When snapshots arrive, we won't necessarily be able to arbitrarily split
existis - when we need to split an existing extent, we'll have to check
if the extent was overwritten in child snapshots and if so emit a
whiteout for the split in the child snapshot.

Because extents couldn't span btree nodes previously, journal replay
would sometimes have to split existing extents. That's no good anymore,
but fortunately since extent handling has already been lifted above most
of the btree code there's no real need for that rule anymore.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

We need to differentiate between the search position of a btree
iterator, vs. what it actually points at (what we found). This matters
for extents, where iter->pos will typically be the start of the key we
found and iter->real_pos will be the end of the key we found (which soon
won't necessarily be in the same btree node!) and it will also matter
for snapshots.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

The upcoming patch to allow extents to span btree nodes will require
this... and this assertion seems to be popping, and it's not a very good
assertion anyways.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

More repair code, now that we can repair extents during initial gc.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

When allocating an erasure coding stripe, bcachefs will always reuse any
partial stripes before reserving a new stripe. This causes unnecessary
read amplification when preparing a stripe for writing. This patch changes
bcachefs to always reserve new stripes first, only relying on stripe reuse
when copygc needs more time to empty buckets from existing stripes.
Signed-off-by: Robbie Litchfield <blam.kiwi@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

Found by UBSAN
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

This makes bch2_btree_iter_peek_prev() and bch2_btree_iter_prev()
consistent with peek() and next(), w.r.t. iter->pos.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This adds a new common helper for advancing past the last key returned
by peek().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

The only reason we were keeping this around was for
BTREE_INSERT_NOUNLOCK semantics - if bch2_btree_iter_set_pos() advances
to the next leaf node, it'll drop the lock on the node that we just
inserted to.

But we don't rely on BTREE_INSERT_NOUNLOCK semantics for the extents
btree, just the inodes btree, and if we do need it for the extents btree
in the future we can do it more cleanly by cloning the iterator - this
lets us delete some special cases in the btree iterator code, which is
complicated enough as it is.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

There's no good reason for these functions to not be using
bch2_btree_iter_set_pos().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

It's possible we're calling hash_redo_key() because of a duplicate key -
easiest fix for that is to just not use BCH_HASH_SET_MUST_CREATE.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Also, make the wait in bch2_journal_flush_seq() interruptible, not just
killable.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

When the replicas mechanism was added, for tracking data by which drives
it's replicated on, the check for whether we have sufficient devices was
never updated to make use of it. This patch finally does that.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

We're using BCH_FEATURE_alloc_v2 to also gate journalling updates to dev
usage - we don't have the code for reconstructing this from buckets
anymore, so we need to run fsck if it's not set.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This fixes some arithmetic bugs in "bcachefs: Journal updates to dev
usage" - additionally, it cleans things up by switching everything that
goes in every journal entry to the journal_entry_res mechanism.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

This eliminates the need to scan every bucket to regenerate dev_usage at
mount time.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Originally, bcachefs - going back to bcache - stored, for each bucket, a
16 bit counter corresponding to how long it had been since the bucket
was read from. But, this required periodically rescaling counters on
every bucket to avoid wraparound. That wasn't an issue in bcache, where
we'd perodically rewrite the per bucket metadata all at once, but in
bcachefs we're trying to avoid having to walk every single bucket.

This patch switches to persisting 64 bit io clocks, corresponding to the
64 bit bucket timestaps introduced in the previous patch with
KEY_TYPE_alloc_v2.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This introduces a new version of KEY_TYPE_alloc, which uses the new
varint encoding introduced for inodes. This means we'll eventually be
able to support much larger bucket sizes (for SMR devices), and the
read/write time fields are expanded to 64 bits - which will be used in
the next patch to get rid of the periodic rescaling of those fields.

Also, for buckets that are members of erasure coded stripes, this adds
persistent fields for the index of the stripe they're members of and the
stripe redundancy. This is part of work to get rid of having to scan and
read into memory the alloc and stripes btrees at mount time.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This fixes a bug introduced by "bcachefs: Improve diagnostics when
journal entries are missing" - devices in a replicas entry are supposed
to be sorted.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

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

Also, make journal writes obey foreground_target and metadata_target.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Now that we can repair metadata during GC, we can handle bad pointers
that would trigger errors being marked, when they need to just be
dropped.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

When we walk the btrees during recovery, part of that is checking that
btree topology is correct: for every interior btree node, its child
nodes should exactly span the range the parent node covers.

Previously, we had checks for this, but not repair code. Now that we
have the ability to do btree updates during initial GC, this patch adds
that repair code.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Some errors may need to be fixed in order for GC to successfully run -
walk and mark all metadata. But we can't start the allocators and do
normal btree updates until after GC has completed, and allocation
information is known to be consistent, so we need a different method of
doing btree updates.

Fortunately, we already have code for walking the btree while overlaying
keys from the journal to be replayed. This patch adds an update path
that adds keys to the list of keys to be replayed by journal replay, and
also fixes up iterators.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This is so that when we discover btree topology issues, we can just
update the pointer to a btree node and signal btree read path that the
min/max keys in the node header should be updated from the node pointer.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Still a lot of work to be done here: we can't yet repair btree topology
issues, but this patch refactors things so that we have better access to
what we need in the topology checks. Next up will be figuring out a way
to do btree updates during gc, before journal replay is done.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

There's an outstanding bug with journal entries being missing in journal
replay. This patch adds code to print out where the journal entries were
physically located that were around the entry(ies) being missing, which
should make debugging easier.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

Ideally, this limit will be going away in the future.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

The userspace bch_err() macro doesn't use the filesystem argument. Could
also be fixed with a better macro.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

If we're invalidating a bucket that has cached data in it, data_type
won't be 0 - oops.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

More work towards getting rid of the in memory struct bucket: this path
adds code for marking superblock and journal buckets via the btree, and
uses it in the device add and journal resize paths.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This patch is working towards eventually getting rid of the in memory
struct bucket, and relying only on the btree representation.

Since bch2_invalidate_bucket() was only used for incrementing gens, not
invalidating cached data, no other counters were being changed as a side
effect - meaning it's safe for the allocator code to increment the
bucket gen directly.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This is to make it more amenable for serialization.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This was useful before we had transactional updates to interior btree
nodes - but now, it's just extra unneeded complexity.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

This was leading to a very strange bug in bch2_bucket_io_time_reset(),
where we'd retry without clearing out the list of updates.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

If journal replay hasn't finished, the journal can't be empty - oops.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

We're transitioning to memalloc_nofs_save/restore instead of GFP flags
with the rest of the kernel, and GFP_NOIO was excessively strict and
causing unnnecessary allocation failures - these allocations are done
with btree locks dropped.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>