linux-block.git
10 months agobcachefs: Fix some memcpy() warnings
Kent Overstreet [Wed, 28 Dec 2022 20:17:07 +0000 (15:17 -0500)]
bcachefs: Fix some memcpy() warnings

With CONFIG_FORTIFY_SOURCE, the compiler attempts to warn about mempcys
that extend past struct field boundaries. This results in some spurious
warnings where we use embedded variable length structs, this patch
switches to unsafe_mecpy() to fix the warnings.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Be less restrictive when validating journal overwrite entries
Kent Overstreet [Wed, 21 Dec 2022 00:43:41 +0000 (19:43 -0500)]
bcachefs: Be less restrictive when validating journal overwrite entries

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix bch2_journal_flush_device_pins()
Kent Overstreet [Mon, 7 Mar 2022 06:35:55 +0000 (01:35 -0500)]
bcachefs: Fix bch2_journal_flush_device_pins()

It's now legal for the pin fifo to be empty, which means this code needs
to be updated in order to not hit an assert.

Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
10 months agobcachefs: Fix for long running btree transactions & key cache
Kent Overstreet [Fri, 16 Dec 2022 02:44:32 +0000 (21:44 -0500)]
bcachefs: Fix for long running btree transactions & key cache

While a btree transaction is running, we hold a SRCU read lock on the
btree key cache that prevents btree key cache keys from being freed -
this is so that relock() operations won't access freed memory.

The downside of this is that long running btree transactions prevent
memory from being freed from the key cache. This adds a check in
bch2_trans_begin() - if the transaction has been running longer than 1
second, drop and retake the SRCU read lock and zero out pointers to
unlock key cache paths.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Add a missing bch2_err_str() call
Kent Overstreet [Fri, 23 Dec 2022 01:51:02 +0000 (20:51 -0500)]
bcachefs: Add a missing bch2_err_str() call

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Add some unlikely() annotations
Kent Overstreet [Thu, 24 Nov 2022 01:11:46 +0000 (20:11 -0500)]
bcachefs: Add some unlikely() annotations

Add a few easy unlikely() optimizations. These are mainly worthwhile
because the compiler will (usually) put the branch-not-taken path at the
end of the function, meaning better icache utilization.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: New btree helpers
Kent Overstreet [Thu, 24 Nov 2022 03:13:19 +0000 (22:13 -0500)]
bcachefs: New btree helpers

This introduces some new conveniences, to help cut down on boilerplate:

 - bch2_trans_kmalloc_nomemzero() - performance optimiation
 - bch2_bkey_make_mut()
 - bch2_bkey_get_mut()
 - bch2_bkey_get_mut_typed()
 - bch2_bkey_alloc()

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Allow for more btrees
Kent Overstreet [Sat, 3 Dec 2022 01:36:06 +0000 (20:36 -0500)]
bcachefs: Allow for more btrees

Expand some bitfields so we can keep adding more btrees.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Recover from blacklisted journal entries
Kent Overstreet [Wed, 14 Dec 2022 19:47:42 +0000 (14:47 -0500)]
bcachefs: Recover from blacklisted journal entries

If it so happens that we crash while dirty, meaning we don't have the
superblock clean section, and we erroneously mark a journal entry we
wrote as blacklisted, we won't be able to recover.

This patch fixes this by adding a fallback: if we've got no superblock
clean section, and no non-ignored journal entries, we try the most
recent ignored journal entry.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix btree_gc when multiple passes required
Kent Overstreet [Tue, 13 Dec 2022 19:43:03 +0000 (14:43 -0500)]
bcachefs: Fix btree_gc when multiple passes required

We weren't resetting filesystem & device usage when restarting gc, which
was spotted when free bucket counters overflowed - whoops.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix error path in bch2_trans_commit_write_locked()
Kent Overstreet [Tue, 13 Dec 2022 23:19:30 +0000 (18:19 -0500)]
bcachefs: Fix error path in bch2_trans_commit_write_locked()

Previously, we were journalling extra_journal_entries (which is used for
new btree roots, and irreversably mutates system state) before calling
bch2_trans_fs_usage_apply(), which can fail - whoops.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: bch2_trans_revalidate_updates_in_node()
Kent Overstreet [Wed, 23 Nov 2022 23:46:03 +0000 (18:46 -0500)]
bcachefs: bch2_trans_revalidate_updates_in_node()

When we started stashing the key being overwritten in
btree_insert_entry, this introduced a typical iterator invalidation
problem, triggered by btree node splits or resorts.

Previously, dealt with this by unconditionally re-validating those
stashed pointers in the transaction commit path. This patch gets rid of
that by doing it only when needed, in bch2_trans_node_add() or
bch2_trans_node_reinit_iter().

