It's currently u64 which gets instantly translated either to LONG_MAX
(if U64_MAX is passed) or cast to an unsigned long (which is in fact,
wrong because writeback_control::nr_to_write is a signed, long type).
Just convert the function's argument to be long time which obviates the
need to manually convert u64 value to a long. Adjust all call sites
which pass U64_MAX to pass LONG_MAX. Finally ensure that in
shrink_delalloc the u64 is converted to a long without overflowing,
resulting in a negative number.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Commit 38d715f494 ("btrfs: use btrfs_start_delalloc_roots in
shrink_delalloc") cleaned up how we do delalloc shrinking by utilizing
some infrastructure we have in place to flush inodes that we use for
device replace and snapshot. However this introduced a pretty serious
performance regression. To reproduce the user untarred the source
tarball of Firefox (360MiB xz compressed/1.5GiB uncompressed), and would
see it take anywhere from 5 to 20 times as long to untar in 5.10
compared to 5.9. This was observed on fast devices (SSD and better) and
not on HDD.
The root cause is because before we would generally use the normal
writeback path to reclaim delalloc space, and for this we would provide
it with the number of pages we wanted to flush. The referenced commit
changed this to flush that many inodes, which drastically increased the
amount of space we were flushing in certain cases, which severely
affected performance.
We cannot revert this patch unfortunately because of 3d45f221ce
("btrfs: fix deadlock when cloning inline extent and low on free
metadata space") which requires the ability to skip flushing inodes that
are being cloned in certain scenarios, which means we need to keep using
our flushing infrastructure or risk re-introducing the deadlock.
Instead to fix this problem we can go back to providing
btrfs_start_delalloc_roots with a number of pages to flush, and then set
up a writeback_control and utilize sync_inode() to handle the flushing
for us. This gives us the same behavior we had prior to the fix, while
still allowing us to avoid the deadlock that was fixed by Filipe. I
redid the users original test and got the following results on one of
our test machines (256GiB of ram, 56 cores, 2TiB Intel NVMe drive)
5.9 0m54.258s
5.10 1m26.212s
5.10+patch 0m38.800s
5.10+patch is significantly faster than plain 5.9 because of my patch
series "Change data reservations to use the ticketing infra" which
contained the patch that introduced the regression, but generally
improved the overall ENOSPC flushing mechanisms.
Additional testing on consumer-grade SSD (8GiB ram, 8 CPU) confirm
the results:
5.10.5 4m00s
5.10.5+patch 1m08s
5.11-rc2 5m14s
5.11-rc2+patch 1m30s
Reported-by: René Rebe <rene@exactcode.de>
Fixes: 38d715f494 ("btrfs: use btrfs_start_delalloc_roots in shrink_delalloc")
CC: stable@vger.kernel.org # 5.10
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Tested-by: David Sterba <dsterba@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ add my test results ]
Signed-off-by: David Sterba <dsterba@suse.com>
When cloning an inline extent there are cases where we can not just copy
the inline extent from the source range to the target range (e.g. when the
target range starts at an offset greater than zero). In such cases we copy
the inline extent's data into a page of the destination inode and then
dirty that page. However, after that we will need to start a transaction
for each processed extent and, if we are ever low on available metadata
space, we may need to flush existing delalloc for all dirty inodes in an
attempt to release metadata space - if that happens we may deadlock:
* the async reclaim task queued a delalloc work to flush delalloc for
the destination inode of the clone operation;
* the task executing that delalloc work gets blocked waiting for the
range with the dirty page to be unlocked, which is currently locked
by the task doing the clone operation;
* the async reclaim task blocks waiting for the delalloc work to complete;
* the cloning task is waiting on the waitqueue of its reservation ticket
while holding the range with the dirty page locked in the inode's
io_tree;
* if metadata space is not released by some other task (like delalloc for
some other inode completing for example), the clone task waits forever
and as a consequence the delalloc work and async reclaim tasks will hang
forever as well. Releasing more space on the other hand may require
starting a transaction, which will hang as well when trying to reserve
metadata space, resulting in a deadlock between all these tasks.
When this happens, traces like the following show up in dmesg/syslog:
[87452.323003] INFO: task kworker/u16:11:1810830 blocked for more than 120 seconds.
