When multiple writes happen at once, we may need to sacrifice a currently
active block group to be zone finished for a new allocation. We choose a
block group with the least free space left, and zone finish it.
To do the finishing, we need to send IOs for already allocated region
and wait for them and on-going IOs. Otherwise, these IOs fail because the
zone is already finished at the time the IO reach a device.
However, if a block group dedicated to the data relocation is zone
finished, there is a chance that finishing it before an ongoing write IO
reaches the device. That is because there is timing gap between an
allocation is done (block_group->reservations == 0, as pre-allocation is
done) and an ordered extent is created when the relocation IO starts.
Thus, if we finish the zone between them, we can fail the IOs.
We cannot simply use "fs_info->data_reloc_bg == block_group->start" to
avoid the zone finishing. Because, the data_reloc_bg may already switch to
a new block group, while there are still ongoing write IOs to the old
data_reloc_bg.
So, this patch reworks the BLOCK_GROUP_FLAG_ZONED_DATA_RELOC bit to
indicate there is a data relocation allocation and/or ongoing write to the
block group. The bit is set on allocation and cleared in end_io function of
the last IO for the currently allocated region.
To change the timing of the bit setting also solves the issue that the bit
being left even after there is no IO going on. With the current code, if
the data_reloc_bg switches after the last IO to the current data_reloc_bg,
the bit is set at this timing and there is no one clearing that bit. As a
result, that block group is kept unallocatable for anything.
Fixes: 343d8a3085 ("btrfs: zoned: prevent allocation from previous data relocation BG")
Fixes: 74e91b12b1 ("btrfs: zoned: zone finish unused block group")
CC: stable@vger.kernel.org # 6.1+
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
One of the CI runs triggered the following panic
assertion failed: PagePrivate(page) && page->private, in fs/btrfs/subpage.c:229
------------[ cut here ]------------
kernel BUG at fs/btrfs/subpage.c:229!
Internal error: Oops - BUG: 00000000f2000800 [#1] SMP
CPU: 0 PID: 923660 Comm: btrfs Not tainted 6.5.0-rc3+ #1
pstate: 61400005 (nZCv daif +PAN -UAO -TCO +DIT -SSBS BTYPE=--)
pc : btrfs_subpage_assert+0xbc/0xf0
lr : btrfs_subpage_assert+0xbc/0xf0
sp : ffff800093213720
x29: ffff800093213720 x28: ffff8000932138b4 x27: 000000000c280000
x26: 00000001b5d00000 x25: 000000000c281000 x24: 000000000c281fff
x23: 0000000000001000 x22: 0000000000000000 x21: ffffff42b95bf880
x20: ffff42b9528e0000 x19: 0000000000001000 x18: ffffffffffffffff
x17: 667274622f736620 x16: 6e69202c65746176 x15: 0000000000000028
x14: 0000000000000003 x13: 00000000002672d7 x12: 0000000000000000
x11: ffffcd3f0ccd9204 x10: ffffcd3f0554ae50 x9 : ffffcd3f0379528c
x8 : ffff800093213428 x7 : 0000000000000000 x6 : ffffcd3f091771e8
x5 : ffff42b97f333948 x4 : 0000000000000000 x3 : 0000000000000000
x2 : 0000000000000000 x1 : ffff42b9556cde80 x0 : 000000000000004f
Call trace:
btrfs_subpage_assert+0xbc/0xf0
btrfs_subpage_set_dirty+0x38/0xa0
btrfs_page_set_dirty+0x58/0x88
relocate_one_page+0x204/0x5f0
relocate_file_extent_cluster+0x11c/0x180
relocate_data_extent+0xd0/0xf8
relocate_block_group+0x3d0/0x4e8
btrfs_relocate_block_group+0x2d8/0x490
btrfs_relocate_chunk+0x54/0x1a8
btrfs_balance+0x7f4/0x1150
btrfs_ioctl+0x10f0/0x20b8
__arm64_sys_ioctl+0x120/0x11d8
invoke_syscall.constprop.0+0x80/0xd8
do_el0_svc+0x6c/0x158
el0_svc+0x50/0x1b0
el0t_64_sync_handler+0x120/0x130
el0t_64_sync+0x194/0x198
Code: 91098021 b0007fa0 91346000 97e9c6d2 (d4210000)
This is the same problem outlined in 17b17fcd6d ("btrfs:
set_page_extent_mapped after read_folio in btrfs_cont_expand") , and the
fix is the same. I originally looked for the same pattern elsewhere in
our code, but mistakenly skipped over this code because I saw the page
cache readahead before we set_page_extent_mapped, not realizing that
this was only in the !page case, that we can still end up with a
!uptodate page and then do the btrfs_read_folio further down.