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: bkey_min(), bkey_max()
Kent Overstreet [Fri, 25 Nov 2022 20:01:36 +0000 (15:01 -0500)]
bcachefs: bkey_min(), bkey_max()

Parallel to bpos_min(), bpos_max() - trivial refactoring.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: More errcode cleanup
Kent Overstreet [Sun, 20 Nov 2022 03:39:08 +0000 (22:39 -0500)]
bcachefs: More errcode cleanup

We shouldn't be overloading standard error codes now that we have
provisions for bcachefs-specific errorcodes: this patch converts super.c
and super-io.c to per error site errcodes, with a bit of cleanup.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Suppress -EROFS messages when shutting down
Kent Overstreet [Wed, 12 Oct 2022 20:11:31 +0000 (16:11 -0400)]
bcachefs: Suppress -EROFS messages when shutting down

This isn't actually an error condition, this just indicates a normal
shutdown - no reason for these to be in the log.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Add a missing bch2_btree_path_traverse() call
Kent Overstreet [Fri, 9 Dec 2022 21:22:36 +0000 (16:22 -0500)]
bcachefs: Add a missing bch2_btree_path_traverse() call

bch2_btree_iter_peek_upto() in snapshots mode may need to keep a
btree_path for the insert position, not just the position of the key
we're returning. The code was incorrectly assuming this would be in the
same btree node - we were missing a bch2_btree_path_traverse() call.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix bch2_journal_keys_peek_upto()
Kent Overstreet [Fri, 9 Dec 2022 19:09:14 +0000 (14:09 -0500)]
bcachefs: Fix bch2_journal_keys_peek_upto()

bch2_journal_keys_peek_upto() was comparing against btree_id & level
incorrectly - fix this by using __journal_key_cmp().

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Kill btree_insert_ret enum
Kent Overstreet [Wed, 7 Dec 2022 16:39:34 +0000 (11:39 -0500)]
bcachefs: Kill btree_insert_ret enum

Replace with standard bcachefs-private error codes.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix a btree iter assertion pop
Kent Overstreet [Fri, 2 Dec 2022 17:45:37 +0000 (12:45 -0500)]
bcachefs: Fix a btree iter assertion pop

This fixes a (harmless) broken invariant in __bch2_btree_path_set_pos():
iterators to interior nodes should point to the first non whiteout.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Simplify journal read path
Kent Overstreet [Fri, 2 Dec 2022 02:59:25 +0000 (21:59 -0500)]
bcachefs: Simplify journal read path

This just cleans up and simplifies the code that decides where to resume
writing in the journal - when the code was originally written we weren't
saving the precise location of every journal write found.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix a "no journal entries found" bug
Kent Overstreet [Fri, 2 Dec 2022 16:45:58 +0000 (11:45 -0500)]
bcachefs: Fix a "no journal entries found" bug

On startup, we need to ensure the first journal entry written is a flush
write: after a clean shutdown we generally don't read the journal, which
means we might be overwriting whatever was there previously, and there
must always be at least one flush entry in the journal or recovery will
fail.

Found by fstests generic/388.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Don't error out when just reading the journal
Kent Overstreet [Thu, 1 Dec 2022 16:17:18 +0000 (11:17 -0500)]
bcachefs: Don't error out when just reading the journal

This tweaks the recovery and journal paths so that we don't error out
before we need to: the list_journal command should work, even if we
wouldn't be able to replay successfully.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix a livelock in key cache fill path
Kent Overstreet [Wed, 25 Jan 2023 15:15:39 +0000 (10:15 -0500)]
bcachefs: Fix a livelock in key cache fill path

We weren't setting path->uptodate before calling
bch2_btree_key_cache_fill() - which causes __bch2_btree_path_upgrade()
to fail.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Key cache now works for snapshots btrees
Kent Overstreet [Wed, 23 Nov 2022 01:15:33 +0000 (20:15 -0500)]
bcachefs: Key cache now works for snapshots btrees

This switches btree_key_cache_fill() to use a btree iterator, not a
btree path, so that it can search for keys in previous snapshots.

We also add another iterator flag, BTREE_ITER_KEY_CACHE_FILL, to avoid
recursion back into the key cache.