[87452.323644] Tainted: G B W 5.10.0-rc4-btrfs-next-73 #1
[87452.324248] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[87452.324852] task:kworker/u16:11 state:D stack: 0 pid:1810830 ppid: 2 flags:0x00004000
[87452.325520] Workqueue: btrfs-flush_delalloc btrfs_work_helper [btrfs]
[87452.326136] Call Trace:
[87452.326737] __schedule+0x5d1/0xcf0
[87452.327390] schedule+0x45/0xe0
[87452.328174] lock_extent_bits+0x1e6/0x2d0 [btrfs]
[87452.328894] ? finish_wait+0x90/0x90
[87452.329474] btrfs_invalidatepage+0x32c/0x390 [btrfs]
[87452.330133] ? __mod_memcg_state+0x8e/0x160
[87452.330738] __extent_writepage+0x2d4/0x400 [btrfs]
[87452.331405] extent_write_cache_pages+0x2b2/0x500 [btrfs]
[87452.332007] ? lock_release+0x20e/0x4c0
[87452.332557] ? trace_hardirqs_on+0x1b/0xf0
[87452.333127] extent_writepages+0x43/0x90 [btrfs]
[87452.333653] ? lock_acquire+0x1a3/0x490
[87452.334177] do_writepages+0x43/0xe0
[87452.334699] ? __filemap_fdatawrite_range+0xa4/0x100
[87452.335720] __filemap_fdatawrite_range+0xc5/0x100
[87452.336500] btrfs_run_delalloc_work+0x17/0x40 [btrfs]
[87452.337216] btrfs_work_helper+0xf1/0x600 [btrfs]
[87452.337838] process_one_work+0x24e/0x5e0
[87452.338437] worker_thread+0x50/0x3b0
[87452.339137] ? process_one_work+0x5e0/0x5e0
[87452.339884] kthread+0x153/0x170
[87452.340507] ? kthread_mod_delayed_work+0xc0/0xc0
[87452.341153] ret_from_fork+0x22/0x30
[87452.341806] INFO: task kworker/u16:1:2426217 blocked for more than 120 seconds.
[87452.342487] Tainted: G B W 5.10.0-rc4-btrfs-next-73 #1
[87452.343274] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[87452.344049] task:kworker/u16:1 state:D stack: 0 pid:2426217 ppid: 2 flags:0x00004000
[87452.344974] Workqueue: events_unbound btrfs_async_reclaim_metadata_space [btrfs]
[87452.345655] Call Trace:
[87452.346305] __schedule+0x5d1/0xcf0
[87452.346947] ? kvm_clock_read+0x14/0x30
[87452.347676] ? wait_for_completion+0x81/0x110
[87452.348389] schedule+0x45/0xe0
[87452.349077] schedule_timeout+0x30c/0x580
[87452.349718] ? _raw_spin_unlock_irqrestore+0x3c/0x60
[87452.350340] ? lock_acquire+0x1a3/0x490
[87452.351006] ? try_to_wake_up+0x7a/0xa20
[87452.351541] ? lock_release+0x20e/0x4c0
[87452.352040] ? lock_acquired+0x199/0x490
[87452.352517] ? wait_for_completion+0x81/0x110
[87452.353000] wait_for_completion+0xab/0x110
[87452.353490] start_delalloc_inodes+0x2af/0x390 [btrfs]
[87452.353973] btrfs_start_delalloc_roots+0x12d/0x250 [btrfs]
[87452.354455] flush_space+0x24f/0x660 [btrfs]
[87452.355063] btrfs_async_reclaim_metadata_space+0x1bb/0x480 [btrfs]
[87452.355565] process_one_work+0x24e/0x5e0
[87452.356024] worker_thread+0x20f/0x3b0
[87452.356487] ? process_one_work+0x5e0/0x5e0
[87452.356973] kthread+0x153/0x170
[87452.357434] ? kthread_mod_delayed_work+0xc0/0xc0
[87452.357880] ret_from_fork+0x22/0x30
(...)
< stack traces of several tasks waiting for the locks of the inodes of the
clone operation >
(...)