The fix here is the same as the above mentioned patch, move the
set_page_extent_mapped call to after the btrfs_read_folio() block to
make sure that we have the subpage blocksize stuff setup properly before
using the page.
CC: stable@vger.kernel.org # 6.1+
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
My initial fix for the generic/475 hangs was related to metadata, but
our CI testing uncovered another case where we hang for similar reasons.
We again have a task with a plug that is holding an outstanding request
that is keeping the dm device from finishing it's suspend, and that task
is stuck in the allocator.
This time it is stuck trying to allocate data, but we do not have a
block group that matches the size class. The larger loop in the
allocator looks like this (simplified of course)
find_free_extent
for_each_block_group {
ffe_ctl->cached == btrfs_block_group_cache_done(bg)
if (!ffe_ctl->cached)
ffe_ctl->have_caching_bg = true;
do_allocation()
btrfs_wait_block_group_cache_progress();
}
if (loop == LOOP_CACHING_WAIT && ffe_ctl->have_caching_bg)
go search again;
In my earlier fix we were trying to allocate from the block group, but
we weren't waiting for the progress because we were only waiting for the
free space to be >= the amount of free space we wanted. My fix made it
so we waited for forward progress to be made as well, so we would be
sure to wait.
This time however we did not have a block group that matched our size
class, so what was happening was this
find_free_extent
for_each_block_group {
ffe_ctl->cached == btrfs_block_group_cache_done(bg)
if (!ffe_ctl->cached)
ffe_ctl->have_caching_bg = true;
if (size_class_doesn't_match())
goto loop;
do_allocation()
btrfs_wait_block_group_cache_progress();
loop:
release_block_group(block_group);
}
if (loop == LOOP_CACHING_WAIT && ffe_ctl->have_caching_bg)
go search again;
The size_class_doesn't_match() part was true, so we'd just skip this
block group and never wait for caching, and then because we found a
caching block group we'd just go back and do the loop again. We never
sleep and thus never flush the plug and we have the same deadlock.
Fix the logic for waiting on the block group caching to instead do it
unconditionally when we goto loop. This takes the logic out of the
allocation step, so now the loop looks more like this
find_free_extent
for_each_block_group {
ffe_ctl->cached == btrfs_block_group_cache_done(bg)
if (!ffe_ctl->cached)
ffe_ctl->have_caching_bg = true;
if (size_class_doesn't_match())
goto loop;
do_allocation()
btrfs_wait_block_group_cache_progress();
loop:
if (loop > LOOP_CACHING_NOWAIT && !ffe_ctl->retry_uncached &&
!ffe_ctl->cached) {
ffe_ctl->retry_uncached = true;
btrfs_wait_block_group_cache_progress();
}
release_block_group(block_group);
}
if (loop == LOOP_CACHING_WAIT && ffe_ctl->have_caching_bg)
go search again;
This simplifies the logic a lot, and makes sure that if we're hitting
uncached block groups we're always waiting on them at some point.
I ran this through 100 iterations of generic/475, as this particular
case was harder to hit than the previous one.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Use LIST_HEAD() to initialize the list_head instead of open-coding it.
Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[PROBLEM]
Inside function update_inline_extent_backref(), we have several
BUG_ON()s along with some ASSERT()s which can be triggered by corrupted
filesystem.
[ANAYLYSE]
Most of those BUG_ON()s and ASSERT()s are just a way of handling
unexpected on-disk data.
Although we have tree-checker to rule out obviously incorrect extent
tree blocks, it's not enough for these ones. Thus we need proper error
handling for them.
[FIX]
Thankfully all the callers of update_inline_extent_backref() would
eventually handle the errror by aborting the current transaction.
So this patch would do the proper error handling by:
- Make update_inline_extent_backref() to return int
The return value would be either 0 or -EUCLEAN.
- Replace BUG_ON()s and ASSERT()s with proper error handling
This includes:
* Dump the bad extent tree leaf
* Output an error message for the cause
This would include the extent bytenr, num_bytes (if needed), the bad
values and expected good values.
* Return -EUCLEAN
Note here we remove all the WARN_ON()s, as eventually the transaction
would be aborted, thus a backtrace would be triggered anyway.
- Better comments on why we expect refs == 1 and refs_to_mode == -1 for
tree blocks
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now that, we can re-enable metadata over-commit. As we moved the activation
from the reservation time to the write time, we no longer need to ensure
all the reserved bytes is properly activated.
Without the metadata over-commit, it suffers from lower performance because
it needs to flush the delalloc items more often and allocate more block
groups. Re-enabling metadata over-commit will solve the issue.
