korg/linux
0
0
Fork 0
mirror of https://mirrors.bfsu.edu.cn/git/linux.git synced 2024-11-24 12:44:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

bcachefs changes for 6.10-rc1

Browse Source 
- More safety fixes, primarily found by syzbot
 
 - Run the upgrade/downgrade paths in nochnages mode. Nochanges mode is
   primarily for testing fsck/recovery in dry run mode, so it shouldn't
   change anything besides disabling writes and holding dirty metadata in
   memory.
 
   The idea here was to reduce the amount of activity if we can't write
   anything out, so that bringing up a filesystem in "super ro" mode
   would be more lilkely to work for data recovery - but norecovery is
   the correct option for this.
 
 - btree_trans->locked; we now track whether a btree_trans has any btree
   nodes locked, and this is used for improved assertions related to
   trans_unlock() and trans_relock(). We'll also be using it for
   improving how we work with lockdep in the future: we don't want
   lockdep to be tracking individual btree node locks because we take too
   many for lockdep to track, and it's not necessary since we have a
   cycle detector.
 
 - Trigger improvements that are prep work for online fsck
 
 - BTREE_TRIGGER_check_repair; this regularizes how we do some repair
   work for extents that goes with running triggers in fsck, and fixes
   some subtle issues with transaction restarts there.
 
 - bch2_snapshot_equiv() has now been ripped out of fsck.c; snapshot
   equivalence classes are for when snapshot deletion leaves behind
   redundant snapshot nodes, but snapshot deletion now cleans this up
   right away, so the abstraction doesn't need to leak.
 
 - Improvements to how we resume writing to the journal in recovery. The
   code for picking the new place to write when reading the journal is
   greatly simplified and we also store the position in the superblock
   for when we don't read the journal; this means that we preserve more
   of the journal for list_journal debugging.
 
 - Improvements to sysfs btree_cache and btree_node_cache, for debugging
   memory reclaim.
 
 - We now detect when we've blocked for 10 seconds on the allocator in
   the write path and dump some useful info.
 
 - Safety fixes for devices references: this is a big series that changes
   almost all device lookups to properly check if the device exists and
   take a reference to it.
 
   Previously we assumed that if a bkey exists that references a device
   then the device must exist, and this was enforced in .invalid methods,
   but this was incorrect because it meant device removal relied on
   accounting being correct to not leave keys pointing to invalid
   devices, and that's not something we can assume.
 
   Getting the "pointer to invalid device" checks out of our .invalid()
   methods fixes some long standing device removal bugs; the only
   outstanding bug with device removal now is a race between the discard
   path and deleting alloc info, which should be easily fixed.
 
 - The allocator now prefers not to expand the new
   member_info.btree_allocated bitmap, meaning if repair ever requires
   scanning for btree nodes (because of a corrupt interior nodes) we
   won't have to scan the whole device(s).
 
 - New coding style document, which among other things talks about the
   correct usage of assertions
 -----BEGIN PGP SIGNATURE-----
 
 iQIzBAABCgAdFiEEKnAFLkS8Qha+jvQrE6szbY3KbnYFAmZKJQgACgkQE6szbY3K
 bnZETg//SU9H0OHnBSMB/cteF6PKo9QR+dhT+3n+gWTxl0o/egbGTqwbzVqGtd2f
 J6II1BsDk8VoTOb/gFfLRShlmJfnj2jpRThU265faR/7LQYeSaqndDPkjOpTayAD
 Nj/DJyiSUTL753rZh3yUhOpOIHf7iapH6wuaZCPfhdfk+yvZNW8iz07JHjHLKRp8
 I2cFH0r6kN916NdRkt9oDCz68WouT8eWTqwcKra04XsLEZjNJHxLpKMq4M8UdPc7
 YynJPVt+aP8+VduGIq6pV8Co3afCP2oUywo11JpRmvLsw4tex/59wxOYtpMfgn6k
 4H+9WqiBwkbmnLDrfFHWRameS6F/7+GRAOVuz9nkmfk61UPU15gLjSRffqZ6u2YC
 7vbrXgebId/sZXtBpQd83RMMX52BnEJah0upNJ54IsSqfDYkU9lwl6CEyYpcX1hf
 YNBGBTbspZztc3AB13b3ow421FMhaySUg0FDmntMR9O8Z6/BXk7Ykc7b8DPEfrFs
 W6JY7q+ARBxr+EgFcV74fvMCf7NJTAhyv80AKryo7NFU2JZOyyaTxcTGSnolX4Mi
 lyHiOgicmOX+vy3vbC1dZoDcmIDJ4Uc0vixYcpKiZqxlR8XJ+wpevC50TEhxrcW+
 ZO4SloQvgyjI34xu/gZgjRYb3BhXK3x+ougVFpRG8V8zQ/+ccWg=
 =MKrF
 -----END PGP SIGNATURE-----

Merge tag 'bcachefs-2024-05-19' of https://evilpiepirate.org/git/bcachefs

Pull bcachefs updates from Kent Overstreet:

 - More safety fixes, primarily found by syzbot

 - Run the upgrade/downgrade paths in nochnages mode. Nochanges mode is
   primarily for testing fsck/recovery in dry run mode, so it shouldn't
   change anything besides disabling writes and holding dirty metadata
   in memory.

   The idea here was to reduce the amount of activity if we can't write
   anything out, so that bringing up a filesystem in "super ro" mode
   would be more lilkely to work for data recovery - but norecovery is
   the correct option for this.

 - btree_trans->locked; we now track whether a btree_trans has any btree
   nodes locked, and this is used for improved assertions related to
   trans_unlock() and trans_relock(). We'll also be using it for
   improving how we work with lockdep in the future: we don't want
   lockdep to be tracking individual btree node locks because we take
   too many for lockdep to track, and it's not necessary since we have a
   cycle detector.

 - Trigger improvements that are prep work for online fsck

 - BTREE_TRIGGER_check_repair; this regularizes how we do some repair
   work for extents that goes with running triggers in fsck, and fixes
   some subtle issues with transaction restarts there.

 - bch2_snapshot_equiv() has now been ripped out of fsck.c; snapshot
   equivalence classes are for when snapshot deletion leaves behind
   redundant snapshot nodes, but snapshot deletion now cleans this up
   right away, so the abstraction doesn't need to leak.

 - Improvements to how we resume writing to the journal in recovery. The
   code for picking the new place to write when reading the journal is
   greatly simplified and we also store the position in the superblock
   for when we don't read the journal; this means that we preserve more
   of the journal for list_journal debugging.

 - Improvements to sysfs btree_cache and btree_node_cache, for debugging
   memory reclaim.

 - We now detect when we've blocked for 10 seconds on the allocator in
   the write path and dump some useful info.

 - Safety fixes for devices references: this is a big series that
   changes almost all device lookups to properly check if the device
   exists and take a reference to it.

   Previously we assumed that if a bkey exists that references a device
   then the device must exist, and this was enforced in .invalid
   methods, but this was incorrect because it meant device removal
   relied on accounting being correct to not leave keys pointing to
   invalid devices, and that's not something we can assume.

   Getting the "pointer to invalid device" checks out of our .invalid()
   methods fixes some long standing device removal bugs; the only
   outstanding bug with device removal now is a race between the discard
   path and deleting alloc info, which should be easily fixed.

 - The allocator now prefers not to expand the new
   member_info.btree_allocated bitmap, meaning if repair ever requires
   scanning for btree nodes (because of a corrupt interior nodes) we
   won't have to scan the whole device(s).

 - New coding style document, which among other things talks about the
   correct usage of assertions

* tag 'bcachefs-2024-05-19' of https://evilpiepirate.org/git/bcachefs: (155 commits)
  bcachefs: add no_invalid_checks flag
  bcachefs: add counters for failed shrinker reclaim
  bcachefs: Fix sb_field_downgrade validation
  bcachefs: Plumb bch_validate_flags to sb_field_ops.validate()
  bcachefs: s/bkey_invalid_flags/bch_validate_flags
  bcachefs: fsync() should not return -EROFS
  bcachefs: Invalid devices are now checked for by fsck, not .invalid methods
  bcachefs: kill bch2_dev_bkey_exists() in bch2_check_fix_ptrs()
  bcachefs: kill bch2_dev_bkey_exists() in bch2_read_endio()
  bcachefs: bch2_dev_get_ioref() checks for device not present
  bcachefs: bch2_dev_get_ioref2(); io_read.c
  bcachefs: bch2_dev_get_ioref2(); debug.c
  bcachefs: bch2_dev_get_ioref2(); journal_io.c
  bcachefs: bch2_dev_get_ioref2(); io_write.c
  bcachefs: bch2_dev_get_ioref2(); btree_io.c
  bcachefs: bch2_dev_get_ioref2(); backpointers.c
  bcachefs: bch2_dev_get_ioref2(); alloc_background.c
  bcachefs: for_each_bset() declares loop iter
  bcachefs: Move BCACHEFS_STATFS_MAGIC value to UAPI magic.h
  bcachefs: Improve sysfs internal/btree_cache
  ...
This commit is contained in:
Linus Torvalds 2024-05-19 13:45:48 -07:00
commit 16dbfae867
123 changed files with 4692 additions and 4384 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

186
Documentation/filesystems/bcachefs/CodingStyle.rst Normal file
View File

@ -0,0 +1,186 @@
.. SPDX-License-Identifier: GPL-2.0
bcachefs coding style
=====================
Good development is like gardening, and codebases are our gardens. Tend to them
every day; look for little things that are out of place or in need of tidying.
A little weeding here and there goes a long way; don't wait until things have
spiraled out of control.
Things don't always have to be perfect - nitpicking often does more harm than
good. But appreciate beauty when you see it - and let people know.
The code that you are afraid to touch is the code most in need of refactoring.
A little organizing here and there goes a long way.
Put real thought into how you organize things.
Good code is readable code, where the structure is simple and leaves nowhere
for bugs to hide.
Assertions are one of our most important tools for writing reliable code. If in
the course of writing a patchset you encounter a condition that shouldn't
happen (and will have unpredictable or undefined behaviour if it does), or
you're not sure if it can happen and not sure how to handle it yet - make it a
BUG_ON(). Don't leave undefined or unspecified behavior lurking in the codebase.
By the time you finish the patchset, you should understand better which
assertions need to be handled and turned into checks with error paths, and
which should be logically impossible. Leave the BUG_ON()s in for the ones which
are logically impossible. (Or, make them debug mode assertions if they're
expensive - but don't turn everything into a debug mode assertion, so that
we're not stuck debugging undefined behaviour should it turn out that you were
wrong).
Assertions are documentation that can't go out of date. Good assertions are
wonderful.
Good assertions drastically and dramatically reduce the amount of testing
required to shake out bugs.
Good assertions are based on state, not logic. To write good assertions, you
have to think about what the invariants on your state are.
Good invariants and assertions will hold everywhere in your codebase. This
means that you can run them in only a few places in the checked in version, but
should you need to debug something that caused the assertion to fail, you can
quickly shotgun them everywhere to find the codepath that broke the invariant.
A good assertion checks something that the compiler could check for us, and
elide - if we were working in a language with embedded correctness proofs that
the compiler could check. This is something that exists today, but it'll likely
still be a few decades before it comes to systems programming languages. But we
can still incorporate that kind of thinking into our code and document the
invariants with runtime checks - much like the way people working in
dynamically typed languages may add type annotations, gradually making their
code statically typed.
Looking for ways to make your assertions simpler - and higher level - will
often nudge you towards making the entire system simpler and more robust.
Good code is code where you can poke around and see what it's doing -
introspection. We can't debug anything if we can't see what's going on.
Whenever we're debugging, and the solution isn't immediately obvious, if the
issue is that we don't know where the issue is because we can't see what's
going on - fix that first.
We have the tools to make anything visible at runtime, efficiently - RCU and
percpu data structures among them. Don't let things stay hidden.
The most important tool for introspection is the humble pretty printer - in
bcachefs, this means `*_to_text()` functions, which output to printbufs.
Pretty printers are wonderful, because they compose and you can use them
everywhere. Having functions to print whatever object you're working with will
make your error messages much easier to write (therefore they will actually
exist) and much more informative. And they can be used from sysfs/debugfs, as
well as tracepoints.
Runtime info and debugging tools should come with clear descriptions and
labels, and good structure - we don't want files with a list of bare integers,
like in procfs. Part of the job of the debugging tools is to educate users and
new developers as to how the system works.
Error messages should, whenever possible, tell you everything you need to debug
the issue. It's worth putting effort into them.
Tracepoints shouldn't be the first thing you reach for. They're an important
tool, but always look for more immediate ways to make things visible. When we
have to rely on tracing, we have to know which tracepoints we're looking for,
and then we have to run the troublesome workload, and then we have to sift
through logs. This is a lot of steps to go through when a user is hitting
something, and if it's intermittent it may not even be possible.
The humble counter is an incredibly useful tool. They're cheap and simple to
use, and many complicated internal operations with lots of things that can
behave weirdly (anything involving memory reclaim, for example) become
shockingly easy to debug once you have counters on every distinct codepath.
Persistent counters are even better.
When debugging, try to get the most out of every bug you come across; don't
rush to fix the initial issue. Look for things that will make related bugs
easier the next time around - introspection, new assertions, better error
messages, new debug tools, and do those first. Look for ways to make the system
better behaved; often one bug will uncover several other bugs through
downstream effects.
Fix all that first, and then the original bug last - even if that means keeping
a user waiting. They'll thank you in the long run, and when they understand
what you're doing you'll be amazed at how patient they're happy to be. Users
like to help - otherwise they wouldn't be reporting the bug in the first place.
Talk to your users. Don't isolate yourself.
Users notice all sorts of interesting things, and by just talking to them and
interacting with them you can benefit from their experience.
Spend time doing support and helpdesk stuff. Don't just write code - code isn't
finished until it's being used trouble free.
This will also motivate you to make your debugging tools as good as possible,
and perhaps even your documentation, too. Like anything else in life, the more
time you spend at it the better you'll get, and you the developer are the
person most able to improve the tools to make debugging quick and easy.
Be wary of how you take on and commit to big projects. Don't let development
become product-manager focused. Often time an idea is a good one but needs to
wait for its proper time - but you won't know if it's the proper time for an
idea until you start writing code.
Expect to throw a lot of things away, or leave them half finished for later.
Nobody writes all perfect code that all gets shipped, and you'll be much more
productive in the long run if you notice this early and shift to something
else. The experience gained and lessons learned will be valuable for all the
other work you do.
But don't be afraid to tackle projects that require significant rework of
existing code. Sometimes these can be the best projects, because they can lead
us to make existing code more general, more flexible, more multipurpose and
perhaps more robust. Just don't hesitate to abandon the idea if it looks like
it's going to make a mess of things.
Complicated features can often be done as a series of refactorings, with the
final change that actually implements the feature as a quite small patch at the
end. It's wonderful when this happens, especially when those refactorings are
things that improve the codebase in their own right. When that happens there's
much less risk of wasted effort if the feature you were going for doesn't work
out.
Always strive to work incrementally. Always strive to turn the big projects
into little bite sized projects that can prove their own merits.
Instead of always tackling those big projects, look for little things that
will be useful, and make the big projects easier.
The question of what's likely to be useful is where junior developers most
often go astray - doing something because it seems like it'll be useful often
leads to overengineering. Knowing what's useful comes from many years of
experience, or talking with people who have that experience - or from simply
reading lots of code and looking for common patterns and issues. Don't be
afraid to throw things away and do something simpler.
Talk about your ideas with your fellow developers; often times the best things
come from relaxed conversations where people aren't afraid to say "what if?".
Don't neglect your tools.
The most important tools (besides the compiler and our text editor) are the
tools we use for testing. The shortest possible edit/test/debug cycle is
essential for working productively. We learn, gain experience, and discover the
errors in our thinking by running our code and seeing what happens. If your
time is being wasted because your tools are bad or too slow - don't accept it,
fix it.
Put effort into your documentation, commmit messages, and code comments - but
don't go overboard. A good commit message is wonderful - but if the information
was important enough to go in a commit message, ask yourself if it would be
even better as a code comment.
A good code comment is wonderful, but even better is the comment that didn't
need to exist because the code was so straightforward as to be obvious;
organized into small clean and tidy modules, with clear and descriptive names
for functions and variable, where every line of code has a clear purpose.

1
Documentation/filesystems/bcachefs/index.rst
View File

@ -8,4 +8,5 @@ bcachefs Documentation
:maxdepth: 2
:numbered:
CodingStyle
errorcodes

45
fs/bcachefs/acl.c
View File

@ -282,18 +282,12 @@ struct posix_acl *bch2_get_acl(struct mnt_idmap *idmap,
struct btree_trans *trans = bch2_trans_get(c);
struct btree_iter iter = { NULL };
struct posix_acl *acl = NULL;
struct bkey_s_c k;
int ret;
retry:
bch2_trans_begin(trans);
ret = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc,
&hash, inode_inum(inode), &search, 0);
if (ret)
goto err;
k = bch2_btree_iter_peek_slot(&iter);
ret = bkey_err(k);
struct bkey_s_c k = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc,
&hash, inode_inum(inode), &search, 0);
int ret = bkey_err(k);
if (ret)
goto err;
@ -366,7 +360,7 @@ retry:
ret = bch2_subvol_is_ro_trans(trans, inode->ei_subvol) ?:
bch2_inode_peek(trans, &inode_iter, &inode_u, inode_inum(inode),
BTREE_ITER_INTENT);
BTREE_ITER_intent);
if (ret)
goto btree_err;
@ -414,39 +408,30 @@ int bch2_acl_chmod(struct btree_trans *trans, subvol_inum inum,
struct bch_hash_info hash_info = bch2_hash_info_init(trans->c, inode);
struct xattr_search_key search = X_SEARCH(KEY_TYPE_XATTR_INDEX_POSIX_ACL_ACCESS, "", 0);
struct btree_iter iter;
struct bkey_s_c_xattr xattr;
struct bkey_i_xattr *new;
struct posix_acl *acl = NULL;
struct bkey_s_c k;
int ret;
ret = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc,
&hash_info, inum, &search, BTREE_ITER_INTENT);
struct bkey_s_c k = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc,
&hash_info, inum, &search, BTREE_ITER_intent);
int ret = bkey_err(k);
if (ret)
return bch2_err_matches(ret, ENOENT) ? 0 : ret;
k = bch2_btree_iter_peek_slot(&iter);
ret = bkey_err(k);
if (ret)
goto err;
xattr = bkey_s_c_to_xattr(k);
struct bkey_s_c_xattr xattr = bkey_s_c_to_xattr(k);
acl = bch2_acl_from_disk(trans, xattr_val(xattr.v),
le16_to_cpu(xattr.v->x_val_len));
ret = PTR_ERR_OR_ZERO(acl);
if (IS_ERR_OR_NULL(acl))
goto err;
ret = allocate_dropping_locks_errcode(trans,
__posix_acl_chmod(&acl, _gfp, mode));
if (ret)
goto err;
new = bch2_acl_to_xattr(trans, acl, ACL_TYPE_ACCESS);
if (IS_ERR(new)) {
ret = PTR_ERR(new);
ret = allocate_dropping_locks_errcode(trans, __posix_acl_chmod(&acl, _gfp, mode));
if (ret)
goto err;
struct bkey_i_xattr *new = bch2_acl_to_xattr(trans, acl, ACL_TYPE_ACCESS);
ret = PTR_ERR_OR_ZERO(new);
if (ret)
goto err;
}
new->k.p = iter.pos;
ret = bch2_trans_update(trans, &iter, &new->k_i, 0);

339
fs/bcachefs/alloc_background.c
View File

@ -195,7 +195,7 @@ static unsigned bch_alloc_v1_val_u64s(const struct bch_alloc *a)
}
int bch2_alloc_v1_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_s_c_alloc a = bkey_s_c_to_alloc(k);
@ -211,7 +211,7 @@ fsck_err:
}
int bch2_alloc_v2_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_alloc_unpacked u;
@ -225,7 +225,7 @@ fsck_err:
}
int bch2_alloc_v3_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_alloc_unpacked u;
@ -239,7 +239,7 @@ fsck_err:
}
int bch2_alloc_v4_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags, struct printbuf *err)
enum bch_validate_flags flags, struct printbuf *err)
{
struct bkey_s_c_alloc_v4 a = bkey_s_c_to_alloc_v4(k);
int ret = 0;
@ -263,7 +263,7 @@ int bch2_alloc_v4_invalid(struct bch_fs *c, struct bkey_s_c k,
case BCH_DATA_free:
case BCH_DATA_need_gc_gens:
case BCH_DATA_need_discard:
bkey_fsck_err_on(bch2_bucket_sectors(*a.v) || a.v->stripe,
bkey_fsck_err_on(bch2_bucket_sectors_total(*a.v) || a.v->stripe,
c, err, alloc_key_empty_but_have_data,
"empty data type free but have data");
break;
@ -330,27 +330,17 @@ void bch2_alloc_to_text(struct printbuf *out, struct bch_fs *c, struct bkey_s_c
prt_printf(out, "gen %u oldest_gen %u data_type ", a->gen, a->oldest_gen);
bch2_prt_data_type(out, a->data_type);
prt_newline(out);
prt_printf(out, "journal_seq %llu", a->journal_seq);
prt_newline(out);
prt_printf(out, "need_discard %llu", BCH_ALLOC_V4_NEED_DISCARD(a));
prt_newline(out);
prt_printf(out, "need_inc_gen %llu", BCH_ALLOC_V4_NEED_INC_GEN(a));
prt_newline(out);
prt_printf(out, "dirty_sectors %u", a->dirty_sectors);
prt_newline(out);
prt_printf(out, "cached_sectors %u", a->cached_sectors);
prt_newline(out);
prt_printf(out, "stripe %u", a->stripe);
prt_newline(out);
prt_printf(out, "stripe_redundancy %u", a->stripe_redundancy);
prt_newline(out);
prt_printf(out, "io_time[READ] %llu", a->io_time[READ]);
prt_newline(out);
prt_printf(out, "io_time[WRITE] %llu", a->io_time[WRITE]);
prt_newline(out);
prt_printf(out, "fragmentation %llu", a->fragmentation_lru);
prt_newline(out);
prt_printf(out, "bp_start %llu", BCH_ALLOC_V4_BACKPOINTERS_START(a));
prt_printf(out, "journal_seq %llu\n", a->journal_seq);
prt_printf(out, "need_discard %llu\n", BCH_ALLOC_V4_NEED_DISCARD(a));
prt_printf(out, "need_inc_gen %llu\n", BCH_ALLOC_V4_NEED_INC_GEN(a));
prt_printf(out, "dirty_sectors %u\n", a->dirty_sectors);
prt_printf(out, "cached_sectors %u\n", a->cached_sectors);
prt_printf(out, "stripe %u\n", a->stripe);
prt_printf(out, "stripe_redundancy %u\n", a->stripe_redundancy);
prt_printf(out, "io_time[READ] %llu\n", a->io_time[READ]);
prt_printf(out, "io_time[WRITE] %llu\n", a->io_time[WRITE]);
prt_printf(out, "fragmentation %llu\n", a->fragmentation_lru);
prt_printf(out, "bp_start %llu\n", BCH_ALLOC_V4_BACKPOINTERS_START(a));
printbuf_indent_sub(out, 2);
}
@ -439,22 +429,18 @@ struct bkey_i_alloc_v4 *bch2_alloc_to_v4_mut(struct btree_trans *trans, struct b
}
struct bkey_i_alloc_v4 *
bch2_trans_start_alloc_update(struct btree_trans *trans, struct btree_iter *iter,
struct bpos pos)
bch2_trans_start_alloc_update_noupdate(struct btree_trans *trans, struct btree_iter *iter,
struct bpos pos)
{
struct bkey_s_c k;
struct bkey_i_alloc_v4 *a;
int ret;
k = bch2_bkey_get_iter(trans, iter, BTREE_ID_alloc, pos,
BTREE_ITER_WITH_UPDATES|
BTREE_ITER_CACHED|
BTREE_ITER_INTENT);
ret = bkey_err(k);
struct bkey_s_c k = bch2_bkey_get_iter(trans, iter, BTREE_ID_alloc, pos,
BTREE_ITER_with_updates|
BTREE_ITER_cached|
BTREE_ITER_intent);
int ret = bkey_err(k);
if (unlikely(ret))
return ERR_PTR(ret);
a = bch2_alloc_to_v4_mut_inlined(trans, k);
struct bkey_i_alloc_v4 *a = bch2_alloc_to_v4_mut_inlined(trans, k);
ret = PTR_ERR_OR_ZERO(a);
if (unlikely(ret))
goto err;
@ -464,6 +450,20 @@ err:
return ERR_PTR(ret);
}
__flatten
struct bkey_i_alloc_v4 *bch2_trans_start_alloc_update(struct btree_trans *trans, struct bpos pos)
{
struct btree_iter iter;
struct bkey_i_alloc_v4 *a = bch2_trans_start_alloc_update_noupdate(trans, &iter, pos);
int ret = PTR_ERR_OR_ZERO(a);
if (ret)
return ERR_PTR(ret);
ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
bch2_trans_iter_exit(trans, &iter);
return unlikely(ret) ? ERR_PTR(ret) : a;
}
static struct bpos alloc_gens_pos(struct bpos pos, unsigned *offset)
{
*offset = pos.offset & KEY_TYPE_BUCKET_GENS_MASK;
@ -487,7 +487,7 @@ static unsigned alloc_gen(struct bkey_s_c k, unsigned offset)
}
int bch2_bucket_gens_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
int ret = 0;
@ -520,7 +520,7 @@ int bch2_bucket_gens_init(struct bch_fs *c)
int ret;
ret = for_each_btree_key(trans, iter, BTREE_ID_alloc, POS_MIN,
BTREE_ITER_PREFETCH, k, ({
BTREE_ITER_prefetch, k, ({
/*
* Not a fsck error because this is checked/repaired by
* bch2_check_alloc_key() which runs later:
@ -567,29 +567,31 @@ iter_err:
int bch2_alloc_read(struct bch_fs *c)
{
struct btree_trans *trans = bch2_trans_get(c);
struct bch_dev *ca = NULL;
int ret;
down_read(&c->gc_lock);
if (c->sb.version_upgrade_complete >= bcachefs_metadata_version_bucket_gens) {
ret = for_each_btree_key(trans, iter, BTREE_ID_bucket_gens, POS_MIN,
BTREE_ITER_PREFETCH, k, ({
BTREE_ITER_prefetch, k, ({
u64 start = bucket_gens_pos_to_alloc(k.k->p, 0).offset;
u64 end = bucket_gens_pos_to_alloc(bpos_nosnap_successor(k.k->p), 0).offset;
if (k.k->type != KEY_TYPE_bucket_gens)
continue;
const struct bch_bucket_gens *g = bkey_s_c_to_bucket_gens(k).v;
ca = bch2_dev_iterate(c, ca, k.k->p.inode);
/*
* Not a fsck error because this is checked/repaired by
* bch2_check_alloc_key() which runs later:
*/
if (!bch2_dev_exists2(c, k.k->p.inode))
if (!ca) {
bch2_btree_iter_set_pos(&iter, POS(k.k->p.inode + 1, 0));
continue;
}
struct bch_dev *ca = bch_dev_bkey_exists(c, k.k->p.inode);
const struct bch_bucket_gens *g = bkey_s_c_to_bucket_gens(k).v;
for (u64 b = max_t(u64, ca->mi.first_bucket, start);
b < min_t(u64, ca->mi.nbuckets, end);
@ -599,15 +601,16 @@ int bch2_alloc_read(struct bch_fs *c)
}));
} else {
ret = for_each_btree_key(trans, iter, BTREE_ID_alloc, POS_MIN,
BTREE_ITER_PREFETCH, k, ({
BTREE_ITER_prefetch, k, ({
ca = bch2_dev_iterate(c, ca, k.k->p.inode);
/*
* Not a fsck error because this is checked/repaired by
* bch2_check_alloc_key() which runs later:
*/
if (!bch2_dev_bucket_exists(c, k.k->p))
if (!ca) {
bch2_btree_iter_set_pos(&iter, POS(k.k->p.inode + 1, 0));
continue;
struct bch_dev *ca = bch_dev_bkey_exists(c, k.k->p.inode);
}
struct bch_alloc_v4 a;
*bucket_gen(ca, k.k->p.offset) = bch2_alloc_to_v4(k, &a)->gen;
@ -615,6 +618,7 @@ int bch2_alloc_read(struct bch_fs *c)
}));
}
bch2_dev_put(ca);
bch2_trans_put(trans);
up_read(&c->gc_lock);
@ -625,12 +629,12 @@ int bch2_alloc_read(struct bch_fs *c)
/* Free space/discard btree: */
static int bch2_bucket_do_index(struct btree_trans *trans,
struct bch_dev *ca,
struct bkey_s_c alloc_k,
const struct bch_alloc_v4 *a,
bool set)
{
struct bch_fs *c = trans->c;
struct bch_dev *ca = bch_dev_bkey_exists(c, alloc_k.k->p.inode);
struct btree_iter iter;
struct bkey_s_c old;
struct bkey_i *k;
@ -667,7 +671,7 @@ static int bch2_bucket_do_index(struct btree_trans *trans,
old = bch2_bkey_get_iter(trans, &iter, btree,
bkey_start_pos(&k->k),
BTREE_ITER_INTENT);
BTREE_ITER_intent);
ret = bkey_err(old);
if (ret)
return ret;
@ -711,8 +715,8 @@ static noinline int bch2_bucket_gen_update(struct btree_trans *trans,
return ret;
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_bucket_gens, pos,
BTREE_ITER_INTENT|
BTREE_ITER_WITH_UPDATES);
BTREE_ITER_intent|
BTREE_ITER_with_updates);
ret = bkey_err(k);
if (ret)
return ret;
@ -734,26 +738,24 @@ static noinline int bch2_bucket_gen_update(struct btree_trans *trans,
int bch2_trigger_alloc(struct btree_trans *trans,
enum btree_id btree, unsigned level,
struct bkey_s_c old, struct bkey_s new,
unsigned flags)
enum btree_iter_update_trigger_flags flags)
{
struct bch_fs *c = trans->c;
int ret = 0;
if (bch2_trans_inconsistent_on(!bch2_dev_bucket_exists(c, new.k->p), trans,
"alloc key for invalid device or bucket"))
struct bch_dev *ca = bch2_dev_bucket_tryget(c, new.k->p);
if (!ca)
return -EIO;
struct bch_dev *ca = bch_dev_bkey_exists(c, new.k->p.inode);
struct bch_alloc_v4 old_a_convert;
const struct bch_alloc_v4 *old_a = bch2_alloc_to_v4(old, &old_a_convert);
if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
if (flags & BTREE_TRIGGER_transactional) {
struct bch_alloc_v4 *new_a = bkey_s_to_alloc_v4(new).v;
new_a->data_type = alloc_data_type(*new_a, new_a->data_type);
alloc_data_type_set(new_a, new_a->data_type);
if (bch2_bucket_sectors(*new_a) > bch2_bucket_sectors(*old_a)) {
if (bch2_bucket_sectors_total(*new_a) > bch2_bucket_sectors_total(*old_a)) {
new_a->io_time[READ] = max_t(u64, 1, atomic64_read(&c->io_clock[READ].now));
new_a->io_time[WRITE]= max_t(u64, 1, atomic64_read(&c->io_clock[WRITE].now));
SET_BCH_ALLOC_V4_NEED_INC_GEN(new_a, true);
@ -770,10 +772,10 @@ int bch2_trigger_alloc(struct btree_trans *trans,
if (old_a->data_type != new_a->data_type ||
(new_a->data_type == BCH_DATA_free &&
alloc_freespace_genbits(*old_a) != alloc_freespace_genbits(*new_a))) {
ret = bch2_bucket_do_index(trans, old, old_a, false) ?:
bch2_bucket_do_index(trans, new.s_c, new_a, true);
ret = bch2_bucket_do_index(trans, ca, old, old_a, false) ?:
bch2_bucket_do_index(trans, ca, new.s_c, new_a, true);
if (ret)
return ret;
goto err;
}
if (new_a->data_type == BCH_DATA_cached &&
@ -787,24 +789,23 @@ int bch2_trigger_alloc(struct btree_trans *trans,
bucket_to_u64(new.k->p),
old_lru, new_lru);
if (ret)
return ret;
goto err;
}
new_a->fragmentation_lru = alloc_lru_idx_fragmentation(*new_a,
bch_dev_bkey_exists(c, new.k->p.inode));
new_a->fragmentation_lru = alloc_lru_idx_fragmentation(*new_a, ca);
if (old_a->fragmentation_lru != new_a->fragmentation_lru) {
ret = bch2_lru_change(trans,
BCH_LRU_FRAGMENTATION_START,
bucket_to_u64(new.k->p),
old_a->fragmentation_lru, new_a->fragmentation_lru);
if (ret)
return ret;
goto err;
}
if (old_a->gen != new_a->gen) {
ret = bch2_bucket_gen_update(trans, new.k->p, new_a->gen);
if (ret)
return ret;
goto err;
}
/*
@ -812,21 +813,21 @@ int bch2_trigger_alloc(struct btree_trans *trans,
* not:
*/
if ((flags & BTREE_TRIGGER_BUCKET_INVALIDATE) &&
if ((flags & BTREE_TRIGGER_bucket_invalidate) &&
old_a->cached_sectors) {
ret = bch2_update_cached_sectors_list(trans, new.k->p.inode,
-((s64) old_a->cached_sectors));
if (ret)
return ret;
goto err;
}
}
if ((flags & BTREE_TRIGGER_ATOMIC) && (flags & BTREE_TRIGGER_INSERT)) {
if ((flags & BTREE_TRIGGER_atomic) && (flags & BTREE_TRIGGER_insert)) {
struct bch_alloc_v4 *new_a = bkey_s_to_alloc_v4(new).v;
u64 journal_seq = trans->journal_res.seq;
u64 bucket_journal_seq = new_a->journal_seq;
if ((flags & BTREE_TRIGGER_INSERT) &&
if ((flags & BTREE_TRIGGER_insert) &&
data_type_is_empty(old_a->data_type) !=
data_type_is_empty(new_a->data_type) &&
new.k->type == KEY_TYPE_alloc_v4) {
@ -854,7 +855,7 @@ int bch2_trigger_alloc(struct btree_trans *trans,
if (ret) {
bch2_fs_fatal_error(c,
"setting bucket_needs_journal_commit: %s", bch2_err_str(ret));
return ret;
goto err;
}
}
@ -884,11 +885,11 @@ int bch2_trigger_alloc(struct btree_trans *trans,
bch2_do_invalidates(c);
if (statechange(a->data_type == BCH_DATA_need_gc_gens))
bch2_do_gc_gens(c);
bch2_gc_gens_async(c);
}
if ((flags & BTREE_TRIGGER_GC) &&
(flags & BTREE_TRIGGER_BUCKET_INVALIDATE)) {
if ((flags & BTREE_TRIGGER_gc) &&
(flags & BTREE_TRIGGER_bucket_invalidate)) {
struct bch_alloc_v4 new_a_convert;
const struct bch_alloc_v4 *new_a = bch2_alloc_to_v4(new.s_c, &new_a_convert);
@ -908,12 +909,13 @@ int bch2_trigger_alloc(struct btree_trans *trans,
bucket_unlock(g);
percpu_up_read(&c->mark_lock);
}
return 0;
err:
bch2_dev_put(ca);
return ret;
}
/*
* This synthesizes deleted extents for holes, similar to BTREE_ITER_SLOTS for
* This synthesizes deleted extents for holes, similar to BTREE_ITER_slots for
* extents style btrees, but works on non-extents btrees:
*/
static struct bkey_s_c bch2_get_key_or_hole(struct btree_iter *iter, struct bpos end, struct bkey *hole)
@ -958,35 +960,34 @@ static struct bkey_s_c bch2_get_key_or_hole(struct btree_iter *iter, struct bpos
}
}
static bool next_bucket(struct bch_fs *c, struct bpos *bucket)
static bool next_bucket(struct bch_fs *c, struct bch_dev **ca, struct bpos *bucket)
{
struct bch_dev *ca;
if (*ca) {
if (bucket->offset < (*ca)->mi.first_bucket)
bucket->offset = (*ca)->mi.first_bucket;
if (bch2_dev_bucket_exists(c, *bucket))
return true;
if (bch2_dev_exists2(c, bucket->inode)) {
ca = bch_dev_bkey_exists(c, bucket->inode);
if (bucket->offset < ca->mi.first_bucket) {
bucket->offset = ca->mi.first_bucket;
if (bucket->offset < (*ca)->mi.nbuckets)
return true;
}
bch2_dev_put(*ca);
*ca = NULL;
bucket->inode++;
bucket->offset = 0;
}
rcu_read_lock();
ca = __bch2_next_dev_idx(c, bucket->inode, NULL);
if (ca)
*bucket = POS(ca->dev_idx, ca->mi.first_bucket);
*ca = __bch2_next_dev_idx(c, bucket->inode, NULL);
if (*ca) {
*bucket = POS((*ca)->dev_idx, (*ca)->mi.first_bucket);
bch2_dev_get(*ca);
}
rcu_read_unlock();
return ca != NULL;
return *ca != NULL;
}
static struct bkey_s_c bch2_get_key_or_real_bucket_hole(struct btree_iter *iter, struct bkey *hole)
static struct bkey_s_c bch2_get_key_or_real_bucket_hole(struct btree_iter *iter,
struct bch_dev **ca, struct bkey *hole)
{
struct bch_fs *c = iter->trans->c;
struct bkey_s_c k;
@ -995,22 +996,21 @@ again:
if (bkey_err(k))
return k;
if (!k.k->type) {
struct bpos bucket = bkey_start_pos(k.k);
*ca = bch2_dev_iterate_noerror(c, *ca, k.k->p.inode);
if (!bch2_dev_bucket_exists(c, bucket)) {
if (!next_bucket(c, &bucket))
if (!k.k->type) {
struct bpos hole_start = bkey_start_pos(k.k);
if (!*ca || !bucket_valid(*ca, hole_start.offset)) {
if (!next_bucket(c, ca, &hole_start))
return bkey_s_c_null;
bch2_btree_iter_set_pos(iter, bucket);
bch2_btree_iter_set_pos(iter, hole_start);
goto again;
}
if (!bch2_dev_bucket_exists(c, k.k->p)) {
struct bch_dev *ca = bch_dev_bkey_exists(c, bucket.inode);
bch2_key_resize(hole, ca->mi.nbuckets - bucket.offset);
}
if (k.k->p.offset > (*ca)->mi.nbuckets)
bch2_key_resize(hole, (*ca)->mi.nbuckets - hole_start.offset);
}
return k;
@ -1025,24 +1025,25 @@ int bch2_check_alloc_key(struct btree_trans *trans,
struct btree_iter *bucket_gens_iter)
{
struct bch_fs *c = trans->c;
struct bch_dev *ca;
struct bch_alloc_v4 a_convert;
const struct bch_alloc_v4 *a;
unsigned discard_key_type, freespace_key_type;
unsigned gens_offset;
struct bkey_s_c k;
struct printbuf buf = PRINTBUF;
int ret;
int ret = 0;
if (fsck_err_on(!bch2_dev_bucket_exists(c, alloc_k.k->p), c,
alloc_key_to_missing_dev_bucket,
struct bch_dev *ca = bch2_dev_bucket_tryget_noerror(c, alloc_k.k->p);
if (fsck_err_on(!ca,
c, alloc_key_to_missing_dev_bucket,
"alloc key for invalid device:bucket %llu:%llu",
alloc_k.k->p.inode, alloc_k.k->p.offset))
return bch2_btree_delete_at(trans, alloc_iter, 0);
ret = bch2_btree_delete_at(trans, alloc_iter, 0);
if (!ca)
return ret;
ca = bch_dev_bkey_exists(c, alloc_k.k->p.inode);
if (!ca->mi.freespace_initialized)
return 0;
goto out;
a = bch2_alloc_to_v4(alloc_k, &a_convert);
@ -1141,25 +1142,26 @@ int bch2_check_alloc_key(struct btree_trans *trans,
if (ret)
goto err;
}
out:
err:
fsck_err:
bch2_dev_put(ca);
printbuf_exit(&buf);
return ret;
}
static noinline_for_stack
int bch2_check_alloc_hole_freespace(struct btree_trans *trans,
struct bch_dev *ca,
struct bpos start,
struct bpos *end,
struct btree_iter *freespace_iter)
{
struct bch_fs *c = trans->c;
struct bch_dev *ca;
struct bkey_s_c k;
struct printbuf buf = PRINTBUF;
int ret;
ca = bch_dev_bkey_exists(c, start.inode);
if (!ca->mi.freespace_initialized)
return 0;
@ -1313,7 +1315,7 @@ static noinline_for_stack int bch2_check_discard_freespace_key(struct btree_tran
goto delete;
out:
fsck_err:
set_btree_iter_dontneed(&alloc_iter);
bch2_set_btree_iter_dontneed(&alloc_iter);
bch2_trans_iter_exit(trans, &alloc_iter);
printbuf_exit(&buf);
return ret;
@ -1337,30 +1339,25 @@ int bch2_check_bucket_gens_key(struct btree_trans *trans,
{
struct bch_fs *c = trans->c;
struct bkey_i_bucket_gens g;
struct bch_dev *ca;
u64 start = bucket_gens_pos_to_alloc(k.k->p, 0).offset;
u64 end = bucket_gens_pos_to_alloc(bpos_nosnap_successor(k.k->p), 0).offset;
u64 b;
bool need_update = false, dev_exists;
bool need_update = false;
struct printbuf buf = PRINTBUF;
int ret = 0;
BUG_ON(k.k->type != KEY_TYPE_bucket_gens);
bkey_reassemble(&g.k_i, k);
/* if no bch_dev, skip out whether we repair or not */
dev_exists = bch2_dev_exists2(c, k.k->p.inode);
if (!dev_exists) {
if (fsck_err_on(!dev_exists, c,
bucket_gens_to_invalid_dev,
"bucket_gens key for invalid device:\n %s",
(bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
struct bch_dev *ca = bch2_dev_tryget_noerror(c, k.k->p.inode);
if (!ca) {
if (fsck_err(c, bucket_gens_to_invalid_dev,
"bucket_gens key for invalid device:\n %s",
(bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
ret = bch2_btree_delete_at(trans, iter, 0);
}
goto out;
}
ca = bch_dev_bkey_exists(c, k.k->p.inode);
if (fsck_err_on(end <= ca->mi.first_bucket ||
start >= ca->mi.nbuckets, c,
bucket_gens_to_invalid_buckets,
@ -1398,6 +1395,7 @@ int bch2_check_bucket_gens_key(struct btree_trans *trans,
}
out:
fsck_err:
bch2_dev_put(ca);
printbuf_exit(&buf);
return ret;
}
@ -1406,25 +1404,26 @@ int bch2_check_alloc_info(struct bch_fs *c)
{
struct btree_trans *trans = bch2_trans_get(c);
struct btree_iter iter, discard_iter, freespace_iter, bucket_gens_iter;
struct bch_dev *ca = NULL;
struct bkey hole;
struct bkey_s_c k;
int ret = 0;
bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc, POS_MIN,
BTREE_ITER_PREFETCH);
BTREE_ITER_prefetch);
bch2_trans_iter_init(trans, &discard_iter, BTREE_ID_need_discard, POS_MIN,
BTREE_ITER_PREFETCH);
BTREE_ITER_prefetch);
bch2_trans_iter_init(trans, &freespace_iter, BTREE_ID_freespace, POS_MIN,
BTREE_ITER_PREFETCH);
BTREE_ITER_prefetch);
bch2_trans_iter_init(trans, &bucket_gens_iter, BTREE_ID_bucket_gens, POS_MIN,
BTREE_ITER_PREFETCH);
BTREE_ITER_prefetch);
while (1) {
struct bpos next;
bch2_trans_begin(trans);
k = bch2_get_key_or_real_bucket_hole(&iter, &hole);
k = bch2_get_key_or_real_bucket_hole(&iter, &ca, &hole);
ret = bkey_err(k);
if (ret)
goto bkey_err;
@ -1445,7 +1444,7 @@ int bch2_check_alloc_info(struct bch_fs *c)
} else {
next = k.k->p;
ret = bch2_check_alloc_hole_freespace(trans,
ret = bch2_check_alloc_hole_freespace(trans, ca,
bkey_start_pos(k.k),
&next,
&freespace_iter) ?:
@ -1473,19 +1472,21 @@ bkey_err:
bch2_trans_iter_exit(trans, &freespace_iter);
bch2_trans_iter_exit(trans, &discard_iter);
bch2_trans_iter_exit(trans, &iter);
bch2_dev_put(ca);
ca = NULL;
if (ret < 0)
goto err;
ret = for_each_btree_key(trans, iter,
BTREE_ID_need_discard, POS_MIN,
BTREE_ITER_PREFETCH, k,
BTREE_ITER_prefetch, k,
bch2_check_discard_freespace_key(trans, &iter));
if (ret)
goto err;
bch2_trans_iter_init(trans, &iter, BTREE_ID_freespace, POS_MIN,
BTREE_ITER_PREFETCH);
BTREE_ITER_prefetch);
while (1) {
bch2_trans_begin(trans);
k = bch2_btree_iter_peek(&iter);
@ -1515,7 +1516,7 @@ bkey_err:
ret = for_each_btree_key_commit(trans, iter,
BTREE_ID_bucket_gens, POS_MIN,
BTREE_ITER_PREFETCH, k,
BTREE_ITER_prefetch, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
bch2_check_bucket_gens_key(trans, &iter, k));
err:
@ -1562,7 +1563,7 @@ static int bch2_check_alloc_to_lru_ref(struct btree_trans *trans,
a_mut->v.io_time[READ] = atomic64_read(&c->io_clock[READ].now);
ret = bch2_trans_update(trans, alloc_iter,
&a_mut->k_i, BTREE_TRIGGER_NORUN);
&a_mut->k_i, BTREE_TRIGGER_norun);
if (ret)
goto err;
@ -1601,7 +1602,7 @@ int bch2_check_alloc_to_lru_refs(struct bch_fs *c)
{
int ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter, BTREE_ID_alloc,
POS_MIN, BTREE_ITER_PREFETCH, k,
POS_MIN, BTREE_ITER_prefetch, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
bch2_check_alloc_to_lru_ref(trans, &iter)));
bch_err_fn(c, ret);
@ -1657,9 +1658,7 @@ static void discard_buckets_next_dev(struct bch_fs *c, struct discard_buckets_st
bch2_journal_flush_async(&c->journal, NULL);
if (s->ca)
percpu_ref_put(&s->ca->ref);
if (ca)
percpu_ref_get(&ca->ref);
percpu_ref_put(&s->ca->io_ref);
s->ca = ca;
s->need_journal_commit_this_dev = 0;
}
@ -1673,15 +1672,15 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
struct bpos pos = need_discard_iter->pos;
struct btree_iter iter = { NULL };
struct bkey_s_c k;
struct bch_dev *ca;
struct bkey_i_alloc_v4 *a;
struct printbuf buf = PRINTBUF;
bool discard_locked = false;
int ret = 0;
ca = bch_dev_bkey_exists(c, pos.inode);
if (!percpu_ref_tryget(&ca->io_ref)) {
struct bch_dev *ca = s->ca && s->ca->dev_idx == pos.inode
? s->ca
: bch2_dev_get_ioref(c, pos.inode, WRITE);
if (!ca) {
bch2_btree_iter_set_pos(need_discard_iter, POS(pos.inode + 1, 0));
return 0;
}
@ -1703,7 +1702,7 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_alloc,
need_discard_iter->pos,
BTREE_ITER_CACHED);
BTREE_ITER_cached);
ret = bkey_err(k);
if (ret)
goto out;
@ -1713,7 +1712,7 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
if (ret)
goto out;
if (a->v.dirty_sectors) {
if (bch2_bucket_sectors_total(a->v)) {
if (bch2_trans_inconsistent_on(c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info,
trans, "attempting to discard bucket with dirty data\n%s",
(bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
@ -1771,7 +1770,7 @@ static int bch2_discard_one_bucket(struct btree_trans *trans,
}
SET_BCH_ALLOC_V4_NEED_DISCARD(&a->v, false);
a->v.data_type = alloc_data_type(a->v, a->v.data_type);
alloc_data_type_set(&a->v, a->v.data_type);
write:
ret = bch2_trans_update(trans, &iter, &a->k_i, 0) ?:
bch2_trans_commit(trans, NULL, NULL,
@ -1787,7 +1786,6 @@ out:
discard_in_flight_remove(c, iter.pos);
s->seen++;
bch2_trans_iter_exit(trans, &iter);
percpu_ref_put(&ca->io_ref);
printbuf_exit(&buf);
return ret;
}
@ -1827,7 +1825,7 @@ void bch2_do_discards(struct bch_fs *c)
static int bch2_clear_bucket_needs_discard(struct btree_trans *trans, struct bpos bucket)
{
struct btree_iter iter;
bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc, bucket, BTREE_ITER_INTENT);
bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc, bucket, BTREE_ITER_intent);
struct bkey_s_c k = bch2_btree_iter_peek_slot(&iter);
int ret = bkey_err(k);
if (ret)
@ -1840,7 +1838,7 @@ static int bch2_clear_bucket_needs_discard(struct btree_trans *trans, struct bpo
BUG_ON(a->v.dirty_sectors);
SET_BCH_ALLOC_V4_NEED_DISCARD(&a->v, false);
a->v.data_type = alloc_data_type(a->v, a->v.data_type);
alloc_data_type_set(&a->v, a->v.data_type);
ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
err:
@ -1862,9 +1860,8 @@ static void bch2_do_discards_fast_work(struct work_struct *work)
if (i->snapshot)
continue;
ca = bch_dev_bkey_exists(c, i->inode);
if (!percpu_ref_tryget(&ca->io_ref)) {
ca = bch2_dev_get_ioref(c, i->inode, WRITE);
if (!ca) {
darray_remove_item(&c->discard_buckets_in_flight, i);
continue;
}
@ -1903,9 +1900,12 @@ static void bch2_do_discards_fast_work(struct work_struct *work)
static void bch2_discard_one_bucket_fast(struct bch_fs *c, struct bpos bucket)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, bucket.inode);
rcu_read_lock();
struct bch_dev *ca = bch2_dev_rcu(c, bucket.inode);
bool dead = !ca || percpu_ref_is_dying(&ca->io_ref);
rcu_read_unlock();
if (!percpu_ref_is_dying(&ca->io_ref) &&
if (!dead &&
!discard_in_flight_add(c, bucket) &&
bch2_write_ref_tryget(c, BCH_WRITE_REF_discard_fast) &&
!queue_work(c->write_ref_wq, &c->discard_fast_work))
@ -1918,7 +1918,6 @@ static int invalidate_one_bucket(struct btree_trans *trans,
s64 *nr_to_invalidate)
{
struct bch_fs *c = trans->c;
struct btree_iter alloc_iter = { NULL };
struct bkey_i_alloc_v4 *a = NULL;
struct printbuf buf = PRINTBUF;
struct bpos bucket = u64_to_bucket(lru_k.k->p.offset);
@ -1936,7 +1935,7 @@ static int invalidate_one_bucket(struct btree_trans *trans,
if (bch2_bucket_is_open_safe(c, bucket.inode, bucket.offset))
return 0;
a = bch2_trans_start_alloc_update(trans, &alloc_iter, bucket);
a = bch2_trans_start_alloc_update(trans, bucket);
ret = PTR_ERR_OR_ZERO(a);
if (ret)
goto out;
@ -1961,18 +1960,15 @@ static int invalidate_one_bucket(struct btree_trans *trans,
a->v.io_time[READ] = atomic64_read(&c->io_clock[READ].now);
a->v.io_time[WRITE] = atomic64_read(&c->io_clock[WRITE].now);
ret = bch2_trans_update(trans, &alloc_iter, &a->k_i,
BTREE_TRIGGER_BUCKET_INVALIDATE) ?:
bch2_trans_commit(trans, NULL, NULL,
BCH_WATERMARK_btree|
BCH_TRANS_COMMIT_no_enospc);
ret = bch2_trans_commit(trans, NULL, NULL,
BCH_WATERMARK_btree|
BCH_TRANS_COMMIT_no_enospc);
if (ret)
goto out;
trace_and_count(c, bucket_invalidate, c, bucket.inode, bucket.offset, cached_sectors);
--*nr_to_invalidate;
out:
bch2_trans_iter_exit(trans, &alloc_iter);
printbuf_exit(&buf);
return ret;
err:
@ -2014,11 +2010,11 @@ static void bch2_do_invalidates_work(struct work_struct *work)
ret = for_each_btree_key_upto(trans, iter, BTREE_ID_lru,
lru_pos(ca->dev_idx, 0, 0),
lru_pos(ca->dev_idx, U64_MAX, LRU_TIME_MAX),
BTREE_ITER_INTENT, k,
BTREE_ITER_intent, k,
invalidate_one_bucket(trans, &iter, k, &nr_to_invalidate));
if (ret < 0) {
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
break;
}
}
@ -2051,7 +2047,7 @@ int bch2_dev_freespace_init(struct bch_fs *c, struct bch_dev *ca,
bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc,
POS(ca->dev_idx, max_t(u64, ca->mi.first_bucket, bucket_start)),
BTREE_ITER_PREFETCH);
BTREE_ITER_prefetch);
/*
* Scan the alloc btree for every bucket on @ca, and add buckets to the
* freespace/need_discard/need_gc_gens btrees as needed:
@ -2083,7 +2079,7 @@ int bch2_dev_freespace_init(struct bch_fs *c, struct bch_dev *ca,
struct bch_alloc_v4 a_convert;
const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &a_convert);
ret = bch2_bucket_do_index(trans, k, a, true) ?:
ret = bch2_bucket_do_index(trans, ca, k, a, true) ?:
bch2_trans_commit(trans, NULL, NULL,
BCH_TRANS_COMMIT_no_enospc);
if (ret)
@ -2155,7 +2151,7 @@ int bch2_fs_freespace_init(struct bch_fs *c)
ret = bch2_dev_freespace_init(c, ca, 0, ca->mi.nbuckets);
if (ret) {
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
bch_err_fn(c, ret);
return ret;
}
@ -2182,7 +2178,10 @@ int bch2_bucket_io_time_reset(struct btree_trans *trans, unsigned dev,
u64 now;
int ret = 0;
a = bch2_trans_start_alloc_update(trans, &iter, POS(dev, bucket_nr));
if (bch2_trans_relock(trans))
bch2_trans_begin(trans);
a = bch2_trans_start_alloc_update_noupdate(trans, &iter, POS(dev, bucket_nr));
ret = PTR_ERR_OR_ZERO(a);
if (ret)
return ret;

109
fs/bcachefs/alloc_background.h
View File

@ -8,21 +8,18 @@
#include "debug.h"
#include "super.h"
enum bkey_invalid_flags;
enum bch_validate_flags;
/* How out of date a pointer gen is allowed to be: */
#define BUCKET_GC_GEN_MAX 96U
static inline bool bch2_dev_bucket_exists(struct bch_fs *c, struct bpos pos)
{
struct bch_dev *ca;
if (!bch2_dev_exists2(c, pos.inode))
return false;
ca = bch_dev_bkey_exists(c, pos.inode);
return pos.offset >= ca->mi.first_bucket &&
pos.offset < ca->mi.nbuckets;
rcu_read_lock();
struct bch_dev *ca = bch2_dev_rcu(c, pos.inode);
bool ret = ca && bucket_valid(ca, pos.offset);
rcu_read_unlock();
return ret;
}
static inline u64 bucket_to_u64(struct bpos bucket)
@ -40,38 +37,50 @@ static inline u8 alloc_gc_gen(struct bch_alloc_v4 a)
return a.gen - a.oldest_gen;
}
static inline enum bch_data_type __alloc_data_type(u32 dirty_sectors,
u32 cached_sectors,
u32 stripe,
struct bch_alloc_v4 a,
enum bch_data_type data_type)
static inline void alloc_to_bucket(struct bucket *dst, struct bch_alloc_v4 src)
{
if (stripe)
return data_type == BCH_DATA_parity ? data_type : BCH_DATA_stripe;
if (dirty_sectors)
return data_type;
if (cached_sectors)
return BCH_DATA_cached;
if (BCH_ALLOC_V4_NEED_DISCARD(&a))
return BCH_DATA_need_discard;
if (alloc_gc_gen(a) >= BUCKET_GC_GEN_MAX)
return BCH_DATA_need_gc_gens;
return BCH_DATA_free;
dst->gen = src.gen;
dst->data_type = src.data_type;
dst->dirty_sectors = src.dirty_sectors;
dst->cached_sectors = src.cached_sectors;
dst->stripe = src.stripe;
}
static inline enum bch_data_type alloc_data_type(struct bch_alloc_v4 a,
enum bch_data_type data_type)
static inline void __bucket_m_to_alloc(struct bch_alloc_v4 *dst, struct bucket src)
{
return __alloc_data_type(a.dirty_sectors, a.cached_sectors,
a.stripe, a, data_type);
dst->gen = src.gen;
dst->data_type = src.data_type;
dst->dirty_sectors = src.dirty_sectors;
dst->cached_sectors = src.cached_sectors;
dst->stripe = src.stripe;
}
static inline struct bch_alloc_v4 bucket_m_to_alloc(struct bucket b)
{
struct bch_alloc_v4 ret = {};
__bucket_m_to_alloc(&ret, b);
return ret;
}
static inline enum bch_data_type bucket_data_type(enum bch_data_type data_type)
{
return data_type == BCH_DATA_stripe ? BCH_DATA_user : data_type;
switch (data_type) {
case BCH_DATA_cached:
case BCH_DATA_stripe:
return BCH_DATA_user;
default:
return data_type;
}
}
static inline unsigned bch2_bucket_sectors(struct bch_alloc_v4 a)
static inline bool bucket_data_type_mismatch(enum bch_data_type bucket,
enum bch_data_type ptr)
{
return !data_type_is_empty(bucket) &&
bucket_data_type(bucket) != bucket_data_type(ptr);
}
static inline unsigned bch2_bucket_sectors_total(struct bch_alloc_v4 a)
{
return a.dirty_sectors + a.cached_sectors;
}
@ -89,6 +98,27 @@ static inline unsigned bch2_bucket_sectors_fragmented(struct bch_dev *ca,
return d ? max(0, ca->mi.bucket_size - d) : 0;
}
static inline enum bch_data_type alloc_data_type(struct bch_alloc_v4 a,
enum bch_data_type data_type)
{
if (a.stripe)
return data_type == BCH_DATA_parity ? data_type : BCH_DATA_stripe;
if (a.dirty_sectors)
return data_type;
if (a.cached_sectors)
return BCH_DATA_cached;
if (BCH_ALLOC_V4_NEED_DISCARD(&a))
return BCH_DATA_need_discard;
if (alloc_gc_gen(a) >= BUCKET_GC_GEN_MAX)
return BCH_DATA_need_gc_gens;
return BCH_DATA_free;
}
static inline void alloc_data_type_set(struct bch_alloc_v4 *a, enum bch_data_type data_type)
{
a->data_type = alloc_data_type(*a, data_type);
}
static inline u64 alloc_lru_idx_read(struct bch_alloc_v4 a)
{
return a.data_type == BCH_DATA_cached ? a.io_time[READ] : 0;
@ -147,7 +177,9 @@ static inline void set_alloc_v4_u64s(struct bkey_i_alloc_v4 *a)
}
struct bkey_i_alloc_v4 *
bch2_trans_start_alloc_update(struct btree_trans *, struct btree_iter *, struct bpos);
bch2_trans_start_alloc_update_noupdate(struct btree_trans *, struct btree_iter *, struct bpos);
struct bkey_i_alloc_v4 *
bch2_trans_start_alloc_update(struct btree_trans *, struct bpos);
void __bch2_alloc_to_v4(struct bkey_s_c, struct bch_alloc_v4 *);
@ -173,13 +205,13 @@ struct bkey_i_alloc_v4 *bch2_alloc_to_v4_mut(struct btree_trans *, struct bkey_s
int bch2_bucket_io_time_reset(struct btree_trans *, unsigned, size_t, int);
int bch2_alloc_v1_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
int bch2_alloc_v2_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
int bch2_alloc_v3_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
int bch2_alloc_v4_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_alloc_v4_swab(struct bkey_s);
void bch2_alloc_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
@ -213,7 +245,7 @@ void bch2_alloc_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
})
int bch2_bucket_gens_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_bucket_gens_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
#define bch2_bkey_ops_bucket_gens ((struct bkey_ops) { \
@ -233,7 +265,8 @@ static inline bool bkey_is_alloc(const struct bkey *k)
int bch2_alloc_read(struct bch_fs *);
int bch2_trigger_alloc(struct btree_trans *, enum btree_id, unsigned,
struct bkey_s_c, struct bkey_s, unsigned);
struct bkey_s_c, struct bkey_s,
enum btree_iter_update_trigger_flags);
int bch2_check_alloc_info(struct bch_fs *);
int bch2_check_alloc_to_lru_refs(struct bch_fs *);
void bch2_do_discards(struct bch_fs *);

