This adds a debug mode where we split up the c->writes refcount into
distinct refcounts for every codepath that takes a reference, and adds
sysfs code to print the value of each ref.
This will make it easier to debug shutdown hangs due to refcount leaks.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The btree_node_write_blocked bit was a later addition to this code,
it only mirrors the state of the b->write_blocked list (empty or
nonempty) - unfortunately, when it was added it wasn't correctly kept in
sync - oops.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
In debug mode, we now track where btree iterators and paths are
initialized/allocated - helpful in tracking down btree path overflows.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
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>
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>
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>
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>
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>
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>
btree nodes can be written by other threads (shrinker, journal reclaim)
with only a read lock, but brand new nodes should only be written by the
thread doing the split/interior update. bch2_btree_update_add_new_node()
sets btree node flags to indicate that this is a new node and should not
be written out by other threads, thus we need to call it before dropping
our write lock.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Previously, bch2_btree_update_start() would always take all intent
locks, all the way up to the root.
We've finally got data from users where this became a scalability issue
- so, this patch fixes bch2_btree_update_start() to only take the locks
we need.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Now that we have an error path plumbed through, there's no need to be
using bch2_btree_node_lock_write_nofail().
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The next patch in the series is (finally!) going to change btree splits
(and interior updates in general) to not take intent locks all the way
up to the root - instead only locking the nodes they'll need to modify.
However, this will be introducing a race since if we're not holding a
write lock on a btree node it can be written out by another thread, and
then we might not have enough space for a new bset entry.
We can handle this by retrying - we just need to introduce a new error
path.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
In order to avoid locking all btree nodes up to the root for btree node
splits, we're going to have to introduce a new error path into
bch2_btree_insert_node(); this mean we can't have done any writes or
modified global state before that point.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
btree_node_lock_nopath() is something we'd like to get rid of, it's
always prone to deadlocks if we accidentally are holding other locks,
because it doesn't mark the lock it's taking in a path: we'll want to
get rid of it in the future, but for now this patch works it by calling
bch2_trans_unlock().
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This function is fairly small and only used in two places: one very hot,
the other cold, so it should definitely be inlined.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This is a major oopsy - we should always be unlocking before calling
closure_sync(), else we'll cause a deadlock.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We were checking for -EAGAIN, but we're not returned that when we didn't
pass a closure to wait with - oops.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
With the new deadlock cycle detector, it's critical that all held locks
be marked in a btree_path, because that's what the cycle detector
traverses - any locks that aren't correctly marked will cause deadlocks.
This changes the btree_path to allocate some btree_paths for the new
nodes, since until the final update is done we otherwise don't have a
path referencing them.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Centralizing the transaction restart/tracepoint in
bch2_btree_path_upgrade() lets us improve the tracepoint - now it emits
old and new locks_want.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
With the upcoming cycle detector, we have to be careful about using
btree_node_lock_nopath - in particular, using it to take write locks can
cause deadlocks.
All held locks need to be tracked in a btree_path, so that the cycle
detector knows about them - unless we know that we cannot cause
deadlocks for other reasons: e.g. we are only taking read locks, or
we're in very early fsck (topology repair).
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Ideally, all the code in btree_locking.c should be converted, but then
we'd want to convert btree_path to point to btree_key_cached_common too,
and then we'd be in for a much bigger cleanup - but a bit of incremental
cleanup will still be helpful for the next patches.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Taking a write lock will be able to fail, with the new cycle detector -
unless we pass it nofail, which is possible but not preferred.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
In the future, with the new deadlock cycle detector, we won't be using
bare six_lock_* anymore: lock wait entries will all be embedded in
btree_trans, and we will need a btree_trans context whenever locking a
btree node.
This patch plumbs a btree_trans to the few places that need it, and adds
two new locking functions
- btree_node_lock_nopath, which may fail returning a transaction
restart, and
- btree_node_lock_nopath_nofail, to be used in places where we know we
cannot deadlock (i.e. because we're holding no other locks).
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Also, do some reorganizing/renaming, convert atomic counters in bch_fs
to persistent counters, and add a few missing counters.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
On failure to get a journal pre-reservation because we're called from
journal reclaim we're not supposed to return a transaction restart error
- this fixes a livelock.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Our types are exported to the tracepoint code, so it's not necessary to
break things out individually when passing them to tracepoints - we can
also call other functions from TP_fast_assign().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
It doesn't make any sense to set should_be_locked on btree_paths that
aren't locked, and is often a bug - this patch adds assertions and fixes
some of those bugs.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Now that we have error codes, with subtypes, we can switch to our own
error code for transaction restarts - and even better, a distinct error
code for each transaction restart reason: clearer code and better
debugging.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
We need the caller name and a place to store our results, btree_trans provides this.
