The issue can be easily reproduced using the following reproducer:
root@q:linux# cat repro.sh
set -e
mkfs.btrfs -q -f /dev/sdb
mount /dev/sdb /mnt/btrfs
btrfs quota enable -s /mnt/btrfs
umount /mnt/btrfs
mount /dev/sdb /mnt/btrfs
The issue is that when enabling quotas, at btrfs_quota_enable(), we set
BTRFS_QGROUP_STATUS_FLAG_SIMPLE_MODE at fs_info->qgroup_flags and persist
it in the quota root in the item with the key BTRFS_QGROUP_STATUS_KEY, but
we only set the incompat bit BTRFS_FEATURE_INCOMPAT_SIMPLE_QUOTA after we
commit the transaction used to enable simple quotas.
This means that if after that transaction commit we unmount the filesystem
without starting and committing any other transaction, or we have a power
failure, the next time we mount the filesystem we will find the flag
BTRFS_QGROUP_STATUS_FLAG_SIMPLE_MODE set in the item with the key
BTRFS_QGROUP_STATUS_KEY but we will not find the incompat bit
BTRFS_FEATURE_INCOMPAT_SIMPLE_QUOTA set in the superblock, triggering an
assertion failure at:
To fix this issue, set the BTRFS_FEATURE_INCOMPAT_SIMPLE_QUOTA flag
immediately after setting the BTRFS_QGROUP_STATUS_FLAG_SIMPLE_MODE.
This ensures that both flags are flushed to disk within the same
transaction.
Fixes: 182940f4f4db ("btrfs: qgroup: add new quota mode for simple quotas") CC: stable@vger.kernel.org # 6.6+ Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Julian Sun <sunjunchao2870@gmail.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 9 Dec 2024 16:43:44 +0000 (16:43 +0000)]
btrfs: avoid monopolizing a core when activating a swap file
During swap activation we iterate over the extents of a file and we can
have many thousands of them, so we can end up in a busy loop monopolizing
a core. Avoid this by doing a voluntary reschedule after processing each
extent.
CC: stable@vger.kernel.org # 5.4+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 9 Dec 2024 16:31:41 +0000 (16:31 +0000)]
btrfs: allow swap activation to be interruptible
During swap activation we iterate over the extents of a file, then do
several checks for each extent, some of which may take some significant
time such as checking if an extent is shared. Since a file can have
many thousands of extents, this can be a very slow operation and it's
currently not interruptible. I had a bug during development of a previous
patch that resulted in an infinite loop when iterating the extents, so
a core was busy looping and I couldn't cancel the operation, which is very
annoying and requires a reboot. So make the loop interruptible by checking
for fatal signals at the end of each iteration and stopping immediately if
there is one.
CC: stable@vger.kernel.org # 5.4+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 9 Dec 2024 12:54:14 +0000 (12:54 +0000)]
btrfs: fix swap file activation failure due to extents that used to be shared
When activating a swap file, to determine if an extent is shared we use
can_nocow_extent(), which ends up at btrfs_cross_ref_exist(). That helper
is meant to be quick because it's used in the NOCOW write path, when
flushing delalloc and when doing a direct IO write, however it does return
some false positives, meaning it may indicate that an extent is shared
even if it's no longer the case. For the write path this is fine, we just
do a unnecessary COW operation instead of doing a more rigorous check
which would be too heavy (calling btrfs_is_data_extent_shared()).
However when activating a swap file, the false positives simply result
in a failure, which is confusing for users/applications. One particular
case where this happens is when a data extent only has 1 reference but
that reference is not inlined in the extent item located in the extent
tree - this happens when we create more than 33 references for an extent
and then delete those 33 references plus every other non-inline reference
except one. The function check_committed_ref() assumes that if the size
of an extent item doesn't match the size of struct btrfs_extent_item
plus the size of an inline reference (plus an owner reference in case
simple quotas are enabled), then the extent is shared - that is not the
case however, we can have a single reference but it's not inlined - the
reason we do this is to be fast and avoid inspecting non-inline references
which may be located in another leaf of the extent tree, slowing down
write paths.
The following test script reproduces the bug:
$ cat test.sh
#!/bin/bash
DEV=/dev/sdi
MNT=/mnt/sdi
NUM_CLONES=50
umount $DEV &> /dev/null
run_test()
{
local sync_after_add_reflinks=$1
local sync_after_remove_reflinks=$2
touch $MNT/foo
chmod 0600 $MNT/foo
# On btrfs the file must be NOCOW.
chattr +C $MNT/foo &> /dev/null
xfs_io -s -c "pwrite -b 1M 0 1M" $MNT/foo
mkswap $MNT/foo
for ((i = 1; i <= $NUM_CLONES; i++)); do
touch $MNT/foo_clone_$i
chmod 0600 $MNT/foo_clone_$i
# On btrfs the file must be NOCOW.
chattr +C $MNT/foo_clone_$i &> /dev/null
cp --reflink=always $MNT/foo $MNT/foo_clone_$i
done
if [ $sync_after_add_reflinks -ne 0 ]; then
# Flush delayed refs and commit current transaction.
sync -f $MNT
fi
# Remove the original file and all clones except the last.
rm -f $MNT/foo
for ((i = 1; i < $NUM_CLONES; i++)); do
rm -f $MNT/foo_clone_$i
done
if [ $sync_after_remove_reflinks -ne 0 ]; then
# Flush delayed refs and commit current transaction.
sync -f $MNT
fi
# Now use the last clone as a swap file. It should work since
# its extent are not shared anymore.
swapon $MNT/foo_clone_${NUM_CLONES}
swapoff $MNT/foo_clone_${NUM_CLONES}
umount $MNT
}
echo -e "\nTest without sync after creating and removing clones"
run_test 0 0
echo -e "\nTest with sync after creating clones"
run_test 1 0
echo -e "\nTest with sync after removing clones"
run_test 0 1
echo -e "\nTest with sync after creating and removing clones"
run_test 1 1
Running the test:
$ ./test.sh
Test without sync after creating and removing clones
wrote 1048576/1048576 bytes at offset 0
1 MiB, 1 ops; 0.0017 sec (556.793 MiB/sec and 556.7929 ops/sec)
Setting up swapspace version 1, size = 1020 KiB (1044480 bytes)
no label, UUID=a6b9c29e-5ef4-4689-a8ac-bc199c750f02
swapon: /mnt/sdi/foo_clone_50: swapon failed: Invalid argument
swapoff: /mnt/sdi/foo_clone_50: swapoff failed: Invalid argument
Test with sync after creating clones
wrote 1048576/1048576 bytes at offset 0
1 MiB, 1 ops; 0.0036 sec (271.739 MiB/sec and 271.7391 ops/sec)
Setting up swapspace version 1, size = 1020 KiB (1044480 bytes)
no label, UUID=5e9008d6-1f7a-4948-a1b4-3f30aba20a33
swapon: /mnt/sdi/foo_clone_50: swapon failed: Invalid argument
swapoff: /mnt/sdi/foo_clone_50: swapoff failed: Invalid argument
Test with sync after removing clones
wrote 1048576/1048576 bytes at offset 0
1 MiB, 1 ops; 0.0103 sec (96.665 MiB/sec and 96.6651 ops/sec)
Setting up swapspace version 1, size = 1020 KiB (1044480 bytes)
no label, UUID=916c2740-fa9f-4385-9f06-29c3f89e4764
Test with sync after creating and removing clones
wrote 1048576/1048576 bytes at offset 0
1 MiB, 1 ops; 0.0031 sec (314.268 MiB/sec and 314.2678 ops/sec)
Setting up swapspace version 1, size = 1020 KiB (1044480 bytes)
no label, UUID=06aab1dd-4d90-49c0-bd9f-3a8db4e2f912
swapon: /mnt/sdi/foo_clone_50: swapon failed: Invalid argument
swapoff: /mnt/sdi/foo_clone_50: swapoff failed: Invalid argument
Fix this by reworking btrfs_swap_activate() to instead of using extent
maps and checking for shared extents with can_nocow_extent(), iterate
over the inode's file extent items and use the accurate
btrfs_is_data_extent_shared().
