Now that bch2_btree_iter_peek_with_updates() has been removed in favor
of BTREE_ITER_WITH_UPDATES, we need to make sure it's not used where we
don't want it.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Commit c42bca92be "bio: don't copy bvec
for direct IO" changed bio_iov_iter_get_pages() to point bio->bi_iovec
at the incoming biovec, meaning if we already allocated one, it'll be
leaked.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This fixes some rare cases where the metadata checksum option specified
may map to the wrong actual checksum type.
Signed-off-by: Janpieter Sollie <janpieter.sollie@edpnet.be>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
If the last entry(ies) would be all zeros, there's no need to write them
out - the read path already handles that.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
- Fix null deref on mount when given a null device name.
- Move the dev_name checks to return EINVAL when it is invalid.
Signed-off-by: Dan Robertson <dan@dlrobertson.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
If filesystem on disk was used by a version with a larger BCH_DATA_NR
thas the currently running version, we don't want this to cause a buffer
overrun.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Previously, checksummed extents could only be merged when the checksum
covered only the currently live data.
xfstest generic/064 creates a test file, then uses finsert calls to
split the extent, then collapse calls to see if they get merged. But
without any reads to trigger the narrow_crcs path, each of the split
extents will still have a checksum for the entire original extent.
This patch improves the extent merge path so that if either of the
extents we're attempting to merge has a checksum that covers the entire
merged extent, we just use that checksum.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
We haven't had extent merging in quite some time. It used to be done by
the btree code when sorting btree nodes, but that was eliminated as part
of the work to separate extent handling from core btree code.
This patch re-implements extent merging in the transaction commit path.
We don't currently have the ability to merge reflink pointers, we need
to do some work on the triggers code to be able to do that without
ending up with incorrect refcounts.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This patch simplifies the key merging code by getting rid of partial
merges - it's simpler and saner if we just don't merge extents when
they'd overflow k->size.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Now that extent handling has been lifted to bch2_trans_update(), we
don't need to keep two different lists of updates.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This lifts handling of overlapping extents out of __bch2_trans_commit()
and moves it to where we first do the update - which means that
BTREE_ITER_WITH_UPDATES can now work correctly in extents mode.
Also, this patch reworks how extent triggers work: previously, on
partial extent overwrite we would pass this information to the trigger,
telling it what part of the extent was being overwritten. But, this
approach has had too many subtle corner cases - now, we only mark whole
extents, meaning on partial extent overwrite we unmark the old extent
and mark the new extent.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This codepath won't just be for extents in the future, it'll also be for
BTREE_ITER_FILTER_SNAPSHOTS mode.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This drops bch2_btree_iter_peek_with_updates() and replaces it with a
new flag, BTREE_ITER_WITH_UPDATES, and also reworks
bch2_btree_iter_peek_slot() to respect it too.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This adds the ability for btree iterators to own child iterators - to be
used by an upcoming rework of bch2_btree_iter_peek_slot(), so we can
scan forwards while maintaining our current position.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
As a rule we don't want to be holding btree locks while submitting IO -
this will improve overall filesystem latency.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This improves the handling of overlapping btree nodes; now, we handle
the case where one btree node completely overwrites another.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
In make_extent_indirect(), we were allocating too small of a buffer for
the new indirect extent.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
bcachefs-tools recently started putting a backup superblock at the end
of the device. This causes a problem if the bucket size doesn't divide
the device size - but we can fix it by just skipping marking that part.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
When we switched to using bch2_btree_bset_insert_key() for extents it
turned out it started leaving invalid keys around - of type deleted but
nonzero size - but this is fine (if ugly) because they're never written
out.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Do not compile the acl.o target if BCACHEFS_POSIX_ACL is not enabled.
Signed-off-by: Dan Robertson <dan@dlrobertson.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Add a field to struct btree_iter for tracking whether it should be
locked - this fixes spurious transaction restarts in
bch2_trans_relock().
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>
Upcoming refactoring is going to change bch2_trans_update() to start
returning transaction restarts.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
We were missing a kthread_should_stop() check in the loop in
bch2_invalidate_buckets(), very occasionally leading to us getting stuck
while shutting down.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
When devices have different bucket sizes, we may accumulate a journal
write that doesn't fit on some of our devices - previously, we'd
underflow when calculating space on that device and then everything
would get weird.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This fixes a "disk usage increased without a reservation" bug, when
reflinking compressed extents. Also, there's no good reason for reflink
to be fragmenting extents anyways.