Signed-off-by: Daniel Hill <daniel@gluo.nz>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Better/more descriptive naming, and prep for adding
nested_lockrestart_do() and nested_commit_do().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
If we're trying to get a ref and the refcount has been killed, it means
we're doing an emergency shutdown - we always want tryget_live().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This converts bcachefs to the modern printbuf interface/implementation,
synced with the version to be submitted upstream.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Delete some obsolete tracepoints, organize alloc tracepoints better,
make a few tracepoints more consistent.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Can't take btree node locks while holding btree_reserve_cache_lock - it
would be nice if we could check this with lockdep.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
We're seeing occasional firings of the assertion in the key cache
shutdown code that nr_dirty == 0, which means we must sometimes be doing
transaction commits after we've gone read only.
Cleanups & changes:
- BCH_FS_ALLOC_CLEAN renamed to BCH_FS_CLEAN_SHUTDOWN
- new helper bch2_btree_interior_updates_flush(), which returns true if
it had to wait
- bch2_btree_flush_writes() now also returns true if there were btree
writes in flight
- __bch2_fs_read_only now checks if btree writes were in flight in the
shutdown loop: btree write completion does a transaction update, to
update the pointer in the parent node
- assert that !BCH_FS_CLEAN_SHUTDOWN in __bch2_trans_commit
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Previously, btree_update_interior.c passed keys to bch2_trans_mark_*
that hadn't been fully initialized - they didn't have the key field
filled out, just the value.
With backpointers, we need to make sure keys are fully initialized
before marking them - because the backpointer points back to the
original key.
This patch tweaks the interior update paths to fix this.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
For backpointers, we'll need the full key location - that means btree_id
and btree level. This patch plumbs it through.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This adds a new parameter to .key_invalid() methods for whether the key
is being read or written; the idea being that methods can do more
aggressive checks when a key is newly created and being written, when we
wouldn't want to delete the key because of those checks.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Now that we have new persistent data structures for the allocator, this
patch converts the allocator to use them.
Now, foreground bucket allocation uses the freespace btree to find
buckets to allocate, instead of popping buckets off the freelist.
The background allocator threads are no longer needed and are deleted,
as well as the allocator freelists. Now we only need background tasks
for invalidating buckets containing cached data (when we are low on
empty buckets), and for issuing discards.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
In the write path, after the write to the block device(s) complete we
have to punt to process context to do the btree update.
Instead of using the work item embedded in op->cl, this patch switches
to a per write-point work item. This helps with two different issues:
- lock contention: btree updates to the same writepoint will (usually)
be updating the same alloc keys
- context switch overhead: when we're bottlenecked on btree updates,
having a thread (running out of a work item) checking the write point
for completed ops is cheaper than queueing up a new work item and
waking up a kworker.
In an arbitrary benchmark, 4k random writes with fio running inside a
VM, this patch resulted in a 10% improvement in total iops.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This simplifies the logic in bch2_btree_update_start() a bit, handling
the unlock/block logic more locally.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Since journal reclaim -> btree key cache flushing may require the
allocation of new btree nodes, it has an implicit dependency on copygc
in order to make forward progress - so we should avoid blocking copygc
unless the journal is really close to full.
This introduces watermarks to replace our single MAY_GET_UNRESERVED bit
in the journal, and adds a watermark for copygc and plumbs it through.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
bch2_btree_iter_next_node() was mucking with other btree_path state
without setting path->update to be consistent with the fact that the
path is very much no longer uptodate - oops.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Six locks have a percpu mode, which we use for interior btree nodes, as
well as btree key cache keys for the subvolumes btree. We've been
switching locks back and forth between percpu and non percpu mode as
needed, but it turns out this is racy - when we're reusing an existing
node, other threads could be attempting to lock it while we're switching
it between modes.
This patch fixes this by never switching 'struct btree' between the two
modes, and instead segragating them between two different freed lists.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We don't need to pass the number of nodes required to
bch2_btree_update_start, just whether we're doing a split at @level.
This is prep work for a fix to our usage of six lock's percpu mode,
which is going to require us to count up and allocate interior nodes and
leaf nodes seperately.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
When we do an interior btree update, we create new btree nodes and link
them into the btree in memory, but they don't become reachable on disk
until later, when btree_update_nodes_written_trans() runs.