CC: stable@vger.kernel.org # 5.4+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Fri, 29 Nov 2024 12:25:30 +0000 (12:25 +0000)]
btrfs: fix race with memory mapped writes when activating swap file
When activating the swap file we flush all delalloc and wait for ordered
extent completion, so that we don't miss any delalloc and extents before
we check that the file's extent layout is usable for a swap file and
activate the swap file. We are called with the inode's VFS lock acquired,
so we won't race with buffered and direct IO writes, however we can still
race with memory mapped writes since they don't acquire the inode's VFS
lock. The race window is between flushing all delalloc and locking the
whole file's extent range, since memory mapped writes lock an extent range
with the length of a page.
Fix this by acquiring the inode's mmap lock before we flush delalloc.
CC: stable@vger.kernel.org # 5.4+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Boris Burkov [Fri, 13 Dec 2024 20:33:22 +0000 (12:33 -0800)]
btrfs: check folio mapping after unlock in put_file_data()
When we call btrfs_read_folio() we get an unlocked folio, so it is possible
for a different thread to concurrently modify folio->mapping. We must
check that this hasn't happened once we do have the lock.
CC: stable@vger.kernel.org # 6.12+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Boris Burkov <boris@bur.io> Signed-off-by: David Sterba <dsterba@suse.com>
Boris Burkov [Fri, 13 Dec 2024 20:22:32 +0000 (12:22 -0800)]
btrfs: check folio mapping after unlock in relocate_one_folio()
When we call btrfs_read_folio() to bring a folio uptodate, we unlock the
folio. The result of that is that a different thread can modify the
mapping (like remove it with invalidate) before we call folio_lock().
This results in an invalid page and we need to try again.
In particular, if we are relocating concurrently with aborting a
transaction, this can result in a crash like the following:
This occurs because cleanup_one_transaction() calls
destroy_delalloc_inodes() which calls invalidate_inode_pages2() which
takes the folio_lock before setting mapping to NULL. We fail to check
this, and subsequently call set_extent_mapping(), which assumes that
mapping != NULL (in fact it asserts that in debug mode)
Note that the "fixes" patch here is not the one that introduced the
race (the very first iteration of this code from 2009) but a more recent
change that made this particular crash happen in practice.
Fixes: e7f1326cc24e ("btrfs: set page extent mapped after read_folio in relocate_one_page") CC: stable@vger.kernel.org # 6.1+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Boris Burkov <boris@bur.io> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 11 Dec 2024 16:08:07 +0000 (16:08 +0000)]
btrfs: fix use-after-free when COWing tree bock and tracing is enabled
When a COWing a tree block, at btrfs_cow_block(), and we have the
tracepoint trace_btrfs_cow_block() enabled and preemption is also enabled
(CONFIG_PREEMPT=y), we can trigger a use-after-free in the COWed extent
buffer while inside the tracepoint code. This is because in some paths
that call btrfs_cow_block(), such as btrfs_search_slot(), we are holding
the last reference on the extent buffer @buf so btrfs_force_cow_block()
drops the last reference on the @buf extent buffer when it calls
free_extent_buffer_stale(buf), which schedules the release of the extent
buffer with RCU. This means that if we are on a kernel with preemption,
the current task may be preempted before calling trace_btrfs_cow_block()
and the extent buffer already released by the time trace_btrfs_cow_block()
is called, resulting in a use-after-free.
Fix this by moving the trace_btrfs_cow_block() from btrfs_cow_block() to
btrfs_force_cow_block() before the COWed extent buffer is freed.
This also has a side effect of invoking the tracepoint in the tree defrag
code, at defrag.c:btrfs_realloc_node(), since btrfs_force_cow_block() is
called there, but this is fine and it was actually missing there.
btrfs: fix use-after-free waiting for encoded read endios
Fix a use-after-free in the I/O completion path for encoded reads by
using a completion instead of a wait_queue for synchronizing the
destruction of 'struct btrfs_encoded_read_private'.
Fixes: 1881fba89bd5 ("btrfs: add BTRFS_IOC_ENCODED_READ ioctl") CC: stable@vger.kernel.org # 6.1+ Reviewed-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Tue, 3 Dec 2024 11:53:27 +0000 (11:53 +0000)]
btrfs: flush delalloc workers queue before stopping cleaner kthread during unmount
During the unmount path, at close_ctree(), we first stop the cleaner
kthread, using kthread_stop() which frees the associated task_struct, and
then stop and destroy all the work queues. However after we stopped the
cleaner we may still have a worker from the delalloc_workers queue running
inode.c:submit_compressed_extents(), which calls btrfs_add_delayed_iput(),
which in turn tries to wake up the cleaner kthread - which was already
destroyed before, resulting in a use-after-free on the task_struct.
Syzbot reported this with the following stack traces:
BUG: KASAN: slab-use-after-free in __lock_acquire+0x78/0x2100 kernel/locking/lockdep.c:5089
Read of size 8 at addr ffff8880259d2818 by task kworker/u8:3/52
The buggy address belongs to the object at ffff8880259d1e00
which belongs to the cache task_struct of size 7424
The buggy address is located 2584 bytes inside of
freed 7424-byte region [ffff8880259d1e00, ffff8880259d3b00)
Commit e546fe1da9bd ("block: Rework bio_split() return value") changed
bio_split() so that it can return errors.
Add error handling for it in btrfs_split_bio() and ultimately
btrfs_submit_chunk(). As the bio is not submitted, the bio counter must
be decremented to pair btrfs_bio_counter_inc_blocked().
Reviewed-by: John Garry <john.g.garry@oracle.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Mon, 2 Dec 2024 21:47:53 +0000 (08:17 +1030)]
btrfs: properly wait for writeback before buffered write
[BUG]
Before commit e820dbeb6ad1 ("btrfs: convert btrfs_buffered_write() to
use folios"), function prepare_one_folio() will always wait for folio
writeback to finish before returning the folio.
However commit e820dbeb6ad1 ("btrfs: convert btrfs_buffered_write() to
use folios") changed to use FGP_STABLE to do the writeback wait, but
FGP_STABLE is calling folio_wait_stable(), which only calls
folio_wait_writeback() if the address space has AS_STABLE_WRITES, which
is not set for btrfs inodes.
This means we will not wait for the folio writeback at all.
[CAUSE]
The cause is FGP_STABLE is not waiting for writeback unconditionally, but
only for address spaces with AS_STABLE_WRITES, normally such flag is set
when the super block has SB_I_STABLE_WRITES flag.
Such super block flag is set when the block device has hardware digest
support or has internal checksum requirement.
I'd argue btrfs should set such super block due to its default data
checksum behavior, but it is not set yet, so this means FGP_STABLE flag
will have no effect at all.