[92867.444138] RSP: 002b:00007ffc3371bbe8 EFLAGS: 00000246 ORIG_RAX: 0000000000000052
[92867.444624] RAX: ffffffffffffffda RBX: 00007ffc3371bea0 RCX: 00007f61efe73f97
[92867.445116] RDX: 0000000000000000 RSI: 0000560fbd5d7a40 RDI: 0000560fbd5d8960
[92867.445595] RBP: 00007ffc3371beb0 R08: 0000000000000001 R09: 0000000000000003
[92867.446070] R10: 00007ffc3371b996 R11: 0000000000000246 R12: 0000000000000000
[92867.446820] R13: 000000000000001f R14: 00007ffc3371bea0 R15: 00007ffc3371beb0
[92867.447361] task:fsstress state:D stack: 0 pid:2508238 ppid:2508153 flags:0x00004000
[92867.447920] Call Trace:
[92867.448435] __schedule+0x5d1/0xcf0
[92867.448934] ? _raw_spin_unlock_irqrestore+0x3c/0x60
[92867.449423] schedule+0x45/0xe0
[92867.449916] __reserve_bytes+0x4a4/0xb10 [btrfs]
[92867.450576] ? finish_wait+0x90/0x90
[92867.451202] btrfs_reserve_metadata_bytes+0x29/0x190 [btrfs]
[92867.451815] btrfs_block_rsv_add+0x1f/0x50 [btrfs]
[92867.452412] start_transaction+0x2d1/0x760 [btrfs]
[92867.453216] clone_copy_inline_extent+0x333/0x490 [btrfs]
[92867.453848] ? lock_release+0x20e/0x4c0
[92867.454539] ? btrfs_search_slot+0x9a7/0xc30 [btrfs]
[92867.455218] btrfs_clone+0x569/0x7e0 [btrfs]
[92867.455952] btrfs_clone_files+0xf6/0x150 [btrfs]
[92867.456588] btrfs_remap_file_range+0x324/0x3d0 [btrfs]
[92867.457213] do_clone_file_range+0xd4/0x1f0
[92867.457828] vfs_clone_file_range+0x4d/0x230
[92867.458355] ? lock_release+0x20e/0x4c0
[92867.458890] ioctl_file_clone+0x8f/0xc0
[92867.459377] do_vfs_ioctl+0x342/0x750
[92867.459913] __x64_sys_ioctl+0x62/0xb0
[92867.460377] do_syscall_64+0x33/0x80
[92867.460842] entry_SYSCALL_64_after_hwframe+0x44/0xa9
(...)
< stack traces of more tasks blocked on metadata reservation like the clone
task above, because the async reclaim task has deadlocked >
(...)
Another thing to notice is that the worker task that is deadlocked when
trying to flush the destination inode of the clone operation is at
btrfs_invalidatepage(). This is simply because the clone operation has a
destination offset greater than the i_size and we only update the i_size
of the destination file after cloning an extent (just like we do in the
buffered write path).
Since the async reclaim path uses btrfs_start_delalloc_roots() to trigger
the flushing of delalloc for all inodes that have delalloc, add a runtime
flag to an inode to signal it should not be flushed, and for inodes with
that flag set, start_delalloc_inodes() will simply skip them. When the
cloning code needs to dirty a page to copy an inline extent, set that flag
on the inode and then clear it when the clone operation finishes.
This could be sporadically triggered with test case generic/269 from
fstests, which exercises many fsstress processes running in parallel with
several dd processes filling up the entire filesystem.
CC: stable@vger.kernel.org # 5.9+
Fixes: 05a5a7621c ("Btrfs: implement full reflink support for inline extents")
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have this thing wrapped in an RCU lock, but it's really not needed.
We create all the space_info's on mount, and we destroy them on unmount.
The list never changes and we're protected from messing with it by the
normal mount/umount path, so kill the RCU stuff around it.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
It is not used since commit 0096420adb ("btrfs: do not
account global reserve in can_overcommit").
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Dave reported an issue where generic/102 would sometimes hang. This
turned out to be because we'd get into this spot where we were no longer
making progress on data reservations because our exit condition was not
met. The log is basically
while (!space_info->full && !list_empty(&space_info->tickets))
flush_space(space_info, flush_state);
where flush state is our various flush states, but doesn't include
ALLOC_CHUNK_FORCE. This is because we actually lead with allocating
chunks, and so the assumption was that once you got to the actual
flushing states you could no longer allocate chunks. This was a stupid
assumption, because you could have deleted block groups that would be
reclaimed by a transaction commit, thus unsetting space_info->full.
This is essentially what happens with generic/102, and so sometimes
you'd get stuck in the flushing loop because we weren't allocating
chunks, but flushing space wasn't giving us what we needed to make
progress.
Fix this by adding ALLOC_CHUNK_FORCE to the end of our flushing states,
that way we will eventually bail out because we did end up with
space_info->full if we free'd a chunk previously. Otherwise, as is the
case for this test, we'll allocate our chunk and continue on our happy
merry way.
Reported-by: David Sterba <dsterba@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The data flushing steps are not obvious to people other than myself and
Chris. Write a giant comment explaining the reasoning behind each flush
step for data as well as why it is in that particular order.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now that we have the data ticketing stuff in place, move normal data
reservations to use an async reclaim helper to satisfy tickets. Before
we could have multiple tasks race in and both allocate chunks, resulting
in more data chunks than we would necessarily need. Serializing these
allocations and making a single thread responsible for flushing will
only allocate chunks as needed, as well as cut down on transaction
commits and other flush related activities.