Fixes: 79417d040f ("btrfs: zoned: disable metadata overcommit for zoned")
CC: stable@vger.kernel.org # 6.1+
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now that a non-DATA block group is activated at write time, don't
activate it on allocation time.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now that we switched to write time activation, we no longer need to (and
must not) count the fresh region as zone unusable. This commit is similar
to revert of commit fa2068d7e9 ("btrfs: zoned: count fresh BG
region as zone unusable").
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In the current implementation, block groups are activated at reservation
time to ensure that all reserved bytes can be written to an active metadata
block group. However, this approach has proven to be less efficient, as it
activates block groups more frequently than necessary, putting pressure on
the active zone resource and leading to potential issues such as early
ENOSPC or hung_task.
Another drawback of the current method is that it hampers metadata
over-commit, and necessitates additional flush operations and block group
allocations, resulting in decreased overall performance.
To address these issues, this commit introduces a write-time activation of
metadata and system block group. This involves reserving at least one
active block group specifically for a metadata and system block group.
Since metadata write-out is always allocated sequentially, when we need to
write to a non-active block group, we can wait for the ongoing IOs to
complete, activate a new block group, and then proceed with writing to the
new block group.
Fixes: b093151391 ("btrfs: zoned: activate metadata block group on flush_space")
CC: stable@vger.kernel.org # 6.1+
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Ensure a metadata and system block group can be activated on write time, by
leaving a certain number of active zones when trying to activate a data
block group.
Zones for two metadata block groups (normal and tree-log) and one system
block group are reserved, according to the profile type: two zones per
block group on the DUP profile and one zone per block group otherwise.
The reservation must be freed once a non-data block group is allocated. If
not, we over-reserve the active zones and data block group activation will
suffer. For the dynamic reservation count, we need to manage the
reservation count per device.
The reservation count variable is protected by
fs_info->zone_active_bgs_lock.
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
On finishing a zone, the meta_write_pointer should be set of the end of the
zone to reflect the actual write pointer position.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We currently advance the meta_write_pointer in
btrfs_check_meta_write_pointer(). That makes it necessary to revert it
when locking the buffer failed. Instead, we can advance it just before
sending the buffer.
Also, this is necessary for the following commit. In the commit, it needs
to release the zoned_meta_io_lock to allow IOs to come in and wait for them
to fill the currently active block group. If we advance the
meta_write_pointer before locking the extent buffer, the following extent
buffer can pass the meta_write_pointer check, resulting in an unaligned
write failure.
Advancing the pointer is still thread-safe as the extent buffer is locked.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now that we have writeback_control passed to
btrfs_check_meta_write_pointer(), we can move the wbc condition in
submit_eb_page() to btrfs_check_meta_write_pointer() and return int.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For metadata write out on the zoned mode, we call
btrfs_check_meta_write_pointer() to check if an extent buffer to be written
is aligned to the write pointer.
We look up a block group containing the extent buffer for every extent
buffer, which takes unnecessary effort as the writing extent buffers are
mostly contiguous.
Introduce "zoned_bg" to cache the block group working on. Also, while
at it, rename "cache" to "block_group".
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Introduce btrfs_eb_write_context to consolidate writeback_control and the
exntent buffer context. This will help adding a block group context as
well.
While at it, move the eb context setting before
btrfs_check_meta_write_pointer(). We can set it here because we anyway need
to skip pages in the same eb if that eb is rejected by
btrfs_check_meta_write_pointer().
Suggested-by: Christoph Hellwig <hch@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When flushing qgroups, we try to join a running transaction, with
btrfs_join_transaction(), and then commit the transaction. However using
btrfs_join_transaction() will result in creating a new transaction in case
there isn't any running or if there's an existing one already committing.
This is pointless as we only need to attach to an existing one that is
not committing and in case there's an existing one committing, wait for
its commit to complete. Creating and committing an empty transaction is
wasteful, pointless IO and unnecessary rotation of the backup roots.
So use btrfs_attach_transaction_barrier() instead, to avoid creating and
committing empty transactions.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When starting a qgroup rescan, we try to join a running transaction, with
btrfs_join_transaction(), and then commit the transaction. However using
btrfs_join_transaction() will result in creating a new transaction in case
there isn't any running or if there's an existing one already committing.
This is pointless as we only need to attach to an existing one that is
not committing and in case there's an existing one committing, wait for
its commit to complete. Creating and committing an empty transaction is
wasteful, pointless IO and unnecessary rotation of the backup roots.
So use btrfs_attach_transaction_barrier() instead, to avoid creating and
committing empty transactions.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When flushing space and we are in the COMMIT_TRANS state, we join a
transaction with btrfs_join_transaction() and then commit the returned
transaction. However btrfs_join_transaction() starts a new transaction if
there is none currently open, which is pointless since comitting a new,
empty transaction, doesn't achieve anything, it only wastes time, IO and
creates an unnecessary rotation of the backup roots.
