Valgrind was complaining about a jump depending on uninitialized memory
- we weren't, but this change makes the code less confusing for valgrind
to follow.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

This makes the flow control in bch2_btree_iter_peek() and
bch2_btree_iter_peek_prev() a bit cleaner.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Should have been using bpos_cmp(), not bkey_cmp().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

__bch2_read() -> __bch2_read_extent() -> bch2_bucket_io_time_reset() may
cause a transaction restart, which we don't return an error for because
it doesn't prevent us from making forward progress on the read we're
submitting.

Instead, change __bch2_read() and bchfs_read() to check for transaction
restarts.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Inode creation is done with non-cached btree iterators, but then in the
same transaction the inode may be updated again with a cached iterator -
it makes cache coherency easier if new inodes always land in the
underlying btree.

This patch adds a check to bch2_trans_update() - if the same key is
updated multiple times in the same transaction with both cached and non
cache iterators, use the non cached iterator.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

This will be used to make other operations on btree iterators within a
transaction more efficient, and enable some other improvements to how we
manage btree iterators.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This fixes a bad ptr deref on recovery from unclean shutdown in
bch2_btree_node_get_noiter().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

We're only called by bch2_trans_begin() now.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

With the recent transaction restart changes, it's no longer needed - all
transaction commits have BTREE_INSERT_NOUNLOCK semantics.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

With the new and improved handling of transaction restarts, this should
finally be safe.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

It's now the caller's responsibility to call bch2_trans_begin.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

iter->should_be_locked means that if bch2_btree_iter_relock() fails, we
need to restart the transaction.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

If there's more than one iterator in the btree_trans, it's requried to
call bch2_trans_begin() to handle transaction restarts.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Start tracking when btree transactions have been restarted - and assert
that we're always calling bch2_trans_begin() immediately after
transaction restart.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

More consistent behaviour means less likely to trip over ourselves in
silly ways.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Btree node merging now happens prior to transaction commit, not after,
so we don't need to pay attention to BTREE_INSERT_NOUNLOCK.

Also, foreground_maybe_merge shouldn't be calling
bch2_btree_iter_traverse_all() - this is becoming private to the btree
iterator code and should only be called by bch2_trans_begin().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Upcoming patch will require that a transaction restart is always
immediately followed by bch2_trans_begin().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

On transaction restart iterators won't be locked anymore - make sure
we're always checking for errors.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

bch2_btree_iter_traverse_all() may loop, and it needs to clear
iter->should_be_locked on every iteration.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

This avoids unexpected lock restarts in bch2_btree_iter_traverse_all().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

They should already be traversed, and we're asserting that since the
introduction of iter->should_be_locked
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

bch2_btree_node_ptr_v2 has a field for stashing a pointer to the in
memory btree node; this is safe because we clear this field when reading
in nodes from disk and we never free in memory btree nodes - but, we
have bug reports that indicate something might be faulty with this
optimization, so let's add an option for it.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Btree iterator tracepoints should print whether they're for the key
cache.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

This adds a new helper for btree_cache.c that does what we want where
the iterator is still being traverse - and also eliminates some
unnecessary transaction restarts.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

We're working to standardize handling of transaction restarts.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Don't print out the ": " when there isn't a value to print.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

We were squashing BCH_FSCK_ERRORS_NOT_FIXED.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

This is needed for snapshots because we need to start handling lock
restarts even when just calling bch2_inode_peek().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Snapshots add another btree lookup, thus we need to handle lock
restarts.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Downgrading of btree iterators is something that should only happen
explicitly.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

Add a field to struct bset for the sector offset within the btree node
where it was written.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

This reverts commit 664f9847bec525d396d62d2db094ca9020289ae0.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This closes a significant hole (and last known hole) in our ability to
verify metadata. Previously, since btree nodes are log structured, we
couldn't detect lost btree writes that weren't the first write to a
given node. Additionally, this seems to have lead to some significant
metadata corruption on multi device filesystems with metadata
replication: since a write may have made it to one device and not
another, if we read that btree node back from the replica that did have
that write and started appending after that point, the other replica
would have a gap in the bset entries and reading from that replica
wouldn't find the rest of the bsets.

But, since updates to interior btree nodes are now journalled, we can
close this hole by updating pointers to btree nodes after every write
with the currently written number of sectors, without negatively
affecting performance. This means we will always detect lost or corrupt
metadata - it also means that our btree is now a curious hybrid of COW
and non COW btrees, with all the benefits of both (excluding
complexity).
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

We should always print out the full btree node ptr.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

The unit tests hadn't been updated for various recent btree changes -
this patch makes them work again.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

We'd hit a BUG() when rewinding at the start of the btree on btrees with
snapshots.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

The fsck code handles transaction restarts in a very ad hoc way, and not
always correctly. This patch makes some improvements to check_dirents(),
but more work needs to be done to figure out how this kind of code
should be structured.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>