btrfs: unexport btrfs_run_discard_work and make it static

Mark btrfs_run_discard_work static and move it above its callers.

Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: reinterpret async discard iops_limit=0 as no delay

Currently, a limit of 0 results in a hard coded metering over 6 hours.
Since the default is a set limit, I suspect no one truly depends on this
rather arbitrary setting. Repurpose it for an arguably more useful
"unlimited" mode, where the delay is 0.

Note that if block groups are too new, or go fully empty, there is still
a delay associated with those conditions. Those delays implement
heuristics for not trimming a region we are relatively likely to fully
overwrite soon.

CC: stable@vger.kernel.org # 6.2+
Reviewed-by: Neal Gompa <neal@gompa.dev>
Signed-off-by: Boris Burkov <boris@bur.io>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: set default discard iops_limit to 1000

Previously, the default was a relatively conservative 10. This results
in a 100ms delay, so with ~300 discards in a commit, it takes the full
30s till the next commit to finish the discards. On a workstation, this
results in the disk never going idle, wasting power/battery, etc.

Set the default to 1000, which results in using the smallest possible
delay, currently, which is 1ms. This has shown to not pathologically
keep the disk busy by the original reporter.

Link: https://lore.kernel.org/linux-btrfs/Y%2F+n1wS%2F4XAH7X1p@nz/
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2182228
CC: stable@vger.kernel.org # 6.2+
Reviewed-by: Neal Gompa <neal@gompa.dev
Signed-off-by: Boris Burkov <boris@bur.io>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: reinterpret async discard iops_limit=0 as no delay

Currently, a limit of 0 results in a hard coded metering over 6 hours.
Since the default is a set limit, I suspect no one truly depends on this
rather arbitrary setting. Repurpose it for an arguably more useful
"unlimited" mode, where the delay is 0.

Note that if block groups are too new, or go fully empty, there is still
a delay associated with those conditions. Those delays implement
heuristics for not trimming a region we are relatively likely to fully
overwrite soon.

CC: stable@vger.kernel.org # 6.2+
Reviewed-by: Neal Gompa <neal@gompa.dev>
Signed-off-by: Boris Burkov <boris@bur.io>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: set default discard iops_limit to 1000

Previously, the default was a relatively conservative 10. This results
in a 100ms delay, so with ~300 discards in a commit, it takes the full
30s till the next commit to finish the discards. On a workstation, this
results in the disk never going idle, wasting power/battery, etc.

Set the default to 1000, which results in using the smallest possible
delay, currently, which is 1ms. This has shown to not pathologically
keep the disk busy by the original reporter.

Link: https://lore.kernel.org/linux-btrfs/Y%2F+n1wS%2F4XAH7X1p@nz/
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2182228
CC: stable@vger.kernel.org # 6.2+
Reviewed-by: Neal Gompa <neal@gompa.dev
Signed-off-by: Boris Burkov <boris@bur.io>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: hold block group refcount during async discard

Async discard does not acquire the block group reference count while it
holds a reference on the discard list. This is generally OK, as the
paths which destroy block groups tend to try to synchronize on
cancelling async discard work. However, relying on cancelling work
requires careful analysis to be sure it is safe from races with
unpinning scheduling more work.

While I am unable to find a race with unpinning in the current code for
either the unused bgs or relocation paths, I believe we have one in an
older version of auto relocation in a Meta internal build. This suggests
that this is in fact an error prone model, and could be fragile to
future changes to these bg deletion paths.

To make this ownership more clear, add a refcount for async discard. If
work is queued for a block group, its refcount should be incremented,
and when work is completed or canceled, it should be decremented.

CC: stable@vger.kernel.org # 5.15+
Signed-off-by: Boris Burkov <boris@bur.io>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: update function comments

Update, reformat or reword function comments. This also removes the kdoc
marker so we don't get reports when the function name is missing.

Changes made:

- remove kdoc markers
- reformat the brief description to be a proper sentence
- reword to imperative voice
- align parameter list
- fix typos

Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: move mount option definitions to fs.h

These are fs wide definitions and helpers, move them out of ctree.h and
into fs.h.

Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: fix typos in comments

Fix typos that have snuck in since the last round. Found by codespell.

Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: document now parameter of peek_discard_list

Fixes fs/btrfs/discard.c:203: warning: Function parameter or member 'now' not described in 'peek_discard_list'

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: merge critical sections of discard lock in workfn

btrfs_discard_workfn() drops discard_ctl->lock just to take it again in
a moment in btrfs_discard_schedule_work(). Avoid that and also reuse
ktime.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: fix racy access to discard_ctl data

Because only one discard worker may be running at any given point, it
could have been safe to modify ->prev_discard, etc. without
synchronization, if not for @override flag in
btrfs_discard_schedule_work() and delayed_work_pending() returning false
while workfn is running.

That may lead to torn reads of u64 for some architectures, but that's
not a big problem as only slightly affects the discard rate.

Suggested-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: fix async discard stall

Might happen that bg->discard_eligible_time was changed without
rescheduling, so btrfs_discard_workfn() wakes up earlier than that new
time, peek_discard_list() returns NULL, and all work halts and goes to
sleep without further rescheduling even there are block groups to
discard.

It happens pretty often, but not so visible from the userspace because
after some time it usually will be kicked off anyway by someone else
calling btrfs_discard_reschedule_work().

Fix it by continue rescheduling if block group discard lists are not
empty.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: don't miss async discards after scheduled work override

If btrfs_discard_schedule_work() is called with override=true, it sets
delay anew regardless how much time is left until the timer should have
fired. If delays are long (that can happen, for example, with low
kbps_limit), they might get constantly overridden without having a
chance to run the discard work.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: discard: store async discard delay as ns not as jiffies

Most delay calculations are done in ns or ms, so store
discard_ctl->delay in ms and convert the final delay to jiffies only at
the end.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: discard: speed up async discard up to iops_limit

Instead of using iops_limit only for cutting off extremes, calculate the
discard delay directly from it, so it closely follows iops_limit and
doesn't under-discard even though quotas are not saturated.

The iops limit could be hit more often in some cases and could increase
the discard rate.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: cleanup btrfs_discard_update_discardable usage

This passes in the block_group and the free_space_ctl, but we can get
this from the block group itself. Part of this is because we call it
from __load_free_space_cache, which can be called for the inode cache as
well.

Move that call into the block group specific load section, wrap it in
the right lock that we need for the assertion (but otherwise this is
safe without the lock because this happens in single-thread context).

Fix up the arguments to only take the block group. Add a lockdep_assert
as well for good measure to make sure we don't mess up the locking
again.

Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: discard: add missing put when grabbing block group from unused list

[BUG]
The following small test script can trigger ASSERT() at unmount time:

mkfs.btrfs -f $dev
mount $dev $mnt
mount -o remount,discard=async $mnt
umount $mnt

The call trace:
assertion failed: atomic_read(&block_group->count) == 1, in fs/btrfs/block-group.c:3431
------------[ cut here ]------------
kernel BUG at fs/btrfs/ctree.h:3204!
invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
CPU: 4 PID: 10389 Comm: umount Tainted: G O 5.8.0-rc3-custom+ #68
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 0.0.0 02/06/2015
Call Trace:
btrfs_free_block_groups.cold+0x22/0x55 [btrfs]
close_ctree+0x2cb/0x323 [btrfs]
btrfs_put_super+0x15/0x17 [btrfs]
generic_shutdown_super+0x72/0x110
kill_anon_super+0x18/0x30
btrfs_kill_super+0x17/0x30 [btrfs]
deactivate_locked_super+0x3b/0xa0
deactivate_super+0x40/0x50
cleanup_mnt+0x135/0x190
__cleanup_mnt+0x12/0x20
task_work_run+0x64/0xb0
__prepare_exit_to_usermode+0x1bc/0x1c0
__syscall_return_slowpath+0x47/0x230
do_syscall_64+0x64/0xb0
entry_SYSCALL_64_after_hwframe+0x44/0xa9

The code:
ASSERT(atomic_read(&block_group->count) == 1);
btrfs_put_block_group(block_group);

[CAUSE]
Obviously it's some btrfs_get_block_group() call doesn't get its put
call.