So use btrfs_attach_transaction_barrier() to avoid starting a new
transaction. This also waits for any ongoing transaction that is
committing (state >= TRANS_STATE_COMMIT_DOING) to fully complete, and
therefore wait for all the extents that were pinned during the
transaction's lifetime to be unpinned.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When flushing space we join a transaction to flush delayed items and
delayed references, in order to try to release space. However using
btrfs_join_transaction() not only joins an existing transaction as well
as it starts a new transaction if there is none open. If there is no
transaction open, we don't have neither delayed items nor delayed
references, so creating a new transaction is a waste of time, IO and
creates an unnecessary rotation of the backup roots without gaining any
benefits (including releasing space).
So use btrfs_join_transaction_nostart() when attempting to flush delayed
items and references.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There is no point in having find_free_dev_extent() because it's just a
simple wrapper around find_free_dev_extent_start() which always passes a
value of 0 for the search_start argument. Since there are no other callers
of find_free_dev_extent_start(), remove find_free_dev_extent() and rename
find_free_dev_extent_start() to find_free_dev_extent(), removing its
search_start argument because it's always 0.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The function find_free_dev_extent() is only used within volumes.c, so make
it static and remove its prototype from volumes.h.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_cleanup_fs_roots() is not used outside disk-io.c, so make it static,
remove its prototype from disk-io.h and move its definition above the
where it's used in disk-io.c
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
At priority_reclaim_metadata_space(), if we were not able to satisfy the
the ticket after going through the various flushing states and we notice
the fs went into an error state, likely due to a transaction abort during
the flushing, set the ticket's error to the error that caused the
transaction abort instead of an unconditional -EROFS.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
During mount we will call btrfs_orphan_cleanup() to remove any inodes that
were previously deleted (have a link count of 0) but for which we were not
able before to remove their items from the subvolume tree. The removal of
the items will happen by triggering eviction, when we do the final iput()
on them at btrfs_orphan_cleanup(), which will end in the loop at
btrfs_evict_inode() that truncates inode items.
In a dire situation we may have a transaction abort due to -ENOSPC when
attempting to truncate the inode items, and in that case the orphan item
(key type BTRFS_ORPHAN_ITEM_KEY) will remain in the subvolume tree and
when we hit the next iteration of the while loop at btrfs_orphan_cleanup()
we will find the same orphan item as before, and then we will return
-EINVAL from btrfs_orphan_cleanup() through the following if statement:
if (found_key.offset == last_objectid) {
btrfs_err(fs_info,
"Error removing orphan entry, stopping orphan cleanup");
ret = -EINVAL;
goto out;
}
This makes the mount operation fail with -EINVAL, when it should have been
-ENOSPC. This is confusing because -EINVAL might lead a user into thinking
it provided invalid mount options for example.
An example where this happens:
$ mount test.img /mnt
mount: /mnt: wrong fs type, bad option, bad superblock on /dev/loop0, missing codepage or helper program, or other error.
$ dmesg
[ 2542.356934] BTRFS: device fsid 977fff75-1181-4d2b-a739-384fa710d16e devid 1 transid 47409973 /dev/loop0 scanned by mount (4459)
[ 2542.357451] BTRFS info (device loop0): using crc32c (crc32c-intel) checksum algorithm
[ 2542.357461] BTRFS info (device loop0): disk space caching is enabled
[ 2542.742287] BTRFS info (device loop0): auto enabling async discard
[ 2542.764554] BTRFS info (device loop0): checking UUID tree
[ 2551.743065] ------------[ cut here ]------------
[ 2551.743068] BTRFS: Transaction aborted (error -28)
[ 2551.743149] WARNING: CPU: 7 PID: 215 at fs/btrfs/block-group.c:3494 btrfs_write_dirty_block_groups+0x397/0x3d0 [btrfs]
[ 2551.743311] Modules linked in: btrfs blake2b_generic (...)
[ 2551.743353] CPU: 7 PID: 215 Comm: kworker/u24:5 Not tainted 6.4.0-rc6-btrfs-next-134+ #1
[ 2551.743356] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.16.2-0-gea1b7a073390-prebuilt.qemu.org 04/01/2014
[ 2551.743357] Workqueue: events_unbound btrfs_async_reclaim_metadata_space [btrfs]
[ 2551.743405] RIP: 0010:btrfs_write_dirty_block_groups+0x397/0x3d0 [btrfs]
[ 2551.743449] Code: 8b 43 0c (...)