This will allow us to re-enable the key cache for inodes in the next
patch.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Bring back BTREE_ITER_CACHED_NOFILL
Kent Overstreet [Tue, 20 Dec 2022 16:26:57 +0000 (11:26 -0500)]
bcachefs: Bring back BTREE_ITER_CACHED_NOFILL

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Kill __btree_trans_peek_key_cache()
Kent Overstreet [Tue, 20 Dec 2022 21:02:09 +0000 (16:02 -0500)]
bcachefs: Kill __btree_trans_peek_key_cache()

There was no reason for this to be a separate helper - we always want
the relock call that btree_trans_peek_key_cache() did.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: bch2_btree_path_peek_slot_exact()
Kent Overstreet [Tue, 20 Dec 2022 16:13:19 +0000 (11:13 -0500)]
bcachefs: bch2_btree_path_peek_slot_exact()

When we start using the key cache for inodes again, it'll be possible
for bch2_btree_path_peek_slot() to return a key in a different snapshot
with a key cache path.

This isn't what we want when triggers are checking what they're
overwriting, so introduce a new helper for the commit path.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix __btree_trans_peek_key_cache()
Kent Overstreet [Tue, 20 Dec 2022 15:51:20 +0000 (10:51 -0500)]
bcachefs: Fix __btree_trans_peek_key_cache()

We were returning a pointer to a variable on the stack - oops.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: New bpos_cmp(), bkey_cmp() replacements
Kent Overstreet [Thu, 24 Nov 2022 08:12:22 +0000 (03:12 -0500)]
bcachefs: New bpos_cmp(), bkey_cmp() replacements

This patch introduces
 - bpos_eq()
 - bpos_lt()
 - bpos_le()
 - bpos_gt()
 - bpos_ge()

and equivalent replacements for bkey_cmp().

Looking at the generated assembly these could probably be improved
further, but we already see a significant code size improvement with
this patch.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: New magic number
Kent Overstreet [Sat, 3 Dec 2022 00:46:49 +0000 (19:46 -0500)]
bcachefs: New magic number

Add a new bcachefs-specific magic number for the superblock, instead of
continuing to use the old bcache magic number3

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: extents no longer require special handling for packing
Kent Overstreet [Sat, 26 Nov 2022 09:36:56 +0000 (04:36 -0500)]
bcachefs: extents no longer require special handling for packing

Extent overwrite used to  be handled differently, underneath the
journaling layer and within the core btree code. This imposed
restrictions on bkey packing/packed formats, which no longer apply.

This patch deletes those restrictions.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix BCH_IOCTL_DISK_SET_STATE
Kent Overstreet [Fri, 25 Nov 2022 23:29:36 +0000 (18:29 -0500)]
bcachefs: Fix BCH_IOCTL_DISK_SET_STATE

 - Ensure we print an error message if necessary.

   Ideally we'd return the precise error code to userspace and leave
   printing the error message to the userspace tool, but we haven't
   decided to make our private error codes ABI-stable yet.

 - Return standard error code to userspace

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Don't set accessed bit on btree node fill
Kent Overstreet [Fri, 25 Nov 2022 21:04:42 +0000 (16:04 -0500)]
bcachefs: Don't set accessed bit on btree node fill

Btree nodes shouldn't have their accessed bit set when entering the
btree cache by being read in from disk - this fixes linear scans
thrashing the cache.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix an include
Kent Overstreet [Fri, 25 Nov 2022 21:15:14 +0000 (16:15 -0500)]
bcachefs: Fix an include

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Kill BCH_FEATURE_incompressible
Kent Overstreet [Fri, 25 Nov 2022 17:56:49 +0000 (12:56 -0500)]
bcachefs: Kill BCH_FEATURE_incompressible

This isn't needed anymore, we only support metadata versions that have
this.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Better inlining in bch2_time_stats_update()
Kent Overstreet [Wed, 23 Nov 2022 23:23:48 +0000 (18:23 -0500)]
bcachefs: Better inlining in bch2_time_stats_update()

Move the actual slowpath off into a new function -
bch2_time_stats_clear_buffer() - and inline
bch2_time_stats_update_one().

Alo, use the new inlined update functions from mean_and_variance.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Optimize bch2_trans_iter_init()
Kent Overstreet [Fri, 25 Nov 2022 05:40:27 +0000 (00:40 -0500)]
bcachefs: Optimize bch2_trans_iter_init()

When flags & btree_id are constants, we can constant fold the entire
calculation of the actual iterator flags - and the whole thing becomes
small enough to inline.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: More dio inlining
Kent Overstreet [Fri, 25 Nov 2022 04:52:28 +0000 (23:52 -0500)]
bcachefs: More dio inlining

Eliminate another function call in the O_DIRECT write path.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Kill some unneeded references to c->flags
Kent Overstreet [Wed, 23 Nov 2022 23:51:27 +0000 (18:51 -0500)]
bcachefs: Kill some unneeded references to c->flags

This drops some unneeded references to JOURNAL_REPLAY_DONE in c->flags:
we're already mirroring it in btree_trans, we just weren't using it
consistently.