304
fs/bcachefs/alloc_foreground.c
View File

@ -71,7 +71,7 @@ void bch2_reset_alloc_cursors(struct bch_fs *c)
{
rcu_read_lock();
for_each_member_device_rcu(c, ca, NULL)
ca->alloc_cursor = 0;
memset(ca->alloc_cursor, 0, sizeof(ca->alloc_cursor));
rcu_read_unlock();
}
@ -100,7 +100,7 @@ static void bch2_open_bucket_hash_remove(struct bch_fs *c, struct open_bucket *o
void __bch2_open_bucket_put(struct bch_fs *c, struct open_bucket *ob)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
struct bch_dev *ca = ob_dev(c, ob);
if (ob->ec) {
ec_stripe_new_put(c, ob->ec, STRIPE_REF_io);
@ -300,7 +300,7 @@ static struct open_bucket *try_alloc_bucket(struct btree_trans *trans, struct bc
k = bch2_bkey_get_iter(trans, &iter,
BTREE_ID_alloc, POS(ca->dev_idx, b),
BTREE_ITER_CACHED);
BTREE_ITER_cached);
ret = bkey_err(k);
if (ret) {
ob = ERR_PTR(ret);
@ -342,9 +342,9 @@ static struct open_bucket *try_alloc_bucket(struct btree_trans *trans, struct bc
struct bch_backpointer bp;
struct bpos bp_pos = POS_MIN;
ret = bch2_get_next_backpointer(trans, POS(ca->dev_idx, b), -1,
ret = bch2_get_next_backpointer(trans, ca, POS(ca->dev_idx, b), -1,
&bp_pos, &bp,
BTREE_ITER_NOPRESERVE);
BTREE_ITER_nopreserve);
if (ret) {
ob = ERR_PTR(ret);
goto err;
@ -363,10 +363,10 @@ static struct open_bucket *try_alloc_bucket(struct btree_trans *trans, struct bc
ob = __try_alloc_bucket(c, ca, b, watermark, a, s, cl);
if (!ob)
set_btree_iter_dontneed(&iter);
bch2_set_btree_iter_dontneed(&iter);
err:
if (iter.path)
set_btree_iter_dontneed(&iter);
bch2_set_btree_iter_dontneed(&iter);
bch2_trans_iter_exit(trans, &iter);
printbuf_exit(&buf);
return ob;
@ -389,7 +389,8 @@ bch2_bucket_alloc_early(struct btree_trans *trans,
struct bkey_s_c k, ck;
struct open_bucket *ob = NULL;
u64 first_bucket = max_t(u64, ca->mi.first_bucket, ca->new_fs_bucket_idx);
u64 alloc_start = max(first_bucket, READ_ONCE(ca->alloc_cursor));
u64 *dev_alloc_cursor = &ca->alloc_cursor[s->btree_bitmap];
u64 alloc_start = max(first_bucket, *dev_alloc_cursor);
u64 alloc_cursor = alloc_start;
int ret;
@ -404,9 +405,8 @@ bch2_bucket_alloc_early(struct btree_trans *trans,
*/
again:
for_each_btree_key_norestart(trans, iter, BTREE_ID_alloc, POS(ca->dev_idx, alloc_cursor),
BTREE_ITER_SLOTS, k, ret) {
struct bch_alloc_v4 a_convert;
const struct bch_alloc_v4 *a;
BTREE_ITER_slots, k, ret) {
u64 bucket = k.k->p.offset;
if (bkey_ge(k.k->p, POS(ca->dev_idx, ca->mi.nbuckets)))
break;
@ -415,12 +415,29 @@ again:
is_superblock_bucket(ca, k.k->p.offset))
continue;
a = bch2_alloc_to_v4(k, &a_convert);
if (s->btree_bitmap != BTREE_BITMAP_ANY &&
s->btree_bitmap != bch2_dev_btree_bitmap_marked_sectors(ca,
bucket_to_sector(ca, bucket), ca->mi.bucket_size)) {
if (s->btree_bitmap == BTREE_BITMAP_YES &&
bucket_to_sector(ca, bucket) > 64ULL << ca->mi.btree_bitmap_shift)
break;
bucket = sector_to_bucket(ca,
round_up(bucket_to_sector(ca, bucket) + 1,
1ULL << ca->mi.btree_bitmap_shift));
bch2_btree_iter_set_pos(&iter, POS(ca->dev_idx, bucket));
s->buckets_seen++;
s->skipped_mi_btree_bitmap++;
continue;
}
struct bch_alloc_v4 a_convert;
const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &a_convert);
if (a->data_type != BCH_DATA_free)
continue;
/* now check the cached key to serialize concurrent allocs of the bucket */
ck = bch2_bkey_get_iter(trans, &citer, BTREE_ID_alloc, k.k->p, BTREE_ITER_CACHED);
ck = bch2_bkey_get_iter(trans, &citer, BTREE_ID_alloc, k.k->p, BTREE_ITER_cached);
ret = bkey_err(ck);
if (ret)
break;
@ -433,7 +450,7 @@ again:
ob = __try_alloc_bucket(trans->c, ca, k.k->p.offset, watermark, a, s, cl);
next:
set_btree_iter_dontneed(&citer);
bch2_set_btree_iter_dontneed(&citer);
bch2_trans_iter_exit(trans, &citer);
if (ob)
break;
@ -441,7 +458,6 @@ next:
bch2_trans_iter_exit(trans, &iter);
alloc_cursor = iter.pos.offset;
ca->alloc_cursor = alloc_cursor;
if (!ob && ret)
ob = ERR_PTR(ret);
@ -451,6 +467,8 @@ next:
goto again;
}
*dev_alloc_cursor = alloc_cursor;
return ob;
}
@ -463,7 +481,8 @@ static struct open_bucket *bch2_bucket_alloc_freelist(struct btree_trans *trans,
struct btree_iter iter;
struct bkey_s_c k;
struct open_bucket *ob = NULL;
u64 alloc_start = max_t(u64, ca->mi.first_bucket, READ_ONCE(ca->alloc_cursor));
u64 *dev_alloc_cursor = &ca->alloc_cursor[s->btree_bitmap];
u64 alloc_start = max_t(u64, ca->mi.first_bucket, READ_ONCE(*dev_alloc_cursor));
u64 alloc_cursor = alloc_start;
int ret;
@ -485,10 +504,30 @@ again:
s->buckets_seen++;
u64 bucket = alloc_cursor & ~(~0ULL << 56);
if (s->btree_bitmap != BTREE_BITMAP_ANY &&
s->btree_bitmap != bch2_dev_btree_bitmap_marked_sectors(ca,
bucket_to_sector(ca, bucket), ca->mi.bucket_size)) {
if (s->btree_bitmap == BTREE_BITMAP_YES &&
bucket_to_sector(ca, bucket) > 64ULL << ca->mi.btree_bitmap_shift)
goto fail;
bucket = sector_to_bucket(ca,
round_up(bucket_to_sector(ca, bucket) + 1,
1ULL << ca->mi.btree_bitmap_shift));
u64 genbits = alloc_cursor >> 56;
alloc_cursor = bucket | (genbits << 56);
if (alloc_cursor > k.k->p.offset)
bch2_btree_iter_set_pos(&iter, POS(ca->dev_idx, alloc_cursor));
s->skipped_mi_btree_bitmap++;
continue;
}
ob = try_alloc_bucket(trans, ca, watermark,
alloc_cursor, s, k, cl);
if (ob) {
set_btree_iter_dontneed(&iter);
bch2_set_btree_iter_dontneed(&iter);
break;
}
}
@ -496,10 +535,9 @@ again:
if (ob || ret)
break;
}
fail:
bch2_trans_iter_exit(trans, &iter);
ca->alloc_cursor = alloc_cursor;
if (!ob && ret)
ob = ERR_PTR(ret);
@ -508,14 +546,56 @@ again:
goto again;
}
*dev_alloc_cursor = alloc_cursor;
return ob;
}
static noinline void trace_bucket_alloc2(struct bch_fs *c, struct bch_dev *ca,
enum bch_watermark watermark,
enum bch_data_type data_type,
struct closure *cl,
struct bch_dev_usage *usage,
struct bucket_alloc_state *s,
struct open_bucket *ob)
{
struct printbuf buf = PRINTBUF;
printbuf_tabstop_push(&buf, 24);
prt_printf(&buf, "dev\t%s (%u)\n", ca->name, ca->dev_idx);
prt_printf(&buf, "watermark\t%s\n", bch2_watermarks[watermark]);
prt_printf(&buf, "data type\t%s\n", __bch2_data_types[data_type]);
prt_printf(&buf, "blocking\t%u\n", cl != NULL);
prt_printf(&buf, "free\t%llu\n", usage->d[BCH_DATA_free].buckets);
prt_printf(&buf, "avail\t%llu\n", dev_buckets_free(ca, *usage, watermark));
prt_printf(&buf, "copygc_wait\t%lu/%lli\n",
bch2_copygc_wait_amount(c),
c->copygc_wait - atomic64_read(&c->io_clock[WRITE].now));
prt_printf(&buf, "seen\t%llu\n", s->buckets_seen);
prt_printf(&buf, "open\t%llu\n", s->skipped_open);
prt_printf(&buf, "need journal commit\t%llu\n", s->skipped_need_journal_commit);
prt_printf(&buf, "nocow\t%llu\n", s->skipped_nocow);
prt_printf(&buf, "nouse\t%llu\n", s->skipped_nouse);
prt_printf(&buf, "mi_btree_bitmap\t%llu\n", s->skipped_mi_btree_bitmap);
if (!IS_ERR(ob)) {
prt_printf(&buf, "allocated\t%llu\n", ob->bucket);
trace_bucket_alloc(c, buf.buf);
} else {
prt_printf(&buf, "err\t%s\n", bch2_err_str(PTR_ERR(ob)));
trace_bucket_alloc_fail(c, buf.buf);
}
printbuf_exit(&buf);
}
/**
* bch2_bucket_alloc_trans - allocate a single bucket from a specific device
* @trans: transaction object
* @ca: device to allocate from
* @watermark: how important is this allocation?
* @data_type: BCH_DATA_journal, btree, user...
* @cl: if not NULL, closure to be used to wait if buckets not available
* @usage: for secondarily also returning the current device usage
*
@ -524,6 +604,7 @@ again:
static struct open_bucket *bch2_bucket_alloc_trans(struct btree_trans *trans,
struct bch_dev *ca,
enum bch_watermark watermark,
enum bch_data_type data_type,
struct closure *cl,
struct bch_dev_usage *usage)
{
@ -531,7 +612,9 @@ static struct open_bucket *bch2_bucket_alloc_trans(struct btree_trans *trans,
struct open_bucket *ob = NULL;
bool freespace = READ_ONCE(ca->mi.freespace_initialized);
u64 avail;
struct bucket_alloc_state s = { 0 };
struct bucket_alloc_state s = {
.btree_bitmap = data_type == BCH_DATA_btree,
};
bool waiting = false;
again:
bch2_dev_usage_read_fast(ca, usage);
@ -541,7 +624,7 @@ again:
bch2_do_discards(c);
if (usage->d[BCH_DATA_need_gc_gens].buckets > avail)
bch2_do_gc_gens(c);
bch2_gc_gens_async(c);
if (should_invalidate_buckets(ca, *usage))
bch2_do_invalidates(c);
@ -569,6 +652,11 @@ alloc:
if (s.skipped_need_journal_commit * 2 > avail)
bch2_journal_flush_async(&c->journal, NULL);
if (!ob && s.btree_bitmap != BTREE_BITMAP_ANY) {
s.btree_bitmap = BTREE_BITMAP_ANY;
goto alloc;
}
if (!ob && freespace && c->curr_recovery_pass <= BCH_RECOVERY_PASS_check_alloc_info) {
freespace = false;
goto alloc;
@ -578,33 +666,24 @@ err:
ob = ERR_PTR(-BCH_ERR_no_buckets_found);
if (!IS_ERR(ob))
trace_and_count(c, bucket_alloc, ca,
bch2_watermarks[watermark],
ob->bucket,
usage->d[BCH_DATA_free].buckets,
avail,
bch2_copygc_wait_amount(c),
c->copygc_wait - atomic64_read(&c->io_clock[WRITE].now),
&s,
cl == NULL,
"");
ob->data_type = data_type;
if (!IS_ERR(ob))
count_event(c, bucket_alloc);
else if (!bch2_err_matches(PTR_ERR(ob), BCH_ERR_transaction_restart))
trace_and_count(c, bucket_alloc_fail, ca,
bch2_watermarks[watermark],
0,
usage->d[BCH_DATA_free].buckets,
avail,
bch2_copygc_wait_amount(c),
c->copygc_wait - atomic64_read(&c->io_clock[WRITE].now),
&s,
cl == NULL,
bch2_err_str(PTR_ERR(ob)));
count_event(c, bucket_alloc_fail);
if (!IS_ERR(ob)
? trace_bucket_alloc_enabled()
: trace_bucket_alloc_fail_enabled())
trace_bucket_alloc2(c, ca, watermark, data_type, cl, usage, &s, ob);
return ob;
}
struct open_bucket *bch2_bucket_alloc(struct bch_fs *c, struct bch_dev *ca,
enum bch_watermark watermark,
enum bch_data_type data_type,
struct closure *cl)
{
struct bch_dev_usage usage;
@ -612,7 +691,7 @@ struct open_bucket *bch2_bucket_alloc(struct bch_fs *c, struct bch_dev *ca,
bch2_trans_do(c, NULL, NULL, 0,
PTR_ERR_OR_ZERO(ob = bch2_bucket_alloc_trans(trans, ca, watermark,
cl, &usage)));
data_type, cl, &usage)));
return ob;
}
@ -678,8 +757,7 @@ static int add_new_bucket(struct bch_fs *c,
unsigned flags,
struct open_bucket *ob)
{
unsigned durability =
bch_dev_bkey_exists(c, ob->dev)->mi.durability;
unsigned durability = ob_dev(c, ob)->mi.durability;
BUG_ON(*nr_effective >= nr_replicas);
@ -711,37 +789,28 @@ int bch2_bucket_alloc_set_trans(struct btree_trans *trans,
struct bch_fs *c = trans->c;
struct dev_alloc_list devs_sorted =
bch2_dev_alloc_list(c, stripe, devs_may_alloc);
unsigned dev;
struct bch_dev *ca;
int ret = -BCH_ERR_insufficient_devices;
unsigned i;
BUG_ON(*nr_effective >= nr_replicas);
for (i = 0; i < devs_sorted.nr; i++) {
for (unsigned i = 0; i < devs_sorted.nr; i++) {
struct bch_dev_usage usage;
struct open_bucket *ob;
dev = devs_sorted.devs[i];
rcu_read_lock();
ca = rcu_dereference(c->devs[dev]);
if (ca)
percpu_ref_get(&ca->ref);
rcu_read_unlock();
unsigned dev = devs_sorted.devs[i];
struct bch_dev *ca = bch2_dev_tryget_noerror(c, dev);
if (!ca)
continue;
if (!ca->mi.durability && *have_cache) {
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
continue;
}
ob = bch2_bucket_alloc_trans(trans, ca, watermark, cl, &usage);
ob = bch2_bucket_alloc_trans(trans, ca, watermark, data_type, cl, &usage);
if (!IS_ERR(ob))
bch2_dev_stripe_increment_inlined(ca, stripe, &usage);
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
if (IS_ERR(ob)) {
ret = PTR_ERR(ob);
@ -750,8 +819,6 @@ int bch2_bucket_alloc_set_trans(struct btree_trans *trans,
continue;
}
ob->data_type = data_type;
if (add_new_bucket(c, ptrs, devs_may_alloc,
nr_replicas, nr_effective,
have_cache, flags, ob)) {
@ -836,7 +903,7 @@ static bool want_bucket(struct bch_fs *c,
bool *have_cache, bool ec,
struct open_bucket *ob)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
struct bch_dev *ca = ob_dev(c, ob);
if (!test_bit(ob->dev, devs_may_alloc->d))
return false;
@ -906,7 +973,7 @@ static int bucket_alloc_set_partial(struct bch_fs *c,
struct open_bucket *ob = c->open_buckets + c->open_buckets_partial[i];
if (want_bucket(c, wp, devs_may_alloc, have_cache, ec, ob)) {
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
struct bch_dev *ca = ob_dev(c, ob);
struct bch_dev_usage usage;
u64 avail;
@ -1291,7 +1358,7 @@ deallocate_extra_replicas(struct bch_fs *c,
unsigned i;
open_bucket_for_each(c, ptrs, ob, i) {
unsigned d = bch_dev_bkey_exists(c, ob->dev)->mi.durability;
unsigned d = ob_dev(c, ob)->mi.durability;
if (d && d <= extra_replicas) {
extra_replicas -= d;
@ -1342,6 +1409,10 @@ retry:
*wp_ret = wp = writepoint_find(trans, write_point.v);
ret = bch2_trans_relock(trans);
if (ret)
goto err;
/* metadata may not allocate on cache devices: */
if (wp->data_type != BCH_DATA_user)
have_cache = true;
@ -1444,7 +1515,7 @@ err:
struct bch_extent_ptr bch2_ob_ptr(struct bch_fs *c, struct open_bucket *ob)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
struct bch_dev *ca = ob_dev(c, ob);
return (struct bch_extent_ptr) {
.type = 1 << BCH_EXTENT_ENTRY_ptr,
@ -1520,7 +1591,7 @@ void bch2_fs_allocator_foreground_init(struct bch_fs *c)
static void bch2_open_bucket_to_text(struct printbuf *out, struct bch_fs *c, struct open_bucket *ob)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
struct bch_dev *ca = ob_dev(c, ob);
unsigned data_type = ob->data_type;
barrier(); /* READ_ONCE() doesn't work on bitfields */
@ -1622,3 +1693,104 @@ void bch2_write_points_to_text(struct printbuf *out, struct bch_fs *c)
prt_str(out, "Btree write point\n");
bch2_write_point_to_text(out, c, &c->btree_write_point);
}
void bch2_fs_alloc_debug_to_text(struct printbuf *out, struct bch_fs *c)
{
unsigned nr[BCH_DATA_NR];
memset(nr, 0, sizeof(nr));
for (unsigned i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
nr[c->open_buckets[i].data_type]++;
printbuf_tabstop_push(out, 24);
percpu_down_read(&c->mark_lock);
prt_printf(out, "hidden\t%llu\n", bch2_fs_usage_read_one(c, &c->usage_base->b.hidden));
prt_printf(out, "btree\t%llu\n", bch2_fs_usage_read_one(c, &c->usage_base->b.btree));
prt_printf(out, "data\t%llu\n", bch2_fs_usage_read_one(c, &c->usage_base->b.data));
prt_printf(out, "cached\t%llu\n", bch2_fs_usage_read_one(c, &c->usage_base->b.cached));
prt_printf(out, "reserved\t%llu\n", bch2_fs_usage_read_one(c, &c->usage_base->b.reserved));
prt_printf(out, "online_reserved\t%llu\n", percpu_u64_get(c->online_reserved));
prt_printf(out, "nr_inodes\t%llu\n", bch2_fs_usage_read_one(c, &c->usage_base->b.nr_inodes));
percpu_up_read(&c->mark_lock);
prt_newline(out);
prt_printf(out, "freelist_wait\t%s\n", c->freelist_wait.list.first ? "waiting" : "empty");
prt_printf(out, "open buckets allocated\t%i\n", OPEN_BUCKETS_COUNT - c->open_buckets_nr_free);
prt_printf(out, "open buckets total\t%u\n", OPEN_BUCKETS_COUNT);
prt_printf(out, "open_buckets_wait\t%s\n", c->open_buckets_wait.list.first ? "waiting" : "empty");
prt_printf(out, "open_buckets_btree\t%u\n", nr[BCH_DATA_btree]);
prt_printf(out, "open_buckets_user\t%u\n", nr[BCH_DATA_user]);
prt_printf(out, "btree reserve cache\t%u\n", c->btree_reserve_cache_nr);
}
void bch2_dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
{
struct bch_fs *c = ca->fs;
struct bch_dev_usage stats = bch2_dev_usage_read(ca);
unsigned nr[BCH_DATA_NR];
memset(nr, 0, sizeof(nr));
for (unsigned i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
nr[c->open_buckets[i].data_type]++;
printbuf_tabstop_push(out, 12);
printbuf_tabstop_push(out, 16);
printbuf_tabstop_push(out, 16);
printbuf_tabstop_push(out, 16);
printbuf_tabstop_push(out, 16);
bch2_dev_usage_to_text(out, &stats);
prt_newline(out);
prt_printf(out, "reserves:\n");
for (unsigned i = 0; i < BCH_WATERMARK_NR; i++)
prt_printf(out, "%s\t%llu\r\n", bch2_watermarks[i], bch2_dev_buckets_reserved(ca, i));
prt_newline(out);
printbuf_tabstops_reset(out);
printbuf_tabstop_push(out, 12);
printbuf_tabstop_push(out, 16);
prt_printf(out, "open buckets\t%i\r\n", ca->nr_open_buckets);
prt_printf(out, "buckets to invalidate\t%llu\r\n", should_invalidate_buckets(ca, stats));
}
void bch2_print_allocator_stuck(struct bch_fs *c)
{
struct printbuf buf = PRINTBUF;
prt_printf(&buf, "Allocator stuck? Waited for 10 seconds\n");
prt_printf(&buf, "Allocator debug:\n");
printbuf_indent_add(&buf, 2);
bch2_fs_alloc_debug_to_text(&buf, c);
printbuf_indent_sub(&buf, 2);
prt_newline(&buf);
for_each_online_member(c, ca) {
prt_printf(&buf, "Dev %u:\n", ca->dev_idx);
printbuf_indent_add(&buf, 2);
bch2_dev_alloc_debug_to_text(&buf, ca);
printbuf_indent_sub(&buf, 2);
prt_newline(&buf);
}
prt_printf(&buf, "Copygc debug:\n");
printbuf_indent_add(&buf, 2);
bch2_copygc_wait_to_text(&buf, c);
printbuf_indent_sub(&buf, 2);
prt_newline(&buf);
prt_printf(&buf, "Journal debug:\n");
printbuf_indent_add(&buf, 2);
bch2_journal_debug_to_text(&buf, &c->journal);
printbuf_indent_sub(&buf, 2);
bch2_print_string_as_lines(KERN_ERR, buf.buf);
printbuf_exit(&buf);
}

15
fs/bcachefs/alloc_foreground.h
View File

@ -30,8 +30,14 @@ void bch2_dev_stripe_increment(struct bch_dev *, struct dev_stripe_state *);
long bch2_bucket_alloc_new_fs(struct bch_dev *);
static inline struct bch_dev *ob_dev(struct bch_fs *c, struct open_bucket *ob)
{
return bch2_dev_have_ref(c, ob->dev);
}
struct open_bucket *bch2_bucket_alloc(struct bch_fs *, struct bch_dev *,
enum bch_watermark, struct closure *);
enum bch_watermark, enum bch_data_type,
struct closure *);
static inline void ob_push(struct bch_fs *c, struct open_buckets *obs,
struct open_bucket *ob)
@ -184,7 +190,7 @@ bch2_alloc_sectors_append_ptrs_inlined(struct bch_fs *c, struct write_point *wp,
wp->sectors_allocated += sectors;
open_bucket_for_each(c, &wp->ptrs, ob, i) {
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
struct bch_dev *ca = ob_dev(c, ob);
struct bch_extent_ptr ptr = bch2_ob_ptr(c, ob);
ptr.cached = cached ||
@ -221,4 +227,9 @@ void bch2_open_buckets_partial_to_text(struct printbuf *, struct bch_fs *);
void bch2_write_points_to_text(struct printbuf *, struct bch_fs *);
void bch2_fs_alloc_debug_to_text(struct printbuf *, struct bch_fs *);
void bch2_dev_alloc_debug_to_text(struct printbuf *, struct bch_dev *);
void bch2_print_allocator_stuck(struct bch_fs *);
#endif /* _BCACHEFS_ALLOC_FOREGROUND_H */

7
fs/bcachefs/alloc_types.h
View File

@ -9,11 +9,18 @@
#include "fifo.h"
struct bucket_alloc_state {
enum {
BTREE_BITMAP_NO,
BTREE_BITMAP_YES,
BTREE_BITMAP_ANY,
} btree_bitmap;
u64 buckets_seen;
u64 skipped_open;
u64 skipped_need_journal_commit;
u64 skipped_nocow;
u64 skipped_nouse;
u64 skipped_mi_btree_bitmap;
};
#define BCH_WATERMARKS() \

158
fs/bcachefs/backpointers.c
View File

@ -23,6 +23,7 @@ static bool extent_matches_bp(struct bch_fs *c,
const union bch_extent_entry *entry;
struct extent_ptr_decoded p;
rcu_read_lock();
bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
struct bpos bucket2;
struct bch_backpointer bp2;
@ -30,31 +31,43 @@ static bool extent_matches_bp(struct bch_fs *c,
if (p.ptr.cached)
continue;
bch2_extent_ptr_to_bp(c, btree_id, level, k, p, entry, &bucket2, &bp2);
struct bch_dev *ca = bch2_dev_rcu(c, p.ptr.dev);
if (!ca)
continue;
bch2_extent_ptr_to_bp(c, ca, btree_id, level, k, p, entry, &bucket2, &bp2);
if (bpos_eq(bucket, bucket2) &&
!memcmp(&bp, &bp2, sizeof(bp)))
!memcmp(&bp, &bp2, sizeof(bp))) {
rcu_read_unlock();
return true;
}
}
rcu_read_unlock();
return false;
}
int bch2_backpointer_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_s_c_backpointer bp = bkey_s_c_to_backpointer(k);
/* these will be caught by fsck */
if (!bch2_dev_exists2(c, bp.k->p.inode))
rcu_read_lock();
struct bch_dev *ca = bch2_dev_rcu(c, bp.k->p.inode);
if (!ca) {
/* these will be caught by fsck */
rcu_read_unlock();
return 0;
}
struct bch_dev *ca = bch_dev_bkey_exists(c, bp.k->p.inode);
struct bpos bucket = bp_pos_to_bucket(c, bp.k->p);
struct bpos bucket = bp_pos_to_bucket(ca, bp.k->p);
struct bpos bp_pos = bucket_pos_to_bp_noerror(ca, bucket, bp.v->bucket_offset);
rcu_read_unlock();
int ret = 0;
bkey_fsck_err_on((bp.v->bucket_offset >> MAX_EXTENT_COMPRESS_RATIO_SHIFT) >= ca->mi.bucket_size ||
!bpos_eq(bp.k->p, bucket_pos_to_bp_noerror(ca, bucket, bp.v->bucket_offset)),
!bpos_eq(bp.k->p, bp_pos),
c, err,
backpointer_bucket_offset_wrong,
"backpointer bucket_offset wrong");
@ -75,10 +88,16 @@ void bch2_backpointer_to_text(struct printbuf *out, const struct bch_backpointer
void bch2_backpointer_k_to_text(struct printbuf *out, struct bch_fs *c, struct bkey_s_c k)
{
if (bch2_dev_exists2(c, k.k->p.inode)) {
rcu_read_lock();
struct bch_dev *ca = bch2_dev_rcu(c, k.k->p.inode);
if (ca) {
struct bpos bucket = bp_pos_to_bucket(ca, k.k->p);
rcu_read_unlock();
prt_str(out, "bucket=");
bch2_bpos_to_text(out, bp_pos_to_bucket(c, k.k->p));
bch2_bpos_to_text(out, bucket);
prt_str(out, " ");
} else {
rcu_read_unlock();
}
bch2_backpointer_to_text(out, bkey_s_c_to_backpointer(k).v);
@ -117,8 +136,7 @@ static noinline int backpointer_mod_err(struct btree_trans *trans,
bch_err(c, "%s", buf.buf);
} else if (c->curr_recovery_pass > BCH_RECOVERY_PASS_check_extents_to_backpointers) {
prt_printf(&buf, "backpointer not found when deleting");
prt_newline(&buf);
prt_printf(&buf, "backpointer not found when deleting\n");
printbuf_indent_add(&buf, 2);
prt_printf(&buf, "searching for ");
@ -145,6 +163,7 @@ static noinline int backpointer_mod_err(struct btree_trans *trans,
}
int bch2_bucket_backpointer_mod_nowritebuffer(struct btree_trans *trans,
struct bch_dev *ca,
struct bpos bucket,
struct bch_backpointer bp,
struct bkey_s_c orig_k,
@ -161,7 +180,7 @@ int bch2_bucket_backpointer_mod_nowritebuffer(struct btree_trans *trans,
return ret;
bkey_backpointer_init(&bp_k->k_i);
bp_k->k.p = bucket_pos_to_bp(trans->c, bucket, bp.bucket_offset);
bp_k->k.p = bucket_pos_to_bp(ca, bucket, bp.bucket_offset);
bp_k->v = bp;
if (!insert) {
@ -171,9 +190,9 @@ int bch2_bucket_backpointer_mod_nowritebuffer(struct btree_trans *trans,
k = bch2_bkey_get_iter(trans, &bp_iter, BTREE_ID_backpointers,
bp_k->k.p,
BTREE_ITER_INTENT|
BTREE_ITER_SLOTS|
BTREE_ITER_WITH_UPDATES);
BTREE_ITER_intent|
BTREE_ITER_slots|
BTREE_ITER_with_updates);
ret = bkey_err(k);
if (ret)
goto err;
@ -197,13 +216,13 @@ err:
* Find the next backpointer >= *bp_offset:
*/
int bch2_get_next_backpointer(struct btree_trans *trans,
struct bch_dev *ca,
struct bpos bucket, int gen,
struct bpos *bp_pos,
struct bch_backpointer *bp,
unsigned iter_flags)
{
struct bch_fs *c = trans->c;
struct bpos bp_end_pos = bucket_pos_to_bp(c, bpos_nosnap_successor(bucket), 0);
struct bpos bp_end_pos = bucket_pos_to_bp(ca, bpos_nosnap_successor(bucket), 0);
struct btree_iter alloc_iter = { NULL }, bp_iter = { NULL };
struct bkey_s_c k;
int ret = 0;
@ -213,7 +232,7 @@ int bch2_get_next_backpointer(struct btree_trans *trans,
if (gen >= 0) {
k = bch2_bkey_get_iter(trans, &alloc_iter, BTREE_ID_alloc,
bucket, BTREE_ITER_CACHED|iter_flags);
bucket, BTREE_ITER_cached|iter_flags);
ret = bkey_err(k);
if (ret)
goto out;
@ -223,7 +242,7 @@ int bch2_get_next_backpointer(struct btree_trans *trans,
goto done;
}
*bp_pos = bpos_max(*bp_pos, bucket_pos_to_bp(c, bucket, 0));
*bp_pos = bpos_max(*bp_pos, bucket_pos_to_bp(ca, bucket, 0));
for_each_btree_key_norestart(trans, bp_iter, BTREE_ID_backpointers,
*bp_pos, iter_flags, k, ret) {
@ -249,7 +268,6 @@ static void backpointer_not_found(struct btree_trans *trans,
{
struct bch_fs *c = trans->c;
struct printbuf buf = PRINTBUF;
struct bpos bucket = bp_pos_to_bucket(c, bp_pos);
/*
* If we're using the btree write buffer, the backpointer we were
@ -259,6 +277,10 @@ static void backpointer_not_found(struct btree_trans *trans,
if (likely(!bch2_backpointers_no_use_write_buffer))
return;
struct bpos bucket;
if (!bp_pos_to_bucket_nodev(c, bp_pos, &bucket))
return;
prt_printf(&buf, "backpointer doesn't match %s it points to:\n ",
bp.level ? "btree node" : "extent");
prt_printf(&buf, "bucket: ");
@ -288,15 +310,17 @@ struct bkey_s_c bch2_backpointer_get_key(struct btree_trans *trans,
{
if (likely(!bp.level)) {
struct bch_fs *c = trans->c;
struct bpos bucket = bp_pos_to_bucket(c, bp_pos);
struct bkey_s_c k;
struct bpos bucket;
if (!bp_pos_to_bucket_nodev(c, bp_pos, &bucket))
return bkey_s_c_err(-EIO);
bch2_trans_node_iter_init(trans, iter,
bp.btree_id,
bp.pos,
0, 0,
iter_flags);
k = bch2_btree_iter_peek_slot(iter);
struct bkey_s_c k = bch2_btree_iter_peek_slot(iter);
if (bkey_err(k)) {
bch2_trans_iter_exit(trans, iter);
return k;
@ -325,18 +349,20 @@ struct btree *bch2_backpointer_get_node(struct btree_trans *trans,
struct bch_backpointer bp)
{
struct bch_fs *c = trans->c;
struct bpos bucket = bp_pos_to_bucket(c, bp_pos);
struct btree *b;
BUG_ON(!bp.level);
struct bpos bucket;
if (!bp_pos_to_bucket_nodev(c, bp_pos, &bucket))
return ERR_PTR(-EIO);
bch2_trans_node_iter_init(trans, iter,
bp.btree_id,
bp.pos,
0,
bp.level - 1,
0);
b = bch2_btree_iter_peek_node(iter);
struct btree *b = bch2_btree_iter_peek_node(iter);
if (IS_ERR_OR_NULL(b))
goto err;
@ -367,16 +393,16 @@ static int bch2_check_btree_backpointer(struct btree_trans *trans, struct btree_
struct printbuf buf = PRINTBUF;
int ret = 0;
if (fsck_err_on(!bch2_dev_exists2(c, k.k->p.inode), c,
backpointer_to_missing_device,
"backpointer for missing device:\n%s",
(bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
ret = bch2_btree_delete_at(trans, bp_iter, 0);
struct bpos bucket;
if (!bp_pos_to_bucket_nodev_noerror(c, k.k->p, &bucket)) {
if (fsck_err(c, backpointer_to_missing_device,
"backpointer for missing device:\n%s",
(bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
ret = bch2_btree_delete_at(trans, bp_iter, 0);
goto out;
}
alloc_k = bch2_bkey_get_iter(trans, &alloc_iter, BTREE_ID_alloc,
bp_pos_to_bucket(c, k.k->p), 0);
alloc_k = bch2_bkey_get_iter(trans, &alloc_iter, BTREE_ID_alloc, bucket, 0);
ret = bkey_err(alloc_k);
if (ret)
goto out;
@ -460,8 +486,8 @@ found:
bytes = p.crc.compressed_size << 9;
struct bch_dev *ca = bch_dev_bkey_exists(c, dev);
if (!bch2_dev_get_ioref(ca, READ))
struct bch_dev *ca = bch2_dev_get_ioref(c, dev, READ);
if (!ca)
return false;
data_buf = kvmalloc(bytes, GFP_KERNEL);
@ -511,25 +537,27 @@ static int check_bp_exists(struct btree_trans *trans,
struct printbuf buf = PRINTBUF;
struct bkey_s_c bp_k;
struct bkey_buf tmp;
int ret;
int ret = 0;
bch2_bkey_buf_init(&tmp);
if (!bch2_dev_bucket_exists(c, bucket)) {
struct bch_dev *ca = bch2_dev_bucket_tryget(c, bucket);
if (!ca) {
prt_str(&buf, "extent for nonexistent device:bucket ");
bch2_bpos_to_text(&buf, bucket);
prt_str(&buf, "\n ");
bch2_bkey_val_to_text(&buf, c, orig_k);
bch_err(c, "%s", buf.buf);
return -BCH_ERR_fsck_repair_unimplemented;
ret = -BCH_ERR_fsck_repair_unimplemented;
goto err;
}
if (bpos_lt(bucket, s->bucket_start) ||
bpos_gt(bucket, s->bucket_end))
return 0;
goto out;
bp_k = bch2_bkey_get_iter(trans, &bp_iter, BTREE_ID_backpointers,
bucket_pos_to_bp(c, bucket, bp.bucket_offset),
bucket_pos_to_bp(ca, bucket, bp.bucket_offset),
0);
ret = bkey_err(bp_k);
if (ret)
@ -562,6 +590,7 @@ fsck_err:
bch2_trans_iter_exit(trans, &other_extent_iter);
bch2_trans_iter_exit(trans, &bp_iter);
bch2_bkey_buf_exit(&tmp, c);
bch2_dev_put(ca);
printbuf_exit(&buf);
return ret;
check_existing_bp:
@ -637,13 +666,13 @@ missing:
struct bkey_i_backpointer n_bp_k;
bkey_backpointer_init(&n_bp_k.k_i);
n_bp_k.k.p = bucket_pos_to_bp(trans->c, bucket, bp.bucket_offset);
n_bp_k.k.p = bucket_pos_to_bp(ca, bucket, bp.bucket_offset);
n_bp_k.v = bp;
prt_printf(&buf, "\n want: ");
bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&n_bp_k.k_i));
if (fsck_err(c, ptr_to_missing_backpointer, "%s", buf.buf))
ret = bch2_bucket_backpointer_mod(trans, bucket, bp, orig_k, true);
ret = bch2_bucket_backpointer_mod(trans, ca, bucket, bp, orig_k, true);
goto out;
}
@ -667,7 +696,14 @@ static int check_extent_to_backpointers(struct btree_trans *trans,
if (p.ptr.cached)
continue;
bch2_extent_ptr_to_bp(c, btree, level, k, p, entry, &bucket_pos, &bp);
rcu_read_lock();
struct bch_dev *ca = bch2_dev_rcu(c, p.ptr.dev);
if (ca)
bch2_extent_ptr_to_bp(c, ca, btree, level, k, p, entry, &bucket_pos, &bp);
rcu_read_unlock();
if (!ca)
continue;
ret = check_bp_exists(trans, s, bucket_pos, bp, k);
if (ret)
@ -760,7 +796,7 @@ static int bch2_get_btree_in_memory_pos(struct btree_trans *trans,
__for_each_btree_node(trans, iter, btree,
btree == start.btree ? start.pos : POS_MIN,
0, depth, BTREE_ITER_PREFETCH, b, ret) {
0, depth, BTREE_ITER_prefetch, b, ret) {
mem_may_pin -= btree_buf_bytes(b);
if (mem_may_pin <= 0) {
c->btree_cache.pinned_nodes_end = *end =
@ -794,31 +830,13 @@ static int bch2_check_extents_to_backpointers_pass(struct btree_trans *trans,
while (level >= depth) {
struct btree_iter iter;
bch2_trans_node_iter_init(trans, &iter, btree_id, POS_MIN, 0,
level,
BTREE_ITER_PREFETCH);
while (1) {
bch2_trans_begin(trans);
struct bkey_s_c k = bch2_btree_iter_peek(&iter);
if (!k.k)
break;
ret = bkey_err(k) ?:
check_extent_to_backpointers(trans, s, btree_id, level, k) ?:
bch2_trans_commit(trans, NULL, NULL,
BCH_TRANS_COMMIT_no_enospc);
if (bch2_err_matches(ret, BCH_ERR_transaction_restart)) {
ret = 0;
continue;
}
if (ret)
break;
if (bpos_eq(iter.pos, SPOS_MAX))
break;
bch2_btree_iter_advance(&iter);
}
bch2_trans_iter_exit(trans, &iter);
bch2_trans_node_iter_init(trans, &iter, btree_id, POS_MIN, 0, level,
BTREE_ITER_prefetch);
ret = for_each_btree_key_continue(trans, iter, 0, k, ({
check_extent_to_backpointers(trans, s, btree_id, level, k) ?:
bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
}));
if (ret)
return ret;
@ -936,7 +954,7 @@ static int bch2_check_backpointers_to_extents_pass(struct btree_trans *trans,
struct bpos last_flushed_pos = SPOS_MAX;
return for_each_btree_key_commit(trans, iter, BTREE_ID_backpointers,
POS_MIN, BTREE_ITER_PREFETCH, k,
POS_MIN, BTREE_ITER_prefetch, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
check_one_backpointer(trans, start, end,
bkey_s_c_to_backpointer(k),

43
fs/bcachefs/backpointers.h
View File

@ -6,6 +6,7 @@
#include "btree_iter.h"
#include "btree_update.h"
#include "buckets.h"
#include "error.h"
#include "super.h"
static inline u64 swab40(u64 x)
@ -18,7 +19,7 @@ static inline u64 swab40(u64 x)
}
int bch2_backpointer_invalid(struct bch_fs *, struct bkey_s_c k,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_backpointer_to_text(struct printbuf *, const struct bch_backpointer *);
void bch2_backpointer_k_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
void bch2_backpointer_swab(struct bkey_s);
@ -36,15 +37,29 @@ void bch2_backpointer_swab(struct bkey_s);
* Convert from pos in backpointer btree to pos of corresponding bucket in alloc
* btree:
*/
static inline struct bpos bp_pos_to_bucket(const struct bch_fs *c,
struct bpos bp_pos)
static inline struct bpos bp_pos_to_bucket(const struct bch_dev *ca, struct bpos bp_pos)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, bp_pos.inode);
u64 bucket_sector = bp_pos.offset >> MAX_EXTENT_COMPRESS_RATIO_SHIFT;
return POS(bp_pos.inode, sector_to_bucket(ca, bucket_sector));
}
static inline bool bp_pos_to_bucket_nodev_noerror(struct bch_fs *c, struct bpos bp_pos, struct bpos *bucket)
{
rcu_read_lock();
struct bch_dev *ca = bch2_dev_rcu(c, bp_pos.inode);
if (ca)
*bucket = bp_pos_to_bucket(ca, bp_pos);
rcu_read_unlock();
return ca != NULL;
}
static inline bool bp_pos_to_bucket_nodev(struct bch_fs *c, struct bpos bp_pos, struct bpos *bucket)
{
return !bch2_fs_inconsistent_on(!bp_pos_to_bucket_nodev_noerror(c, bp_pos, bucket),
c, "backpointer for missing device %llu", bp_pos.inode);
}
static inline struct bpos bucket_pos_to_bp_noerror(const struct bch_dev *ca,
struct bpos bucket,
u64 bucket_offset)
@ -57,32 +72,32 @@ static inline struct bpos bucket_pos_to_bp_noerror(const struct bch_dev *ca,
/*
* Convert from pos in alloc btree + bucket offset to pos in backpointer btree:
*/
static inline struct bpos bucket_pos_to_bp(const struct bch_fs *c,
static inline struct bpos bucket_pos_to_bp(const struct bch_dev *ca,
struct bpos bucket,
u64 bucket_offset)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, bucket.inode);
struct bpos ret = bucket_pos_to_bp_noerror(ca, bucket, bucket_offset);
EBUG_ON(!bkey_eq(bucket, bp_pos_to_bucket(c, ret)));
EBUG_ON(!bkey_eq(bucket, bp_pos_to_bucket(ca, ret)));
return ret;
}
int bch2_bucket_backpointer_mod_nowritebuffer(struct btree_trans *, struct bpos bucket,
struct bch_backpointer, struct bkey_s_c, bool);
int bch2_bucket_backpointer_mod_nowritebuffer(struct btree_trans *, struct bch_dev *,
struct bpos bucket, struct bch_backpointer, struct bkey_s_c, bool);
static inline int bch2_bucket_backpointer_mod(struct btree_trans *trans,
struct bch_dev *ca,
struct bpos bucket,
struct bch_backpointer bp,
struct bkey_s_c orig_k,
bool insert)
{
if (unlikely(bch2_backpointers_no_use_write_buffer))
return bch2_bucket_backpointer_mod_nowritebuffer(trans, bucket, bp, orig_k, insert);
return bch2_bucket_backpointer_mod_nowritebuffer(trans, ca, bucket, bp, orig_k, insert);
struct bkey_i_backpointer bp_k;
bkey_backpointer_init(&bp_k.k_i);
bp_k.k.p = bucket_pos_to_bp(trans->c, bucket, bp.bucket_offset);
bp_k.k.p = bucket_pos_to_bp(ca, bucket, bp.bucket_offset);
bp_k.v = bp;
if (!insert) {
@ -120,7 +135,7 @@ static inline enum bch_data_type bch2_bkey_ptr_data_type(struct bkey_s_c k,
}
}
static inline void bch2_extent_ptr_to_bp(struct bch_fs *c,
static inline void bch2_extent_ptr_to_bp(struct bch_fs *c, struct bch_dev *ca,
enum btree_id btree_id, unsigned level,
struct bkey_s_c k, struct extent_ptr_decoded p,
const union bch_extent_entry *entry,
@ -130,7 +145,7 @@ static inline void bch2_extent_ptr_to_bp(struct bch_fs *c,
s64 sectors = level ? btree_sectors(c) : k.k->size;
u32 bucket_offset;
*bucket_pos = PTR_BUCKET_POS_OFFSET(c, &p.ptr, &bucket_offset);
*bucket_pos = PTR_BUCKET_POS_OFFSET(ca, &p.ptr, &bucket_offset);
*bp = (struct bch_backpointer) {
.btree_id = btree_id,
.level = level,
@ -142,7 +157,7 @@ static inline void bch2_extent_ptr_to_bp(struct bch_fs *c,
};
}
int bch2_get_next_backpointer(struct btree_trans *, struct bpos, int,
int bch2_get_next_backpointer(struct btree_trans *, struct bch_dev *ca, struct bpos, int,
struct bpos *, struct bch_backpointer *, unsigned);
struct bkey_s_c bch2_backpointer_get_key(struct btree_trans *, struct btree_iter *,
struct bpos, struct bch_backpointer,

34
fs/bcachefs/bcachefs.h
View File

@ -359,6 +359,8 @@ do { \
#define BCH_DEBUG_PARAMS_ALWAYS() \
BCH_DEBUG_PARAM(key_merging_disabled, \
"Disables merging of extents") \
BCH_DEBUG_PARAM(btree_node_merging_disabled, \
"Disables merging of btree nodes") \
BCH_DEBUG_PARAM(btree_gc_always_rewrite, \
"Causes mark and sweep to compact and rewrite every " \
"btree node it traverses") \
@ -468,6 +470,7 @@ enum bch_time_stats {
#include "quota_types.h"
#include "rebalance_types.h"
#include "replicas_types.h"
#include "sb-members_types.h"
#include "subvolume_types.h"
#include "super_types.h"
#include "thread_with_file_types.h"
@ -516,8 +519,8 @@ enum gc_phase {
struct gc_pos {
enum gc_phase phase;
u16 level;
struct bpos pos;
unsigned level;
};
struct reflink_gc {
@ -534,7 +537,13 @@ struct io_count {
struct bch_dev {
struct kobject kobj;
#ifdef CONFIG_BCACHEFS_DEBUG
atomic_long_t ref;
bool dying;
unsigned long last_put;
#else
struct percpu_ref ref;
#endif
struct completion ref_completion;
struct percpu_ref io_ref;
struct completion io_ref_completion;
@ -560,14 +569,11 @@ struct bch_dev {
struct bch_devs_mask self;
/* biosets used in cloned bios for writing multiple replicas */
struct bio_set replica_set;
/*
* Buckets:
* Per-bucket arrays are protected by c->mark_lock, bucket_lock and
* gc_lock, for device resize - holding any is sufficient for access:
* Or rcu_read_lock(), but only for ptr_stale():
* Or rcu_read_lock(), but only for dev_ptr_stale():
*/
struct bucket_array __rcu *buckets_gc;
struct bucket_gens __rcu *bucket_gens;
@ -581,7 +587,7 @@ struct bch_dev {
/* Allocator: */
u64 new_fs_bucket_idx;
u64 alloc_cursor;
u64 alloc_cursor[3];
unsigned nr_open_buckets;
unsigned nr_btree_reserve;
@ -627,12 +633,12 @@ struct bch_dev {
x(clean_shutdown) \
x(fsck_running) \
x(initial_gc_unfixed) \
x(need_another_gc) \
x(need_delete_dead_snapshots) \
x(error) \
x(topology_error) \
x(errors_fixed) \
x(errors_not_fixed)
x(errors_not_fixed) \
x(no_invalid_checks)
enum bch_fs_flags {
#define x(n) BCH_FS_##n,
@ -715,6 +721,7 @@ struct btree_trans_buf {
x(discard_fast) \
x(invalidate) \
x(delete_dead_snapshots) \
x(gc_gens) \
x(snapshot_delete_pagecache) \
x(sysfs) \
x(btree_write_buffer)
@ -926,7 +933,6 @@ struct bch_fs {
/* JOURNAL SEQ BLACKLIST */
struct journal_seq_blacklist_table *
journal_seq_blacklist_table;
struct work_struct journal_seq_blacklist_gc_work;
/* ALLOCATOR */
spinlock_t freelist_lock;
@ -957,8 +963,7 @@ struct bch_fs {
struct work_struct discard_fast_work;
/* GARBAGE COLLECTION */
struct task_struct *gc_thread;
atomic_t kick_gc;
struct work_struct gc_gens_work;
unsigned long gc_count;
enum btree_id gc_gens_btree;
@ -988,6 +993,7 @@ struct bch_fs {
struct bio_set bio_read;
struct bio_set bio_read_split;
struct bio_set bio_write;
struct bio_set replica_set;
struct mutex bio_bounce_pages_lock;
mempool_t bio_bounce_pages;
struct bucket_nocow_lock_table
@ -1115,7 +1121,6 @@ struct bch_fs {
u64 counters_on_mount[BCH_COUNTER_NR];
u64 __percpu *counters;
unsigned btree_gc_periodic:1;
unsigned copy_gc_enabled:1;
bool promote_whole_extents;
@ -1250,11 +1255,6 @@ static inline s64 bch2_current_time(const struct bch_fs *c)
return timespec_to_bch2_time(c, now);
}
static inline bool bch2_dev_exists2(const struct bch_fs *c, unsigned dev)
{
return dev < c->sb.nr_devices && c->devs[dev];
}
static inline struct stdio_redirect *bch2_fs_stdio_redirect(struct bch_fs *c)
{
struct stdio_redirect *stdio = c->stdio;

10
fs/bcachefs/bcachefs_format.h
View File

@ -76,6 +76,7 @@
#include <asm/byteorder.h>
#include <linux/kernel.h>
#include <linux/uuid.h>
#include <uapi/linux/magic.h>
#include "vstructs.h"
#ifdef __KERNEL__
@ -589,6 +590,13 @@ struct bch_member {
__le64 errors_reset_time;
__le64 seq;
__le64 btree_allocated_bitmap;
/*
* On recovery from a clean shutdown we don't normally read the journal,
* but we still want to resume writing from where we left off so we
* don't overwrite more than is necessary, for list journal debugging:
*/
__le32 last_journal_bucket;
__le32 last_journal_bucket_offset;
};
/*
@ -1283,7 +1291,7 @@ enum bch_compression_opts {
UUID_INIT(0xc68573f6, 0x66ce, 0x90a9, \
0xd9, 0x6a, 0x60, 0xcf, 0x80, 0x3d, 0xf7, 0xef)
#define BCACHEFS_STATFS_MAGIC 0xca451a4e
#define BCACHEFS_STATFS_MAGIC BCACHEFS_SUPER_MAGIC
#define JSET_MAGIC __cpu_to_le64(0x245235c1a3625032ULL)
#define BSET_MAGIC __cpu_to_le64(0x90135c78b99e07f5ULL)

15
fs/bcachefs/bkey.c
View File

@ -640,7 +640,7 @@ struct bkey_format bch2_bkey_format_done(struct bkey_format_state *s)
int bch2_bkey_format_invalid(struct bch_fs *c,
struct bkey_format *f,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
unsigned i, bits = KEY_PACKED_BITS_START;
@ -656,20 +656,17 @@ int bch2_bkey_format_invalid(struct bch_fs *c,
* unpacked format:
*/
for (i = 0; i < f->nr_fields; i++) {
if (!c || c->sb.version_min >= bcachefs_metadata_version_snapshot) {
if ((!c || c->sb.version_min >= bcachefs_metadata_version_snapshot) &&
bch2_bkey_format_field_overflows(f, i)) {
unsigned unpacked_bits = bch2_bkey_format_current.bits_per_field[i];
u64 unpacked_max = ~((~0ULL << 1) << (unpacked_bits - 1));
u64 packed_max = f->bits_per_field[i]
? ~((~0ULL << 1) << (f->bits_per_field[i] - 1))
: 0;
u64 field_offset = le64_to_cpu(f->field_offset[i]);
if (packed_max + field_offset < packed_max ||
packed_max + field_offset > unpacked_max) {
prt_printf(err, "field %u too large: %llu + %llu > %llu",
i, packed_max, field_offset, unpacked_max);
return -BCH_ERR_invalid;
}
prt_printf(err, "field %u too large: %llu + %llu > %llu",
i, packed_max, le64_to_cpu(f->field_offset[i]), unpacked_max);
return -BCH_ERR_invalid;
}
bits += f->bits_per_field[i];

33
fs/bcachefs/bkey.h
View File

@ -9,10 +9,10 @@
#include "util.h"
#include "vstructs.h"
enum bkey_invalid_flags {
BKEY_INVALID_WRITE = (1U << 0),
BKEY_INVALID_COMMIT = (1U << 1),
BKEY_INVALID_JOURNAL = (1U << 2),
enum bch_validate_flags {
BCH_VALIDATE_write = (1U << 0),
BCH_VALIDATE_commit = (1U << 1),
BCH_VALIDATE_journal = (1U << 2),
};
#if 0
@ -574,8 +574,31 @@ static inline void bch2_bkey_format_add_key(struct bkey_format_state *s, const s
void bch2_bkey_format_add_pos(struct bkey_format_state *, struct bpos);
struct bkey_format bch2_bkey_format_done(struct bkey_format_state *);
static inline bool bch2_bkey_format_field_overflows(struct bkey_format *f, unsigned i)
{
unsigned f_bits = f->bits_per_field[i];
unsigned unpacked_bits = bch2_bkey_format_current.bits_per_field[i];
u64 unpacked_mask = ~((~0ULL << 1) << (unpacked_bits - 1));
u64 field_offset = le64_to_cpu(f->field_offset[i]);
if (f_bits > unpacked_bits)
return true;
if ((f_bits == unpacked_bits) && field_offset)
return true;
u64 f_mask = f_bits
? ~((~0ULL << (f_bits - 1)) << 1)
: 0;
if (((field_offset + f_mask) & unpacked_mask) < field_offset)
return true;
return false;
}
int bch2_bkey_format_invalid(struct bch_fs *, struct bkey_format *,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_bkey_format_to_text(struct printbuf *, const struct bkey_format *);
#endif /* _BCACHEFS_BKEY_H */

22
fs/bcachefs/bkey_methods.c
View File

@ -27,7 +27,7 @@ const char * const bch2_bkey_types[] = {
};
static int deleted_key_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags, struct printbuf *err)
enum bch_validate_flags flags, struct printbuf *err)
{
return 0;
}
@ -41,7 +41,7 @@ static int deleted_key_invalid(struct bch_fs *c, struct bkey_s_c k,
})
static int empty_val_key_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags, struct printbuf *err)
enum bch_validate_flags flags, struct printbuf *err)
{
int ret = 0;
@ -58,7 +58,7 @@ fsck_err:
})
static int key_type_cookie_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags, struct printbuf *err)
enum bch_validate_flags flags, struct printbuf *err)
{
return 0;
}
@ -82,7 +82,7 @@ static void key_type_cookie_to_text(struct printbuf *out, struct bch_fs *c,
})
static int key_type_inline_data_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags, struct printbuf *err)
enum bch_validate_flags flags, struct printbuf *err)
{
return 0;
}
@ -123,9 +123,12 @@ const struct bkey_ops bch2_bkey_null_ops = {
};
int bch2_bkey_val_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
if (test_bit(BCH_FS_no_invalid_checks, &c->flags))
return 0;
const struct bkey_ops *ops = bch2_bkey_type_ops(k.k->type);
int ret = 0;
@ -159,9 +162,12 @@ const char *bch2_btree_node_type_str(enum btree_node_type type)
int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
enum btree_node_type type,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
if (test_bit(BCH_FS_no_invalid_checks, &c->flags))
return 0;
int ret = 0;
bkey_fsck_err_on(k.k->u64s < BKEY_U64s, c, err,
@ -172,7 +178,7 @@ int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
return 0;
bkey_fsck_err_on(k.k->type < KEY_TYPE_MAX &&
(type == BKEY_TYPE_btree || (flags & BKEY_INVALID_COMMIT)) &&
(type == BKEY_TYPE_btree || (flags & BCH_VALIDATE_commit)) &&
!(bch2_key_types_allowed[type] & BIT_ULL(k.k->type)), c, err,
bkey_invalid_type_for_btree,
"invalid key type for btree %s (%s)",
@ -224,7 +230,7 @@ fsck_err:
int bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
enum btree_node_type type,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
return __bch2_bkey_invalid(c, k, type, flags, err) ?:

73
fs/bcachefs/bkey_methods.h
View File

@ -22,14 +22,15 @@ extern const struct bkey_ops bch2_bkey_null_ops;
*/
struct bkey_ops {
int (*key_invalid)(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags, struct printbuf *err);
enum bch_validate_flags flags, struct printbuf *err);
void (*val_to_text)(struct printbuf *, struct bch_fs *,
struct bkey_s_c);
void (*swab)(struct bkey_s);
bool (*key_normalize)(struct bch_fs *, struct bkey_s);
bool (*key_merge)(struct bch_fs *, struct bkey_s, struct bkey_s_c);
int (*trigger)(struct btree_trans *, enum btree_id, unsigned,
struct bkey_s_c, struct bkey_s, unsigned);
struct bkey_s_c, struct bkey_s,
enum btree_iter_update_trigger_flags);
void (*compat)(enum btree_id id, unsigned version,
unsigned big_endian, int write,
struct bkey_s);
@ -48,11 +49,11 @@ static inline const struct bkey_ops *bch2_bkey_type_ops(enum bch_bkey_type type)
}
int bch2_bkey_val_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
int __bch2_bkey_invalid(struct bch_fs *, struct bkey_s_c, enum btree_node_type,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
int bch2_bkey_invalid(struct bch_fs *, struct bkey_s_c, enum btree_node_type,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
int bch2_bkey_in_btree_node(struct bch_fs *, struct btree *,
struct bkey_s_c, struct printbuf *);
@ -76,56 +77,10 @@ static inline bool bch2_bkey_maybe_mergable(const struct bkey *l, const struct b
bool bch2_bkey_merge(struct bch_fs *, struct bkey_s, struct bkey_s_c);
enum btree_update_flags {
__BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE = __BTREE_ITER_FLAGS_END,
__BTREE_UPDATE_NOJOURNAL,
__BTREE_UPDATE_KEY_CACHE_RECLAIM,
__BTREE_TRIGGER_NORUN,
__BTREE_TRIGGER_TRANSACTIONAL,
__BTREE_TRIGGER_ATOMIC,
__BTREE_TRIGGER_GC,
__BTREE_TRIGGER_INSERT,
__BTREE_TRIGGER_OVERWRITE,
__BTREE_TRIGGER_BUCKET_INVALIDATE,
};
#define BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE (1U << __BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE)
#define BTREE_UPDATE_NOJOURNAL (1U << __BTREE_UPDATE_NOJOURNAL)
#define BTREE_UPDATE_KEY_CACHE_RECLAIM (1U << __BTREE_UPDATE_KEY_CACHE_RECLAIM)
/* Don't run triggers at all */
#define BTREE_TRIGGER_NORUN (1U << __BTREE_TRIGGER_NORUN)
/*
* If set, we're running transactional triggers as part of a transaction commit:
* triggers may generate new updates
*
* If cleared, and either BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE are set,
* we're running atomic triggers during a transaction commit: we have our
* journal reservation, we're holding btree node write locks, and we know the
* transaction is going to commit (returning an error here is a fatal error,
* causing us to go emergency read-only)
*/
#define BTREE_TRIGGER_TRANSACTIONAL (1U << __BTREE_TRIGGER_TRANSACTIONAL)
#define BTREE_TRIGGER_ATOMIC (1U << __BTREE_TRIGGER_ATOMIC)
/* We're in gc/fsck: running triggers to recalculate e.g. disk usage */
#define BTREE_TRIGGER_GC (1U << __BTREE_TRIGGER_GC)
/* @new is entering the btree */
#define BTREE_TRIGGER_INSERT (1U << __BTREE_TRIGGER_INSERT)
/* @old is leaving the btree */
#define BTREE_TRIGGER_OVERWRITE (1U << __BTREE_TRIGGER_OVERWRITE)
/* signal from bucket invalidate path to alloc trigger */
#define BTREE_TRIGGER_BUCKET_INVALIDATE (1U << __BTREE_TRIGGER_BUCKET_INVALIDATE)
static inline int bch2_key_trigger(struct btree_trans *trans,
enum btree_id btree, unsigned level,
struct bkey_s_c old, struct bkey_s new,
unsigned flags)
enum btree_iter_update_trigger_flags flags)
{
const struct bkey_ops *ops = bch2_bkey_type_ops(old.k->type ?: new.k->type);
@ -135,8 +90,9 @@ static inline int bch2_key_trigger(struct btree_trans *trans,
}
static inline int bch2_key_trigger_old(struct btree_trans *trans,
enum btree_id btree_id, unsigned level,
struct bkey_s_c old, unsigned flags)
enum btree_id btree_id, unsigned level,
struct bkey_s_c old,
enum btree_iter_update_trigger_flags flags)
{
struct bkey_i deleted;
@ -144,12 +100,13 @@ static inline int bch2_key_trigger_old(struct btree_trans *trans,
deleted.k.p = old.k->p;
return bch2_key_trigger(trans, btree_id, level, old, bkey_i_to_s(&deleted),
BTREE_TRIGGER_OVERWRITE|flags);
BTREE_TRIGGER_overwrite|flags);
}
static inline int bch2_key_trigger_new(struct btree_trans *trans,
enum btree_id btree_id, unsigned level,
struct bkey_s new, unsigned flags)
enum btree_id btree_id, unsigned level,
struct bkey_s new,
enum btree_iter_update_trigger_flags flags)
{
struct bkey_i deleted;
@ -157,7 +114,7 @@ static inline int bch2_key_trigger_new(struct btree_trans *trans,
deleted.k.p = new.k->p;
return bch2_key_trigger(trans, btree_id, level, bkey_i_to_s_c(&deleted), new,
BTREE_TRIGGER_INSERT|flags);
BTREE_TRIGGER_insert|flags);
}
void bch2_bkey_renumber(enum btree_node_type, struct bkey_packed *, int);

