinned by dirty inodes when inodegc runs. 3. #2 implies that the inode cluster buffer can be reclaimed by memory pressure at any time. 4. The assert failure implies that the cluster buffer was attached to the transaction, but not marked done. It had been accessed earlier in the transaction, but not marked done. 5. #4 implies the cluster buffer has been invalidated (i.e. marked stale). 6. #5 implies that the inode cluster buffer was instantiated uninitialised in the transaction in xfs_ifree_cluster(), which only instantiates the buffers to invalidate them and never marks them as done. Given factors 1-3, this issue is highly dependent on timing and environmental factors. Hence the issue can be very difficult to reproduce in some situations, but highly reliable in others. Luis has an environment where it can be reproduced easily by g/531 but, OTOH, I've reproduced it only once in ~2000 cycles of g/531. I think the fix is to have xfs_ifree_cluster() set the XBF_DONE flag on the cluster buffers, even though they may not be initialised. The reasons why I think this is safe are: 1. A buffer cache lookup hit on a XBF_STALE buffer will clear the XBF_DONE flag. Hence all future users of the buffer know they have to re-initialise the contents before use and mark it done themselves. 2. xfs_trans_binval() sets the XFS_BLI_STALE flag, which means the buffer remains locked until the journal commit completes and the buffer is unpinned. Hence once marked XBF_STALE/XFS_BLI_STALE by xfs_ifree_cluster(), the only context that can access the freed buffer is the currently running transaction. 3. #2 implies that future buffer lookups in the currently running transaction will hit the transaction match code and not the buffer cache. Hence XBF_STALE and XFS_BLI_STALE will not be cleared unless the transaction initialises and logs the buffer with valid contents again. At which point, the buffer will be marked marked XBF_DONE again, so having XBF_DONE already set on the stale buffer is a moot point. 4. #2 also implies that any concurrent access to that cluster buffer will block waiting on the buffer lock until the inode cluster has been fully freed and is no longer an active inode cluster buffer. 5. #4 + #1 means that any future user of the disk range of that buffer will always see the range of disk blocks covered by the cluster buffer as not done, and hence must initialise the contents themselves. 6. Setting XBF_DONE in xfs_ifree_cluster() then means the unlinked inode precommit code will see a XBF_DONE buffer from the transaction match as it expects. It can then attach the stale but newly dirtied inode to the stale but newly dirtied cluster buffer without unexpected failures. The stale buffer will then sail through the journal and do the right thing with the attached stale inode during unpin. Hence the fix is just one line of extra code. The explanation of why we have to set XBF_DONE in xfs_ifree_cluster, OTOH, is long and complex.... Fixes: 82842fee6e59 ("xfs: fix AGF vs inode cluster buffer deadlock") Signed-off-by: Dave Chinner Tested-by: Luis Chamberlain Reviewed-by: Christoph Hellwig Reviewed-by: Darrick J. Wong Signed-off-by: Chandan Babu R

xfs: make sure sb_fdblocks is non-negative

2024-06-10T06:08:12+00:00

A user with a completely full filesystem experienced an unexpected shutdown when the filesystem tried to write the superblock during runtime. kernel shows the following dmesg: [ 8.176281] XFS (dm-4): Metadata corruption detected at xfs_sb_write_verify+0x60/0x120 [xfs], xfs_sb block 0x0 [ 8.177417] XFS (dm-4): Unmount and run xfs_repair [ 8.178016] XFS (dm-4): First 128 bytes of corrupted metadata buffer: [ 8.178703] 00000000: 58 46 53 42 00 00 10 00 00 00 00 00 01 90 00 00 XFSB............ [ 8.179487] 00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ [ 8.180312] 00000020: cf 12 dc 89 ca 26 45 29 92 e6 e3 8d 3b b8 a2 c3 .....&E)....;... [ 8.181150] 00000030: 00 00 00 00 01 00 00 06 00 00 00 00 00 00 00 80 ................ [ 8.182003] 00000040: 00 00 00 00 00 00 00 81 00 00 00 00 00 00 00 82 ................ [ 8.182004] 00000050: 00 00 00 01 00 64 00 00 00 00 00 04 00 00 00 00 .....d.......... [ 8.182004] 00000060: 00 00 64 00 b4 a5 02 00 02 00 00 08 00 00 00 00 ..d............. [ 8.182005] 00000070: 00 00 00 00 00 00 00 00 0c 09 09 03 17 00 00 19 ................ [ 8.182008] XFS (dm-4): Corruption of in-memory data detected. Shutting down filesystem [ 8.182010] XFS (dm-4): Please unmount the filesystem and rectify the problem(s) When xfs_log_sb writes super block to disk, b_fdblocks is fetched from m_fdblocks without any lock. As m_fdblocks can experience a positive -> negative -> positive changing when the FS reaches fullness (see xfs_mod_fdblocks). So there is a chance that sb_fdblocks is negative, and because sb_fdblocks is type of unsigned long long, it reads super big. And sb_fdblocks being bigger than sb_dblocks is a problem during log recovery, xfs_validate_sb_write() complains. Fix: As sb_fdblocks will be re-calculated during mount when lazysbcount is enabled, We just need to make xfs_validate_sb_write() happy -- make sure sb_fdblocks is not nenative. This patch also takes care of other percpu counters in xfs_log_sb. Signed-off-by: Wengang Wang Reviewed-by: Darrick J. Wong Signed-off-by: Chandan Babu R