We may want to do this with more flags:
  btree_iter.c:   unsigned nr = test_bit(BCH_FS_STARTED, &c->flags)
  btree_update_leaf.c:    if (unlikely(!test_bit(BCH_FS_MAY_GO_RW, &c->flags))) {

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Improve bch2_inode_opts_to_opts()
Kent Overstreet [Thu, 24 Nov 2022 01:28:15 +0000 (20:28 -0500)]
bcachefs: Improve bch2_inode_opts_to_opts()

It turns out the *_defined entries of bch_io_opts are only used in one
place - in the xattr get path - and there we immediately convert to a
bch_opts struct, which also has the *_defined entries.

This patch changes bch2_inode_opts_to_opts() to go directly from
bch_inode_unpacked to bch_opts, which is a minor simplification and will
also let us slim down struct bch_io_opts in another patch.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Better inlining in bch2_subvolume_get_snapshot()
Kent Overstreet [Wed, 23 Nov 2022 23:22:59 +0000 (18:22 -0500)]
bcachefs: Better inlining in bch2_subvolume_get_snapshot()

This provides an inlined version of bch2_subvolume_get() and uses it in
bch2_subvolume_get_snapshot(), since this is the version that's used all
over the place and in fast paths (e.g. IO paths).

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Inline bch2_bkey_format_add_key()
Kent Overstreet [Thu, 24 Nov 2022 08:38:31 +0000 (03:38 -0500)]
bcachefs: Inline bch2_bkey_format_add_key()

This is only called in two places, and when it's used we use it in a
tight loop - it's definitely worth inlining.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Tiny bch2_trans_update_by_path_trace() optimization
Kent Overstreet [Wed, 23 Nov 2022 04:47:22 +0000 (23:47 -0500)]
bcachefs: Tiny bch2_trans_update_by_path_trace() optimization

This just removes a redundant comparison - there's more work we could do
here to remove some redundant copying.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Move some asserts behind CONFIG_BCACHEFS_DEBUG
Kent Overstreet [Wed, 23 Nov 2022 03:06:04 +0000 (22:06 -0500)]
bcachefs: Move some asserts behind CONFIG_BCACHEFS_DEBUG

Convert some non-critical asserts in long-stable code to debug asserts.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Split out __bch2_btree_node_get()
Kent Overstreet [Wed, 23 Nov 2022 03:05:45 +0000 (22:05 -0500)]
bcachefs: Split out __bch2_btree_node_get()

Standard splitting out of the slow path from the fast path of a
function. We may follow this up in another patch with inlining the fast
path into btree_iter.c.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Handle last journal write being torn
Kent Overstreet [Sun, 20 Nov 2022 02:40:35 +0000 (21:40 -0500)]
bcachefs: Handle last journal write being torn

If the last journal write didn't complete sucessfully due to a torn
write, we'll detect it as a checksum error. In that case, we should just
pretend that journal entry was never written.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Improve journal_read() logging
Kent Overstreet [Sun, 20 Nov 2022 02:20:58 +0000 (21:20 -0500)]
bcachefs: Improve journal_read() logging

Print out the journal entries we read and will replay as soon as
possible - if we get an error walidating keys it's helpful to know where
it was in the journal.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix a transaction path overflow
Kent Overstreet [Fri, 18 Nov 2022 00:07:40 +0000 (19:07 -0500)]
bcachefs: Fix a transaction path overflow

It turns out we need bch2_extent_trim_atomi() even when we're deleting
extents one at a time because it's possible for one reflink_p to
reference arbitrarily many reflink_v extents. This doesn't normally
happen, but the data move path can fragment existing extents in the
background.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix a race with b->write_type
Kent Overstreet [Thu, 17 Nov 2022 21:03:15 +0000 (16:03 -0500)]
bcachefs: Fix a race with b->write_type

b->write_type needs to be set atomically with setting the
btree_node_need_write flag, so move it into b->flags.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Error message improvement
Kent Overstreet [Wed, 16 Nov 2022 01:25:08 +0000 (20:25 -0500)]
bcachefs: Error message improvement

 - Centralize format strings in bcachefs.h
 - Add bch2_fmt_inum_offset() and related helpers
 - Switch error messages for inodes to also print out the offset, in
   bytes

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Improve a few warnings
Kent Overstreet [Tue, 15 Nov 2022 20:57:07 +0000 (15:57 -0500)]
bcachefs: Improve a few warnings

Warnings ought to always have a format string/log message - makes them
considerably more useful.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix for_each_btree_key2()
Kent Overstreet [Wed, 16 Nov 2022 04:17:55 +0000 (23:17 -0500)]
bcachefs: Fix for_each_btree_key2()

Previously, when we exited from the loop body with a break statement
_ret wouldn't have been assigned to yet, and we could spuriously return
a transaction restart error.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Btree split improvement
Kent Overstreet [Wed, 16 Nov 2022 02:52:12 +0000 (21:52 -0500)]
bcachefs: Btree split improvement

This improves the bkey_format calculation when splitting btree nodes.
Previously, we'd use a format calculated for the original node for the
lower of the two new nodes.