81
fs/bcachefs/bkey_sort.c
View File

@ -6,9 +6,9 @@
#include "bset.h"
#include "extents.h"
typedef int (*sort_cmp_fn)(struct btree *,
struct bkey_packed *,
struct bkey_packed *);
typedef int (*sort_cmp_fn)(const struct btree *,
const struct bkey_packed *,
const struct bkey_packed *);
static inline bool sort_iter_end(struct sort_iter *iter)
{
@ -70,9 +70,9 @@ static inline struct bkey_packed *sort_iter_next(struct sort_iter *iter,
/*
* If keys compare equal, compare by pointer order:
*/
static inline int key_sort_fix_overlapping_cmp(struct btree *b,
struct bkey_packed *l,
struct bkey_packed *r)
static inline int key_sort_fix_overlapping_cmp(const struct btree *b,
const struct bkey_packed *l,
const struct bkey_packed *r)
{
return bch2_bkey_cmp_packed(b, l, r) ?:
cmp_int((unsigned long) l, (unsigned long) r);
@ -154,46 +154,59 @@ bch2_sort_repack(struct bset *dst, struct btree *src,
return nr;
}
static inline int sort_keys_cmp(struct btree *b,
struct bkey_packed *l,
struct bkey_packed *r)
static inline int keep_unwritten_whiteouts_cmp(const struct btree *b,
const struct bkey_packed *l,
const struct bkey_packed *r)
{
return bch2_bkey_cmp_packed_inlined(b, l, r) ?:
(int) bkey_deleted(r) - (int) bkey_deleted(l) ?:
(int) l->needs_whiteout - (int) r->needs_whiteout;
(long) l - (long) r;
}
unsigned bch2_sort_keys(struct bkey_packed *dst,
struct sort_iter *iter,
bool filter_whiteouts)
#include "btree_update_interior.h"
/*
* For sorting in the btree node write path: whiteouts not in the unwritten
* whiteouts area are dropped, whiteouts in the unwritten whiteouts area are
* dropped if overwritten by real keys:
*/
unsigned bch2_sort_keys_keep_unwritten_whiteouts(struct bkey_packed *dst, struct sort_iter *iter)
{
const struct bkey_format *f = &iter->b->format;
struct bkey_packed *in, *next, *out = dst;
sort_iter_sort(iter, sort_keys_cmp);
sort_iter_sort(iter, keep_unwritten_whiteouts_cmp);
while ((in = sort_iter_next(iter, sort_keys_cmp))) {
bool needs_whiteout = false;
if (bkey_deleted(in) &&
(filter_whiteouts || !in->needs_whiteout))
while ((in = sort_iter_next(iter, keep_unwritten_whiteouts_cmp))) {
if (bkey_deleted(in) && in < unwritten_whiteouts_start(iter->b))
continue;
while ((next = sort_iter_peek(iter)) &&
!bch2_bkey_cmp_packed_inlined(iter->b, in, next)) {
BUG_ON(in->needs_whiteout &&
next->needs_whiteout);
needs_whiteout |= in->needs_whiteout;
in = sort_iter_next(iter, sort_keys_cmp);
}
if ((next = sort_iter_peek(iter)) &&
!bch2_bkey_cmp_packed_inlined(iter->b, in, next))
continue;
if (bkey_deleted(in)) {
memcpy_u64s_small(out, in, bkeyp_key_u64s(f, in));
set_bkeyp_val_u64s(f, out, 0);
} else {
bkey_p_copy(out, in);
}
out->needs_whiteout |= needs_whiteout;
bkey_p_copy(out, in);
out = bkey_p_next(out);
}
return (u64 *) out - (u64 *) dst;
}
/*
* Main sort routine for compacting a btree node in memory: we always drop
* whiteouts because any whiteouts that need to be written are in the unwritten
* whiteouts area:
*/
unsigned bch2_sort_keys(struct bkey_packed *dst, struct sort_iter *iter)
{
struct bkey_packed *in, *out = dst;
sort_iter_sort(iter, bch2_bkey_cmp_packed_inlined);
while ((in = sort_iter_next(iter, bch2_bkey_cmp_packed_inlined))) {
if (bkey_deleted(in))
continue;
bkey_p_copy(out, in);
out = bkey_p_next(out);
}

4
fs/bcachefs/bkey_sort.h
View File

@ -48,7 +48,7 @@ bch2_sort_repack(struct bset *, struct btree *,
struct btree_node_iter *,
struct bkey_format *, bool);
unsigned bch2_sort_keys(struct bkey_packed *,
struct sort_iter *, bool);
unsigned bch2_sort_keys_keep_unwritten_whiteouts(struct bkey_packed *, struct sort_iter *);
unsigned bch2_sort_keys(struct bkey_packed *, struct sort_iter *);
#endif /* _BCACHEFS_BKEY_SORT_H */

29
fs/bcachefs/bset.c
View File

@ -103,8 +103,6 @@ void bch2_dump_bset(struct bch_fs *c, struct btree *b,
void bch2_dump_btree_node(struct bch_fs *c, struct btree *b)
{
struct bset_tree *t;
console_lock();
for_each_bset(b, t)
bch2_dump_bset(c, b, bset(b, t), t - b->set);
@ -136,7 +134,6 @@ void bch2_dump_btree_node_iter(struct btree *b,
struct btree_nr_keys bch2_btree_node_count_keys(struct btree *b)
{
struct bset_tree *t;
struct bkey_packed *k;
struct btree_nr_keys nr = {};
@ -198,7 +195,6 @@ void bch2_btree_node_iter_verify(struct btree_node_iter *iter,
{
struct btree_node_iter_set *set, *s2;
struct bkey_packed *k, *p;
struct bset_tree *t;
if (bch2_btree_node_iter_end(iter))
return;
@ -213,12 +209,14 @@ void bch2_btree_node_iter_verify(struct btree_node_iter *iter,
/* Verify that set->end is correct: */
btree_node_iter_for_each(iter, set) {
for_each_bset(b, t)
if (set->end == t->end_offset)
if (set->end == t->end_offset) {
BUG_ON(set->k < btree_bkey_first_offset(t) ||
set->k >= t->end_offset);
goto found;
}
BUG();
found:
BUG_ON(set->k < btree_bkey_first_offset(t) ||
set->k >= t->end_offset);
do {} while (0);
}
/* Verify iterator is sorted: */
@ -377,11 +375,9 @@ static struct bkey_float *bkey_float(const struct btree *b,
return ro_aux_tree_base(b, t)->f + idx;
}
static void bset_aux_tree_verify(const struct btree *b)
static void bset_aux_tree_verify(struct btree *b)
{
#ifdef CONFIG_BCACHEFS_DEBUG
const struct bset_tree *t;
for_each_bset(b, t) {
if (t->aux_data_offset == U16_MAX)
continue;
@ -685,20 +681,20 @@ static __always_inline void make_bfloat(struct btree *b, struct bset_tree *t,
}
/* bytes remaining - only valid for last bset: */
static unsigned __bset_tree_capacity(const struct btree *b, const struct bset_tree *t)
static unsigned __bset_tree_capacity(struct btree *b, const struct bset_tree *t)
{
bset_aux_tree_verify(b);
return btree_aux_data_bytes(b) - t->aux_data_offset * sizeof(u64);
}
static unsigned bset_ro_tree_capacity(const struct btree *b, const struct bset_tree *t)
static unsigned bset_ro_tree_capacity(struct btree *b, const struct bset_tree *t)
{
return __bset_tree_capacity(b, t) /
(sizeof(struct bkey_float) + sizeof(u8));
}
static unsigned bset_rw_tree_capacity(const struct btree *b, const struct bset_tree *t)
static unsigned bset_rw_tree_capacity(struct btree *b, const struct bset_tree *t)
{
return __bset_tree_capacity(b, t) / sizeof(struct rw_aux_tree);
}
@ -1374,8 +1370,6 @@ void bch2_btree_node_iter_init(struct btree_node_iter *iter,
void bch2_btree_node_iter_init_from_start(struct btree_node_iter *iter,
struct btree *b)
{
struct bset_tree *t;
memset(iter, 0, sizeof(*iter));
for_each_bset(b, t)
@ -1481,7 +1475,6 @@ struct bkey_packed *bch2_btree_node_iter_prev_all(struct btree_node_iter *iter,
{
struct bkey_packed *k, *prev = NULL;
struct btree_node_iter_set *set;
struct bset_tree *t;
unsigned end = 0;
if (bch2_expensive_debug_checks)
@ -1550,9 +1543,7 @@ struct bkey_s_c bch2_btree_node_iter_peek_unpack(struct btree_node_iter *iter,
void bch2_btree_keys_stats(const struct btree *b, struct bset_stats *stats)
{
const struct bset_tree *t;
for_each_bset(b, t) {
for_each_bset_c(b, t) {
enum bset_aux_tree_type type = bset_aux_tree_type(t);
size_t j;

6
fs/bcachefs/bset.h
View File

@ -206,7 +206,10 @@ static inline size_t btree_aux_data_u64s(const struct btree *b)
}
#define for_each_bset(_b, _t) \
for (_t = (_b)->set; _t < (_b)->set + (_b)->nsets; _t++)
for (struct bset_tree *_t = (_b)->set; _t < (_b)->set + (_b)->nsets; _t++)
#define for_each_bset_c(_b, _t) \
for (const struct bset_tree *_t = (_b)->set; _t < (_b)->set + (_b)->nsets; _t++)
#define bset_tree_for_each_key(_b, _t, _k) \
for (_k = btree_bkey_first(_b, _t); \
@ -294,7 +297,6 @@ static inline struct bset_tree *
bch2_bkey_to_bset_inlined(struct btree *b, struct bkey_packed *k)
{
unsigned offset = __btree_node_key_to_offset(b, k);
struct bset_tree *t;
for_each_bset(b, t)
if (offset <= t->end_offset) {

149
fs/bcachefs/btree_cache.c
View File

@ -16,6 +16,12 @@
#include <linux/prefetch.h>
#include <linux/sched/mm.h>
#define BTREE_CACHE_NOT_FREED_INCREMENT(counter) \
do { \
if (shrinker_counter) \
bc->not_freed_##counter++; \
} while (0)
const char * const bch2_btree_node_flags[] = {
#define x(f) #f,
BTREE_FLAGS()
@ -162,6 +168,9 @@ void bch2_btree_node_hash_remove(struct btree_cache *bc, struct btree *b)
/* Cause future lookups for this node to fail: */
b->hash_val = 0;
if (b->c.btree_id < BTREE_ID_NR)
--bc->used_by_btree[b->c.btree_id];
}
int __bch2_btree_node_hash_insert(struct btree_cache *bc, struct btree *b)
@ -169,8 +178,11 @@ int __bch2_btree_node_hash_insert(struct btree_cache *bc, struct btree *b)
BUG_ON(b->hash_val);
b->hash_val = btree_ptr_hash_val(&b->key);
return rhashtable_lookup_insert_fast(&bc->table, &b->hash,
bch_btree_cache_params);
int ret = rhashtable_lookup_insert_fast(&bc->table, &b->hash,
bch_btree_cache_params);
if (!ret && b->c.btree_id < BTREE_ID_NR)
bc->used_by_btree[b->c.btree_id]++;
return ret;
}
int bch2_btree_node_hash_insert(struct btree_cache *bc, struct btree *b,
@ -190,6 +202,35 @@ int bch2_btree_node_hash_insert(struct btree_cache *bc, struct btree *b,
return ret;
}
void bch2_btree_node_update_key_early(struct btree_trans *trans,
enum btree_id btree, unsigned level,
struct bkey_s_c old, struct bkey_i *new)
{
struct bch_fs *c = trans->c;
struct btree *b;
struct bkey_buf tmp;
int ret;
bch2_bkey_buf_init(&tmp);
bch2_bkey_buf_reassemble(&tmp, c, old);
b = bch2_btree_node_get_noiter(trans, tmp.k, btree, level, true);
if (!IS_ERR_OR_NULL(b)) {
mutex_lock(&c->btree_cache.lock);
bch2_btree_node_hash_remove(&c->btree_cache, b);
bkey_copy(&b->key, new);
ret = __bch2_btree_node_hash_insert(&c->btree_cache, b);
BUG_ON(ret);
mutex_unlock(&c->btree_cache.lock);
six_unlock_read(&b->c.lock);
}
bch2_bkey_buf_exit(&tmp, c);
}
__flatten
static inline struct btree *btree_cache_find(struct btree_cache *bc,
const struct bkey_i *k)
@ -203,7 +244,7 @@ static inline struct btree *btree_cache_find(struct btree_cache *bc,
* this version is for btree nodes that have already been freed (we're not
* reaping a real btree node)
*/
static int __btree_node_reclaim(struct bch_fs *c, struct btree *b, bool flush)
static int __btree_node_reclaim(struct bch_fs *c, struct btree *b, bool flush, bool shrinker_counter)
{
struct btree_cache *bc = &c->btree_cache;
int ret = 0;
@ -225,38 +266,64 @@ wait_on_io:
if (b->flags & ((1U << BTREE_NODE_dirty)|
(1U << BTREE_NODE_read_in_flight)|
(1U << BTREE_NODE_write_in_flight))) {
if (!flush)
if (!flush) {
if (btree_node_dirty(b))
BTREE_CACHE_NOT_FREED_INCREMENT(dirty);
else if (btree_node_read_in_flight(b))
BTREE_CACHE_NOT_FREED_INCREMENT(read_in_flight);
else if (btree_node_write_in_flight(b))
BTREE_CACHE_NOT_FREED_INCREMENT(write_in_flight);
return -BCH_ERR_ENOMEM_btree_node_reclaim;
}
/* XXX: waiting on IO with btree cache lock held */
bch2_btree_node_wait_on_read(b);
bch2_btree_node_wait_on_write(b);
}
if (!six_trylock_intent(&b->c.lock))
if (!six_trylock_intent(&b->c.lock)) {
BTREE_CACHE_NOT_FREED_INCREMENT(lock_intent);
return -BCH_ERR_ENOMEM_btree_node_reclaim;
}
if (!six_trylock_write(&b->c.lock))
if (!six_trylock_write(&b->c.lock)) {
BTREE_CACHE_NOT_FREED_INCREMENT(lock_write);
goto out_unlock_intent;
}
/* recheck under lock */
if (b->flags & ((1U << BTREE_NODE_read_in_flight)|
(1U << BTREE_NODE_write_in_flight))) {
if (!flush)
if (!flush) {
if (btree_node_read_in_flight(b))
BTREE_CACHE_NOT_FREED_INCREMENT(read_in_flight);
else if (btree_node_write_in_flight(b))
BTREE_CACHE_NOT_FREED_INCREMENT(write_in_flight);
goto out_unlock;
}
six_unlock_write(&b->c.lock);
six_unlock_intent(&b->c.lock);
goto wait_on_io;
}
if (btree_node_noevict(b) ||
btree_node_write_blocked(b) ||
btree_node_will_make_reachable(b))
if (btree_node_noevict(b)) {
BTREE_CACHE_NOT_FREED_INCREMENT(noevict);
goto out_unlock;
}
if (btree_node_write_blocked(b)) {
BTREE_CACHE_NOT_FREED_INCREMENT(write_blocked);
goto out_unlock;
}
if (btree_node_will_make_reachable(b)) {
BTREE_CACHE_NOT_FREED_INCREMENT(will_make_reachable);
goto out_unlock;
}
if (btree_node_dirty(b)) {
if (!flush)
if (!flush) {
BTREE_CACHE_NOT_FREED_INCREMENT(dirty);
goto out_unlock;
}
/*
* Using the underscore version because we don't want to compact
* bsets after the write, since this node is about to be evicted
@ -286,14 +353,14 @@ out_unlock_intent:
goto out;
}
static int btree_node_reclaim(struct bch_fs *c, struct btree *b)
static int btree_node_reclaim(struct bch_fs *c, struct btree *b, bool shrinker_counter)
{
return __btree_node_reclaim(c, b, false);
return __btree_node_reclaim(c, b, false, shrinker_counter);
}
static int btree_node_write_and_reclaim(struct bch_fs *c, struct btree *b)
{
return __btree_node_reclaim(c, b, true);
return __btree_node_reclaim(c, b, true, false);
}
static unsigned long bch2_btree_cache_scan(struct shrinker *shrink,
@ -341,11 +408,12 @@ static unsigned long bch2_btree_cache_scan(struct shrinker *shrink,
if (touched >= nr)
goto out;
if (!btree_node_reclaim(c, b)) {
if (!btree_node_reclaim(c, b, true)) {
btree_node_data_free(c, b);
six_unlock_write(&b->c.lock);
six_unlock_intent(&b->c.lock);
freed++;
bc->freed++;
}
}
restart:
@ -354,9 +422,11 @@ restart:
if (btree_node_accessed(b)) {
clear_btree_node_accessed(b);
} else if (!btree_node_reclaim(c, b)) {
bc->not_freed_access_bit++;
} else if (!btree_node_reclaim(c, b, true)) {
freed++;
btree_node_data_free(c, b);
bc->freed++;
bch2_btree_node_hash_remove(bc, b);
six_unlock_write(&b->c.lock);
@ -564,7 +634,7 @@ static struct btree *btree_node_cannibalize(struct bch_fs *c)
struct btree *b;
list_for_each_entry_reverse(b, &bc->live, list)
if (!btree_node_reclaim(c, b))
if (!btree_node_reclaim(c, b, false))
return b;
while (1) {
@ -600,7 +670,7 @@ struct btree *bch2_btree_node_mem_alloc(struct btree_trans *trans, bool pcpu_rea
* disk node. Check the freed list before allocating a new one:
*/
list_for_each_entry(b, freed, list)
if (!btree_node_reclaim(c, b)) {
if (!btree_node_reclaim(c, b, false)) {
list_del_init(&b->list);
goto got_node;
}
@ -626,7 +696,7 @@ got_node:
* the list. Check if there's any freed nodes there:
*/
list_for_each_entry(b2, &bc->freeable, list)
if (!btree_node_reclaim(c, b2)) {
if (!btree_node_reclaim(c, b2, false)) {
swap(b->data, b2->data);
swap(b->aux_data, b2->aux_data);
btree_node_to_freedlist(bc, b2);
@ -846,7 +916,6 @@ static struct btree *__bch2_btree_node_get(struct btree_trans *trans, struct btr
struct bch_fs *c = trans->c;
struct btree_cache *bc = &c->btree_cache;
struct btree *b;
struct bset_tree *t;
bool need_relock = false;
int ret;
@ -966,7 +1035,6 @@ struct btree *bch2_btree_node_get(struct btree_trans *trans, struct btree_path *
{
struct bch_fs *c = trans->c;
struct btree *b;
struct bset_tree *t;
int ret;
EBUG_ON(level >= BTREE_MAX_DEPTH);
@ -1043,7 +1111,6 @@ struct btree *bch2_btree_node_get_noiter(struct btree_trans *trans,
struct bch_fs *c = trans->c;
struct btree_cache *bc = &c->btree_cache;
struct btree *b;
struct bset_tree *t;
int ret;
EBUG_ON(level >= BTREE_MAX_DEPTH);
@ -1240,9 +1307,39 @@ void bch2_btree_node_to_text(struct printbuf *out, struct bch_fs *c, const struc
stats.failed);
}
void bch2_btree_cache_to_text(struct printbuf *out, const struct bch_fs *c)
static void prt_btree_cache_line(struct printbuf *out, const struct bch_fs *c,
const char *label, unsigned nr)
{
prt_printf(out, "nr nodes:\t\t%u\n", c->btree_cache.used);
prt_printf(out, "nr dirty:\t\t%u\n", atomic_read(&c->btree_cache.dirty));
prt_printf(out, "cannibalize lock:\t%p\n", c->btree_cache.alloc_lock);
prt_printf(out, "%s\t", label);
prt_human_readable_u64(out, nr * c->opts.btree_node_size);
prt_printf(out, " (%u)\n", nr);
}
void bch2_btree_cache_to_text(struct printbuf *out, const struct btree_cache *bc)
{
struct bch_fs *c = container_of(bc, struct bch_fs, btree_cache);
if (!out->nr_tabstops)
printbuf_tabstop_push(out, 32);
prt_btree_cache_line(out, c, "total:", bc->used);
prt_btree_cache_line(out, c, "nr dirty:", atomic_read(&bc->dirty));
prt_printf(out, "cannibalize lock:\t%p\n", bc->alloc_lock);
prt_newline(out);
for (unsigned i = 0; i < ARRAY_SIZE(bc->used_by_btree); i++)
prt_btree_cache_line(out, c, bch2_btree_id_str(i), bc->used_by_btree[i]);
prt_newline(out);
prt_printf(out, "freed:\t%u\n", bc->freed);
prt_printf(out, "not freed:\n");
prt_printf(out, " dirty\t%u\n", bc->not_freed_dirty);
prt_printf(out, " write in flight\t%u\n", bc->not_freed_write_in_flight);
prt_printf(out, " read in flight\t%u\n", bc->not_freed_read_in_flight);
prt_printf(out, " lock intent failed\t%u\n", bc->not_freed_lock_intent);
prt_printf(out, " lock write failed\t%u\n", bc->not_freed_lock_write);
prt_printf(out, " access bit\t%u\n", bc->not_freed_access_bit);
prt_printf(out, " no evict failed\t%u\n", bc->not_freed_noevict);
prt_printf(out, " write blocked\t%u\n", bc->not_freed_write_blocked);
prt_printf(out, " will make reachable\t%u\n", bc->not_freed_will_make_reachable);
}

5
fs/bcachefs/btree_cache.h
View File

@ -17,6 +17,9 @@ int __bch2_btree_node_hash_insert(struct btree_cache *, struct btree *);
int bch2_btree_node_hash_insert(struct btree_cache *, struct btree *,
unsigned, enum btree_id);
void bch2_btree_node_update_key_early(struct btree_trans *, enum btree_id, unsigned,
struct bkey_s_c, struct bkey_i *);
void bch2_btree_cache_cannibalize_unlock(struct btree_trans *);
int bch2_btree_cache_cannibalize_lock(struct btree_trans *, struct closure *);
@ -131,6 +134,6 @@ static inline struct btree *btree_node_root(struct bch_fs *c, struct btree *b)
const char *bch2_btree_id_str(enum btree_id);
void bch2_btree_pos_to_text(struct printbuf *, struct bch_fs *, const struct btree *);
void bch2_btree_node_to_text(struct printbuf *, struct bch_fs *, const struct btree *);
void bch2_btree_cache_to_text(struct printbuf *, const struct bch_fs *);
void bch2_btree_cache_to_text(struct printbuf *, const struct btree_cache *);
#endif /* _BCACHEFS_BTREE_CACHE_H */

1050
fs/bcachefs/btree_gc.c
View File

File diff suppressed because it is too large Load Diff

44
fs/bcachefs/btree_gc.h
View File

@ -6,10 +6,7 @@
#include "btree_types.h"
int bch2_check_topology(struct bch_fs *);
int bch2_gc(struct bch_fs *, bool, bool);
int bch2_gc_gens(struct bch_fs *);
void bch2_gc_thread_stop(struct bch_fs *);
int bch2_gc_thread_start(struct bch_fs *);
int bch2_check_allocations(struct bch_fs *);
/*
* For concurrent mark and sweep (with other index updates), we define a total
@ -37,16 +34,16 @@ static inline struct gc_pos gc_phase(enum gc_phase phase)
{
return (struct gc_pos) {
.phase = phase,
.pos = POS_MIN,
.level = 0,
.pos = POS_MIN,
};
}
static inline int gc_pos_cmp(struct gc_pos l, struct gc_pos r)
{
return cmp_int(l.phase, r.phase) ?:
bpos_cmp(l.pos, r.pos) ?:
cmp_int(l.level, r.level);
return cmp_int(l.phase, r.phase) ?:
-cmp_int(l.level, r.level) ?:
bpos_cmp(l.pos, r.pos);
}
static inline enum gc_phase btree_id_to_gc_phase(enum btree_id id)
@ -60,13 +57,13 @@ static inline enum gc_phase btree_id_to_gc_phase(enum btree_id id)
}
}
static inline struct gc_pos gc_pos_btree(enum btree_id id,
struct bpos pos, unsigned level)
static inline struct gc_pos gc_pos_btree(enum btree_id btree, unsigned level,
struct bpos pos)
{
return (struct gc_pos) {
.phase = btree_id_to_gc_phase(id),
.pos = pos,
.phase = btree_id_to_gc_phase(btree),
.level = level,
.pos = pos,
};
}
@ -76,19 +73,7 @@ static inline struct gc_pos gc_pos_btree(enum btree_id id,
*/
static inline struct gc_pos gc_pos_btree_node(struct btree *b)
{
return gc_pos_btree(b->c.btree_id, b->key.k.p, b->c.level);
}
/*
* GC position of the pointer to a btree root: we don't use
* gc_pos_pointer_to_btree_node() here to avoid a potential race with
* btree_split() increasing the tree depth - the new root will have level > the
* old root and thus have a greater gc position than the old root, but that
* would be incorrect since once gc has marked the root it's not coming back.
*/
static inline struct gc_pos gc_pos_btree_root(enum btree_id id)
{
return gc_pos_btree(id, SPOS_MAX, BTREE_MAX_DEPTH);
return gc_pos_btree(b->c.btree_id, b->c.level, b->key.k.p);
}
static inline bool gc_visited(struct bch_fs *c, struct gc_pos pos)
@ -104,11 +89,8 @@ static inline bool gc_visited(struct bch_fs *c, struct gc_pos pos)
return ret;
}
static inline void bch2_do_gc_gens(struct bch_fs *c)
{
atomic_inc(&c->kick_gc);
if (c->gc_thread)
wake_up_process(c->gc_thread);
}
int bch2_gc_gens(struct bch_fs *);
void bch2_gc_gens_async(struct bch_fs *);
void bch2_fs_gc_init(struct bch_fs *);
#endif /* _BCACHEFS_BTREE_GC_H */

117
fs/bcachefs/btree_io.c
View File

@ -23,6 +23,18 @@
#include <linux/sched/mm.h>
static void bch2_btree_node_header_to_text(struct printbuf *out, struct btree_node *bn)
{
prt_printf(out, "btree=%s l=%u seq %llux\n",
bch2_btree_id_str(BTREE_NODE_ID(bn)),
(unsigned) BTREE_NODE_LEVEL(bn), bn->keys.seq);
prt_str(out, "min: ");
bch2_bpos_to_text(out, bn->min_key);
prt_newline(out);
prt_str(out, "max: ");
bch2_bpos_to_text(out, bn->max_key);
}
void bch2_btree_node_io_unlock(struct btree *b)
{
EBUG_ON(!btree_node_write_in_flight(b));
@ -217,7 +229,6 @@ static bool should_compact_bset(struct btree *b, struct bset_tree *t,
static bool bch2_drop_whiteouts(struct btree *b, enum compact_mode mode)
{
struct bset_tree *t;
bool ret = false;
for_each_bset(b, t) {
@ -288,8 +299,7 @@ bool bch2_compact_whiteouts(struct bch_fs *c, struct btree *b,
static void btree_node_sort(struct bch_fs *c, struct btree *b,
unsigned start_idx,
unsigned end_idx,
bool filter_whiteouts)
unsigned end_idx)
{
struct btree_node *out;
struct sort_iter_stack sort_iter;
@ -320,7 +330,7 @@ static void btree_node_sort(struct bch_fs *c, struct btree *b,
start_time = local_clock();
u64s = bch2_sort_keys(out->keys.start, &sort_iter.iter, filter_whiteouts);
u64s = bch2_sort_keys(out->keys.start, &sort_iter.iter);
out->keys.u64s = cpu_to_le16(u64s);
@ -426,13 +436,12 @@ static bool btree_node_compact(struct bch_fs *c, struct btree *b)
break;
if (b->nsets - unwritten_idx > 1) {
btree_node_sort(c, b, unwritten_idx,
b->nsets, false);
btree_node_sort(c, b, unwritten_idx, b->nsets);
ret = true;
}
if (unwritten_idx > 1) {
btree_node_sort(c, b, 0, unwritten_idx, false);
btree_node_sort(c, b, 0, unwritten_idx);
ret = true;
}
@ -441,8 +450,6 @@ static bool btree_node_compact(struct bch_fs *c, struct btree *b)
void bch2_btree_build_aux_trees(struct btree *b)
{
struct bset_tree *t;
for_each_bset(b, t)
bch2_bset_build_aux_tree(b, t,
!bset_written(b, bset(b, t)) &&
@ -524,7 +531,9 @@ static void btree_err_msg(struct printbuf *out, struct bch_fs *c,
prt_printf(out, "at btree ");
bch2_btree_pos_to_text(out, c, b);
prt_printf(out, "\n node offset %u/%u",
printbuf_indent_add(out, 2);
prt_printf(out, "\nnode offset %u/%u",
b->written, btree_ptr_sectors_written(&b->key));
if (i)
prt_printf(out, " bset u64s %u", le16_to_cpu(i->u64s));
@ -543,6 +552,7 @@ static int __btree_err(int ret,
const char *fmt, ...)
{
struct printbuf out = PRINTBUF;
bool silent = c->curr_recovery_pass == BCH_RECOVERY_PASS_scan_for_btree_nodes;
va_list args;
btree_err_msg(&out, c, ca, b, i, b->written, write);
@ -564,12 +574,14 @@ static int __btree_err(int ret,
if (!have_retry && ret == -BCH_ERR_btree_node_read_err_must_retry)
ret = -BCH_ERR_btree_node_read_err_bad_node;
if (ret != -BCH_ERR_btree_node_read_err_fixable)
if (!silent && ret != -BCH_ERR_btree_node_read_err_fixable)
bch2_sb_error_count(c, err_type);
switch (ret) {
case -BCH_ERR_btree_node_read_err_fixable:
ret = bch2_fsck_err(c, FSCK_CAN_FIX, err_type, "%s", out.buf);
ret = !silent
? bch2_fsck_err(c, FSCK_CAN_FIX, err_type, "%s", out.buf)
: -BCH_ERR_fsck_fix;
if (ret != -BCH_ERR_fsck_fix &&
ret != -BCH_ERR_fsck_ignore)
goto fsck_err;
@ -577,14 +589,17 @@ static int __btree_err(int ret,
break;
case -BCH_ERR_btree_node_read_err_want_retry:
case -BCH_ERR_btree_node_read_err_must_retry:
bch2_print_string_as_lines(KERN_ERR, out.buf);
if (!silent)
bch2_print_string_as_lines(KERN_ERR, out.buf);
break;
case -BCH_ERR_btree_node_read_err_bad_node:
bch2_print_string_as_lines(KERN_ERR, out.buf);
if (!silent)
bch2_print_string_as_lines(KERN_ERR, out.buf);
ret = bch2_topology_error(c);
break;
case -BCH_ERR_btree_node_read_err_incompatible:
bch2_print_string_as_lines(KERN_ERR, out.buf);
if (!silent)
bch2_print_string_as_lines(KERN_ERR, out.buf);
ret = -BCH_ERR_fsck_errors_not_fixed;
break;
default:
@ -619,8 +634,6 @@ fsck_err:
__cold
void bch2_btree_node_drop_keys_outside_node(struct btree *b)
{
struct bset_tree *t;
for_each_bset(b, t) {
struct bset *i = bset(b, t);
struct bkey_packed *k;
@ -1021,18 +1034,19 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca,
-BCH_ERR_btree_node_read_err_must_retry,
c, ca, b, NULL,
btree_node_bad_seq,
"got wrong btree node (want %llx got %llx)\n"
"got btree %s level %llu pos %s",
bp->seq, b->data->keys.seq,
bch2_btree_id_str(BTREE_NODE_ID(b->data)),
BTREE_NODE_LEVEL(b->data),
buf.buf);
"got wrong btree node: got\n%s",
(printbuf_reset(&buf),
bch2_btree_node_header_to_text(&buf, b->data),
buf.buf));
} else {
btree_err_on(!b->data->keys.seq,
-BCH_ERR_btree_node_read_err_must_retry,
c, ca, b, NULL,
btree_node_bad_seq,
"bad btree header: seq 0");
"bad btree header: seq 0\n%s",
(printbuf_reset(&buf),
bch2_btree_node_header_to_text(&buf, b->data),
buf.buf));
}
while (b->written < (ptr_written ?: btree_sectors(c))) {
@ -1095,7 +1109,7 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca,
nonce = btree_nonce(i, b->written << 9);
struct bch_csum csum = csum_vstruct(c, BSET_CSUM_TYPE(i), nonce, bne);
csum_bad = bch2_crc_cmp(bne->csum, csum);
if (csum_bad)
if (ca && csum_bad)
bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
btree_err_on(csum_bad,
@ -1249,12 +1263,14 @@ int bch2_btree_node_read_done(struct bch_fs *c, struct bch_dev *ca,
btree_node_reset_sib_u64s(b);
rcu_read_lock();
bkey_for_each_ptr(bch2_bkey_ptrs(bkey_i_to_s(&b->key)), ptr) {
struct bch_dev *ca2 = bch_dev_bkey_exists(c, ptr->dev);
struct bch_dev *ca2 = bch2_dev_rcu(c, ptr->dev);
if (ca2->mi.state != BCH_MEMBER_STATE_rw)
if (!ca2 || ca2->mi.state != BCH_MEMBER_STATE_rw)
set_btree_node_need_rewrite(b);
}
rcu_read_unlock();
if (!ptr_written)
set_btree_node_need_rewrite(b);
@ -1279,8 +1295,8 @@ static void btree_node_read_work(struct work_struct *work)
struct btree_read_bio *rb =
container_of(work, struct btree_read_bio, work);
struct bch_fs *c = rb->c;
struct bch_dev *ca = rb->have_ioref ? bch2_dev_have_ref(c, rb->pick.ptr.dev) : NULL;
struct btree *b = rb->b;
struct bch_dev *ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
struct bio *bio = &rb->bio;
struct bch_io_failures failed = { .nr = 0 };
struct printbuf buf = PRINTBUF;
@ -1292,8 +1308,8 @@ static void btree_node_read_work(struct work_struct *work)
while (1) {
retry = true;
bch_info(c, "retrying read");
ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
rb->have_ioref = bch2_dev_get_ioref(ca, READ);
ca = bch2_dev_get_ioref(c, rb->pick.ptr.dev, READ);
rb->have_ioref = ca != NULL;
bio_reset(bio, NULL, REQ_OP_READ|REQ_SYNC|REQ_META);
bio->bi_iter.bi_sector = rb->pick.ptr.offset;
bio->bi_iter.bi_size = btree_buf_bytes(b);
@ -1307,7 +1323,7 @@ static void btree_node_read_work(struct work_struct *work)
start:
printbuf_reset(&buf);
bch2_btree_pos_to_text(&buf, c, b);
bch2_dev_io_err_on(bio->bi_status, ca, BCH_MEMBER_ERROR_read,
bch2_dev_io_err_on(ca && bio->bi_status, ca, BCH_MEMBER_ERROR_read,
"btree read error %s for %s",
bch2_blk_status_to_str(bio->bi_status), buf.buf);
if (rb->have_ioref)
@ -1363,7 +1379,7 @@ static void btree_node_read_endio(struct bio *bio)
struct bch_fs *c = rb->c;
if (rb->have_ioref) {
struct bch_dev *ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
struct bch_dev *ca = bch2_dev_have_ref(c, rb->pick.ptr.dev);
bch2_latency_acct(ca, rb->start_time, READ);
}
@ -1560,7 +1576,7 @@ static void btree_node_read_all_replicas_endio(struct bio *bio)
struct btree_node_read_all *ra = rb->ra;
if (rb->have_ioref) {
struct bch_dev *ca = bch_dev_bkey_exists(c, rb->pick.ptr.dev);
struct bch_dev *ca = bch2_dev_have_ref(c, rb->pick.ptr.dev);
bch2_latency_acct(ca, rb->start_time, READ);
}
@ -1602,14 +1618,14 @@ static int btree_node_read_all_replicas(struct bch_fs *c, struct btree *b, bool
i = 0;
bkey_for_each_ptr_decode(k.k, ptrs, pick, entry) {
struct bch_dev *ca = bch_dev_bkey_exists(c, pick.ptr.dev);
struct bch_dev *ca = bch2_dev_get_ioref(c, pick.ptr.dev, READ);
struct btree_read_bio *rb =
container_of(ra->bio[i], struct btree_read_bio, bio);
rb->c = c;
rb->b = b;
rb->ra = ra;
rb->start_time = local_clock();
rb->have_ioref = bch2_dev_get_ioref(ca, READ);
rb->have_ioref = ca != NULL;
rb->idx = i;
rb->pick = pick;
rb->bio.bi_iter.bi_sector = pick.ptr.offset;
@ -1679,7 +1695,7 @@ void bch2_btree_node_read(struct btree_trans *trans, struct btree *b,
return;
}
ca = bch_dev_bkey_exists(c, pick.ptr.dev);
ca = bch2_dev_get_ioref(c, pick.ptr.dev, READ);
bio = bio_alloc_bioset(NULL,
buf_pages(b->data, btree_buf_bytes(b)),
@ -1691,7 +1707,7 @@ void bch2_btree_node_read(struct btree_trans *trans, struct btree *b,
rb->b = b;
rb->ra = NULL;
rb->start_time = local_clock();
rb->have_ioref = bch2_dev_get_ioref(ca, READ);
rb->have_ioref = ca != NULL;
rb->pick = pick;
INIT_WORK(&rb->work, btree_node_read_work);
bio->bi_iter.bi_sector = pick.ptr.offset;
@ -1846,7 +1862,6 @@ static void btree_node_write_work(struct work_struct *work)
container_of(work, struct btree_write_bio, work);
struct bch_fs *c = wbio->wbio.c;
struct btree *b = wbio->wbio.bio.bi_private;
struct bch_extent_ptr *ptr;
int ret = 0;
btree_bounce_free(c,
@ -1896,13 +1911,14 @@ static void btree_node_write_endio(struct bio *bio)
struct btree_write_bio *wb = container_of(orig, struct btree_write_bio, wbio);
struct bch_fs *c = wbio->c;
struct btree *b = wbio->bio.bi_private;
struct bch_dev *ca = bch_dev_bkey_exists(c, wbio->dev);
struct bch_dev *ca = wbio->have_ioref ? bch2_dev_have_ref(c, wbio->dev) : NULL;
unsigned long flags;
if (wbio->have_ioref)
bch2_latency_acct(ca, wbio->submit_time, WRITE);
if (bch2_dev_io_err_on(bio->bi_status, ca, BCH_MEMBER_ERROR_write,
if (!ca ||
bch2_dev_io_err_on(bio->bi_status, ca, BCH_MEMBER_ERROR_write,
"btree write error: %s",
bch2_blk_status_to_str(bio->bi_status)) ||
bch2_meta_write_fault("btree")) {
@ -1969,7 +1985,6 @@ static void btree_write_submit(struct work_struct *work)
void __bch2_btree_node_write(struct bch_fs *c, struct btree *b, unsigned flags)
{
struct btree_write_bio *wbio;
struct bset_tree *t;
struct bset *i;
struct btree_node *bn = NULL;
struct btree_node_entry *bne = NULL;
@ -2095,11 +2110,11 @@ do_write:
unwritten_whiteouts_end(b));
SET_BSET_SEPARATE_WHITEOUTS(i, false);
b->whiteout_u64s = 0;
u64s = bch2_sort_keys(i->start, &sort_iter.iter, false);
u64s = bch2_sort_keys_keep_unwritten_whiteouts(i->start, &sort_iter.iter);
le16_add_cpu(&i->u64s, u64s);
b->whiteout_u64s = 0;
BUG_ON(!b->written && i->u64s != b->data->keys.u64s);
set_needs_whiteout(i, false);
@ -2226,7 +2241,6 @@ bool bch2_btree_post_write_cleanup(struct bch_fs *c, struct btree *b)
{
bool invalidated_iter = false;
struct btree_node_entry *bne;
struct bset_tree *t;
if (!btree_node_just_written(b))
return false;
@ -2249,7 +2263,7 @@ bool bch2_btree_post_write_cleanup(struct bch_fs *c, struct btree *b)
* single bset:
*/
if (b->nsets > 1) {
btree_node_sort(c, b, 0, b->nsets, true);
btree_node_sort(c, b, 0, b->nsets);
invalidated_iter = true;
} else {
invalidated_iter = bch2_drop_whiteouts(b, COMPACT_ALL);
@ -2346,20 +2360,13 @@ void bch2_btree_write_stats_to_text(struct printbuf *out, struct bch_fs *c)
printbuf_tabstop_push(out, 20);
printbuf_tabstop_push(out, 10);
prt_tab(out);
prt_str(out, "nr");
prt_tab(out);
prt_str(out, "size");
prt_newline(out);
prt_printf(out, "\tnr\tsize\n");
for (unsigned i = 0; i < BTREE_WRITE_TYPE_NR; i++) {
u64 nr = atomic64_read(&c->btree_write_stats[i].nr);
u64 bytes = atomic64_read(&c->btree_write_stats[i].bytes);
prt_printf(out, "%s:", bch2_btree_write_types[i]);
prt_tab(out);
prt_u64(out, nr);
prt_tab(out);
prt_printf(out, "%s:\t%llu\t", bch2_btree_write_types[i], nr);
prt_human_readable_u64(out, nr ? div64_u64(bytes, nr) : 0);
prt_newline(out);
}

2
fs/bcachefs/btree_io.h
View File

@ -81,8 +81,6 @@ static inline bool should_compact_bset_lazy(struct btree *b,
static inline bool bch2_maybe_compact_whiteouts(struct bch_fs *c, struct btree *b)
{
struct bset_tree *t;
for_each_bset(b, t)
if (should_compact_bset_lazy(b, t))
return bch2_compact_whiteouts(c, b, COMPACT_LAZY);

347
fs/bcachefs/btree_iter.c
View File

@ -61,7 +61,7 @@ static inline int btree_path_cmp(const struct btree_path *l,
static inline struct bpos bkey_successor(struct btree_iter *iter, struct bpos p)
{
/* Are we iterating over keys in all snapshots? */
if (iter->flags & BTREE_ITER_ALL_SNAPSHOTS) {
if (iter->flags & BTREE_ITER_all_snapshots) {
p = bpos_successor(p);
} else {
p = bpos_nosnap_successor(p);
@ -74,7 +74,7 @@ static inline struct bpos bkey_successor(struct btree_iter *iter, struct bpos p)
static inline struct bpos bkey_predecessor(struct btree_iter *iter, struct bpos p)
{
/* Are we iterating over keys in all snapshots? */
if (iter->flags & BTREE_ITER_ALL_SNAPSHOTS) {
if (iter->flags & BTREE_ITER_all_snapshots) {
p = bpos_predecessor(p);
} else {
p = bpos_nosnap_predecessor(p);
@ -88,7 +88,7 @@ static inline struct bpos btree_iter_search_key(struct btree_iter *iter)
{
struct bpos pos = iter->pos;
if ((iter->flags & BTREE_ITER_IS_EXTENTS) &&
if ((iter->flags & BTREE_ITER_is_extents) &&
!bkey_eq(pos, POS_MAX))
pos = bkey_successor(iter, pos);
return pos;
@ -253,13 +253,13 @@ static void bch2_btree_iter_verify(struct btree_iter *iter)
BUG_ON(iter->btree_id >= BTREE_ID_NR);
BUG_ON(!!(iter->flags & BTREE_ITER_CACHED) != btree_iter_path(trans, iter)->cached);
BUG_ON(!!(iter->flags & BTREE_ITER_cached) != btree_iter_path(trans, iter)->cached);
BUG_ON((iter->flags & BTREE_ITER_IS_EXTENTS) &&
(iter->flags & BTREE_ITER_ALL_SNAPSHOTS));
BUG_ON((iter->flags & BTREE_ITER_is_extents) &&
(iter->flags & BTREE_ITER_all_snapshots));
BUG_ON(!(iter->flags & __BTREE_ITER_ALL_SNAPSHOTS) &&
(iter->flags & BTREE_ITER_ALL_SNAPSHOTS) &&
BUG_ON(!(iter->flags & BTREE_ITER_snapshot_field) &&
(iter->flags & BTREE_ITER_all_snapshots) &&
!btree_type_has_snapshot_field(iter->btree_id));
if (iter->update_path)
@ -269,10 +269,10 @@ static void bch2_btree_iter_verify(struct btree_iter *iter)
static void bch2_btree_iter_verify_entry_exit(struct btree_iter *iter)
{
BUG_ON((iter->flags & BTREE_ITER_FILTER_SNAPSHOTS) &&
BUG_ON((iter->flags & BTREE_ITER_filter_snapshots) &&
!iter->pos.snapshot);
BUG_ON(!(iter->flags & BTREE_ITER_ALL_SNAPSHOTS) &&
BUG_ON(!(iter->flags & BTREE_ITER_all_snapshots) &&
iter->pos.snapshot != iter->snapshot);
BUG_ON(bkey_lt(iter->pos, bkey_start_pos(&iter->k)) ||
@ -289,7 +289,7 @@ static int bch2_btree_iter_verify_ret(struct btree_iter *iter, struct bkey_s_c k
if (!bch2_debug_check_iterators)
return 0;
if (!(iter->flags & BTREE_ITER_FILTER_SNAPSHOTS))
if (!(iter->flags & BTREE_ITER_filter_snapshots))
return 0;
if (bkey_err(k) || !k.k)
@ -300,8 +300,8 @@ static int bch2_btree_iter_verify_ret(struct btree_iter *iter, struct bkey_s_c k
k.k->p.snapshot));
bch2_trans_iter_init(trans, &copy, iter->btree_id, iter->pos,
BTREE_ITER_NOPRESERVE|
BTREE_ITER_ALL_SNAPSHOTS);
BTREE_ITER_nopreserve|
BTREE_ITER_all_snapshots);
prev = bch2_btree_iter_prev(&copy);
if (!prev.k)
goto out;
@ -897,7 +897,7 @@ static noinline int btree_node_iter_and_journal_peek(struct btree_trans *trans,
bch2_bkey_buf_reassemble(out, c, k);
if ((flags & BTREE_ITER_PREFETCH) &&
if ((flags & BTREE_ITER_prefetch) &&
c->opts.btree_node_prefetch)
ret = btree_path_prefetch_j(trans, path, &jiter);
@ -944,7 +944,7 @@ static __always_inline int btree_path_down(struct btree_trans *trans,
bch2_bkey_buf_unpack(&tmp, c, l->b, k);
if ((flags & BTREE_ITER_PREFETCH) &&
if ((flags & BTREE_ITER_prefetch) &&
c->opts.btree_node_prefetch) {
ret = btree_path_prefetch(trans, path);
if (ret)
@ -999,6 +999,7 @@ retry_all:
bch2_trans_unlock(trans);
cond_resched();
trans->locked = true;
if (unlikely(trans->memory_allocation_failure)) {
struct closure cl;
@ -1162,6 +1163,7 @@ int bch2_btree_path_traverse_one(struct btree_trans *trans,
goto out_uptodate;
path->level = btree_path_up_until_good_node(trans, path, 0);
unsigned max_level = path->level;
EBUG_ON(btree_path_node(path, path->level) &&
!btree_node_locked(path, path->level));
@ -1192,6 +1194,16 @@ int bch2_btree_path_traverse_one(struct btree_trans *trans,
goto out;
}
}
if (unlikely(max_level > path->level)) {
struct btree_path *linked;
unsigned iter;
trans_for_each_path_with_node(trans, path_l(path)->b, linked, iter)
for (unsigned j = path->level + 1; j < max_level; j++)
linked->l[j] = path->l[j];
}
out_uptodate:
path->uptodate = BTREE_ITER_UPTODATE;
out:
@ -1221,11 +1233,14 @@ static inline void btree_path_copy(struct btree_trans *trans, struct btree_path
}
static btree_path_idx_t btree_path_clone(struct btree_trans *trans, btree_path_idx_t src,
bool intent)
bool intent, unsigned long ip)
{
btree_path_idx_t new = btree_path_alloc(trans, src);
btree_path_copy(trans, trans->paths + new, trans->paths + src);
__btree_path_get(trans->paths + new, intent);
#ifdef TRACK_PATH_ALLOCATED
trans->paths[new].ip_allocated = ip;
#endif
return new;
}
@ -1234,7 +1249,7 @@ btree_path_idx_t __bch2_btree_path_make_mut(struct btree_trans *trans,
btree_path_idx_t path, bool intent, unsigned long ip)
{
__btree_path_put(trans->paths + path, intent);
path = btree_path_clone(trans, path, intent);
path = btree_path_clone(trans, path, intent, ip);
trans->paths[path].preserve = false;
return path;
}
@ -1334,6 +1349,26 @@ static inline void __bch2_path_free(struct btree_trans *trans, btree_path_idx_t
__clear_bit(path, trans->paths_allocated);
}
static bool bch2_btree_path_can_relock(struct btree_trans *trans, struct btree_path *path)
{
unsigned l = path->level;
do {
if (!btree_path_node(path, l))
break;
if (!is_btree_node(path, l))
return false;
if (path->l[l].lock_seq != path->l[l].b->c.lock.seq)
return false;
l++;
} while (l < path->locks_want);
return true;
}
void bch2_path_put(struct btree_trans *trans, btree_path_idx_t path_idx, bool intent)
{
struct btree_path *path = trans->paths + path_idx, *dup;
@ -1348,10 +1383,15 @@ void bch2_path_put(struct btree_trans *trans, btree_path_idx_t path_idx, bool in
if (!dup && !(!path->preserve && !is_btree_node(path, path->level)))
return;
if (path->should_be_locked &&
!trans->restarted &&
(!dup || !bch2_btree_path_relock_norestart(trans, dup)))
return;
if (path->should_be_locked && !trans->restarted) {
if (!dup)
return;
if (!(trans->locked
? bch2_btree_path_relock_norestart(trans, dup)
: bch2_btree_path_can_relock(trans, dup)))
return;
}
if (dup) {
dup->preserve |= path->preserve;
@ -1384,22 +1424,26 @@ void __noreturn bch2_trans_in_restart_error(struct btree_trans *trans)
(void *) trans->last_restarted_ip);
}
void __noreturn bch2_trans_unlocked_error(struct btree_trans *trans)
{
panic("trans should be locked, unlocked by %pS\n",
(void *) trans->last_unlock_ip);
}
noinline __cold
void bch2_trans_updates_to_text(struct printbuf *buf, struct btree_trans *trans)
{
prt_printf(buf, "transaction updates for %s journal seq %llu",
prt_printf(buf, "transaction updates for %s journal seq %llu\n",
trans->fn, trans->journal_res.seq);
prt_newline(buf);
printbuf_indent_add(buf, 2);
trans_for_each_update(trans, i) {
struct bkey_s_c old = { &i->old_k, i->old_v };
prt_printf(buf, "update: btree=%s cached=%u %pS",
prt_printf(buf, "update: btree=%s cached=%u %pS\n",
bch2_btree_id_str(i->btree_id),
i->cached,
(void *) i->ip_allocated);
prt_newline(buf);
prt_printf(buf, " old ");
bch2_bkey_val_to_text(buf, trans->c, old);
@ -1428,23 +1472,63 @@ void bch2_dump_trans_updates(struct btree_trans *trans)
printbuf_exit(&buf);
}
static void bch2_btree_path_to_text(struct printbuf *out, struct btree_trans *trans, btree_path_idx_t path_idx)
static void bch2_btree_path_to_text_short(struct printbuf *out, struct btree_trans *trans, btree_path_idx_t path_idx)
{
struct btree_path *path = trans->paths + path_idx;
prt_printf(out, "path: idx %2u ref %u:%u %c %c btree=%s l=%u pos ",
prt_printf(out, "path: idx %2u ref %u:%u %c %c %c btree=%s l=%u pos ",
path_idx, path->ref, path->intent_ref,
path->preserve ? 'P' : ' ',
path->should_be_locked ? 'S' : ' ',
path->cached ? 'C' : 'B',
bch2_btree_id_str(path->btree_id),
path->level);
bch2_bpos_to_text(out, path->pos);
prt_printf(out, " locks %u", path->nodes_locked);
#ifdef TRACK_PATH_ALLOCATED
prt_printf(out, " %pS", (void *) path->ip_allocated);
#endif
}
static const char *btree_node_locked_str(enum btree_node_locked_type t)
{
switch (t) {
case BTREE_NODE_UNLOCKED:
return "unlocked";
case BTREE_NODE_READ_LOCKED:
return "read";
case BTREE_NODE_INTENT_LOCKED:
return "intent";
case BTREE_NODE_WRITE_LOCKED:
return "write";
default:
return NULL;
}
}
void bch2_btree_path_to_text(struct printbuf *out, struct btree_trans *trans, btree_path_idx_t path_idx)
{
bch2_btree_path_to_text_short(out, trans, path_idx);
struct btree_path *path = trans->paths + path_idx;
prt_printf(out, " uptodate %u locks_want %u", path->uptodate, path->locks_want);
prt_newline(out);
printbuf_indent_add(out, 2);
for (unsigned l = 0; l < BTREE_MAX_DEPTH; l++) {
prt_printf(out, "l=%u locks %s seq %u node ", l,
btree_node_locked_str(btree_node_locked_type(path, l)),
path->l[l].lock_seq);
int ret = PTR_ERR_OR_ZERO(path->l[l].b);
if (ret)
prt_str(out, bch2_err_str(ret));
else
prt_printf(out, "%px", path->l[l].b);
prt_newline(out);
}
printbuf_indent_sub(out, 2);
}
static noinline __cold
@ -1456,8 +1540,10 @@ void __bch2_trans_paths_to_text(struct printbuf *out, struct btree_trans *trans,
if (!nosort)
btree_trans_sort_paths(trans);
trans_for_each_path_idx_inorder(trans, iter)
bch2_btree_path_to_text(out, trans, iter.path_idx);
trans_for_each_path_idx_inorder(trans, iter) {
bch2_btree_path_to_text_short(out, trans, iter.path_idx);
prt_newline(out);
}
}
noinline __cold
@ -1608,11 +1694,12 @@ btree_path_idx_t bch2_path_get(struct btree_trans *trans,
unsigned flags, unsigned long ip)
{
struct btree_path *path;
bool cached = flags & BTREE_ITER_CACHED;
bool intent = flags & BTREE_ITER_INTENT;
bool cached = flags & BTREE_ITER_cached;
bool intent = flags & BTREE_ITER_intent;
struct trans_for_each_path_inorder_iter iter;
btree_path_idx_t path_pos = 0, path_idx;
bch2_trans_verify_not_unlocked(trans);
bch2_trans_verify_not_in_restart(trans);
bch2_trans_verify_locks(trans);
@ -1657,7 +1744,7 @@ btree_path_idx_t bch2_path_get(struct btree_trans *trans,
trans->paths_sorted = false;
}
if (!(flags & BTREE_ITER_NOPRESERVE))
if (!(flags & BTREE_ITER_nopreserve))
path->preserve = true;
if (path->intent_ref)
@ -1678,6 +1765,22 @@ btree_path_idx_t bch2_path_get(struct btree_trans *trans,
return path_idx;
}
btree_path_idx_t bch2_path_get_unlocked_mut(struct btree_trans *trans,
enum btree_id btree_id,
unsigned level,
struct bpos pos)
{
btree_path_idx_t path_idx = bch2_path_get(trans, btree_id, pos, level + 1, level,
BTREE_ITER_nopreserve|
BTREE_ITER_intent, _RET_IP_);
path_idx = bch2_btree_path_make_mut(trans, path_idx, true, _RET_IP_);
struct btree_path *path = trans->paths + path_idx;
bch2_btree_path_downgrade(trans, path);
__bch2_btree_path_unlock(trans, path);
return path_idx;
}
struct bkey_s_c bch2_btree_path_peek_slot(struct btree_path *path, struct bkey *u)
{
@ -1719,6 +1822,19 @@ hole:
return (struct bkey_s_c) { u, NULL };
}
void bch2_set_btree_iter_dontneed(struct btree_iter *iter)
{
struct btree_trans *trans = iter->trans;
if (!iter->path || trans->restarted)
return;
struct btree_path *path = btree_iter_path(trans, iter);
path->preserve = false;
if (path->ref == 1)
path->should_be_locked = false;
}
/* Btree iterators: */
int __must_check
@ -1733,9 +1849,11 @@ bch2_btree_iter_traverse(struct btree_iter *iter)
struct btree_trans *trans = iter->trans;
int ret;
bch2_trans_verify_not_unlocked(trans);
iter->path = bch2_btree_path_set_pos(trans, iter->path,
btree_iter_search_key(iter),
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
btree_iter_ip_allocated(iter));
ret = bch2_btree_path_traverse(iter->trans, iter->path, iter->flags);
@ -1774,7 +1892,7 @@ struct btree *bch2_btree_iter_peek_node(struct btree_iter *iter)
iter->k.p = iter->pos = b->key.k.p;
iter->path = bch2_btree_path_set_pos(trans, iter->path, b->key.k.p,
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
btree_iter_ip_allocated(iter));
btree_path_set_should_be_locked(btree_iter_path(trans, iter));
out:
@ -1835,13 +1953,16 @@ struct btree *bch2_btree_iter_next_node(struct btree_iter *iter)
if (bpos_eq(iter->pos, b->key.k.p)) {
__btree_path_set_level_up(trans, path, path->level++);
} else {
if (btree_lock_want(path, path->level + 1) == BTREE_NODE_UNLOCKED)
btree_node_unlock(trans, path, path->level + 1);
/*
* Haven't gotten to the end of the parent node: go back down to
* the next child node
*/
iter->path = bch2_btree_path_set_pos(trans, iter->path,
bpos_successor(iter->pos),
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
btree_iter_ip_allocated(iter));
path = btree_iter_path(trans, iter);
@ -1859,7 +1980,7 @@ struct btree *bch2_btree_iter_next_node(struct btree_iter *iter)
iter->k.p = iter->pos = b->key.k.p;
iter->path = bch2_btree_path_set_pos(trans, iter->path, b->key.k.p,
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
btree_iter_ip_allocated(iter));
btree_path_set_should_be_locked(btree_iter_path(trans, iter));
EBUG_ON(btree_iter_path(trans, iter)->uptodate);
@ -1878,11 +1999,11 @@ err:
inline bool bch2_btree_iter_advance(struct btree_iter *iter)
{
struct bpos pos = iter->k.p;
bool ret = !(iter->flags & BTREE_ITER_ALL_SNAPSHOTS
bool ret = !(iter->flags & BTREE_ITER_all_snapshots
? bpos_eq(pos, SPOS_MAX)
: bkey_eq(pos, SPOS_MAX));
if (ret && !(iter->flags & BTREE_ITER_IS_EXTENTS))
if (ret && !(iter->flags & BTREE_ITER_is_extents))
pos = bkey_successor(iter, pos);
bch2_btree_iter_set_pos(iter, pos);
return ret;
@ -1891,11 +2012,11 @@ inline bool bch2_btree_iter_advance(struct btree_iter *iter)
inline bool bch2_btree_iter_rewind(struct btree_iter *iter)
{
struct bpos pos = bkey_start_pos(&iter->k);
bool ret = !(iter->flags & BTREE_ITER_ALL_SNAPSHOTS
bool ret = !(iter->flags & BTREE_ITER_all_snapshots
? bpos_eq(pos, POS_MIN)
: bkey_eq(pos, POS_MIN));
if (ret && !(iter->flags & BTREE_ITER_IS_EXTENTS))
if (ret && !(iter->flags & BTREE_ITER_is_extents))
pos = bkey_predecessor(iter, pos);
bch2_btree_iter_set_pos(iter, pos);
return ret;
@ -2006,7 +2127,10 @@ struct bkey_s_c btree_trans_peek_key_cache(struct btree_iter *iter, struct bpos
struct bkey_s_c k;
int ret;
if ((iter->flags & BTREE_ITER_KEY_CACHE_FILL) &&
bch2_trans_verify_not_in_restart(trans);
bch2_trans_verify_not_unlocked(trans);
if ((iter->flags & BTREE_ITER_key_cache_fill) &&
bpos_eq(iter->pos, pos))
return bkey_s_c_null;
@ -2015,17 +2139,17 @@ struct bkey_s_c btree_trans_peek_key_cache(struct btree_iter *iter, struct bpos
if (!iter->key_cache_path)
iter->key_cache_path = bch2_path_get(trans, iter->btree_id, pos,
iter->flags & BTREE_ITER_INTENT, 0,
iter->flags|BTREE_ITER_CACHED|
BTREE_ITER_CACHED_NOFILL,
iter->flags & BTREE_ITER_intent, 0,
iter->flags|BTREE_ITER_cached|
BTREE_ITER_cached_nofill,
_THIS_IP_);
iter->key_cache_path = bch2_btree_path_set_pos(trans, iter->key_cache_path, pos,
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
btree_iter_ip_allocated(iter));
ret = bch2_btree_path_traverse(trans, iter->key_cache_path,
iter->flags|BTREE_ITER_CACHED) ?:
iter->flags|BTREE_ITER_cached) ?:
bch2_btree_path_relock(trans, btree_iter_path(trans, iter), _THIS_IP_);
if (unlikely(ret))
return bkey_s_c_err(ret);
@ -2053,7 +2177,7 @@ static struct bkey_s_c __bch2_btree_iter_peek(struct btree_iter *iter, struct bp
struct btree_path_level *l;
iter->path = bch2_btree_path_set_pos(trans, iter->path, search_key,
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
btree_iter_ip_allocated(iter));
ret = bch2_btree_path_traverse(trans, iter->path, iter->flags);
@ -2078,7 +2202,7 @@ static struct bkey_s_c __bch2_btree_iter_peek(struct btree_iter *iter, struct bp
k = btree_path_level_peek_all(trans->c, l, &iter->k);
if (unlikely(iter->flags & BTREE_ITER_WITH_KEY_CACHE) &&
if (unlikely(iter->flags & BTREE_ITER_with_key_cache) &&
k.k &&
(k2 = btree_trans_peek_key_cache(iter, k.k->p)).k) {
k = k2;
@ -2089,10 +2213,10 @@ static struct bkey_s_c __bch2_btree_iter_peek(struct btree_iter *iter, struct bp
}
}
if (unlikely(iter->flags & BTREE_ITER_WITH_JOURNAL))
if (unlikely(iter->flags & BTREE_ITER_with_journal))
k = btree_trans_peek_journal(trans, iter, k);
if (unlikely((iter->flags & BTREE_ITER_WITH_UPDATES) &&
if (unlikely((iter->flags & BTREE_ITER_with_updates) &&
trans->nr_updates))
bch2_btree_trans_peek_updates(trans, iter, &k);
@ -2144,11 +2268,12 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
struct bpos iter_pos;
int ret;
EBUG_ON((iter->flags & BTREE_ITER_FILTER_SNAPSHOTS) && bkey_eq(end, POS_MAX));
bch2_trans_verify_not_unlocked(trans);
EBUG_ON((iter->flags & BTREE_ITER_filter_snapshots) && bkey_eq(end, POS_MAX));
if (iter->update_path) {
bch2_path_put_nokeep(trans, iter->update_path,
iter->flags & BTREE_ITER_INTENT);
iter->flags & BTREE_ITER_intent);
iter->update_path = 0;
}
@ -2171,7 +2296,7 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
* isn't monotonically increasing before FILTER_SNAPSHOTS, and
* that's what we check against in extents mode:
*/
if (unlikely(!(iter->flags & BTREE_ITER_IS_EXTENTS)
if (unlikely(!(iter->flags & BTREE_ITER_is_extents)
? bkey_gt(k.k->p, end)
: k.k->p.inode > end.inode))
goto end;
@ -2179,13 +2304,13 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
if (iter->update_path &&
!bkey_eq(trans->paths[iter->update_path].pos, k.k->p)) {
bch2_path_put_nokeep(trans, iter->update_path,
iter->flags & BTREE_ITER_INTENT);
iter->flags & BTREE_ITER_intent);
iter->update_path = 0;
}
if ((iter->flags & BTREE_ITER_FILTER_SNAPSHOTS) &&
(iter->flags & BTREE_ITER_INTENT) &&
!(iter->flags & BTREE_ITER_IS_EXTENTS) &&
if ((iter->flags & BTREE_ITER_filter_snapshots) &&
(iter->flags & BTREE_ITER_intent) &&
!(iter->flags & BTREE_ITER_is_extents) &&
!iter->update_path) {
struct bpos pos = k.k->p;
@ -2200,12 +2325,12 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
* advance, same as on exit for iter->path, but only up
* to snapshot
*/
__btree_path_get(trans->paths + iter->path, iter->flags & BTREE_ITER_INTENT);
__btree_path_get(trans->paths + iter->path, iter->flags & BTREE_ITER_intent);
iter->update_path = iter->path;
iter->update_path = bch2_btree_path_set_pos(trans,
iter->update_path, pos,
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
_THIS_IP_);
ret = bch2_btree_path_traverse(trans, iter->update_path, iter->flags);
if (unlikely(ret)) {
@ -2218,7 +2343,7 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
* We can never have a key in a leaf node at POS_MAX, so
* we don't have to check these successor() calls:
*/
if ((iter->flags & BTREE_ITER_FILTER_SNAPSHOTS) &&
if ((iter->flags & BTREE_ITER_filter_snapshots) &&
!bch2_snapshot_is_ancestor(trans->c,
iter->snapshot,
k.k->p.snapshot)) {
@ -2227,7 +2352,7 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
}
if (bkey_whiteout(k.k) &&
!(iter->flags & BTREE_ITER_ALL_SNAPSHOTS)) {
!(iter->flags & BTREE_ITER_all_snapshots)) {
search_key = bkey_successor(iter, k.k->p);
continue;
}
@ -2237,12 +2362,12 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
* equal to the key we just returned - except extents can
* straddle iter->pos:
*/
if (!(iter->flags & BTREE_ITER_IS_EXTENTS))
if (!(iter->flags & BTREE_ITER_is_extents))
iter_pos = k.k->p;
else
iter_pos = bkey_max(iter->pos, bkey_start_pos(k.k));
if (unlikely(!(iter->flags & BTREE_ITER_IS_EXTENTS)
if (unlikely(!(iter->flags & BTREE_ITER_is_extents)
? bkey_gt(iter_pos, end)
: bkey_ge(iter_pos, end)))
goto end;
@ -2253,7 +2378,7 @@ struct bkey_s_c bch2_btree_iter_peek_upto(struct btree_iter *iter, struct bpos e
iter->pos = iter_pos;
iter->path = bch2_btree_path_set_pos(trans, iter->path, k.k->p,
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
btree_iter_ip_allocated(iter));
btree_path_set_should_be_locked(btree_iter_path(trans, iter));
@ -2266,7 +2391,7 @@ out_no_locked:
btree_path_set_should_be_locked(trans->paths + iter->update_path);
}
if (!(iter->flags & BTREE_ITER_ALL_SNAPSHOTS))
if (!(iter->flags & BTREE_ITER_all_snapshots))
iter->pos.snapshot = iter->snapshot;
ret = bch2_btree_iter_verify_ret(iter, k);
@ -2316,21 +2441,22 @@ struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *iter)
btree_path_idx_t saved_path = 0;
int ret;
bch2_trans_verify_not_unlocked(trans);
EBUG_ON(btree_iter_path(trans, iter)->cached ||
btree_iter_path(trans, iter)->level);
if (iter->flags & BTREE_ITER_WITH_JOURNAL)
if (iter->flags & BTREE_ITER_with_journal)
return bkey_s_c_err(-BCH_ERR_btree_iter_with_journal_not_supported);
bch2_btree_iter_verify(iter);
bch2_btree_iter_verify_entry_exit(iter);
if (iter->flags & BTREE_ITER_FILTER_SNAPSHOTS)
if (iter->flags & BTREE_ITER_filter_snapshots)
search_key.snapshot = U32_MAX;
while (1) {
iter->path = bch2_btree_path_set_pos(trans, iter->path, search_key,
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
btree_iter_ip_allocated(iter));
ret = bch2_btree_path_traverse(trans, iter->path, iter->flags);
@ -2345,17 +2471,17 @@ struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *iter)
k = btree_path_level_peek(trans, path, &path->l[0], &iter->k);
if (!k.k ||
((iter->flags & BTREE_ITER_IS_EXTENTS)
((iter->flags & BTREE_ITER_is_extents)
? bpos_ge(bkey_start_pos(k.k), search_key)
: bpos_gt(k.k->p, search_key)))
k = btree_path_level_prev(trans, path, &path->l[0], &iter->k);
if (unlikely((iter->flags & BTREE_ITER_WITH_UPDATES) &&
if (unlikely((iter->flags & BTREE_ITER_with_updates) &&
trans->nr_updates))
bch2_btree_trans_peek_prev_updates(trans, iter, &k);
if (likely(k.k)) {
if (iter->flags & BTREE_ITER_FILTER_SNAPSHOTS) {
if (iter->flags & BTREE_ITER_filter_snapshots) {
if (k.k->p.snapshot == iter->snapshot)
goto got_key;
@ -2366,7 +2492,7 @@ struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *iter)
*/
if (saved_path && !bkey_eq(k.k->p, saved_k.p)) {
bch2_path_put_nokeep(trans, iter->path,
iter->flags & BTREE_ITER_INTENT);
iter->flags & BTREE_ITER_intent);
iter->path = saved_path;
saved_path = 0;
iter->k = saved_k;
@ -2379,9 +2505,10 @@ struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *iter)
k.k->p.snapshot)) {
if (saved_path)
bch2_path_put_nokeep(trans, saved_path,
iter->flags & BTREE_ITER_INTENT);
iter->flags & BTREE_ITER_intent);
saved_path = btree_path_clone(trans, iter->path,
iter->flags & BTREE_ITER_INTENT);
iter->flags & BTREE_ITER_intent,
_THIS_IP_);
path = btree_iter_path(trans, iter);
saved_k = *k.k;
saved_v = k.v;
@ -2392,9 +2519,9 @@ struct bkey_s_c bch2_btree_iter_peek_prev(struct btree_iter *iter)
}
got_key:
if (bkey_whiteout(k.k) &&
!(iter->flags & BTREE_ITER_ALL_SNAPSHOTS)) {
!(iter->flags & BTREE_ITER_all_snapshots)) {
search_key = bkey_predecessor(iter, k.k->p);
if (iter->flags & BTREE_ITER_FILTER_SNAPSHOTS)
if (iter->flags & BTREE_ITER_filter_snapshots)
search_key.snapshot = U32_MAX;
continue;
}
@ -2418,11 +2545,11 @@ got_key:
if (bkey_lt(k.k->p, iter->pos))
iter->pos = k.k->p;
if (iter->flags & BTREE_ITER_FILTER_SNAPSHOTS)
if (iter->flags & BTREE_ITER_filter_snapshots)
iter->pos.snapshot = iter->snapshot;
out_no_locked:
if (saved_path)
bch2_path_put_nokeep(trans, saved_path, iter->flags & BTREE_ITER_INTENT);
bch2_path_put_nokeep(trans, saved_path, iter->flags & BTREE_ITER_intent);
bch2_btree_iter_verify_entry_exit(iter);
bch2_btree_iter_verify(iter);
@ -2452,12 +2579,13 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
struct bkey_s_c k;
int ret;
bch2_trans_verify_not_unlocked(trans);
bch2_btree_iter_verify(iter);
bch2_btree_iter_verify_entry_exit(iter);
EBUG_ON(btree_iter_path(trans, iter)->level && (iter->flags & BTREE_ITER_WITH_KEY_CACHE));
EBUG_ON(btree_iter_path(trans, iter)->level && (iter->flags & BTREE_ITER_with_key_cache));
/* extents can't span inode numbers: */
if ((iter->flags & BTREE_ITER_IS_EXTENTS) &&
if ((iter->flags & BTREE_ITER_is_extents) &&
unlikely(iter->pos.offset == KEY_OFFSET_MAX)) {
if (iter->pos.inode == KEY_INODE_MAX)
return bkey_s_c_null;
@ -2467,7 +2595,7 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
search_key = btree_iter_search_key(iter);
iter->path = bch2_btree_path_set_pos(trans, iter->path, search_key,
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
btree_iter_ip_allocated(iter));
ret = bch2_btree_path_traverse(trans, iter->path, iter->flags);
@ -2476,22 +2604,22 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
goto out_no_locked;
}
if ((iter->flags & BTREE_ITER_CACHED) ||
!(iter->flags & (BTREE_ITER_IS_EXTENTS|BTREE_ITER_FILTER_SNAPSHOTS))) {
if ((iter->flags & BTREE_ITER_cached) ||
!(iter->flags & (BTREE_ITER_is_extents|BTREE_ITER_filter_snapshots))) {
k = bkey_s_c_null;
if (unlikely((iter->flags & BTREE_ITER_WITH_UPDATES) &&
if (unlikely((iter->flags & BTREE_ITER_with_updates) &&
trans->nr_updates)) {
bch2_btree_trans_peek_slot_updates(trans, iter, &k);
if (k.k)
goto out;
}
if (unlikely(iter->flags & BTREE_ITER_WITH_JOURNAL) &&
if (unlikely(iter->flags & BTREE_ITER_with_journal) &&
(k = btree_trans_peek_slot_journal(trans, iter)).k)
goto out;
if (unlikely(iter->flags & BTREE_ITER_WITH_KEY_CACHE) &&
if (unlikely(iter->flags & BTREE_ITER_with_key_cache) &&
(k = btree_trans_peek_key_cache(iter, iter->pos)).k) {
if (!bkey_err(k))
iter->k = *k.k;
@ -2506,12 +2634,12 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
struct bpos next;
struct bpos end = iter->pos;
if (iter->flags & BTREE_ITER_IS_EXTENTS)
if (iter->flags & BTREE_ITER_is_extents)
end.offset = U64_MAX;
EBUG_ON(btree_iter_path(trans, iter)->level);
if (iter->flags & BTREE_ITER_INTENT) {
if (iter->flags & BTREE_ITER_intent) {
struct btree_iter iter2;
bch2_trans_copy_iter(&iter2, iter);
@ -2542,7 +2670,7 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
bkey_init(&iter->k);
iter->k.p = iter->pos;
if (iter->flags & BTREE_ITER_IS_EXTENTS) {
if (iter->flags & BTREE_ITER_is_extents) {
bch2_key_resize(&iter->k,
min_t(u64, KEY_SIZE_MAX,
(next.inode == iter->pos.inode
@ -2726,13 +2854,13 @@ void bch2_trans_iter_exit(struct btree_trans *trans, struct btree_iter *iter)
{
if (iter->update_path)
bch2_path_put_nokeep(trans, iter->update_path,
iter->flags & BTREE_ITER_INTENT);
iter->flags & BTREE_ITER_intent);
if (iter->path)
bch2_path_put(trans, iter->path,
iter->flags & BTREE_ITER_INTENT);
iter->flags & BTREE_ITER_intent);
if (iter->key_cache_path)
bch2_path_put(trans, iter->key_cache_path,
iter->flags & BTREE_ITER_INTENT);
iter->flags & BTREE_ITER_intent);
iter->path = 0;
iter->update_path = 0;
iter->key_cache_path = 0;
@ -2757,9 +2885,9 @@ void bch2_trans_node_iter_init(struct btree_trans *trans,
unsigned depth,
unsigned flags)
{
flags |= BTREE_ITER_NOT_EXTENTS;
flags |= __BTREE_ITER_ALL_SNAPSHOTS;
flags |= BTREE_ITER_ALL_SNAPSHOTS;
flags |= BTREE_ITER_not_extents;
flags |= BTREE_ITER_snapshot_field;
flags |= BTREE_ITER_all_snapshots;
bch2_trans_iter_init_common(trans, iter, btree_id, pos, locks_want, depth,
__bch2_btree_iter_flags(trans, btree_id, flags),
@ -2782,9 +2910,9 @@ void bch2_trans_copy_iter(struct btree_iter *dst, struct btree_iter *src)
dst->ip_allocated = _RET_IP_;
#endif
if (src->path)
__btree_path_get(trans->paths + src->path, src->flags & BTREE_ITER_INTENT);
__btree_path_get(trans->paths + src->path, src->flags & BTREE_ITER_intent);
if (src->update_path)
__btree_path_get(trans->paths + src->update_path, src->flags & BTREE_ITER_INTENT);
__btree_path_get(trans->paths + src->update_path, src->flags & BTREE_ITER_intent);
dst->key_cache_path = 0;
}
@ -2953,7 +3081,8 @@ u32 bch2_trans_begin(struct btree_trans *trans)
if (!trans->restarted &&
(need_resched() ||
time_after64(now, trans->last_begin_time + BTREE_TRANS_MAX_LOCK_HOLD_TIME_NS))) {
drop_locks_do(trans, (cond_resched(), 0));
bch2_trans_unlock(trans);
cond_resched();
now = local_clock();
}
trans->last_begin_time = now;
@ -2963,11 +3092,14 @@ u32 bch2_trans_begin(struct btree_trans *trans)
bch2_trans_srcu_unlock(trans);
trans->last_begin_ip = _RET_IP_;
trans->locked = true;
if (trans->restarted) {
bch2_btree_path_traverse_all(trans);
trans->notrace_relock_fail = false;
}
bch2_trans_verify_not_unlocked(trans);
return trans->restart_count;
}
@ -3020,7 +3152,7 @@ struct btree_trans *__bch2_trans_get(struct bch_fs *c, unsigned fn_idx)
*/
BUG_ON(pos_task &&
pid == pos_task->pid &&
bch2_trans_locked(pos));
pos->locked);
if (pos_task && pid < pos_task->pid) {
list_add_tail(&trans->list, &pos->list);
@ -3036,8 +3168,9 @@ got_trans:
trans->last_begin_time = local_clock();
trans->fn_idx = fn_idx;
trans->locking_wait.task = current;
trans->locked = true;
trans->journal_replay_not_finished =
unlikely(!test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags)) &&
unlikely(!test_bit(JOURNAL_replay_done, &c->journal.flags)) &&
atomic_inc_not_zero(&c->journal_keys.ref);
trans->nr_paths = ARRAY_SIZE(trans->_paths);
trans->paths_allocated = trans->_paths_allocated;
@ -3166,13 +3299,11 @@ bch2_btree_bkey_cached_common_to_text(struct printbuf *out,
pid = owner ? owner->pid : 0;
rcu_read_unlock();
prt_tab(out);
prt_printf(out, "%px %c l=%u %s:", b, b->cached ? 'c' : 'b',
prt_printf(out, "\t%px %c l=%u %s:", b, b->cached ? 'c' : 'b',
b->level, bch2_btree_id_str(b->btree_id));
bch2_bpos_to_text(out, btree_node_pos(b));
prt_tab(out);
prt_printf(out, " locks %u:%u:%u held by pid %u",
prt_printf(out, "\t locks %u:%u:%u held by pid %u",
c.n[0], c.n[1], c.n[2], pid);
}
@ -3229,10 +3360,8 @@ void bch2_btree_trans_to_text(struct printbuf *out, struct btree_trans *trans)
b = READ_ONCE(trans->locking);
if (b) {
prt_printf(out, " blocked for %lluus on",
div_u64(local_clock() - trans->locking_wait.start_time,
1000));
prt_newline(out);
prt_printf(out, " blocked for %lluus on\n",
div_u64(local_clock() - trans->locking_wait.start_time, 1000));
prt_printf(out, " %c", lock_types[trans->locking_wait.lock_want]);
bch2_btree_bkey_cached_common_to_text(out, b);
prt_newline(out);