Priority reservations will still work as they have before, simply
trying to allocate a chunk until they can make their reservation.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We can end up with freed extents in the delayed refs, and thus
may_commit_transaction() may not think we have enough pinned space to
commit the transaction and we'll ENOSPC early. Handle this by running
the delayed refs in order to make sure pinned is uptodate before we try
to commit the transaction.
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Before we were waiting on iputs after we committed the transaction, but
this doesn't really make much sense. We want to reclaim any space we
may have in order to be more likely to commit the transaction, due to
pinned space being added by running the delayed iputs. Fix this by
making delayed iputs run before committing the transaction.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We used to unconditionally commit the transaction at least 2 times and
then on the 3rd try check against pinned space to make sure committing
the transaction was worth the effort. This is overkill, we know nobody
is going to steal our reservation, and if we can't make our reservation
with the pinned amount simply bail out.
This also cleans up the passing of bytes_needed to
may_commit_transaction, as that was the thing we added into place in
order to accomplish this behavior. We no longer need it so remove that
mess.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This was an old wart left over from how we previously did data
reservations. Before we could have people race in and take a
reservation while we were flushing space, so we needed to make sure we
looped a few times before giving up. Now that we're using the ticketing
infrastructure we don't have to worry about this and can drop the logic
altogether.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now that data reservations follow the same pattern as metadata
reservations we can simply rename __reserve_metadata_bytes to
__reserve_bytes and use that helper for data reservations.
Things to keep in mind, btrfs_can_overcommit() returns 0 for data,
because we can never overcommit. We also will never pass in FLUSH_ALL
for data, so we'll simply be added to the priority list and go straight
into handle_reserve_ticket.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Nikolay reported a problem where generic/371 would fail sometimes with a
slow drive. The gist of the test is that we fallocate a file in
parallel with a pwrite of a different file. These two files combined
are smaller than the file system, but sometimes the pwrite would ENOSPC.
A fair bit of investigation uncovered the fact that the fallocate
workload was racing in and grabbing the free space that the pwrite
workload was trying to free up so it could make its own reservation.
After a few loops of this eventually the pwrite workload would error out
with an ENOSPC.
We've had the same problem with metadata as well, and we serialized all
metadata allocations to satisfy this problem. This wasn't usually a
problem with data because data reservations are more straightforward,
but obviously could still happen.
Fix this by not allowing reservations to occur if there are any pending
tickets waiting to be satisfied on the space info.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now that we have all the infrastructure in place, use the ticketing
infrastructure to make data allocations. This still maintains the exact
same flushing behavior, but now we're using tickets to get our
reservations satisfied.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Create a new function btrfs_reserve_data_bytes() in order to handle data
reservations. This uses the new flush types and flush states to handle
making data reservations.
This patch specifically does not change any functionality, and is
purposefully not cleaned up in order to make bisection easier for the
future patches. The new helper is identical to the old helper in how it
handles data reservations. We first try to force a chunk allocation,
and then we run through the flush states all at once and in the same
order that they were done with the old helper.
Subsequent patches will clean this up and change the behavior of the
flushing, and it is important to keep those changes separate so we can
easily bisect down to the patch that caused the regression, rather than
the patch that made us start using the new infrastructure.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Data space flushing currently unconditionally commits the transaction
twice in a row, and the last time it checks if there's enough pinned
extents to satisfy its reservation before deciding to commit the
transaction for the 3rd and final time.
Encode this logic into may_commit_transaction(). In the next patch we
will pass in U64_MAX for bytes_needed the first two times, and the final
time we will pass in the actual bytes we need so the normal logic will
apply.
This patch exists solely to make the logical changes I will make to the
flushing state machine separate to make it easier to bisect any
performance related regressions.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently the way we do data reservations is by seeing if we have enough
space in our space_info. If we do not and we're a normal inode we'll
1) Attempt to force a chunk allocation until we can't anymore.
2) If that fails we'll flush delalloc, then commit the transaction, then
run the delayed iputs.
If we are a free space inode we're only allowed to force a chunk
allocation. In order to use the normal flushing mechanism we need to
encode this into a flush state array for normal inodes. Since both will
start with allocating chunks until the space info is full there is no
need to add this as a flush state, this will be handled specially.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Right now if the space is freed up after the ordered extents complete
(which is likely since the reservations are held until they complete),
we would do extra delalloc flushing before we'd notice that we didn't
have any more tickets. Fix this by moving the tickets check after our
wait_ordered_extents check.