[ 2551.743451] RSP: 0018:ffff982c005a7c40 EFLAGS: 00010286
[ 2551.743452] RAX: 0000000000000000 RBX: ffff88fc6e44b400 RCX: 0000000000000000
[ 2551.743453] RDX: 0000000000000002 RSI: ffffffff8dff0878 RDI: 00000000ffffffff
[ 2551.743454] RBP: ffff88fc51817208 R08: 0000000000000000 R09: ffff982c005a7ae0
[ 2551.743455] R10: 0000000000000001 R11: 0000000000000001 R12: ffff88fc43d2e570
[ 2551.743456] R13: ffff88fc43d2e400 R14: ffff88fc8fb08ee0 R15: ffff88fc6e44b530
[ 2551.743457] FS: 0000000000000000(0000) GS:ffff89035fbc0000(0000) knlGS:0000000000000000
[ 2551.743458] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 2551.743459] CR2: 00007fa8cdf2f6f4 CR3: 0000000124850003 CR4: 0000000000370ee0
[ 2551.743462] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 2551.743463] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 2551.743464] Call Trace:
[ 2551.743472] <TASK>
[ 2551.743474] ? __warn+0x80/0x130
[ 2551.743478] ? btrfs_write_dirty_block_groups+0x397/0x3d0 [btrfs]
[ 2551.743520] ? report_bug+0x1f4/0x200
[ 2551.743523] ? handle_bug+0x42/0x70
[ 2551.743526] ? exc_invalid_op+0x14/0x70
[ 2551.743528] ? asm_exc_invalid_op+0x16/0x20
[ 2551.743532] ? btrfs_write_dirty_block_groups+0x397/0x3d0 [btrfs]
[ 2551.743574] ? _raw_spin_unlock+0x15/0x30
[ 2551.743576] ? btrfs_run_delayed_refs+0x1bd/0x200 [btrfs]
[ 2551.743609] commit_cowonly_roots+0x1e9/0x260 [btrfs]
[ 2551.743652] btrfs_commit_transaction+0x42e/0xfa0 [btrfs]
[ 2551.743693] ? __pfx_autoremove_wake_function+0x10/0x10
[ 2551.743697] flush_space+0xf1/0x5d0 [btrfs]
[ 2551.743743] ? _raw_spin_unlock+0x15/0x30
[ 2551.743745] ? finish_task_switch+0x91/0x2a0
[ 2551.743748] ? _raw_spin_unlock+0x15/0x30
[ 2551.743750] ? btrfs_get_alloc_profile+0xc9/0x1f0 [btrfs]
[ 2551.743793] btrfs_async_reclaim_metadata_space+0xe1/0x230 [btrfs]
[ 2551.743837] process_one_work+0x1d9/0x3e0
[ 2551.743844] worker_thread+0x4a/0x3b0
[ 2551.743847] ? __pfx_worker_thread+0x10/0x10
[ 2551.743849] kthread+0xee/0x120
[ 2551.743852] ? __pfx_kthread+0x10/0x10
[ 2551.743854] ret_from_fork+0x29/0x50
[ 2551.743860] </TASK>
[ 2551.743861] ---[ end trace 0000000000000000 ]---
[ 2551.743863] BTRFS info (device loop0: state A): dumping space info:
[ 2551.743866] BTRFS info (device loop0: state A): space_info DATA has 126976 free, is full
[ 2551.743868] BTRFS info (device loop0: state A): space_info total=13458472960, used=13458137088, pinned=143360, reserved=0, may_use=0, readonly=65536 zone_unusable=0
[ 2551.743870] BTRFS info (device loop0: state A): space_info METADATA has -51625984 free, is full
[ 2551.743872] BTRFS info (device loop0: state A): space_info total=771751936, used=770146304, pinned=1605632, reserved=0, may_use=51625984, readonly=0 zone_unusable=0
[ 2551.743874] BTRFS info (device loop0: state A): space_info SYSTEM has 14663680 free, is not full
[ 2551.743875] BTRFS info (device loop0: state A): space_info total=14680064, used=16384, pinned=0, reserved=0, may_use=0, readonly=0 zone_unusable=0
[ 2551.743877] BTRFS info (device loop0: state A): global_block_rsv: size 53231616 reserved 51544064
[ 2551.743878] BTRFS info (device loop0: state A): trans_block_rsv: size 0 reserved 0
[ 2551.743879] BTRFS info (device loop0: state A): chunk_block_rsv: size 0 reserved 0
[ 2551.743880] BTRFS info (device loop0: state A): delayed_block_rsv: size 0 reserved 0
[ 2551.743881] BTRFS info (device loop0: state A): delayed_refs_rsv: size 786432 reserved 0
[ 2551.743886] BTRFS: error (device loop0: state A) in btrfs_write_dirty_block_groups:3494: errno=-28 No space left
[ 2551.743911] BTRFS info (device loop0: state EA): forced readonly
[ 2551.743951] BTRFS warning (device loop0: state EA): could not allocate space for delete; will truncate on mount
[ 2551.743962] BTRFS error (device loop0: state EA): Error removing orphan entry, stopping orphan cleanup
[ 2551.743973] BTRFS warning (device loop0: state EA): Skipping commit of aborted transaction.