(For NODATASUM inodes, we can skip the waiting in theory but that should
be an optimization in the future.)
This can lead to data checksum mismatch, as we can modify the folio
while it's still under writeback, this will make the contents differ
from the contents at submission and checksum calculation.
[FIX]
Instead of fully relying on FGP_STABLE, manually do the folio writeback
waiting, until we set the address space or super flag.
Fixes: e820dbeb6ad1 ("btrfs: convert btrfs_buffered_write() to use folios") Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Fri, 29 Nov 2024 13:33:03 +0000 (13:33 +0000)]
btrfs: fix missing snapshot drew unlock when root is dead during swap activation
When activating a swap file we acquire the root's snapshot drew lock and
then check if the root is dead, failing and returning with -EPERM if it's
dead but without unlocking the root's snapshot lock. Fix this by adding
the missing unlock.
Fixes: 60021bd754c6 ("btrfs: prevent subvol with swapfile from being deleted") Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Reviewed-by: David Sterba <dsterba@suse.com> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
mount $dev $mnt1 -o subvol=subvol1
while mount -o remount,ro $mnt1; do mount -o remount,rw $mnt1; done &
bg=$!
while mount $dev $mnt2 -o subvol=subvol2; do umount $mnt2; done
kill $bg
wait
umount -R $mnt1
umount -R $mnt2
The script will fail with the following error:
mount: /mnt/scratch: /dev/mapper/test-scratch1 already mounted on /mnt/test.
dmesg(1) may have more information after failed mount system call.
umount: /mnt/test: target is busy.
umount: /mnt/scratch/: not mounted
And there is no kernel error message.
[CAUSE]
During the btrfs mount, to support mounting different subvolumes with
different RO/RW flags, we need to detect that and retry if needed:
Retry with matching RO flags if the initial mount fail with -EBUSY.
The problem is, during that retry we do not hold any super block lock
(s_umount), this means there can be a remount process changing the RO
flags of the original fs super block.
If so, we can have an EBUSY error during retry. And this time we treat
any failure as an error, without any retry and cause the above EBUSY
mount failure.
[FIX]
The current retry behavior is racy because we do not have a super block
thus no way to hold s_umount to prevent the race with remount.
Solve the root problem by allowing fc->sb_flags to mismatch from the
sb->s_flags at btrfs_get_tree_super().
Then at the re-entry point btrfs_get_tree_subvol(), manually check the
fc->s_flags against sb->s_flags, if it's a RO->RW mismatch, then
reconfigure with s_umount lock hold.
Reported-by: Enno Gotthold <egotthold@suse.com> Reported-by: Fabian Vogt <fvogt@suse.com>
[ Special thanks for the reproducer and early analysis pointing to btrfs. ] Fixes: f044b318675f ("btrfs: handle the ro->rw transition for mounting different subvolumes") Link: https://bugzilla.suse.com/show_bug.cgi?id=1231836 Signed-off-by: Qu Wenruo <wqu@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Mark Harmstone [Fri, 15 Nov 2024 15:49:17 +0000 (15:49 +0000)]
btrfs: fix lockdep warnings on io_uring encoded reads
Lockdep doesn't like the fact that btrfs_uring_read_extent() returns to
userspace still holding the inode lock, even though we release it once
the I/O finishes. Add calls to rwsem_release() and rwsem_acquire_read() to
work round this.
Reported-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> 34310c442e17 ("btrfs: add io_uring command for encoded reads (ENCODED_READ ioctl)") Signed-off-by: Mark Harmstone <maharmstone@fb.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Fri, 15 Nov 2024 11:29:21 +0000 (11:29 +0000)]
btrfs: ref-verify: fix use-after-free after invalid ref action
At btrfs_ref_tree_mod() after we successfully inserted the new ref entry
(local variable 'ref') into the respective block entry's rbtree (local
variable 'be'), if we find an unexpected action of BTRFS_DROP_DELAYED_REF,
we error out and free the ref entry without removing it from the block
entry's rbtree. Then in the error path of btrfs_ref_tree_mod() we call
btrfs_free_ref_cache(), which iterates over all block entries and then
calls free_block_entry() for each one, and there we will trigger a
use-after-free when we are called against the block entry to which we
added the freed ref entry to its rbtree, since the rbtree still points
to the block entry, as we didn't remove it from the rbtree before freeing
it in the error path at btrfs_ref_tree_mod(). Fix this by removing the
new ref entry from the rbtree before freeing it.
Syzbot report this with the following stack traces:
The buggy address belongs to the object at ffff888042d1af00
which belongs to the cache kmalloc-64 of size 64
The buggy address is located 56 bytes inside of
freed 64-byte region [ffff888042d1af00, ffff888042d1af40)
Memory state around the buggy address: ffff888042d1ae00: fa fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc ffff888042d1ae80: 00 00 00 00 00 fc fc fc fc fc fc fc fc fc fc fc
>ffff888042d1af00: fa fb fb fb fb fb fb fb fc fc fc fc fc fc fc fc
^ ffff888042d1af80: 00 00 00 00 00 00 fc fc fc fc fc fc fc fc fc fc ffff888042d1b000: 00 00 00 00 00 fc fc 00 00 00 00 00 fc fc 00 00
Reported-by: syzbot+7325f164162e200000c1@syzkaller.appspotmail.com Link: https://lore.kernel.org/linux-btrfs/673723eb.050a0220.1324f8.00a8.GAE@google.com/T/#u Fixes: fd708b81d972 ("Btrfs: add a extent ref verify tool") CC: stable@vger.kernel.org # 4.19+ Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Lizhi Xu [Fri, 25 Oct 2024 04:55:53 +0000 (12:55 +0800)]
btrfs: add a sanity check for btrfs root in btrfs_search_slot()
Syzbot reports a null-ptr-deref in btrfs_search_slot().
The reproducer is using rescue=ibadroots, and the extent tree root is
corrupted thus the extent tree is NULL.
When scrub tries to search the extent tree to gather the needed extent
info, btrfs_search_slot() doesn't check if the target root is NULL or
not, resulting the null-ptr-deref.
Add sanity check for btrfs root before using it in btrfs_search_slot().
Filipe Manana [Fri, 15 Nov 2024 15:46:13 +0000 (15:46 +0000)]
btrfs: don't loop for nowait writes when checking for cross references
When checking for delayed refs when verifying if there are cross
references for a data extent, we stop if the path has nowait set and we
can't try lock the delayed ref head's mutex, returning -EAGAIN with the
goal of making a write fallback to a blocking context. However we ignore
the -EAGAIN at btrfs_cross_ref_exist() when check_delayed_ref() returns
it, and keep looping instead of immediately returning the -EAGAIN to the
caller.
Fix this by not looping if we get -EAGAIN and we have a nowait path.
Fixes: 26ce91144631 ("btrfs: make can_nocow_extent nowait compatible") CC: stable@vger.kernel.org # 6.1+ Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Thu, 14 Nov 2024 15:38:27 +0000 (15:38 +0000)]
btrfs: sysfs: advertise experimental features only if CONFIG_BTRFS_EXPERIMENTAL=y
We are advertising experimental features through sysfs if
CONFIG_BTRFS_DEBUG is set, without looking if CONFIG_BTRFS_EXPERIMENTAL
is set. This is wrong as it will result in reporting experimental
features as supported when CONFIG_BTRFS_EXPERIMENTAL is not set but
CONFIG_BTRFS_DEBUG is set.