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The original iteration of flushing had us flushing delalloc and then
checking to see if we could make our reservation, thus we were very
careful about how many pages we would flush at once.
But now that everything is async and we satisfy tickets as the space
becomes available we don't have to keep track of any of this, simply
try and flush the number of dirty inodes we may have in order to
reclaim space to make our reservation. This cleans up our delalloc
flushing significantly.
The async_pages stuff is dropped because btrfs_start_delalloc_roots()
handles the case that we generate async extents for us, so we no longer
require this extra logic.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have traditionally used flush_space() to flush metadata space, so
we've been unconditionally using btrfs_metadata_alloc_profile() for our
profile to allocate a chunk. However if we're going to use this for
data we need to use btrfs_get_alloc_profile() on the space_info we pass
in.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently shrink_delalloc just looks up the metadata space info, but
this won't work if we're trying to reclaim space for data chunks. We
get the right space_info we want passed into flush_space, so simply pass
that along to shrink_delalloc.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Data allocations are going to want to pass in U64_MAX for flushing
space, adjust shrink_delalloc to handle this properly.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We don't use this anywhere inside of shrink_delalloc since 17024ad0a0
("Btrfs: fix early ENOSPC due to delalloc"), remove it.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have btrfs_wait_ordered_roots() which takes a u64 for nr, but
btrfs_start_delalloc_roots() that takes an int for nr, which makes using
them in conjunction, especially for something like (u64)-1, annoying and
inconsistent. Fix btrfs_start_delalloc_roots() to take a u64 for nr and
adjust start_delalloc_inodes() and it's callers appropriately.
This means we've adjusted start_delalloc_inodes() to take a pointer of
nr since we want to preserve the ability for start-delalloc_inodes() to
return an error, so simply make it do the nr adjusting as necessary.
Part of adjusting the callers to this means changing
btrfs_writeback_inodes_sb_nr() to take a u64 for items. This may be
confusing because it seems unrelated, but the caller of
btrfs_writeback_inodes_sb_nr() already passes in a u64, it's just the
function variable that needs to be changed.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The reclaim_size counter of a space_info object is unsigned. So its value
can never be negative, it's pointless to have an assertion that checks
its value is >= 0, therefore remove it.
Reviewed-by: Nikolay Borisov <nborisov@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>
With normal tickets we could have a large reservation at the front of
the list that is unable to be satisfied, but a smaller ticket later on
that can be satisfied. The way we handle this is to run
btrfs_try_granting_tickets() in maybe_fail_all_tickets().
However no such protection exists for priority tickets. Fix this by
handling it in handle_reserve_ticket(). If we've returned after
attempting to flush space in a priority related way, we'll still be on
the priority list and need to be removed.
We rely on the flushing to free up space and wake the ticket, but if
there is not enough space to reclaim _but_ there's enough space in the
space_info to handle subsequent reservations then we would have gotten
an ENOSPC erroneously.
Address this by catching where we are still on the list, meaning we were
a priority ticket, and removing ourselves and then running
btrfs_try_granting_tickets(). This will handle this particular corner
case.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In debugging a generic/320 failure on ppc64, Nikolay noticed that
sometimes we'd ENOSPC out with plenty of space to reclaim if we had
committed the transaction. He further discovered that this was because
there was a priority ticket that was small enough to fit in the free
space currently in the space_info.
Consider the following scenario. There is no more space to reclaim in
the fs without committing the transaction. Assume there's 1MiB of space
free in the space info, but there are pending normal tickets with 2MiB
reservations.
Now a priority ticket comes in with a .5MiB reservation. Because we
have normal tickets pending we add ourselves to the priority list,
despite the fact that we could satisfy this reservation.
The flushing machinery now gets to the point where it wants to commit
the transaction, but because there's a .5MiB ticket on the priority list
and we have 1MiB of free space we assume the ticket will be granted
soon, so we bail without committing the transaction.
Meanwhile the priority flushing does not commit the transaction, and
eventually fails with an ENOSPC. Then all other tickets are failed with
ENOSPC because we were never able to actually commit the transaction.
The fix for this is we should have simply granted the priority flusher
his reservation, because there was space to make the reservation.
Priority flushers by definition take priority, so they are allowed to
make their reservations before any previous normal tickets. By not
adding this priority ticket to the list the normal flushing mechanisms
will then commit the transaction and everything will continue normally.
We still need to serialize ourselves with other priority tickets, so if
there are any tickets on the priority list then we need to add ourselves
to that list in order to maintain the serialization between priority
tickets.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
On ppc64le with 64k page size (respectively 64k block size) generic/320
was failing and debug output showed we were getting a premature ENOSPC
with a bunch of space in btrfs_fs_info::trans_block_rsv.