95
fs/bcachefs/btree_iter.h
View File

@ -216,9 +216,13 @@ int __must_check bch2_btree_path_traverse_one(struct btree_trans *,
btree_path_idx_t,
unsigned, unsigned long);
static inline void bch2_trans_verify_not_unlocked(struct btree_trans *);
static inline int __must_check bch2_btree_path_traverse(struct btree_trans *trans,
btree_path_idx_t path, unsigned flags)
{
bch2_trans_verify_not_unlocked(trans);
if (trans->paths[path].uptodate < BTREE_ITER_NEED_RELOCK)
return 0;
@ -227,6 +231,9 @@ static inline int __must_check bch2_btree_path_traverse(struct btree_trans *tran
btree_path_idx_t bch2_path_get(struct btree_trans *, enum btree_id, struct bpos,
unsigned, unsigned, unsigned, unsigned long);
btree_path_idx_t bch2_path_get_unlocked_mut(struct btree_trans *, enum btree_id,
unsigned, struct bpos);
struct bkey_s_c bch2_btree_path_peek_slot(struct btree_path *, struct bkey *);
/*
@ -283,7 +290,6 @@ int bch2_trans_relock(struct btree_trans *);
int bch2_trans_relock_notrace(struct btree_trans *);
void bch2_trans_unlock(struct btree_trans *);
void bch2_trans_unlock_long(struct btree_trans *);
bool bch2_trans_locked(struct btree_trans *);
static inline int trans_was_restarted(struct btree_trans *trans, u32 restart_count)
{
@ -309,6 +315,14 @@ static inline void bch2_trans_verify_not_in_restart(struct btree_trans *trans)
bch2_trans_in_restart_error(trans);
}
void __noreturn bch2_trans_unlocked_error(struct btree_trans *);
static inline void bch2_trans_verify_not_unlocked(struct btree_trans *trans)
{
if (!trans->locked)
bch2_trans_unlocked_error(trans);
}
__always_inline
static int btree_trans_restart_nounlock(struct btree_trans *trans, int err)
{
@ -386,10 +400,10 @@ static inline void bch2_btree_iter_set_pos(struct btree_iter *iter, struct bpos
if (unlikely(iter->update_path))
bch2_path_put(trans, iter->update_path,
iter->flags & BTREE_ITER_INTENT);
iter->flags & BTREE_ITER_intent);
iter->update_path = 0;
if (!(iter->flags & BTREE_ITER_ALL_SNAPSHOTS))
if (!(iter->flags & BTREE_ITER_all_snapshots))
new_pos.snapshot = iter->snapshot;
__bch2_btree_iter_set_pos(iter, new_pos);
@ -397,7 +411,7 @@ static inline void bch2_btree_iter_set_pos(struct btree_iter *iter, struct bpos
static inline void bch2_btree_iter_set_pos_to_extent_start(struct btree_iter *iter)
{
BUG_ON(!(iter->flags & BTREE_ITER_IS_EXTENTS));
BUG_ON(!(iter->flags & BTREE_ITER_is_extents));
iter->pos = bkey_start_pos(&iter->k);
}
@ -416,20 +430,20 @@ static inline unsigned __bch2_btree_iter_flags(struct btree_trans *trans,
unsigned btree_id,
unsigned flags)
{
if (!(flags & (BTREE_ITER_ALL_SNAPSHOTS|BTREE_ITER_NOT_EXTENTS)) &&
if (!(flags & (BTREE_ITER_all_snapshots|BTREE_ITER_not_extents)) &&
btree_id_is_extents(btree_id))
flags |= BTREE_ITER_IS_EXTENTS;
flags |= BTREE_ITER_is_extents;
if (!(flags & __BTREE_ITER_ALL_SNAPSHOTS) &&
if (!(flags & BTREE_ITER_snapshot_field) &&
!btree_type_has_snapshot_field(btree_id))
flags &= ~BTREE_ITER_ALL_SNAPSHOTS;
flags &= ~BTREE_ITER_all_snapshots;
if (!(flags & BTREE_ITER_ALL_SNAPSHOTS) &&
if (!(flags & BTREE_ITER_all_snapshots) &&
btree_type_has_snapshots(btree_id))
flags |= BTREE_ITER_FILTER_SNAPSHOTS;
flags |= BTREE_ITER_filter_snapshots;
if (trans->journal_replay_not_finished)
flags |= BTREE_ITER_WITH_JOURNAL;
flags |= BTREE_ITER_with_journal;
return flags;
}
@ -439,10 +453,10 @@ static inline unsigned bch2_btree_iter_flags(struct btree_trans *trans,
unsigned flags)
{
if (!btree_id_cached(trans->c, btree_id)) {
flags &= ~BTREE_ITER_CACHED;
flags &= ~BTREE_ITER_WITH_KEY_CACHE;
} else if (!(flags & BTREE_ITER_CACHED))
flags |= BTREE_ITER_WITH_KEY_CACHE;
flags &= ~BTREE_ITER_cached;
flags &= ~BTREE_ITER_with_key_cache;
} else if (!(flags & BTREE_ITER_cached))
flags |= BTREE_ITER_with_key_cache;
return __bch2_btree_iter_flags(trans, btree_id, flags);
}
@ -494,18 +508,7 @@ void bch2_trans_node_iter_init(struct btree_trans *, struct btree_iter *,
unsigned, unsigned, unsigned);
void bch2_trans_copy_iter(struct btree_iter *, struct btree_iter *);
static inline void set_btree_iter_dontneed(struct btree_iter *iter)
{
struct btree_trans *trans = iter->trans;
if (!iter->path || trans->restarted)
return;
struct btree_path *path = btree_iter_path(trans, iter);
path->preserve = false;
if (path->ref == 1)
path->should_be_locked = false;
}
void bch2_set_btree_iter_dontneed(struct btree_iter *);
void *__bch2_trans_kmalloc(struct btree_trans *, size_t);
@ -619,14 +622,14 @@ u32 bch2_trans_begin(struct btree_trans *);
static inline struct bkey_s_c bch2_btree_iter_peek_prev_type(struct btree_iter *iter,
unsigned flags)
{
return flags & BTREE_ITER_SLOTS ? bch2_btree_iter_peek_slot(iter) :
return flags & BTREE_ITER_slots ? bch2_btree_iter_peek_slot(iter) :
bch2_btree_iter_peek_prev(iter);
}
static inline struct bkey_s_c bch2_btree_iter_peek_type(struct btree_iter *iter,
unsigned flags)
{
return flags & BTREE_ITER_SLOTS ? bch2_btree_iter_peek_slot(iter) :
return flags & BTREE_ITER_slots ? bch2_btree_iter_peek_slot(iter) :
bch2_btree_iter_peek(iter);
}
@ -634,7 +637,7 @@ static inline struct bkey_s_c bch2_btree_iter_peek_upto_type(struct btree_iter *
struct bpos end,
unsigned flags)
{
if (!(flags & BTREE_ITER_SLOTS))
if (!(flags & BTREE_ITER_slots))
return bch2_btree_iter_peek_upto(iter, end);
if (bkey_gt(iter->pos, end))
@ -699,16 +702,12 @@ transaction_restart: \
_ret2 ?: trans_was_restarted(_trans, _restart_count); \
})
#define for_each_btree_key_upto(_trans, _iter, _btree_id, \
_start, _end, _flags, _k, _do) \
#define for_each_btree_key_upto_continue(_trans, _iter, \
_end, _flags, _k, _do) \
({ \
struct btree_iter _iter; \
struct bkey_s_c _k; \
int _ret3 = 0; \
\
bch2_trans_iter_init((_trans), &(_iter), (_btree_id), \
(_start), (_flags)); \
\
do { \
_ret3 = lockrestart_do(_trans, ({ \
(_k) = bch2_btree_iter_peek_upto_type(&(_iter), \
@ -724,6 +723,21 @@ transaction_restart: \
_ret3; \
})
#define for_each_btree_key_continue(_trans, _iter, _flags, _k, _do) \
for_each_btree_key_upto_continue(_trans, _iter, SPOS_MAX, _flags, _k, _do)
#define for_each_btree_key_upto(_trans, _iter, _btree_id, \
_start, _end, _flags, _k, _do) \
({ \
bch2_trans_begin(trans); \
\
struct btree_iter _iter; \
bch2_trans_iter_init((_trans), &(_iter), (_btree_id), \
(_start), (_flags)); \
\
for_each_btree_key_upto_continue(_trans, _iter, _end, _flags, _k, _do);\
})
#define for_each_btree_key(_trans, _iter, _btree_id, \
_start, _flags, _k, _do) \
for_each_btree_key_upto(_trans, _iter, _btree_id, _start, \
@ -794,14 +808,6 @@ __bch2_btree_iter_peek_and_restart(struct btree_trans *trans,
return k;
}
#define for_each_btree_key_old(_trans, _iter, _btree_id, \
_start, _flags, _k, _ret) \
for (bch2_trans_iter_init((_trans), &(_iter), (_btree_id), \
(_start), (_flags)); \
(_k) = __bch2_btree_iter_peek_and_restart((_trans), &(_iter), _flags),\
!((_ret) = bkey_err(_k)) && (_k).k; \
bch2_btree_iter_advance(&(_iter)))
#define for_each_btree_key_upto_norestart(_trans, _iter, _btree_id, \
_start, _end, _flags, _k, _ret) \
for (bch2_trans_iter_init((_trans), &(_iter), (_btree_id), \
@ -861,6 +867,7 @@ __bch2_btree_iter_peek_and_restart(struct btree_trans *trans,
})
void bch2_trans_updates_to_text(struct printbuf *, struct btree_trans *);
void bch2_btree_path_to_text(struct printbuf *, struct btree_trans *, btree_path_idx_t);
void bch2_trans_paths_to_text(struct printbuf *, struct btree_trans *);
void bch2_dump_trans_updates(struct btree_trans *);
void bch2_dump_trans_paths_updates(struct btree_trans *);

17
fs/bcachefs/btree_journal_iter.c
View File

@ -623,3 +623,20 @@ void bch2_shoot_down_journal_keys(struct bch_fs *c, enum btree_id btree,
keys->data[dst++] = *i;
keys->nr = keys->gap = dst;
}
void bch2_journal_keys_dump(struct bch_fs *c)
{
struct journal_keys *keys = &c->journal_keys;
struct printbuf buf = PRINTBUF;
pr_info("%zu keys:", keys->nr);
move_gap(keys, keys->nr);
darray_for_each(*keys, i) {
printbuf_reset(&buf);
bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(i->k));
pr_err("%s l=%u %s", bch2_btree_id_str(i->btree_id), i->level, buf.buf);
}
printbuf_exit(&buf);
}

2
fs/bcachefs/btree_journal_iter.h
View File

@ -70,4 +70,6 @@ void bch2_shoot_down_journal_keys(struct bch_fs *, enum btree_id,
unsigned, unsigned,
struct bpos, struct bpos);
void bch2_journal_keys_dump(struct bch_fs *);
#endif /* _BCACHEFS_BTREE_JOURNAL_ITER_H */

107
fs/bcachefs/btree_key_cache.c
View File

@ -383,9 +383,9 @@ static int btree_key_cache_fill(struct btree_trans *trans,
int ret;
bch2_trans_iter_init(trans, &iter, ck->key.btree_id, ck->key.pos,
BTREE_ITER_KEY_CACHE_FILL|
BTREE_ITER_CACHED_NOFILL);
iter.flags &= ~BTREE_ITER_WITH_JOURNAL;
BTREE_ITER_key_cache_fill|
BTREE_ITER_cached_nofill);
iter.flags &= ~BTREE_ITER_with_journal;
k = bch2_btree_iter_peek_slot(&iter);
ret = bkey_err(k);
if (ret)
@ -456,7 +456,7 @@ static int btree_key_cache_fill(struct btree_trans *trans,
bch2_btree_node_unlock_write(trans, ck_path, ck_path->l[0].b);
/* We're not likely to need this iterator again: */
set_btree_iter_dontneed(&iter);
bch2_set_btree_iter_dontneed(&iter);
err:
bch2_trans_iter_exit(trans, &iter);
return ret;
@ -515,23 +515,10 @@ retry:
fill:
path->uptodate = BTREE_ITER_UPTODATE;
if (!ck->valid && !(flags & BTREE_ITER_CACHED_NOFILL)) {
/*
* Using the underscore version because we haven't set
* path->uptodate yet:
*/
if (!path->locks_want &&
!__bch2_btree_path_upgrade(trans, path, 1, NULL)) {
trace_and_count(trans->c, trans_restart_key_cache_upgrade, trans, _THIS_IP_);
ret = btree_trans_restart(trans, BCH_ERR_transaction_restart_key_cache_upgrade);
goto err;
}
ret = btree_key_cache_fill(trans, path, ck);
if (ret)
goto err;
ret = bch2_btree_path_relock(trans, path, _THIS_IP_);
if (!ck->valid && !(flags & BTREE_ITER_cached_nofill)) {
ret = bch2_btree_path_upgrade(trans, path, 1) ?:
btree_key_cache_fill(trans, path, ck) ?:
bch2_btree_path_relock(trans, path, _THIS_IP_);
if (ret)
goto err;
@ -622,13 +609,13 @@ static int btree_key_cache_flush_pos(struct btree_trans *trans,
int ret;
bch2_trans_iter_init(trans, &b_iter, key.btree_id, key.pos,
BTREE_ITER_SLOTS|
BTREE_ITER_INTENT|
BTREE_ITER_ALL_SNAPSHOTS);
BTREE_ITER_slots|
BTREE_ITER_intent|
BTREE_ITER_all_snapshots);
bch2_trans_iter_init(trans, &c_iter, key.btree_id, key.pos,
BTREE_ITER_CACHED|
BTREE_ITER_INTENT);
b_iter.flags &= ~BTREE_ITER_WITH_KEY_CACHE;
BTREE_ITER_cached|
BTREE_ITER_intent);
b_iter.flags &= ~BTREE_ITER_with_key_cache;
ret = bch2_btree_iter_traverse(&c_iter);
if (ret)
@ -661,14 +648,14 @@ static int btree_key_cache_flush_pos(struct btree_trans *trans,
commit_flags |= BCH_WATERMARK_reclaim;
if (ck->journal.seq != journal_last_seq(j) ||
!test_bit(JOURNAL_SPACE_LOW, &c->journal.flags))
!test_bit(JOURNAL_space_low, &c->journal.flags))
commit_flags |= BCH_TRANS_COMMIT_no_journal_res;
ret = bch2_btree_iter_traverse(&b_iter) ?:
bch2_trans_update(trans, &b_iter, ck->k,
BTREE_UPDATE_KEY_CACHE_RECLAIM|
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|
BTREE_TRIGGER_NORUN) ?:
BTREE_UPDATE_key_cache_reclaim|
BTREE_UPDATE_internal_snapshot_node|
BTREE_TRIGGER_norun) ?:
bch2_trans_commit(trans, NULL, NULL,
BCH_TRANS_COMMIT_no_check_rw|
BCH_TRANS_COMMIT_no_enospc|
@ -790,7 +777,7 @@ bool bch2_btree_insert_key_cached(struct btree_trans *trans,
* flushing. The flush callback will not proceed unless ->seq matches
* the latest pin, so make sure it starts with a consistent value.
*/
if (!(insert_entry->flags & BTREE_UPDATE_NOJOURNAL) ||
if (!(insert_entry->flags & BTREE_UPDATE_nojournal) ||
!journal_pin_active(&ck->journal)) {
ck->seq = trans->journal_res.seq;
}
@ -835,6 +822,8 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
int srcu_idx;
mutex_lock(&bc->lock);
bc->requested_to_free += sc->nr_to_scan;
srcu_idx = srcu_read_lock(&c->btree_trans_barrier);
flags = memalloc_nofs_save();
@ -853,6 +842,7 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
atomic_long_dec(&bc->nr_freed);
freed++;
bc->nr_freed_nonpcpu--;
bc->freed++;
}
list_for_each_entry_safe(ck, t, &bc->freed_pcpu, list) {
@ -866,6 +856,7 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
atomic_long_dec(&bc->nr_freed);
freed++;
bc->nr_freed_pcpu--;
bc->freed++;
}
rcu_read_lock();
@ -884,13 +875,18 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
ck = container_of(pos, struct bkey_cached, hash);
if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
bc->skipped_dirty++;
goto next;
} else if (test_bit(BKEY_CACHED_ACCESSED, &ck->flags)) {
clear_bit(BKEY_CACHED_ACCESSED, &ck->flags);
bc->skipped_accessed++;
goto next;
} else if (bkey_cached_lock_for_evict(ck)) {
bkey_cached_evict(bc, ck);
bkey_cached_free(bc, ck);
bc->moved_to_freelist++;
} else {
bc->skipped_lock_fail++;
}
scanned++;
@ -1037,14 +1033,47 @@ int bch2_fs_btree_key_cache_init(struct btree_key_cache *bc)
return 0;
}
void bch2_btree_key_cache_to_text(struct printbuf *out, struct btree_key_cache *c)
void bch2_btree_key_cache_to_text(struct printbuf *out, struct btree_key_cache *bc)
{
prt_printf(out, "nr_freed:\t%lu", atomic_long_read(&c->nr_freed));
prt_newline(out);
prt_printf(out, "nr_keys:\t%lu", atomic_long_read(&c->nr_keys));
prt_newline(out);
prt_printf(out, "nr_dirty:\t%lu", atomic_long_read(&c->nr_dirty));
prt_newline(out);
struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
printbuf_tabstop_push(out, 24);
printbuf_tabstop_push(out, 12);
unsigned flags = memalloc_nofs_save();
mutex_lock(&bc->lock);
prt_printf(out, "keys:\t%lu\r\n", atomic_long_read(&bc->nr_keys));
prt_printf(out, "dirty:\t%lu\r\n", atomic_long_read(&bc->nr_dirty));
prt_printf(out, "freelist:\t%lu\r\n", atomic_long_read(&bc->nr_freed));
prt_printf(out, "nonpcpu freelist:\t%zu\r\n", bc->nr_freed_nonpcpu);
prt_printf(out, "pcpu freelist:\t%zu\r\n", bc->nr_freed_pcpu);
prt_printf(out, "\nshrinker:\n");
prt_printf(out, "requested_to_free:\t%lu\r\n", bc->requested_to_free);
prt_printf(out, "freed:\t%lu\r\n", bc->freed);
prt_printf(out, "moved_to_freelist:\t%lu\r\n", bc->moved_to_freelist);
prt_printf(out, "skipped_dirty:\t%lu\r\n", bc->skipped_dirty);
prt_printf(out, "skipped_accessed:\t%lu\r\n", bc->skipped_accessed);
prt_printf(out, "skipped_lock_fail:\t%lu\r\n", bc->skipped_lock_fail);
prt_printf(out, "srcu seq:\t%lu\r\n", get_state_synchronize_srcu(&c->btree_trans_barrier));
struct bkey_cached *ck;
unsigned iter = 0;
list_for_each_entry(ck, &bc->freed_nonpcpu, list) {
prt_printf(out, "freed_nonpcpu:\t%lu\r\n", ck->btree_trans_barrier_seq);
if (++iter > 10)
break;
}
iter = 0;
list_for_each_entry(ck, &bc->freed_pcpu, list) {
prt_printf(out, "freed_pcpu:\t%lu\r\n", ck->btree_trans_barrier_seq);
if (++iter > 10)
break;
}
mutex_unlock(&bc->lock);
memalloc_flags_restore(flags);
}
void bch2_btree_key_cache_exit(void)

8
fs/bcachefs/btree_key_cache_types.h
View File

@ -24,6 +24,14 @@ struct btree_key_cache {
atomic_long_t nr_freed;
atomic_long_t nr_keys;
atomic_long_t nr_dirty;
/* shrinker stats */
unsigned long requested_to_free;
unsigned long freed;
unsigned long moved_to_freelist;
unsigned long skipped_dirty;
unsigned long skipped_accessed;
unsigned long skipped_lock_fail;
};
struct bkey_cached_key {

183
fs/bcachefs/btree_locking.c
View File

@ -83,8 +83,7 @@ static noinline void print_cycle(struct printbuf *out, struct lock_graph *g)
{
struct trans_waiting_for_lock *i;
prt_printf(out, "Found lock cycle (%u entries):", g->nr);
prt_newline(out);
prt_printf(out, "Found lock cycle (%u entries):\n", g->nr);
for (i = g->g; i < g->g + g->nr; i++) {
struct task_struct *task = READ_ONCE(i->trans->locking_wait.task);
@ -224,8 +223,7 @@ static noinline int break_cycle(struct lock_graph *g, struct printbuf *cycle)
bch2_btree_trans_to_text(&buf, trans);
prt_printf(&buf, "backtrace:");
prt_newline(&buf);
prt_printf(&buf, "backtrace:\n");
printbuf_indent_add(&buf, 2);
bch2_prt_task_backtrace(&buf, trans->locking_wait.task, 2, GFP_NOWAIT);
printbuf_indent_sub(&buf, 2);
@ -492,8 +490,6 @@ static inline bool btree_path_get_locks(struct btree_trans *trans,
if (path->uptodate == BTREE_ITER_NEED_RELOCK)
path->uptodate = BTREE_ITER_UPTODATE;
bch2_trans_verify_locks(trans);
return path->uptodate < BTREE_ITER_NEED_RELOCK;
}
@ -609,7 +605,9 @@ bool bch2_btree_path_relock_norestart(struct btree_trans *trans, struct btree_pa
{
struct get_locks_fail f;
return btree_path_get_locks(trans, path, false, &f);
bool ret = btree_path_get_locks(trans, path, false, &f);
bch2_trans_verify_locks(trans);
return ret;
}
int __bch2_btree_path_relock(struct btree_trans *trans,
@ -632,7 +630,9 @@ bool bch2_btree_path_upgrade_noupgrade_sibs(struct btree_trans *trans,
path->locks_want = new_locks_want;
return btree_path_get_locks(trans, path, true, f);
bool ret = btree_path_get_locks(trans, path, true, f);
bch2_trans_verify_locks(trans);
return ret;
}
bool __bch2_btree_path_upgrade(struct btree_trans *trans,
@ -640,8 +640,9 @@ bool __bch2_btree_path_upgrade(struct btree_trans *trans,
unsigned new_locks_want,
struct get_locks_fail *f)
{
if (bch2_btree_path_upgrade_noupgrade_sibs(trans, path, new_locks_want, f))
return true;
bool ret = bch2_btree_path_upgrade_noupgrade_sibs(trans, path, new_locks_want, f);
if (ret)
goto out;
/*
* XXX: this is ugly - we'd prefer to not be mucking with other
@ -675,8 +676,9 @@ bool __bch2_btree_path_upgrade(struct btree_trans *trans,
btree_path_get_locks(trans, linked, true, NULL);
}
}
return false;
out:
bch2_trans_verify_locks(trans);
return ret;
}
void __bch2_btree_path_downgrade(struct btree_trans *trans,
@ -725,82 +727,100 @@ void bch2_trans_downgrade(struct btree_trans *trans)
bch2_btree_path_downgrade(trans, path);
}
int bch2_trans_relock(struct btree_trans *trans)
static inline void __bch2_trans_unlock(struct btree_trans *trans)
{
struct btree_path *path;
unsigned i;
trans_for_each_path(trans, path, i)
__bch2_btree_path_unlock(trans, path);
}
static noinline __cold int bch2_trans_relock_fail(struct btree_trans *trans, struct btree_path *path,
struct get_locks_fail *f, bool trace)
{
if (!trace)
goto out;
if (trace_trans_restart_relock_enabled()) {
struct printbuf buf = PRINTBUF;
bch2_bpos_to_text(&buf, path->pos);
prt_printf(&buf, " l=%u seq=%u node seq=", f->l, path->l[f->l].lock_seq);
if (IS_ERR_OR_NULL(f->b)) {
prt_str(&buf, bch2_err_str(PTR_ERR(f->b)));
} else {
prt_printf(&buf, "%u", f->b->c.lock.seq);
struct six_lock_count c =
bch2_btree_node_lock_counts(trans, NULL, &f->b->c, f->l);
prt_printf(&buf, " self locked %u.%u.%u", c.n[0], c.n[1], c.n[2]);
c = six_lock_counts(&f->b->c.lock);
prt_printf(&buf, " total locked %u.%u.%u", c.n[0], c.n[1], c.n[2]);
}
trace_trans_restart_relock(trans, _RET_IP_, buf.buf);
printbuf_exit(&buf);
}
count_event(trans->c, trans_restart_relock);
out:
__bch2_trans_unlock(trans);
bch2_trans_verify_locks(trans);
return btree_trans_restart(trans, BCH_ERR_transaction_restart_relock);
}
static inline int __bch2_trans_relock(struct btree_trans *trans, bool trace)
{
bch2_trans_verify_locks(trans);
if (unlikely(trans->restarted))
return -((int) trans->restarted);
if (unlikely(trans->locked))
goto out;
struct btree_path *path;
unsigned i;
trans_for_each_path(trans, path, i) {
struct get_locks_fail f;
if (path->should_be_locked &&
!btree_path_get_locks(trans, path, false, &f)) {
if (trace_trans_restart_relock_enabled()) {
struct printbuf buf = PRINTBUF;
bch2_bpos_to_text(&buf, path->pos);
prt_printf(&buf, " l=%u seq=%u node seq=",
f.l, path->l[f.l].lock_seq);
if (IS_ERR_OR_NULL(f.b)) {
prt_str(&buf, bch2_err_str(PTR_ERR(f.b)));
} else {
prt_printf(&buf, "%u", f.b->c.lock.seq);
struct six_lock_count c =
bch2_btree_node_lock_counts(trans, NULL, &f.b->c, f.l);
prt_printf(&buf, " self locked %u.%u.%u", c.n[0], c.n[1], c.n[2]);
c = six_lock_counts(&f.b->c.lock);
prt_printf(&buf, " total locked %u.%u.%u", c.n[0], c.n[1], c.n[2]);
}
trace_trans_restart_relock(trans, _RET_IP_, buf.buf);
printbuf_exit(&buf);
}
count_event(trans->c, trans_restart_relock);
return btree_trans_restart(trans, BCH_ERR_transaction_restart_relock);
}
!btree_path_get_locks(trans, path, false, &f))
return bch2_trans_relock_fail(trans, path, &f, trace);
}
trans->locked = true;
out:
bch2_trans_verify_locks(trans);
return 0;
}
int bch2_trans_relock(struct btree_trans *trans)
{
return __bch2_trans_relock(trans, true);
}
int bch2_trans_relock_notrace(struct btree_trans *trans)
{
struct btree_path *path;
unsigned i;
if (unlikely(trans->restarted))
return -((int) trans->restarted);
trans_for_each_path(trans, path, i)
if (path->should_be_locked &&
!bch2_btree_path_relock_norestart(trans, path)) {
return btree_trans_restart(trans, BCH_ERR_transaction_restart_relock);
}
return 0;
return __bch2_trans_relock(trans, false);
}
void bch2_trans_unlock_noassert(struct btree_trans *trans)
{
struct btree_path *path;
unsigned i;
__bch2_trans_unlock(trans);
trans_for_each_path(trans, path, i)
__bch2_btree_path_unlock(trans, path);
trans->locked = false;
trans->last_unlock_ip = _RET_IP_;
}
void bch2_trans_unlock(struct btree_trans *trans)
{
struct btree_path *path;
unsigned i;
__bch2_trans_unlock(trans);
trans_for_each_path(trans, path, i)
__bch2_btree_path_unlock(trans, path);
trans->locked = false;
trans->last_unlock_ip = _RET_IP_;
}
void bch2_trans_unlock_long(struct btree_trans *trans)
@ -809,17 +829,6 @@ void bch2_trans_unlock_long(struct btree_trans *trans)
bch2_trans_srcu_unlock(trans);
}
bool bch2_trans_locked(struct btree_trans *trans)
{
struct btree_path *path;
unsigned i;
trans_for_each_path(trans, path, i)
if (path->nodes_locked)
return true;
return false;
}
int __bch2_trans_mutex_lock(struct btree_trans *trans,
struct mutex *lock)
{
@ -836,15 +845,19 @@ int __bch2_trans_mutex_lock(struct btree_trans *trans,
void bch2_btree_path_verify_locks(struct btree_path *path)
{
unsigned l;
/*
* A path may be uptodate and yet have nothing locked if and only if
* there is no node at path->level, which generally means we were
* iterating over all nodes and got to the end of the btree
*/
BUG_ON(path->uptodate == BTREE_ITER_UPTODATE &&
btree_path_node(path, path->level) &&
!path->nodes_locked);
if (!path->nodes_locked) {
BUG_ON(path->uptodate == BTREE_ITER_UPTODATE &&
btree_path_node(path, path->level));
if (!path->nodes_locked)
return;
}
for (l = 0; l < BTREE_MAX_DEPTH; l++) {
for (unsigned l = 0; l < BTREE_MAX_DEPTH; l++) {
int want = btree_lock_want(path, l);
int have = btree_node_locked_type(path, l);
@ -857,8 +870,24 @@ void bch2_btree_path_verify_locks(struct btree_path *path)
}
}
static bool bch2_trans_locked(struct btree_trans *trans)
{
struct btree_path *path;
unsigned i;
trans_for_each_path(trans, path, i)
if (path->nodes_locked)
return true;
return false;
}
void bch2_trans_verify_locks(struct btree_trans *trans)
{
if (!trans->locked) {
BUG_ON(bch2_trans_locked(trans));
return;
}
struct btree_path *path;
unsigned i;

4
fs/bcachefs/btree_locking.h
View File

@ -364,14 +364,14 @@ static inline int bch2_btree_path_upgrade(struct btree_trans *trans,
struct btree_path *path,
unsigned new_locks_want)
{
struct get_locks_fail f;
struct get_locks_fail f = {};
unsigned old_locks_want = path->locks_want;
new_locks_want = min(new_locks_want, BTREE_MAX_DEPTH);
if (path->locks_want < new_locks_want
? __bch2_btree_path_upgrade(trans, path, new_locks_want, &f)
: path->uptodate == BTREE_ITER_UPTODATE)
: path->nodes_locked)
return 0;
trace_and_count(trans->c, trans_restart_upgrade, trans, _THIS_IP_, path,

70
fs/bcachefs/btree_trans_commit.c
View File

@ -1,6 +1,7 @@
// SPDX-License-Identifier: GPL-2.0
#include "bcachefs.h"
#include "alloc_foreground.h"
#include "btree_gc.h"
#include "btree_io.h"
#include "btree_iter.h"
@ -19,6 +20,26 @@
#include <linux/prefetch.h>
static const char * const trans_commit_flags_strs[] = {
#define x(n, ...) #n,
BCH_TRANS_COMMIT_FLAGS()
#undef x
NULL
};
void bch2_trans_commit_flags_to_text(struct printbuf *out, enum bch_trans_commit_flags flags)
{
enum bch_watermark watermark = flags & BCH_WATERMARK_MASK;
prt_printf(out, "watermark=%s", bch2_watermarks[watermark]);
flags >>= BCH_WATERMARK_BITS;
if (flags) {
prt_char(out, ' ');
bch2_prt_bitflags(out, trans_commit_flags_strs, flags);
}
}
static void verify_update_old_key(struct btree_trans *trans, struct btree_insert_entry *i)
{
#ifdef CONFIG_BCACHEFS_DEBUG
@ -315,8 +336,8 @@ static inline void btree_insert_entry_checks(struct btree_trans *trans,
BUG_ON(i->btree_id != path->btree_id);
EBUG_ON(!i->level &&
btree_type_has_snapshots(i->btree_id) &&
!(i->flags & BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) &&
test_bit(JOURNAL_REPLAY_DONE, &trans->c->journal.flags) &&
!(i->flags & BTREE_UPDATE_internal_snapshot_node) &&
test_bit(JOURNAL_replay_done, &trans->c->journal.flags) &&
i->k->k.p.snapshot &&
bch2_snapshot_is_internal_node(trans->c, i->k->k.p.snapshot) > 0);
}
@ -443,13 +464,13 @@ static int run_one_mem_trigger(struct btree_trans *trans,
verify_update_old_key(trans, i);
if (unlikely(flags & BTREE_TRIGGER_NORUN))
if (unlikely(flags & BTREE_TRIGGER_norun))
return 0;
if (old_ops->trigger == new_ops->trigger) {
ret = bch2_key_trigger(trans, i->btree_id, i->level,
old, bkey_i_to_s(new),
BTREE_TRIGGER_INSERT|BTREE_TRIGGER_OVERWRITE|flags);
BTREE_TRIGGER_insert|BTREE_TRIGGER_overwrite|flags);
} else {
ret = bch2_key_trigger_new(trans, i->btree_id, i->level,
bkey_i_to_s(new), flags) ?:
@ -472,11 +493,11 @@ static int run_one_trans_trigger(struct btree_trans *trans, struct btree_insert_
struct bkey_s_c old = { &old_k, i->old_v };
const struct bkey_ops *old_ops = bch2_bkey_type_ops(old.k->type);
const struct bkey_ops *new_ops = bch2_bkey_type_ops(i->k->k.type);
unsigned flags = i->flags|BTREE_TRIGGER_TRANSACTIONAL;
unsigned flags = i->flags|BTREE_TRIGGER_transactional;
verify_update_old_key(trans, i);
if ((i->flags & BTREE_TRIGGER_NORUN) ||
if ((i->flags & BTREE_TRIGGER_norun) ||
!(BTREE_NODE_TYPE_HAS_TRANS_TRIGGERS & (1U << i->bkey_type)))
return 0;
@ -486,8 +507,8 @@ static int run_one_trans_trigger(struct btree_trans *trans, struct btree_insert_
i->overwrite_trigger_run = true;
i->insert_trigger_run = true;
return bch2_key_trigger(trans, i->btree_id, i->level, old, bkey_i_to_s(i->k),
BTREE_TRIGGER_INSERT|
BTREE_TRIGGER_OVERWRITE|flags) ?: 1;
BTREE_TRIGGER_insert|
BTREE_TRIGGER_overwrite|flags) ?: 1;
} else if (overwrite && !i->overwrite_trigger_run) {
i->overwrite_trigger_run = true;
return bch2_key_trigger_old(trans, i->btree_id, i->level, old, flags) ?: 1;
@ -572,7 +593,7 @@ static int bch2_trans_commit_run_triggers(struct btree_trans *trans)
#ifdef CONFIG_BCACHEFS_DEBUG
trans_for_each_update(trans, i)
BUG_ON(!(i->flags & BTREE_TRIGGER_NORUN) &&
BUG_ON(!(i->flags & BTREE_TRIGGER_norun) &&
(BTREE_NODE_TYPE_HAS_TRANS_TRIGGERS & (1U << i->bkey_type)) &&
(!i->insert_trigger_run || !i->overwrite_trigger_run));
#endif
@ -590,7 +611,7 @@ static noinline int bch2_trans_commit_run_gc_triggers(struct btree_trans *trans)
if (btree_node_type_needs_gc(__btree_node_type(i->level, i->btree_id)) &&
gc_visited(trans->c, gc_pos_btree_node(insert_l(trans, i)->b))) {
int ret = run_one_mem_trigger(trans, i, i->flags|BTREE_TRIGGER_GC);
int ret = run_one_mem_trigger(trans, i, i->flags|BTREE_TRIGGER_gc);
if (ret)
return ret;
}
@ -609,6 +630,9 @@ bch2_trans_commit_write_locked(struct btree_trans *trans, unsigned flags,
unsigned u64s = 0;
int ret;
bch2_trans_verify_not_unlocked(trans);
bch2_trans_verify_not_in_restart(trans);
if (race_fault()) {
trace_and_count(c, trans_restart_fault_inject, trans, trace_ip);
return btree_trans_restart_nounlock(trans, BCH_ERR_transaction_restart_fault_inject);
@ -686,7 +710,7 @@ bch2_trans_commit_write_locked(struct btree_trans *trans, unsigned flags,
trans_for_each_update(trans, i)
if (BTREE_NODE_TYPE_HAS_ATOMIC_TRIGGERS & (1U << i->bkey_type)) {
ret = run_one_mem_trigger(trans, i, BTREE_TRIGGER_ATOMIC|i->flags);
ret = run_one_mem_trigger(trans, i, BTREE_TRIGGER_atomic|i->flags);
if (ret)
goto fatal_err;
}
@ -705,7 +729,7 @@ bch2_trans_commit_write_locked(struct btree_trans *trans, unsigned flags,
if (i->key_cache_already_flushed)
continue;
if (i->flags & BTREE_UPDATE_NOJOURNAL)
if (i->flags & BTREE_UPDATE_nojournal)
continue;
verify_update_old_key(trans, i);
@ -766,16 +790,15 @@ static noinline void bch2_drop_overwrites_from_journal(struct btree_trans *trans
}
static noinline int bch2_trans_commit_bkey_invalid(struct btree_trans *trans,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct btree_insert_entry *i,
struct printbuf *err)
{
struct bch_fs *c = trans->c;
printbuf_reset(err);
prt_printf(err, "invalid bkey on insert from %s -> %ps",
prt_printf(err, "invalid bkey on insert from %s -> %ps\n",
trans->fn, (void *) i->ip_allocated);
prt_newline(err);
printbuf_indent_add(err, 2);
bch2_bkey_val_to_text(err, c, bkey_i_to_s_c(i->k));
@ -796,8 +819,7 @@ static noinline int bch2_trans_commit_journal_entry_invalid(struct btree_trans *
struct bch_fs *c = trans->c;
struct printbuf buf = PRINTBUF;
prt_printf(&buf, "invalid bkey on insert from %s", trans->fn);
prt_newline(&buf);
prt_printf(&buf, "invalid bkey on insert from %s\n", trans->fn);
printbuf_indent_add(&buf, 2);
bch2_journal_entry_to_text(&buf, c, i);
@ -988,6 +1010,9 @@ int __bch2_trans_commit(struct btree_trans *trans, unsigned flags)
struct bch_fs *c = trans->c;
int ret = 0;
bch2_trans_verify_not_unlocked(trans);
bch2_trans_verify_not_in_restart(trans);
if (!trans->nr_updates &&
!trans->journal_entries_u64s)
goto out_reset;
@ -1000,10 +1025,10 @@ int __bch2_trans_commit(struct btree_trans *trans, unsigned flags)
trans_for_each_update(trans, i) {
struct printbuf buf = PRINTBUF;
enum bkey_invalid_flags invalid_flags = 0;
enum bch_validate_flags invalid_flags = 0;
if (!(flags & BCH_TRANS_COMMIT_no_journal_res))
invalid_flags |= BKEY_INVALID_WRITE|BKEY_INVALID_COMMIT;
invalid_flags |= BCH_VALIDATE_write|BCH_VALIDATE_commit;
if (unlikely(bch2_bkey_invalid(c, bkey_i_to_s_c(i->k),
i->bkey_type, invalid_flags, &buf)))
@ -1018,10 +1043,10 @@ int __bch2_trans_commit(struct btree_trans *trans, unsigned flags)
for (struct jset_entry *i = trans->journal_entries;
i != (void *) ((u64 *) trans->journal_entries + trans->journal_entries_u64s);
i = vstruct_next(i)) {
enum bkey_invalid_flags invalid_flags = 0;
enum bch_validate_flags invalid_flags = 0;
if (!(flags & BCH_TRANS_COMMIT_no_journal_res))
invalid_flags |= BKEY_INVALID_WRITE|BKEY_INVALID_COMMIT;
invalid_flags |= BCH_VALIDATE_write|BCH_VALIDATE_commit;
if (unlikely(bch2_journal_entry_validate(c, NULL, i,
bcachefs_metadata_version_current,
@ -1065,7 +1090,7 @@ int __bch2_trans_commit(struct btree_trans *trans, unsigned flags)
if (i->key_cache_already_flushed)
continue;
if (i->flags & BTREE_UPDATE_NOJOURNAL)
if (i->flags & BTREE_UPDATE_nojournal)
continue;
/* we're going to journal the key being updated: */
@ -1086,6 +1111,7 @@ int __bch2_trans_commit(struct btree_trans *trans, unsigned flags)
}
retry:
errored_at = NULL;
bch2_trans_verify_not_unlocked(trans);
bch2_trans_verify_not_in_restart(trans);
if (likely(!(flags & BCH_TRANS_COMMIT_no_journal_res)))
memset(&trans->journal_res, 0, sizeof(trans->journal_res));

127
fs/bcachefs/btree_types.h
View File

@ -163,9 +163,21 @@ struct btree_cache {
/* Number of elements in live + freeable lists */
unsigned used;
unsigned reserve;
unsigned freed;
unsigned not_freed_lock_intent;
unsigned not_freed_lock_write;
unsigned not_freed_dirty;
unsigned not_freed_read_in_flight;
unsigned not_freed_write_in_flight;
unsigned not_freed_noevict;
unsigned not_freed_write_blocked;
unsigned not_freed_will_make_reachable;
unsigned not_freed_access_bit;
atomic_t dirty;
struct shrinker *shrink;
unsigned used_by_btree[BTREE_ID_NR];
/*
* If we need to allocate memory for a new btree node and that
* allocation fails, we can cannibalize another node in the btree cache
@ -187,36 +199,89 @@ struct btree_node_iter {
} data[MAX_BSETS];
};
#define BTREE_ITER_FLAGS() \
x(slots) \
x(intent) \
x(prefetch) \
x(is_extents) \
x(not_extents) \
x(cached) \
x(with_key_cache) \
x(with_updates) \
x(with_journal) \
x(snapshot_field) \
x(all_snapshots) \
x(filter_snapshots) \
x(nopreserve) \
x(cached_nofill) \
x(key_cache_fill) \
#define STR_HASH_FLAGS() \
x(must_create) \
x(must_replace)
#define BTREE_UPDATE_FLAGS() \
x(internal_snapshot_node) \
x(nojournal) \
x(key_cache_reclaim)
/*
* Iterate over all possible positions, synthesizing deleted keys for holes:
* BTREE_TRIGGER_norun - don't run triggers at all
*
* BTREE_TRIGGER_transactional - we're running transactional triggers as part of
* a transaction commit: triggers may generate new updates
*
* BTREE_TRIGGER_atomic - we're running atomic triggers during a transaction
* commit: we have our journal reservation, we're holding btree node write
* locks, and we know the transaction is going to commit (returning an error
* here is a fatal error, causing us to go emergency read-only)
*
* BTREE_TRIGGER_gc - we're in gc/fsck: running triggers to recalculate e.g. disk usage
*
* BTREE_TRIGGER_insert - @new is entering the btree
* BTREE_TRIGGER_overwrite - @old is leaving the btree
*
* BTREE_TRIGGER_bucket_invalidate - signal from bucket invalidate path to alloc
* trigger
*/
static const __maybe_unused u16 BTREE_ITER_SLOTS = 1 << 0;
/*
* Indicates that intent locks should be taken on leaf nodes, because we expect
* to be doing updates:
*/
static const __maybe_unused u16 BTREE_ITER_INTENT = 1 << 1;
/*
* Causes the btree iterator code to prefetch additional btree nodes from disk:
*/
static const __maybe_unused u16 BTREE_ITER_PREFETCH = 1 << 2;
/*
* Used in bch2_btree_iter_traverse(), to indicate whether we're searching for
* @pos or the first key strictly greater than @pos
*/
static const __maybe_unused u16 BTREE_ITER_IS_EXTENTS = 1 << 3;
static const __maybe_unused u16 BTREE_ITER_NOT_EXTENTS = 1 << 4;
static const __maybe_unused u16 BTREE_ITER_CACHED = 1 << 5;
static const __maybe_unused u16 BTREE_ITER_WITH_KEY_CACHE = 1 << 6;
static const __maybe_unused u16 BTREE_ITER_WITH_UPDATES = 1 << 7;
static const __maybe_unused u16 BTREE_ITER_WITH_JOURNAL = 1 << 8;
static const __maybe_unused u16 __BTREE_ITER_ALL_SNAPSHOTS = 1 << 9;
static const __maybe_unused u16 BTREE_ITER_ALL_SNAPSHOTS = 1 << 10;
static const __maybe_unused u16 BTREE_ITER_FILTER_SNAPSHOTS = 1 << 11;
static const __maybe_unused u16 BTREE_ITER_NOPRESERVE = 1 << 12;
static const __maybe_unused u16 BTREE_ITER_CACHED_NOFILL = 1 << 13;
static const __maybe_unused u16 BTREE_ITER_KEY_CACHE_FILL = 1 << 14;
#define __BTREE_ITER_FLAGS_END 15
#define BTREE_TRIGGER_FLAGS() \
x(norun) \
x(transactional) \
x(atomic) \
x(check_repair) \
x(gc) \
x(insert) \
x(overwrite) \
x(is_root) \
x(bucket_invalidate)
enum {
#define x(n) BTREE_ITER_FLAG_BIT_##n,
BTREE_ITER_FLAGS()
STR_HASH_FLAGS()
BTREE_UPDATE_FLAGS()
BTREE_TRIGGER_FLAGS()
#undef x
};
/* iter flags must fit in a u16: */
//BUILD_BUG_ON(BTREE_ITER_FLAG_BIT_key_cache_fill > 15);
enum btree_iter_update_trigger_flags {
#define x(n) BTREE_ITER_##n = 1U << BTREE_ITER_FLAG_BIT_##n,
BTREE_ITER_FLAGS()
#undef x
#define x(n) STR_HASH_##n = 1U << BTREE_ITER_FLAG_BIT_##n,
STR_HASH_FLAGS()
#undef x
#define x(n) BTREE_UPDATE_##n = 1U << BTREE_ITER_FLAG_BIT_##n,
BTREE_UPDATE_FLAGS()
#undef x
#define x(n) BTREE_TRIGGER_##n = 1U << BTREE_ITER_FLAG_BIT_##n,
BTREE_TRIGGER_FLAGS()
#undef x
};
enum btree_path_uptodate {
BTREE_ITER_UPTODATE = 0,
@ -307,7 +372,7 @@ struct btree_iter {
*/
struct bkey k;
/* BTREE_ITER_WITH_JOURNAL: */
/* BTREE_ITER_with_journal: */
size_t journal_idx;
#ifdef TRACK_PATH_ALLOCATED
unsigned long ip_allocated;
@ -418,6 +483,8 @@ struct btree_trans {
u8 lock_must_abort;
bool lock_may_not_fail:1;
bool srcu_held:1;
bool locked:1;
bool write_locked:1;
bool used_mempool:1;
bool in_traverse_all:1;
bool paths_sorted:1;
@ -425,13 +492,13 @@ struct btree_trans {
bool journal_transaction_names:1;
bool journal_replay_not_finished:1;
bool notrace_relock_fail:1;
bool write_locked:1;
enum bch_errcode restarted:16;
u32 restart_count;
u64 last_begin_time;
unsigned long last_begin_ip;
unsigned long last_restarted_ip;
unsigned long last_unlock_ip;
unsigned long srcu_lock_time;
const char *fn;