[ 2551.743989] BTRFS error (device loop0: state EA): could not do orphan cleanup -22
So make the btrfs_orphan_cleanup() return the value of BTRFS_FS_ERROR(),
if it's set, and -EINVAL otherwise.
For that same example, after this change, the mount operation fails with
-ENOSPC:
$ mount test.img /mnt
mount: /mnt: mount(2) system call failed: No space left on device.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently when we turn the fs into an error state, typically after a
transaction abort, we don't store the error anywhere, we just set a bit
(BTRFS_FS_STATE_ERROR) at struct btrfs_fs_info::fs_state to signal the
error state.
There are cases where it would be useful to have access to the specific
error in order to provide a more meaningful error to users/applications.
This change adds a member to struct btrfs_fs_info to store the error and
removes the BTRFS_FS_STATE_ERROR bit. When there's no error, the new
member (fs_error) has a value of 0, otherwise its value is a negative
errno value.
Followup changes will make use of this new member.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When doing a priority metadata space reclaim, while we are going through
the flush states and running their respective operations, it's possible
that a transaction abort happened, for example when running delayed refs
we hit -ENOSPC or in the critical section of transaction commit we failed
with -ENOSPC or some other error. In these cases a transaction was aborted
and the fs turned into error state. If that happened, then it makes no
sense to steal from the global block reserve and return success to the
caller if the stealing was successful - the caller will later get an
error when attempting to modify the fs. Instead make the ticket fail if
we have the fs in error state and don't attempt to steal from the global
rsv, as it's not only it's pointless, it also simplifies debugging some
-ENOSPC problems.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When dumping a space info also sum the available space for all block
groups and then print it. This often useful for debugging -ENOSPC
related problems.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When dumping a space info, we iterate over all its block groups and then
print their size and the amounts of bytes used, reserved, pinned, etc.
When debugging -ENOSPC problems it's also useful to know how much space
is available (free), so calculate that and print it as well.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When dumping a space info's block groups, also print the number of bytes
used for super blocks and delalloc. This is often useful for debugging
-ENOSPC problems.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When dumping free space, with btrfs_dump_free_space(), we pass a bytes
argument in order to count how many free space entries in the block group
have a size greater than or equal to that number of bytes. We then print
how many suitable entries we found, but we don't print the target number
of bytes, we just say "bytes". Change the message to actually print the
number of bytes, which makes debugging -ENOSPC issues a bit easier.
Also sligthly change the odd grammar and terminology: the sentence is
ending with 'is', which doesn't make sense, and the term 'blocks' is
confusing as we are referring to free space entries within the block
group's free space cache.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Update the comment for btrfs_join_transaction_nostart() to be more clear
about how it works and how it's different from btrfs_attach_transaction().
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When joining a transaction with TRANS_JOIN_NOSTART, if we don't find a
running transaction we end up creating one. This goes against the purpose
of TRANS_JOIN_NOSTART which is to join a running transaction if its state
is at or below the state TRANS_STATE_COMMIT_START, otherwise return an
-ENOENT error and don't start a new transaction. So fix this to not create
a new transaction if there's no running transaction at or below that
state.
CC: stable@vger.kernel.org # 4.14+
Fixes: a6d155d2e3 ("Btrfs: fix deadlock between fiemap and transaction commits")
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BACKGROUND]
Currently memove_extent_buffer() does a loop where it strop at any page
boundary inside [dst_offset, dst_offset + len) or [src_offset,
src_offset + len).
This is mostly allowing us to do copy_pages(), but if we're going to use
folios we will need to handle multi-page (the old behavior) or single
folio (the new optimization).
The current code would be a burden for future changes.
[ENHANCEMENT]
Instead of sticking with copy_pages(), here we utilize the new
__write_extent_buffer() helper to handle the writes.
Unlike the refactoring in memcpy_extent_buffer(), we can not just rely
on the write_extent_buffer() and only handle page boundaries inside src
range.
The function write_extent_buffer() itself is still doing forward
writing, thus it cannot handle the following case: (already in the
extent buffer memory operation tests, cross page overlapping run 2)
Src Page boundary
|///////|
|///|////|
Dst
In the above case, if we just follow page boundary in the src range, we
have no need to do any split, just one __write_extent_buffer() with
use_memmove = true.