Fix this by checking for CONFIG_BTRFS_EXPERIMENTAL instead of
CONFIG_BTRFS_DEBUG.
Fixes: 67cd3f221769 ("btrfs: split out CONFIG_BTRFS_EXPERIMENTAL from CONFIG_BTRFS_DEBUG") Reviewed-by: Neal Gompa <neal@gompa.dev> Reviewed-by: Qu Wenruo <wqu@suse.com> Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Thu, 14 Nov 2024 12:11:26 +0000 (12:11 +0000)]
btrfs: fix deadlock between transaction commits and extent locks
When running a workload with fsstress and duperemove (generic/561) we can
hit a deadlock related to transaction commits and locking extent ranges,
as described below.
Task A hanging during a transaction commit, waiting for all other writers
to complete:
$ cat -n fs/btrfs/transaction.c
(...)
2162 int btrfs_commit_transaction(struct btrfs_trans_handle *trans)
2163 {
(...)
2349 spin_lock(&fs_info->trans_lock);
2350 add_pending_snapshot(trans);
2351 cur_trans->state = TRANS_STATE_COMMIT_DOING;
2352 spin_unlock(&fs_info->trans_lock);
2353
2354 /*
2355 * The thread has started/joined the transaction thus it holds the
2356 * lockdep map as a reader. It has to release it before acquiring the
2357 * lockdep map as a writer.
2358 */
2359 btrfs_lockdep_release(fs_info, btrfs_trans_num_writers);
2360 btrfs_might_wait_for_event(fs_info, btrfs_trans_num_writers);
2361 wait_event(cur_trans->writer_wait,
2362 atomic_read(&cur_trans->num_writers) == 1);
(...)
The transaction is in the TRANS_STATE_COMMIT_DOING state and so it's
waiting for all other existing writers to complete and release their
transaction handle.
Task B is running ordered extent completion and blocked waiting to lock an
extent range in an inode's io tree:
This extent range locking happens after joining the current transaction,
so task A is waiting for task B to release its transaction handle
(decrementing the transaction's num_writers counter).
Task C while doing a fiemap it tries to join the current transaction:
It tries to join the current transaction, but it can't because the
transaction is in the TRANS_STATE_COMMIT_DOING state, so
join_transaction() returns -EBUSY to start_transaction() and makes it
wait for the current transaction to complete. And while it's waiting
for the transaction to complete, it's holding an extent range locked
in the same inode that task B is operating, which causes a deadlock
between these 3 tasks. The extent range for the inode was locked at
the start of the fiemap operation, early at extent_fiemap().
In short these tasks deadlock because:
1) Task A is waiting for task B to release its transaction handle;
2) Task B is waiting to lock an extent range for an inode while holding a
transaction handle open;
3) Task C is waiting for the current transaction to complete (for task A
to finish the transaction commit) while holding the extent range for
the inode locked, so task B can't progress and release its transaction
handle.
This results in an ABBA deadlock involving transaction commits and extent
locks. Extent locks are higher level locks, like inode VFS locks, and
should always be acquired before joining or starting a transaction, but
recently commit 2206265f41e9 ("btrfs: remove code duplication in ordered
extent finishing") accidentally changed btrfs_finish_one_ordered() to do
the transaction join before locking the extent range.
Fix this by making sure that btrfs_finish_one_ordered() always locks the
extent before joining a transaction and add an explicit comment about the
need for this order.
Fixes: 2206265f41e9 ("btrfs: remove code duplication in ordered extent finishing") Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
The buggy address belongs to the object at ffff888106a83f00
which belongs to the cache kmalloc-rnd-07-96 of size 96
The buggy address is located 24 bytes inside of
freed 96-byte region [ffff888106a83f00, ffff888106a83f60)
Memory state around the buggy address: ffff888106a83e00: fa fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc ffff888106a83e80: fa fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc
>ffff888106a83f00: fa fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc
^ ffff888106a83f80: fa fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc ffff888106a84000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
==================================================================
Further analyzing the trace and the crash dump's vmcore file shows that
the wake_up() call in btrfs_encoded_read_endio() is calling wake_up() on
the wait_queue that is in the private data passed to the end_io handler.
Commit 4ff47df40447 ("btrfs: move priv off stack in
btrfs_encoded_read_regular_fill_pages()") moved 'struct
btrfs_encoded_read_private' off the stack.
Before that commit one can see a corruption of the private data when
analyzing the vmcore after a crash:
Here we can see several indicators of in-memory data corruption, e.g. the
large negative atomic values of ->pending or
->wait->lock->rlock->raw_lock->val, as well as the bogus spinlock magic
0x1ff7ae32 (decimal 536325682 above) instead of 0xdead4ead or the bogus
pointer values for ->wait->head.
To fix this, change atomic_dec_return() to atomic_dec_and_test() to fix the
corruption, as atomic_dec_return() is defined as two instructions on
x86_64, whereas atomic_dec_and_test() is defined as a single atomic
operation. This can lead to a situation where counter value is already
decremented but the if statement in btrfs_encoded_read_endio() is not
completely processed, i.e. the 0 test has not completed. If another thread
continues executing btrfs_encoded_read_regular_fill_pages() the
atomic_dec_return() there can see an already updated ->pending counter and
continues by freeing the private data. Continuing in the endio handler the
test for 0 succeeds and the wait_queue is woken up, resulting in a
use-after-free.
Reported-by: Shinichiro Kawasaki <shinichiro.kawasaki@wdc.com> Suggested-by: Damien Le Moal <Damien.LeMoal@wdc.com> Fixes: 1881fba89bd5 ("btrfs: add BTRFS_IOC_ENCODED_READ ioctl") CC: stable@vger.kernel.org # 6.1+ Reviewed-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 6 Nov 2024 11:26:07 +0000 (11:26 +0000)]
btrfs: send: check for read-only send root under critical section
We're checking if the send root is read-only without being under the
protection of the root's root_item_lock spinlock, which is what protects
the root's flags when clearing the read-only flag, done at
btrfs_ioctl_subvol_setflags(). Furthermore, it should be done in the
same critical section that increments the root's send_in_progress counter,
as btrfs_ioctl_subvol_setflags() clears the read-only flag in the same
critical section that checks the counter's value.
So fix this by moving the read-only check under the critical section
delimited by the root's root_item_lock which also increments the root's
send_in_progress counter.
Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 6 Nov 2024 10:32:09 +0000 (10:32 +0000)]
btrfs: send: check for dead send root under critical section
We're checking if the send root is dead without the protection of the
root's root_item_lock spinlock, which is what protects the root's flags.
The inverse, setting the dead flag on a root, is done under the protection
of that lock, at btrfs_delete_subvolume(). Also checking and updating the
root's send_in_progress counter is supposed to be done in the same
critical section as checking for or setting the root dead flag, so that
these operations are done atomically as a single step (which is correctly
done by btrfs_delete_subvolume()).
So fix this by checking if the send root is dead in the same critical
section that updates the send_in_progress counter, which is protected by
the root's root_item_lock spinlock.
Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 6 Nov 2024 12:21:13 +0000 (12:21 +0000)]
btrfs: remove check for NULL fs_info at btrfs_folio_end_lock_bitmap()
Smatch complains about possibly dereferencing a NULL fs_info at
btrfs_folio_end_lock_bitmap():
fs/btrfs/subpage.c:332 btrfs_folio_end_lock_bitmap() warn: variable dereferenced before check 'fs_info' (see line 326)
because we access fs_info to set the 'start_bit' variable before doing the
check for a NULL fs_info.