This meant there were still open transaction handles holding space, yet
the flusher didn't commit the transaction because it deemed the freed
space won't be enough to satisfy the current reserve ticket. Fix this
by accounting for space in trans_block_rsv when deciding whether the
current transaction should be committed or not.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We previously had a limit of stealing 50% of the global reserve for
unlink. This was from a time when the global reserve was used for the
delayed refs as well. However now those reservations are kept separate,
so the global reserve can be depleted much more to allow us to make
progress for space restoring operations like unlink. Change the minimum
amount of space required to be left in the global reserve to 10%.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For unlink transactions and block group removal
btrfs_start_transaction_fallback_global_rsv will first try to start an
ordinary transaction and if it fails it will fall back to reserving the
required amount by stealing from the global reserve. This is problematic
because of all the same reasons we had with previous iterations of the
ENOSPC handling, thundering herd. We get a bunch of failures all at
once, everybody tries to allocate from the global reserve, some win and
some lose, we get an ENSOPC.
Fix this behavior by introducing BTRFS_RESERVE_FLUSH_ALL_STEAL. It's
used to mark unlink reservation. To fix this we need to integrate this
logic into the normal ENOSPC infrastructure. We still go through all of
the normal flushing work, and at the moment we begin to fail all the
tickets we try to satisfy any tickets that are allowed to steal by
stealing from the global reserve. If this works we start the flushing
system over again just like we would with a normal ticket satisfaction.
This serializes our global reserve stealing, so we don't have the
thundering herd problem.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Whenever we add a ticket to a space_info object we increment the object's
reclaim_size counter witht the ticket's bytes, and we decrement it with
the corresponding amount only when we are able to grant the requested
space to the ticket. When we are not able to grant the space to a ticket,
or when the ticket is removed due to a signal (e.g. an application has
received sigterm from the terminal) we never decrement the counter with
the corresponding bytes from the ticket. This leak can result in the
space reclaim code to later do much more work than necessary. So fix it
by decrementing the counter when those two cases happen as well.
Fixes: db161806dc ("btrfs: account ticket size at add/delete time")
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Instead of iterating all pending tickets on the normal/priority list to
sum their total size the cost can be amortized across ticket addition/
removal. This turns O(n) + O(m) (where n is the size of the normal list
and m of the priority list) into O(1). This will mostly have effect in
workloads that experience heavy flushing.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
I noticed while running my snapshot torture test that we were getting a
lot of metadata chunks allocated with very little actually used.
Digging into this we would commit the transaction, still not have enough
space, and then force a chunk allocation.
I noticed that we were barely flushing any delalloc at all, despite the
fact that we had around 13gib of outstanding delalloc reservations. It
turns out this is because of our btrfs_calc_reclaim_metadata_size()
calculation. It _only_ takes into account the outstanding ticket sizes,
which isn't the whole story. In this particular workload we're slowly
filling up the disk, which means our overcommit space will suddenly
become a lot less, and our outstanding reservations will be well more
than what we can handle. However we are only flushing based on our
ticket size, which is much less than we need to actually reclaim.
So fix btrfs_calc_reclaim_metadata_size() to take into account the
overage in the case that we've gotten less available space suddenly.
This makes it so we attempt to reclaim a lot more delalloc space, which
allows us to make our reservations and we no longer are allocating a
bunch of needless metadata chunks.
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Add another comment to cover how the space reservation system works
generally. This covers the actual reservation flow, as well as how
flushing is handled.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
inc_block_group_ro does a calculation to see if we have enough room left
over if we mark this block group as read only in order to see if it's ok
to mark the block group as read only.
The problem is this calculation _only_ works for data, where our used is
always less than our total. For metadata we will overcommit, so this
will almost always fail for metadata.
Fix this by exporting btrfs_can_overcommit, and then see if we have
enough space to remove the remaining free space in the block group we
are trying to mark read only. If we do then we can mark this block
group as read only.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have the space_info, we can just check its flags to see if it's the
system chunk space info.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The type name is misleading, a single entry is named 'cache' while this
normally means a collection of objects. Rename that everywhere. Also the
identifier was quite long, making function prototypes harder to format.
Suggested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
It is not used anymore since commit 957780eb27 ("Btrfs: introduce
ticketed enospc infrastructure"), so just remove it.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The on-disk format of block group item makes use of the key that stores
the offset and length. This is further used in the code, although this
makes thing harder to understand. The key is also packed so the
offset/length is not properly aligned as u64.