Updates to the new nodes can thus happen before they're reachable on
disk, and if the updates to those new nodes are written before the nodes
become reachable, we would then drop the journal pin for those updates
before the btree has them.
This is what the journal pin in bch2_btree_update_start() was protecting
against. However, it's not actually needed because we don't allow
subsequent append writes to btree nodes until the node is reachable on
disk.
Dropping this unneeded pin also fixes a bug introduced by "bcachefs:
Journal seq now incremented at entry open, not close" - in the new code,
if the journal is completely empty a journal pin list for
journal_cur_seq() won't exist.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Checking btree_node_may_write() isn't atomic with the other btree flags,
dirty and need_write in particular. There was a rare race where we'd
unblock a node from writing while __btree_node_flush() was setting
need_write, and no thread would notice that the node was now both able
to write and needed to be written.
Fix this by adding btree node flags for will_make_reachable and
write_blocked that can be checked in the cmpxchg loop in
__bch2_btree_node_write.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
btree_node_write_if_need() kicks off a btree node write only if
need_write is set; this makes the locking easier to reason about by
moving the check into the cmpxchg loop in __bch2_btree_node_write().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
In btree_update_interior.c, we were changing a path's level directly -
which affects path sort order - without re-sorting paths, leading to
assertions when bch2_path_get() verified paths were sorted correctly.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This patch changes printbufs dynamically allocate and reallocate a
buffer as needed. Stack usage has become a bit of a problem, and a major
cause of that has been static size string buffers on the stack.
The most involved part of this refactoring is that printbufs must now be
exited with printbuf_exit().
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We're now coming up with triggers that modify the update being done. A
bkey_s_c is const - bkey_i is the correct type to be using here.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
BTREE_NODE_SEQ() is supposed to give us a time ordering of btree nodes
on disk, so that we can tell which btree node is newer if we ever have
to scan the entire device to find btree nodes.
The btree node merge path wasn't setting it correctly on the new node -
oops.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This consolidates some of the btree node lock path, so that when we're
blocked taking a write lock on a node it shows up in
bch2_btree_trans_to_text(), along with intent and read locks.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
These nodes aren't reachable by other threads, so there's no need to
keep it locked - and this fixes a bug with the assertion in
bch2_trans_unlock() firing on transaction restart.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
bch2_btree_update_key() is used in the btree node write path - before
delivering the completion we have to update the parent pointer with the
number of sectors written.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Symbol decoding, via %ps, isn't supported in userspace - this will also
be faster when we're using trans->fn in the fast path, as with the new
BCH_JSET_ENTRY_log journal messages.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
With BTREE_ITER_WITH_JOURNAL, there's no longer any restrictions on the
order we have to replay keys from the journal in, and we can also start
up journal reclaim right away - and delete a bunch of code.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This adds a new btree iterator flag, BTREE_ITER_WITH_JOURNAL, that is
automatically enabled when initializing a btree iterator before journal
replay has completed - it overlays the contents of the journal with the
btree.
This lets us delete bch2_btree_and_journal_walk() and just use the
normal btree iterator interface instead - which also lets us delete a
significant amount of duplicated code.
Note that BTREE_ITER_WITH_JOURNAL is still unoptimized in this patch -
we're redoing the binary search over keys in the journal every time we
call bch2_btree_iter_peek().
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This adds flags for options that must be a power of two (block size and
btree node size), and options that are stored in the superblock as a
power of two (encoded extent max).
Also: options are now stored in memory in the same units they're
displayed in (bytes): we now convert when getting and setting from the
superblock.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This adds a flag to not mark the initial btree_path as preserve, for
paths that we expect to be cheap to reconstitute if necessary - this
solves a btree_path overflow caused by need_whiteout_for_snapshot().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This adds more latency/event measurements and breaks some apart into
more events. Journal writes are broken apart into flush writes and
noflush writes, btree compactions are broken out from btree splits,
btree mergers are added, as well as btree_interior_updates - foreground
and total.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
We have been getting away from handling transaction restarts locally -
convert bch2_btree_node_rewrite() to the newer style.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This changes for_each_btree_node() to work like for_each_btree_key(),
and to that end bch2_btree_iter_peek_node() and next_node() also return
error ptrs.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
- check for getting to the end of the btree in bch2_path_verify_locks
and __btree_path_traverse_all(), this fixes an infinite loop in
__btree_path_traverse_all().
- relax requirement in bch2_btree_node_upgrade() that we must want an
intent lock, this fixes bugs with paths that point to interior nodes
(nonzero level).