This was particularly bad on sequential insertions, where we iteratively
split the last btree node, whos format has to include KEY_MAX.

Now, we calculate formats precisely for the keys the two new nodes will
contain. This also should make splitting a bit more efficient, since
we're only copying keys once (from the original node to the new node,
instead of new node, replacement node, then upper split).

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix return code from btree_path_traverse_one()
Kent Overstreet [Wed, 16 Nov 2022 03:48:03 +0000 (22:48 -0500)]
bcachefs: Fix return code from btree_path_traverse_one()

trans->restarted is a positive error code, not the usual negative

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Minor dio write path improvements
Kent Overstreet [Mon, 14 Nov 2022 03:43:37 +0000 (22:43 -0500)]
bcachefs: Minor dio write path improvements

This switches where we take quota reservations to be per bch_wirte_op
instead of per dio_write, so we can drop the quota reservation in the
same place as we call i_sectors_acct(), and only take/release
ei_quota_lock once.

In the future we'd like ei_quota_lock to not be a mutex, so that we can
avoid punting to process context before deliving write completions in
nocow mode.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Quota: Don't allocate memory under lock
Kent Overstreet [Mon, 14 Nov 2022 03:35:55 +0000 (22:35 -0500)]
bcachefs: Quota: Don't allocate memory under lock

The genradix code can handle multiple threads trying to allocate at the
same time - we don't need the genradix_ptr_alloc() call to happen under
a lock.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix a use after free
Kent Overstreet [Mon, 14 Nov 2022 07:22:30 +0000 (02:22 -0500)]
bcachefs: Fix a use after free

This fixes a regression from percpu freedlists in the btree key cache
code: in a rare error path, we were immediately freeing a bkey_cached
that had been used before and should've waited for an SRCU barrier.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Delete atomic_inc_bug()
Kent Overstreet [Mon, 14 Nov 2022 01:53:24 +0000 (20:53 -0500)]
bcachefs: Delete atomic_inc_bug()

These were wrappers around atomic operations that verified that the
counter wasn't negative, but they're dead code - delete.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fixes for building in userspace
Kent Overstreet [Mon, 14 Nov 2022 01:01:42 +0000 (20:01 -0500)]
bcachefs: Fixes for building in userspace

 - Marking a non-static function as inline doesn't actually work and is
   now causing problems - drop that

 - Introduce BCACHEFS_LOG_PREFIX for when we want to prefix log messages
   with bcachefs (filesystem name)

 - Userspace doesn't have real percpu variables (maybe we can get this
   fixed someday), put an #ifdef around bch2_disk_reservation_add()
   fastpath

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Factor out two_state_shared_lock
Kent Overstreet [Fri, 4 Nov 2022 17:25:57 +0000 (13:25 -0400)]
bcachefs: Factor out two_state_shared_lock

We have a unique lock used for controlling adding to the pagecache: the
lock has two states, where both states are shared - the lock may be held
multiple times for either state - but not both states at the same time.

This is exactly what we need for nocow mode locking, so this patch pulls
it out of fs.c into its own file.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Kill BCH_WRITE_FLUSH
Kent Overstreet [Thu, 3 Nov 2022 04:29:43 +0000 (00:29 -0400)]
bcachefs: Kill BCH_WRITE_FLUSH

BCH_WRITE_FLUSH is a write flag that causes a journal flush.  It's only
used in the direct IO path, and this will allow for some consolidation
with the regular fsync path, which will help with the upcoming nocow
mode.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: bch2_trans_commit_bkey_invalid()
Kent Overstreet [Thu, 3 Nov 2022 15:14:04 +0000 (11:14 -0400)]
bcachefs: bch2_trans_commit_bkey_invalid()

 - factor out more slowpath code into non-inline function
 - use bch2_print_string_as_lines(), so our error message doesn't get
   truncated

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Kill bch2_alloc_sectors_start()
Kent Overstreet [Wed, 2 Nov 2022 19:41:32 +0000 (15:41 -0400)]
bcachefs: Kill bch2_alloc_sectors_start()

Only used in one place, just inline it there.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Improve __bch2_btree_path_make_mut()
Kent Overstreet [Tue, 1 Nov 2022 08:23:24 +0000 (04:23 -0400)]
bcachefs: Improve __bch2_btree_path_make_mut()

btree_path_copy() doesn't need to call
bch2_btree_path_check_sort_fast() - the newly allocated path will always
be in the correct position, post copy; also delete some redundant
branches from __bch2_btree_path_make_mut().

