proper lock ordering is: fs_reclaim -> btree node locks
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

not using unlock_long() blocks key cache reclaim, and the allocator may
take awhile
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This reverts commit 86d81ec5.

This wasn't tested with memcg enabled, it immediately hits a null ptr
deref in list_lru_add().
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

Reported-by: syzbot+e74fea078710bbca6f4b@syzkaller.appspotmail.com
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

this fixes a 'transaction should be locked' error in backpointers fsck
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Closes: https://syzkaller.appspot.com/bug?extid=8996d8f176cf946ef641Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Instead of popping an assert in bch2_write(), WARN and print out some
debugging info.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

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

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

After commit 230e9fc2 ("slab: add SLAB_ACCOUNT flag"), we need to mark
the inode cache as SLAB_ACCOUNT, similar to commit 5d097056 ("kmemcg:
account for certain kmem allocations to memcg")
Signed-off-by: Youling Tang <tangyouling@kylinos.cn>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

originally, stack closures were only used synchronously, and with the
original implementation of closure_sync() the ref never hit 0; thus,
closure_put_after_sub() assumes that if the ref hits 0 it's on the debug
list, in debug mode.

that's no longer true with the current implementation of closure_sync,
so we need a new magic so closure_debug_destroy() doesn't pop an assert.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

We don't have a way to flush a timer that's executing the callback, and
this is simple and limited enough in scope that we can just use the lock
instead.

Needed for the next patch that adds direct wakeups from the allocator to
copygc, where we're now more frequently calling io_timer_del() on an
expiring timer.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fragmentation_lru derives from dirty_sectors, and wasn't being checked.
Co-developed-by: Daniel Hill <daniel@gluo.nz>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

We need to make sure we're not missing any fragmenation entries in the
LRU BTREE after repairing ALLOC BTREE

Also, use the new bch2_btree_write_buffer_maybe_flush() helper; this was
only working without it before since bucket invalidation (usually)
wasn't happening while fsck was running.
Co-developed-by: Daniel Hill <daniel@gluo.nz>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Add a new helper for checking references to write buffer btrees, where
we need a flush before we definitively know we have an inconsistency.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Fixes warnings from bch2_print_allocator_stuck()
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Accidental infinite loop; also fix btree_deadlock_to_text()
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

BCH_READ_NODECODE mode - used by the move paths - really wants to use
only the original rbio, but the retry path really wants to clone - oof.

Make sure to copy the crc of the pointer we read from back to the
original rbio, or we'll see spurious checksum errors later.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

On a new filesystem or device we have to allocate the journal with a
bump allocator, because allocation info isn't ready yet - but when
hot-adding a device that doesn't have a journal, we don't want to use
that path.

Reported-by: syzbot+24a867cb90d8315cccff@syzkaller.appspotmail.com
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Reported-by: syzbot+e5292b50f1957164a4b6@syzkaller.appspotmail.com
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

the unlock is now in read_extent, this fixes an assertion pop in
read_from_stale_dirty_pointer()
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

When allocating too huge a snapshot table, we should fail gracefully
in __snapshot_t_mut() instead of fail in kmalloc().

Reported-by: syzbot+770e99b65e26fa023ab1@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=770e99b65e26fa023ab1
Tested-by: syzbot+770e99b65e26fa023ab1@syzkaller.appspotmail.com
Signed-off-by: Pei Li <peili.dev@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

There's no reason for discards to be single threaded across all devices;
this will improve performance on multi device setups.

Additionally, making them per-device simplifies the refcounting on
bch_dev->io_ref; we now hold it for the duration that the discard path
is running, which fixes a race between the discard path and device
removal.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This series fix the shift-out-of-bounds issue in
bch2_blacklist_entries_gc().

Instead of passing 0 to eytzinger0_first() when iterating the entries,
we explicitly check 0 and initialize i to be 0.

syzbot has tested the proposed patch and the reproducer did not trigger
any issue:

Reported-and-tested-by: syzbot+835d255ad6bc7f29ee12@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=835d255ad6bc7f29ee12Signed-off-by: Pei Li <peili.dev@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Acquire fsck_error_counts_lock before accessing the critical section
protected by this lock.

syzbot has tested the proposed patch and the reproducer did not trigger
any issue.

Reported-by: syzbot+a2bc0e838efd7663f4d9@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=a2bc0e838efd7663f4d9Signed-off-by: Pei Li <peili.dev@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This fixes a rare deadlock when we're doing an emergency shutdown due to
failure to do a journal write.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

This fixes filesystem size not changing on device removal.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

The debug code relies on btree_trans_list being ordered so that it can
resume on subsequent calls or lock restarts.

However, it was using trans->locknig_wait.task.pid, which is incorrect
since btree_trans objects are cached and reused - typically by different
tasks.

Fix this by switching to pointer order, and also sort them lazily when
required - speeding up the btree_trans_get() fastpath.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

debug.c was using closure_get() on a different thread's closure where
the we don't know if the object being refcounted is alive.

We keep btree_trans objects on a list so they can be printed by debug
code, and because it is cost prohibitive to touch the btree_trans list
every time we allocate and free btree_trans objects, cached objects are
also on this list.

However, we do not want the debug code to see cached but not in use
btree_trans objects - critically because the btree_paths array will have
been freed (if it was reallocated).

closure_get() is also incorrect to use when that get may race with it
hitting zero, i.e. we must already have a ref on the object or know the
ref can't currently hit 0 for other reasons (as used in the cycle
detector).

to fix this, use the previously introduced closure_get_not_zero(),
closure_return_sync(), and closure_init_stack_release(); the debug code
now can only take a ref on a trans object if it's alive and in use.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Provide new primitives for solving a lifetime issue with bcachefs
btree_trans objects.

closure_sync_return(): like closure_sync(), wait synchronously for any
outstanding gets. like closure_return, the closure is considered
"finished" and the ref left at 0.

closure_get_not_zero(): get a ref on a closure if it's alive, i.e. the
ref is not zero.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

btree_deadlock_to_text() searches the list of btree transactions to find
a deadlock - when it finds one it's done; it's not like other *_read()
functions that's printing each object.

Factor out btree_deadlock_to_text() to make this clearer.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

We were grabbing the sequence number before unlock incremented it - fix
this by moving the increment to seqmutex_lock() (so the seqmutex_relock()
failure path skips the mutex_trylock()), and returning the sequence
number from unlock(), to make the API simpler and safer.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This fixes incorrect/missign checking of strndup_user() returns.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>