- bch2_btree_node_update_key(): fix it to upgrade the path to an intent
lock, if necessary
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
We can end up in a strange situation where a btree_path points to a node
being freed even after pointers to it should have been replaced by
pointers to the new node - if the btree node has been reused since the
pointer to it was created.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
New rule is: if a btree path holds any locks it should be holding
precisely the locks wanted (accoringing to path->level and
path->locks_want).
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This splits btree_iter into two components: btree_iter is now the
externally visible componont, and it points to a btree_path which is now
reference counted.
This means we no longer have to clone iterators up front if they might
be mutated - btree_path can be shared by multiple iterators, and cloned
if an iterator would mutate a shared btree_path. This will help us use
iterators more efficiently, as well as slimming down the main long lived
state in btree_trans, and significantly cleans up the logic for iterator
lifetimes.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
These utility functions are for managing btree node state within a
btree_trans - rename them for consistency, and drop some unneeded
arguments.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This is prep work for splitting btree_path out from btree_iter -
btree_path will not have a pointer to btree_trans.
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>
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>
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>
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>
On fstest generic/388, we were seeing sporadic deadlocks in the
emergency shutdown, where we'd get stuck shutting down the allocator
because bch2_btree_update_start() -> bch2_btree_reserve_get() allocated
and then deallocated some btree nodes, putting them back on the
btree_reserve_cache, after the allocator shutdown code had already
cleared out that cache.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This is something we've attempted to stick to for quite some time, as it
helps guarantee filesystem latency - but there's a few remaining paths
that this patch fixes.
This is also necessary for an upcoming patch to update btree pointers
after every btree write - since the btree write completion path will now
be doing btree operations.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
btree_trans should always be passed when we have one - iter->trans is
disfavoured. This mainly updates old code in btree_update_interior.c,
some of which predates btree_trans.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Internal btree code really wants a POS_MAX with all fields ~0; external
code more likely wants the snapshot field to be 0, because when we're
passing it to bch2_trans_get_iter() it's used for the snapshot we're
operating in, which should be 0 for most btrees that don't use
snapshots.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
- We no longer mark subsets of extents, they're marked like regular
keys now - which means we can drop the offset & sectors arguments
to trigger functions
- Drop other arguments that are no longer needed anymore in various
places - fs_usage
- Drop the logic for handling extents in bch2_mark_update() that isn't
needed anymore, to match bch2_trans_mark_update()
- Better logic for hanlding the BTREE_ITER_CACHED_NOFILL case, where we
don't have an old key to mark
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Adding iter->should_be_locked introduced a regression where it ended up
not being set on the iterator passed to bch2_btree_update_start(), which
is definitely not what we want.
This patch requires it to be set when calling bch2_trans_update(), and
adds various fixups to make that happen.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This patch adds some new tracepoints to the btree iterator code, and
adds new fields to the existing tracepoints - primarily for the iterator
position.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Waiting on a btree node write with btree locks held can deadlock, if the
write errors: the write error path has to do do a btree update to drop
the pointer to the replica that errored.
The interior update path has to wait on in flight btree writes before
freeing nodes on disk. Previously, this was done in
bch2_btree_interior_update_will_free_node(), and could deadlock; now, we
just stash a pointer to the node and do it in
btree_update_nodes_written(), just prior to the transactional part of
the update.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Also, clean up workqueue usage - we shouldn't be using system
workqueues, pretty much everything we do needs to be on our own
WQ_MEM_RECLAIM workqueues.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This splits out btree topology repair into a separate pass, and makes
some improvements:
- When we have to pick which of two overlapping nodes to drop keys
from, we use the btree node header sequence number to preserve the
newer node
- the gc code has been changed so that it doesn't bail out if we're
continuing/ignoring on fsck error - this way the dump tool can skip
running the repair pass but still walk all reachable metadata
- add a new superblock flag indicating when a filesystem is known to
have btree topology issues, and the topology repair pass should be
run
- changing the start/end of a node might mean keys in that node have to
be deleted: this patch handles that better by splitting it out into a
separate function and running it explicitly in the topology repair
code, previously those keys were only being dropped when the btree
node was read in.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Consolidate common parts of bch2_btree_insert_keys_interior() and
btree_split_insert_keys() - prep work for adding some new topology
assertions.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This patch adds self healing functionality for btree nodes - if we
notice a problem when reading a btree node, we just rewrite it.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This replaces an assertion in the btree merge path with a
bch2_inconsistent_error() - fsck will fix it.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Turns out, we weren't waiting on in flight btree writes when freeing
existing btree nodes. This lead to stray btree writes overwriting newly
allocated buckets, but only started showing itself with some of the
recent allocator work and another patch to move submitting of btree
writes to worqueues.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We need to flush the btree key cache when it's too dirty, because
otherwise the shrinker won't be able to reclaim memory - this is done by
journal reclaim. But journal reclaim also kicks btree node writes: this
meant that btree node writes were getting kicked much too often just
because we needed to flush btree key cache keys.