The offending btrfs_get_block_group() happens here:

void btrfs_mark_bg_unused(struct btrfs_block_group *bg)
{
if (list_empty(&bg->bg_list)) {
btrfs_get_block_group(bg);
list_add_tail(&bg->bg_list, &fs_info->unused_bgs);
}
}

So every call sites removing the block group from unused_bgs list should
reduce the ref count of that block group.

However for async discard, it didn't follow the call convention:

void btrfs_discard_punt_unused_bgs_list(struct btrfs_fs_info *fs_info)
{
list_for_each_entry_safe(block_group, next, &fs_info->unused_bgs,
bg_list) {
list_del_init(&block_group->bg_list);
btrfs_discard_queue_work(&fs_info->discard_ctl, block_group);
}
}

And in btrfs_discard_queue_work(), it doesn't call
btrfs_put_block_group() either.

[FIX]
Fix the problem by reducing the reference count when we grab the block
group from unused_bgs list.

Reported-by: Marcos Paulo de Souza <mpdesouza@suse.com>
Fixes: 6e80d4f8c422 ("btrfs: handle empty block_group removal for async discard")
CC: stable@vger.kernel.org # 5.6+
Tested-by: Marcos Paulo de Souza <mpdesouza@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: add correction to handle -1 edge case in async discard

From Dave's testing described below, it's possible to drive a file
system to have bogus values of discardable_extents and _bytes. As
btrfs_discard_calc_delay() is the only user of discardable_extents, we
can correct here for any negative discardable_extents/discardable_bytes.

The problem is not reliably reproducible. The workload that created it
was based on linux git tree, switching between release tags, then
everytihng deleted followed by a full rebalance. At this state the
values of discardable_bytes was 16K and discardable_extents was -1,
expected values 0 and 0.

Repeating the workload again did not correct the bogus values so the
offset seems to be stable once it happens.

Reported-by: David Sterba <dsterba@suse.com>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: add async discard implementation overview

Give a brief overview for how async discard is implemented.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: keep track of discard reuse stats

Keep track of how much we are discarding and how often we are reusing
with async discard. The discard_*_bytes values don't need any special
protection because the work item provides the single threaded access.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: only keep track of data extents for async discard

As mentioned earlier, discarding data can be done either by issuing an
explicit discard or implicitly by reusing the LBA. Metadata block_groups
see much more frequent reuse due to well it being metadata. So instead
of explicitly discarding metadata block_groups, just leave them be and
let the latter implicit discarding be done for them.

For mixed block_groups, block_groups which contain both metadata and
data, we let them be as higher fragmentation is expected.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: have multiple discard lists

Non-block group destruction discarding currently only had a single list
with no minimum discard length. This can lead to caravaning more
meaningful discards behind a heavily fragmented block group.

This adds support for multiple lists with minimum discard lengths to
prevent the caravan effect. We promote block groups back up when we
exceed the BTRFS_ASYNC_DISCARD_MAX_FILTER size, currently we support
only 2 lists with filters of 1MB and 32KB respectively.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: make max async discard size tunable

Expose max_discard_size as a tunable via sysfs and switch the current
fixed maximum to the default value.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: add kbps discard rate limit for async discard

Provide the ability to rate limit based on kbps in addition to iops as
additional guides for the target discard rate. The delay used ends up
being max(kbps_delay, iops_delay).

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: calculate discard delay based on number of extents

An earlier patch keeps track of discardable_extents. These are
undiscarded extents managed by the free space cache. Here, we will use
this to dynamically calculate the discard delay interval.

There are 3 rate to consider. The first is the target convergence rate,
the rate to discard all discardable_extents over the
BTRFS_DISCARD_TARGET_MSEC time frame. This is clamped by the lower
limit, the iops limit or BTRFS_DISCARD_MIN_DELAY (1ms), and the upper
limit, BTRFS_DISCARD_MAX_DELAY (1s). We reevaluate this delay every
transaction commit.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: keep track of discardable_bytes for async discard

Keep track of this metric so that we can understand how ahead or behind
we are in discarding rate. This uses the same accounting method as
discardable_extents, deltas between previous/current values and
propagating them up.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update changelog ]
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: track discardable extents for async discard

The number of discardable extents will serve as the rate limiting metric
for how often we should discard. This keeps track of discardable extents
in the free space caches by maintaining deltas and propagating them to
the global count.