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>
We can't use btree_update_wq becuase btree updates may be waiting on
btree writes to complete.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
If the transactior restarts on a different CPU, it could end up needing
to read in a different btree node, which makes another transaction
restart more likely...
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Journal write errors were racing with the submission path - potentially
causing writes to other replicas to not get submitted.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
__bch2_trans_mark_reflink_p wasn't always correctly returning the number
of sectors processed - the new logic is a bit more straightforward
overall too.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
We're seeing a bug where inode creates end up spinning in
bch2_inode_create - disabling sharding will simplify what we're testing.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
We already had op->end_io as an alternative mechanism to op->cl.parent
for delivering write completions; this switches all code paths to using
op->end_io.
Two reasons:
- op->end_io is more efficient, due to fewer atomic ops, this completes
the conversion that was originally only done for the direct IO path.
- We'll be restructing the write path to use a different mechanism for
punting to process context, refactoring to not use op->cl will make
that easier.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This deletes bch_write_op.index_update_fn: indirect function calls have
gotten considerably more expensive post spectre/meltdown, and we only
have two different index_update_fns - this patch adds a flag to specify
which one to use (normal vs. data move path).
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We seem to have a bug where the copygc thread ends up spinning and
making the system unusable - this will at least prevent it from locking
up the machine, and it's a good thing to have anyways.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
After unclean shutdown, btree writes may have completed on one device
and not others - and this inconsistency could lead us to writing new
bsets with a gap in our btree node in one of our replicas.
Fortunately, this is only an issue with bsets that are newer than the
most recent journal flush, and we already have a mechanism for detecting
and blacklisting those. We just need to make sure to start new btree
writes after the most recent _non_ blacklisted bset.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
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>
There's a new module parameter, verify_all_btree_replicas, that enables
reading from every btree replica when reading in btree nodes and
comparing them against each other. We've been seeing some strange btree
corruption - this will hopefully aid in tracking it down and catching it
more often.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Buffered writes may have to increase their disk reservation at btree
update time, due to compression and erasure coding being unpredictable:
O_DIRECT writes should be checking for -ENOSPC, but buffered writes have
already been accepted and should not.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
When we delete the dirent an inode points to, we need to zero out the
backpointer fields - this was missed in the RENAME_OVERWRITE case.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Currently, we handle multiple overlapping extents in the same
transaction commit by doing fixups in bch2_trans_update() - this patch
extents that to split updates when necessary. The next patch that
changes the reflink code to not fragment extents when making them
indirect will require this.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Writeback throttling is a kernel config option and not always enabled.
When it's not enabled we need a fallback, to avoid unbounded memory
pinning and work item backlogs.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The current implementation of bch_statfs does not scale the number of
available blocks provided in f_bavail by the reserve factor. This causes
an allocation of a file of this size to fail.
Signed-off-by: Dan Robertson <dan@dlrobertson.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This bug led to push_whiteout() generating whiteouts that failed
bch2_bkey_invalid() due to nonzero length fields - oops.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
It was being skipped when hole punching, leading to problems when
splitting compressed extents.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
It's needed when we split an existing compressed extent - we get a null
ptr deref without it.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
fs/bcachefs/bset.c edited prefetch macro to add clang support
fs/bcachefs/btree_iter.c bugfix: initialize iter->real_pos in bch2_btree_iter_init for later use
fs/bcachefs/io.c bugfix: eliminated undefined behavior (negative bitshift)
fs/bcachefs/buckets.c bugfix: invert sign to handle 64bit abs()
Signed-off-by: Brett Holman <bpholman5@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We weren't holding mark_lock correctly - it's needed for the new_fs
path.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
- Ensure the second key value in bch_hash_info is initialized to zero
if the info type is of type BCH_STR_HASH_SIPHASH.
- Initialize the possibly returned value in bch2_inode_create. Assuming
bch2_btree_iter_peek returns bkey_s_c_null, the uninitialized value
of ret could be returned to the user as an error pointer.
- Fix compiler warning in initialization of bkey_s_c_stripe
fs/bcachefs/buckets.c:1646:35: warning: suggest braces around initialization
of subobject [-Wmissing-braces]
struct bkey_s_c_stripe new_s = { NULL };
^~~~
Signed-off-by: Dan Robertson <dan@dlrobertson.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
bch2_check_fix_ptrs() is awkward, we need to find a way to improve it.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Fix a possible read out of bounds if bch2_ioctl_fs_usage is called when
replica_entries_bytes is set to a value that is smaller than the size
of bch_replicas_usage.