However fs_info is never NULL, since in the only caller of
btrfs_folio_end_lock_bitmap() is extent_writepage(), where we have an
inode which always as a non-NULL fs_info.
So remove the check for a NULL fs_info at btrfs_folio_end_lock_bitmap().
Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Tue, 5 Nov 2024 14:38:38 +0000 (14:38 +0000)]
btrfs: simplify logic to decrement snapshot counter at btrfs_mksnapshot()
There's no point in having a 'snapshot_force_cow' variable to track if we
need to decrement the root->snapshot_force_cow counter, as we never jump
to the 'out' label after incrementing the counter. Simplify this by
removing the variable and always decrementing the counter before the 'out'
label, right after the call to btrfs_mksubvol().
Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Move the 'count' field, which is 4 bytes long, to just below the ref count
field, so we eliminate the hole and reduce the structure size from 320
bytes down to 312 bytes:
Filipe Manana [Mon, 4 Nov 2024 11:50:14 +0000 (11:50 +0000)]
btrfs: update stale comment for struct btrfs_delayed_ref_node::add_list
The comment refers to a list in the respective delayed ref head that no
longer exists (ref_list), it was replaced with a rbtree (ref_tree) in
commit 0e0adbcfdc90 ("btrfs: track refs in a rb_tree instead of a list").
So update the stale comment to refer to the rbtree instead of the old
list.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
David Sterba [Mon, 21 Oct 2024 19:28:26 +0000 (21:28 +0200)]
btrfs: add new ioctl to wait for cleaned subvolumes
Add a new unprivileged ioctl that will let the command
'btrfs subvolume sync' work without the (privileged) SEARCH_TREE ioctl.
There are several modes of operation, where the most common ones are to
wait on a specific subvolume or all currently queued for cleaning. This
is utilized e.g. in backup applications that delete subvolumes and wait
until they're cleaned to check for remaining space.
The other modes are for flexibility, e.g. for monitoring or
checkpoints in the queue of deleted subvolumes, again without the need
to use SEARCH_TREE.
Notes:
- waiting is interruptible, the timeout is set to 1 second and is not
configurable
- repeated calls to the ioctl see a different state, so this is
inherently racy when using e.g. the count or peek next/last
Use cases:
- a subvolume A was deleted, wait for cleaning (WAIT_FOR_ONE)
- a bunch of subvolumes were deleted, wait for all (WAIT_FOR_QUEUED or
PEEK_LAST + WAIT_FOR_ONE)
- count how many are queued (not blocking), for monitoring purposes
- report progress (PEEK_NEXT), may miss some if cleaning is quick
- own waiting in user space (PEEK_LAST until it's 0)
Haisu Wang [Fri, 25 Oct 2024 06:54:41 +0000 (14:54 +0800)]
btrfs: simplify range tracking in cow_file_range()
Simplify tracking of the range processed by using cur_alloc_size only to
store the reserved part that may fail to the allocated extent. Remove
the ram_size as well since it is always equal to cur_alloc_size in the
context. Advance the start in normal path until extent allocation
succeeds and keep the start unchanged in the error handling path.
Passed the fstest generic/475 test for a hundred times with quota
enabled. And a modified generic/475 test by removing the sleep time
for a hundred times. About one tenth of the tests do enter the error
handling path due to fail to reserve extent.
Suggested-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Haisu Wang <haisuwang@tencent.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Leo Martins [Fri, 30 Aug 2024 20:24:55 +0000 (13:24 -0700)]
btrfs: remove conditional path allocation in btrfs_read_locked_inode()
Remove conditional path allocation from btrfs_read_locked_inode(). Add
an ASSERT(path) to indicate it should never be called with a NULL path.
Call btrfs_read_locked_inode() directly from btrfs_iget(). This causes
code duplication between btrfs_iget() and btrfs_iget_path(), but I
think this is justifiable as it removes the need for conditionally
allocating the path inside of btrfs_read_locked_inode(). This makes the
code easier to reason about and makes it clear who has the
responsibility of allocating and freeing the path.
Signed-off-by: Leo Martins <loemra.dev@gmail.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Leo Martins [Fri, 30 Aug 2024 20:24:54 +0000 (13:24 -0700)]
btrfs: push cleanup into btrfs_read_locked_inode()
Move btrfs_add_inode_to_root() so it can be called from
btrfs_read_locked_inode(), no changes were made to the function.
Move cleanup code from btrfs_iget_path() to btrfs_read_locked_inode.
This improves readability and improves a leaky abstraction. Previously
btrfs_iget_path() had to handle a positive error case as a result of a
call to btrfs_search_slot(), but it makes more sense to handle this
closer to the source of the call.
Signed-off-by: Leo Martins <loemra.dev@gmail.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Pavel Begunkov [Mon, 4 Nov 2024 16:12:04 +0000 (16:12 +0000)]
io_uring/cmd: let cmds to know about dying task
When the taks that submitted a request is dying, a task work for that
request might get run by a kernel thread or even worse by a half
dismantled task. We can't just cancel the task work without running the
callback as the cmd might need to do some clean up, so pass a flag
instead. If set, it's not safe to access any task resources and the
callback is expected to cancel the cmd ASAP.
Reviewed-by: Jens Axboe <axboe@kernel.dk> Reviewed-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: David Sterba <dsterba@suse.com>
Mark Harmstone [Thu, 31 Oct 2024 16:03:56 +0000 (16:03 +0000)]
btrfs: add struct io_btrfs_cmd as type for io_uring_cmd_to_pdu()
Add struct io_btrfs_cmd as a wrapper type for io_uring_cmd_to_pdu(),
rather than using a raw pointer.
Suggested-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Mark Harmstone <maharmstone@fb.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Mark Harmstone [Tue, 22 Oct 2024 14:50:20 +0000 (15:50 +0100)]
btrfs: add io_uring command for encoded reads (ENCODED_READ ioctl)
Add an io_uring command for encoded reads, using the same interface as
the existing BTRFS_IOC_ENCODED_READ ioctl.
btrfs_uring_encoded_read() is an io_uring version of
btrfs_ioctl_encoded_read(), which validates the user input and calls
btrfs_encoded_read() to read the appropriate metadata. If we determine
that we need to read an extent from disk, we call
btrfs_encoded_read_regular_fill_pages() through
btrfs_uring_read_extent() to prepare the bio.
The existing btrfs_encoded_read_regular_fill_pages() is changed so that
if it is passed a valid uring_ctx, rather than waking up any waiting
threads it calls btrfs_uring_read_extent_endio(). This in turn copies
the read data back to userspace, and calls io_uring_cmd_done() to
complete the io_uring command.
Because we're potentially doing a non-blocking read,
btrfs_uring_read_extent() doesn't clean up after itself if it returns
-EIOCBQUEUED. Instead, it allocates a priv struct, populates the fields
there that we will need to unlock the inode and free our allocations,
and defers this to the btrfs_uring_read_finished() that gets called when
the bio completes.
Signed-off-by: Mark Harmstone <maharmstone@fb.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Mark Harmstone [Tue, 22 Oct 2024 14:50:19 +0000 (15:50 +0100)]
btrfs: move priv off stack in btrfs_encoded_read_regular_fill_pages()
Change btrfs_encoded_read_regular_fill_pages() so that the priv struct
is allocated rather than stored on the stack, in preparation for adding
an asynchronous mode to the function.