95
fs/bcachefs/btree_update.c
View File

@ -25,14 +25,14 @@ static inline int btree_insert_entry_cmp(const struct btree_insert_entry *l,
static int __must_check
bch2_trans_update_by_path(struct btree_trans *, btree_path_idx_t,
struct bkey_i *, enum btree_update_flags,
struct bkey_i *, enum btree_iter_update_trigger_flags,
unsigned long ip);
static noinline int extent_front_merge(struct btree_trans *trans,
struct btree_iter *iter,
struct bkey_s_c k,
struct bkey_i **insert,
enum btree_update_flags flags)
enum btree_iter_update_trigger_flags flags)
{
struct bch_fs *c = trans->c;
struct bkey_i *update;
@ -104,8 +104,8 @@ static int need_whiteout_for_snapshot(struct btree_trans *trans,
pos.snapshot++;
for_each_btree_key_norestart(trans, iter, btree_id, pos,
BTREE_ITER_ALL_SNAPSHOTS|
BTREE_ITER_NOPRESERVE, k, ret) {
BTREE_ITER_all_snapshots|
BTREE_ITER_nopreserve, k, ret) {
if (!bkey_eq(k.k->p, pos))
break;
@ -138,8 +138,8 @@ int __bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
darray_init(&s);
bch2_trans_iter_init(trans, &old_iter, id, old_pos,
BTREE_ITER_NOT_EXTENTS|
BTREE_ITER_ALL_SNAPSHOTS);
BTREE_ITER_not_extents|
BTREE_ITER_all_snapshots);
while ((old_k = bch2_btree_iter_prev(&old_iter)).k &&
!(ret = bkey_err(old_k)) &&
bkey_eq(old_pos, old_k.k->p)) {
@ -151,8 +151,8 @@ int __bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
continue;
new_k = bch2_bkey_get_iter(trans, &new_iter, id, whiteout_pos,
BTREE_ITER_NOT_EXTENTS|
BTREE_ITER_INTENT);
BTREE_ITER_not_extents|
BTREE_ITER_intent);
ret = bkey_err(new_k);
if (ret)
break;
@ -168,7 +168,7 @@ int __bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
update->k.type = KEY_TYPE_whiteout;
ret = bch2_trans_update(trans, &new_iter, update,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
}
bch2_trans_iter_exit(trans, &new_iter);
@ -185,7 +185,7 @@ int __bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
struct btree_iter *iter,
enum btree_update_flags flags,
enum btree_iter_update_trigger_flags flags,
struct bkey_s_c old,
struct bkey_s_c new)
{
@ -218,7 +218,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
ret = bch2_insert_snapshot_whiteouts(trans, btree_id,
old.k->p, update->k.p) ?:
bch2_btree_insert_nonextent(trans, btree_id, update,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|flags);
BTREE_UPDATE_internal_snapshot_node|flags);
if (ret)
return ret;
}
@ -235,7 +235,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
ret = bch2_insert_snapshot_whiteouts(trans, btree_id,
old.k->p, update->k.p) ?:
bch2_btree_insert_nonextent(trans, btree_id, update,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|flags);
BTREE_UPDATE_internal_snapshot_node|flags);
if (ret)
return ret;
}
@ -260,7 +260,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
}
ret = bch2_btree_insert_nonextent(trans, btree_id, update,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|flags);
BTREE_UPDATE_internal_snapshot_node|flags);
if (ret)
return ret;
}
@ -273,7 +273,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
bch2_cut_front(new.k->p, update);
ret = bch2_trans_update_by_path(trans, iter->path, update,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|
BTREE_UPDATE_internal_snapshot_node|
flags, _RET_IP_);
if (ret)
return ret;
@ -285,7 +285,7 @@ int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
static int bch2_trans_update_extent(struct btree_trans *trans,
struct btree_iter *orig_iter,
struct bkey_i *insert,
enum btree_update_flags flags)
enum btree_iter_update_trigger_flags flags)
{
struct btree_iter iter;
struct bkey_s_c k;
@ -293,9 +293,9 @@ static int bch2_trans_update_extent(struct btree_trans *trans,
int ret = 0;
bch2_trans_iter_init(trans, &iter, btree_id, bkey_start_pos(&insert->k),
BTREE_ITER_INTENT|
BTREE_ITER_WITH_UPDATES|
BTREE_ITER_NOT_EXTENTS);
BTREE_ITER_intent|
BTREE_ITER_with_updates|
BTREE_ITER_not_extents);
k = bch2_btree_iter_peek_upto(&iter, POS(insert->k.p.inode, U64_MAX));
if ((ret = bkey_err(k)))
goto err;
@ -346,7 +346,7 @@ err:
static noinline int flush_new_cached_update(struct btree_trans *trans,
struct btree_insert_entry *i,
enum btree_update_flags flags,
enum btree_iter_update_trigger_flags flags,
unsigned long ip)
{
struct bkey k;
@ -354,7 +354,7 @@ static noinline int flush_new_cached_update(struct btree_trans *trans,
btree_path_idx_t path_idx =
bch2_path_get(trans, i->btree_id, i->old_k.p, 1, 0,
BTREE_ITER_INTENT, _THIS_IP_);
BTREE_ITER_intent, _THIS_IP_);
ret = bch2_btree_path_traverse(trans, path_idx, 0);
if (ret)
goto out;
@ -372,7 +372,7 @@ static noinline int flush_new_cached_update(struct btree_trans *trans,
goto out;
i->key_cache_already_flushed = true;
i->flags |= BTREE_TRIGGER_NORUN;
i->flags |= BTREE_TRIGGER_norun;
btree_path_set_should_be_locked(btree_path);
ret = bch2_trans_update_by_path(trans, path_idx, i->k, flags, ip);
@ -383,7 +383,7 @@ out:
static int __must_check
bch2_trans_update_by_path(struct btree_trans *trans, btree_path_idx_t path_idx,
struct bkey_i *k, enum btree_update_flags flags,
struct bkey_i *k, enum btree_iter_update_trigger_flags flags,
unsigned long ip)
{
struct bch_fs *c = trans->c;
@ -479,15 +479,15 @@ static noinline int bch2_trans_update_get_key_cache(struct btree_trans *trans,
if (!iter->key_cache_path)
iter->key_cache_path =
bch2_path_get(trans, path->btree_id, path->pos, 1, 0,
BTREE_ITER_INTENT|
BTREE_ITER_CACHED, _THIS_IP_);
BTREE_ITER_intent|
BTREE_ITER_cached, _THIS_IP_);
iter->key_cache_path =
bch2_btree_path_set_pos(trans, iter->key_cache_path, path->pos,
iter->flags & BTREE_ITER_INTENT,
iter->flags & BTREE_ITER_intent,
_THIS_IP_);
ret = bch2_btree_path_traverse(trans, iter->key_cache_path, BTREE_ITER_CACHED);
ret = bch2_btree_path_traverse(trans, iter->key_cache_path, BTREE_ITER_cached);
if (unlikely(ret))
return ret;
@ -505,17 +505,17 @@ static noinline int bch2_trans_update_get_key_cache(struct btree_trans *trans,
}
int __must_check bch2_trans_update(struct btree_trans *trans, struct btree_iter *iter,
struct bkey_i *k, enum btree_update_flags flags)
struct bkey_i *k, enum btree_iter_update_trigger_flags flags)
{
btree_path_idx_t path_idx = iter->update_path ?: iter->path;
int ret;
if (iter->flags & BTREE_ITER_IS_EXTENTS)
if (iter->flags & BTREE_ITER_is_extents)
return bch2_trans_update_extent(trans, iter, k, flags);
if (bkey_deleted(&k->k) &&
!(flags & BTREE_UPDATE_KEY_CACHE_RECLAIM) &&
(iter->flags & BTREE_ITER_FILTER_SNAPSHOTS)) {
!(flags & BTREE_UPDATE_key_cache_reclaim) &&
(iter->flags & BTREE_ITER_filter_snapshots)) {
ret = need_whiteout_for_snapshot(trans, iter->btree_id, k->k.p);
if (unlikely(ret < 0))
return ret;
@ -528,7 +528,7 @@ int __must_check bch2_trans_update(struct btree_trans *trans, struct btree_iter
* Ensure that updates to cached btrees go to the key cache:
*/
struct btree_path *path = trans->paths + path_idx;
if (!(flags & BTREE_UPDATE_KEY_CACHE_RECLAIM) &&
if (!(flags & BTREE_UPDATE_key_cache_reclaim) &&
!path->cached &&
!path->level &&
btree_id_cached(trans->c, path->btree_id)) {
@ -587,7 +587,7 @@ int bch2_bkey_get_empty_slot(struct btree_trans *trans, struct btree_iter *iter,
struct bkey_s_c k;
int ret = 0;
bch2_trans_iter_init(trans, iter, btree, POS_MAX, BTREE_ITER_INTENT);
bch2_trans_iter_init(trans, iter, btree, POS_MAX, BTREE_ITER_intent);
k = bch2_btree_iter_prev(iter);
ret = bkey_err(k);
if (ret)
@ -621,15 +621,15 @@ void bch2_trans_commit_hook(struct btree_trans *trans,
int bch2_btree_insert_nonextent(struct btree_trans *trans,
enum btree_id btree, struct bkey_i *k,
enum btree_update_flags flags)
enum btree_iter_update_trigger_flags flags)
{
struct btree_iter iter;
int ret;
bch2_trans_iter_init(trans, &iter, btree, k->k.p,
BTREE_ITER_CACHED|
BTREE_ITER_NOT_EXTENTS|
BTREE_ITER_INTENT);
BTREE_ITER_cached|
BTREE_ITER_not_extents|
BTREE_ITER_intent);
ret = bch2_btree_iter_traverse(&iter) ?:
bch2_trans_update(trans, &iter, k, flags);
bch2_trans_iter_exit(trans, &iter);
@ -637,16 +637,13 @@ int bch2_btree_insert_nonextent(struct btree_trans *trans,
}
int bch2_btree_insert_trans(struct btree_trans *trans, enum btree_id id,
struct bkey_i *k, enum btree_update_flags flags)
struct bkey_i *k, enum btree_iter_update_trigger_flags flags)
{
struct btree_iter iter;
int ret;
bch2_trans_iter_init(trans, &iter, id, bkey_start_pos(&k->k),
BTREE_ITER_CACHED|
BTREE_ITER_INTENT);
ret = bch2_btree_iter_traverse(&iter) ?:
bch2_trans_update(trans, &iter, k, flags);
BTREE_ITER_intent|flags);
int ret = bch2_btree_iter_traverse(&iter) ?:
bch2_trans_update(trans, &iter, k, flags);
bch2_trans_iter_exit(trans, &iter);
return ret;
}
@ -698,8 +695,8 @@ int bch2_btree_delete(struct btree_trans *trans,
int ret;
bch2_trans_iter_init(trans, &iter, btree, pos,
BTREE_ITER_CACHED|
BTREE_ITER_INTENT);
BTREE_ITER_cached|
BTREE_ITER_intent);
ret = bch2_btree_iter_traverse(&iter) ?:
bch2_btree_delete_at(trans, &iter, update_flags);
bch2_trans_iter_exit(trans, &iter);
@ -717,7 +714,7 @@ int bch2_btree_delete_range_trans(struct btree_trans *trans, enum btree_id id,
struct bkey_s_c k;
int ret = 0;
bch2_trans_iter_init(trans, &iter, id, start, BTREE_ITER_INTENT);
bch2_trans_iter_init(trans, &iter, id, start, BTREE_ITER_intent);
while ((k = bch2_btree_iter_peek_upto(&iter, end)).k) {
struct disk_reservation disk_res =
bch2_disk_reservation_init(trans->c, 0);
@ -745,7 +742,7 @@ int bch2_btree_delete_range_trans(struct btree_trans *trans, enum btree_id id,
*/
delete.k.p = iter.pos;
if (iter.flags & BTREE_ITER_IS_EXTENTS)
if (iter.flags & BTREE_ITER_is_extents)
bch2_key_resize(&delete.k,
bpos_min(end, k.k->p).offset -
iter.pos.offset);
@ -804,7 +801,7 @@ int bch2_btree_bit_mod(struct btree_trans *trans, enum btree_id btree,
k->k.p = pos;
struct btree_iter iter;
bch2_trans_iter_init(trans, &iter, btree, pos, BTREE_ITER_INTENT);
bch2_trans_iter_init(trans, &iter, btree, pos, BTREE_ITER_intent);
ret = bch2_btree_iter_traverse(&iter) ?:
bch2_trans_update(trans, &iter, k, 0);
@ -852,7 +849,7 @@ __bch2_fs_log_msg(struct bch_fs *c, unsigned commit_flags, const char *fmt,
if (ret)
goto err;
if (!test_bit(JOURNAL_STARTED, &c->journal.flags)) {
if (!test_bit(JOURNAL_running, &c->journal.flags)) {
ret = darray_make_room(&c->journal.early_journal_entries, jset_u64s(u64s));
if (ret)
goto err;

14
fs/bcachefs/btree_update.h
View File

@ -44,16 +44,18 @@ enum bch_trans_commit_flags {
#undef x
};
void bch2_trans_commit_flags_to_text(struct printbuf *, enum bch_trans_commit_flags);
int bch2_btree_delete_extent_at(struct btree_trans *, struct btree_iter *,
unsigned, unsigned);
int bch2_btree_delete_at(struct btree_trans *, struct btree_iter *, unsigned);
int bch2_btree_delete(struct btree_trans *, enum btree_id, struct bpos, unsigned);
int bch2_btree_insert_nonextent(struct btree_trans *, enum btree_id,
struct bkey_i *, enum btree_update_flags);
struct bkey_i *, enum btree_iter_update_trigger_flags);
int bch2_btree_insert_trans(struct btree_trans *, enum btree_id, struct bkey_i *,
enum btree_update_flags);
enum btree_iter_update_trigger_flags);
int bch2_btree_insert(struct bch_fs *, enum btree_id, struct bkey_i *,
struct disk_reservation *, int flags);
@ -94,14 +96,14 @@ static inline int bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
}
int bch2_trans_update_extent_overwrite(struct btree_trans *, struct btree_iter *,
enum btree_update_flags,
enum btree_iter_update_trigger_flags,
struct bkey_s_c, struct bkey_s_c);
int bch2_bkey_get_empty_slot(struct btree_trans *, struct btree_iter *,
enum btree_id, struct bpos);
int __must_check bch2_trans_update(struct btree_trans *, struct btree_iter *,
struct bkey_i *, enum btree_update_flags);
struct bkey_i *, enum btree_iter_update_trigger_flags);
struct jset_entry *__bch2_trans_jset_entry_alloc(struct btree_trans *, unsigned);
@ -276,7 +278,7 @@ static inline struct bkey_i *__bch2_bkey_get_mut_noupdate(struct btree_trans *tr
unsigned flags, unsigned type, unsigned min_bytes)
{
struct bkey_s_c k = __bch2_bkey_get_iter(trans, iter,
btree_id, pos, flags|BTREE_ITER_INTENT, type);
btree_id, pos, flags|BTREE_ITER_intent, type);
struct bkey_i *ret = IS_ERR(k.k)
? ERR_CAST(k.k)
: __bch2_bkey_make_mut_noupdate(trans, k, 0, min_bytes);
@ -299,7 +301,7 @@ static inline struct bkey_i *__bch2_bkey_get_mut(struct btree_trans *trans,
unsigned flags, unsigned type, unsigned min_bytes)
{
struct bkey_i *mut = __bch2_bkey_get_mut_noupdate(trans, iter,
btree_id, pos, flags|BTREE_ITER_INTENT, type, min_bytes);
btree_id, pos, flags|BTREE_ITER_intent, type, min_bytes);
int ret;
if (IS_ERR(mut))

95
fs/bcachefs/btree_update_interior.c
View File

@ -38,22 +38,6 @@ static int bch2_btree_insert_node(struct btree_update *, struct btree_trans *,
btree_path_idx_t, struct btree *, struct keylist *);
static void bch2_btree_update_add_new_node(struct btree_update *, struct btree *);
static btree_path_idx_t get_unlocked_mut_path(struct btree_trans *trans,
enum btree_id btree_id,
unsigned level,
struct bpos pos)
{
btree_path_idx_t path_idx = bch2_path_get(trans, btree_id, pos, level + 1, level,
BTREE_ITER_NOPRESERVE|
BTREE_ITER_INTENT, _RET_IP_);
path_idx = bch2_btree_path_make_mut(trans, path_idx, true, _RET_IP_);
struct btree_path *path = trans->paths + path_idx;
bch2_btree_path_downgrade(trans, path);
__bch2_btree_path_unlock(trans, path);
return path_idx;
}
/*
* Verify that child nodes correctly span parent node's range:
*/
@ -73,6 +57,24 @@ int bch2_btree_node_check_topology(struct btree_trans *trans, struct btree *b)
!bpos_eq(bkey_i_to_btree_ptr_v2(&b->key)->v.min_key,
b->data->min_key));
if (b == btree_node_root(c, b)) {
if (!bpos_eq(b->data->min_key, POS_MIN)) {
printbuf_reset(&buf);
bch2_bpos_to_text(&buf, b->data->min_key);
need_fsck_err(c, btree_root_bad_min_key,
"btree root with incorrect min_key: %s", buf.buf);
goto topology_repair;
}
if (!bpos_eq(b->data->max_key, SPOS_MAX)) {
printbuf_reset(&buf);
bch2_bpos_to_text(&buf, b->data->max_key);
need_fsck_err(c, btree_root_bad_max_key,
"btree root with incorrect max_key: %s", buf.buf);
goto topology_repair;
}
}
if (!b->c.level)
return 0;
@ -158,7 +160,6 @@ topology_repair:
static void __bch2_btree_calc_format(struct bkey_format_state *s, struct btree *b)
{
struct bkey_packed *k;
struct bset_tree *t;
struct bkey uk;
for_each_bset(b, t)
@ -646,7 +647,7 @@ static int btree_update_nodes_written_trans(struct btree_trans *trans,
unsigned level = bkey_i_to_btree_ptr_v2(k)->v.mem_ptr;
ret = bch2_key_trigger_old(trans, as->btree_id, level, bkey_i_to_s_c(k),
BTREE_TRIGGER_TRANSACTIONAL);
BTREE_TRIGGER_transactional);
if (ret)
return ret;
}
@ -655,7 +656,7 @@ static int btree_update_nodes_written_trans(struct btree_trans *trans,
unsigned level = bkey_i_to_btree_ptr_v2(k)->v.mem_ptr;
ret = bch2_key_trigger_new(trans, as->btree_id, level, bkey_i_to_s(k),
BTREE_TRIGGER_TRANSACTIONAL);
BTREE_TRIGGER_transactional);
if (ret)
return ret;
}
@ -735,9 +736,6 @@ err:
*/
b = READ_ONCE(as->b);
if (b) {
btree_path_idx_t path_idx = get_unlocked_mut_path(trans,
as->btree_id, b->c.level, b->key.k.p);
struct btree_path *path = trans->paths + path_idx;
/*
* @b is the node we did the final insert into:
*
@ -755,12 +753,16 @@ err:
* btree_node_lock_nopath() (the use of which is always suspect,
* we need to work on removing this in the future)
*
* It should be, but get_unlocked_mut_path() -> bch2_path_get()
* It should be, but bch2_path_get_unlocked_mut() -> bch2_path_get()
* calls bch2_path_upgrade(), before we call path_make_mut(), so
* we may rarely end up with a locked path besides the one we
* have here:
*/
bch2_trans_unlock(trans);
bch2_trans_begin(trans);
btree_path_idx_t path_idx = bch2_path_get_unlocked_mut(trans,
as->btree_id, b->c.level, b->key.k.p);
struct btree_path *path = trans->paths + path_idx;
btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_intent);
mark_btree_node_locked(trans, path, b->c.level, BTREE_NODE_INTENT_LOCKED);
path->l[b->c.level].lock_seq = six_lock_seq(&b->c.lock);
@ -1154,13 +1156,12 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
flags |= watermark;
if (watermark < BCH_WATERMARK_reclaim &&
test_bit(JOURNAL_SPACE_LOW, &c->journal.flags)) {
test_bit(JOURNAL_space_low, &c->journal.flags)) {
if (flags & BCH_TRANS_COMMIT_journal_reclaim)
return ERR_PTR(-BCH_ERR_journal_reclaim_would_deadlock);
bch2_trans_unlock(trans);
wait_event(c->journal.wait, !test_bit(JOURNAL_SPACE_LOW, &c->journal.flags));
ret = bch2_trans_relock(trans);
ret = drop_locks_do(trans,
({ wait_event(c->journal.wait, !test_bit(JOURNAL_space_low, &c->journal.flags)); 0; }));
if (ret)
return ERR_PTR(ret);
}
@ -1206,7 +1207,7 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
as->start_time = start_time;
as->ip_started = _RET_IP_;
as->mode = BTREE_UPDATE_none;
as->watermark = watermark;
as->flags = flags;
as->took_gc_lock = true;
as->btree_id = path->btree_id;
as->update_level_start = level_start;
@ -1360,7 +1361,7 @@ static void bch2_insert_fixup_btree_ptr(struct btree_update *as,
BUG_ON(insert->k.type == KEY_TYPE_btree_ptr_v2 &&
!btree_ptr_sectors_written(insert));
if (unlikely(!test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags)))
if (unlikely(!test_bit(JOURNAL_replay_done, &c->journal.flags)))
bch2_journal_key_overwritten(c, b->c.btree_id, b->c.level, insert->k.p);
if (bch2_bkey_invalid(c, bkey_i_to_s_c(insert),
@ -1619,12 +1620,12 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans,
six_unlock_write(&n2->c.lock);
six_unlock_write(&n1->c.lock);
path1 = get_unlocked_mut_path(trans, as->btree_id, n1->c.level, n1->key.k.p);
path1 = bch2_path_get_unlocked_mut(trans, as->btree_id, n1->c.level, n1->key.k.p);
six_lock_increment(&n1->c.lock, SIX_LOCK_intent);
mark_btree_node_locked(trans, trans->paths + path1, n1->c.level, BTREE_NODE_INTENT_LOCKED);
bch2_btree_path_level_init(trans, trans->paths + path1, n1);
path2 = get_unlocked_mut_path(trans, as->btree_id, n2->c.level, n2->key.k.p);
path2 = bch2_path_get_unlocked_mut(trans, as->btree_id, n2->c.level, n2->key.k.p);
six_lock_increment(&n2->c.lock, SIX_LOCK_intent);
mark_btree_node_locked(trans, trans->paths + path2, n2->c.level, BTREE_NODE_INTENT_LOCKED);
bch2_btree_path_level_init(trans, trans->paths + path2, n2);
@ -1669,7 +1670,7 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans,
bch2_btree_update_add_new_node(as, n1);
six_unlock_write(&n1->c.lock);
path1 = get_unlocked_mut_path(trans, as->btree_id, n1->c.level, n1->key.k.p);
path1 = bch2_path_get_unlocked_mut(trans, as->btree_id, n1->c.level, n1->key.k.p);
six_lock_increment(&n1->c.lock, SIX_LOCK_intent);
mark_btree_node_locked(trans, trans->paths + path1, n1->c.level, BTREE_NODE_INTENT_LOCKED);
bch2_btree_path_level_init(trans, trans->paths + path1, n1);
@ -1947,6 +1948,8 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
u64 start_time = local_clock();
int ret = 0;
bch2_trans_verify_not_in_restart(trans);
bch2_trans_verify_not_unlocked(trans);
BUG_ON(!trans->paths[path].should_be_locked);
BUG_ON(!btree_node_locked(&trans->paths[path], level));
@ -1979,7 +1982,7 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
: bpos_successor(b->data->max_key);
sib_path = bch2_path_get(trans, btree, sib_pos,
U8_MAX, level, BTREE_ITER_INTENT, _THIS_IP_);
U8_MAX, level, BTREE_ITER_intent, _THIS_IP_);
ret = bch2_btree_path_traverse(trans, sib_path, false);
if (ret)
goto err;
@ -2072,7 +2075,7 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
bch2_btree_update_add_new_node(as, n);
six_unlock_write(&n->c.lock);
new_path = get_unlocked_mut_path(trans, btree, n->c.level, n->key.k.p);
new_path = bch2_path_get_unlocked_mut(trans, btree, n->c.level, n->key.k.p);
six_lock_increment(&n->c.lock, SIX_LOCK_intent);
mark_btree_node_locked(trans, trans->paths + new_path, n->c.level, BTREE_NODE_INTENT_LOCKED);
bch2_btree_path_level_init(trans, trans->paths + new_path, n);
@ -2150,7 +2153,7 @@ int bch2_btree_node_rewrite(struct btree_trans *trans,
bch2_btree_update_add_new_node(as, n);
six_unlock_write(&n->c.lock);
new_path = get_unlocked_mut_path(trans, iter->btree_id, n->c.level, n->key.k.p);
new_path = bch2_path_get_unlocked_mut(trans, iter->btree_id, n->c.level, n->key.k.p);
six_lock_increment(&n->c.lock, SIX_LOCK_intent);
mark_btree_node_locked(trans, trans->paths + new_path, n->c.level, BTREE_NODE_INTENT_LOCKED);
bch2_btree_path_level_init(trans, trans->paths + new_path, n);
@ -2333,10 +2336,10 @@ static int __bch2_btree_node_update_key(struct btree_trans *trans,
if (!skip_triggers) {
ret = bch2_key_trigger_old(trans, b->c.btree_id, b->c.level + 1,
bkey_i_to_s_c(&b->key),
BTREE_TRIGGER_TRANSACTIONAL) ?:
BTREE_TRIGGER_transactional) ?:
bch2_key_trigger_new(trans, b->c.btree_id, b->c.level + 1,
bkey_i_to_s(new_key),
BTREE_TRIGGER_TRANSACTIONAL);
BTREE_TRIGGER_transactional);
if (ret)
return ret;
}
@ -2353,7 +2356,7 @@ static int __bch2_btree_node_update_key(struct btree_trans *trans,
bch2_trans_copy_iter(&iter2, iter);
iter2.path = bch2_btree_path_make_mut(trans, iter2.path,
iter2.flags & BTREE_ITER_INTENT,
iter2.flags & BTREE_ITER_intent,
_THIS_IP_);
struct btree_path *path2 = btree_iter_path(trans, &iter2);
@ -2365,7 +2368,7 @@ static int __bch2_btree_node_update_key(struct btree_trans *trans,
trans->paths_sorted = false;
ret = bch2_btree_iter_traverse(&iter2) ?:
bch2_trans_update(trans, &iter2, new_key, BTREE_TRIGGER_NORUN);
bch2_trans_update(trans, &iter2, new_key, BTREE_TRIGGER_norun);
if (ret)
goto err;
} else {
@ -2473,7 +2476,7 @@ int bch2_btree_node_update_key_get_iter(struct btree_trans *trans,
bch2_trans_node_iter_init(trans, &iter, b->c.btree_id, b->key.k.p,
BTREE_MAX_DEPTH, b->c.level,
BTREE_ITER_INTENT);
BTREE_ITER_intent);
ret = bch2_btree_iter_traverse(&iter);
if (ret)
goto out;
@ -2487,7 +2490,6 @@ int bch2_btree_node_update_key_get_iter(struct btree_trans *trans,
BUG_ON(!btree_node_hashed(b));
struct bch_extent_ptr *ptr;
bch2_bkey_drop_ptrs(bkey_i_to_s(new_key), ptr,
!bch2_bkey_has_device(bkey_i_to_s(&b->key), ptr->dev));
@ -2511,7 +2513,7 @@ void bch2_btree_set_root_for_read(struct bch_fs *c, struct btree *b)
bch2_btree_set_root_inmem(c, b);
}
static int __bch2_btree_root_alloc_fake(struct btree_trans *trans, enum btree_id id, unsigned level)
int bch2_btree_root_alloc_fake_trans(struct btree_trans *trans, enum btree_id id, unsigned level)
{
struct bch_fs *c = trans->c;
struct closure cl;
@ -2559,17 +2561,18 @@ static int __bch2_btree_root_alloc_fake(struct btree_trans *trans, enum btree_id
void bch2_btree_root_alloc_fake(struct bch_fs *c, enum btree_id id, unsigned level)
{
bch2_trans_run(c, __bch2_btree_root_alloc_fake(trans, id, level));
bch2_trans_run(c, bch2_btree_root_alloc_fake_trans(trans, id, level));
}
static void bch2_btree_update_to_text(struct printbuf *out, struct btree_update *as)
{
prt_printf(out, "%ps: btree=%s l=%u-%u watermark=%s mode=%s nodes_written=%u cl.remaining=%u journal_seq=%llu\n",
(void *) as->ip_started,
prt_printf(out, "%ps: ", (void *) as->ip_started);
bch2_trans_commit_flags_to_text(out, as->flags);
prt_printf(out, " btree=%s l=%u-%u mode=%s nodes_written=%u cl.remaining=%u journal_seq=%llu\n",
bch2_btree_id_str(as->btree_id),
as->update_level_start,
as->update_level_end,
bch2_watermarks[as->watermark],
bch2_btree_update_modes[as->mode],
as->nodes_written,
closure_nr_remaining(&as->cl),

7
fs/bcachefs/btree_update_interior.h
View File

@ -52,7 +52,7 @@ struct btree_update {
struct list_head unwritten_list;
enum btree_update_mode mode;
enum bch_watermark watermark;
enum bch_trans_commit_flags flags;
unsigned nodes_written:1;
unsigned took_gc_lock:1;
@ -144,6 +144,9 @@ static inline int bch2_foreground_maybe_merge_sibling(struct btree_trans *trans,
EBUG_ON(!btree_node_locked(path, level));
if (bch2_btree_node_merging_disabled)
return 0;
b = path->l[level].b;
if (b->sib_u64s[sib] > trans->c->btree_foreground_merge_threshold)
return 0;
@ -172,6 +175,8 @@ int bch2_btree_node_update_key_get_iter(struct btree_trans *, struct btree *,
struct bkey_i *, unsigned, bool);
void bch2_btree_set_root_for_read(struct bch_fs *, struct btree *);
int bch2_btree_root_alloc_fake_trans(struct btree_trans *, enum btree_id, unsigned);
void bch2_btree_root_alloc_fake(struct bch_fs *, enum btree_id, unsigned);
static inline unsigned btree_update_reserve_required(struct bch_fs *c,

8
fs/bcachefs/btree_write_buffer.c
View File

@ -122,7 +122,7 @@ static noinline int wb_flush_one_slowpath(struct btree_trans *trans,
trans->journal_res.seq = wb->journal_seq;
return bch2_trans_update(trans, iter, &wb->k,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
BTREE_UPDATE_internal_snapshot_node) ?:
bch2_trans_commit(trans, NULL, NULL,
BCH_TRANS_COMMIT_no_enospc|
BCH_TRANS_COMMIT_no_check_rw|
@ -191,13 +191,13 @@ btree_write_buffered_insert(struct btree_trans *trans,
int ret;
bch2_trans_iter_init(trans, &iter, wb->btree, bkey_start_pos(&wb->k.k),
BTREE_ITER_CACHED|BTREE_ITER_INTENT);
BTREE_ITER_cached|BTREE_ITER_intent);
trans->journal_res.seq = wb->journal_seq;
ret = bch2_btree_iter_traverse(&iter) ?:
bch2_trans_update(trans, &iter, &wb->k,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
bch2_trans_iter_exit(trans, &iter);
return ret;
}
@ -332,7 +332,7 @@ static int bch2_btree_write_buffer_flush_locked(struct btree_trans *trans)
if (!iter.path || iter.btree_id != k->btree) {
bch2_trans_iter_exit(trans, &iter);
bch2_trans_iter_init(trans, &iter, k->btree, k->k.k.p,
BTREE_ITER_INTENT|BTREE_ITER_ALL_SNAPSHOTS);
BTREE_ITER_intent|BTREE_ITER_all_snapshots);
}
bch2_btree_iter_set_pos(&iter, k->k.k.p);

725
fs/bcachefs/buckets.c
View File

File diff suppressed because it is too large Load Diff

70
fs/bcachefs/buckets.h
View File

@ -12,7 +12,7 @@
#include "extents.h"
#include "sb-members.h"
static inline size_t sector_to_bucket(const struct bch_dev *ca, sector_t s)
static inline u64 sector_to_bucket(const struct bch_dev *ca, sector_t s)
{
return div_u64(s, ca->mi.bucket_size);
}
@ -30,8 +30,7 @@ static inline sector_t bucket_remainder(const struct bch_dev *ca, sector_t s)
return remainder;
}
static inline size_t sector_to_bucket_and_offset(const struct bch_dev *ca, sector_t s,
u32 *offset)
static inline u64 sector_to_bucket_and_offset(const struct bch_dev *ca, sector_t s, u32 *offset)
{
return div_u64_rem(s, ca->mi.bucket_size, offset);
}
@ -94,7 +93,7 @@ static inline struct bucket *gc_bucket(struct bch_dev *ca, size_t b)
{
struct bucket_array *buckets = gc_bucket_array(ca);
BUG_ON(b < buckets->first_bucket || b >= buckets->nbuckets);
BUG_ON(!bucket_valid(ca, b));
return buckets->b + b;
}
@ -111,7 +110,7 @@ static inline u8 *bucket_gen(struct bch_dev *ca, size_t b)
{
struct bucket_gens *gens = bucket_gens(ca);
BUG_ON(b < gens->first_bucket || b >= gens->nbuckets);
BUG_ON(!bucket_valid(ca, b));
return gens->b + b;
}
@ -121,20 +120,16 @@ static inline size_t PTR_BUCKET_NR(const struct bch_dev *ca,
return sector_to_bucket(ca, ptr->offset);
}
static inline struct bpos PTR_BUCKET_POS(const struct bch_fs *c,
const struct bch_extent_ptr *ptr)
static inline struct bpos PTR_BUCKET_POS(const struct bch_dev *ca,
const struct bch_extent_ptr *ptr)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
return POS(ptr->dev, PTR_BUCKET_NR(ca, ptr));
}
static inline struct bpos PTR_BUCKET_POS_OFFSET(const struct bch_fs *c,
static inline struct bpos PTR_BUCKET_POS_OFFSET(const struct bch_dev *ca,
const struct bch_extent_ptr *ptr,
u32 *bucket_offset)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
return POS(ptr->dev, sector_to_bucket_and_offset(ca, ptr->offset, bucket_offset));
}
@ -175,17 +170,19 @@ static inline int gen_after(u8 a, u8 b)
return r > 0 ? r : 0;
}
static inline u8 dev_ptr_stale_rcu(struct bch_dev *ca, const struct bch_extent_ptr *ptr)
{
return gen_after(*bucket_gen(ca, PTR_BUCKET_NR(ca, ptr)), ptr->gen);
}
/**
* ptr_stale() - check if a pointer points into a bucket that has been
* dev_ptr_stale() - check if a pointer points into a bucket that has been
* invalidated.
*/
static inline u8 ptr_stale(struct bch_dev *ca,
const struct bch_extent_ptr *ptr)
static inline u8 dev_ptr_stale(struct bch_dev *ca, const struct bch_extent_ptr *ptr)
{
u8 ret;
rcu_read_lock();
ret = gen_after(*bucket_gen(ca, PTR_BUCKET_NR(ca, ptr)), ptr->gen);
u8 ret = dev_ptr_stale_rcu(ca, ptr);
rcu_read_unlock();
return ret;
@ -306,8 +303,6 @@ bch2_fs_usage_read_short(struct bch_fs *);
void bch2_dev_usage_update(struct bch_fs *, struct bch_dev *,
const struct bch_alloc_v4 *,
const struct bch_alloc_v4 *, u64, bool);
void bch2_dev_usage_update_m(struct bch_fs *, struct bch_dev *,
struct bucket *, struct bucket *);
/* key/bucket marking: */
@ -333,27 +328,29 @@ int bch2_replicas_deltas_realloc(struct btree_trans *, unsigned);
void bch2_fs_usage_initialize(struct bch_fs *);
int bch2_check_bucket_ref(struct btree_trans *, struct bkey_s_c,
const struct bch_extent_ptr *,
s64, enum bch_data_type, u8, u8, u32);
int bch2_bucket_ref_update(struct btree_trans *, struct bch_dev *,
struct bkey_s_c, const struct bch_extent_ptr *,
s64, enum bch_data_type, u8, u8, u32 *);
int bch2_mark_metadata_bucket(struct bch_fs *, struct bch_dev *,
size_t, enum bch_data_type, unsigned,
struct gc_pos, unsigned);
int bch2_check_fix_ptrs(struct btree_trans *,
enum btree_id, unsigned, struct bkey_s_c,
enum btree_iter_update_trigger_flags);
int bch2_trigger_extent(struct btree_trans *, enum btree_id, unsigned,
struct bkey_s_c, struct bkey_s, unsigned);
struct bkey_s_c, struct bkey_s,
enum btree_iter_update_trigger_flags);
int bch2_trigger_reservation(struct btree_trans *, enum btree_id, unsigned,
struct bkey_s_c, struct bkey_s, unsigned);
struct bkey_s_c, struct bkey_s,
enum btree_iter_update_trigger_flags);
#define trigger_run_overwrite_then_insert(_fn, _trans, _btree_id, _level, _old, _new, _flags)\
({ \
int ret = 0; \
\
if (_old.k->type) \
ret = _fn(_trans, _btree_id, _level, _old, _flags & ~BTREE_TRIGGER_INSERT); \
ret = _fn(_trans, _btree_id, _level, _old, _flags & ~BTREE_TRIGGER_insert); \
if (!ret && _new.k->type) \
ret = _fn(_trans, _btree_id, _level, _new.s_c, _flags & ~BTREE_TRIGGER_OVERWRITE);\
ret = _fn(_trans, _btree_id, _level, _new.s_c, _flags & ~BTREE_TRIGGER_overwrite);\
ret; \
})
@ -362,9 +359,13 @@ void bch2_trans_account_disk_usage_change(struct btree_trans *);
void bch2_trans_fs_usage_revert(struct btree_trans *, struct replicas_delta_list *);
int bch2_trans_fs_usage_apply(struct btree_trans *, struct replicas_delta_list *);
int bch2_trans_mark_metadata_bucket(struct btree_trans *, struct bch_dev *,
size_t, enum bch_data_type, unsigned);
int bch2_trans_mark_dev_sb(struct bch_fs *, struct bch_dev *);
int bch2_trans_mark_metadata_bucket(struct btree_trans *, struct bch_dev *, u64,
enum bch_data_type, unsigned,
enum btree_iter_update_trigger_flags);
int bch2_trans_mark_dev_sb(struct bch_fs *, struct bch_dev *,
enum btree_iter_update_trigger_flags);
int bch2_trans_mark_dev_sbs_flags(struct bch_fs *,
enum btree_iter_update_trigger_flags);
int bch2_trans_mark_dev_sbs(struct bch_fs *);
static inline bool is_superblock_bucket(struct bch_dev *ca, u64 b)
@ -464,6 +465,9 @@ static inline u64 avail_factor(u64 r)
return div_u64(r << RESERVE_FACTOR, (1 << RESERVE_FACTOR) + 1);
}
void bch2_buckets_nouse_free(struct bch_fs *);
int bch2_buckets_nouse_alloc(struct bch_fs *);
int bch2_dev_buckets_resize(struct bch_fs *, struct bch_dev *, u64);
void bch2_dev_buckets_free(struct bch_dev *);
int bch2_dev_buckets_alloc(struct bch_fs *, struct bch_dev *);

66
fs/bcachefs/chardev.c
View File

@ -32,12 +32,7 @@ static struct bch_dev *bch2_device_lookup(struct bch_fs *c, u64 dev,
if (dev >= c->sb.nr_devices)
return ERR_PTR(-EINVAL);
rcu_read_lock();
ca = rcu_dereference(c->devs[dev]);
if (ca)
percpu_ref_get(&ca->ref);
rcu_read_unlock();
ca = bch2_dev_tryget_noerror(c, dev);
if (!ca)
return ERR_PTR(-EINVAL);
} else {
@ -391,7 +386,7 @@ static long bch2_ioctl_disk_offline(struct bch_fs *c, struct bch_ioctl_disk arg)
return PTR_ERR(ca);
ret = bch2_dev_offline(c, ca, arg.flags);
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
return ret;
}
@ -420,7 +415,7 @@ static long bch2_ioctl_disk_set_state(struct bch_fs *c,
if (ret)
bch_err(c, "Error setting device state: %s", bch2_err_str(ret));
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
return ret;
}
@ -615,7 +610,7 @@ static long bch2_ioctl_dev_usage(struct bch_fs *c,
arg.d[i].fragmented = src.d[i].fragmented;
}
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
return copy_to_user_errcode(user_arg, &arg, sizeof(arg));
}
@ -667,7 +662,7 @@ static long bch2_ioctl_dev_usage_v2(struct bch_fs *c,
goto err;
}
err:
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
return ret;
}
@ -689,11 +684,9 @@ static long bch2_ioctl_read_super(struct bch_fs *c,
if (arg.flags & BCH_READ_DEV) {
ca = bch2_device_lookup(c, arg.dev, arg.flags);
if (IS_ERR(ca)) {
ret = PTR_ERR(ca);
goto err;
}
ret = PTR_ERR_OR_ZERO(ca);
if (ret)
goto err_unlock;
sb = ca->disk_sb.sb;
} else {
@ -708,8 +701,8 @@ static long bch2_ioctl_read_super(struct bch_fs *c,
ret = copy_to_user_errcode((void __user *)(unsigned long)arg.sb, sb,
vstruct_bytes(sb));
err:
if (!IS_ERR_OR_NULL(ca))
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
err_unlock:
mutex_unlock(&c->sb_lock);
return ret;
}
@ -753,7 +746,7 @@ static long bch2_ioctl_disk_resize(struct bch_fs *c,
ret = bch2_dev_resize(c, ca, arg.nbuckets);
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
return ret;
}
@ -779,7 +772,7 @@ static long bch2_ioctl_disk_resize_journal(struct bch_fs *c,
ret = bch2_set_nr_journal_buckets(c, ca, arg.nbuckets);
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
return ret;
}
@ -961,7 +954,9 @@ static const struct file_operations bch_chardev_fops = {
};
static int bch_chardev_major;
static struct class *bch_chardev_class;
static const struct class bch_chardev_class = {
.name = "bcachefs",
};
static struct device *bch_chardev;
void bch2_fs_chardev_exit(struct bch_fs *c)
@ -978,7 +973,7 @@ int bch2_fs_chardev_init(struct bch_fs *c)
if (c->minor < 0)
return c->minor;
c->chardev = device_create(bch_chardev_class, NULL,
c->chardev = device_create(&bch_chardev_class, NULL,
MKDEV(bch_chardev_major, c->minor), c,
"bcachefs%u-ctl", c->minor);
if (IS_ERR(c->chardev))
@ -989,32 +984,39 @@ int bch2_fs_chardev_init(struct bch_fs *c)
void bch2_chardev_exit(void)
{
if (!IS_ERR_OR_NULL(bch_chardev_class))
device_destroy(bch_chardev_class,
MKDEV(bch_chardev_major, U8_MAX));
if (!IS_ERR_OR_NULL(bch_chardev_class))
class_destroy(bch_chardev_class);
device_destroy(&bch_chardev_class, MKDEV(bch_chardev_major, U8_MAX));
class_unregister(&bch_chardev_class);
if (bch_chardev_major > 0)
unregister_chrdev(bch_chardev_major, "bcachefs");
}
int __init bch2_chardev_init(void)
{
int ret;
bch_chardev_major = register_chrdev(0, "bcachefs-ctl", &bch_chardev_fops);
if (bch_chardev_major < 0)
return bch_chardev_major;
bch_chardev_class = class_create("bcachefs");
if (IS_ERR(bch_chardev_class))
return PTR_ERR(bch_chardev_class);
ret = class_register(&bch_chardev_class);
if (ret)
goto major_out;
bch_chardev = device_create(bch_chardev_class, NULL,
bch_chardev = device_create(&bch_chardev_class, NULL,
MKDEV(bch_chardev_major, U8_MAX),
NULL, "bcachefs-ctl");
if (IS_ERR(bch_chardev))
return PTR_ERR(bch_chardev);
if (IS_ERR(bch_chardev)) {
ret = PTR_ERR(bch_chardev);
goto class_out;
}
return 0;
class_out:
class_unregister(&bch_chardev_class);
major_out:
unregister_chrdev(bch_chardev_major, "bcachefs-ctl");
return ret;
}
#endif /* NO_BCACHEFS_CHARDEV */

17
fs/bcachefs/checksum.c
View File

@ -469,9 +469,8 @@ int bch2_rechecksum_bio(struct bch_fs *c, struct bio *bio,
/* BCH_SB_FIELD_crypt: */
static int bch2_sb_crypt_validate(struct bch_sb *sb,
struct bch_sb_field *f,
struct printbuf *err)
static int bch2_sb_crypt_validate(struct bch_sb *sb, struct bch_sb_field *f,
enum bch_validate_flags flags, struct printbuf *err)
{
struct bch_sb_field_crypt *crypt = field_to_type(f, crypt);
@ -494,14 +493,10 @@ static void bch2_sb_crypt_to_text(struct printbuf *out, struct bch_sb *sb,
{
struct bch_sb_field_crypt *crypt = field_to_type(f, crypt);
prt_printf(out, "KFD: %llu", BCH_CRYPT_KDF_TYPE(crypt));
prt_newline(out);
prt_printf(out, "scrypt n: %llu", BCH_KDF_SCRYPT_N(crypt));
prt_newline(out);
prt_printf(out, "scrypt r: %llu", BCH_KDF_SCRYPT_R(crypt));
prt_newline(out);
prt_printf(out, "scrypt p: %llu", BCH_KDF_SCRYPT_P(crypt));
prt_newline(out);
prt_printf(out, "KFD: %llu\n", BCH_CRYPT_KDF_TYPE(crypt));
prt_printf(out, "scrypt n: %llu\n", BCH_KDF_SCRYPT_N(crypt));
prt_printf(out, "scrypt r: %llu\n", BCH_KDF_SCRYPT_R(crypt));
prt_printf(out, "scrypt p: %llu\n", BCH_KDF_SCRYPT_P(crypt));
}
const struct bch_sb_field_ops bch_sb_field_ops_crypt = {

54
fs/bcachefs/data_update.c
View File

@ -106,7 +106,7 @@ static int __bch2_data_update_index_update(struct btree_trans *trans,
bch2_trans_iter_init(trans, &iter, m->btree_id,
bkey_start_pos(&bch2_keylist_front(keys)->k),
BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
BTREE_ITER_slots|BTREE_ITER_intent);
while (1) {
struct bkey_s_c k;
@ -203,6 +203,8 @@ restart_drop_conflicting_replicas:
/* Now, drop excess replicas: */
restart_drop_extra_replicas:
rcu_read_lock();
bkey_for_each_ptr_decode(old.k, bch2_bkey_ptrs(bkey_i_to_s(insert)), p, entry) {
unsigned ptr_durability = bch2_extent_ptr_durability(c, &p);
@ -214,6 +216,7 @@ restart_drop_extra_replicas:
goto restart_drop_extra_replicas;
}
}
rcu_read_unlock();
/* Finally, add the pointers we just wrote: */
extent_for_each_ptr_decode(extent_i_to_s(new), p, entry)
@ -288,7 +291,7 @@ restart_drop_extra_replicas:
k.k->p, insert->k.p) ?:
bch2_bkey_set_needs_rebalance(c, insert, &op->opts) ?:
bch2_trans_update(trans, &iter, insert,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
BTREE_UPDATE_internal_snapshot_node) ?:
bch2_trans_commit(trans, &op->res,
NULL,
BCH_TRANS_COMMIT_no_check_rw|
@ -357,10 +360,11 @@ void bch2_data_update_exit(struct data_update *update)
bch2_bkey_ptrs_c(bkey_i_to_s_c(update->k.k));
bkey_for_each_ptr(ptrs, ptr) {
struct bch_dev *ca = bch2_dev_have_ref(c, ptr->dev);
if (c->opts.nocow_enabled)
bch2_bucket_nocow_unlock(&c->nocow_locks,
PTR_BUCKET_POS(c, ptr), 0);
percpu_ref_put(&bch_dev_bkey_exists(c, ptr->dev)->ref);
PTR_BUCKET_POS(ca, ptr), 0);
bch2_dev_put(ca);
}
bch2_bkey_buf_exit(&update->k, c);
@ -386,8 +390,10 @@ static void bch2_update_unwritten_extent(struct btree_trans *trans,
while (bio_sectors(bio)) {
unsigned sectors = bio_sectors(bio);
bch2_trans_begin(trans);
bch2_trans_iter_init(trans, &iter, update->btree_id, update->op.pos,
BTREE_ITER_SLOTS);
BTREE_ITER_slots);
ret = lockrestart_do(trans, ({
k = bch2_btree_iter_peek_slot(&iter);
bkey_err(k);
@ -465,7 +471,6 @@ int bch2_extent_drop_ptrs(struct btree_trans *trans,
while (data_opts.kill_ptrs) {
unsigned i = 0, drop = __fls(data_opts.kill_ptrs);
struct bch_extent_ptr *ptr;
bch2_bkey_drop_ptrs(bkey_i_to_s(n), ptr, i++ == drop);
data_opts.kill_ptrs ^= 1U << drop;
@ -480,15 +485,15 @@ int bch2_extent_drop_ptrs(struct btree_trans *trans,
/*
* Since we're not inserting through an extent iterator
* (BTREE_ITER_ALL_SNAPSHOTS iterators aren't extent iterators),
* (BTREE_ITER_all_snapshots iterators aren't extent iterators),
* we aren't using the extent overwrite path to delete, we're
* just using the normal key deletion path:
*/
if (bkey_deleted(&n->k) && !(iter->flags & BTREE_ITER_IS_EXTENTS))
if (bkey_deleted(&n->k) && !(iter->flags & BTREE_ITER_is_extents))
n->k.size = 0;
return bch2_trans_relock(trans) ?:
bch2_trans_update(trans, iter, n, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
bch2_trans_update(trans, iter, n, BTREE_UPDATE_internal_snapshot_node) ?:
bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
}
@ -539,15 +544,26 @@ int bch2_data_update_init(struct btree_trans *trans,
m->op.compression_opt = background_compression(io_opts);
m->op.watermark = m->data_opts.btree_insert_flags & BCH_WATERMARK_MASK;
bkey_for_each_ptr(ptrs, ptr)
percpu_ref_get(&bch_dev_bkey_exists(c, ptr->dev)->ref);
bkey_for_each_ptr(ptrs, ptr) {
if (!bch2_dev_tryget(c, ptr->dev)) {
bkey_for_each_ptr(ptrs, ptr2) {
if (ptr2 == ptr)
break;
bch2_dev_put(bch2_dev_have_ref(c, ptr2->dev));
}
return -BCH_ERR_data_update_done;
}
}
unsigned durability_have = 0, durability_removing = 0;
i = 0;
bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
struct bch_dev *ca = bch2_dev_have_ref(c, p.ptr.dev);
struct bpos bucket = PTR_BUCKET_POS(ca, &p.ptr);
bool locked;
rcu_read_lock();
if (((1U << i) & m->data_opts.rewrite_ptrs)) {
BUG_ON(p.ptr.cached);
@ -561,6 +577,7 @@ int bch2_data_update_init(struct btree_trans *trans,
bch2_dev_list_add_dev(&m->op.devs_have, p.ptr.dev);
durability_have += bch2_extent_ptr_durability(c, &p);
}
rcu_read_unlock();
/*
* op->csum_type is normally initialized from the fs/file's
@ -579,15 +596,13 @@ int bch2_data_update_init(struct btree_trans *trans,
if (ctxt) {
move_ctxt_wait_event(ctxt,
(locked = bch2_bucket_nocow_trylock(&c->nocow_locks,
PTR_BUCKET_POS(c, &p.ptr), 0)) ||
bucket, 0)) ||
list_empty(&ctxt->ios));
if (!locked)
bch2_bucket_nocow_lock(&c->nocow_locks,
PTR_BUCKET_POS(c, &p.ptr), 0);
bch2_bucket_nocow_lock(&c->nocow_locks, bucket, 0);
} else {
if (!bch2_bucket_nocow_trylock(&c->nocow_locks,
PTR_BUCKET_POS(c, &p.ptr), 0)) {
if (!bch2_bucket_nocow_trylock(&c->nocow_locks, bucket, 0)) {
ret = -BCH_ERR_nocow_lock_blocked;
goto err;
}
@ -649,10 +664,11 @@ int bch2_data_update_init(struct btree_trans *trans,
err:
i = 0;
bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
struct bch_dev *ca = bch2_dev_have_ref(c, p.ptr.dev);
struct bpos bucket = PTR_BUCKET_POS(ca, &p.ptr);
if ((1U << i) & ptrs_locked)
bch2_bucket_nocow_unlock(&c->nocow_locks,
PTR_BUCKET_POS(c, &p.ptr), 0);
percpu_ref_put(&bch_dev_bkey_exists(c, p.ptr.dev)->ref);
bch2_bucket_nocow_unlock(&c->nocow_locks, bucket, 0);
bch2_dev_put(ca);
i++;
}

80
fs/bcachefs/debug.c
View File

@ -37,11 +37,11 @@ static bool bch2_btree_verify_replica(struct bch_fs *c, struct btree *b,
struct btree_node *n_ondisk = c->verify_ondisk;
struct btree_node *n_sorted = c->verify_data->data;
struct bset *sorted, *inmemory = &b->data->keys;
struct bch_dev *ca = bch_dev_bkey_exists(c, pick.ptr.dev);
struct bio *bio;
bool failed = false, saw_error = false;
if (!bch2_dev_get_ioref(ca, READ))
struct bch_dev *ca = bch2_dev_get_ioref(c, pick.ptr.dev, READ);
if (!ca)
return false;
bio = bio_alloc_bioset(ca->disk_sb.bdev,
@ -194,8 +194,8 @@ void bch2_btree_node_ondisk_to_text(struct printbuf *out, struct bch_fs *c,
return;
}
ca = bch_dev_bkey_exists(c, pick.ptr.dev);
if (!bch2_dev_get_ioref(ca, READ)) {
ca = bch2_dev_get_ioref(c, pick.ptr.dev, READ);
if (!ca) {
prt_printf(out, "error getting device to read from: not online\n");
return;
}
@ -375,8 +375,8 @@ static ssize_t bch2_read_btree(struct file *file, char __user *buf,
return flush_buf(i) ?:
bch2_trans_run(i->c,
for_each_btree_key(trans, iter, i->id, i->from,
BTREE_ITER_PREFETCH|
BTREE_ITER_ALL_SNAPSHOTS, k, ({
BTREE_ITER_prefetch|
BTREE_ITER_all_snapshots, k, ({
bch2_bkey_val_to_text(&i->buf, i->c, k);
prt_newline(&i->buf);
bch2_trans_unlock(trans);
@ -459,8 +459,8 @@ static ssize_t bch2_read_bfloat_failed(struct file *file, char __user *buf,
return flush_buf(i) ?:
bch2_trans_run(i->c,
for_each_btree_key(trans, iter, i->id, i->from,
BTREE_ITER_PREFETCH|
BTREE_ITER_ALL_SNAPSHOTS, k, ({
BTREE_ITER_prefetch|
BTREE_ITER_all_snapshots, k, ({
struct btree_path_level *l =
&btree_iter_path(trans, &iter)->l[0];
struct bkey_packed *_k =
@ -492,51 +492,26 @@ static void bch2_cached_btree_node_to_text(struct printbuf *out, struct bch_fs *
if (!out->nr_tabstops)
printbuf_tabstop_push(out, 32);
prt_printf(out, "%px btree=%s l=%u ",
b,
bch2_btree_id_str(b->c.btree_id),
b->c.level);
prt_newline(out);
prt_printf(out, "%px btree=%s l=%u\n", b, bch2_btree_id_str(b->c.btree_id), b->c.level);
printbuf_indent_add(out, 2);
bch2_bkey_val_to_text(out, c, bkey_i_to_s_c(&b->key));
prt_newline(out);
prt_printf(out, "flags: ");
prt_tab(out);
prt_printf(out, "flags:\t");
prt_bitflags(out, bch2_btree_node_flags, b->flags);
prt_newline(out);
prt_printf(out, "pcpu read locks: ");
prt_tab(out);
prt_printf(out, "%u", b->c.lock.readers != NULL);
prt_newline(out);
prt_printf(out, "pcpu read locks:\t%u\n", b->c.lock.readers != NULL);
prt_printf(out, "written:\t%u\n", b->written);
prt_printf(out, "writes blocked:\t%u\n", !list_empty_careful(&b->write_blocked));
prt_printf(out, "will make reachable:\t%lx\n", b->will_make_reachable);
prt_printf(out, "written:");
prt_tab(out);
prt_printf(out, "%u", b->written);
prt_newline(out);
prt_printf(out, "writes blocked:");
prt_tab(out);
prt_printf(out, "%u", !list_empty_careful(&b->write_blocked));
prt_newline(out);
prt_printf(out, "will make reachable:");
prt_tab(out);
prt_printf(out, "%lx", b->will_make_reachable);
prt_newline(out);
prt_printf(out, "journal pin %px:", &b->writes[0].journal);
prt_tab(out);
prt_printf(out, "%llu", b->writes[0].journal.seq);
prt_newline(out);
prt_printf(out, "journal pin %px:", &b->writes[1].journal);
prt_tab(out);
prt_printf(out, "%llu", b->writes[1].journal.seq);
prt_newline(out);
prt_printf(out, "journal pin %px:\t%llu\n",
&b->writes[0].journal, b->writes[0].journal.seq);
prt_printf(out, "journal pin %px:\t%llu\n",
&b->writes[1].journal, b->writes[1].journal.seq);
printbuf_indent_sub(out, 2);
}
@ -625,8 +600,7 @@ restart:
bch2_btree_trans_to_text(&i->buf, trans);
prt_printf(&i->buf, "backtrace:");
prt_newline(&i->buf);
prt_printf(&i->buf, "backtrace:\n");
printbuf_indent_add(&i->buf, 2);
bch2_prt_task_backtrace(&i->buf, task, 0, GFP_KERNEL);
printbuf_indent_sub(&i->buf, 2);
@ -782,25 +756,20 @@ static ssize_t btree_transaction_stats_read(struct file *file, char __user *buf,
!bch2_btree_transaction_fns[i->iter])
break;
prt_printf(&i->buf, "%s: ", bch2_btree_transaction_fns[i->iter]);
prt_newline(&i->buf);
prt_printf(&i->buf, "%s:\n", bch2_btree_transaction_fns[i->iter]);
printbuf_indent_add(&i->buf, 2);
mutex_lock(&s->lock);
prt_printf(&i->buf, "Max mem used: %u", s->max_mem);
prt_newline(&i->buf);
prt_printf(&i->buf, "Transaction duration:");
prt_newline(&i->buf);
prt_printf(&i->buf, "Max mem used: %u\n", s->max_mem);
prt_printf(&i->buf, "Transaction duration:\n");
printbuf_indent_add(&i->buf, 2);
bch2_time_stats_to_text(&i->buf, &s->duration);
printbuf_indent_sub(&i->buf, 2);
if (IS_ENABLED(CONFIG_BCACHEFS_LOCK_TIME_STATS)) {
prt_printf(&i->buf, "Lock hold times:");
prt_newline(&i->buf);
prt_printf(&i->buf, "Lock hold times:\n");
printbuf_indent_add(&i->buf, 2);
bch2_time_stats_to_text(&i->buf, &s->lock_hold_times);
@ -808,8 +777,7 @@ static ssize_t btree_transaction_stats_read(struct file *file, char __user *buf,
}
if (s->max_paths_text) {
prt_printf(&i->buf, "Maximum allocated btree paths (%u):", s->nr_max_paths);
prt_newline(&i->buf);
prt_printf(&i->buf, "Maximum allocated btree paths (%u):\n", s->nr_max_paths);
printbuf_indent_add(&i->buf, 2);
prt_str_indented(&i->buf, s->max_paths_text);