But __write_extent_buffer() would split the dst range into two,
so it first copies the beginning part of the src range into the first half
of the dst range.
After this operation, the beginning of the dst range is already updated,
causing corruption.
So we have to follow the old behavior of handling both page boundaries.
And since we're the last caller of copy_pages(), we can remove it
completely.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BACKGROUND]
Currently memcpy_extent_buffer() does a loop where it would stop at
any page boundary inside [dst_offset, dst_offset + len) or [src_offset,
src_offset + len).
This is mostly allowing us to do copy_pages(), but if we're going to use
folios we will need to handle multi-page (the old behavior) or single
folio (the new optimization).
The current code would be a burden for future changes.
[ENHANCEMENT]
There is a hidden pitfall of the naming memcpy_extent_buffer(), unlike
regular memcpy(), this function can handle overlapping ranges.
So here we extract write_extent_buffer() into a new internal helper,
__write_extent_buffer(), and add a new parameter @use_memmove, to
indicate whether we should use memmove() or regular memcpy().
Now we can go __write_extent_buffer() to handle writing into the dst
range, with proper overlapping detection.
This has a tiny change to the chance of calling memmove().
As the split only happens at the source range page boundaries, the
memcpy/memmove() range would be slightly larger than the old code,
thus slightly increase the chance we call memmove() other than memcopy().
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_clone_extent_buffer() calls copy_page() at each iteration but we
can copy all pages at the end in one go if there were no errors.
This would make later conversion to folios easier.
Reviewed-by: Sweet Tea Dorminy <sweettea-kernel@dorminy.me>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BACKGROUND]
copy_extent_buffer_full() currently does different handling for regular
and subpage cases, for regular cases it does a page by page copying.
For subpage cases, it just copies the content.
This is fine for the page based extent buffer code, but for the incoming
folio conversion, it can be a burden to add a new branch just to handle
all the different combinations (subpage vs regular, one single folio vs
multi pages).
[ENHANCE]
Instead of handling the different combinations, just go one single
handling for all cases, utilizing write_extent_buffer() to do the
copying.
Reviewed-by: Sweet Tea Dorminy <sweettea-kernel@dorminy.me>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Helpers write_extent_buffer_chunk_tree_uuid() and
write_extent_buffer_fsid(), they can be implemented by
write_extent_buffer().
These two helpers are not that frequently used, they only get called
during initialization of a new tree block. There is not much need for
those slightly optimized versions. And since they can be easily
converted to one write_extent_buffer() call, define them as inline
helpers.
This would make later page/folio switch much easier, as all change only
need to happen in write_extent_buffer().
Reviewed-by: Sweet Tea Dorminy <sweettea-kernel@dorminy.me>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BACKGROUND]
Currently we handle extent bitmaps manually in
extent_buffer_bitmap_set() and extent_buffer_bitmap_clear().
Although with various helpers like eb_bitmap_offset() it's still a little
messy to read. The code seems to be a copy of bitmap_set(), but with
all the cross-page handling embedded into the code.
[ENHANCEMENT]
This patch would enhance the readability by introducing two helpers:
- memset_extent_buffer()
To handle the byte aligned range, thus all the cross-page handling is
done there.
- extent_buffer_get_byte()
This for the first and the last byte operations, which only need to
grab one byte, thus no need for any cross-page handling.
So we can split both extent_buffer_bitmap_set() and
extent_buffer_bitmap_clear() into 3 parts:
- Handle the first byte
If the range fits inside the first byte, we can exit early.
- Handle the byte aligned part
This is the part which can have cross-page operations, and it would
be handled by memset_extent_buffer().
- Handle the last byte
This refactoring does not only make the code a little easier to read,
but also makes later folio/page switch much easier, as the switch only
needs to be done inside memset_extent_buffer() and extent_buffer_get_byte().
Reviewed-by: Sweet Tea Dorminy <sweettea-kernel@dorminy.me>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The new self tests would populate a memory range with random bytes, then
copy it to the extent buffer, so that we can verify if the extent buffer
memory operation and memmove()/memcopy() are resulting the same
contents.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Enhance extent bitmap tests for the following aspects:
- Remove unnecessary @len from __test_eb_bitmaps()
We can fetch the length from extent buffer
- Explicitly distinguish bit and byte length
Now every start/len inside bitmap tests would have either "byte_" or
"bit_" prefix to make it more explicit.
- Better error reporting
If we have mismatch bits, the error report would dump the following
contents:
* start bytenr
* bit number
* the full byte from bitmap
* the full byte from the extent
This is to save developers time so obvious problem can be found
immediately
- Extract bitmap set/clear and check operation into two helpers
This is to save some code lines, as we will have more tests to do.