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Inlining improvements
Kent Overstreet [Tue, 1 Nov 2022 07:37:53 +0000 (03:37 -0400)]
bcachefs: Inlining improvements

 - Don't call into bch2_encrypt_bio() when we're not encrypting
 - Pull slowpath out of trans_lock_write()
 - Make sure bc2h_trans_journal_res_get() gets inlined.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: DIO write path optimization
Kent Overstreet [Tue, 1 Nov 2022 00:30:27 +0000 (20:30 -0400)]
bcachefs: DIO write path optimization

 - With BCH_WRITE_SYNC, we no longer need the completion in struct
   dio_write
 - Pull out bch2_dio_write_copy_iov() into a separate non-inline
   function, it's code that doesn't run in the common case
 - Copy mapping and inode pointers into dio_write, avoiding pointer
   chasing at the start of bch2_dio_write_loop()
 - kthread_use_mm() is not needed in the common case; move it into
   bch2_dio_write_loop_async()
 - factor out various helpers from bch2_dio_write_loop() and rework
   control flow for better icache utilization

Other small optimizations:

 - bch2_keylist_free() is only used in one place, at the end of the
   bch2_write() path - drop the reinit
 - in bch2_disk_reservation_put(), check if res->sectors is nonzero
   before touching c->online_reserved, since that will likely be a cache
   miss

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
bcachefs: More DIO write path optimization

Better code prefetching (?)

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: BCH_WRITE_SYNC
Kent Overstreet [Sat, 29 Oct 2022 19:54:17 +0000 (15:54 -0400)]
bcachefs: BCH_WRITE_SYNC

This adds a new flag for the write path, BCH_WRITE_SYNC, and switches
the O_DIRECT write path to use it when we're not running asynchronously.

It runs the btree update after the write in the original thread's
context instead of a kworker, cutting context switches in half.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: More style fixes
Kent Overstreet [Sat, 22 Oct 2022 19:59:53 +0000 (15:59 -0400)]
bcachefs: More style fixes

Fixes for various checkpatch errors.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Kill BCH_WRITE_JOURNAL_SEQ_PTR
Kent Overstreet [Mon, 31 Oct 2022 02:21:11 +0000 (22:21 -0400)]
bcachefs: Kill BCH_WRITE_JOURNAL_SEQ_PTR

Dead code, delete.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: should_compact_all()
Kent Overstreet [Fri, 28 Oct 2022 22:56:31 +0000 (18:56 -0400)]
bcachefs: should_compact_all()

This factors out a properly-documented helper for deciding when we want
to sort a btree node with MAX_BSETS bsets down to a single bset.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Improved btree write statistics
Kent Overstreet [Fri, 28 Oct 2022 21:08:41 +0000 (17:08 -0400)]
bcachefs: Improved btree write statistics

This replaces sysfs btree_avg_write_size with btree_write_stats, which
now breaks out statistics by the source of the btree write.

Btree writes that are too small are a source of inefficiency, and
excessive btree resort overhead - this will let us see what's causing
them.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Improve fs_usage_apply_warn() message
Kent Overstreet [Mon, 24 Oct 2022 17:34:17 +0000 (13:34 -0400)]
bcachefs: Improve fs_usage_apply_warn() message

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix a spurious warning
Kent Overstreet [Mon, 24 Oct 2022 19:10:14 +0000 (15:10 -0400)]
bcachefs: Fix a spurious warning

Fixes fstests generic/648

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix buffered write path for generic/275
Kent Overstreet [Mon, 24 Oct 2022 02:01:50 +0000 (22:01 -0400)]
bcachefs: Fix buffered write path for generic/275

Per fstests generic/275, on -ENOSPC we're supposed write until the
filesystem is full - i.e. do a partial write instead of failing the full
write.