Add start (key.objectid) and length (key.offset) members to block group
and remove the embedded key. When the item is searched or written, a
local variable for key is used.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For unknown reasons, the member 'used' in the block group struct is
stored in the b-tree item and accessed everywhere using the special
accessor helper. Let's unify it and make it a regular member and only
update the item before writing it to the tree.
The item is still being used for flags and chunk_objectid, there's some
duplication until the item is removed in following patches.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The attribute is more relaxed than const and the functions could
dereference pointers, as long as the observable state is not changed. We
do have such functions, based on -Wsuggest-attribute=pure .
The visible effects of this patch are negligible, there are differences
in the assembly but hard to summarize.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When a task that is allocating metadata needs to wait for the async
reclaim job to process its ticket and gets a signal (because it was killed
for example) before doing the wait, the task ends up erroring out but
with space reserved for its ticket, which never gets released, resulting
in a metadata space leak (more specifically a leak in the bytes_may_use
counter of the metadata space_info object).
Here's the sequence of steps leading to the space leak:
1) A task tries to create a file for example, so it ends up trying to
start a transaction at btrfs_create();
2) The filesystem is currently in a state where there is not enough
metadata free space to satisfy the transaction's needs. So at
space-info.c:__reserve_metadata_bytes() we create a ticket and
add it to the list of tickets of the space info object. Also,
because the metadata async reclaim job is not running, we queue
a job ro run metadata reclaim;
3) In the meanwhile the task receives a signal (like SIGTERM from
a kill command for example);
4) After queing the async reclaim job, at __reserve_metadata_bytes(),
we unlock the metadata space info and call handle_reserve_ticket();
5) That last function calls wait_reserve_ticket(), which acquires the
lock from the metadata space info. Then in the first iteration of
its while loop, it calls prepare_to_wait_event(), which returns
-ERESTARTSYS because the task has a pending signal. As a result,
we set the error field of the ticket to -EINTR and exit the while
loop without deleting the ticket from the list of tickets (in the
space info object). After exiting the loop we unlock the space info;
6) The async reclaim job is able to release enough metadata, acquires
the metadata space info's lock and then reserves space for the ticket,
since the ticket is still in the list of (non-priority) tickets. The
space reservation happens at btrfs_try_granting_tickets(), called from
maybe_fail_all_tickets(). This increments the bytes_may_use counter
from the metadata space info object, sets the ticket's bytes field to
zero (meaning success, that space was reserved) and removes it from
the list of tickets;
7) wait_reserve_ticket() returns, with the error field of the ticket
set to -EINTR. Then handle_reserve_ticket() just propagates that error
to the caller. Because an error was returned, the caller does not
release the reserved space, since the expectation is that any error
means no space was reserved.
Fix this by removing the ticket from the list, while holding the space
info lock, at wait_reserve_ticket() when prepare_to_wait_event() returns
an error.
Also add some comments and an assertion to guarantee we never end up with
a ticket that has an error set and a bytes counter field set to zero, to
more easily detect regressions in the future.
This issue could be triggered sporadically by some test cases from fstests
such as generic/269 for example, which tries to fill a filesystem and then
kills fsstress processes running in the background.
When this issue happens, we get a warning in syslog/dmesg when unmounting
the filesystem, like the following:
------------[ cut here ]------------
WARNING: CPU: 0 PID: 13240 at fs/btrfs/block-group.c:3186 btrfs_free_block_groups+0x314/0x470 [btrfs]
(...)
CPU: 0 PID: 13240 Comm: umount Tainted: G W L 5.3.0-rc8-btrfs-next-48+ #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.0-0-ga698c8995f-prebuilt.qemu.org 04/01/2014
RIP: 0010:btrfs_free_block_groups+0x314/0x470 [btrfs]
(...)
RSP: 0018:ffff9910c14cfdb8 EFLAGS: 00010286
RAX: 0000000000000024 RBX: ffff89cd8a4d55f0 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffff89cdf6a178a8 RDI: ffff89cdf6a178a8
RBP: ffff9910c14cfde8 R08: 0000000000000000 R09: 0000000000000001
R10: ffff89cd4d618040 R11: 0000000000000000 R12: ffff89cd8a4d5508
R13: ffff89cde7c4a600 R14: dead000000000122 R15: dead000000000100
FS: 00007f42754432c0(0000) GS:ffff89cdf6a00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007fd25a47f730 CR3: 000000021f8d6006 CR4: 00000000003606f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
close_ctree+0x1ad/0x390 [btrfs]
generic_shutdown_super+0x6c/0x110
kill_anon_super+0xe/0x30
btrfs_kill_super+0x12/0xa0 [btrfs]
deactivate_locked_super+0x3a/0x70
cleanup_mnt+0xb4/0x160
task_work_run+0x7e/0xc0
exit_to_usermode_loop+0xfa/0x100
do_syscall_64+0x1cb/0x220
entry_SYSCALL_64_after_hwframe+0x49/0xbe
RIP: 0033:0x7f4274d2cb37
(...)