Signed-off-by: Mark Harmstone <maharmstone@fb.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Mark Harmstone [Tue, 22 Oct 2024 14:50:18 +0000 (15:50 +0100)]
btrfs: don't sleep in btrfs_encoded_read() if IOCB_NOWAIT is set
Change btrfs_encoded_read() so that it returns -EAGAIN rather than sleeps
if IOCB_NOWAIT is set in iocb->ki_flags. The conditions that require
sleeping are: inode lock, writeback, extent lock, ordered range.
Signed-off-by: Mark Harmstone <maharmstone@fb.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Mark Harmstone [Tue, 22 Oct 2024 14:50:17 +0000 (15:50 +0100)]
btrfs: change btrfs_encoded_read() so that reading of extent is done by caller
Change the behaviour of btrfs_encoded_read() so that if it needs to read
an extent from disk, it leaves the extent and inode locked and returns
-EIOCBQUEUED. The caller is then responsible for doing the I/O via
btrfs_encoded_read_regular() and unlocking the extent and inode.
Signed-off-by: Mark Harmstone <maharmstone@fb.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Thu, 24 Oct 2024 11:07:51 +0000 (12:07 +0100)]
btrfs: remove no longer used delayed ref head search functionality
After the previous patch, which converted the rb-tree used to track
delayed ref heads into an xarray, the find_ref_head() function is now
used only by one caller which always passes false to the 'return_bigger'
argument. So remove the 'return_bigger' logic, simplifying the function,
and move all the function code to the single caller.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 23 Oct 2024 15:27:14 +0000 (16:27 +0100)]
btrfs: track delayed ref heads in an xarray
Currently we use a red black tree (rb-tree) to track the delayed ref
heads (in struct btrfs_delayed_ref_root::href_root). This however is not
very efficient when the number of delayed ref heads is large (and it's
very common to be at least in the order of thousands) since rb-trees are
binary trees. For example for 10K delayed ref heads, the tree has a depth
of 13. Besides that, inserting into the tree requires navigating through
it and pulling useless cache lines in the process since the red black tree
nodes are embedded within the delayed ref head structure - on the other
hand, by being embedded, it requires no extra memory allocations.
We can improve this by using an xarray instead which has a much higher
branching factor than a red black tree (binary balanced tree) and is more
cache friendly and behaves like a resizable array, with a much better
search and insertion complexity than a red black tree. This only has one
small disadvantage which is that insertion will sometimes require
allocating memory for the xarray - which may fail (not that often since
it uses a kmem_cache) - but on the other hand we can reduce the delayed
ref head structure size by 24 bytes (from 152 down to 128 bytes) after
removing the embedded red black tree node, meaning than we can now fit
32 delayed ref heads per 4K page instead of 26, and that gain compensates
for the occasional memory allocations needed for the xarray nodes. We
also end up using only 2 cache lines instead of 3 per delayed ref head.
Running the following fs_mark test showed some improvements:
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 23 Oct 2024 13:03:44 +0000 (14:03 +0100)]
btrfs: add comments regarding locking to struct btrfs_delayed_ref_root
Add some comments to struct btrfs_delayed_ref_root's fields to mention
what its spinlock protects.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 23 Oct 2024 11:48:15 +0000 (12:48 +0100)]
btrfs: assert delayed refs lock is held at add_delayed_ref_head()
The delayed refs lock must be held when calling add_delayed_ref_head(),
so assert that it's being held.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 23 Oct 2024 11:43:35 +0000 (12:43 +0100)]
btrfs: assert delayed refs lock is held at find_first_ref_head()
The delayed refs lock must be held when calling find_first_ref_head(), so
assert that it's being held.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 23 Oct 2024 11:41:17 +0000 (12:41 +0100)]
btrfs: assert delayed refs lock is held at find_ref_head()
We have 3 callers for find_ref_head() so assert at find_ref_head() that we
have the delayed refs lock held, removing the assertion from one of its
callers (btrfs_find_delayed_ref_head()).
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Tue, 22 Oct 2024 12:39:11 +0000 (13:39 +0100)]
btrfs: pass fs_info to btrfs_delete_ref_head()
One of the following patches in the series will need to access fs_info at
btrfs_delete_ref_head(), so pass a fs_info argument to it.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 21 Oct 2024 15:08:18 +0000 (16:08 +0100)]
btrfs: pass fs_info to functions that search for delayed ref heads
One of the following patches in the series will need to access fs_info in
the function find_ref_head(), so pass a fs_info argument to it as well as
to the functions btrfs_select_ref_head() and btrfs_find_delayed_ref_head()
which call find_ref_head().
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 21 Oct 2024 11:52:55 +0000 (12:52 +0100)]
btrfs: move delayed ref head unselection to delayed-ref.c
The unselect_delayed_ref_head() at extent-tree.c doesn't really belong in
that file as it's a delayed refs specific detail and therefore should be
at delayed-ref.c. Further its inverse, btrfs_select_ref_head(), is at
delayed-ref.c, so it only makes sense to have it there too.
So move unselect_delayed_ref_head() into delayed-ref.c and rename it to
btrfs_unselect_ref_head() so that its name closely matches its inverse
(btrfs_select_ref_head()).
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 21 Oct 2024 11:40:08 +0000 (12:40 +0100)]
btrfs: simplify obtaining a delayed ref head
Instead of doing it in two steps outside of delayed-ref.c, leaking low
level details such as locking, move the logic entirely to delayed-ref.c
under btrfs_select_ref_head(), reducing code and making things simpler
for the caller.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 21 Oct 2024 11:32:55 +0000 (12:32 +0100)]
btrfs: change return type of btrfs_delayed_ref_lock() to boolean
The function only returns 0, meaning it was able to lock the delayed ref
head, or -EAGAIN in case it wasn't able to lock it. So simplify this and
use a boolean return type instead, returning true if it was able to lock
and false otherwise.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 23 Oct 2024 13:14:11 +0000 (14:14 +0100)]
btrfs: remove num_entries atomic counter from delayed ref root
The atomic counter 'num_entries' is not used anymore, we increment it
and decrement it but then we don't ever read it to use for any logic.
Its last use was removed with commit 61a56a992fcf ("btrfs: delayed refs
pre-flushing should only run the heads we have"). So remove it.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Fri, 18 Oct 2024 10:29:37 +0000 (11:29 +0100)]
btrfs: use helper to find first ref head at btrfs_destroy_delayed_refs()
Instead of open coding it, use the find_first_ref_head() helper at
btrfs_destroy_delayed_refs(). This avoids duplicating the logic,
specially with the upcoming changes in subsequent patches.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 21 Oct 2024 11:10:26 +0000 (12:10 +0100)]
btrfs: remove duplicated code to drop delayed ref during transaction abort
When destroying delayed refs during a transaction abort, we have open
coded the removal of a delayed ref, which is also done by the static
helper function drop_delayed_ref(). So remove that duplicated code and
use drop_delayed_ref() instead.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 21 Oct 2024 16:12:22 +0000 (17:12 +0100)]
btrfs: remove fs_info parameter from btrfs_cleanup_one_transaction()
The fs_info parameter is redundant because it can be extracted from the
transaction given as another parameter. So remove it and use the fs_info
accessible from the transaction.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Mon, 21 Oct 2024 16:03:46 +0000 (17:03 +0100)]
btrfs: remove fs_info parameter from btrfs_destroy_delayed_refs()
The fs_info parameter is redundant because it can be extracted from the
transaction given as another parameter. So remove it and use the fs_info
accessible from the transaction.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Thu, 17 Oct 2024 15:23:41 +0000 (16:23 +0100)]
btrfs: move btrfs_destroy_delayed_refs() to delayed-ref.c
It's better suited at delayed-ref.c since it's about delayed refs and
contains logic to iterate over them (using the red black tree, doing all
the locking, freeing, etc), so move it from disk-io.c, which is pretty
big, into delayed-ref.c, hiding implementation details of how delayed
refs are tracked and managed. This also facilitates the next patches in
the series.