This is a partial fix for the buffered write path: we'll still fail on a
page boundary.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Journal keys overlay fixes
Kent Overstreet [Sun, 23 Oct 2022 21:37:23 +0000 (17:37 -0400)]
bcachefs: Journal keys overlay fixes

 - In the btree iterator code that overlays keys from the journal, we
   were incorrectly specifying level=0 instead of the btree_path's
   current level in a few places
 - When we didn't do journal replay, we shouldn't free the journal keys:
   this fixes cmd_list and cmd_dump, which run in norecovery mode

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Fix an out-of-bounds shift
Kent Overstreet [Sat, 22 Oct 2022 20:19:27 +0000 (16:19 -0400)]
bcachefs: Fix an out-of-bounds shift

roundup_pow_of_two() is undefined for 0 - oops.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Make error messages more uniform
Kent Overstreet [Sat, 22 Oct 2022 19:10:28 +0000 (15:10 -0400)]
bcachefs: Make error messages more uniform

Use __func__ in error messages that refer to function name, and do so
more uniformly - suggested by checkpatch.pl

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Convert to __packed and __aligned
Kent Overstreet [Sat, 22 Oct 2022 19:00:16 +0000 (15:00 -0400)]
bcachefs: Convert to __packed and __aligned

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: bch2_bkey_cmp_packed_inlined()
Kent Overstreet [Fri, 21 Oct 2022 23:20:09 +0000 (19:20 -0400)]
bcachefs: bch2_bkey_cmp_packed_inlined()

This adds an inlined version of bch2_bkey_cmp_packed(), and uses it in
bch2_sort_keys(), where it's part of the inner loop.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Move bkey bkey_unpack_key() to bkey.h
Kent Overstreet [Fri, 21 Oct 2022 23:15:07 +0000 (19:15 -0400)]
bcachefs: Move bkey bkey_unpack_key() to bkey.h

Long ago, bkey_unpack_key() was added to bset.h instead of bkey.h
because bkey.h didn't include btree_types.h, which it needs for the
compiled unpack function.

This patch finally moves it to the proper location.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Optimize __bch2_btree_node_iter_advance()
Kent Overstreet [Fri, 21 Oct 2022 21:37:42 +0000 (17:37 -0400)]
bcachefs: Optimize __bch2_btree_node_iter_advance()

This replaces an expensive memmove() call with an open-coded version.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Don't touch c->flags in bch2_trans_iter_init()
Kent Overstreet [Fri, 21 Oct 2022 21:26:49 +0000 (17:26 -0400)]
bcachefs: Don't touch c->flags in bch2_trans_iter_init()

This moves the JOURNAL_REPLAY_DONE flag check from
bch2_trans_iter_init() to bch2_trans_init(), where we stash a copy in
btree_trans - gaining us a small performance improvement.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Assorted checkpatch fixes
Kent Overstreet [Wed, 19 Oct 2022 22:31:33 +0000 (18:31 -0400)]
bcachefs: Assorted checkpatch fixes

checkpatch.pl gives lots of warnings that we don't want - suggested
ignore list:

 ASSIGN_IN_IF
 UNSPECIFIED_INT - bcachefs coding style prefers single token type names
 NEW_TYPEDEFS - typedefs are occasionally good
 FUNCTION_ARGUMENTS - we prefer to look at functions in .c files
  (hopefully with docbook documentation), not .h
  file prototypes
 MULTISTATEMENT_MACRO_USE_DO_WHILE
- we have _many_ x-macros and other macros where
  we can't do this

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Optimize bch2_dev_usage_read()
Kent Overstreet [Fri, 21 Oct 2022 18:01:19 +0000 (14:01 -0400)]
bcachefs: Optimize bch2_dev_usage_read()

 - add bch2_dev_usage_read_fast(), which doesn't return by value -
   bch_dev_usage is big enough that we don't want the silent memcpy
 - tweak the allocation path to only call bch2_dev_usage_read() once per
   bucket allocated

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: fix bch2_write_extent() crc corruption.
Daniel Hill [Tue, 11 Oct 2022 08:33:56 +0000 (21:33 +1300)]
bcachefs: fix bch2_write_extent() crc corruption.

crc.compression_type & nouce gets reset to inside bch2_rechecksum_bio(),
we set it back to the previous values calculated. This fixes
incompressible extents being marked as uncompressed.

Signed-off-by: Daniel Hill <daniel@gluo.nz>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: fix security warning in pr_name_and_units
Daniel B. Hill [Mon, 17 Oct 2022 20:54:32 +0000 (09:54 +1300)]
bcachefs: fix security warning in pr_name_and_units

Signed-off-by: Daniel Hill <daniel@gluo.nz>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Don't issue transaction restart on key cache realloc
Kent Overstreet [Mon, 17 Oct 2022 11:32:57 +0000 (07:32 -0400)]
bcachefs: Don't issue transaction restart on key cache realloc

This shouldn't be needed anymore, since we don't rely on the pointer
validity that this was guarding against anymore - we get a new good
reference and save it right after this function.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Separate out flush_new_cached_update()
Kent Overstreet [Mon, 17 Oct 2022 11:31:37 +0000 (07:31 -0400)]
bcachefs: Separate out flush_new_cached_update()