- Add new tests
The following tests are added, mostly for the incoming extent bitmap
accessor refactoring:
* Set bits inside the same byte
* Clear bits inside the same byte
* Cross byte boundary set
* Cross byte boundary clear
* Cross multi-byte boundary set
* Cross multi-byte boundary clear
Those new tests have already saved my backend for the incoming extent
buffer bitmap refactoring.
Reviewed-by: Sweet Tea Dorminy <sweettea-kernel@dorminy.me>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Some of these loop types aren't described, and they should be with the
definitions to make it easier to tell what each of them do.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There is a race between systemd and mount, as both of them try to register
the device in the kernel. When systemd loses the race, it prints the
following message:
BTRFS error: device /dev/sdb7 belongs to fsid 1b3bacbf-14db-49c9-a3ef-547998aacc4e, and the fs is already mounted.
The 'btrfs dev scan' registers one device at a time, so there is no way
for the mount thread to wait in the kernel for all the devices to have
registered as it won't know if all the devices are discovered.
For now, improve the error log by printing the command name and process
ID along with the error message.
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Fold folio_account_redirty into folio_redirty_for_writepage now
that all other users except for the also unused account_page_redirty
wrapper are gone.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For zoned file systems we need to use run_delalloc_zoned to submit
writeback, as we need to write out partial allocations when running into
zone active limits.
submit_uncompressed_range currently always calls cow_file_range to
allocate blocks and thus misses the active zone limits handling. Fix
this by passing the pages_dirty argument to run_delalloc_zoned and always
using it from submit_uncompressed_range as it does the right thing for
zoned and non-zoned file systems.
To account for the fact that run_delalloc_zoned is now also used for
non-zoned file systems rename it to run_delalloc_cow, and add comment
describing it.
Fixes: 42c0110009 ("btrfs: zoned: introduce dedicated data write path for zoned filesystems")
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David Sterba <dsterba@suse.com>
extent_write_locked_range currently expects that either all or no
pages are dirty when it is called. Bur run_delalloc_zoned is called
directly in the writepages path, and has the dirty bit cleared only
for locked_page and which the extent_write_cache_pages currently
operates. It currently works around this by redirtying locked_page,
but that is a bit inefficient and cumbersome. Pass a locked_page
argument to run_delalloc_zoned so that clearing the dirty bit can
be skipped on just that page.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Handling of the done_offset to cow_file_range is a bit confusing, as
it is not updated at all when the function succeeds, and the -EAGAIN
status is used bother for the case where we need to wait for a zone
finish and the one where the allocation was partially successful.
Change the calling convention so that done_offset is always updated,
and 0 is returned if some allocation was successful (partial allocation
can still only happen for zoned devices), and waiting for a zone
finish is done internally in cow_file_range instead of the caller.
Also write a comment explaining the logic.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David Sterba <dsterba@suse.com>
compress_file_range needs to clear the dirty bit before handing off work
to the compression worker threads to prevent processes coming in through
mmap and changing the file contents while the compression is accessing
the data (See commit 4adaa61102 ("Btrfs: fix race between mmap writes
and compression").
But when compress_file_range decides to not compress the data, it falls
back to submit_uncompressed_range which uses extent_write_locked_range
to write the uncompressed data. extent_write_locked_range currently
expects all pages to be marked dirty so that it can clear the dirty
bit itself, and thus compress_file_range has to redirty the page range.
Redirtying the page range is rather inefficient and also pointless,
so instead pass a pages_dirty parameter to extent_write_locked_range
and skip the redirty game entirely.
Note that compress_file_range was even redirtying the locked_page twice
given that extent_range_clear_dirty_for_io already redirties all pages
in the range, which must include locked_page if there is one.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
compress_file_range has two code blocks to free the page array for the
compressed data. Share the code using a goto label.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
compress_file_range can fail to compress either because of resource or
alignment constraints or because the data is incompressible. In the latter
case the inode is marked so that compression isn't tried again. Currently
that check is based on the condition that the pages array has been allocated
which is rather cryptic. Use a separate label to clearly distinguish this
case.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently the logic whether to compress or not in compress_file_range is
a bit convoluted because it tries to share code for creating inline
extents for the compressible [1] path and the bail to uncompressed path.
But the latter isn't needed at all, because cow_file_range as called by
submit_uncompressed_range will already create inline extents as needed,
so there is no need to have special handling for it if we can live with
the fact that it will be called a bit later in the ->ordered_func of the
workqueue instead of right now.
[1] there is undocumented logic that creates an uncompressed inline
extent outside of the shall not compress logic if total_in is too small.
This logic isn't explained in comments or any commit log I could find,
so I've preserved it. Documentation explaining it would be appreciated
if anyone understands this code.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