97
fs/bcachefs/dirent.c
View File

@ -98,7 +98,7 @@ const struct bch_hash_desc bch2_dirent_hash_desc = {
};
int bch2_dirent_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_s_c_dirent d = bkey_s_c_to_dirent(k);
@ -118,7 +118,7 @@ int bch2_dirent_invalid(struct bch_fs *c, struct bkey_s_c k,
* Check new keys don't exceed the max length
* (older keys may be larger.)
*/
bkey_fsck_err_on((flags & BKEY_INVALID_COMMIT) && d_name.len > BCH_NAME_MAX, c, err,
bkey_fsck_err_on((flags & BCH_VALIDATE_commit) && d_name.len > BCH_NAME_MAX, c, err,
dirent_name_too_long,
"dirent name too big (%u > %u)",
d_name.len, BCH_NAME_MAX);
@ -205,7 +205,7 @@ int bch2_dirent_create_snapshot(struct btree_trans *trans,
const struct bch_hash_info *hash_info,
u8 type, const struct qstr *name, u64 dst_inum,
u64 *dir_offset,
bch_str_hash_flags_t str_hash_flags)
enum btree_iter_update_trigger_flags flags)
{
subvol_inum dir_inum = { .subvol = dir_subvol, .inum = dir };
struct bkey_i_dirent *dirent;
@ -220,9 +220,8 @@ int bch2_dirent_create_snapshot(struct btree_trans *trans,
dirent->k.p.snapshot = snapshot;
ret = bch2_hash_set_in_snapshot(trans, bch2_dirent_hash_desc, hash_info,
dir_inum, snapshot,
&dirent->k_i, str_hash_flags,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
dir_inum, snapshot, &dirent->k_i,
flags|BTREE_UPDATE_internal_snapshot_node);
*dir_offset = dirent->k.p.offset;
return ret;
@ -232,7 +231,7 @@ int bch2_dirent_create(struct btree_trans *trans, subvol_inum dir,
const struct bch_hash_info *hash_info,
u8 type, const struct qstr *name, u64 dst_inum,
u64 *dir_offset,
bch_str_hash_flags_t str_hash_flags)
enum btree_iter_update_trigger_flags flags)
{
struct bkey_i_dirent *dirent;
int ret;
@ -243,7 +242,7 @@ int bch2_dirent_create(struct btree_trans *trans, subvol_inum dir,
return ret;
ret = bch2_hash_set(trans, bch2_dirent_hash_desc, hash_info,
dir, &dirent->k_i, str_hash_flags);
dir, &dirent->k_i, flags);
*dir_offset = dirent->k.p.offset;
return ret;
@ -272,7 +271,7 @@ int bch2_dirent_read_target(struct btree_trans *trans, subvol_inum dir,
} else {
target->subvol = le32_to_cpu(d.v->d_child_subvol);
ret = bch2_subvolume_get(trans, target->subvol, true, BTREE_ITER_CACHED, &s);
ret = bch2_subvolume_get(trans, target->subvol, true, BTREE_ITER_cached, &s);
target->inum = le64_to_cpu(s.inode);
}
@ -301,13 +300,9 @@ int bch2_dirent_rename(struct btree_trans *trans,
memset(dst_inum, 0, sizeof(*dst_inum));
/* Lookup src: */
ret = bch2_hash_lookup(trans, &src_iter, bch2_dirent_hash_desc,
src_hash, src_dir, src_name,
BTREE_ITER_INTENT);
if (ret)
goto out;
old_src = bch2_btree_iter_peek_slot(&src_iter);
old_src = bch2_hash_lookup(trans, &src_iter, bch2_dirent_hash_desc,
src_hash, src_dir, src_name,
BTREE_ITER_intent);
ret = bkey_err(old_src);
if (ret)
goto out;
@ -329,13 +324,9 @@ int bch2_dirent_rename(struct btree_trans *trans,
if (ret)
goto out;
} else {
ret = bch2_hash_lookup(trans, &dst_iter, bch2_dirent_hash_desc,
dst_hash, dst_dir, dst_name,
BTREE_ITER_INTENT);
if (ret)
goto out;
old_dst = bch2_btree_iter_peek_slot(&dst_iter);
old_dst = bch2_hash_lookup(trans, &dst_iter, bch2_dirent_hash_desc,
dst_hash, dst_dir, dst_name,
BTREE_ITER_intent);
ret = bkey_err(old_dst);
if (ret)
goto out;
@ -450,7 +441,7 @@ out_set_src:
if (delete_src) {
bch2_btree_iter_set_snapshot(&src_iter, old_src.k->p.snapshot);
ret = bch2_btree_iter_traverse(&src_iter) ?:
bch2_btree_delete_at(trans, &src_iter, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
bch2_btree_delete_at(trans, &src_iter, BTREE_UPDATE_internal_snapshot_node);
if (ret)
goto out;
}
@ -458,7 +449,7 @@ out_set_src:
if (delete_dst) {
bch2_btree_iter_set_snapshot(&dst_iter, old_dst.k->p.snapshot);
ret = bch2_btree_iter_traverse(&dst_iter) ?:
bch2_btree_delete_at(trans, &dst_iter, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
bch2_btree_delete_at(trans, &dst_iter, BTREE_UPDATE_internal_snapshot_node);
if (ret)
goto out;
}
@ -479,13 +470,9 @@ int bch2_dirent_lookup_trans(struct btree_trans *trans,
const struct qstr *name, subvol_inum *inum,
unsigned flags)
{
int ret = bch2_hash_lookup(trans, iter, bch2_dirent_hash_desc,
hash_info, dir, name, flags);
if (ret)
return ret;
struct bkey_s_c k = bch2_btree_iter_peek_slot(iter);
ret = bkey_err(k);
struct bkey_s_c k = bch2_hash_lookup(trans, iter, bch2_dirent_hash_desc,
hash_info, dir, name, flags);
int ret = bkey_err(k);
if (ret)
goto err;
@ -541,16 +528,26 @@ int bch2_empty_dir_trans(struct btree_trans *trans, subvol_inum dir)
bch2_empty_dir_snapshot(trans, dir.inum, dir.subvol, snapshot);
}
static int bch2_dir_emit(struct dir_context *ctx, struct bkey_s_c_dirent d, subvol_inum target)
{
struct qstr name = bch2_dirent_get_name(d);
bool ret = dir_emit(ctx, name.name,
name.len,
target.inum,
vfs_d_type(d.v->d_type));
if (ret)
ctx->pos = d.k->p.offset + 1;
return ret;
}
int bch2_readdir(struct bch_fs *c, subvol_inum inum, struct dir_context *ctx)
{
struct btree_trans *trans = bch2_trans_get(c);
struct btree_iter iter;
struct bkey_s_c k;
struct bkey_s_c_dirent dirent;
subvol_inum target;
u32 snapshot;
struct bkey_buf sk;
struct qstr name;
int ret;
bch2_bkey_buf_init(&sk);
@ -567,7 +564,9 @@ retry:
if (k.k->type != KEY_TYPE_dirent)
continue;
dirent = bkey_s_c_to_dirent(k);
/* dir_emit() can fault and block: */
bch2_bkey_buf_reassemble(&sk, c, k);
struct bkey_s_c_dirent dirent = bkey_i_to_s_c_dirent(sk.k);
ret = bch2_dirent_read_target(trans, inum, dirent, &target);
if (ret < 0)
@ -575,28 +574,22 @@ retry:
if (ret)
continue;
/* dir_emit() can fault and block: */
bch2_bkey_buf_reassemble(&sk, c, k);
dirent = bkey_i_to_s_c_dirent(sk.k);
bch2_trans_unlock(trans);
name = bch2_dirent_get_name(dirent);
ctx->pos = dirent.k->p.offset;
if (!dir_emit(ctx, name.name,
name.len,
target.inum,
vfs_d_type(dirent.v->d_type)))
break;
ctx->pos = dirent.k->p.offset + 1;
/*
* read_target looks up subvolumes, we can overflow paths if the
* directory has many subvolumes in it
*
* XXX: btree_trans_too_many_iters() is something we'd like to
* get rid of, and there's no good reason to be using it here
* except that we don't yet have a for_each_btree_key() helper
* that does subvolume_get_snapshot().
*/
ret = btree_trans_too_many_iters(trans);
if (ret)
ret = drop_locks_do(trans,
bch2_dir_emit(ctx, dirent, target)) ?:
btree_trans_too_many_iters(trans);
if (ret) {
ret = ret < 0 ? ret : 0;
break;
}
}
bch2_trans_iter_exit(trans, &iter);
err:

8
fs/bcachefs/dirent.h
View File

@ -4,11 +4,11 @@
#include "str_hash.h"
enum bkey_invalid_flags;
enum bch_validate_flags;
extern const struct bch_hash_desc bch2_dirent_hash_desc;
int bch2_dirent_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_dirent_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
#define bch2_bkey_ops_dirent ((struct bkey_ops) { \
@ -38,11 +38,11 @@ int bch2_dirent_read_target(struct btree_trans *, subvol_inum,
int bch2_dirent_create_snapshot(struct btree_trans *, u32, u64, u32,
const struct bch_hash_info *, u8,
const struct qstr *, u64, u64 *,
bch_str_hash_flags_t);
enum btree_iter_update_trigger_flags);
int bch2_dirent_create(struct btree_trans *, subvol_inum,
const struct bch_hash_info *, u8,
const struct qstr *, u64, u64 *,
bch_str_hash_flags_t);
enum btree_iter_update_trigger_flags);
static inline unsigned vfs_d_type(unsigned type)
{

11
fs/bcachefs/disk_groups.c
View File

@ -18,9 +18,8 @@ static int group_cmp(const void *_l, const void *_r)
strncmp(l->label, r->label, sizeof(l->label));
}
static int bch2_sb_disk_groups_validate(struct bch_sb *sb,
struct bch_sb_field *f,
struct printbuf *err)
static int bch2_sb_disk_groups_validate(struct bch_sb *sb, struct bch_sb_field *f,
enum bch_validate_flags flags, struct printbuf *err)
{
struct bch_sb_field_disk_groups *groups =
field_to_type(f, disk_groups);
@ -177,7 +176,7 @@ int bch2_sb_disk_groups_to_cpu(struct bch_fs *c)
struct bch_member m = bch2_sb_member_get(c->disk_sb.sb, i);
struct bch_disk_group_cpu *dst;
if (!bch2_member_exists(&m))
if (!bch2_member_alive(&m))
continue;
g = BCH_MEMBER_GROUP(&m);
@ -523,7 +522,7 @@ int bch2_opt_target_parse(struct bch_fs *c, const char *val, u64 *res,
ca = bch2_dev_lookup(c, val);
if (!IS_ERR(ca)) {
*res = dev_to_target(ca->dev_idx);
percpu_ref_put(&ca->ref);
bch2_dev_put(ca);
return 0;
}
@ -588,7 +587,7 @@ static void bch2_target_to_text_sb(struct printbuf *out, struct bch_sb *sb, unsi
case TARGET_DEV: {
struct bch_member m = bch2_sb_member_get(sb, t.dev);
if (bch2_dev_exists(sb, t.dev)) {
if (bch2_member_exists(sb, t.dev)) {
prt_printf(out, "Device ");
pr_uuid(out, m.uuid.b);
prt_printf(out, " (%u)", t.dev);

415
fs/bcachefs/ec.c
View File

@ -107,7 +107,7 @@ struct ec_bio {
/* Stripes btree keys: */
int bch2_stripe_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
@ -163,146 +163,189 @@ void bch2_stripe_to_text(struct printbuf *out, struct bch_fs *c,
/* Triggers: */
static int bch2_trans_mark_stripe_bucket(struct btree_trans *trans,
struct bkey_s_c_stripe s,
unsigned idx, bool deleting)
static int __mark_stripe_bucket(struct btree_trans *trans,
struct bch_dev *ca,
struct bkey_s_c_stripe s,
unsigned ptr_idx, bool deleting,
struct bpos bucket,
struct bch_alloc_v4 *a,
enum btree_iter_update_trigger_flags flags)
{
struct bch_fs *c = trans->c;
const struct bch_extent_ptr *ptr = &s.v->ptrs[idx];
struct btree_iter iter;
struct bkey_i_alloc_v4 *a;
enum bch_data_type data_type = idx >= s.v->nr_blocks - s.v->nr_redundant
? BCH_DATA_parity : 0;
s64 sectors = data_type ? le16_to_cpu(s.v->sectors) : 0;
int ret = 0;
if (deleting)
sectors = -sectors;
a = bch2_trans_start_alloc_update(trans, &iter, PTR_BUCKET_POS(c, ptr));
if (IS_ERR(a))
return PTR_ERR(a);
ret = bch2_check_bucket_ref(trans, s.s_c, ptr, sectors, data_type,
a->v.gen, a->v.data_type,
a->v.dirty_sectors);
if (ret)
goto err;
if (!deleting) {
if (bch2_trans_inconsistent_on(a->v.stripe ||
a->v.stripe_redundancy, trans,
"bucket %llu:%llu gen %u data type %s dirty_sectors %u: multiple stripes using same bucket (%u, %llu)",
iter.pos.inode, iter.pos.offset, a->v.gen,
bch2_data_type_str(a->v.data_type),
a->v.dirty_sectors,
a->v.stripe, s.k->p.offset)) {
ret = -EIO;
goto err;
}
if (bch2_trans_inconsistent_on(data_type && a->v.dirty_sectors, trans,
"bucket %llu:%llu gen %u data type %s dirty_sectors %u: data already in stripe bucket %llu",
iter.pos.inode, iter.pos.offset, a->v.gen,
bch2_data_type_str(a->v.data_type),
a->v.dirty_sectors,
s.k->p.offset)) {
ret = -EIO;
goto err;
}
a->v.stripe = s.k->p.offset;
a->v.stripe_redundancy = s.v->nr_redundant;
a->v.data_type = BCH_DATA_stripe;
} else {
if (bch2_trans_inconsistent_on(a->v.stripe != s.k->p.offset ||
a->v.stripe_redundancy != s.v->nr_redundant, trans,
"bucket %llu:%llu gen %u: not marked as stripe when deleting stripe %llu (got %u)",
iter.pos.inode, iter.pos.offset, a->v.gen,
s.k->p.offset, a->v.stripe)) {
ret = -EIO;
goto err;
}
a->v.stripe = 0;
a->v.stripe_redundancy = 0;
a->v.data_type = alloc_data_type(a->v, BCH_DATA_user);
}
a->v.dirty_sectors += sectors;
if (data_type)
a->v.data_type = !deleting ? data_type : 0;
ret = bch2_trans_update(trans, &iter, &a->k_i, 0);
if (ret)
goto err;
err:
bch2_trans_iter_exit(trans, &iter);
return ret;
}
static int mark_stripe_bucket(struct btree_trans *trans,
struct bkey_s_c k,
unsigned ptr_idx,
unsigned flags)
{
struct bch_fs *c = trans->c;
const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
unsigned nr_data = s->nr_blocks - s->nr_redundant;
const struct bch_extent_ptr *ptr = s.v->ptrs + ptr_idx;
unsigned nr_data = s.v->nr_blocks - s.v->nr_redundant;
bool parity = ptr_idx >= nr_data;
enum bch_data_type data_type = parity ? BCH_DATA_parity : BCH_DATA_stripe;
s64 sectors = parity ? le16_to_cpu(s->sectors) : 0;
const struct bch_extent_ptr *ptr = s->ptrs + ptr_idx;
struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
struct bucket old, new, *g;
s64 sectors = parity ? le16_to_cpu(s.v->sectors) : 0;
struct printbuf buf = PRINTBUF;
int ret = 0;
BUG_ON(!(flags & BTREE_TRIGGER_GC));
struct bch_fs *c = trans->c;
if (deleting)
sectors = -sectors;
/* * XXX doesn't handle deletion */
if (!deleting) {
if (bch2_trans_inconsistent_on(a->stripe ||
a->stripe_redundancy, trans,
"bucket %llu:%llu gen %u data type %s dirty_sectors %u: multiple stripes using same bucket (%u, %llu)\n%s",
bucket.inode, bucket.offset, a->gen,
bch2_data_type_str(a->data_type),
a->dirty_sectors,
a->stripe, s.k->p.offset,
(bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
ret = -EIO;
goto err;
}
percpu_down_read(&c->mark_lock);
g = PTR_GC_BUCKET(ca, ptr);
if (bch2_trans_inconsistent_on(parity && bch2_bucket_sectors_total(*a), trans,
"bucket %llu:%llu gen %u data type %s dirty_sectors %u cached_sectors %u: data already in parity bucket\n%s",
bucket.inode, bucket.offset, a->gen,
bch2_data_type_str(a->data_type),
a->dirty_sectors,
a->cached_sectors,
(bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
ret = -EIO;
goto err;
}
} else {
if (bch2_trans_inconsistent_on(a->stripe != s.k->p.offset ||
a->stripe_redundancy != s.v->nr_redundant, trans,
"bucket %llu:%llu gen %u: not marked as stripe when deleting stripe (got %u)\n%s",
bucket.inode, bucket.offset, a->gen,
a->stripe,
(bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
ret = -EIO;
goto err;
}
if (g->dirty_sectors ||
(g->stripe && g->stripe != k.k->p.offset)) {
bch2_fs_inconsistent(c,
"bucket %u:%zu gen %u: multiple stripes using same bucket\n%s",
ptr->dev, PTR_BUCKET_NR(ca, ptr), g->gen,
(bch2_bkey_val_to_text(&buf, c, k), buf.buf));
ret = -EINVAL;
goto err;
if (bch2_trans_inconsistent_on(a->data_type != data_type, trans,
"bucket %llu:%llu gen %u data type %s: wrong data type when stripe, should be %s\n%s",
bucket.inode, bucket.offset, a->gen,
bch2_data_type_str(a->data_type),
bch2_data_type_str(data_type),
(bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
ret = -EIO;
goto err;
}
if (bch2_trans_inconsistent_on(parity &&
(a->dirty_sectors != -sectors ||
a->cached_sectors), trans,
"bucket %llu:%llu gen %u dirty_sectors %u cached_sectors %u: wrong sectors when deleting parity block of stripe\n%s",
bucket.inode, bucket.offset, a->gen,
a->dirty_sectors,
a->cached_sectors,
(bch2_bkey_val_to_text(&buf, c, s.s_c), buf.buf))) {
ret = -EIO;
goto err;
}
}
bucket_lock(g);
old = *g;
if (sectors) {
ret = bch2_bucket_ref_update(trans, ca, s.s_c, ptr, sectors, data_type,
a->gen, a->data_type, &a->dirty_sectors);
if (ret)
goto err;
}
ret = bch2_check_bucket_ref(trans, k, ptr, sectors, data_type,
g->gen, g->data_type,
g->dirty_sectors);
if (ret)
goto err;
if (!deleting) {
a->stripe = s.k->p.offset;
a->stripe_redundancy = s.v->nr_redundant;
} else {
a->stripe = 0;
a->stripe_redundancy = 0;
}
g->data_type = data_type;
g->dirty_sectors += sectors;
g->stripe = k.k->p.offset;
g->stripe_redundancy = s->nr_redundant;
new = *g;
alloc_data_type_set(a, data_type);
err:
bucket_unlock(g);
if (!ret)
bch2_dev_usage_update_m(c, ca, &old, &new);
percpu_up_read(&c->mark_lock);
printbuf_exit(&buf);
return ret;
}
static int mark_stripe_bucket(struct btree_trans *trans,
struct bkey_s_c_stripe s,
unsigned ptr_idx, bool deleting,
enum btree_iter_update_trigger_flags flags)
{
struct bch_fs *c = trans->c;
const struct bch_extent_ptr *ptr = s.v->ptrs + ptr_idx;
int ret = 0;
struct bch_dev *ca = bch2_dev_tryget(c, ptr->dev);
if (unlikely(!ca)) {
if (!(flags & BTREE_TRIGGER_overwrite))
ret = -EIO;
goto err;
}
struct bpos bucket = PTR_BUCKET_POS(ca, ptr);
if (flags & BTREE_TRIGGER_transactional) {
struct bkey_i_alloc_v4 *a =
bch2_trans_start_alloc_update(trans, bucket);
ret = PTR_ERR_OR_ZERO(a) ?:
__mark_stripe_bucket(trans, ca, s, ptr_idx, deleting, bucket, &a->v, flags);
}
if (flags & BTREE_TRIGGER_gc) {
percpu_down_read(&c->mark_lock);
struct bucket *g = gc_bucket(ca, bucket.offset);
bucket_lock(g);
struct bch_alloc_v4 old = bucket_m_to_alloc(*g), new = old;
ret = __mark_stripe_bucket(trans, ca, s, ptr_idx, deleting, bucket, &new, flags);
if (!ret) {
alloc_to_bucket(g, new);
bch2_dev_usage_update(c, ca, &old, &new, 0, true);
}
bucket_unlock(g);
percpu_up_read(&c->mark_lock);
}
err:
bch2_dev_put(ca);
return ret;
}
static int mark_stripe_buckets(struct btree_trans *trans,
struct bkey_s_c old, struct bkey_s_c new,
enum btree_iter_update_trigger_flags flags)
{
const struct bch_stripe *old_s = old.k->type == KEY_TYPE_stripe
? bkey_s_c_to_stripe(old).v : NULL;
const struct bch_stripe *new_s = new.k->type == KEY_TYPE_stripe
? bkey_s_c_to_stripe(new).v : NULL;
BUG_ON(old_s && new_s && old_s->nr_blocks != new_s->nr_blocks);
unsigned nr_blocks = new_s ? new_s->nr_blocks : old_s->nr_blocks;
for (unsigned i = 0; i < nr_blocks; i++) {
if (new_s && old_s &&
!memcmp(&new_s->ptrs[i],
&old_s->ptrs[i],
sizeof(new_s->ptrs[i])))
continue;
if (new_s) {
int ret = mark_stripe_bucket(trans,
bkey_s_c_to_stripe(new), i, false, flags);
if (ret)
return ret;
}
if (old_s) {
int ret = mark_stripe_bucket(trans,
bkey_s_c_to_stripe(old), i, true, flags);
if (ret)
return ret;
}
}
return 0;
}
int bch2_trigger_stripe(struct btree_trans *trans,
enum btree_id btree_id, unsigned level,
enum btree_id btree, unsigned level,
struct bkey_s_c old, struct bkey_s _new,
unsigned flags)
enum btree_iter_update_trigger_flags flags)
{
struct bkey_s_c new = _new.s_c;
struct bch_fs *c = trans->c;
@ -312,7 +355,10 @@ int bch2_trigger_stripe(struct btree_trans *trans,
const struct bch_stripe *new_s = new.k->type == KEY_TYPE_stripe
? bkey_s_c_to_stripe(new).v : NULL;
if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
if (unlikely(flags & BTREE_TRIGGER_check_repair))
return bch2_check_fix_ptrs(trans, btree, level, _new.s_c, flags);
if (flags & BTREE_TRIGGER_transactional) {
/*
* If the pointers aren't changing, we don't need to do anything:
*/
@ -347,31 +393,12 @@ int bch2_trigger_stripe(struct btree_trans *trans,
return ret;
}
unsigned nr_blocks = new_s ? new_s->nr_blocks : old_s->nr_blocks;
for (unsigned i = 0; i < nr_blocks; i++) {
if (new_s && old_s &&
!memcmp(&new_s->ptrs[i],
&old_s->ptrs[i],
sizeof(new_s->ptrs[i])))
continue;
if (new_s) {
int ret = bch2_trans_mark_stripe_bucket(trans,
bkey_s_c_to_stripe(new), i, false);
if (ret)
return ret;
}
if (old_s) {
int ret = bch2_trans_mark_stripe_bucket(trans,
bkey_s_c_to_stripe(old), i, true);
if (ret)
return ret;
}
}
int ret = mark_stripe_buckets(trans, old, new, flags);
if (ret)
return ret;
}
if (flags & BTREE_TRIGGER_ATOMIC) {
if (flags & BTREE_TRIGGER_atomic) {
struct stripe *m = genradix_ptr(&c->stripes, idx);
if (!m) {
@ -410,7 +437,7 @@ int bch2_trigger_stripe(struct btree_trans *trans,
}
}
if (flags & BTREE_TRIGGER_GC) {
if (flags & BTREE_TRIGGER_gc) {
struct gc_stripe *m =
genradix_ptr_alloc(&c->gc_stripes, idx, GFP_KERNEL);
@ -439,13 +466,11 @@ int bch2_trigger_stripe(struct btree_trans *trans,
*/
memset(m->block_sectors, 0, sizeof(m->block_sectors));
for (unsigned i = 0; i < new_s->nr_blocks; i++) {
int ret = mark_stripe_bucket(trans, new, i, flags);
if (ret)
return ret;
}
int ret = mark_stripe_buckets(trans, old, new, flags);
if (ret)
return ret;
int ret = bch2_update_replicas(c, new, &m->r.e,
ret = bch2_update_replicas(c, new, &m->r.e,
((s64) m->sectors * m->nr_redundant),
0, true);
if (ret) {
@ -608,19 +633,21 @@ static void ec_validate_checksums(struct bch_fs *c, struct ec_stripe_buf *buf)
struct bch_csum got = ec_block_checksum(buf, i, offset);
if (bch2_crc_cmp(want, got)) {
struct printbuf err = PRINTBUF;
struct bch_dev *ca = bch_dev_bkey_exists(c, v->ptrs[i].dev);
struct bch_dev *ca = bch2_dev_tryget(c, v->ptrs[i].dev);
if (ca) {
struct printbuf err = PRINTBUF;
prt_str(&err, "stripe ");
bch2_csum_err_msg(&err, v->csum_type, want, got);
prt_printf(&err, " for %ps at %u of\n ", (void *) _RET_IP_, i);
bch2_bkey_val_to_text(&err, c, bkey_i_to_s_c(&buf->key));
bch_err_ratelimited(ca, "%s", err.buf);
printbuf_exit(&err);
prt_str(&err, "stripe ");
bch2_csum_err_msg(&err, v->csum_type, want, got);
prt_printf(&err, " for %ps at %u of\n ", (void *) _RET_IP_, i);
bch2_bkey_val_to_text(&err, c, bkey_i_to_s_c(&buf->key));
bch_err_ratelimited(ca, "%s", err.buf);
printbuf_exit(&err);
bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
}
clear_bit(i, buf->valid);
bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
break;
}
@ -687,7 +714,7 @@ static void ec_block_endio(struct bio *bio)
bch2_blk_status_to_str(bio->bi_status)))
clear_bit(ec_bio->idx, ec_bio->buf->valid);
if (ptr_stale(ca, ptr)) {
if (dev_ptr_stale(ca, ptr)) {
bch_err_ratelimited(ca->fs,
"error %s stripe: stale pointer after io",
bio_data_dir(bio) == READ ? "reading from" : "writing to");
@ -705,13 +732,18 @@ static void ec_block_io(struct bch_fs *c, struct ec_stripe_buf *buf,
struct bch_stripe *v = &bkey_i_to_stripe(&buf->key)->v;
unsigned offset = 0, bytes = buf->size << 9;
struct bch_extent_ptr *ptr = &v->ptrs[idx];
struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
enum bch_data_type data_type = idx < v->nr_blocks - v->nr_redundant
? BCH_DATA_user
: BCH_DATA_parity;
int rw = op_is_write(opf);
if (ptr_stale(ca, ptr)) {
struct bch_dev *ca = bch2_dev_get_ioref(c, ptr->dev, rw);
if (!ca) {
clear_bit(idx, buf->valid);
return;
}
if (dev_ptr_stale(ca, ptr)) {
bch_err_ratelimited(c,
"error %s stripe: stale pointer",
rw == READ ? "reading from" : "writing to");
@ -719,10 +751,6 @@ static void ec_block_io(struct bch_fs *c, struct ec_stripe_buf *buf,
return;
}
if (!bch2_dev_get_ioref(ca, rw)) {
clear_bit(idx, buf->valid);
return;
}
this_cpu_add(ca->io_done->sectors[rw][data_type], buf->size);
@ -769,7 +797,7 @@ static int get_stripe_key_trans(struct btree_trans *trans, u64 idx,
int ret;
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_stripes,
POS(0, idx), BTREE_ITER_SLOTS);
POS(0, idx), BTREE_ITER_slots);
ret = bkey_err(k);
if (ret)
goto err;
@ -1060,7 +1088,7 @@ static int ec_stripe_delete(struct btree_trans *trans, u64 idx)
int ret;
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_stripes, POS(0, idx),
BTREE_ITER_INTENT);
BTREE_ITER_intent);
ret = bkey_err(k);
if (ret)
goto err;
@ -1131,7 +1159,7 @@ static int ec_stripe_key_update(struct btree_trans *trans,
int ret;
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_stripes,
new->k.p, BTREE_ITER_INTENT);
new->k.p, BTREE_ITER_intent);
ret = bkey_err(k);
if (ret)
goto err;
@ -1173,6 +1201,7 @@ err:
}
static int ec_stripe_update_extent(struct btree_trans *trans,
struct bch_dev *ca,
struct bpos bucket, u8 gen,
struct ec_stripe_buf *s,
struct bpos *bp_pos)
@ -1183,13 +1212,13 @@ static int ec_stripe_update_extent(struct btree_trans *trans,
struct btree_iter iter;
struct bkey_s_c k;
const struct bch_extent_ptr *ptr_c;
struct bch_extent_ptr *ptr, *ec_ptr = NULL;
struct bch_extent_ptr *ec_ptr = NULL;
struct bch_extent_stripe_ptr stripe_ptr;
struct bkey_i *n;
int ret, dev, block;
ret = bch2_get_next_backpointer(trans, bucket, gen,
bp_pos, &bp, BTREE_ITER_CACHED);
ret = bch2_get_next_backpointer(trans, ca, bucket, gen,
bp_pos, &bp, BTREE_ITER_cached);
if (ret)
return ret;
if (bpos_eq(*bp_pos, SPOS_MAX))
@ -1214,7 +1243,7 @@ static int ec_stripe_update_extent(struct btree_trans *trans,
return -EIO;
}
k = bch2_backpointer_get_key(trans, &iter, *bp_pos, bp, BTREE_ITER_INTENT);
k = bch2_backpointer_get_key(trans, &iter, *bp_pos, bp, BTREE_ITER_intent);
ret = bkey_err(k);
if (ret)
return ret;
@ -1272,17 +1301,21 @@ static int ec_stripe_update_bucket(struct btree_trans *trans, struct ec_stripe_b
{
struct bch_fs *c = trans->c;
struct bch_stripe *v = &bkey_i_to_stripe(&s->key)->v;
struct bch_extent_ptr bucket = v->ptrs[block];
struct bpos bucket_pos = PTR_BUCKET_POS(c, &bucket);
struct bch_extent_ptr ptr = v->ptrs[block];
struct bpos bp_pos = POS_MIN;
int ret = 0;
struct bch_dev *ca = bch2_dev_tryget(c, ptr.dev);
if (!ca)
return -EIO;
struct bpos bucket_pos = PTR_BUCKET_POS(ca, &ptr);
while (1) {
ret = commit_do(trans, NULL, NULL,
BCH_TRANS_COMMIT_no_check_rw|
BCH_TRANS_COMMIT_no_enospc,
ec_stripe_update_extent(trans, bucket_pos, bucket.gen,
s, &bp_pos));
ec_stripe_update_extent(trans, ca, bucket_pos, ptr.gen, s, &bp_pos));
if (ret)
break;
if (bkey_eq(bp_pos, POS_MAX))
@ -1291,6 +1324,7 @@ static int ec_stripe_update_bucket(struct btree_trans *trans, struct ec_stripe_b
bp_pos = bpos_nosnap_successor(bp_pos);
}
bch2_dev_put(ca);
return ret;
}
@ -1321,20 +1355,18 @@ static void zero_out_rest_of_ec_bucket(struct bch_fs *c,
unsigned block,
struct open_bucket *ob)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
unsigned offset = ca->mi.bucket_size - ob->sectors_free;
int ret;
if (!bch2_dev_get_ioref(ca, WRITE)) {
struct bch_dev *ca = bch2_dev_get_ioref(c, ob->dev, WRITE);
if (!ca) {
s->err = -BCH_ERR_erofs_no_writes;
return;
}
unsigned offset = ca->mi.bucket_size - ob->sectors_free;
memset(s->new_stripe.data[block] + (offset << 9),
0,
ob->sectors_free << 9);
ret = blkdev_issue_zeroout(ca->disk_sb.bdev,
int ret = blkdev_issue_zeroout(ca->disk_sb.bdev,
ob->bucket * ca->mi.bucket_size + offset,
ob->sectors_free,
GFP_KERNEL, 0);
@ -1519,16 +1551,13 @@ void bch2_ec_bucket_cancel(struct bch_fs *c, struct open_bucket *ob)
void *bch2_writepoint_ec_buf(struct bch_fs *c, struct write_point *wp)
{
struct open_bucket *ob = ec_open_bucket(c, &wp->ptrs);
struct bch_dev *ca;
unsigned offset;
if (!ob)
return NULL;
BUG_ON(!ob->ec->new_stripe.data[ob->ec_idx]);
ca = bch_dev_bkey_exists(c, ob->dev);
offset = ca->mi.bucket_size - ob->sectors_free;
struct bch_dev *ca = ob_dev(c, ob);
unsigned offset = ca->mi.bucket_size - ob->sectors_free;
return ob->ec->new_stripe.data[ob->ec_idx] + (offset << 9);
}
@ -1937,7 +1966,7 @@ static int __bch2_ec_stripe_head_reserve(struct btree_trans *trans, struct ec_st
}
for_each_btree_key_norestart(trans, iter, BTREE_ID_stripes, start_pos,
BTREE_ITER_SLOTS|BTREE_ITER_INTENT, k, ret) {
BTREE_ITER_slots|BTREE_ITER_intent, k, ret) {
if (bkey_gt(k.k->p, POS(0, U32_MAX))) {
if (start_pos.offset) {
start_pos = min_pos;
@ -2127,7 +2156,7 @@ int bch2_stripes_read(struct bch_fs *c)
{
int ret = bch2_trans_run(c,
for_each_btree_key(trans, iter, BTREE_ID_stripes, POS_MIN,
BTREE_ITER_PREFETCH, k, ({
BTREE_ITER_prefetch, k, ({
if (k.k->type != KEY_TYPE_stripe)
continue;

7
fs/bcachefs/ec.h
View File

@ -6,14 +6,15 @@
#include "buckets_types.h"
#include "extents_types.h"
enum bkey_invalid_flags;
enum bch_validate_flags;
int bch2_stripe_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_stripe_to_text(struct printbuf *, struct bch_fs *,
struct bkey_s_c);
int bch2_trigger_stripe(struct btree_trans *, enum btree_id, unsigned,
struct bkey_s_c, struct bkey_s, unsigned);
struct bkey_s_c, struct bkey_s,
enum btree_iter_update_trigger_flags);
#define bch2_bkey_ops_stripe ((struct bkey_ops) { \
.key_invalid = bch2_stripe_invalid, \

59
fs/bcachefs/error.c
View File

@ -176,6 +176,21 @@ static struct fsck_err_state *fsck_err_get(struct bch_fs *c, const char *fmt)
return s;
}
/* s/fix?/fixing/ s/recreate?/recreating/ */
static void prt_actioning(struct printbuf *out, const char *action)
{
unsigned len = strlen(action);
BUG_ON(action[len - 1] != '?');
--len;
if (action[len - 1] == 'e')
--len;
prt_bytes(out, action, len);
prt_str(out, "ing");
}
int bch2_fsck_err(struct bch_fs *c,
enum bch_fsck_flags flags,
enum bch_sb_error_id err,
@ -186,6 +201,7 @@ int bch2_fsck_err(struct bch_fs *c,
bool print = true, suppressing = false, inconsistent = false;
struct printbuf buf = PRINTBUF, *out = &buf;
int ret = -BCH_ERR_fsck_ignore;
const char *action_orig = "fix?", *action = action_orig;
if ((flags & FSCK_CAN_FIX) &&
test_bit(err, c->sb.errors_silent))
@ -197,6 +213,19 @@ int bch2_fsck_err(struct bch_fs *c,
prt_vprintf(out, fmt, args);
va_end(args);
/* Custom fix/continue/recreate/etc.? */
if (out->buf[out->pos - 1] == '?') {
const char *p = strrchr(out->buf, ',');
if (p) {
out->pos = p - out->buf;
action = kstrdup(p + 2, GFP_KERNEL);
if (!action) {
ret = -ENOMEM;
goto err;
}
}
}
mutex_lock(&c->fsck_error_msgs_lock);
s = fsck_err_get(c, fmt);
if (s) {
@ -208,12 +237,16 @@ int bch2_fsck_err(struct bch_fs *c,
if (s->last_msg && !strcmp(buf.buf, s->last_msg)) {
ret = s->ret;
mutex_unlock(&c->fsck_error_msgs_lock);
printbuf_exit(&buf);
return ret;
goto err;
}
kfree(s->last_msg);
s->last_msg = kstrdup(buf.buf, GFP_KERNEL);
if (!s->last_msg) {
mutex_unlock(&c->fsck_error_msgs_lock);
ret = -ENOMEM;
goto err;
}
if (c->opts.ratelimit_errors &&
!(flags & FSCK_NO_RATELIMIT) &&
@ -239,7 +272,8 @@ int bch2_fsck_err(struct bch_fs *c,
inconsistent = true;
ret = -BCH_ERR_fsck_errors_not_fixed;
} else if (flags & FSCK_CAN_FIX) {
prt_str(out, ", fixing");
prt_str(out, ", ");
prt_actioning(out, action);
ret = -BCH_ERR_fsck_fix;
} else {
prt_str(out, ", continuing");
@ -254,16 +288,16 @@ int bch2_fsck_err(struct bch_fs *c,
: c->opts.fix_errors;
if (fix == FSCK_FIX_ask) {
int ask;
prt_str(out, ", ");
prt_str(out, action);
prt_str(out, ": fix?");
if (bch2_fs_stdio_redirect(c))
bch2_print(c, "%s", out->buf);
else
bch2_print_string_as_lines(KERN_ERR, out->buf);
print = false;
ask = bch2_fsck_ask_yn(c);
int ask = bch2_fsck_ask_yn(c);
if (ask >= YN_ALLNO && s)
s->fix = ask == YN_ALLNO
@ -276,10 +310,12 @@ int bch2_fsck_err(struct bch_fs *c,
} else if (fix == FSCK_FIX_yes ||
(c->opts.nochanges &&
!(flags & FSCK_CAN_IGNORE))) {
prt_str(out, ", fixing");
prt_str(out, ", ");
prt_actioning(out, action);
ret = -BCH_ERR_fsck_fix;
} else {
prt_str(out, ", not fixing");
prt_str(out, ", not ");
prt_actioning(out, action);
}
} else if (flags & FSCK_NEED_FSCK) {
prt_str(out, " (run fsck to correct)");
@ -311,8 +347,6 @@ int bch2_fsck_err(struct bch_fs *c,
mutex_unlock(&c->fsck_error_msgs_lock);
printbuf_exit(&buf);
if (inconsistent)
bch2_inconsistent_error(c);
@ -322,7 +356,10 @@ int bch2_fsck_err(struct bch_fs *c,
set_bit(BCH_FS_errors_not_fixed, &c->flags);
set_bit(BCH_FS_error, &c->flags);
}
err:
if (action != action_orig)
kfree(action);
printbuf_exit(&buf);
return ret;
}

2
fs/bcachefs/extent_update.c
View File

@ -72,7 +72,7 @@ static int count_iters_for_insert(struct btree_trans *trans,
for_each_btree_key_norestart(trans, iter,
BTREE_ID_reflink, POS(0, idx + offset),
BTREE_ITER_SLOTS, r_k, ret2) {
BTREE_ITER_slots, r_k, ret2) {
if (bkey_ge(bkey_start_pos(r_k.k), POS(0, idx + sectors)))
break;

151
fs/bcachefs/extents.c
View File

@ -71,6 +71,12 @@ void bch2_mark_io_failure(struct bch_io_failures *failed,
}
}
static inline u64 dev_latency(struct bch_fs *c, unsigned dev)
{
struct bch_dev *ca = bch2_dev_rcu(c, dev);
return ca ? atomic64_read(&ca->cur_latency[READ]) : S64_MAX;
}
/*
* returns true if p1 is better than p2:
*/
@ -79,11 +85,8 @@ static inline bool ptr_better(struct bch_fs *c,
const struct extent_ptr_decoded p2)
{
if (likely(!p1.idx && !p2.idx)) {
struct bch_dev *dev1 = bch_dev_bkey_exists(c, p1.ptr.dev);
struct bch_dev *dev2 = bch_dev_bkey_exists(c, p2.ptr.dev);
u64 l1 = atomic64_read(&dev1->cur_latency[READ]);
u64 l2 = atomic64_read(&dev2->cur_latency[READ]);
u64 l1 = dev_latency(c, p1.ptr.dev);
u64 l2 = dev_latency(c, p2.ptr.dev);
/* Pick at random, biased in favor of the faster device: */
@ -109,21 +112,21 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k,
const union bch_extent_entry *entry;
struct extent_ptr_decoded p;
struct bch_dev_io_failures *f;
struct bch_dev *ca;
int ret = 0;
if (k.k->type == KEY_TYPE_error)
return -EIO;
rcu_read_lock();
bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
/*
* Unwritten extent: no need to actually read, treat it as a
* hole and return 0s:
*/
if (p.ptr.unwritten)
return 0;
ca = bch_dev_bkey_exists(c, p.ptr.dev);
if (p.ptr.unwritten) {
ret = 0;
break;
}
/*
* If there are any dirty pointers it's an error if we can't
@ -132,7 +135,9 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k,
if (!ret && !p.ptr.cached)
ret = -EIO;
if (p.ptr.cached && ptr_stale(ca, &p.ptr))
struct bch_dev *ca = bch2_dev_rcu(c, p.ptr.dev);
if (p.ptr.cached && (!ca || dev_ptr_stale(ca, &p.ptr)))
continue;
f = failed ? dev_io_failures(failed, p.ptr.dev) : NULL;
@ -141,12 +146,13 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k,
? f->idx
: f->idx + 1;
if (!p.idx &&
!bch2_dev_is_readable(ca))
if (!p.idx && !ca)
p.idx++;
if (bch2_force_reconstruct_read &&
!p.idx && p.has_ec)
if (!p.idx && p.has_ec && bch2_force_reconstruct_read)
p.idx++;
if (!p.idx && !bch2_dev_is_readable(ca))
p.idx++;
if (p.idx >= (unsigned) p.has_ec + 1)
@ -158,6 +164,7 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k,
*pick = p;
ret = 1;
}
rcu_read_unlock();
return ret;
}
@ -165,7 +172,7 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k,
/* KEY_TYPE_btree_ptr: */
int bch2_btree_ptr_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
int ret = 0;
@ -186,7 +193,7 @@ void bch2_btree_ptr_to_text(struct printbuf *out, struct bch_fs *c,
}
int bch2_btree_ptr_v2_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_s_c_btree_ptr_v2 bp = bkey_s_c_to_btree_ptr_v2(k);
@ -201,6 +208,11 @@ int bch2_btree_ptr_v2_invalid(struct bch_fs *c, struct bkey_s_c k,
c, err, btree_ptr_v2_min_key_bad,
"min_key > key");
if (flags & BCH_VALIDATE_write)
bkey_fsck_err_on(!bp.v->sectors_written,
c, err, btree_ptr_v2_written_0,
"sectors_written == 0");
ret = bch2_bkey_ptrs_invalid(c, k, flags, err);
fsck_err:
return ret;
@ -247,7 +259,6 @@ bool bch2_extent_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r)
const union bch_extent_entry *en_r;
struct extent_ptr_decoded lp, rp;
bool use_right_ptr;
struct bch_dev *ca;
en_l = l_ptrs.start;
en_r = r_ptrs.start;
@ -278,8 +289,12 @@ bool bch2_extent_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r)
return false;
/* Extents may not straddle buckets: */
ca = bch_dev_bkey_exists(c, lp.ptr.dev);
if (PTR_BUCKET_NR(ca, &lp.ptr) != PTR_BUCKET_NR(ca, &rp.ptr))
rcu_read_lock();
struct bch_dev *ca = bch2_dev_rcu(c, lp.ptr.dev);
bool same_bucket = ca && PTR_BUCKET_NR(ca, &lp.ptr) == PTR_BUCKET_NR(ca, &rp.ptr);
rcu_read_unlock();
if (!same_bucket)
return false;
if (lp.has_ec != rp.has_ec ||
@ -385,7 +400,7 @@ bool bch2_extent_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r)
/* KEY_TYPE_reservation: */
int bch2_reservation_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_s_c_reservation r = bkey_s_c_to_reservation(k);
@ -667,16 +682,16 @@ static inline unsigned __extent_ptr_durability(struct bch_dev *ca, struct extent
unsigned bch2_extent_ptr_desired_durability(struct bch_fs *c, struct extent_ptr_decoded *p)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, p->ptr.dev);
struct bch_dev *ca = bch2_dev_rcu(c, p->ptr.dev);
return __extent_ptr_durability(ca, p);
return ca ? __extent_ptr_durability(ca, p) : 0;
}
unsigned bch2_extent_ptr_durability(struct bch_fs *c, struct extent_ptr_decoded *p)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, p->ptr.dev);
struct bch_dev *ca = bch2_dev_rcu(c, p->ptr.dev);
if (ca->mi.state == BCH_MEMBER_STATE_failed)
if (!ca || ca->mi.state == BCH_MEMBER_STATE_failed)
return 0;
return __extent_ptr_durability(ca, p);
@ -689,8 +704,10 @@ unsigned bch2_bkey_durability(struct bch_fs *c, struct bkey_s_c k)
struct extent_ptr_decoded p;
unsigned durability = 0;
rcu_read_lock();
bkey_for_each_ptr_decode(k.k, ptrs, p, entry)
durability += bch2_extent_ptr_durability(c, &p);
rcu_read_unlock();
return durability;
}
@ -702,9 +719,11 @@ static unsigned bch2_bkey_durability_safe(struct bch_fs *c, struct bkey_s_c k)
struct extent_ptr_decoded p;
unsigned durability = 0;
rcu_read_lock();
bkey_for_each_ptr_decode(k.k, ptrs, p, entry)
if (p.ptr.dev < c->sb.nr_devices && c->devs[p.ptr.dev])
durability += bch2_extent_ptr_durability(c, &p);
rcu_read_unlock();
return durability;
}
@ -833,8 +852,6 @@ union bch_extent_entry *bch2_bkey_drop_ptr(struct bkey_s k,
void bch2_bkey_drop_device(struct bkey_s k, unsigned dev)
{
struct bch_extent_ptr *ptr;
bch2_bkey_drop_ptrs(k, ptr, ptr->dev == dev);
}
@ -860,14 +877,21 @@ const struct bch_extent_ptr *bch2_bkey_has_device_c(struct bkey_s_c k, unsigned
bool bch2_bkey_has_target(struct bch_fs *c, struct bkey_s_c k, unsigned target)
{
struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
struct bch_dev *ca;
bool ret = false;
rcu_read_lock();
bkey_for_each_ptr(ptrs, ptr)
if (bch2_dev_in_target(c, ptr->dev, target) &&
(ca = bch2_dev_rcu(c, ptr->dev)) &&
(!ptr->cached ||
!ptr_stale(bch_dev_bkey_exists(c, ptr->dev), ptr)))
return true;
!dev_ptr_stale_rcu(ca, ptr))) {
ret = true;
break;
}
rcu_read_unlock();
return false;
return ret;
}
bool bch2_bkey_matches_ptr(struct bch_fs *c, struct bkey_s_c k,
@ -969,21 +993,23 @@ void bch2_extent_ptr_set_cached(struct bkey_s k, struct bch_extent_ptr *ptr)
*/
bool bch2_extent_normalize(struct bch_fs *c, struct bkey_s k)
{
struct bch_extent_ptr *ptr;
struct bch_dev *ca;
rcu_read_lock();
bch2_bkey_drop_ptrs(k, ptr,
ptr->cached &&
ptr_stale(bch_dev_bkey_exists(c, ptr->dev), ptr));
(ca = bch2_dev_rcu(c, ptr->dev)) &&
dev_ptr_stale_rcu(ca, ptr));
rcu_read_unlock();
return bkey_deleted(k.k);
}
void bch2_extent_ptr_to_text(struct printbuf *out, struct bch_fs *c, const struct bch_extent_ptr *ptr)
{
struct bch_dev *ca = c && ptr->dev < c->sb.nr_devices && c->devs[ptr->dev]
? bch_dev_bkey_exists(c, ptr->dev)
: NULL;
out->atomic++;
rcu_read_lock();
struct bch_dev *ca = bch2_dev_rcu(c, ptr->dev);
if (!ca) {
prt_printf(out, "ptr: %u:%llu gen %u%s", ptr->dev,
(u64) ptr->offset, ptr->gen,
@ -998,11 +1024,11 @@ void bch2_extent_ptr_to_text(struct printbuf *out, struct bch_fs *c, const struc
prt_str(out, " cached");
if (ptr->unwritten)
prt_str(out, " unwritten");
if (b >= ca->mi.first_bucket &&
b < ca->mi.nbuckets &&
ptr_stale(ca, ptr))
if (bucket_valid(ca, b) && dev_ptr_stale_rcu(ca, ptr))
prt_printf(out, " stale");
}
rcu_read_unlock();
--out->atomic;
}
void bch2_bkey_ptrs_to_text(struct printbuf *out, struct bch_fs *c,
@ -1069,55 +1095,50 @@ void bch2_bkey_ptrs_to_text(struct printbuf *out, struct bch_fs *c,
static int extent_ptr_invalid(struct bch_fs *c,
struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
const struct bch_extent_ptr *ptr,
unsigned size_ondisk,
bool metadata,
struct printbuf *err)
{
struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
u64 bucket;
u32 bucket_offset;
struct bch_dev *ca;
int ret = 0;
if (!bch2_dev_exists2(c, ptr->dev)) {
/*
* If we're in the write path this key might have already been
* overwritten, and we could be seeing a device that doesn't
* exist anymore due to racing with device removal:
*/
if (flags & BKEY_INVALID_WRITE)
return 0;
bkey_fsck_err(c, err, ptr_to_invalid_device,
"pointer to invalid device (%u)", ptr->dev);
rcu_read_lock();
struct bch_dev *ca = bch2_dev_rcu(c, ptr->dev);
if (!ca) {
rcu_read_unlock();
return 0;
}
u32 bucket_offset;
u64 bucket = sector_to_bucket_and_offset(ca, ptr->offset, &bucket_offset);
unsigned first_bucket = ca->mi.first_bucket;
u64 nbuckets = ca->mi.nbuckets;
unsigned bucket_size = ca->mi.bucket_size;
rcu_read_unlock();
ca = bch_dev_bkey_exists(c, ptr->dev);
struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
bkey_for_each_ptr(ptrs, ptr2)
bkey_fsck_err_on(ptr != ptr2 && ptr->dev == ptr2->dev, c, err,
ptr_to_duplicate_device,
"multiple pointers to same device (%u)", ptr->dev);
bucket = sector_to_bucket_and_offset(ca, ptr->offset, &bucket_offset);
bkey_fsck_err_on(bucket >= ca->mi.nbuckets, c, err,
bkey_fsck_err_on(bucket >= nbuckets, c, err,
ptr_after_last_bucket,
"pointer past last bucket (%llu > %llu)", bucket, ca->mi.nbuckets);
bkey_fsck_err_on(ptr->offset < bucket_to_sector(ca, ca->mi.first_bucket), c, err,
"pointer past last bucket (%llu > %llu)", bucket, nbuckets);
bkey_fsck_err_on(bucket < first_bucket, c, err,
ptr_before_first_bucket,
"pointer before first bucket (%llu < %u)", bucket, ca->mi.first_bucket);
bkey_fsck_err_on(bucket_offset + size_ondisk > ca->mi.bucket_size, c, err,
"pointer before first bucket (%llu < %u)", bucket, first_bucket);
bkey_fsck_err_on(bucket_offset + size_ondisk > bucket_size, c, err,
ptr_spans_multiple_buckets,
"pointer spans multiple buckets (%u + %u > %u)",
bucket_offset, size_ondisk, ca->mi.bucket_size);
bucket_offset, size_ondisk, bucket_size);
fsck_err:
return ret;
}
int bch2_bkey_ptrs_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
@ -1193,7 +1214,7 @@ int bch2_bkey_ptrs_invalid(struct bch_fs *c, struct bkey_s_c k,
bkey_fsck_err_on(crc_is_encoded(crc) &&
(crc.uncompressed_size > c->opts.encoded_extent_max >> 9) &&
(flags & (BKEY_INVALID_WRITE|BKEY_INVALID_COMMIT)), c, err,
(flags & (BCH_VALIDATE_write|BCH_VALIDATE_commit)), c, err,
ptr_crc_uncompressed_size_too_big,
"too large encoded extent");

12
fs/bcachefs/extents.h
View File

@ -8,7 +8,7 @@
struct bch_fs;
struct btree_trans;
enum bkey_invalid_flags;
enum bch_validate_flags;
/* extent entries: */
@ -406,12 +406,12 @@ int bch2_bkey_pick_read_device(struct bch_fs *, struct bkey_s_c,
/* KEY_TYPE_btree_ptr: */
int bch2_btree_ptr_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_btree_ptr_to_text(struct printbuf *, struct bch_fs *,
struct bkey_s_c);
int bch2_btree_ptr_v2_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_btree_ptr_v2_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
void bch2_btree_ptr_v2_compat(enum btree_id, unsigned, unsigned,
int, struct bkey_s);
@ -448,7 +448,7 @@ bool bch2_extent_merge(struct bch_fs *, struct bkey_s, struct bkey_s_c);
/* KEY_TYPE_reservation: */
int bch2_reservation_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_reservation_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
bool bch2_reservation_merge(struct bch_fs *, struct bkey_s, struct bkey_s_c);
@ -654,7 +654,7 @@ union bch_extent_entry *bch2_bkey_drop_ptr(struct bkey_s,
do { \
struct bkey_ptrs _ptrs = bch2_bkey_ptrs(_k); \
\
_ptr = &_ptrs.start->ptr; \
struct bch_extent_ptr *_ptr = &_ptrs.start->ptr; \
\
while ((_ptr = bkey_ptr_next(_ptrs, _ptr))) { \
if (_cond) { \
@ -680,7 +680,7 @@ void bch2_extent_ptr_to_text(struct printbuf *out, struct bch_fs *, const struct
void bch2_bkey_ptrs_to_text(struct printbuf *, struct bch_fs *,
struct bkey_s_c);
int bch2_bkey_ptrs_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_ptr_swab(struct bkey_s);

105
fs/bcachefs/eytzinger.c
View File

@ -171,7 +171,7 @@ void eytzinger0_sort_r(void *base, size_t n, size_t size,
swap_r_func_t swap_func,
const void *priv)
{
int i, c, r;
int i, j, k;
/* called from 'sort' without swap function, let's pick the default */
if (swap_func == SWAP_WRAPPER && !((struct wrapper *)priv)->swap_func)
@ -188,17 +188,22 @@ void eytzinger0_sort_r(void *base, size_t n, size_t size,
/* heapify */
for (i = n / 2 - 1; i >= 0; --i) {
for (r = i; r * 2 + 1 < n; r = c) {
c = r * 2 + 1;
/* Find the sift-down path all the way to the leaves. */
for (j = i; k = j * 2 + 1, k + 1 < n;)
j = eytzinger0_do_cmp(base, n, size, cmp_func, priv, k, k + 1) > 0 ? k : k + 1;
if (c + 1 < n &&
eytzinger0_do_cmp(base, n, size, cmp_func, priv, c, c + 1) < 0)
c++;
/* Special case for the last leaf with no sibling. */
if (j * 2 + 2 == n)
j = j * 2 + 1;
if (eytzinger0_do_cmp(base, n, size, cmp_func, priv, r, c) >= 0)
break;
/* Backtrack to the correct location. */
while (j != i && eytzinger0_do_cmp(base, n, size, cmp_func, priv, i, j) >= 0)
j = (j - 1) / 2;
eytzinger0_do_swap(base, n, size, swap_func, priv, r, c);
/* Shift the element into its correct place. */
for (k = j; j != i;) {
j = (j - 1) / 2;
eytzinger0_do_swap(base, n, size, swap_func, priv, j, k);
}
}
@ -206,17 +211,22 @@ void eytzinger0_sort_r(void *base, size_t n, size_t size,
for (i = n - 1; i > 0; --i) {
eytzinger0_do_swap(base, n, size, swap_func, priv, 0, i);
for (r = 0; r * 2 + 1 < i; r = c) {
c = r * 2 + 1;
/* Find the sift-down path all the way to the leaves. */
for (j = 0; k = j * 2 + 1, k + 1 < i;)
j = eytzinger0_do_cmp(base, n, size, cmp_func, priv, k, k + 1) > 0 ? k : k + 1;
if (c + 1 < i &&
eytzinger0_do_cmp(base, n, size, cmp_func, priv, c, c + 1) < 0)
c++;
/* Special case for the last leaf with no sibling. */
if (j * 2 + 2 == i)
j = j * 2 + 1;
if (eytzinger0_do_cmp(base, n, size, cmp_func, priv, r, c) >= 0)
break;
/* Backtrack to the correct location. */
while (j && eytzinger0_do_cmp(base, n, size, cmp_func, priv, 0, j) >= 0)
j = (j - 1) / 2;
eytzinger0_do_swap(base, n, size, swap_func, priv, r, c);
/* Shift the element into its correct place. */
for (k = j; j;) {
j = (j - 1) / 2;
eytzinger0_do_swap(base, n, size, swap_func, priv, j, k);
}
}
}
@ -232,3 +242,64 @@ void eytzinger0_sort(void *base, size_t n, size_t size,
return eytzinger0_sort_r(base, n, size, _CMP_WRAPPER, SWAP_WRAPPER, &w);
}
#if 0
#include <linux/slab.h>
#include <linux/random.h>
#include <linux/ktime.h>
static u64 cmp_count;
static int mycmp(const void *a, const void *b)
{
u32 _a = *(u32 *)a;
u32 _b = *(u32 *)b;
cmp_count++;
if (_a < _b)
return -1;
else if (_a > _b)
return 1;
else
return 0;
}
static int test(void)
{
size_t N, i;
ktime_t start, end;
s64 delta;
u32 *arr;
for (N = 10000; N <= 100000; N += 10000) {
arr = kmalloc_array(N, sizeof(u32), GFP_KERNEL);
cmp_count = 0;
for (i = 0; i < N; i++)
arr[i] = get_random_u32();
start = ktime_get();
eytzinger0_sort(arr, N, sizeof(u32), mycmp, NULL);
end = ktime_get();
delta = ktime_us_delta(end, start);
printk(KERN_INFO "time: %lld\n", delta);
printk(KERN_INFO "comparisons: %lld\n", cmp_count);
u32 prev = 0;
eytzinger0_for_each(i, N) {
if (prev > arr[i])
goto err;
prev = arr[i];
}
kfree(arr);
}
return 0;
err:
kfree(arr);
return -1;
}
#endif

38
fs/bcachefs/fs-common.c
View File

@ -42,7 +42,7 @@ int bch2_create_trans(struct btree_trans *trans,
if (ret)
goto err;
ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_INTENT);
ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_intent);
if (ret)
goto err;
@ -70,7 +70,7 @@ int bch2_create_trans(struct btree_trans *trans,
struct bch_subvolume s;
ret = bch2_subvolume_get(trans, snapshot_src.subvol, true,
BTREE_ITER_CACHED, &s);
BTREE_ITER_cached, &s);
if (ret)
goto err;
@ -78,7 +78,7 @@ int bch2_create_trans(struct btree_trans *trans,
}
ret = bch2_inode_peek(trans, &inode_iter, new_inode, snapshot_src,
BTREE_ITER_INTENT);
BTREE_ITER_intent);
if (ret)
goto err;
@ -163,7 +163,7 @@ int bch2_create_trans(struct btree_trans *trans,
name,
dir_target,
&dir_offset,
BCH_HASH_SET_MUST_CREATE);
STR_HASH_must_create);
if (ret)
goto err;
@ -171,7 +171,7 @@ int bch2_create_trans(struct btree_trans *trans,
new_inode->bi_dir_offset = dir_offset;
}
inode_iter.flags &= ~BTREE_ITER_ALL_SNAPSHOTS;
inode_iter.flags &= ~BTREE_ITER_all_snapshots;
bch2_btree_iter_set_snapshot(&inode_iter, snapshot);
ret = bch2_btree_iter_traverse(&inode_iter) ?:
@ -198,16 +198,16 @@ int bch2_link_trans(struct btree_trans *trans,
if (dir.subvol != inum.subvol)
return -EXDEV;
ret = bch2_inode_peek(trans, &inode_iter, inode_u, inum, BTREE_ITER_INTENT);
ret = bch2_inode_peek(trans, &inode_iter, inode_u, inum, BTREE_ITER_intent);
if (ret)
goto err;
return ret;
inode_u->bi_ctime = now;
ret = bch2_inode_nlink_inc(inode_u);
if (ret)
return ret;
goto err;
ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_INTENT);
ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_intent);
if (ret)
goto err;
@ -223,7 +223,7 @@ int bch2_link_trans(struct btree_trans *trans,
ret = bch2_dirent_create(trans, dir, &dir_hash,
mode_to_type(inode_u->bi_mode),
name, inum.inum, &dir_offset,
BCH_HASH_SET_MUST_CREATE);
STR_HASH_must_create);
if (ret)
goto err;
@ -255,19 +255,19 @@ int bch2_unlink_trans(struct btree_trans *trans,
struct bkey_s_c k;
int ret;
ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_INTENT);
ret = bch2_inode_peek(trans, &dir_iter, dir_u, dir, BTREE_ITER_intent);
if (ret)
goto err;
dir_hash = bch2_hash_info_init(c, dir_u);
ret = bch2_dirent_lookup_trans(trans, &dirent_iter, dir, &dir_hash,
name, &inum, BTREE_ITER_INTENT);
name, &inum, BTREE_ITER_intent);
if (ret)
goto err;
ret = bch2_inode_peek(trans, &inode_iter, inode_u, inum,
BTREE_ITER_INTENT);
BTREE_ITER_intent);
if (ret)
goto err;
@ -322,7 +322,7 @@ int bch2_unlink_trans(struct btree_trans *trans,
ret = bch2_hash_delete_at(trans, bch2_dirent_hash_desc,
&dir_hash, &dirent_iter,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
BTREE_UPDATE_internal_snapshot_node) ?:
bch2_inode_write(trans, &dir_iter, dir_u) ?:
bch2_inode_write(trans, &inode_iter, inode_u);
err:
@ -363,7 +363,7 @@ static int subvol_update_parent(struct btree_trans *trans, u32 subvol, u32 new_p
struct bkey_i_subvolume *s =
bch2_bkey_get_mut_typed(trans, &iter,
BTREE_ID_subvolumes, POS(0, subvol),
BTREE_ITER_CACHED, subvolume);
BTREE_ITER_cached, subvolume);
int ret = PTR_ERR_OR_ZERO(s);
if (ret)
return ret;
@ -394,7 +394,7 @@ int bch2_rename_trans(struct btree_trans *trans,
int ret;
ret = bch2_inode_peek(trans, &src_dir_iter, src_dir_u, src_dir,
BTREE_ITER_INTENT);
BTREE_ITER_intent);
if (ret)
goto err;
@ -403,7 +403,7 @@ int bch2_rename_trans(struct btree_trans *trans,
if (dst_dir.inum != src_dir.inum ||
dst_dir.subvol != src_dir.subvol) {
ret = bch2_inode_peek(trans, &dst_dir_iter, dst_dir_u, dst_dir,
BTREE_ITER_INTENT);
BTREE_ITER_intent);
if (ret)
goto err;
@ -423,13 +423,13 @@ int bch2_rename_trans(struct btree_trans *trans,
goto err;
ret = bch2_inode_peek(trans, &src_inode_iter, src_inode_u, src_inum,
BTREE_ITER_INTENT);
BTREE_ITER_intent);
if (ret)
goto err;
if (dst_inum.inum) {
ret = bch2_inode_peek(trans, &dst_inode_iter, dst_inode_u, dst_inum,
BTREE_ITER_INTENT);
BTREE_ITER_intent);
if (ret)
goto err;
}