This change moves the code between files but also does the following
simple cleanups:
1) Rename the 'cache' variable to 'bg', since it's a block group
(the 'cache' logic comes from old days where the block group
structure was named 'btrfs_block_group_cache');
2) Move the 'ref' variable declaration to the scope of the inner
while loop, since it's not used outside that loop.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Thu, 17 Oct 2024 14:07:45 +0000 (15:07 +0100)]
btrfs: remove BUG_ON() at btrfs_destroy_delayed_refs()
At btrfs_destroy_delayed_refs() it's unexpected to not find the block
group to which a delayed reference's extent belongs to, so we have this
BUG_ON(), not just because it's highly unexpected but also because we
don't know what to do there.
Since we are in the transaction abort path, there's nothing we can do
other than proceed and cleanup all used resources we can. So remove
the BUG_ON() and deal with a missing block group by logging an error
message and continuing to cleanup all we can related to the current
delayed ref head and moving to other delayed refs.
Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Robbie Ko [Thu, 24 Oct 2024 02:31:42 +0000 (10:31 +0800)]
btrfs: reduce extent tree lock contention when searching for inline backref
When inserting extent backref, in order to check whether refs other than
inline refs are used, we always use path keep locks for tree search, which
will increase the lock contention of extent tree.
We do not need the parent node every time to determine whether normal
refs are used. It is only needed when the extent item is the last item
in a leaf.
Therefore, we change it to first use keep_locks=0 for search. If the
extent item happens to be the last item in the leaf, we then change to
keep_locks=1 for the second search to reduce lock contention.
Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Robbie Ko <robbieko@synology.com> Signed-off-by: David Sterba <dsterba@suse.com>
btrfs: tests: implement case for partial RAID stripe-tree delete
Implement self-tests for partial deletion of RAID stripe-tree entries.
These two new tests cover both the deletion of the front of a RAID
stripe-tree stripe extent as well as truncation of an item to make it
smaller.
Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com>
This ASSERT()ion triggers, because for the initial design of RAID
stripe-tree, I had the "one ordered-extent equals one bio" rule of zoned
btrfs in mind.
But for a RAID stripe-tree based system, that is not hosted on a zoned
storage device, but on a regular device this rule doesn't apply.
So in case the range we want to delete starts in the middle of the
previous item, grab the item and "truncate" it's length. That is, clone
the item, subtract the deleted portion from the key's offset, delete the
old item and insert the new one.
In case the range to delete ends in the middle of an item, we have to
adjust both the item's key as well as the stripe extents and then
re-insert the modified clone into the tree after deleting the old stripe
extent.
Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo [Thu, 10 Oct 2024 04:46:13 +0000 (15:16 +1030)]
btrfs: convert btrfs_buffered_write() to use folios
The buffered write path is still heavily utilizing the page interface.
Since we have converted it to do a page-by-page copying, it's much easier
to convert all involved functions to folio interface, this involves:
All function are changed to accept a folio parameter, and if the word
"page" is in the function name, change that to "folio" too.
The function btrfs_dirty_page() is exported for v1 space cache, convert
v1 cache call site to convert its page to folio for the new interface.
And there is a small enhancement for prepare_one_folio(), instead of
manually waiting for the page writeback, let __filemap_get_folio() to
handle that by using FGP_WRITEBEGIN, which implies
(FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE).
Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Mark Harmstone [Tue, 15 Oct 2024 11:37:29 +0000 (12:37 +0100)]
btrfs: fix wrong sizeof in btrfs_do_encoded_write()
btrfs_do_encoded_write() was converted to use folios in 400b172b8cdc,
but we're still allocating based on sizeof(struct page *) rather than
sizeof(struct folio *). There's no functional change.
Reviewed-by: Boris Burkov <boris@bur.io> Signed-off-by: Mark Harmstone <maharmstone@fb.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Boris Burkov [Tue, 15 Oct 2024 21:27:32 +0000 (14:27 -0700)]
btrfs: do not clear read-only when adding sprout device
If you follow the seed/sprout wiki, it suggests the following workflow:
btrfstune -S 1 seed_dev
mount seed_dev mnt
btrfs device add sprout_dev
mount -o remount,rw mnt
The first mount mounts the FS readonly, which results in not setting
BTRFS_FS_OPEN, and setting the readonly bit on the sb. The device add
somewhat surprisingly clears the readonly bit on the sb (though the
mount is still practically readonly, from the users perspective...).
Finally, the remount checks the readonly bit on the sb against the flag
and sees no change, so it does not run the code intended to run on
ro->rw transitions, leaving BTRFS_FS_OPEN unset.
As a result, when the cleaner_kthread runs, it sees no BTRFS_FS_OPEN and
does no work. This results in leaking deleted snapshots until we run out
of space.
I propose fixing it at the first departure from what feels reasonable:
when we clear the readonly bit on the sb during device add.
A new fstest I have written reproduces the bug and confirms the fix.
Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Boris Burkov <boris@bur.io> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 16 Oct 2024 10:13:24 +0000 (11:13 +0100)]
btrfs: remove local generation variable from read_block_for_search()
It's redundant to have the 'gen' variable since we already have the same
value in the local btrfs_tree_parent_check structure. So remove it and
instead use the structure's field.
Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 16 Oct 2024 10:09:26 +0000 (11:09 +0100)]
btrfs: remove redundant initializations for struct btrfs_tree_parent_check
It's pointless to initialize the has_first_key field of the stack local
btrfs_tree_parent_check structure at btrfs_tree_parent_check() and at
btrfs_qgroup_trace_subtree() since all fields not explicitly initialized
are zeroed out. In the case of the first function it's a bit odd because
we are assigning 0 and the field is of type bool, however not incorrect
since a 0 is converted to false.
Just remove the explicit initializations due to their redundancy.
Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 16 Oct 2024 09:57:48 +0000 (10:57 +0100)]
btrfs: simplify arguments for btrfs_verify_level_key()
The only caller of btrfs_verify_level_key() is read_block_for_search() and
it's passing 3 arguments to it that can be extracted from its on stack
variable of type struct btrfs_tree_parent_check.
So change btrfs_verify_level_key() to accept an argument of type
struct btrfs_tree_parent_check instead of level, first key and parent
transid arguments.
Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 16 Oct 2024 09:45:47 +0000 (10:45 +0100)]
btrfs: remove redundant level argument from read_block_for_search()
The level parameter passed to read_block_for_search() always matches the
level of the extent buffer passed in the "eb_ret" parameter, which we are
also extracting into the "parent_level" local variable.
So remove the level parameter and instead use the "parent_level" variable
which in fact has a better name (it's the level of the parent node from
which we are reading a child node/leaf).
Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Tue, 17 Sep 2024 11:03:11 +0000 (12:03 +0100)]
btrfs: re-enable the extent map shrinker
Now that the extent map shrinker can only be run by a single task and runs
asynchronously as a work queue job, enable it as it can no longer cause
stalls on tasks allocating memory and entering the extent map shrinker
through the fs shrinker (implemented by btrfs_free_cached_objects()).