The deltas are calculated from 2 values stored in PREV and CURR entries,
then propagated up to the global discard ctl. The current counter value
becomes the previous counter value after update.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update changelog ]
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: discard one region at a time in async discard

The prior two patches added discarding via a background workqueue. This
just piggybacked off of the fstrim code to trim the whole block at once.
Well inevitably this is worse performance wise and will aggressively
overtrim. But it was nice to plumb the other infrastructure to keep the
patches easier to review.

This adds the real goal of this series which is discarding slowly (ie. a
slow long running fstrim). The discarding is split into two phases,
extents and then bitmaps. The reason for this is two fold. First, the
bitmap regions overlap the extent regions. Second, discarding the
extents first will let the newly trimmed bitmaps have the highest chance
of coalescing when being readded to the free space cache.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: handle empty block_group removal for async discard

block_group removal is a little tricky. It can race with the extent
allocator, the cleaner thread, and balancing. The current path is for a
block_group to be added to the unused_bgs list. Then, when the cleaner
thread comes around, it starts a transaction and then proceeds with
removing the block_group. Extents that are pinned are subsequently
removed from the pinned trees and then eventually a discard is issued
for the entire block_group.

Async discard introduces another player into the game, the discard
workqueue. While it has none of the racing issues, the new problem is
ensuring we don't leave free space untrimmed prior to forgetting the
block_group. This is handled by placing fully free block_groups on a
separate discard queue. This is necessary to maintain discarding order
as in the future we will slowly trim even fully free block_groups. The
ordering helps us make progress on the same block_group rather than say
the last fully freed block_group or needing to search through the fully
freed block groups at the beginning of a list and insert after.

The new order of events is a fully freed block group gets placed on the
unused discard queue first. Once it's processed, it will be placed on
the unusued_bgs list and then the original sequence of events will
happen, just without the final whole block_group discard.

The mount flags can change when processing unused_bgs, so when flipping
from DISCARD to DISCARD_ASYNC, the unused_bgs must be punted to the
discard_list to be trimmed. If we flip off DISCARD_ASYNC, we punt
free block groups on the discard_list to the unused_bg queue which will
do the final discard for us.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs: add the beginning of async discard, discard workqueue

When discard is enabled, everytime a pinned extent is released back to
the block_group's free space cache, a discard is issued for the extent.
This is an overeager approach when it comes to discarding and helping
the SSD maintain enough free space to prevent severe garbage collection
situations.

This adds the beginning of async discard. Instead of issuing a discard
prior to returning it to the free space, it is just marked as untrimmed.
The block_group is then added to a LRU which then feeds into a workqueue
to issue discards at a much slower rate. Full discarding of unused block
groups is still done and will be addressed in a future patch of the
series.

For now, we don't persist the discard state of extents and bitmaps.
Therefore, our failure recovery mode will be to consider extents
untrimmed. This lets us handle failure and unmounting as one in the
same.

On a number of Facebook webservers, I collected data every minute
accounting the time we spent in btrfs_finish_extent_commit() (col. 1)
and in btrfs_commit_transaction() (col. 2). btrfs_finish_extent_commit()
is where we discard extents synchronously before returning them to the
free space cache.

discard=sync:
p99 total per minute p99 total per minute
Drive | extent_commit() (ms) | commit_trans() (ms)
---------------------------------------------------------------
Drive A | 434 | 1170
Drive B | 880 | 2330
Drive C | 2943 | 3920
Drive D | 4763 | 5701

discard=async:
p99 total per minute p99 total per minute
Drive | extent_commit() (ms) | commit_trans() (ms)
--------------------------------------------------------------
Drive A | 134 | 956
Drive B | 64 | 1972
Drive C | 59 | 1032
Drive D | 62 | 1200

While it's not great that the stats are cumulative over 1m, all of these
servers are running the same workload and and the delta between the two
are substantial. We are spending significantly less time in
btrfs_finish_extent_commit() which is responsible for discarding.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>