14
fs/bcachefs/fs-io-buffered.c
View File

@ -30,15 +30,8 @@ static void bch2_readpages_end_io(struct bio *bio)
{
struct folio_iter fi;
bio_for_each_folio_all(fi, bio) {
if (!bio->bi_status) {
folio_mark_uptodate(fi.folio);
} else {
folio_clear_uptodate(fi.folio);
folio_set_error(fi.folio);
}
folio_unlock(fi.folio);
}
bio_for_each_folio_all(fi, bio)
folio_end_read(fi.folio, bio->bi_status == BLK_STS_OK);
bio_put(bio);
}
@ -176,7 +169,7 @@ retry:
bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
SPOS(inum.inum, rbio->bio.bi_iter.bi_sector, snapshot),
BTREE_ITER_SLOTS);
BTREE_ITER_slots);
while (1) {
struct bkey_s_c k;
unsigned bytes, sectors, offset_into_extent;
@ -408,7 +401,6 @@ static void bch2_writepage_io_done(struct bch_write_op *op)
bio_for_each_folio_all(fi, bio) {
struct bch_folio *s;
folio_set_error(fi.folio);
mapping_set_error(fi.folio->mapping, -EIO);
s = __bch2_folio(fi.folio);

2
fs/bcachefs/fs-io-direct.c
View File

@ -254,7 +254,7 @@ retry:
for_each_btree_key_norestart(trans, iter, BTREE_ID_extents,
SPOS(inum.inum, offset, snapshot),
BTREE_ITER_SLOTS, k, err) {
BTREE_ITER_slots, k, err) {
if (bkey_ge(bkey_start_pos(k.k), POS(inum.inum, end)))
break;

2
fs/bcachefs/fs-io-pagecache.c
View File

@ -214,7 +214,7 @@ retry:
for_each_btree_key_norestart(trans, iter, BTREE_ID_extents,
SPOS(inum.inum, offset, snapshot),
BTREE_ITER_SLOTS, k, ret) {
BTREE_ITER_slots, k, ret) {
unsigned nr_ptrs = bch2_bkey_nr_ptrs_fully_allocated(k);
unsigned state = bkey_to_sector_state(k);

9
fs/bcachefs/fs-io.c
View File

@ -202,7 +202,10 @@ int bch2_fsync(struct file *file, loff_t start, loff_t end, int datasync)
goto out;
ret = bch2_flush_inode(c, inode);
out:
return bch2_err_class(ret);
ret = bch2_err_class(ret);
if (ret == -EROFS)
ret = -EIO;
return ret;
}
/* truncate: */
@ -594,7 +597,7 @@ static int __bchfs_fallocate(struct bch_inode_info *inode, int mode,
bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
POS(inode->v.i_ino, start_sector),
BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
BTREE_ITER_slots|BTREE_ITER_intent);
while (!ret && bkey_lt(iter.pos, end_pos)) {
s64 i_sectors_delta = 0;
@ -1009,7 +1012,7 @@ retry:
for_each_btree_key_norestart(trans, iter, BTREE_ID_extents,
SPOS(inode->v.i_ino, offset >> 9, snapshot),
BTREE_ITER_SLOTS, k, ret) {
BTREE_ITER_slots, k, ret) {
if (k.k->p.inode != inode->v.i_ino) {
next_hole = bch2_seek_pagecache_hole(&inode->v,
offset, MAX_LFS_FILESIZE, 0, false);

2
fs/bcachefs/fs-ioctl.c
View File

@ -548,7 +548,7 @@ long bch2_compat_fs_ioctl(struct file *file, unsigned cmd, unsigned long arg)
{
/* These are just misnamed, they actually get/put from/to user an int */
switch (cmd) {
case FS_IOC_GETFLAGS:
case FS_IOC32_GETFLAGS:
cmd = FS_IOC_GETFLAGS;
break;
case FS_IOC32_SETFLAGS:

109
fs/bcachefs/fs.c
View File

@ -90,7 +90,7 @@ retry:
bch2_trans_begin(trans);
ret = bch2_inode_peek(trans, &iter, &inode_u, inode_inum(inode),
BTREE_ITER_INTENT) ?:
BTREE_ITER_intent) ?:
(set ? set(trans, inode, &inode_u, p) : 0) ?:
bch2_inode_write(trans, &iter, &inode_u) ?:
bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
@ -213,19 +213,43 @@ static struct bch_inode_info *bch2_inode_insert(struct bch_fs *c, struct bch_ino
_ret; \
})
static struct inode *bch2_alloc_inode(struct super_block *sb)
{
BUG();
}
static struct bch_inode_info *__bch2_new_inode(struct bch_fs *c)
{
struct bch_inode_info *inode = kmem_cache_alloc(bch2_inode_cache, GFP_NOFS);
if (!inode)
return NULL;
inode_init_once(&inode->v);
mutex_init(&inode->ei_update_lock);
two_state_lock_init(&inode->ei_pagecache_lock);
INIT_LIST_HEAD(&inode->ei_vfs_inode_list);
mutex_init(&inode->ei_quota_lock);
inode->v.i_state = 0;
if (unlikely(inode_init_always(c->vfs_sb, &inode->v))) {
kmem_cache_free(bch2_inode_cache, inode);
return NULL;
}
return inode;
}
/*
* Allocate a new inode, dropping/retaking btree locks if necessary:
*/
static struct bch_inode_info *bch2_new_inode(struct btree_trans *trans)
{
struct bch_fs *c = trans->c;
struct bch_inode_info *inode =
memalloc_flags_do(PF_MEMALLOC_NORECLAIM|PF_MEMALLOC_NOWARN,
to_bch_ei(new_inode(c->vfs_sb)));
__bch2_new_inode(trans->c));
if (unlikely(!inode)) {
int ret = drop_locks_do(trans, (inode = to_bch_ei(new_inode(c->vfs_sb))) ? 0 : -ENOMEM);
int ret = drop_locks_do(trans, (inode = __bch2_new_inode(trans->c)) ? 0 : -ENOMEM);
if (ret && inode) {
__destroy_inode(&inode->v);
kmem_cache_free(bch2_inode_cache, inode);
@ -290,7 +314,7 @@ __bch2_create(struct mnt_idmap *idmap,
if (ret)
return ERR_PTR(ret);
#endif
inode = to_bch_ei(new_inode(c->vfs_sb));
inode = __bch2_new_inode(c);
if (unlikely(!inode)) {
inode = ERR_PTR(-ENOMEM);
goto err;
@ -323,7 +347,7 @@ retry:
inum.inum = inode_u.bi_inum;
ret = bch2_subvolume_get(trans, inum.subvol, true,
BTREE_ITER_WITH_UPDATES, &subvol) ?:
BTREE_ITER_with_updates, &subvol) ?:
bch2_trans_commit(trans, NULL, &journal_seq, 0);
if (unlikely(ret)) {
bch2_quota_acct(c, bch_qid(&inode_u), Q_INO, -1,
@ -376,17 +400,14 @@ static struct bch_inode_info *bch2_lookup_trans(struct btree_trans *trans,
struct bch_fs *c = trans->c;
struct btree_iter dirent_iter = {};
subvol_inum inum = {};
struct printbuf buf = PRINTBUF;
int ret = bch2_hash_lookup(trans, &dirent_iter, bch2_dirent_hash_desc,
dir_hash_info, dir, name, 0);
struct bkey_s_c k = bch2_hash_lookup(trans, &dirent_iter, bch2_dirent_hash_desc,
dir_hash_info, dir, name, 0);
int ret = bkey_err(k);
if (ret)
return ERR_PTR(ret);
struct bkey_s_c k = bch2_btree_iter_peek_slot(&dirent_iter);
ret = bkey_err(k);
if (ret)
goto err;
ret = bch2_dirent_read_target(trans, dir, bkey_s_c_to_dirent(k), &inum);
if (ret > 0)
ret = -ENOENT;
@ -406,20 +427,31 @@ static struct bch_inode_info *bch2_lookup_trans(struct btree_trans *trans,
ret = bch2_subvolume_get(trans, inum.subvol, true, 0, &subvol) ?:
bch2_inode_find_by_inum_nowarn_trans(trans, inum, &inode_u) ?:
PTR_ERR_OR_ZERO(inode = bch2_new_inode(trans));
if (bch2_err_matches(ret, ENOENT)) {
struct printbuf buf = PRINTBUF;
bch2_bkey_val_to_text(&buf, c, k);
bch_err(c, "%s points to missing inode", buf.buf);
printbuf_exit(&buf);
}
bch2_fs_inconsistent_on(bch2_err_matches(ret, ENOENT),
c, "dirent to missing inode:\n %s",
(bch2_bkey_val_to_text(&buf, c, k), buf.buf));
if (ret)
goto err;
/* regular files may have hardlinks: */
if (bch2_fs_inconsistent_on(bch2_inode_should_have_bp(&inode_u) &&
!bkey_eq(k.k->p, POS(inode_u.bi_dir, inode_u.bi_dir_offset)),
c,
"dirent points to inode that does not point back:\n %s",
(bch2_bkey_val_to_text(&buf, c, k),
prt_printf(&buf, "\n "),
bch2_inode_unpacked_to_text(&buf, &inode_u),
buf.buf))) {
ret = -ENOENT;
goto err;
}
bch2_vfs_inode_init(trans, inum, inode, &inode_u, &subvol);
inode = bch2_inode_insert(c, inode);
out:
bch2_trans_iter_exit(trans, &dirent_iter);
printbuf_exit(&buf);
return inode;
err:
inode = ERR_PTR(ret);
@ -787,7 +819,7 @@ retry:
acl = NULL;
ret = bch2_inode_peek(trans, &inode_iter, &inode_u, inode_inum(inode),
BTREE_ITER_INTENT);
BTREE_ITER_intent);
if (ret)
goto btree_err;
@ -1043,6 +1075,10 @@ retry:
bch2_btree_iter_set_pos(&iter,
POS(iter.pos.inode, iter.pos.offset + sectors));
ret = bch2_trans_relock(trans);
if (ret)
break;
}
start = iter.pos.offset;
bch2_trans_iter_exit(trans, &iter);
@ -1490,34 +1526,9 @@ static void bch2_vfs_inode_init(struct btree_trans *trans, subvol_inum inum,
mapping_set_large_folios(inode->v.i_mapping);
}
static struct inode *bch2_alloc_inode(struct super_block *sb)
static void bch2_free_inode(struct inode *vinode)
{
struct bch_inode_info *inode;
inode = kmem_cache_alloc(bch2_inode_cache, GFP_NOFS);
if (!inode)
return NULL;
inode_init_once(&inode->v);
mutex_init(&inode->ei_update_lock);
two_state_lock_init(&inode->ei_pagecache_lock);
INIT_LIST_HEAD(&inode->ei_vfs_inode_list);
mutex_init(&inode->ei_quota_lock);
return &inode->v;
}
static void bch2_i_callback(struct rcu_head *head)
{
struct inode *vinode = container_of(head, struct inode, i_rcu);
struct bch_inode_info *inode = to_bch_ei(vinode);
kmem_cache_free(bch2_inode_cache, inode);
}
static void bch2_destroy_inode(struct inode *vinode)
{
call_rcu(&vinode->i_rcu, bch2_i_callback);
kmem_cache_free(bch2_inode_cache, to_bch_ei(vinode));
}
static int inode_update_times_fn(struct btree_trans *trans,
@ -1825,7 +1836,7 @@ static int bch2_unfreeze(struct super_block *sb)
static const struct super_operations bch_super_operations = {
.alloc_inode = bch2_alloc_inode,
.destroy_inode = bch2_destroy_inode,
.free_inode = bch2_free_inode,
.write_inode = bch2_vfs_write_inode,
.evict_inode = bch2_evict_inode,
.sync_fs = bch2_sync_fs,

212
fs/bcachefs/fsck.c
View File

@ -79,7 +79,7 @@ static int lookup_first_inode(struct btree_trans *trans, u64 inode_nr,
bch2_trans_iter_init(trans, &iter, BTREE_ID_inodes,
POS(0, inode_nr),
BTREE_ITER_ALL_SNAPSHOTS);
BTREE_ITER_all_snapshots);
k = bch2_btree_iter_peek(&iter);
ret = bkey_err(k);
if (ret)
@ -127,13 +127,13 @@ static int lookup_dirent_in_snapshot(struct btree_trans *trans,
u64 *target, unsigned *type, u32 snapshot)
{
struct btree_iter iter;
struct bkey_s_c_dirent d;
int ret = bch2_hash_lookup_in_snapshot(trans, &iter, bch2_dirent_hash_desc,
&hash_info, dir, name, 0, snapshot);
struct bkey_s_c k = bch2_hash_lookup_in_snapshot(trans, &iter, bch2_dirent_hash_desc,
&hash_info, dir, name, 0, snapshot);
int ret = bkey_err(k);
if (ret)
return ret;
d = bkey_s_c_to_dirent(bch2_btree_iter_peek_slot(&iter));
struct bkey_s_c_dirent d = bkey_s_c_to_dirent(bch2_btree_iter_peek_slot(&iter));
*target = le64_to_cpu(d.v->d_inum);
*type = d.v->d_type;
bch2_trans_iter_exit(trans, &iter);
@ -154,12 +154,12 @@ static int __remove_dirent(struct btree_trans *trans, struct bpos pos)
dir_hash_info = bch2_hash_info_init(c, &dir_inode);
bch2_trans_iter_init(trans, &iter, BTREE_ID_dirents, pos, BTREE_ITER_INTENT);
bch2_trans_iter_init(trans, &iter, BTREE_ID_dirents, pos, BTREE_ITER_intent);
ret = bch2_btree_iter_traverse(&iter) ?:
bch2_hash_delete_at(trans, bch2_dirent_hash_desc,
&dir_hash_info, &iter,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
bch2_trans_iter_exit(trans, &iter);
err:
bch_err_fn(c, ret);
@ -274,9 +274,9 @@ create_lostfound:
&lostfound_str,
lostfound->bi_inum,
&lostfound->bi_dir_offset,
BCH_HASH_SET_MUST_CREATE) ?:
STR_HASH_must_create) ?:
bch2_inode_write_flags(trans, &lostfound_iter, lostfound,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
err:
bch_err_msg(c, ret, "creating lost+found");
bch2_trans_iter_exit(trans, &lostfound_iter);
@ -333,7 +333,7 @@ static int reattach_inode(struct btree_trans *trans,
&name,
inode->bi_subvol ?: inode->bi_inum,
&dir_offset,
BCH_HASH_SET_MUST_CREATE);
STR_HASH_must_create);
if (ret)
return ret;
@ -486,14 +486,9 @@ static int reconstruct_reg_inode(struct btree_trans *trans, u32 snapshot, u64 in
return reconstruct_inode(trans, snapshot, inum, k.k->p.offset << 9, S_IFREG);
}
struct snapshots_seen_entry {
u32 id;
u32 equiv;
};
struct snapshots_seen {
struct bpos pos;
DARRAY(struct snapshots_seen_entry) ids;
snapshot_id_list ids;
};
static inline void snapshots_seen_exit(struct snapshots_seen *s)
@ -508,20 +503,15 @@ static inline void snapshots_seen_init(struct snapshots_seen *s)
static int snapshots_seen_add_inorder(struct bch_fs *c, struct snapshots_seen *s, u32 id)
{
struct snapshots_seen_entry *i, n = {
.id = id,
.equiv = bch2_snapshot_equiv(c, id),
};
int ret = 0;
u32 *i;
__darray_for_each(s->ids, i) {
if (i->id == id)
if (*i == id)
return 0;
if (i->id > id)
if (*i > id)
break;
}
ret = darray_insert_item(&s->ids, i - s->ids.data, n);
int ret = darray_insert_item(&s->ids, i - s->ids.data, id);
if (ret)
bch_err(c, "error reallocating snapshots_seen table (size %zu)",
s->ids.size);
@ -531,42 +521,11 @@ static int snapshots_seen_add_inorder(struct bch_fs *c, struct snapshots_seen *s
static int snapshots_seen_update(struct bch_fs *c, struct snapshots_seen *s,
enum btree_id btree_id, struct bpos pos)
{
struct snapshots_seen_entry n = {
.id = pos.snapshot,
.equiv = bch2_snapshot_equiv(c, pos.snapshot),
};
int ret = 0;
if (!bkey_eq(s->pos, pos))
s->ids.nr = 0;
s->pos = pos;
s->pos.snapshot = n.equiv;
darray_for_each(s->ids, i) {
if (i->id == n.id)
return 0;
/*
* We currently don't rigorously track for snapshot cleanup
* needing to be run, so it shouldn't be a fsck error yet:
*/
if (i->equiv == n.equiv) {
bch_err(c, "snapshot deletion did not finish:\n"
" duplicate keys in btree %s at %llu:%llu snapshots %u, %u (equiv %u)\n",
bch2_btree_id_str(btree_id),
pos.inode, pos.offset,
i->id, n.id, n.equiv);
set_bit(BCH_FS_need_delete_dead_snapshots, &c->flags);
return bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_delete_dead_snapshots);
}
}
ret = darray_push(&s->ids, n);
if (ret)
bch_err(c, "error reallocating snapshots_seen table (size %zu)",
s->ids.size);
return ret;
return snapshot_list_add_nodup(c, &s->ids, pos.snapshot);
}
/**
@ -586,12 +545,10 @@ static bool key_visible_in_snapshot(struct bch_fs *c, struct snapshots_seen *see
ssize_t i;
EBUG_ON(id > ancestor);
EBUG_ON(!bch2_snapshot_is_equiv(c, id));
EBUG_ON(!bch2_snapshot_is_equiv(c, ancestor));
/* @ancestor should be the snapshot most recently added to @seen */
EBUG_ON(ancestor != seen->pos.snapshot);
EBUG_ON(ancestor != seen->ids.data[seen->ids.nr - 1].equiv);
EBUG_ON(ancestor != darray_last(seen->ids));
if (id == ancestor)
return true;
@ -610,9 +567,9 @@ static bool key_visible_in_snapshot(struct bch_fs *c, struct snapshots_seen *see
*/
for (i = seen->ids.nr - 2;
i >= 0 && seen->ids.data[i].equiv >= id;
i >= 0 && seen->ids.data[i] >= id;
--i)
if (bch2_snapshot_is_ancestor(c, id, seen->ids.data[i].equiv))
if (bch2_snapshot_is_ancestor(c, id, seen->ids.data[i]))
return false;
return true;
@ -643,9 +600,6 @@ static int ref_visible2(struct bch_fs *c,
u32 src, struct snapshots_seen *src_seen,
u32 dst, struct snapshots_seen *dst_seen)
{
src = bch2_snapshot_equiv(c, src);
dst = bch2_snapshot_equiv(c, dst);
if (dst > src) {
swap(dst, src);
swap(dst_seen, src_seen);
@ -692,7 +646,7 @@ static int add_inode(struct bch_fs *c, struct inode_walker *w,
return darray_push(&w->inodes, ((struct inode_walker_entry) {
.inode = u,
.snapshot = bch2_snapshot_equiv(c, inode.k->p.snapshot),
.snapshot = inode.k->p.snapshot,
}));
}
@ -708,7 +662,7 @@ static int get_inodes_all_snapshots(struct btree_trans *trans,
w->inodes.nr = 0;
for_each_btree_key_norestart(trans, iter, BTREE_ID_inodes, POS(0, inum),
BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
BTREE_ITER_all_snapshots, k, ret) {
if (k.k->p.offset != inum)
break;
@ -728,21 +682,20 @@ static struct inode_walker_entry *
lookup_inode_for_snapshot(struct bch_fs *c, struct inode_walker *w, struct bkey_s_c k)
{
bool is_whiteout = k.k->type == KEY_TYPE_whiteout;
u32 snapshot = bch2_snapshot_equiv(c, k.k->p.snapshot);
struct inode_walker_entry *i;
__darray_for_each(w->inodes, i)
if (bch2_snapshot_is_ancestor(c, snapshot, i->snapshot))
if (bch2_snapshot_is_ancestor(c, k.k->p.snapshot, i->snapshot))
goto found;
return NULL;
found:
BUG_ON(snapshot > i->snapshot);
BUG_ON(k.k->p.snapshot > i->snapshot);
if (snapshot != i->snapshot && !is_whiteout) {
if (k.k->p.snapshot != i->snapshot && !is_whiteout) {
struct inode_walker_entry new = *i;
new.snapshot = snapshot;
new.snapshot = k.k->p.snapshot;
new.count = 0;
struct printbuf buf = PRINTBUF;
@ -751,10 +704,10 @@ found:
bch_info(c, "have key for inode %llu:%u but have inode in ancestor snapshot %u\n"
"unexpected because we should always update the inode when we update a key in that inode\n"
"%s",
w->last_pos.inode, snapshot, i->snapshot, buf.buf);
w->last_pos.inode, k.k->p.snapshot, i->snapshot, buf.buf);
printbuf_exit(&buf);
while (i > w->inodes.data && i[-1].snapshot > snapshot)
while (i > w->inodes.data && i[-1].snapshot > k.k->p.snapshot)
--i;
size_t pos = i - w->inodes.data;
@ -786,10 +739,10 @@ static struct inode_walker_entry *walk_inode(struct btree_trans *trans,
return lookup_inode_for_snapshot(trans->c, w, k);
}
static int __get_visible_inodes(struct btree_trans *trans,
struct inode_walker *w,
struct snapshots_seen *s,
u64 inum)
static int get_visible_inodes(struct btree_trans *trans,
struct inode_walker *w,
struct snapshots_seen *s,
u64 inum)
{
struct bch_fs *c = trans->c;
struct btree_iter iter;
@ -799,19 +752,17 @@ static int __get_visible_inodes(struct btree_trans *trans,
w->inodes.nr = 0;
for_each_btree_key_norestart(trans, iter, BTREE_ID_inodes, POS(0, inum),
BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
u32 equiv = bch2_snapshot_equiv(c, k.k->p.snapshot);
BTREE_ITER_all_snapshots, k, ret) {
if (k.k->p.offset != inum)
break;
if (!ref_visible(c, s, s->pos.snapshot, equiv))
if (!ref_visible(c, s, s->pos.snapshot, k.k->p.snapshot))
continue;
if (bkey_is_inode(k.k))
add_inode(c, w, k);
if (equiv >= s->pos.snapshot)
if (k.k->p.snapshot >= s->pos.snapshot)
break;
}
bch2_trans_iter_exit(trans, &iter);
@ -832,7 +783,7 @@ static int check_key_has_snapshot(struct btree_trans *trans,
"key in missing snapshot: %s",
(bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
ret = bch2_btree_delete_at(trans, iter,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?: 1;
BTREE_UPDATE_internal_snapshot_node) ?: 1;
fsck_err:
printbuf_exit(&buf);
return ret;
@ -861,8 +812,8 @@ static int hash_redo_key(struct btree_trans *trans,
bch2_hash_set_in_snapshot(trans, desc, hash_info,
(subvol_inum) { 0, k.k->p.inode },
k.k->p.snapshot, tmp,
BCH_HASH_SET_MUST_CREATE,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
STR_HASH_must_create|
BTREE_UPDATE_internal_snapshot_node) ?:
bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
}
@ -891,7 +842,7 @@ static int hash_check_key(struct btree_trans *trans,
for_each_btree_key_norestart(trans, iter, desc.btree_id,
SPOS(hash_k.k->p.inode, hash, hash_k.k->p.snapshot),
BTREE_ITER_SLOTS, k, ret) {
BTREE_ITER_slots, k, ret) {
if (bkey_eq(k.k->p, hash_k.k->p))
break;
@ -1233,7 +1184,7 @@ int bch2_check_inodes(struct bch_fs *c)
int ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter, BTREE_ID_inodes,
POS_MIN,
BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
check_inode(trans, &iter, k, &prev, &s, full)));
@ -1362,8 +1313,8 @@ static int overlapping_extents_found(struct btree_trans *trans,
BUG_ON(bkey_le(pos1, bkey_start_pos(&pos2)));
bch2_trans_iter_init(trans, &iter1, btree, pos1,
BTREE_ITER_ALL_SNAPSHOTS|
BTREE_ITER_NOT_EXTENTS);
BTREE_ITER_all_snapshots|
BTREE_ITER_not_extents);
k1 = bch2_btree_iter_peek_upto(&iter1, POS(pos1.inode, U64_MAX));
ret = bkey_err(k1);
if (ret)
@ -1425,7 +1376,7 @@ static int overlapping_extents_found(struct btree_trans *trans,
trans->extra_disk_res += bch2_bkey_sectors_compressed(k2);
ret = bch2_trans_update_extent_overwrite(trans, old_iter,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE,
BTREE_UPDATE_internal_snapshot_node,
k1, k2) ?:
bch2_trans_commit(trans, &res, NULL, BCH_TRANS_COMMIT_no_enospc);
bch2_disk_reservation_put(c, &res);
@ -1466,7 +1417,6 @@ static int check_overlapping_extents(struct btree_trans *trans,
struct snapshots_seen *seen,
struct extent_ends *extent_ends,
struct bkey_s_c k,
u32 equiv,
struct btree_iter *iter,
bool *fixed)
{
@ -1535,11 +1485,8 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
struct bch_fs *c = trans->c;
struct inode_walker_entry *i;
struct printbuf buf = PRINTBUF;
struct bpos equiv = k.k->p;
int ret = 0;
equiv.snapshot = bch2_snapshot_equiv(c, k.k->p.snapshot);
ret = check_key_has_snapshot(trans, iter, k);
if (ret) {
ret = ret < 0 ? ret : 0;
@ -1589,8 +1536,7 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
goto delete;
ret = check_overlapping_extents(trans, s, extent_ends, k,
equiv.snapshot, iter,
ret = check_overlapping_extents(trans, s, extent_ends, k, iter,
&inode->recalculate_sums);
if (ret)
goto err;
@ -1607,8 +1553,8 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
for (;
inode->inodes.data && i >= inode->inodes.data;
--i) {
if (i->snapshot > equiv.snapshot ||
!key_visible_in_snapshot(c, s, i->snapshot, equiv.snapshot))
if (i->snapshot > k.k->p.snapshot ||
!key_visible_in_snapshot(c, s, i->snapshot, k.k->p.snapshot))
continue;
if (k.k->type != KEY_TYPE_whiteout) {
@ -1625,7 +1571,7 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
bch2_btree_iter_set_snapshot(&iter2, i->snapshot);
ret = bch2_btree_iter_traverse(&iter2) ?:
bch2_btree_delete_at(trans, &iter2,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
bch2_trans_iter_exit(trans, &iter2);
if (ret)
goto err;
@ -1652,7 +1598,7 @@ fsck_err:
bch_err_fn(c, ret);
return ret;
delete:
ret = bch2_btree_delete_at(trans, iter, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
ret = bch2_btree_delete_at(trans, iter, BTREE_UPDATE_internal_snapshot_node);
goto out;
}
@ -1673,7 +1619,7 @@ int bch2_check_extents(struct bch_fs *c)
int ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter, BTREE_ID_extents,
POS(BCACHEFS_ROOT_INO, 0),
BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
&res, NULL,
BCH_TRANS_COMMIT_no_enospc, ({
bch2_disk_reservation_put(c, &res);
@ -1698,7 +1644,7 @@ int bch2_check_indirect_extents(struct bch_fs *c)
int ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter, BTREE_ID_reflink,
POS_MIN,
BTREE_ITER_PREFETCH, k,
BTREE_ITER_prefetch, k,
&res, NULL,
BCH_TRANS_COMMIT_no_enospc, ({
bch2_disk_reservation_put(c, &res);
@ -1767,6 +1713,15 @@ static int check_dirent_inode_dirent(struct btree_trans *trans,
if (inode_points_to_dirent(target, d))
return 0;
if (bch2_inode_should_have_bp(target) &&
!fsck_err(c, inode_wrong_backpointer,
"dirent points to inode that does not point back:\n %s",
(bch2_bkey_val_to_text(&buf, c, d.s_c),
prt_printf(&buf, "\n "),
bch2_inode_unpacked_to_text(&buf, target),
buf.buf)))
goto out_noiter;
if (!target->bi_dir &&
!target->bi_dir_offset) {
target->bi_dir = d.k->p.inode;
@ -1835,6 +1790,7 @@ out:
err:
fsck_err:
bch2_trans_iter_exit(trans, &bp_iter);
out_noiter:
printbuf_exit(&buf);
bch_err_fn(c, ret);
return ret;
@ -2052,7 +2008,6 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
struct bch_fs *c = trans->c;
struct inode_walker_entry *i;
struct printbuf buf = PRINTBUF;
struct bpos equiv;
int ret = 0;
ret = check_key_has_snapshot(trans, iter, k);
@ -2061,9 +2016,6 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
goto out;
}
equiv = k.k->p;
equiv.snapshot = bch2_snapshot_equiv(c, k.k->p.snapshot);
ret = snapshots_seen_update(c, s, iter->btree_id, k.k->p);
if (ret)
goto err;
@ -2104,7 +2056,7 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
(printbuf_reset(&buf),
bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
ret = bch2_btree_delete_at(trans, iter,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
goto out;
}
@ -2140,14 +2092,13 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
if (ret)
goto err;
} else {
ret = __get_visible_inodes(trans, target, s, le64_to_cpu(d.v->d_inum));
ret = get_visible_inodes(trans, target, s, le64_to_cpu(d.v->d_inum));
if (ret)
goto err;
if (fsck_err_on(!target->inodes.nr,
c, dirent_to_missing_inode,
"dirent points to missing inode: (equiv %u)\n%s",
equiv.snapshot,
"dirent points to missing inode:\n%s",
(printbuf_reset(&buf),
bch2_bkey_val_to_text(&buf, c, k),
buf.buf))) {
@ -2164,7 +2115,7 @@ static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
}
if (d.v->d_type == DT_DIR)
for_each_visible_inode(c, s, dir, equiv.snapshot, i)
for_each_visible_inode(c, s, dir, d.k->p.snapshot, i)
i->count++;
}
out:
@ -2191,7 +2142,7 @@ int bch2_check_dirents(struct bch_fs *c)
int ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter, BTREE_ID_dirents,
POS(BCACHEFS_ROOT_INO, 0),
BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS,
BTREE_ITER_prefetch|BTREE_ITER_all_snapshots,
k,
NULL, NULL,
BCH_TRANS_COMMIT_no_enospc,
@ -2255,7 +2206,7 @@ int bch2_check_xattrs(struct bch_fs *c)
ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter, BTREE_ID_xattrs,
POS(BCACHEFS_ROOT_INO, 0),
BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS,
BTREE_ITER_prefetch|BTREE_ITER_all_snapshots,
k,
NULL, NULL,
BCH_TRANS_COMMIT_no_enospc,
@ -2422,7 +2373,7 @@ int bch2_check_subvolume_structure(struct bch_fs *c)
{
int ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter,
BTREE_ID_subvolumes, POS_MIN, BTREE_ITER_PREFETCH, k,
BTREE_ID_subvolumes, POS_MIN, BTREE_ITER_prefetch, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
check_subvol_path(trans, &iter, k)));
bch_err_fn(c, ret);
@ -2457,7 +2408,7 @@ static int check_path(struct btree_trans *trans, pathbuf *p, struct bkey_s_c ino
struct btree_iter inode_iter = {};
struct bch_inode_unpacked inode;
struct printbuf buf = PRINTBUF;
u32 snapshot = bch2_snapshot_equiv(c, inode_k.k->p.snapshot);
u32 snapshot = inode_k.k->p.snapshot;
int ret = 0;
p->nr = 0;
@ -2559,9 +2510,9 @@ int bch2_check_directory_structure(struct bch_fs *c)
ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter, BTREE_ID_inodes, POS_MIN,
BTREE_ITER_INTENT|
BTREE_ITER_PREFETCH|
BTREE_ITER_ALL_SNAPSHOTS, k,
BTREE_ITER_intent|
BTREE_ITER_prefetch|
BTREE_ITER_all_snapshots, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc, ({
if (!bkey_is_inode(k.k))
continue;
@ -2661,9 +2612,9 @@ static int check_nlinks_find_hardlinks(struct bch_fs *c,
int ret = bch2_trans_run(c,
for_each_btree_key(trans, iter, BTREE_ID_inodes,
POS(0, start),
BTREE_ITER_INTENT|
BTREE_ITER_PREFETCH|
BTREE_ITER_ALL_SNAPSHOTS, k, ({
BTREE_ITER_intent|
BTREE_ITER_prefetch|
BTREE_ITER_all_snapshots, k, ({
if (!bkey_is_inode(k.k))
continue;
@ -2704,9 +2655,9 @@ static int check_nlinks_walk_dirents(struct bch_fs *c, struct nlink_table *links
int ret = bch2_trans_run(c,
for_each_btree_key(trans, iter, BTREE_ID_dirents, POS_MIN,
BTREE_ITER_INTENT|
BTREE_ITER_PREFETCH|
BTREE_ITER_ALL_SNAPSHOTS, k, ({
BTREE_ITER_intent|
BTREE_ITER_prefetch|
BTREE_ITER_all_snapshots, k, ({
ret = snapshots_seen_update(c, &s, iter.btree_id, k.k->p);
if (ret)
break;
@ -2717,8 +2668,7 @@ static int check_nlinks_walk_dirents(struct bch_fs *c, struct nlink_table *links
if (d.v->d_type != DT_DIR &&
d.v->d_type != DT_SUBVOL)
inc_link(c, &s, links, range_start, range_end,
le64_to_cpu(d.v->d_inum),
bch2_snapshot_equiv(c, d.k->p.snapshot));
le64_to_cpu(d.v->d_inum), d.k->p.snapshot);
}
0;
})));
@ -2781,7 +2731,7 @@ static int check_nlinks_update_hardlinks(struct bch_fs *c,
int ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter, BTREE_ID_inodes,
POS(0, range_start),
BTREE_ITER_INTENT|BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
BTREE_ITER_intent|BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
check_nlinks_update_inode(trans, &iter, k, links, &idx, range_end)));
if (ret < 0) {
@ -2849,7 +2799,7 @@ static int fix_reflink_p_key(struct btree_trans *trans, struct btree_iter *iter,
u->v.front_pad = 0;
u->v.back_pad = 0;
return bch2_trans_update(trans, iter, &u->k_i, BTREE_TRIGGER_NORUN);
return bch2_trans_update(trans, iter, &u->k_i, BTREE_TRIGGER_norun);
}
int bch2_fix_reflink_p(struct bch_fs *c)
@ -2860,8 +2810,8 @@ int bch2_fix_reflink_p(struct bch_fs *c)
int ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter,
BTREE_ID_extents, POS_MIN,
BTREE_ITER_INTENT|BTREE_ITER_PREFETCH|
BTREE_ITER_ALL_SNAPSHOTS, k,
BTREE_ITER_intent|BTREE_ITER_prefetch|
BTREE_ITER_all_snapshots, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
fix_reflink_p_key(trans, &iter, k)));
bch_err_fn(c, ret);

64
fs/bcachefs/inode.c
View File

@ -339,7 +339,7 @@ int bch2_inode_peek_nowarn(struct btree_trans *trans,
k = bch2_bkey_get_iter(trans, iter, BTREE_ID_inodes,
SPOS(0, inum.inum, snapshot),
flags|BTREE_ITER_CACHED);
flags|BTREE_ITER_cached);
ret = bkey_err(k);
if (ret)
return ret;
@ -371,7 +371,7 @@ int bch2_inode_peek(struct btree_trans *trans,
int bch2_inode_write_flags(struct btree_trans *trans,
struct btree_iter *iter,
struct bch_inode_unpacked *inode,
enum btree_update_flags flags)
enum btree_iter_update_trigger_flags flags)
{
struct bkey_inode_buf *inode_p;
@ -399,7 +399,7 @@ int __bch2_fsck_write_inode(struct btree_trans *trans,
return bch2_btree_insert_nonextent(trans, BTREE_ID_inodes,
&inode_p->inode.k_i,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
}
int bch2_fsck_write_inode(struct btree_trans *trans,
@ -473,7 +473,7 @@ fsck_err:
}
int bch2_inode_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_s_c_inode inode = bkey_s_c_to_inode(k);
@ -490,7 +490,7 @@ fsck_err:
}
int bch2_inode_v2_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_s_c_inode_v2 inode = bkey_s_c_to_inode_v2(k);
@ -507,7 +507,7 @@ fsck_err:
}
int bch2_inode_v3_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_s_c_inode_v3 inode = bkey_s_c_to_inode_v3(k);
@ -535,29 +535,19 @@ static void __bch2_inode_unpacked_to_text(struct printbuf *out,
struct bch_inode_unpacked *inode)
{
printbuf_indent_add(out, 2);
prt_printf(out, "mode=%o", inode->bi_mode);
prt_newline(out);
prt_printf(out, "mode=%o\n", inode->bi_mode);
prt_str(out, "flags=");
prt_bitflags(out, bch2_inode_flag_strs, inode->bi_flags & ((1U << 20) - 1));
prt_printf(out, " (%x)", inode->bi_flags);
prt_newline(out);
prt_printf(out, " (%x)\n", inode->bi_flags);
prt_printf(out, "journal_seq=%llu", inode->bi_journal_seq);
prt_newline(out);
prt_printf(out, "bi_size=%llu", inode->bi_size);
prt_newline(out);
prt_printf(out, "bi_sectors=%llu", inode->bi_sectors);
prt_newline(out);
prt_printf(out, "bi_version=%llu", inode->bi_version);
prt_newline(out);
prt_printf(out, "journal_seq=%llu\n", inode->bi_journal_seq);
prt_printf(out, "bi_size=%llu\n", inode->bi_size);
prt_printf(out, "bi_sectors=%llu\n", inode->bi_sectors);
prt_printf(out, "bi_version=%llu\n", inode->bi_version);
#define x(_name, _bits) \
prt_printf(out, #_name "=%llu", (u64) inode->_name); \
prt_newline(out);
prt_printf(out, #_name "=%llu\n", (u64) inode->_name);
BCH_INODE_FIELDS_v3()
#undef x
printbuf_indent_sub(out, 2);
@ -604,11 +594,11 @@ int bch2_trigger_inode(struct btree_trans *trans,
enum btree_id btree_id, unsigned level,
struct bkey_s_c old,
struct bkey_s new,
unsigned flags)
enum btree_iter_update_trigger_flags flags)
{
s64 nr = (s64) bkey_is_inode(new.k) - (s64) bkey_is_inode(old.k);
if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
if (flags & BTREE_TRIGGER_transactional) {
if (nr) {
int ret = bch2_replicas_deltas_realloc(trans, 0);
if (ret)
@ -627,13 +617,13 @@ int bch2_trigger_inode(struct btree_trans *trans,
}
}
if ((flags & BTREE_TRIGGER_ATOMIC) && (flags & BTREE_TRIGGER_INSERT)) {
if ((flags & BTREE_TRIGGER_atomic) && (flags & BTREE_TRIGGER_insert)) {
BUG_ON(!trans->journal_res.seq);
bkey_s_to_inode_v3(new).v->bi_journal_seq = cpu_to_le64(trans->journal_res.seq);
}
if (flags & BTREE_TRIGGER_GC) {
if (flags & BTREE_TRIGGER_gc) {
struct bch_fs *c = trans->c;
percpu_down_read(&c->mark_lock);
@ -645,7 +635,7 @@ int bch2_trigger_inode(struct btree_trans *trans,
}
int bch2_inode_generation_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
int ret = 0;
@ -762,8 +752,8 @@ int bch2_inode_create(struct btree_trans *trans,
pos = start;
bch2_trans_iter_init(trans, iter, BTREE_ID_inodes, POS(0, pos),
BTREE_ITER_ALL_SNAPSHOTS|
BTREE_ITER_INTENT);
BTREE_ITER_all_snapshots|
BTREE_ITER_intent);
again:
while ((k = bch2_btree_iter_peek(iter)).k &&
!(ret = bkey_err(k)) &&
@ -824,7 +814,7 @@ static int bch2_inode_delete_keys(struct btree_trans *trans,
* extent iterator:
*/
bch2_trans_iter_init(trans, &iter, id, POS(inum.inum, 0),
BTREE_ITER_INTENT);
BTREE_ITER_intent);
while (1) {
bch2_trans_begin(trans);
@ -846,7 +836,7 @@ static int bch2_inode_delete_keys(struct btree_trans *trans,
bkey_init(&delete.k);
delete.k.p = iter.pos;
if (iter.flags & BTREE_ITER_IS_EXTENTS)
if (iter.flags & BTREE_ITER_is_extents)
bch2_key_resize(&delete.k,
bpos_min(end, k.k->p).offset -
iter.pos.offset);
@ -895,7 +885,7 @@ retry:
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
SPOS(0, inum.inum, snapshot),
BTREE_ITER_INTENT|BTREE_ITER_CACHED);
BTREE_ITER_intent|BTREE_ITER_cached);
ret = bkey_err(k);
if (ret)
goto err;
@ -1055,7 +1045,7 @@ retry:
bch2_trans_begin(trans);
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
SPOS(0, inum, snapshot), BTREE_ITER_INTENT);
SPOS(0, inum, snapshot), BTREE_ITER_intent);
ret = bkey_err(k);
if (ret)
goto err;
@ -1100,7 +1090,7 @@ static int may_delete_deleted_inode(struct btree_trans *trans,
struct bch_inode_unpacked inode;
int ret;
k = bch2_bkey_get_iter(trans, &inode_iter, BTREE_ID_inodes, pos, BTREE_ITER_CACHED);
k = bch2_bkey_get_iter(trans, &inode_iter, BTREE_ID_inodes, pos, BTREE_ITER_cached);
ret = bkey_err(k);
if (ret)
return ret;
@ -1152,7 +1142,7 @@ static int may_delete_deleted_inode(struct btree_trans *trans,
inode.bi_flags &= ~BCH_INODE_unlinked;
ret = bch2_inode_write_flags(trans, &inode_iter, &inode,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
bch_err_msg(c, ret, "clearing inode unlinked flag");
if (ret)
goto out;
@ -1199,7 +1189,7 @@ again:
* flushed and we'd spin:
*/
ret = for_each_btree_key_commit(trans, iter, BTREE_ID_deleted_inodes, POS_MIN,
BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc, ({
ret = may_delete_deleted_inode(trans, &iter, k.k->p, &need_another_pass);
if (ret > 0) {

23
fs/bcachefs/inode.h
View File

@ -6,19 +6,20 @@
#include "bkey_methods.h"
#include "opts.h"
enum bkey_invalid_flags;
enum bch_validate_flags;
extern const char * const bch2_inode_opts[];
int bch2_inode_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
int bch2_inode_v2_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
int bch2_inode_v3_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_inode_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
int bch2_trigger_inode(struct btree_trans *, enum btree_id, unsigned,
struct bkey_s_c, struct bkey_s, unsigned);
struct bkey_s_c, struct bkey_s,
enum btree_iter_update_trigger_flags);
#define bch2_bkey_ops_inode ((struct bkey_ops) { \
.key_invalid = bch2_inode_invalid, \
@ -49,7 +50,7 @@ static inline bool bkey_is_inode(const struct bkey *k)
}
int bch2_inode_generation_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_inode_generation_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
#define bch2_bkey_ops_inode_generation ((struct bkey_ops) { \
@ -101,7 +102,7 @@ int bch2_inode_peek(struct btree_trans *, struct btree_iter *,
struct bch_inode_unpacked *, subvol_inum, unsigned);
int bch2_inode_write_flags(struct btree_trans *, struct btree_iter *,
struct bch_inode_unpacked *, enum btree_update_flags);
struct bch_inode_unpacked *, enum btree_iter_update_trigger_flags);
static inline int bch2_inode_write(struct btree_trans *trans,
struct btree_iter *iter,
@ -220,6 +221,14 @@ static inline void bch2_inode_nlink_set(struct bch_inode_unpacked *bi,
int bch2_inode_nlink_inc(struct bch_inode_unpacked *);
void bch2_inode_nlink_dec(struct btree_trans *, struct bch_inode_unpacked *);
static inline bool bch2_inode_should_have_bp(struct bch_inode_unpacked *inode)
{
bool inode_has_bp = inode->bi_dir || inode->bi_dir_offset;
return S_ISDIR(inode->bi_mode) ||
(!inode->bi_nlink && inode_has_bp);
}
struct bch_opts bch2_inode_opts_to_opts(struct bch_inode_unpacked *);
void bch2_inode_opts_get(struct bch_io_opts *, struct bch_fs *,
struct bch_inode_unpacked *);

10
fs/bcachefs/io_misc.c
View File

@ -198,7 +198,7 @@ int bch2_fpunch(struct bch_fs *c, subvol_inum inum, u64 start, u64 end,
bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
POS(inum.inum, start),
BTREE_ITER_INTENT);
BTREE_ITER_intent);
ret = bch2_fpunch_at(trans, &iter, inum, end, i_sectors_delta);
@ -230,7 +230,7 @@ static int truncate_set_isize(struct btree_trans *trans,
struct bch_inode_unpacked inode_u;
int ret;
ret = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_INTENT) ?:
ret = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_intent) ?:
(inode_u.bi_size = new_i_size, 0) ?:
bch2_inode_write(trans, &iter, &inode_u);
@ -256,7 +256,7 @@ static int __bch2_resume_logged_op_truncate(struct btree_trans *trans,
bch2_trans_iter_init(trans, &fpunch_iter, BTREE_ID_extents,
POS(inum.inum, round_up(new_i_size, block_bytes(c)) >> 9),
BTREE_ITER_INTENT);
BTREE_ITER_intent);
ret = bch2_fpunch_at(trans, &fpunch_iter, inum, U64_MAX, i_sectors_delta);
bch2_trans_iter_exit(trans, &fpunch_iter);
@ -317,7 +317,7 @@ static int adjust_i_size(struct btree_trans *trans, subvol_inum inum, u64 offset
offset <<= 9;
len <<= 9;
ret = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_INTENT);
ret = bch2_inode_peek(trans, &iter, &inode_u, inum, BTREE_ITER_intent);
if (ret)
return ret;
@ -365,7 +365,7 @@ static int __bch2_resume_logged_op_finsert(struct btree_trans *trans,
bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
POS(inum.inum, 0),
BTREE_ITER_INTENT);
BTREE_ITER_intent);
switch (op->v.state) {
case LOGGED_OP_FINSERT_start:

68
fs/bcachefs/io_read.c
View File

@ -378,7 +378,7 @@ static void bch2_read_retry_nodecode(struct bch_fs *c, struct bch_read_bio *rbio
bch2_bkey_buf_init(&sk);
bch2_trans_iter_init(trans, &iter, rbio->data_btree,
rbio->read_pos, BTREE_ITER_SLOTS);
rbio->read_pos, BTREE_ITER_slots);
retry:
rbio->bio.bi_status = 0;
@ -487,7 +487,7 @@ static int __bch2_rbio_narrow_crcs(struct btree_trans *trans,
return 0;
k = bch2_bkey_get_iter(trans, &iter, rbio->data_btree, rbio->data_pos,
BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
BTREE_ITER_slots|BTREE_ITER_intent);
if ((ret = bkey_err(k)))
goto out;
@ -523,7 +523,7 @@ static int __bch2_rbio_narrow_crcs(struct btree_trans *trans,
goto out;
ret = bch2_trans_update(trans, &iter, new,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
out:
bch2_trans_iter_exit(trans, &iter);
return ret;
@ -541,7 +541,6 @@ static void __bch2_read_endio(struct work_struct *work)
struct bch_read_bio *rbio =
container_of(work, struct bch_read_bio, work);
struct bch_fs *c = rbio->c;
struct bch_dev *ca = bch_dev_bkey_exists(c, rbio->pick.ptr.dev);
struct bio *src = &rbio->bio;
struct bio *dst = &bch2_rbio_parent(rbio)->bio;
struct bvec_iter dst_iter = rbio->bvec_iter;
@ -647,13 +646,15 @@ csum_err:
prt_str(&buf, "data ");
bch2_csum_err_msg(&buf, crc.csum_type, rbio->pick.crc.csum, csum);
bch_err_inum_offset_ratelimited(ca,
rbio->read_pos.inode,
rbio->read_pos.offset << 9,
"data %s", buf.buf);
struct bch_dev *ca = rbio->have_ioref ? bch2_dev_have_ref(c, rbio->pick.ptr.dev) : NULL;
if (ca) {
bch_err_inum_offset_ratelimited(ca,
rbio->read_pos.inode,
rbio->read_pos.offset << 9,
"data %s", buf.buf);
bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
}
printbuf_exit(&buf);
bch2_io_error(ca, BCH_MEMBER_ERROR_checksum);
bch2_rbio_error(rbio, READ_RETRY_AVOID, BLK_STS_IOERR);
goto out;
decompression_err:
@ -675,7 +676,7 @@ static void bch2_read_endio(struct bio *bio)
struct bch_read_bio *rbio =
container_of(bio, struct bch_read_bio, bio);
struct bch_fs *c = rbio->c;
struct bch_dev *ca = bch_dev_bkey_exists(c, rbio->pick.ptr.dev);
struct bch_dev *ca = rbio->have_ioref ? bch2_dev_have_ref(c, rbio->pick.ptr.dev) : NULL;
struct workqueue_struct *wq = NULL;
enum rbio_context context = RBIO_CONTEXT_NULL;
@ -687,17 +688,21 @@ static void bch2_read_endio(struct bio *bio)
if (!rbio->split)
rbio->bio.bi_end_io = rbio->end_io;
if (bch2_dev_inum_io_err_on(bio->bi_status, ca, BCH_MEMBER_ERROR_read,
rbio->read_pos.inode,
rbio->read_pos.offset,
"data read error: %s",
bch2_blk_status_to_str(bio->bi_status))) {
if (bio->bi_status) {
if (ca) {
bch_err_inum_offset_ratelimited(ca,
rbio->read_pos.inode,
rbio->read_pos.offset,
"data read error: %s",
bch2_blk_status_to_str(bio->bi_status));
bch2_io_error(ca, BCH_MEMBER_ERROR_read);
}
bch2_rbio_error(rbio, READ_RETRY_AVOID, bio->bi_status);
return;
}
if (((rbio->flags & BCH_READ_RETRY_IF_STALE) && race_fault()) ||
ptr_stale(ca, &rbio->pick.ptr)) {
(ca && dev_ptr_stale(ca, &rbio->pick.ptr))) {
trace_and_count(c, read_reuse_race, &rbio->bio);
if (rbio->flags & BCH_READ_RETRY_IF_STALE)
@ -758,22 +763,21 @@ err:
}
static noinline void read_from_stale_dirty_pointer(struct btree_trans *trans,
struct bch_dev *ca,
struct bkey_s_c k,
struct bch_extent_ptr ptr)
{
struct bch_fs *c = trans->c;
struct bch_dev *ca = bch_dev_bkey_exists(c, ptr.dev);
struct btree_iter iter;
struct printbuf buf = PRINTBUF;
int ret;
bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc,
PTR_BUCKET_POS(c, &ptr),
BTREE_ITER_CACHED);
PTR_BUCKET_POS(ca, &ptr),
BTREE_ITER_cached);
prt_printf(&buf, "Attempting to read from stale dirty pointer:");
prt_printf(&buf, "Attempting to read from stale dirty pointer:\n");
printbuf_indent_add(&buf, 2);
prt_newline(&buf);
bch2_bkey_val_to_text(&buf, c, k);
prt_newline(&buf);
@ -801,7 +805,6 @@ int __bch2_read_extent(struct btree_trans *trans, struct bch_read_bio *orig,
struct bch_fs *c = trans->c;
struct extent_ptr_decoded pick;
struct bch_read_bio *rbio = NULL;
struct bch_dev *ca = NULL;
struct promote_op *promote = NULL;
bool bounce = false, read_full = false, narrow_crcs = false;
struct bpos data_pos = bkey_start_pos(k.k);
@ -832,7 +835,7 @@ retry_pick:
goto err;
}
ca = bch_dev_bkey_exists(c, pick.ptr.dev);
struct bch_dev *ca = bch2_dev_get_ioref(c, pick.ptr.dev, READ);
/*
* Stale dirty pointers are treated as IO errors, but @failed isn't
@ -842,9 +845,11 @@ retry_pick:
*/
if ((flags & BCH_READ_IN_RETRY) &&
!pick.ptr.cached &&
unlikely(ptr_stale(ca, &pick.ptr))) {
read_from_stale_dirty_pointer(trans, k, pick.ptr);
ca &&
unlikely(dev_ptr_stale(ca, &pick.ptr))) {
read_from_stale_dirty_pointer(trans, ca, k, pick.ptr);
bch2_mark_io_failure(failed, &pick);
percpu_ref_put(&ca->io_ref);
goto retry_pick;
}
@ -859,8 +864,11 @@ retry_pick:
* can happen if we retry, and the extent we were going to read
* has been merged in the meantime:
*/
if (pick.crc.compressed_size > orig->bio.bi_vcnt * PAGE_SECTORS)
if (pick.crc.compressed_size > orig->bio.bi_vcnt * PAGE_SECTORS) {
if (ca)
percpu_ref_put(&ca->io_ref);
goto hole;
}
iter.bi_size = pick.crc.compressed_size << 9;
goto get_bio;
@ -965,7 +973,7 @@ get_bio:
rbio->bvec_iter = iter;
rbio->offset_into_extent= offset_into_extent;
rbio->flags = flags;
rbio->have_ioref = pick_ret > 0 && bch2_dev_get_ioref(ca, READ);
rbio->have_ioref = ca != NULL;
rbio->narrow_crcs = narrow_crcs;
rbio->hole = 0;
rbio->retry = 0;
@ -995,7 +1003,7 @@ get_bio:
* If it's being moved internally, we don't want to flag it as a cache
* hit:
*/
if (pick.ptr.cached && !(flags & BCH_READ_NODECODE))
if (ca && pick.ptr.cached && !(flags & BCH_READ_NODECODE))
bch2_bucket_io_time_reset(trans, pick.ptr.dev,
PTR_BUCKET_NR(ca, &pick.ptr), READ);
@ -1113,7 +1121,7 @@ retry:
bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
SPOS(inum.inum, bvec_iter.bi_sector, snapshot),
BTREE_ITER_SLOTS);
BTREE_ITER_slots);
while (1) {
unsigned bytes, sectors, offset_into_extent;
enum btree_id data_btree = BTREE_ID_extents;

95
fs/bcachefs/io_write.c
View File

@ -166,7 +166,7 @@ int bch2_sum_sector_overwrites(struct btree_trans *trans,
bch2_trans_copy_iter(&iter, extent_iter);
for_each_btree_key_upto_continue_norestart(iter,
new->k.p, BTREE_ITER_SLOTS, old, ret) {
new->k.p, BTREE_ITER_slots, old, ret) {
s64 sectors = min(new->k.p.offset, old.k->p.offset) -
max(bkey_start_offset(&new->k),
bkey_start_offset(old.k));
@ -210,14 +210,14 @@ static inline int bch2_extent_update_i_size_sectors(struct btree_trans *trans,
* to be journalled - if we crash, the bi_journal_seq update will be
* lost, but that's fine.
*/
unsigned inode_update_flags = BTREE_UPDATE_NOJOURNAL;
unsigned inode_update_flags = BTREE_UPDATE_nojournal;
struct btree_iter iter;
struct bkey_s_c k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
SPOS(0,
extent_iter->pos.inode,
extent_iter->snapshot),
BTREE_ITER_CACHED);
BTREE_ITER_cached);
int ret = bkey_err(k);
if (unlikely(ret))
return ret;
@ -259,7 +259,7 @@ static inline int bch2_extent_update_i_size_sectors(struct btree_trans *trans,
}
ret = bch2_trans_update(trans, &iter, &inode->k_i,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|
BTREE_UPDATE_internal_snapshot_node|
inode_update_flags);
err:
bch2_trans_iter_exit(trans, &iter);
@ -368,7 +368,7 @@ static int bch2_write_index_default(struct bch_write_op *op)
bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
bkey_start_pos(&sk.k->k),
BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
BTREE_ITER_slots|BTREE_ITER_intent);
ret = bch2_bkey_set_needs_rebalance(c, sk.k, &op->opts) ?:
bch2_extent_update(trans, inum, &iter, sk.k,
@ -407,13 +407,12 @@ void bch2_submit_wbio_replicas(struct bch_write_bio *wbio, struct bch_fs *c,
BUG_ON(c->opts.nochanges);
bkey_for_each_ptr(ptrs, ptr) {
BUG_ON(!bch2_dev_exists2(c, ptr->dev));
struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
struct bch_dev *ca = nocow
? bch2_dev_have_ref(c, ptr->dev)
: bch2_dev_get_ioref(c, ptr->dev, type == BCH_DATA_btree ? READ : WRITE);
if (to_entry(ptr + 1) < ptrs.end) {
n = to_wbio(bio_alloc_clone(NULL, &wbio->bio,
GFP_NOFS, &ca->replica_set));
n = to_wbio(bio_alloc_clone(NULL, &wbio->bio, GFP_NOFS, &c->replica_set));
n->bio.bi_end_io = wbio->bio.bi_end_io;
n->bio.bi_private = wbio->bio.bi_private;
@ -430,11 +429,12 @@ void bch2_submit_wbio_replicas(struct bch_write_bio *wbio, struct bch_fs *c,
n->c = c;
n->dev = ptr->dev;
n->have_ioref = nocow || bch2_dev_get_ioref(ca,
type == BCH_DATA_btree ? READ : WRITE);
n->have_ioref = ca != NULL;
n->nocow = nocow;
n->submit_time = local_clock();
n->inode_offset = bkey_start_offset(&k->k);
if (nocow)
n->nocow_bucket = PTR_BUCKET_NR(ca, ptr);
n->bio.bi_iter.bi_sector = ptr->offset;
if (likely(n->have_ioref)) {
@ -481,7 +481,6 @@ static void bch2_write_done(struct closure *cl)
static noinline int bch2_write_drop_io_error_ptrs(struct bch_write_op *op)
{
struct keylist *keys = &op->insert_keys;
struct bch_extent_ptr *ptr;
struct bkey_i *src, *dst = keys->keys, *n;
for (src = keys->keys; src != keys->top; src = n) {
@ -650,7 +649,9 @@ static void bch2_write_endio(struct bio *bio)
struct bch_write_bio *wbio = to_wbio(bio);
struct bch_write_bio *parent = wbio->split ? wbio->parent : NULL;
struct bch_fs *c = wbio->c;
struct bch_dev *ca = bch_dev_bkey_exists(c, wbio->dev);
struct bch_dev *ca = wbio->have_ioref
? bch2_dev_have_ref(c, wbio->dev)
: NULL;
if (bch2_dev_inum_io_err_on(bio->bi_status, ca, BCH_MEMBER_ERROR_write,
op->pos.inode,
@ -661,8 +662,12 @@ static void bch2_write_endio(struct bio *bio)
op->flags |= BCH_WRITE_IO_ERROR;
}
if (wbio->nocow)
if (wbio->nocow) {
bch2_bucket_nocow_unlock(&c->nocow_locks,
POS(ca->dev_idx, wbio->nocow_bucket),
BUCKET_NOCOW_LOCK_UPDATE);
set_bit(wbio->dev, op->devs_need_flush->d);
}
if (wbio->have_ioref) {
bch2_latency_acct(ca, wbio->submit_time, WRITE);
@ -1101,30 +1106,21 @@ static bool bch2_extent_is_writeable(struct bch_write_op *op,
return false;
e = bkey_s_c_to_extent(k);
rcu_read_lock();
extent_for_each_ptr_decode(e, p, entry) {
if (crc_is_encoded(p.crc) || p.has_ec)
if (crc_is_encoded(p.crc) || p.has_ec) {
rcu_read_unlock();
return false;
}
replicas += bch2_extent_ptr_durability(c, &p);
}
rcu_read_unlock();
return replicas >= op->opts.data_replicas;
}
static inline void bch2_nocow_write_unlock(struct bch_write_op *op)
{
struct bch_fs *c = op->c;
for_each_keylist_key(&op->insert_keys, k) {
struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(bkey_i_to_s_c(k));
bkey_for_each_ptr(ptrs, ptr)
bch2_bucket_nocow_unlock(&c->nocow_locks,
PTR_BUCKET_POS(c, ptr),
BUCKET_NOCOW_LOCK_UPDATE);
}
}
static int bch2_nocow_write_convert_one_unwritten(struct btree_trans *trans,
struct btree_iter *iter,
struct bkey_i *orig,
@ -1158,7 +1154,7 @@ static int bch2_nocow_write_convert_one_unwritten(struct btree_trans *trans,
return bch2_extent_update_i_size_sectors(trans, iter,
min(new->k.p.offset << 9, new_i_size), 0) ?:
bch2_trans_update(trans, iter, new,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
}
static void bch2_nocow_write_convert_unwritten(struct bch_write_op *op)
@ -1169,7 +1165,7 @@ static void bch2_nocow_write_convert_unwritten(struct bch_write_op *op)
for_each_keylist_key(&op->insert_keys, orig) {
int ret = for_each_btree_key_upto_commit(trans, iter, BTREE_ID_extents,
bkey_start_pos(&orig->k), orig->k.p,
BTREE_ITER_INTENT, k,
BTREE_ITER_intent, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc, ({
bch2_nocow_write_convert_one_unwritten(trans, &iter, orig, k, op->new_i_size);
}));
@ -1195,8 +1191,6 @@ static void bch2_nocow_write_convert_unwritten(struct bch_write_op *op)
static void __bch2_nocow_write_done(struct bch_write_op *op)
{
bch2_nocow_write_unlock(op);
if (unlikely(op->flags & BCH_WRITE_IO_ERROR)) {
op->error = -EIO;
} else if (unlikely(op->flags & BCH_WRITE_CONVERT_UNWRITTEN))
@ -1242,12 +1236,16 @@ retry:
bch2_trans_iter_init(trans, &iter, BTREE_ID_extents,
SPOS(op->pos.inode, op->pos.offset, snapshot),
BTREE_ITER_SLOTS);
BTREE_ITER_slots);
while (1) {
struct bio *bio = &op->wbio.bio;
buckets.nr = 0;
ret = bch2_trans_relock(trans);
if (ret)
break;
k = bch2_btree_iter_peek_slot(&iter);
ret = bkey_err(k);
if (ret)
@ -1267,14 +1265,15 @@ retry:
/* Get iorefs before dropping btree locks: */
struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
bkey_for_each_ptr(ptrs, ptr) {
struct bpos b = PTR_BUCKET_POS(c, ptr);
struct bch_dev *ca = bch2_dev_get_ioref(c, ptr->dev, WRITE);
if (unlikely(!ca))
goto err_get_ioref;
struct bpos b = PTR_BUCKET_POS(ca, ptr);
struct nocow_lock_bucket *l =
bucket_nocow_lock(&c->nocow_locks, bucket_to_u64(b));
prefetch(l);
if (unlikely(!bch2_dev_get_ioref(bch_dev_bkey_exists(c, ptr->dev), WRITE)))
goto err_get_ioref;
/* XXX allocating memory with btree locks held - rare */
darray_push_gfp(&buckets, ((struct bucket_to_lock) {
.b = b, .gen = ptr->gen, .l = l,
@ -1293,7 +1292,7 @@ retry:
bch2_cut_back(POS(op->pos.inode, op->pos.offset + bio_sectors(bio)), op->insert_keys.top);
darray_for_each(buckets, i) {
struct bch_dev *ca = bch_dev_bkey_exists(c, i->b.inode);
struct bch_dev *ca = bch2_dev_have_ref(c, i->b.inode);
__bch2_bucket_nocow_lock(&c->nocow_locks, i->l,
bucket_to_u64(i->b),
@ -1370,7 +1369,7 @@ err:
return;
err_get_ioref:
darray_for_each(buckets, i)
percpu_ref_put(&bch_dev_bkey_exists(c, i->b.inode)->io_ref);
percpu_ref_put(&bch2_dev_have_ref(c, i->b.inode)->io_ref);
/* Fall back to COW path: */
goto out;
@ -1491,7 +1490,11 @@ err:
if ((op->flags & BCH_WRITE_SYNC) ||
(!(op->flags & BCH_WRITE_DONE) &&
!(op->flags & BCH_WRITE_IN_WORKER))) {
closure_sync(&op->cl);
if (closure_sync_timeout(&op->cl, HZ * 10)) {
bch2_print_allocator_stuck(c);
closure_sync(&op->cl);
}
__bch2_write_index(op);
if (!(op->flags & BCH_WRITE_DONE))
@ -1649,8 +1652,7 @@ void bch2_write_op_to_text(struct printbuf *out, struct bch_write_op *op)
prt_bitflags(out, bch2_write_flags, op->flags);
prt_newline(out);
prt_printf(out, "ref: %u", closure_nr_remaining(&op->cl));
prt_newline(out);
prt_printf(out, "ref: %u\n", closure_nr_remaining(&op->cl));
printbuf_indent_sub(out, 2);
}
@ -1658,13 +1660,14 @@ void bch2_write_op_to_text(struct printbuf *out, struct bch_write_op *op)
void bch2_fs_io_write_exit(struct bch_fs *c)
{
mempool_exit(&c->bio_bounce_pages);
bioset_exit(&c->replica_set);
bioset_exit(&c->bio_write);
}
int bch2_fs_io_write_init(struct bch_fs *c)
{
if (bioset_init(&c->bio_write, 1, offsetof(struct bch_write_bio, bio),
BIOSET_NEED_BVECS))
if (bioset_init(&c->bio_write, 1, offsetof(struct bch_write_bio, bio), BIOSET_NEED_BVECS) ||
bioset_init(&c->replica_set, 4, offsetof(struct bch_write_bio, bio), 0))
return -BCH_ERR_ENOMEM_bio_write_init;
if (mempool_init_page_pool(&c->bio_bounce_pages,