This is crucial to prevent exhaustion of memory due to unbounded extent
map creation, primarily with direct IO but also for buffered IO on files
with holes. This problem, for the direct IO case, was first reported in
the Link tag below. That report was added to a Link tag of the first patch
that introduced the extent map shrinker, commit 956a17d9d050 ("btrfs: add
a shrinker for extent maps"), however the Link tag disappeared somehow
from the committed patch (but was included in the submitted patch to the
mailing list), so adding it below for future reference.
Filipe Manana [Thu, 5 Sep 2024 10:31:49 +0000 (11:31 +0100)]
btrfs: rename extent map shrinker members from struct btrfs_fs_info
The names for the members of struct btrfs_fs_info related to the extent
map shrinker are a bit too long, so rename them to be shorter by replacing
the "extent_map_" prefix with the "em_" prefix.
Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Wed, 4 Sep 2024 16:03:43 +0000 (17:03 +0100)]
btrfs: simplify tracking progress for the extent map shrinker
Now that the extent map shrinker can only be run by a single task (as a
work queue item) there is no need to keep the progress of the shrinker
protected by a spinlock and passing the progress to trace events as
parameters. So remove the lock and simplify the arguments for the trace
events.
Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Filipe Manana [Thu, 29 Aug 2024 14:23:32 +0000 (15:23 +0100)]
btrfs: make the extent map shrinker run asynchronously as a work queue job
Currently the extent map shrinker is run synchronously for kswapd tasks
that end up calling the fs shrinker (fs/super.c:super_cache_scan()).
This has some disadvantages and for some heavy workloads with memory
pressure it can cause some delays and stalls that make a machine
unresponsive for some periods. This happens because:
1) We can have several kswapd tasks on machines with multiple NUMA zones,
and running the extent map shrinker concurrently can cause high
contention on some spin locks, namely the spin locks that protect
the radix tree that tracks roots, the per root xarray that tracks
open inodes and the list of delayed iputs. This not only delays the
shrinker but also causes high CPU consumption and makes the task
running the shrinker monopolize a core, resulting in the symptoms
of an unresponsive system. This was noted in previous commits such as
commit ae1e766f623f ("btrfs: only run the extent map shrinker from
kswapd tasks");
2) The extent map shrinker's iteration over inodes can often be slow, even
after changing the data structure that tracks open inodes for a root
from a red black tree (up to kernel 6.10) to an xarray (kernel 6.10+).
The transition to the xarray while it made things a bit faster, it's
still somewhat slow - for example in a test scenario with 10000 inodes
that have no extent maps loaded, the extent map shrinker took between
5ms to 8ms, using a release, non-debug kernel. Iterating over the
extent maps of an inode can also be slow if have an inode with many
thousands of extent maps, since we use a red black tree to track and
search extent maps. So having the extent map shrinker run synchronously
adds extra delay for other things a kswapd task does.
So make the extent map shrinker run asynchronously as a job for the
system unbounded workqueue, just like what we do for data and metadata
space reclaim jobs.
Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Robbie Ko [Tue, 15 Oct 2024 07:41:37 +0000 (15:41 +0800)]
btrfs: reduce lock contention when eb cache miss for btree search
When crawling btree, if an eb cache miss occurs, we change to use the eb
read lock and release all previous locks (including the parent lock) to
reduce lock contention.
If an eb cache miss occurs in a leaf and needs to execute IO, before this
change we released locks only from level 2 and up and we read a leaf's
content from disk while holding a lock on its parent (level 1), causing
the unnecessary lock contention on the parent, after this change we
release locks from level 1 and up, but we lock level 0, and read leaf's
content from disk.
Because we have prepared the check parameters and the read lock of eb we
hold, we can ensure that no race will occur during the check and cause
unexpected errors.
Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Robbie Ko <robbieko@synology.com> Signed-off-by: David Sterba <dsterba@suse.com>
David Sterba [Wed, 9 Oct 2024 14:32:17 +0000 (16:32 +0200)]
btrfs: drop unused parameter data from btrfs_fill_super()
The only caller passes NULL, we can drop the parameter. This is since
the new mount option parser done in 3bb17a25bcb09a ("btrfs: add get_tree
callback for new mount API").
Reviewed-by: Anand Jain <anand.jain@oracle.com> Signed-off-by: David Sterba <dsterba@suse.com>
David Sterba [Wed, 9 Oct 2024 14:32:05 +0000 (16:32 +0200)]
btrfs: drop unused parameter fs_info from folio_range_has_eb()
The parameter was added in 8ff8466d29efc2 ("btrfs: support subpage for
extent buffer page release") for page but hasn't been used since, so we
can drop it.
Reviewed-by: Anand Jain <anand.jain@oracle.com> Signed-off-by: David Sterba <dsterba@suse.com>
David Sterba [Wed, 9 Oct 2024 14:31:58 +0000 (16:31 +0200)]
btrfs: drop unused parameter mask from try_release_extent_state()
The mask parameter used for allocations got unified to GFP_NOFS and
removed from relevant functions in 1d1268004430 ("btrfs: drop gfp from
parameter extent state helpers").
Reviewed-by: Anand Jain <anand.jain@oracle.com> Signed-off-by: David Sterba <dsterba@suse.com>
David Sterba [Wed, 9 Oct 2024 14:31:56 +0000 (16:31 +0200)]
btrfs: drop unused parameter refs from visit_node_for_delete()
The parameter duplicates what can be effectively obtained from
wc->refs[level - 1] and this is what's actually used inside. Added in
commit 2b73c7e761c4 ("btrfs: unify logic to decide if we need to walk
down into a node during snapshot delete").
Reviewed-by: Anand Jain <anand.jain@oracle.com> Signed-off-by: David Sterba <dsterba@suse.com>
David Sterba [Wed, 9 Oct 2024 14:31:49 +0000 (16:31 +0200)]
btrfs: drop unused parameter iov_iter from btrfs_write_check()
The parameter 'from' has never been used since commit b8d8e1fd570a
("btrfs: introduce btrfs_write_check()"), this is for buffered write.
Direct io write needs it so it was probably an interface thing, but we
can drop it.
Reviewed-by: Anand Jain <anand.jain@oracle.com> Signed-off-by: David Sterba <dsterba@suse.com>
David Sterba [Wed, 9 Oct 2024 14:31:47 +0000 (16:31 +0200)]
btrfs: drop unused parameter file_offset from btrfs_encoded_read_regular_fill_pages()
The file_offset parameter used to be passed to encoded read struct but
was removed in commit b665affe93d8 ("btrfs: remove unused members from
struct btrfs_encoded_read_private").
Reviewed-by: Anand Jain <anand.jain@oracle.com> Signed-off-by: David Sterba <dsterba@suse.com>
David Sterba [Wed, 9 Oct 2024 14:31:18 +0000 (16:31 +0200)]
btrfs: drop unused parameter map from scrub_simple_mirror()
The parameter map used to be passed to scrub_extent() until e02ee89baa66c4 ("btrfs: scrub: switch scrub_simple_mirror() to
scrub_stripe infrastructure"), where the scrub implementation was
completely reworked.
Reviewed-by: Anand Jain <anand.jain@oracle.com> Signed-off-by: David Sterba <dsterba@suse.com>
projects / thirdparty / linux.git / log
