bch2_inode_delete_keys() was using BTREE_ITER_NOT_EXTENTS, on the
assumption that it would never need to split extents.

But that caused a race with extents being split by other threads -
specifically, the data move path. Extents iterators have the iterator
position pointing to the start of the extent, which avoids the race.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

The entire btree will be lost, but that is better than the entire
filesystem not being recoverable.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

We can only do this in userspace, unfortunately - but kernel keyrings
have never seemed to worked reliably, this is a useful fallback.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

These errors aren't actual errors, and should never be printed - do this
in the common helpers.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

 - assert in shutdown path that no nocow locks are held
 - check for overflow when taking nocow locks
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>

This makes mount option handling consistent with other filesystems -
options may be handled at different layers, so an option we don't know
about might not be intended for us.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Previously, we would check for invalid bkeys at transaction commit time,
but only if CONFIG_BCACHEFS_DEBUG=y.

This check is important enough to always be on - it appears there's been
corruption making it into the journal that would have been caught by it.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Previously, equiv was set in the snapshot deletion path, which is where
it's needed - equiv, for snapshot ID equivalence classes, would ideally
be a private data structure to the snapshot deletion path.

But if a new snapshot is created while snapshot deletion is running,
move_key_to_correct_snapshot() moves a key to snapshot id 0 - oops.

Fixes: https://github.com/koverstreet/bcachefs/issues/593Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

Signed-off-by: Torge Matthies <openglfreak@googlemail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Initial support for the vfs superblock freeze and unfreeze
operations. Superblock freeze occurs in stages, where the vfs
attempts to quiesce high level write operations, page faults, fs
internal operations, and then finally calls into the filesystem for
any last stage steps (i.e. log flushing, etc.) before marking the
superblock frozen.

The majority of write paths are covered by freeze protection (i.e.
sb_start_write() and friends) in higher level common code, with the
exception of the fs-internal SB_FREEZE_FS stage (i.e.
sb_start_intwrite()). This typically maps to active filesystem
transactions in a manner that allows the vfs to implement a barrier
of internal fs operations during the freeze sequence. This is not a
viable model for bcachefs, however, because it utilizes transactions
both to populate the journal as well as to perform journal reclaim.
This means that mapping intwrite protection to transaction lifecycle
or transaction commit is likely to deadlock freeze, as quiescing the
journal requires transactional operations blocked by the final stage
of freeze.

The flipside of this is that bcachefs does already maintain its own
internal sets of write references for similar purposes, currently
utilized for transitions from read-write to read-only mode. Since
this largely mirrors the high level sequence involved with freeze,
we can simply invoke this mechanism in the freeze callback to fully
quiesce the filesystem in the final stage. This means that while the
SB_FREEZE_FS stage is essentially a no-op, the ->freeze_fs()
callback that immediately follows begins by performing effectively
the same step by quiescing all internal write references.

One caveat to this approach is that without integration of internal
freeze protection, write operations gated on internal write refs
will fail with an internal -EROFS error rather than block on
acquiring freeze protection. IOW, this is roughly equivalent to only
having support for sb_start_intwrite_trylock(), and not the blocking
variant. Many of these paths already use non-blocking internal write
refs and so would map into an sb_start_intwrite_trylock() anyways.
The only instance of this I've been able to uncover that doesn't
explicitly rely on a higher level non-blocking write ref is the
bch2_rbio_narrow_crcs() path, which updates crcs in certain read
cases, and Kent has pointed out isn't critical if it happens to fail
due to read-only status.

Given that, implement basic freeze support as described above and
leave tighter integration with internal freeze protection as a
possible future enhancement. There are multiple potential ideas
worth exploring here. For example, we could implement a multi-stage
freeze callback that might allow bcachefs to quiesce its internal
write references without deadlocks, we could integrate intwrite
protection with bcachefs' internal write references somehow or
another, or perhaps consider implementing blocking support for
internal write refs to be used specifically for freeze, etc. In the
meantime, this enables functional freeze support and the associated
test coverage that comes with it.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

 - fix a few uninitialized return values
 - return a proper error code in lookup_lostfound()
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

 - it's no longer possible for trans to be NULL
 - also, move "wait for read to complete" to the slowpath,
   __bch2_btree_node_get().
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

kvfree_rcu() was renamed - not removed.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

strndup_user() returns an error pointer, not NULL.
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

bch2_journal_write() expects process context, it takes journal_lock as
needed.
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

crypto_alloc_sync_skcipher() returns an ERR_PTR, not NULL.
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

When bucket sector counts were changed from u16s to u32s, a few things
were missed. This fixes an overflow check, and a truncation that
prevented the overflow check from firing.
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

copy_to_user() returns the number of bytes successfully copied - not an
errcode.
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

bcachefs freeze testing via fstests generic/390 occasionally
reproduces the following BUG from bch2_fs_read_only():

  BUG_ON(atomic_long_read(&c->btree_key_cache.nr_dirty));

This indicates that one or more dirty key cache keys still exist
after the attempt to flush and quiesce the fs. The sequence that
leads to this problem actually occurs on unfreeze (ro->rw), and
looks something like the following:

- Task A begins a transaction commit and acquires journal_res for
  the current seq. This transaction intends to perform key cache
  insertion.
- Task B begins a bch2_journal_flush() via bch2_sync_fs(). This ends
  up in journal_entry_want_write(), which closes the current journal
  entry and drops the reference to the pin list created on entry open.
  The pin put pops the front of the journal via fast reclaim since the
  reference count has dropped to 0.
- Task A attempts to set the journal pin for the associated cached
  key, but bch2_journal_pin_set() skips the pin insert because the
  seq of the transaction reservation is behind the front of the pin
  list fifo.