RSP: 002b:00007ffcff701d38 EFLAGS: 00000246 ORIG_RAX: 00000000000000a6
RAX: 0000000000000000 RBX: 0000557ebde2f060 RCX: 00007f4274d2cb37
RDX: 0000000000000001 RSI: 0000000000000000 RDI: 0000557ebde2f240
RBP: 0000557ebde2f240 R08: 0000557ebde2f270 R09: 0000000000000015
R10: 00000000000006b4 R11: 0000000000000246 R12: 00007f427522ee64
R13: 0000000000000000 R14: 0000000000000000 R15: 00007ffcff701fc0
irq event stamp: 0
hardirqs last enabled at (0): [<0000000000000000>] 0x0
hardirqs last disabled at (0): [<ffffffffb12b561e>] copy_process+0x75e/0x1fd0
softirqs last enabled at (0): [<ffffffffb12b561e>] copy_process+0x75e/0x1fd0
softirqs last disabled at (0): [<0000000000000000>] 0x0
---[ end trace bcf4b235461b26f6 ]---
BTRFS info (device sdb): space_info 4 has 19116032 free, is full
BTRFS info (device sdb): space_info total=33554432, used=14176256, pinned=0, reserved=0, may_use=196608, readonly=65536
BTRFS info (device sdb): global_block_rsv: size 0 reserved 0
BTRFS info (device sdb): trans_block_rsv: size 0 reserved 0
BTRFS info (device sdb): chunk_block_rsv: size 0 reserved 0
BTRFS info (device sdb): delayed_block_rsv: size 0 reserved 0
BTRFS info (device sdb): delayed_refs_rsv: size 0 reserved 0
Fixes: 374bf9c5cd ("btrfs: unify error handling for ticket flushing")
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When debugging weird enospc problems it's handy to be able to dump the
space info when we wake up all tickets, and see what the ticket values
are. This helped me figure out cases where we were enospc'ing when we
shouldn't have been.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We ran into a problem in production where a box with plenty of space was
getting wedged doing ENOSPC flushing. These boxes only had 20% of the
disk allocated, but their metadata space + global reserve was right at
the size of their metadata chunk.
In this case can_overcommit should be allowing allocations without
problem, but there's logic in can_overcommit that doesn't allow us to
overcommit if there's not enough real space to satisfy the global
reserve.
This is for historical reasons. Before there were only certain places
we could allocate chunks. We could go to commit the transaction and not
have enough space for our pending delayed refs and such and be unable to
allocate a new chunk. This would result in a abort because of ENOSPC.
This code was added to solve this problem.
However since then we've gained the ability to always be able to
allocate a chunk. So we can easily overcommit in these cases without
risking a transaction abort because of ENOSPC.
Also prior to now the global reserve really would be used because that's
the space we relied on for delayed refs. With delayed refs being
tracked separately we no longer have to worry about running out of
delayed refs space while committing. We are much less likely to
exhaust our global reserve space during transaction commit.
Fix the can_overcommit code to simply see if our current usage + what we
want is less than our current free space plus whatever slack space we
have in the disk is. This solves the problem we were seeing in
production and keeps us from flushing as aggressively as we approach our
actual metadata size usage.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now that we do not do partial filling of tickets simply remove
orig_bytes, it is no longer needed.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now that we aren't partially filling tickets we may have some slack
space left in the space_info. We need to account for this in
may_commit_transaction, otherwise we may choose to not commit the
transaction despite it actually having enough space to satisfy our
ticket.
Calculate the free space we have in the space_info, if any, and subtract
this from the ticket we have and use that amount to determine if we will
need to commit to reclaim enough space.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Now that we no longer partially fill tickets we need to rework
wake_all_tickets to call btrfs_try_to_wakeup_tickets() in order to see
if any subsequent tickets are able to be satisfied. If our tickets_id
changes we know something happened and we can keep flushing.
Also if we find a ticket that is smaller than the first ticket in our
queue then we want to retry the flushing loop again in case
may_commit_transaction() decides we could satisfy the ticket by
committing the transaction.
Rename this to maybe_fail_all_tickets() while we're at it, to better
reflect what the function is actually doing.
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>