This patch splits journal pins into two different lists, and teaches
journal reclaim to not flush btree node writes when it only needs to
flush key cache keys.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This is an important cleanup, eliminating an unnecessary copy in the
transaction commit path.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
JOURNAL_RES_GET_RESERVED should only be used for updatse that need to be
done to free up space in the journal. In particular, when we're flushing
keys from the key cache, if we're flushing them out of order we
shouldn't be using it, since we're using up our remaining space in the
journal without dropping a pin that will let us make forward progress.
With this patch, BTREE_INSERT_JOURNAL_RECLAIM without
BTREE_INSERT_JOURNAL_RESERVED may return -EAGAIN - we can't wait on
journal reclaim if we're already in journal reclaim.
This means we need to propagate these errors up to journal reclaim,
indicating that flushing a journal pin should be retried in the future.
This is prep work for a patch to change the way journal reclaim works,
to split out flushing key cache keys because the btree key cache is too
dirty from journal reclaim because we need space in the journal.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This means that btree node splits don't have to automatically trigger a
transaction restart.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This patch reworks the btree node merge path to use a second btree
iterator to get the sibling node - which means
bch2_btree_iter_get_sibling() can be deleted. Also, it uses
bch2_btree_iter_traverse_all() if necessary - which means it should be
more reliable. We don't currently even try to make it work when
trans->nounlock is set - after a BTREE_INSERT_NOUNLOCK transaction
commit, hopefully this will be a worthwhile tradeoff.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Previously, we were doing btree node merging from
bch2_btree_insert_node() - but this is called from the split path, when
we're in the middle of creating new nodes and deleting new nodes and the
iterators are in a weird state.
Also, this means we're starting a new btree_update while in the middle
of an existing one, and that's asking for deadlocks.
Much simpler and saner to trigger btree node merging _after_ the whole
btree node split path is finished.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
bch2_btree_update_start() is now responsible for taking gc_lock and
upgrading the iterator to lock parent nodes - greatly simplifying error
handling and all of the callers.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This patch starts treating the bpos.snapshot field like part of the key
in the btree code:
* bpos_successor() and bpos_predecessor() now include the snapshot field
* Keys in btrees that will be using snapshots (extents, inodes, dirents
and xattrs) now always have their snapshot field set to U32_MAX
The btree iterator code gets a new flag, BTREE_ITER_ALL_SNAPSHOTS, that
determines whether we're iterating over keys in all snapshots or not -
internally, this controlls whether bkey_(successor|predecessor)
increment/decrement the snapshot field, or only the higher bits of the
key.
We add a new member to struct btree_iter, iter->snapshot: when
BTREE_ITER_ALL_SNAPSHOTS is not set, iter->pos.snapshot should always
equal iter->snapshot, which will be 0 for btrees that don't use
snapshots, and alsways U32_MAX for btrees that will use snapshots
(until we enable snapshot creation).
This patch also introduces a new metadata version number, and compat
code for reading from/writing to older versions - this isn't a forced
upgrade (yet).
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
With snapshots, we're going to need to differentiate between comparisons
that should and shouldn't include the snapshot field. bpos_cmp is now
the comparison function that does include the snapshot field, used by
core btree code.
Upper level filesystem code generally does _not_ want to compare against
the snapshot field - that code wants keys to compare as equal even when
one of them is in an ancestor snapshot.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
On btree node split, we weren't ensuring the min_key of the new larger
node packs in the new format for this node. This triggers some painful
slowpaths in the bset.c aux search tree code - this patch fixes that by
calculating a new format for the new node with the new min_key.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Bkey noops were introduced to deal with trimming inline data extents in
place in the btree: if the u64s field of a bkey was 0, that u64 was a
noop and we'd start looking for the next bkey immediately after it.