xfs: Add cond_resched to block unmap range and reflink remap path

2024-05-27T15:20:35+00:00

An async dio write to a sparse file can generate a lot of extents and when we unlink this file (using rm), the kernel can be busy in umapping and freeing those extents as part of transaction processing. Similarly xfs reflink remapping path can also iterate over a million extent entries in xfs_reflink_remap_blocks(). Since we can busy loop in these two functions, so let's add cond_resched() to avoid softlockup messages like these. watchdog: BUG: soft lockup - CPU#1 stuck for 22s! [kworker/1:0:82435] CPU: 1 PID: 82435 Comm: kworker/1:0 Tainted: G S L 6.9.0-rc5-0-default #1 Workqueue: xfs-inodegc/sda2 xfs_inodegc_worker NIP [c000000000beea10] xfs_extent_busy_trim+0x100/0x290 LR [c000000000bee958] xfs_extent_busy_trim+0x48/0x290 Call Trace: xfs_alloc_get_rec+0x54/0x1b0 (unreliable) xfs_alloc_compute_aligned+0x5c/0x144 xfs_alloc_ag_vextent_size+0x238/0x8d4 xfs_alloc_fix_freelist+0x540/0x694 xfs_free_extent_fix_freelist+0x84/0xe0 __xfs_free_extent+0x74/0x1ec xfs_extent_free_finish_item+0xcc/0x214 xfs_defer_finish_one+0x194/0x388 xfs_defer_finish_noroll+0x1b4/0x5c8 xfs_defer_finish+0x2c/0xc4 xfs_bunmapi_range+0xa4/0x100 xfs_itruncate_extents_flags+0x1b8/0x2f4 xfs_inactive_truncate+0xe0/0x124 xfs_inactive+0x30c/0x3e0 xfs_inodegc_worker+0x140/0x234 process_scheduled_works+0x240/0x57c worker_thread+0x198/0x468 kthread+0x138/0x140 start_kernel_thread+0x14/0x18 run fstests generic/175 at 2024-02-02 04:40:21 [ C17] watchdog: BUG: soft lockup - CPU#17 stuck for 23s! [xfs_io:7679] watchdog: BUG: soft lockup - CPU#17 stuck for 23s! [xfs_io:7679] CPU: 17 PID: 7679 Comm: xfs_io Kdump: loaded Tainted: G X 6.4.0 NIP [c008000005e3ec94] xfs_rmapbt_diff_two_keys+0x54/0xe0 [xfs] LR [c008000005e08798] xfs_btree_get_leaf_keys+0x110/0x1e0 [xfs] Call Trace: 0xc000000014107c00 (unreliable) __xfs_btree_updkeys+0x8c/0x2c0 [xfs] xfs_btree_update_keys+0x150/0x170 [xfs] xfs_btree_lshift+0x534/0x660 [xfs] xfs_btree_make_block_unfull+0x19c/0x240 [xfs] xfs_btree_insrec+0x4e4/0x630 [xfs] xfs_btree_insert+0x104/0x2d0 [xfs] xfs_rmap_insert+0xc4/0x260 [xfs] xfs_rmap_map_shared+0x228/0x630 [xfs] xfs_rmap_finish_one+0x2d4/0x350 [xfs] xfs_rmap_update_finish_item+0x44/0xc0 [xfs] xfs_defer_finish_noroll+0x2e4/0x740 [xfs] __xfs_trans_commit+0x1f4/0x400 [xfs] xfs_reflink_remap_extent+0x2d8/0x650 [xfs] xfs_reflink_remap_blocks+0x154/0x320 [xfs] xfs_file_remap_range+0x138/0x3a0 [xfs] do_clone_file_range+0x11c/0x2f0 vfs_clone_file_range+0x60/0x1c0 ioctl_file_clone+0x78/0x140 sys_ioctl+0x934/0x1270 system_call_exception+0x158/0x320 system_call_vectored_common+0x15c/0x2ec Cc: Ojaswin Mujoo Signed-off-by: Ritesh Harjani (IBM) Reviewed-by: Darrick J. Wong Tested-by: Disha Goel Signed-off-by: Chandan Babu R

xfs: don't open-code u64_to_user_ptr

2024-05-27T10:25:52+00:00

Don't open-code what the kernel already provides. Signed-off-by: "Darrick J. Wong" Reviewed-by: Christoph Hellwig Signed-off-by: Chandan Babu R