Signed-off-by: Dan Robertson <dan@dlrobertson.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Do not attempt to cleanup the returned value of bch2_device_lookup if
the returned value was an error pointer. We currently check to see if
the returned value is null and run the cleanup otherwise. As a result,
we attempt to run the cleanup on a error pointer.
Signed-off-by: Dan Robertson <dan@dlrobertson.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Ensure that the block device pointer in a superblock handle is not
null before dereferencing it in bch2_dev_to_fs. The block device pointer
may be null when mounting a new bcachefs filesystem given another mounted
bcachefs filesystem exists that has at least one device that is offline.
Signed-off-by: Dan Robertson <dan@dlrobertson.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
When parsing the mount options duplicate the given options. This is
required as the options are parsed twice and strsep is used in parsing.
The options will be modified into a possibly invalid options set for the
second round of parsing if the options are not duplicated before
parsing.
Signed-off-by: Dan Robertson <dan@dlrobertson.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Calling mount with an empty source string causes an out-of-bounds error
in split_devs. Check the length of the source string to avoid this.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
By not re-fetching the next update we were going into an infinite loop.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The trigger for reflink pointers wasn't always incrementing/decrementing
the refcounts correctly - this patch fixes that logic.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We really need debug mode assertions that ca->ref and ca->io_ref are
used correctly.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Fix a possible out of bounds write in __bch2_btree_node_write when
the data buffer padding is cleared up to the block size. The out of
bounds write is possible if the data buffers size is not a multiple
of the block size.
Signed-off-by: Dan Robertson <dan@dlrobertson.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
jset->last_seq is in the region that's encrypted - on journal write
completion, we were using it and getting garbage. This patch shadows it
to fix.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This was a major oversight - this means under memory pressure we can end
up reading in a btree node, then having it evicted before we get to use
it.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
There were some overflows in the time conversion functions - fix this by
converting tv_sec and tv_nsec separately. Also, set sb->time_min and
sb->time_max.
Fixes xfstest generic/258.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
If the journal reclaim thread makes it to the timeout without ever
initializing j->last_flushed, we could end up sleeping for a very long
time.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We're seeing a filesystem get stuck when all devices but one have no
more reclaimable buckets - because the copygc wait amount is curretly
filesystem wide.
This patch should fix that, possibly at the expensive of running too
much when only one or a few devices is full and the rebalance thread
needs to move data around.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We're seeing livelocks that appear to be due to
bch2_btree_key_cache_scan repeatedly scanning and blocking other tasks
from using the key cache lock - we probably shouldn't be reporting
objects that can't actually be freed yet.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Trying to debug an issue where after traverse_all() we shouldn't have to
traverse any iterators... yet we are
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We need to also set iter->uptodate to indicate it needs to be traversed.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
There was a bug that led to duplicate btree node pointers being inserted
at the wrong level. The new topology repair code can fix that, except
that the btree cache code gets confused when we read in a btree node
from the pointer that was at the wrong level. This patch evicts nodes
that we're deleting to, which nicely solves the problem.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
With snapshots, using a radix tree for the table of link counts won't
work anymore because we also need to distinguish between inodes with
different snapshot IDs. Instead, this patch builds up a sorted array of
inodes that have hardlinks that we can binary search on - taking
advantage of the fact that with inode backpointers, the check_nlinks()
pass _only_ needs to concern itself with inodes that have hardlinks now.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Fix a few memory safety issues, found by asan in userspace.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
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>
bch2_varint_decode() can read up to 7 bytes past the end of the buffer,
which means we need to allocate slightly larger key cache buffers.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Calling mmap() directly is much better than malloc() then mprotect(), we
end up with much less address space fragmentation.
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>
bch2_check_fix_ptrs() was being called after checking if the replicas
set was marked - but repair could change which replicas set needed to be
marked. Oops.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This is prep work for subvolumes - each subvolume will have its own
lost+found.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Apparently, we have a bug where in mark and sweep while accounting for a
key, a replicas entry isn't found. Change the code to print out the key
we couldn't mark and halt instead of a BUG_ON().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Flushing the btree key cache needs to use allocation reserves - journal
reclaim depends on flushing the btree key cache for making forward
progress, and the allocator and copygc depend on journal reclaim making
forward progress.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
bch2_btree_verify() verifies that the btree node on disk matches what we
have in memory. This patch changes it to verify every replica, and also
fixes it for interior btree nodes - there's a mem_ptr field which is
used as a scratch space and needs to be zeroed out for comparing with
what's on disk.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We don't want to be submitting IO with btree locks held, and btree
writes usually aren't latency sensitive.