But extent handling has been lifted above the btree - we no longer
modify existing extents in place in the btree, and the compatibilty code
for old style extent btree nodes is gone, so we can completely drop this
code.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
KEY_TYPE_discard used to be used for extent whiteouts, but when handling
over overlapping extents was lifted above the core btree code it became
unused. This patch updates various code to reflect that.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
bcachefs has been aggressively migrating filesystems and btree nodes to
the new format for quite some time - this shouldn't affect anyone
anymore, and lets us delete a _lot_ of code. Also, it frees up
KEY_TYPE_discard for a new whiteout key type for snapshots.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This adds a new data job type to scan for btree nodes in the old extent
format, and rewrite them.
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>
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>
A btree node merge operation deletes a key in the parent node; if when
inserting into the parent node we split the parent node, we can end up
with a whiteout in the parent node that we don't want.
The existing code drops them before doing the split, because they can
screw up picking the pivot, but we forgot about the unwritten writeouts
area - that needs to be cleared out too.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This reverts part of the change from "bcachefs: Don't use
BTREE_INSERT_USE_RESERVE so much" - it turns out we still should be
reserving open buckets for btree node allocations, because otherwise
data bucket allocations (especially with erasure coding enabled) can use
up all our open buckets and we won't be able to do the metadata update
that lets us release those open bucket references. Oops.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
With various newer key types - stripe keys, inline data extents - the
old approach of calculating the maximum size of the value is becoming
more and more error prone. Better to switch to bkey_on_stack, which can
dynamically allocate if necessary to handle any size bkey.
In particular we also want to get rid of BKEY_EXTENT_VAL_U64s_MAX.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Previously, we were using BTREE_INSERT_RESERVE in a lot of places where
it no longer makes sense.
- we now have more open_buckets than we used to, and the reserves work
better, so we shouldn't need to use BTREE_INSERT_RESERVE just because
we're holding open_buckets pinned anymore.
- We have the btree key cache for updates to the alloc btree, meaning
we no longer need the btree reserve to ensure the allocator can make
forward progress.
This means that we should only need a reserve for btree updates to
ensure that copygc can make forward progress.
Since it's now just for copygc, we can also fold RESERVE_BTREE into
RESERVE_MOVINGGC (the allocator's freelist reserve).
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This is needed to fix a bug where we're overflowing iterators within a
btree transaction, because we're updating the stripes btree (to update
block counts) and the stripes btree trigger is unnecessarily updating
the alloc btree - it doesn't need to update the alloc btree when the
pointers within a stripe aren't changing.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
If we have an error in the btree interior update path that prevents us
from journalling the update, we can't issue the corresponding btree node
write - we didn't get a journal sequence number that would cause it to
be ignored in recovery.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
For the new nodes an interior btree update makes reachable, updates to
those nodes may be journalled after the btree update starts but before
the transactional part - where we make those nodes reachable. Those
updates need to be kept in the journal until after the btree update
completes, hence we should always get a journal pin at the start of the
interior update.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The interior btree node update path has changed, this is no longer
needed.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
tracking down a bug where we see a btree node pointer in the wrong node
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
this will reduce transaction restarts, from observation of tracepoints.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This lets us improve journal reclaim, so that it now tries to make sure
no more than 3/4s of the btree node cache and btree key cache are dirty
- ensuring the shrinkers can free memory.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This only did anything in two places, and those can just be replaced
wiht bkey_cmp_left_packed()).
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
printbufs know how big the buffer is that was allocated, so we can get
rid of the random PAGE_SIZEs all over the place.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We were missing a 'goto retry' and continuing on with an error pointer.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Now that updates to interior nodes are journalled, we shouldn't be
checking topology of interior nodes until we've finished replaying
updates to that node.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This fixes a bug where recovery fails when one of the devices is read
only.
Also - consolidate the "must rewrite this node to insert it" behind a
new btree node flag.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Whenever we're doing an update that has pointers, that generally means
we need to do the update in order to release open bucket references - so
we should be using the btree open bucket reserve.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This introduces a new kind of btree iterator, cached iterators, which
point to keys cached in a hash table. The cache also acts as a write
cache - in the update path, we journal the update but defer updating the
btree until the cached entry is flushed by journal reclaim.
Cache coherency is for now up to the users to handle, which isn't ideal
but should be good enough for now.
These new iterators will be used for updating inodes and alloc info (the
alloc and stripes btrees).
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Btree node lock ordering is based on the logical key. However, 'struct
btree' may be reused for a different btree node under memory pressure.
This patch uses the new six lock callback to check if a btree node is no
longer the node we wanted to lock before blocking.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Kent Overstreet
Kent Overstreet
Daniel Hill