The end result is that the pin associated with the cached key is not
added, which prevents a subsequent reclaim from processing the key
and thus leaves it dangling at freeze time. The fundamental cause of
this problem is that the front of the journal is allowed to pop
before a transaction with outstanding reservation on the associated
journal seq is able to add a pin. The count for the pin list
associated with the seq drops to zero and is prematurely reclaimed
as a result.

The logical fix for this problem lies in how the journal buffer is
managed in similar scenarios where the entry might have been closed
before a transaction with outstanding reservations happens to be
committed.

When a journal entry is opened, the current sequence number is
bumped, the associated pin list is initialized with a reference
count of 1, and the journal buffer reference count is bumped (via
journal_state_inc()). When a journal reservation is acquired, the
reservation also acquires a reference on the associated buffer. If
the journal entry is closed in the meantime, it drops both the pin
and buffer references held by the open entry, but the buffer still
has references held by outstanding reservation. After the associated
transaction commits, the reservation release drops the associated
buffer references and the buffer is written out once the reference
count has dropped to zero.

The fundamental problem here is that the lifecycle of the pin list
reference held by an open journal entry is too short to cover the
processing of transactions with outstanding reservations. The
simplest way to address this is to expand the pin list reference to
the lifecycle of the buffer vs. the shorter lifecycle of the open
journal entry. This ensures the pin list for a seq with outstanding
reservation cannot be popped and reclaimed before all outstanding
reservations have been released, even if the associated journal
entry has been closed for further reservations.

Move the pin put from journal entry close to where final processing
of the journal buffer occurs. Create a duplicate helper to cover the
case where the caller doesn't already hold the journal lock. This
allows generic/390 to pass reliably.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

bcachefs freeze testing has uncovered some raciness between journal
entry open/close and pin list reference count management. The
details of the problem are described in a separate patch. In
preparation for the associated fix, refactor the journal buffer put
path a bit to allow it to eventually handle dropping the pin list
reference currently held by an open journal entry.

Retain the journal write dispatch helper since the closure code is
inlined and we don't want to increase the amount of inline code in
the transaction commit path, but rename the function to reflect
the purpose of final processing of the journal buffer.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

We have a couple journal pin put helpers to handle cases where the
journal lock is already held or not. Refactor the helpers to lock
and reclaim from the highest level and open code the reclaim from
the one caller of the internal variant. The latter call will be
moved into the journal buf release helper in a later patch.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This code accidentally left out the "ret = " assignment so the errors
from for_each_btree_key2() are not checked.

Fixes: 53534482a250 ("bcachefs: for_each_btree_key2()")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

The copy_to_user() function returns the number of bytes that it wasn't
able to copy but we want to return -EFAULT to the user.

Fixes: e0750d947352 ("bcachefs: Initial commit")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

The "ret = bkey_err(k);" assignment was accidentally left out so the
call to bch2_btree_iter_peek_slot() is not checked for errors.

Fixes: 53306e096d91 ("bcachefs: Always check for transaction restarts")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

The clean up code at the end of the function uses "acl" so it needs
to be initialized to NULL.

Fixes: 53306e096d91 ("bcachefs: Always check for transaction restarts")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Fixes the following observed error reported by Nathan on IRC.

  fs/bcachefs/io_misc.c:467:6: error: explicitly assigning value of
  variable of type 'int' to itself [-Werror,-Wself-assign]
    467 |         ret = ret;
        |         ~~~ ^ ~~~
Reported-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

./fs/bcachefs/btree_update.h: journal.h is included more than once.
Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Closes: https://bugzilla.openanolis.cn/show_bug.cgi?id=6573Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

There is a typo here where it uses ";" instead of "?:".  The result is
that bch2_fs_fs_io_direct_init() is called unconditionally and the errors
from it are not checked.

Fixes: 0060c68159fc ("bcachefs: Split up fs-io.[ch]")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Reviewed-by: Brian Foster <bfoster@redhat.com>

On 32 bit systems, "sizeof(*arg) + replica_entries_bytes" can have an
integer overflow leading to memory corruption.  Use size_add() to
prevent this.

Fixes: b44dd3797034 ("bcachefs: Redo filesystem usage ioctls")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

The copy_to_user() function returns the number of bytes remaining but
we want to return -EFAULT to the user.

Fixes: e0750d947352 ("bcachefs: Initial commit")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

bucket_lock() previously open coded a spinlock, because we need to cram
a spinlock into a single byte.

But it turns out not all archs support xchg() on a single byte; since we
need struct bucket to be small, this means we have to play fun games
with casts and ifdefs for endianness.

This fixes building on 32 bit arm, and likely other architectures.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Cc: linux-bcachefs@vger.kernel.org
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Per previous commit, bare unreachable() considered harmful, convert to
BUG()
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

In general it's a good idea to avoid using bare unreachable() because it
introduces undefined behavior in compiled code.  In this case it even
confuses GCC into emitting an empty unused
bch2_dev_buckets_reserved.part.0() function.

Use BUG() instead, which is nice and defined.  While in theory it should
never trigger, if something were to go awry and the BCH_WATERMARK_NR
case were to actually hit, the failure mode is much more robust.

Fixes the following warnings:

  vmlinux.o: warning: objtool: bch2_bucket_alloc_trans() falls through to next function bch2_reset_alloc_cursors()
  vmlinux.o: warning: objtool: bch2_dev_buckets_reserved.part.0() is missing an ELF size annotation
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Remove a redundant call to bch2_free_super().

This is harmless because bch2_free_super() has a memset() at its end. So
a second call would only lead to from kfree(NULL).

Remove the redundant call and only rely on the error handling path.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>