This separates out the slowpath of bch2_trans_update_by_path_trace()
into a new non-inlined helper.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Optimize __bkey_unpack_key_format_checked()
Kent Overstreet [Mon, 17 Oct 2022 11:19:34 +0000 (07:19 -0400)]
bcachefs: Optimize __bkey_unpack_key_format_checked()

Delete some code when CONFIG_BCACHEFS_DEBUG=n

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Inline bch2_inode_pack()
Kent Overstreet [Mon, 17 Oct 2022 11:09:02 +0000 (07:09 -0400)]
bcachefs: Inline bch2_inode_pack()

It's mainly used from bch2_inode_write(), so inline it there.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: bucket_alloc_fail tracepoint should only fire when we have to block
Kent Overstreet [Mon, 17 Oct 2022 11:07:28 +0000 (07:07 -0400)]
bcachefs: bucket_alloc_fail tracepoint should only fire when we have to block

We don't want to fire the bucket_alloc_fail tracepoint on transaction
restart, when we can retry immediately - only when we the allocation
actually has to block.

Also, switch from sched_clock() to local_clock(), as we've been doing
elsewhere.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Optimize bch2_trans_init()
Kent Overstreet [Mon, 17 Oct 2022 11:03:11 +0000 (07:03 -0400)]
bcachefs: Optimize bch2_trans_init()

Now we store the transaction's fn idx in a local variable, instead of
redoing the lookup every time we call bch2_trans_init().

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Split out __btree_path_up_until_good_node()
Kent Overstreet [Mon, 17 Oct 2022 06:04:31 +0000 (02:04 -0400)]
bcachefs: Split out __btree_path_up_until_good_node()

This breaks up btree_path_up_until_good_node() so that only the fastpath
gets inlined.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Btree key cache shrinker fix
Kent Overstreet [Mon, 17 Oct 2022 06:08:07 +0000 (02:08 -0400)]
bcachefs: Btree key cache shrinker fix

The shrinker assumes freed key cache items are ordered by age, so that
it doesn't have to scan the full list to find items that are old enough
(according to the srcu code) to be freed.

But percpu freelists broke this ordering; this patch fixes this by
ensuring we insert items into the proper position.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: __bio_compress() fix up.
Daniel Hill [Sat, 15 Oct 2022 13:25:54 +0000 (02:25 +1300)]
bcachefs: __bio_compress() fix up.

A single block can't be compressed, so it's incompressible.
This stops rebalance repeatably marking extents as uncompressed.

Signed-off-by: Daniel Hill <daniel@gluo.nz>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: make durability a read-write sysfs option
Daniel Hill [Wed, 25 May 2022 04:11:56 +0000 (16:11 +1200)]
bcachefs: make durability a read-write sysfs option

Sometimes the user may need to change durability after formatting to
match current hardware setup, this option provides a quick and flexible
alternative to removing then adding the device.
It is HIGHLY ADVISED TO RUN REREPLICATE after changing this value so the
system doesn't remain degraded.

Signed-off-by: Daniel Hill <daniel@gluo.nz>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: improve behaviour of btree_cache_scan()
Daniel Hill [Thu, 6 Oct 2022 02:53:36 +0000 (15:53 +1300)]
bcachefs: improve behaviour of btree_cache_scan()

Appending new nodes to the end of the list means we're more likely to
evict old entries when btree_cache_scan() is started.

Signed-off-by: Daniel Hill <daniel@gluo.nz>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Quota fixes
Kent Overstreet [Sat, 15 Oct 2022 07:52:28 +0000 (03:52 -0400)]
bcachefs: Quota fixes

 - We now correctly allow soft limits to be exceeded, instead of always
   returning -EDQUOT
 - Disk quota grate times/warnings can now be set, not just the
   systemwide defaults

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Switch to local_clock() for fastpath time source
Kent Overstreet [Sat, 15 Oct 2022 05:03:14 +0000 (01:03 -0400)]
bcachefs: Switch to local_clock() for fastpath time source

local_clock() isn't always completely accurate - e.g. on machines with
TSC drift - but ktime_get_ns() overhead is too high, unfortunately.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
10 months agobcachefs: Btree key cache improvements
Kent Overstreet [Sat, 15 Oct 2022 04:47:21 +0000 (00:47 -0400)]
bcachefs: Btree key cache improvements

 - In userspace, we don't have real percpu variables; this patch
   disables the percpu freelists in userspace
 - add some error messages for the asserts in
   bch2_fs_btree_key_cache_exit(); we've been hitting this (only in
   userspace, oddly), perhaps this will help us track down the error.
 - bkey_cached_reuse() should likely be taking the key cache lock, and
   it's a slowpath so it doesn't hurt to

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>