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>
Using the normal transaction commit path to insert and journal updates
to interior nodes hadn't been done before this repair code was written,
not surprising that there was a bug.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This will help avoid transaction restarts.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This uses the kthread_wait_freezable() macro to simplify a lot of the
allocator thread code, along with cleaning up bch2_invalidate_bucket2().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We check for this prior to metadata being written, but we're seeing some
strange bugs lately, and this will help catch those closer to where they
occur.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We've started seeing bug reports of pointers to btree nodes being
detected in leaf nodes. This should catch that before it's happened, and
it's something we should've been checking anyways.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
It's not actually the case that iterators are always checked here -
__bch2_trans_commit() checks for that after running triggers.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Even with runtime gc (which currently isn't supported), runtime gc no
longer clears/recalculates the main set of bucket marks - it allocates
and calculates another set, updating the primary at the end.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The owned_by_allocator field is a purely in memory thing, even if/when
we bring back GC at runtime there's no need for it to be recalculating
this field. This is prep work for pulling it out of struct bucket, and
eventually getting rid of the bucket array.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Upcoming patch is going to disallow multiple btree_trans on the stack.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This fixes a regression from
52d86202fd bcachefs: Improve bch2_btree_iter_traverse_all()
We want to avoid mucking with other iterators in the btree transaction
in operations that are only supposed to be touching individual iterators
- that patch was a cleanup to move lock ordering handling to
bch2_btree_iter_traverse_all(). But it broke upgrading of cloned
iterators.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
When dirty key cache keys were separated from other journal pins, we
broke the loop conditional in __bch2_journal_reclaim() - it's supposed
to keep looping as long as there's work to do.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Writepoints are never deallocated so the rcu_read_lock() isn't really
needed, but we are doing lockless list traversal.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This fixes a livelock with btree node splits.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
By changing it to upgrade iterators to intent locks to avoid lock
restarts we can simplify __bch2_btree_node_lock() quite a bit - this
fixes a probable bug where it could potentially drop a lock on an
unrelated error but still succeed instead of causing a transaction
restart.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Can't run arbitrary code inside a wait_event() conditional, due to
task state being weird...
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
If we let bch2_trans_commit() do it, it'll traverse iterators in sorted
order which means we'll get fewer lock restarts.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Avoid cloning iterators if we don't have to.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We have foreground btree node merging now, and any future btree node
merging improvements are going to be based off of that code.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We don't want it to block, if it can't allocate it should just continue
instead of possibly deadlocking.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Now that we have inode backpointers, we can simplify checking directory
structure: instead of doing a DFS from the filesystem root and then
checking if we found everything, we can iterate over every inode and see
if we can go up until we get to the root.
This patch also has a number of fixes and simplifications for the inode
backpointer checks. Also, it turns out we don't actually need the
BCH_INODE_BACKPTR_UNTRUSTED flag.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
bch2_link_trans() uses the btree key cache for inode updates, and fsck
isn't supposed to - also, it's not really what we want for reattaching
unreachable inodes anyways.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The patch that changed bch2_trans_relock() to not look at iter->uptodate
also tried to add an optimization by only having it relock
btree_iter_key() iterators (iterators that are live or have been marked
as keep). But, this wasn't thought through - this pops internal iterator
assertions because on transaction restart, when we're traversing
iterators we traverse all iterators marked as linked, and having
bch2_trans_relock() skip some of those mean that it can skil the
iterator that bch2_btree_iter_traverse_one() is currently traversing.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Now that we have inode backpointers the check_nlink pass only is
concerned with files that have hardlinks, and can be simplified.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This lets us simplify fsck quite a bit, which we need for making fsck
snapshot aware.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Very early on there was a period where we were accidentally generating
dirents with trailing garbage; we've since dropped support for
filesystems that old and the fsck code can be dropped.
Also, this patch switches to a simpler algorithm for checking hash
tables. It's less efficient on hash collision - but with 64 bit keys,
those are very rare.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This splits out checking inode nlinks from the rest of the inode checks
and moves most of the inode checks to the start of fsck, so that other
fsck passes can depend on it.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
bch2_btree_iter_peek() wasn't properly checking for
BTREE_ITER_IS_EXTENTS when updating iter->pos.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Since we're using a NOT_EXTENTS iterator, we shouldn't be setting the
iter pos to the start of the extent.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
It used to be necessary for the allocator thread to batch up
invalidating buckets when possible - but since we added the btree key
cache that hasn't been a concern, and now it's causing the allocator
thread to livelock when the filesystem is nearly full.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We've had BCH_FEATURE_atomic_nlink for quite some time, we can drop this
now.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The previous patch that fixed btree nodes being written too aggressively
now meant that we weren't sorting btree node bsets optimally - this
patch fixes that.
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>
In userspace, we don't really have a well defined PAGE_SIZE and shouln't
be relying on it. This is some more incremental work to remove
references to it.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Linear searches have gotten cheaper relative to binary searches on
modern hardware, due to better branch prediction behaviour.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Fortunately it was just used in an error message
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Since we now ensure a btree node's max key fits in its packed format,
this isn't needed for the reasons it used to be - and, it was being used
inconsistently.
Also reorder struct btree a bit for performance, and kill some dead
code.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
gcc is emitting rep stos here, which is silly (and slow) for an 8 byte
memset.
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>
The bug was that we were trying to find a replicas entry that wasn't
sorted - but, we can also simplify the code by not using
bch2_mark_bkey_replicas and instead ensuring the list of replicas
entries exists directly.
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>
After we get a journal reservation, we need to use it - if we erorr out
of a transaction commit, we'll be eating into space in the journal and
if our transaction needs to make forward progress in order to reclaim
space in the journal, we'll deadlock.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Eventually BTREE_ITER_NODES should be going away. This patch is to fix a
transaction iterator overflow in the btree node merge path because
BTREE_ITER_NODES iterators couldn't be reused.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We were multiplying instead of dividing - oops.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Since we're no longer doing btree node merging post commit, we can now
delete a bunch of code.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Currently, BTREE_INSERT_NOUNLOCK makes it hard to ensure btree node
merging happens reliably - since btree node merging happens after
transaction commit, we can't drop btree locks and block when starting
the btree update.
This patch moves it to before transaction commit - and failure to do a
merge that we wanted to do just restarts the transaction.
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>
Useful number for performance tuning.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We're getting away from relying on iter->uptodate - this changes
bch2_trans_relock() to more directly specify which iterators should be
relocked.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This just drops the offending key - in the bug report where this was
seen, it was clearly a single bit memory error, and fsck will fix the
missing key.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This is specifically to speed up bch2_inode_rm(), so that we're not
traversing iterators we're done with.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This adds a new watermark for the journal reclaim when flushing btree
key cache entries - it should try and stay ahead of where foreground
threads doing transaction commits will enter direct journal reclaim.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Since move.c isn't aware of what subvolume we're in, we can't use the
standard inode lookup code - fortunately, we're just using it for
reading IO options.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
For snapshots, when we allocate a new inode we want to allocate an inode
number that isn't in use in any other subvolume. We won't be able to use
ITER_SLOTS for this, inode allocation needs to change to use
BTREE_ITER_ALL_SNAPSHOTS.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This patch adds two new inode fields, bi_dir and bi_dir_offset, that
point back to the inode's dirent.
Since we're only adding fields for a single backpointer, files that have
been hardlinked won't necessarily have valid backpointers: we also add a
new inode flag, BCH_INODE_BACKPTR_UNTRUSTED, that's set if an inode has
ever had multiple links to it. That's ok, because we only really need
this functionality for directories, which can never have multiple
hardlinks - when we add subvolumes, we'll need a way to enemurate and
print subvolumes, and this will let us reconstruct a path to a subvolume
root given a subvolume root inode.
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>
The btree key cache mutex was becoming a significant bottleneck - it was
mainly used to protect the lists of dirty, clean and freed cached keys.
This patch eliminates the dirty and clean lists - instead, when we need
to scan for keys to drop from the cache we iterate over the rhashtable,
and thus we're able to remove most uses of that lock.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Since we now make sure to always generate packed bkey formats that can
pack the min_key of a btree node, this path should actually never
happen.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
When we pass BTREE_INSERT_NOUNLOCK bch2_trans_commit isn't supposed to
unlock after a successful commit, but it was calling
bch2_trans_cond_resched() - oops.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We weren't packing the min/max keys, which was a major oversight and
completely disabled generating bkey_floats for adjacent nodes.
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>