1
fs/bcachefs/io_write_types.h
View File

@ -20,6 +20,7 @@ struct bch_write_bio {
u64 submit_time;
u64 inode_offset;
u64 nocow_bucket;
struct bch_devs_list failed;
u8 dev;

131
fs/bcachefs/journal.c
View File

@ -53,29 +53,19 @@ static void bch2_journal_buf_to_text(struct printbuf *out, struct journal *j, u6
unsigned i = seq & JOURNAL_BUF_MASK;
struct journal_buf *buf = j->buf + i;
prt_str(out, "seq:");
prt_tab(out);
prt_printf(out, "%llu", seq);
prt_newline(out);
prt_printf(out, "seq:\t%llu\n", seq);
printbuf_indent_add(out, 2);
prt_str(out, "refcount:");
prt_tab(out);
prt_printf(out, "%u", journal_state_count(s, i));
prt_newline(out);
prt_printf(out, "refcount:\t%u\n", journal_state_count(s, i));
prt_str(out, "size:");
prt_tab(out);
prt_printf(out, "size:\t");
prt_human_readable_u64(out, vstruct_bytes(buf->data));
prt_newline(out);
prt_str(out, "expires:");
prt_tab(out);
prt_printf(out, "%li jiffies", buf->expires - jiffies);
prt_newline(out);
prt_printf(out, "expires:\t");
prt_printf(out, "%li jiffies\n", buf->expires - jiffies);
prt_str(out, "flags:");
prt_tab(out);
prt_printf(out, "flags:\t");
if (buf->noflush)
prt_str(out, "noflush ");
if (buf->must_flush)
@ -87,9 +77,9 @@ static void bch2_journal_buf_to_text(struct printbuf *out, struct journal *j, u6
if (buf->write_started)
prt_str(out, "write_started ");
if (buf->write_allocated)
prt_str(out, "write allocated ");
prt_str(out, "write_allocated ");
if (buf->write_done)
prt_str(out, "write done");
prt_str(out, "write_done");
prt_newline(out);
printbuf_indent_sub(out, 2);
@ -948,7 +938,8 @@ static int __bch2_set_nr_journal_buckets(struct bch_dev *ca, unsigned nr,
break;
}
} else {
ob[nr_got] = bch2_bucket_alloc(c, ca, BCH_WATERMARK_normal, cl);
ob[nr_got] = bch2_bucket_alloc(c, ca, BCH_WATERMARK_normal,
BCH_DATA_journal, cl);
ret = PTR_ERR_OR_ZERO(ob[nr_got]);
if (ret)
break;
@ -956,7 +947,7 @@ static int __bch2_set_nr_journal_buckets(struct bch_dev *ca, unsigned nr,
ret = bch2_trans_run(c,
bch2_trans_mark_metadata_bucket(trans, ca,
ob[nr_got]->bucket, BCH_DATA_journal,
ca->mi.bucket_size));
ca->mi.bucket_size, BTREE_TRIGGER_transactional));
if (ret) {
bch2_open_bucket_put(c, ob[nr_got]);
bch_err_msg(c, ret, "marking new journal buckets");
@ -1036,7 +1027,8 @@ err_unblock:
for (i = 0; i < nr_got; i++)
bch2_trans_run(c,
bch2_trans_mark_metadata_bucket(trans, ca,
bu[i], BCH_DATA_free, 0));
bu[i], BCH_DATA_free, 0,
BTREE_TRIGGER_transactional));
err_free:
if (!new_fs)
for (i = 0; i < nr_got; i++)
@ -1187,12 +1179,14 @@ void bch2_fs_journal_stop(struct journal *j)
bch2_journal_meta(j);
journal_quiesce(j);
cancel_delayed_work_sync(&j->write_work);
BUG_ON(!bch2_journal_error(j) &&
test_bit(JOURNAL_REPLAY_DONE, &j->flags) &&
test_bit(JOURNAL_replay_done, &j->flags) &&
j->last_empty_seq != journal_cur_seq(j));
cancel_delayed_work_sync(&j->write_work);
if (!bch2_journal_error(j))
clear_bit(JOURNAL_running, &j->flags);
}
int bch2_fs_journal_start(struct journal *j, u64 cur_seq)
@ -1266,7 +1260,7 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq)
spin_lock(&j->lock);
set_bit(JOURNAL_STARTED, &j->flags);
set_bit(JOURNAL_running, &j->flags);
j->last_flush_write = jiffies;
j->reservations.idx = j->reservations.unwritten_idx = journal_cur_seq(j);
@ -1407,6 +1401,13 @@ int bch2_fs_journal_init(struct journal *j)
/* debug: */
static const char * const bch2_journal_flags_strs[] = {
#define x(n) #n,
JOURNAL_FLAGS()
#undef x
NULL
};
void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
{
struct bch_fs *c = container_of(j, struct bch_fs, journal);
@ -1415,19 +1416,22 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
u64 nr_writes = j->nr_flush_writes + j->nr_noflush_writes;
if (!out->nr_tabstops)
printbuf_tabstop_push(out, 24);
printbuf_tabstop_push(out, 28);
out->atomic++;
rcu_read_lock();
s = READ_ONCE(j->reservations);
prt_printf(out, "flags:\t");
prt_bitflags(out, bch2_journal_flags_strs, j->flags);
prt_newline(out);
prt_printf(out, "dirty journal entries:\t%llu/%llu\n", fifo_used(&j->pin), j->pin.size);
prt_printf(out, "seq:\t\t\t%llu\n", journal_cur_seq(j));
prt_printf(out, "seq_ondisk:\t\t%llu\n", j->seq_ondisk);
prt_printf(out, "last_seq:\t\t%llu\n", journal_last_seq(j));
prt_printf(out, "seq:\t%llu\n", journal_cur_seq(j));
prt_printf(out, "seq_ondisk:\t%llu\n", j->seq_ondisk);
prt_printf(out, "last_seq:\t%llu\n", journal_last_seq(j));
prt_printf(out, "last_seq_ondisk:\t%llu\n", j->last_seq_ondisk);
prt_printf(out, "flushed_seq_ondisk:\t%llu\n", j->flushed_seq_ondisk);
prt_printf(out, "watermark:\t\t%s\n", bch2_watermarks[j->watermark]);
prt_printf(out, "watermark:\t%s\n", bch2_watermarks[j->watermark]);
prt_printf(out, "each entry reserved:\t%u\n", j->entry_u64s_reserved);
prt_printf(out, "nr flush writes:\t%llu\n", j->nr_flush_writes);
prt_printf(out, "nr noflush writes:\t%llu\n", j->nr_noflush_writes);
@ -1436,48 +1440,44 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
prt_newline(out);
prt_printf(out, "nr direct reclaim:\t%llu\n", j->nr_direct_reclaim);
prt_printf(out, "nr background reclaim:\t%llu\n", j->nr_background_reclaim);
prt_printf(out, "reclaim kicked:\t\t%u\n", j->reclaim_kicked);
prt_printf(out, "reclaim kicked:\t%u\n", j->reclaim_kicked);
prt_printf(out, "reclaim runs in:\t%u ms\n", time_after(j->next_reclaim, now)
? jiffies_to_msecs(j->next_reclaim - jiffies) : 0);
prt_printf(out, "blocked:\t\t%u\n", j->blocked);
prt_printf(out, "blocked:\t%u\n", j->blocked);
prt_printf(out, "current entry sectors:\t%u\n", j->cur_entry_sectors);
prt_printf(out, "current entry error:\t%s\n", bch2_journal_errors[j->cur_entry_error]);
prt_printf(out, "current entry:\t\t");
prt_printf(out, "current entry:\t");
switch (s.cur_entry_offset) {
case JOURNAL_ENTRY_ERROR_VAL:
prt_printf(out, "error");
prt_printf(out, "error\n");
break;
case JOURNAL_ENTRY_CLOSED_VAL:
prt_printf(out, "closed");
prt_printf(out, "closed\n");
break;
default:
prt_printf(out, "%u/%u", s.cur_entry_offset, j->cur_entry_u64s);
prt_printf(out, "%u/%u\n", s.cur_entry_offset, j->cur_entry_u64s);
break;
}
prt_newline(out);
prt_printf(out, "unwritten entries:");
prt_newline(out);
prt_printf(out, "unwritten entries:\n");
bch2_journal_bufs_to_text(out, j);
prt_printf(out,
"replay done:\t\t%i\n",
test_bit(JOURNAL_REPLAY_DONE, &j->flags));
prt_printf(out, "space:\n");
prt_printf(out, "\tdiscarded\t%u:%u\n",
printbuf_indent_add(out, 2);
prt_printf(out, "discarded\t%u:%u\n",
j->space[journal_space_discarded].next_entry,
j->space[journal_space_discarded].total);
prt_printf(out, "\tclean ondisk\t%u:%u\n",
prt_printf(out, "clean ondisk\t%u:%u\n",
j->space[journal_space_clean_ondisk].next_entry,
j->space[journal_space_clean_ondisk].total);
prt_printf(out, "\tclean\t\t%u:%u\n",
prt_printf(out, "clean\t%u:%u\n",
j->space[journal_space_clean].next_entry,
j->space[journal_space_clean].total);
prt_printf(out, "\ttotal\t\t%u:%u\n",
prt_printf(out, "total\t%u:%u\n",
j->space[journal_space_total].next_entry,
j->space[journal_space_total].total);
printbuf_indent_sub(out, 2);
for_each_member_device_rcu(c, ca, &c->rw_devs[BCH_DATA_journal]) {
struct journal_device *ja = &ca->journal;
@ -1488,14 +1488,16 @@ void __bch2_journal_debug_to_text(struct printbuf *out, struct journal *j)
if (!ja->nr)
continue;
prt_printf(out, "dev %u:\n", ca->dev_idx);
prt_printf(out, "\tnr\t\t%u\n", ja->nr);
prt_printf(out, "\tbucket size\t%u\n", ca->mi.bucket_size);
prt_printf(out, "\tavailable\t%u:%u\n", bch2_journal_dev_buckets_available(j, ja, journal_space_discarded), ja->sectors_free);
prt_printf(out, "\tdiscard_idx\t%u\n", ja->discard_idx);
prt_printf(out, "\tdirty_ondisk\t%u (seq %llu)\n", ja->dirty_idx_ondisk, ja->bucket_seq[ja->dirty_idx_ondisk]);
prt_printf(out, "\tdirty_idx\t%u (seq %llu)\n", ja->dirty_idx, ja->bucket_seq[ja->dirty_idx]);
prt_printf(out, "\tcur_idx\t\t%u (seq %llu)\n", ja->cur_idx, ja->bucket_seq[ja->cur_idx]);
prt_printf(out, "dev %u:\n", ca->dev_idx);
printbuf_indent_add(out, 2);
prt_printf(out, "nr\t%u\n", ja->nr);
prt_printf(out, "bucket size\t%u\n", ca->mi.bucket_size);
prt_printf(out, "available\t%u:%u\n", bch2_journal_dev_buckets_available(j, ja, journal_space_discarded), ja->sectors_free);
prt_printf(out, "discard_idx\t%u\n", ja->discard_idx);
prt_printf(out, "dirty_ondisk\t%u (seq %llu)\n",ja->dirty_idx_ondisk, ja->bucket_seq[ja->dirty_idx_ondisk]);
prt_printf(out, "dirty_idx\t%u (seq %llu)\n", ja->dirty_idx, ja->bucket_seq[ja->dirty_idx]);
prt_printf(out, "cur_idx\t%u (seq %llu)\n", ja->cur_idx, ja->bucket_seq[ja->cur_idx]);
printbuf_indent_sub(out, 2);
}
rcu_read_unlock();
@ -1527,25 +1529,18 @@ bool bch2_journal_seq_pins_to_text(struct printbuf *out, struct journal *j, u64
pin_list = journal_seq_pin(j, *seq);
prt_printf(out, "%llu: count %u", *seq, atomic_read(&pin_list->count));
prt_newline(out);
prt_printf(out, "%llu: count %u\n", *seq, atomic_read(&pin_list->count));
printbuf_indent_add(out, 2);
for (unsigned i = 0; i < ARRAY_SIZE(pin_list->list); i++)
list_for_each_entry(pin, &pin_list->list[i], list) {
prt_printf(out, "\t%px %ps", pin, pin->flush);
prt_newline(out);
}
list_for_each_entry(pin, &pin_list->list[i], list)
prt_printf(out, "\t%px %ps\n", pin, pin->flush);
if (!list_empty(&pin_list->flushed)) {
prt_printf(out, "flushed:");
prt_newline(out);
}
if (!list_empty(&pin_list->flushed))
prt_printf(out, "flushed:\n");
list_for_each_entry(pin, &pin_list->flushed, list) {
prt_printf(out, "\t%px %ps", pin, pin->flush);
prt_newline(out);
}
list_for_each_entry(pin, &pin_list->flushed, list)
prt_printf(out, "\t%px %ps\n", pin, pin->flush);
printbuf_indent_sub(out, 2);

6
fs/bcachefs/journal.h
View File

@ -372,7 +372,7 @@ static inline int bch2_journal_res_get(struct journal *j, struct journal_res *re
int ret;
EBUG_ON(res->ref);
EBUG_ON(!test_bit(JOURNAL_STARTED, &j->flags));
EBUG_ON(!test_bit(JOURNAL_running, &j->flags));
res->u64s = u64s;
@ -418,8 +418,8 @@ struct bch_dev;
static inline void bch2_journal_set_replay_done(struct journal *j)
{
BUG_ON(!test_bit(JOURNAL_STARTED, &j->flags));
set_bit(JOURNAL_REPLAY_DONE, &j->flags);
BUG_ON(!test_bit(JOURNAL_running, &j->flags));
set_bit(JOURNAL_replay_done, &j->flags);
}
void bch2_journal_unblock(struct journal *);

163
fs/bcachefs/journal_io.c
View File

@ -17,15 +17,38 @@
#include "sb-clean.h"
#include "trace.h"
void bch2_journal_pos_from_member_info_set(struct bch_fs *c)
{
lockdep_assert_held(&c->sb_lock);
for_each_member_device(c, ca) {
struct bch_member *m = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
m->last_journal_bucket = cpu_to_le32(ca->journal.cur_idx);
m->last_journal_bucket_offset = cpu_to_le32(ca->mi.bucket_size - ca->journal.sectors_free);
}
}
void bch2_journal_pos_from_member_info_resume(struct bch_fs *c)
{
mutex_lock(&c->sb_lock);
for_each_member_device(c, ca) {
struct bch_member m = bch2_sb_member_get(c->disk_sb.sb, ca->dev_idx);
unsigned idx = le32_to_cpu(m.last_journal_bucket);
if (idx < ca->journal.nr)
ca->journal.cur_idx = idx;
unsigned offset = le32_to_cpu(m.last_journal_bucket_offset);
if (offset <= ca->mi.bucket_size)
ca->journal.sectors_free = ca->mi.bucket_size - offset;
}
mutex_unlock(&c->sb_lock);
}
void bch2_journal_ptrs_to_text(struct printbuf *out, struct bch_fs *c,
struct journal_replay *j)
{
darray_for_each(j->ptrs, i) {
struct bch_dev *ca = bch_dev_bkey_exists(c, i->dev);
u64 offset;
div64_u64_rem(i->sector, ca->mi.bucket_size, &offset);
if (i != j->ptrs.data)
prt_printf(out, " ");
prt_printf(out, "%u:%u:%u (sector %llu)",
@ -122,6 +145,10 @@ static int journal_entry_add(struct bch_fs *c, struct bch_dev *ca,
struct printbuf buf = PRINTBUF;
int ret = JOURNAL_ENTRY_ADD_OK;
if (!c->journal.oldest_seq_found_ondisk ||
le64_to_cpu(j->seq) < c->journal.oldest_seq_found_ondisk)
c->journal.oldest_seq_found_ondisk = le64_to_cpu(j->seq);
/* Is this entry older than the range we need? */
if (!c->opts.read_entire_journal &&
le64_to_cpu(j->seq) < jlist->last_seq)
@ -272,7 +299,7 @@ static void journal_entry_err_msg(struct printbuf *out,
journal_entry_err_msg(&_buf, version, jset, entry); \
prt_printf(&_buf, msg, ##__VA_ARGS__); \
\
switch (flags & BKEY_INVALID_WRITE) { \
switch (flags & BCH_VALIDATE_write) { \
case READ: \
mustfix_fsck_err(c, _err, "%s", _buf.buf); \
break; \
@ -301,9 +328,9 @@ static int journal_validate_key(struct bch_fs *c,
unsigned level, enum btree_id btree_id,
struct bkey_i *k,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
int write = flags & BKEY_INVALID_WRITE;
int write = flags & BCH_VALIDATE_write;
void *next = vstruct_next(entry);
struct printbuf buf = PRINTBUF;
int ret = 0;
@ -376,7 +403,7 @@ static int journal_entry_btree_keys_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
struct bkey_i *k = entry->start;
@ -385,7 +412,7 @@ static int journal_entry_btree_keys_validate(struct bch_fs *c,
entry->level,
entry->btree_id,
k, version, big_endian,
flags|BKEY_INVALID_JOURNAL);
flags|BCH_VALIDATE_journal);
if (ret == FSCK_DELETED_KEY)
continue;
@ -416,7 +443,7 @@ static int journal_entry_btree_root_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
struct bkey_i *k = entry->start;
int ret = 0;
@ -455,7 +482,7 @@ static int journal_entry_prio_ptrs_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
/* obsolete, don't care: */
return 0;
@ -470,7 +497,7 @@ static int journal_entry_blacklist_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
int ret = 0;
@ -497,7 +524,7 @@ static int journal_entry_blacklist_v2_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
struct jset_entry_blacklist_v2 *bl_entry;
int ret = 0;
@ -539,7 +566,7 @@ static int journal_entry_usage_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
struct jset_entry_usage *u =
container_of(entry, struct jset_entry_usage, entry);
@ -573,7 +600,7 @@ static int journal_entry_data_usage_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
struct jset_entry_data_usage *u =
container_of(entry, struct jset_entry_data_usage, entry);
@ -617,7 +644,7 @@ static int journal_entry_clock_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
struct jset_entry_clock *clock =
container_of(entry, struct jset_entry_clock, entry);
@ -657,13 +684,12 @@ static int journal_entry_dev_usage_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
struct jset_entry_dev_usage *u =
container_of(entry, struct jset_entry_dev_usage, entry);
unsigned bytes = jset_u64s(le16_to_cpu(entry->u64s)) * sizeof(u64);
unsigned expected = sizeof(*u);
unsigned dev;
int ret = 0;
if (journal_entry_err_on(bytes < expected,
@ -675,16 +701,6 @@ static int journal_entry_dev_usage_validate(struct bch_fs *c,
return ret;
}
dev = le32_to_cpu(u->dev);
if (journal_entry_err_on(!bch2_dev_exists2(c, dev),
c, version, jset, entry,
journal_entry_dev_usage_bad_dev,
"bad dev")) {
journal_entry_null_range(entry, vstruct_next(entry));
return ret;
}
if (journal_entry_err_on(u->pad,
c, version, jset, entry,
journal_entry_dev_usage_bad_pad,
@ -719,7 +735,7 @@ static int journal_entry_log_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
return 0;
}
@ -737,7 +753,7 @@ static int journal_entry_overwrite_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
return journal_entry_btree_keys_validate(c, jset, entry,
version, big_endian, READ);
@ -753,7 +769,7 @@ static int journal_entry_write_buffer_keys_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
return journal_entry_btree_keys_validate(c, jset, entry,
version, big_endian, READ);
@ -769,7 +785,7 @@ static int journal_entry_datetime_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
unsigned bytes = vstruct_bytes(entry);
unsigned expected = 16;
@ -799,7 +815,7 @@ static void journal_entry_datetime_to_text(struct printbuf *out, struct bch_fs *
struct jset_entry_ops {
int (*validate)(struct bch_fs *, struct jset *,
struct jset_entry *, unsigned, int,
enum bkey_invalid_flags);
enum bch_validate_flags);
void (*to_text)(struct printbuf *, struct bch_fs *, struct jset_entry *);
};
@ -817,7 +833,7 @@ int bch2_journal_entry_validate(struct bch_fs *c,
struct jset *jset,
struct jset_entry *entry,
unsigned version, int big_endian,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
return entry->type < BCH_JSET_ENTRY_NR
? bch2_jset_entry_ops[entry->type].validate(c, jset, entry,
@ -837,7 +853,7 @@ void bch2_journal_entry_to_text(struct printbuf *out, struct bch_fs *c,
}
static int jset_validate_entries(struct bch_fs *c, struct jset *jset,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
unsigned version = le32_to_cpu(jset->version);
int ret = 0;
@ -863,7 +879,7 @@ fsck_err:
static int jset_validate(struct bch_fs *c,
struct bch_dev *ca,
struct jset *jset, u64 sector,
enum bkey_invalid_flags flags)
enum bch_validate_flags flags)
{
unsigned version;
int ret = 0;
@ -918,7 +934,7 @@ static int jset_validate_early(struct bch_fs *c,
{
size_t bytes = vstruct_bytes(jset);
unsigned version;
enum bkey_invalid_flags flags = BKEY_INVALID_JOURNAL;
enum bch_validate_flags flags = BCH_VALIDATE_journal;
int ret = 0;
if (le64_to_cpu(jset->magic) != jset_magic(c))
@ -1057,6 +1073,13 @@ reread:
goto err;
}
if (le64_to_cpu(j->seq) > ja->highest_seq_found) {
ja->highest_seq_found = le64_to_cpu(j->seq);
ja->cur_idx = bucket;
ja->sectors_free = ca->mi.bucket_size -
bucket_remainder(ca, offset) - sectors;
}
/*
* This happens sometimes if we don't have discards on -
* when we've partially overwritten a bucket with new
@ -1125,8 +1148,6 @@ static CLOSURE_CALLBACK(bch2_journal_read_device)
struct bch_fs *c = ca->fs;
struct journal_list *jlist =
container_of(cl->parent, struct journal_list, cl);
struct journal_replay *r, **_r;
struct genradix_iter iter;
struct journal_read_buf buf = { NULL, 0 };
unsigned i;
int ret = 0;
@ -1146,47 +1167,6 @@ static CLOSURE_CALLBACK(bch2_journal_read_device)
goto err;
}
ja->sectors_free = ca->mi.bucket_size;
mutex_lock(&jlist->lock);
genradix_for_each_reverse(&c->journal_entries, iter, _r) {
r = *_r;
if (!r)
continue;
darray_for_each(r->ptrs, i)
if (i->dev == ca->dev_idx) {
unsigned wrote = bucket_remainder(ca, i->sector) +
vstruct_sectors(&r->j, c->block_bits);
ja->cur_idx = i->bucket;
ja->sectors_free = ca->mi.bucket_size - wrote;
goto found;
}
}
found:
mutex_unlock(&jlist->lock);
if (ja->bucket_seq[ja->cur_idx] &&
ja->sectors_free == ca->mi.bucket_size) {
#if 0
/*
* Debug code for ZNS support, where we (probably) want to be
* correlated where we stopped in the journal to the zone write
* points:
*/
bch_err(c, "ja->sectors_free == ca->mi.bucket_size");
bch_err(c, "cur_idx %u/%u", ja->cur_idx, ja->nr);
for (i = 0; i < 3; i++) {
unsigned idx = (ja->cur_idx + ja->nr - 1 + i) % ja->nr;
bch_err(c, "bucket_seq[%u] = %llu", idx, ja->bucket_seq[idx]);
}
#endif
ja->sectors_free = 0;
}
/*
* Set dirty_idx to indicate the entire journal is full and needs to be
* reclaimed - journal reclaim will immediately reclaim whatever isn't
@ -1255,7 +1235,7 @@ int bch2_journal_read(struct bch_fs *c,
* those entries will be blacklisted:
*/
genradix_for_each_reverse(&c->journal_entries, radix_iter, _i) {
enum bkey_invalid_flags flags = BKEY_INVALID_JOURNAL;
enum bch_validate_flags flags = BCH_VALIDATE_journal;
i = *_i;
@ -1366,7 +1346,7 @@ int bch2_journal_read(struct bch_fs *c,
fsck_err(c, journal_entries_missing,
"journal entries %llu-%llu missing! (replaying %llu-%llu)\n"
" prev at %s\n"
" next at %s",
" next at %s, continue?",
missing_start, missing_end,
*last_seq, *blacklist_seq - 1,
buf1.buf, buf2.buf);
@ -1390,7 +1370,7 @@ int bch2_journal_read(struct bch_fs *c,
continue;
darray_for_each(i->ptrs, ptr) {
struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
struct bch_dev *ca = bch2_dev_have_ref(c, ptr->dev);
if (!ptr->csum_good)
bch_err_dev_offset(ca, ptr->sector,
@ -1400,7 +1380,7 @@ int bch2_journal_read(struct bch_fs *c,
}
ret = jset_validate(c,
bch_dev_bkey_exists(c, i->ptrs.data[0].dev),
bch2_dev_have_ref(c, i->ptrs.data[0].dev),
&i->j,
i->ptrs.data[0].sector,
READ);
@ -1731,10 +1711,8 @@ static CLOSURE_CALLBACK(journal_write_submit)
unsigned sectors = vstruct_sectors(w->data, c->block_bits);
extent_for_each_ptr(bkey_i_to_s_extent(&w->key), ptr) {
struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
struct journal_device *ja = &ca->journal;
if (!percpu_ref_tryget(&ca->io_ref)) {
struct bch_dev *ca = bch2_dev_get_ioref(c, ptr->dev, WRITE);
if (!ca) {
/* XXX: fix this */
bch_err(c, "missing device for journal write\n");
continue;
@ -1743,6 +1721,7 @@ static CLOSURE_CALLBACK(journal_write_submit)
this_cpu_add(ca->io_done->sectors[WRITE][BCH_DATA_journal],
sectors);
struct journal_device *ja = &ca->journal;
struct bio *bio = &ja->bio[w->idx]->bio;
bio_reset(bio, ca->disk_sb.bdev, REQ_OP_WRITE|REQ_SYNC|REQ_META);
bio->bi_iter.bi_sector = ptr->offset;
@ -1958,14 +1937,14 @@ static int bch2_journal_write_pick_flush(struct journal *j, struct journal_buf *
* So if we're in an error state, and we're still starting up, we don't
* write anything at all.
*/
if (error && test_bit(JOURNAL_NEED_FLUSH_WRITE, &j->flags))
if (error && test_bit(JOURNAL_need_flush_write, &j->flags))
return -EIO;
if (error ||
w->noflush ||
(!w->must_flush &&
(jiffies - j->last_flush_write) < msecs_to_jiffies(c->opts.journal_flush_delay) &&
test_bit(JOURNAL_MAY_SKIP_FLUSH, &j->flags))) {
test_bit(JOURNAL_may_skip_flush, &j->flags))) {
w->noflush = true;
SET_JSET_NO_FLUSH(w->data, true);
w->data->last_seq = 0;
@ -1976,7 +1955,7 @@ static int bch2_journal_write_pick_flush(struct journal *j, struct journal_buf *
w->must_flush = true;
j->last_flush_write = jiffies;
j->nr_flush_writes++;
clear_bit(JOURNAL_NEED_FLUSH_WRITE, &j->flags);
clear_bit(JOURNAL_need_flush_write, &j->flags);
}
return 0;

5
fs/bcachefs/journal_io.h
View File

@ -4,6 +4,9 @@
#include "darray.h"
void bch2_journal_pos_from_member_info_set(struct bch_fs *);
void bch2_journal_pos_from_member_info_resume(struct bch_fs *);
struct journal_ptr {
bool csum_good;
u8 dev;
@ -60,7 +63,7 @@ static inline struct jset_entry *__jset_entry_type_next(struct jset *jset,
int bch2_journal_entry_validate(struct bch_fs *, struct jset *,
struct jset_entry *, unsigned, int,
enum bkey_invalid_flags);
enum bch_validate_flags);
void bch2_journal_entry_to_text(struct printbuf *, struct bch_fs *,
struct jset_entry *);

10
fs/bcachefs/journal_reclaim.c
View File

@ -67,7 +67,7 @@ void bch2_journal_set_watermark(struct journal *j)
track_event_change(&c->times[BCH_TIME_blocked_write_buffer_full], low_on_wb))
trace_and_count(c, journal_full, c);
mod_bit(JOURNAL_SPACE_LOW, &j->flags, low_on_space || low_on_pin);
mod_bit(JOURNAL_space_low, &j->flags, low_on_space || low_on_pin);
swap(watermark, j->watermark);
if (watermark > j->watermark)
@ -225,9 +225,9 @@ void bch2_journal_space_available(struct journal *j)
j->space[journal_space_clean_ondisk].total) &&
(clean - clean_ondisk <= total / 8) &&
(clean_ondisk * 2 > clean))
set_bit(JOURNAL_MAY_SKIP_FLUSH, &j->flags);
set_bit(JOURNAL_may_skip_flush, &j->flags);
else
clear_bit(JOURNAL_MAY_SKIP_FLUSH, &j->flags);
clear_bit(JOURNAL_may_skip_flush, &j->flags);
bch2_journal_set_watermark(j);
out:
@ -818,7 +818,7 @@ static int journal_flush_done(struct journal *j, u64 seq_to_flush,
* If journal replay hasn't completed, the unreplayed journal entries
* hold refs on their corresponding sequence numbers
*/
ret = !test_bit(JOURNAL_REPLAY_DONE, &j->flags) ||
ret = !test_bit(JOURNAL_replay_done, &j->flags) ||
journal_last_seq(j) > seq_to_flush ||
!fifo_used(&j->pin);
@ -833,7 +833,7 @@ bool bch2_journal_flush_pins(struct journal *j, u64 seq_to_flush)
/* time_stats this */
bool did_work = false;
if (!test_bit(JOURNAL_STARTED, &j->flags))
if (!test_bit(JOURNAL_running, &j->flags))
return false;
closure_wait_event(&j->async_wait,

10
fs/bcachefs/journal_sb.c
View File

@ -16,9 +16,8 @@ static int u64_cmp(const void *_l, const void *_r)
return cmp_int(*l, *r);
}
static int bch2_sb_journal_validate(struct bch_sb *sb,
struct bch_sb_field *f,
struct printbuf *err)
static int bch2_sb_journal_validate(struct bch_sb *sb, struct bch_sb_field *f,
enum bch_validate_flags flags, struct printbuf *err)
{
struct bch_sb_field_journal *journal = field_to_type(f, journal);
struct bch_member m = bch2_sb_member_get(sb, sb->dev_idx);
@ -99,9 +98,8 @@ static int u64_range_cmp(const void *_l, const void *_r)
return cmp_int(l->start, r->start);
}
static int bch2_sb_journal_v2_validate(struct bch_sb *sb,
struct bch_sb_field *f,
struct printbuf *err)
static int bch2_sb_journal_v2_validate(struct bch_sb *sb, struct bch_sb_field *f,
enum bch_validate_flags flags, struct printbuf *err)
{
struct bch_sb_field_journal_v2 *journal = field_to_type(f, journal_v2);
struct bch_member m = bch2_sb_member_get(sb, sb->dev_idx);

77
fs/bcachefs/journal_seq_blacklist.c
View File

@ -1,8 +1,8 @@
// SPDX-License-Identifier: GPL-2.0
#include "bcachefs.h"
#include "btree_iter.h"
#include "eytzinger.h"
#include "journal.h"
#include "journal_seq_blacklist.h"
#include "super-io.h"
@ -162,9 +162,8 @@ int bch2_blacklist_table_initialize(struct bch_fs *c)
return 0;
}
static int bch2_sb_journal_seq_blacklist_validate(struct bch_sb *sb,
struct bch_sb_field *f,
struct printbuf *err)
static int bch2_sb_journal_seq_blacklist_validate(struct bch_sb *sb, struct bch_sb_field *f,
enum bch_validate_flags flags, struct printbuf *err)
{
struct bch_sb_field_journal_seq_blacklist *bl =
field_to_type(f, journal_seq_blacklist);
@ -217,78 +216,40 @@ const struct bch_sb_field_ops bch_sb_field_ops_journal_seq_blacklist = {
.to_text = bch2_sb_journal_seq_blacklist_to_text
};
void bch2_blacklist_entries_gc(struct work_struct *work)
bool bch2_blacklist_entries_gc(struct bch_fs *c)
{
struct bch_fs *c = container_of(work, struct bch_fs,
journal_seq_blacklist_gc_work);
struct journal_seq_blacklist_table *t;
struct bch_sb_field_journal_seq_blacklist *bl;
struct journal_seq_blacklist_entry *src, *dst;
struct btree_trans *trans = bch2_trans_get(c);
unsigned i, nr, new_nr;
int ret;
for (i = 0; i < BTREE_ID_NR; i++) {
struct btree_iter iter;
struct btree *b;
bch2_trans_node_iter_init(trans, &iter, i, POS_MIN,
0, 0, BTREE_ITER_PREFETCH);
retry:
bch2_trans_begin(trans);
b = bch2_btree_iter_peek_node(&iter);
while (!(ret = PTR_ERR_OR_ZERO(b)) &&
b &&
!test_bit(BCH_FS_stopping, &c->flags))
b = bch2_btree_iter_next_node(&iter);
if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
goto retry;
bch2_trans_iter_exit(trans, &iter);
}
bch2_trans_put(trans);
if (ret)
return;
mutex_lock(&c->sb_lock);
bl = bch2_sb_field_get(c->disk_sb.sb, journal_seq_blacklist);
struct bch_sb_field_journal_seq_blacklist *bl =
bch2_sb_field_get(c->disk_sb.sb, journal_seq_blacklist);
if (!bl)
goto out;
return false;
nr = blacklist_nr_entries(bl);
unsigned nr = blacklist_nr_entries(bl);
dst = bl->start;
t = c->journal_seq_blacklist_table;
struct journal_seq_blacklist_table *t = c->journal_seq_blacklist_table;
BUG_ON(nr != t->nr);
unsigned i;
for (src = bl->start, i = eytzinger0_first(t->nr);
src < bl->start + nr;
src++, i = eytzinger0_next(i, nr)) {
BUG_ON(t->entries[i].start != le64_to_cpu(src->start));
BUG_ON(t->entries[i].end != le64_to_cpu(src->end));
if (t->entries[i].dirty)
if (t->entries[i].dirty || t->entries[i].end >= c->journal.oldest_seq_found_ondisk)
*dst++ = *src;
}
new_nr = dst - bl->start;
unsigned new_nr = dst - bl->start;
if (new_nr == nr)
return false;
bch_info(c, "nr blacklist entries was %u, now %u", nr, new_nr);
bch_verbose(c, "nr blacklist entries was %u, now %u", nr, new_nr);
if (new_nr != nr) {
bl = bch2_sb_field_resize(&c->disk_sb, journal_seq_blacklist,
new_nr ? sb_blacklist_u64s(new_nr) : 0);
BUG_ON(new_nr && !bl);
if (!new_nr)
c->disk_sb.sb->features[0] &= cpu_to_le64(~(1ULL << BCH_FEATURE_journal_seq_blacklist_v3));
bch2_write_super(c);
}
out:
mutex_unlock(&c->sb_lock);
bl = bch2_sb_field_resize(&c->disk_sb, journal_seq_blacklist,
new_nr ? sb_blacklist_u64s(new_nr) : 0);
BUG_ON(new_nr && !bl);
return true;
}

2
fs/bcachefs/journal_seq_blacklist.h
View File

@ -17,6 +17,6 @@ int bch2_blacklist_table_initialize(struct bch_fs *);
extern const struct bch_sb_field_ops bch_sb_field_ops_journal_seq_blacklist;
void bch2_blacklist_entries_gc(struct work_struct *);
bool bch2_blacklist_entries_gc(struct bch_fs *);
#endif /* _BCACHEFS_JOURNAL_SEQ_BLACKLIST_H */

17
fs/bcachefs/journal_types.h
View File

@ -129,12 +129,17 @@ enum journal_space_from {
journal_space_nr,
};
#define JOURNAL_FLAGS() \
x(replay_done) \
x(running) \
x(may_skip_flush) \
x(need_flush_write) \
x(space_low)
enum journal_flags {
JOURNAL_REPLAY_DONE,
JOURNAL_STARTED,
JOURNAL_MAY_SKIP_FLUSH,
JOURNAL_NEED_FLUSH_WRITE,
JOURNAL_SPACE_LOW,
#define x(n) JOURNAL_##n,
JOURNAL_FLAGS()
#undef x
};
/* Reasons we may fail to get a journal reservation: */
@ -229,6 +234,7 @@ struct journal {
u64 last_seq_ondisk;
u64 err_seq;
u64 last_empty_seq;
u64 oldest_seq_found_ondisk;
/*
* FIFO of journal entries whose btree updates have not yet been
@ -326,6 +332,7 @@ struct journal_device {
/* for bch_journal_read_device */
struct closure read;
u64 highest_seq_found;
};
/*

2
fs/bcachefs/logged_ops.c
View File

@ -56,7 +56,7 @@ int bch2_resume_logged_ops(struct bch_fs *c)
int ret = bch2_trans_run(c,
for_each_btree_key(trans, iter,
BTREE_ID_logged_ops, POS_MIN,
BTREE_ITER_PREFETCH, k,
BTREE_ITER_prefetch, k,
resume_logged_op(trans, &iter, k)));
bch_err_fn(c, ret);
return ret;

4
fs/bcachefs/lru.c
View File

@ -11,7 +11,7 @@
/* KEY_TYPE_lru is obsolete: */
int bch2_lru_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
int ret = 0;
@ -149,7 +149,7 @@ int bch2_check_lrus(struct bch_fs *c)
struct bpos last_flushed_pos = POS_MIN;
int ret = bch2_trans_run(c,
for_each_btree_key_commit(trans, iter,
BTREE_ID_lru, POS_MIN, BTREE_ITER_PREFETCH, k,
BTREE_ID_lru, POS_MIN, BTREE_ITER_prefetch, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc|BCH_TRANS_COMMIT_lazy_rw,
bch2_check_lru_key(trans, &iter, k, &last_flushed_pos)));
bch_err_fn(c, ret);

2
fs/bcachefs/lru.h
View File

@ -49,7 +49,7 @@ static inline enum bch_lru_type lru_type(struct bkey_s_c l)
}
int bch2_lru_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_lru_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
void bch2_lru_pos_to_text(struct printbuf *, struct bpos);

8
fs/bcachefs/migrate.c
View File

@ -49,7 +49,7 @@ static int bch2_dev_usrdata_drop_key(struct btree_trans *trans,
if (!bch2_bkey_has_device_c(k, dev_idx))
return 0;
n = bch2_bkey_make_mut(trans, iter, &k, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
n = bch2_bkey_make_mut(trans, iter, &k, BTREE_UPDATE_internal_snapshot_node);
ret = PTR_ERR_OR_ZERO(n);
if (ret)
return ret;
@ -67,7 +67,7 @@ static int bch2_dev_usrdata_drop_key(struct btree_trans *trans,
/*
* Since we're not inserting through an extent iterator
* (BTREE_ITER_ALL_SNAPSHOTS iterators aren't extent iterators),
* (BTREE_ITER_all_snapshots iterators aren't extent iterators),
* we aren't using the extent overwrite path to delete, we're
* just using the normal key deletion path:
*/
@ -87,7 +87,7 @@ static int bch2_dev_usrdata_drop(struct bch_fs *c, unsigned dev_idx, int flags)
continue;
ret = for_each_btree_key_commit(trans, iter, id, POS_MIN,
BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
NULL, NULL, BCH_TRANS_COMMIT_no_enospc,
bch2_dev_usrdata_drop_key(trans, &iter, k, dev_idx, flags));
if (ret)
@ -119,7 +119,7 @@ static int bch2_dev_metadata_drop(struct bch_fs *c, unsigned dev_idx, int flags)
for (id = 0; id < BTREE_ID_NR; id++) {
bch2_trans_node_iter_init(trans, &iter, id, POS_MIN, 0, 0,
BTREE_ITER_PREFETCH);
BTREE_ITER_prefetch);
retry:
ret = 0;
while (bch2_trans_begin(trans),

82
fs/bcachefs/move.c
View File

@ -41,28 +41,23 @@ static void bch2_data_update_opts_to_text(struct printbuf *out, struct bch_fs *c
struct data_update_opts *data_opts)
{
printbuf_tabstop_push(out, 20);
prt_str(out, "rewrite ptrs:");
prt_tab(out);
prt_str(out, "rewrite ptrs:\t");
bch2_prt_u64_base2(out, data_opts->rewrite_ptrs);
prt_newline(out);
prt_str(out, "kill ptrs: ");
prt_tab(out);
prt_str(out, "kill ptrs:\t");
bch2_prt_u64_base2(out, data_opts->kill_ptrs);
prt_newline(out);
prt_str(out, "target: ");
prt_tab(out);
prt_str(out, "target:\t");
bch2_target_to_text(out, c, data_opts->target);
prt_newline(out);
prt_str(out, "compression: ");
prt_tab(out);
prt_str(out, "compression:\t");
bch2_compression_opt_to_text(out, background_compression(*io_opts));
prt_newline(out);
prt_str(out, "extra replicas: ");
prt_tab(out);
prt_str(out, "extra replicas:\t");
prt_u64(out, data_opts->extra_replicas);
}
@ -421,7 +416,7 @@ struct bch_io_opts *bch2_move_get_io_opts(struct btree_trans *trans,
io_opts->d.nr = 0;
ret = for_each_btree_key(trans, iter, BTREE_ID_inodes, POS(0, extent_k.k->p.inode),
BTREE_ITER_ALL_SNAPSHOTS, k, ({
BTREE_ITER_all_snapshots, k, ({
if (k.k->p.offset != extent_k.k->p.inode)
break;
@ -467,7 +462,7 @@ int bch2_move_get_io_opts_one(struct btree_trans *trans,
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
SPOS(0, extent_k.k->p.inode, extent_k.k->p.snapshot),
BTREE_ITER_CACHED);
BTREE_ITER_cached);
ret = bkey_err(k);
if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
return ret;
@ -553,8 +548,8 @@ static int bch2_move_data_btree(struct moving_context *ctxt,
}
bch2_trans_iter_init(trans, &iter, btree_id, start,
BTREE_ITER_PREFETCH|
BTREE_ITER_ALL_SNAPSHOTS);
BTREE_ITER_prefetch|
BTREE_ITER_all_snapshots);
if (ctxt->rate)
bch2_ratelimit_reset(ctxt->rate);
@ -695,6 +690,10 @@ int bch2_evacuate_bucket(struct moving_context *ctxt,
struct bpos bp_pos = POS_MIN;
int ret = 0;
struct bch_dev *ca = bch2_dev_tryget(c, bucket.inode);
if (!ca)
return 0;
trace_bucket_evacuate(c, &bucket);
bch2_bkey_buf_init(&sk);
@ -705,7 +704,7 @@ int bch2_evacuate_bucket(struct moving_context *ctxt,
bch2_trans_begin(trans);
bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc,
bucket, BTREE_ITER_CACHED);
bucket, BTREE_ITER_cached);
ret = lockrestart_do(trans,
bkey_err(k = bch2_btree_iter_peek_slot(&iter)));
bch2_trans_iter_exit(trans, &iter);
@ -716,7 +715,7 @@ int bch2_evacuate_bucket(struct moving_context *ctxt,
a = bch2_alloc_to_v4(k, &a_convert);
dirty_sectors = bch2_bucket_sectors_dirty(*a);
bucket_size = bch_dev_bkey_exists(c, bucket.inode)->mi.bucket_size;
bucket_size = ca->mi.bucket_size;
fragmentation = a->fragmentation_lru;
ret = bch2_btree_write_buffer_tryflush(trans);
@ -730,9 +729,9 @@ int bch2_evacuate_bucket(struct moving_context *ctxt,
bch2_trans_begin(trans);
ret = bch2_get_next_backpointer(trans, bucket, gen,
ret = bch2_get_next_backpointer(trans, ca, bucket, gen,
&bp_pos, &bp,
BTREE_ITER_CACHED);
BTREE_ITER_cached);
if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
continue;
if (ret)
@ -828,6 +827,7 @@ next:
trace_evacuate_bucket(c, &bucket, dirty_sectors, bucket_size, fragmentation, ret);
err:
bch2_dev_put(ca);
bch2_bkey_buf_exit(&sk, c);
return ret;
}
@ -868,7 +868,7 @@ static int bch2_move_btree(struct bch_fs *c,
continue;
bch2_trans_node_iter_init(trans, &iter, btree, POS_MIN, 0, 0,
BTREE_ITER_PREFETCH);
BTREE_ITER_prefetch);
retry:
ret = 0;
while (bch2_trans_begin(trans),
@ -975,26 +975,10 @@ static bool migrate_btree_pred(struct bch_fs *c, void *arg,
*/
static bool bformat_needs_redo(struct bkey_format *f)
{
for (unsigned i = 0; i < f->nr_fields; i++) {
unsigned f_bits = f->bits_per_field[i];
unsigned unpacked_bits = bch2_bkey_format_current.bits_per_field[i];
u64 unpacked_mask = ~((~0ULL << 1) << (unpacked_bits - 1));
u64 field_offset = le64_to_cpu(f->field_offset[i]);
if (f_bits > unpacked_bits)
for (unsigned i = 0; i < f->nr_fields; i++)
if (bch2_bkey_format_field_overflows(f, i))
return true;
if ((f_bits == unpacked_bits) && field_offset)
return true;
u64 f_mask = f_bits
? ~((~0ULL << (f_bits - 1)) << 1)
: 0;
if (((field_offset + f_mask) & unpacked_mask) < field_offset)
return true;
}
return false;
}
@ -1049,6 +1033,7 @@ static bool drop_extra_replicas_pred(struct bch_fs *c, void *arg,
struct extent_ptr_decoded p;
unsigned i = 0;
rcu_read_lock();
bkey_for_each_ptr_decode(k.k, bch2_bkey_ptrs_c(k), p, entry) {
unsigned d = bch2_extent_ptr_durability(c, &p);
@ -1059,6 +1044,7 @@ static bool drop_extra_replicas_pred(struct bch_fs *c, void *arg,
i++;
}
rcu_read_unlock();
return data_opts->kill_ptrs != 0;
}
@ -1143,23 +1129,17 @@ void bch2_move_stats_to_text(struct printbuf *out, struct bch_move_stats *stats)
prt_newline(out);
printbuf_indent_add(out, 2);
prt_str(out, "keys moved: ");
prt_u64(out, atomic64_read(&stats->keys_moved));
prt_newline(out);
prt_str(out, "keys raced: ");
prt_u64(out, atomic64_read(&stats->keys_raced));
prt_newline(out);
prt_str(out, "bytes seen: ");
prt_printf(out, "keys moved: %llu\n", atomic64_read(&stats->keys_moved));
prt_printf(out, "keys raced: %llu\n", atomic64_read(&stats->keys_raced));
prt_printf(out, "bytes seen: ");
prt_human_readable_u64(out, atomic64_read(&stats->sectors_seen) << 9);
prt_newline(out);
prt_str(out, "bytes moved: ");
prt_printf(out, "bytes moved: ");
prt_human_readable_u64(out, atomic64_read(&stats->sectors_moved) << 9);
prt_newline(out);
prt_str(out, "bytes raced: ");
prt_printf(out, "bytes raced: ");
prt_human_readable_u64(out, atomic64_read(&stats->sectors_raced) << 9);
prt_newline(out);
@ -1173,19 +1153,17 @@ static void bch2_moving_ctxt_to_text(struct printbuf *out, struct bch_fs *c, str
bch2_move_stats_to_text(out, ctxt->stats);
printbuf_indent_add(out, 2);
prt_printf(out, "reads: ios %u/%u sectors %u/%u",
prt_printf(out, "reads: ios %u/%u sectors %u/%u\n",
atomic_read(&ctxt->read_ios),
c->opts.move_ios_in_flight,
atomic_read(&ctxt->read_sectors),
c->opts.move_bytes_in_flight >> 9);
prt_newline(out);
prt_printf(out, "writes: ios %u/%u sectors %u/%u",
prt_printf(out, "writes: ios %u/%u sectors %u/%u\n",
atomic_read(&ctxt->write_ios),
c->opts.move_ios_in_flight,
atomic_read(&ctxt->write_sectors),
c->opts.move_bytes_in_flight >> 9);
prt_newline(out);
printbuf_indent_add(out, 2);

4
fs/bcachefs/movinggc.c
View File

@ -84,7 +84,7 @@ static int bch2_bucket_is_movable(struct btree_trans *trans,
return 0;
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_alloc,
b->k.bucket, BTREE_ITER_CACHED);
b->k.bucket, BTREE_ITER_cached);
ret = bkey_err(k);
if (ret)
return ret;
@ -158,6 +158,8 @@ static int bch2_copygc_get_buckets(struct moving_context *ctxt,
if (bch2_fs_fatal_err_on(ret, c, "%s: from bch2_btree_write_buffer_tryflush()", bch2_err_str(ret)))
return ret;
bch2_trans_begin(trans);
ret = for_each_btree_key_upto(trans, iter, BTREE_ID_lru,
lru_pos(BCH_LRU_FRAGMENTATION_START, 0, 0),
lru_pos(BCH_LRU_FRAGMENTATION_START, U64_MAX, LRU_TIME_MAX),

7
fs/bcachefs/opts.h
View File

@ -426,11 +426,6 @@ enum fsck_err_opts {
BCH_SB_VERSION_UPGRADE, BCH_VERSION_UPGRADE_compatible, \
NULL, "Set superblock to latest version,\n" \
"allowing any new features to be used") \
x(buckets_nouse, u8, \
0, \
OPT_BOOL(), \
BCH2_NO_SB_OPT, false, \
NULL, "Allocate the buckets_nouse bitmap") \
x(stdio, u64, \
0, \
OPT_UINT(0, S64_MAX), \
@ -480,7 +475,7 @@ enum fsck_err_opts {
OPT_FS|OPT_MOUNT|OPT_RUNTIME, \
OPT_BOOL(), \
BCH2_NO_SB_OPT, true, \
NULL, "BTREE_ITER_PREFETCH casuse btree nodes to be\n"\
NULL, "BTREE_ITER_prefetch casuse btree nodes to be\n"\
" prefetched sequentially")
struct bch_opts {

232
fs/bcachefs/printbuf.c
View File

@ -10,35 +10,50 @@
#include "printbuf.h"
static inline unsigned __printbuf_linelen(struct printbuf *buf, unsigned pos)
{
return pos - buf->last_newline;
}
static inline unsigned printbuf_linelen(struct printbuf *buf)
{
return buf->pos - buf->last_newline;
return __printbuf_linelen(buf, buf->pos);
}
/*
* Returns spaces from start of line, if set, or 0 if unset:
*/
static inline unsigned cur_tabstop(struct printbuf *buf)
{
return buf->cur_tabstop < buf->nr_tabstops
? buf->_tabstops[buf->cur_tabstop]
: 0;
}
int bch2_printbuf_make_room(struct printbuf *out, unsigned extra)
{
unsigned new_size;
char *buf;
if (!out->heap_allocated)
return 0;
/* Reserved space for terminating nul: */
extra += 1;
if (out->pos + extra < out->size)
if (out->pos + extra <= out->size)
return 0;
new_size = roundup_pow_of_two(out->size + extra);
if (!out->heap_allocated) {
out->overflow = true;
return 0;
}
unsigned new_size = roundup_pow_of_two(out->size + extra);
/*
* Note: output buffer must be freeable with kfree(), it's not required
* that the user use printbuf_exit().
*/
buf = krealloc(out->buf, new_size, !out->atomic ? GFP_KERNEL : GFP_NOWAIT);
char *buf = krealloc(out->buf, new_size, !out->atomic ? GFP_KERNEL : GFP_NOWAIT);
if (!buf) {
out->allocation_failure = true;
out->overflow = true;
return -ENOMEM;
}
@ -47,6 +62,92 @@ int bch2_printbuf_make_room(struct printbuf *out, unsigned extra)
return 0;
}
static void printbuf_advance_pos(struct printbuf *out, unsigned len)
{
out->pos += min(len, printbuf_remaining(out));
}
static void printbuf_insert_spaces(struct printbuf *out, unsigned pos, unsigned nr)
{
unsigned move = out->pos - pos;
bch2_printbuf_make_room(out, nr);
if (pos + nr < out->size)
memmove(out->buf + pos + nr,
out->buf + pos,
min(move, out->size - 1 - pos - nr));
if (pos < out->size)
memset(out->buf + pos, ' ', min(nr, out->size - pos));
printbuf_advance_pos(out, nr);
printbuf_nul_terminate_reserved(out);
}
static void __printbuf_do_indent(struct printbuf *out, unsigned pos)
{
while (true) {
int pad;
unsigned len = out->pos - pos;
char *p = out->buf + pos;
char *n = memscan(p, '\n', len);
if (cur_tabstop(out)) {
n = min(n, (char *) memscan(p, '\r', len));
n = min(n, (char *) memscan(p, '\t', len));
}
pos = n - out->buf;
if (pos == out->pos)
break;
switch (*n) {
case '\n':
pos++;
out->last_newline = pos;
printbuf_insert_spaces(out, pos, out->indent);
pos = min(pos + out->indent, out->pos);
out->last_field = pos;
out->cur_tabstop = 0;
break;
case '\r':
memmove(n, n + 1, out->pos - pos);
--out->pos;
pad = (int) cur_tabstop(out) - (int) __printbuf_linelen(out, pos);
if (pad > 0) {
printbuf_insert_spaces(out, out->last_field, pad);
pos += pad;
}
out->last_field = pos;
out->cur_tabstop++;
break;
case '\t':
pad = (int) cur_tabstop(out) - (int) __printbuf_linelen(out, pos) - 1;
if (pad > 0) {
*n = ' ';
printbuf_insert_spaces(out, pos, pad - 1);
pos += pad;
} else {
memmove(n, n + 1, out->pos - pos);
--out->pos;
}
out->last_field = pos;
out->cur_tabstop++;
break;
}
}
}
static inline void printbuf_do_indent(struct printbuf *out, unsigned pos)
{
if (out->has_indent_or_tabstops && !out->suppress_indent_tabstop_handling)
__printbuf_do_indent(out, pos);
}
void bch2_prt_vprintf(struct printbuf *out, const char *fmt, va_list args)
{
int len;
@ -55,14 +156,14 @@ void bch2_prt_vprintf(struct printbuf *out, const char *fmt, va_list args)
va_list args2;
va_copy(args2, args);
len = vsnprintf(out->buf + out->pos, printbuf_remaining(out), fmt, args2);
len = vsnprintf(out->buf + out->pos, printbuf_remaining_size(out), fmt, args2);
va_end(args2);
} while (len + 1 >= printbuf_remaining(out) &&
!bch2_printbuf_make_room(out, len + 1));
} while (len > printbuf_remaining(out) &&
!bch2_printbuf_make_room(out, len));
len = min_t(size_t, len,
printbuf_remaining(out) ? printbuf_remaining(out) - 1 : 0);
out->pos += len;
unsigned indent_pos = out->pos;
printbuf_advance_pos(out, len);
printbuf_do_indent(out, indent_pos);
}
void bch2_prt_printf(struct printbuf *out, const char *fmt, ...)
@ -72,14 +173,14 @@ void bch2_prt_printf(struct printbuf *out, const char *fmt, ...)
do {
va_start(args, fmt);
len = vsnprintf(out->buf + out->pos, printbuf_remaining(out), fmt, args);
len = vsnprintf(out->buf + out->pos, printbuf_remaining_size(out), fmt, args);
va_end(args);
} while (len + 1 >= printbuf_remaining(out) &&
!bch2_printbuf_make_room(out, len + 1));
} while (len > printbuf_remaining(out) &&
!bch2_printbuf_make_room(out, len));
len = min_t(size_t, len,
printbuf_remaining(out) ? printbuf_remaining(out) - 1 : 0);
out->pos += len;
unsigned indent_pos = out->pos;
printbuf_advance_pos(out, len);
printbuf_do_indent(out, indent_pos);
}
/**
@ -194,33 +295,20 @@ void bch2_printbuf_indent_sub(struct printbuf *buf, unsigned spaces)
void bch2_prt_newline(struct printbuf *buf)
{
unsigned i;
bch2_printbuf_make_room(buf, 1 + buf->indent);
__prt_char(buf, '\n');
__prt_char_reserved(buf, '\n');
buf->last_newline = buf->pos;
for (i = 0; i < buf->indent; i++)
__prt_char(buf, ' ');
__prt_chars_reserved(buf, ' ', buf->indent);
printbuf_nul_terminate(buf);
printbuf_nul_terminate_reserved(buf);
buf->last_field = buf->pos;
buf->cur_tabstop = 0;
}
/*
* Returns spaces from start of line, if set, or 0 if unset:
*/
static inline unsigned cur_tabstop(struct printbuf *buf)
{
return buf->cur_tabstop < buf->nr_tabstops
? buf->_tabstops[buf->cur_tabstop]
: 0;
}
static void __prt_tab(struct printbuf *out)
{
int spaces = max_t(int, 0, cur_tabstop(out) - printbuf_linelen(out));
@ -247,24 +335,9 @@ void bch2_prt_tab(struct printbuf *out)
static void __prt_tab_rjust(struct printbuf *buf)
{
unsigned move = buf->pos - buf->last_field;
int pad = (int) cur_tabstop(buf) - (int) printbuf_linelen(buf);
if (pad > 0) {
bch2_printbuf_make_room(buf, pad);
if (buf->last_field + pad < buf->size)
memmove(buf->buf + buf->last_field + pad,
buf->buf + buf->last_field,
min(move, buf->size - 1 - buf->last_field - pad));
if (buf->last_field < buf->size)
memset(buf->buf + buf->last_field, ' ',
min((unsigned) pad, buf->size - buf->last_field));
buf->pos += pad;
printbuf_nul_terminate(buf);
}
if (pad > 0)
printbuf_insert_spaces(buf, buf->last_field, pad);
buf->last_field = buf->pos;
buf->cur_tabstop++;
@ -301,41 +374,9 @@ void bch2_prt_tab_rjust(struct printbuf *buf)
*/
void bch2_prt_bytes_indented(struct printbuf *out, const char *str, unsigned count)
{
const char *unprinted_start = str;
const char *end = str + count;
if (!out->has_indent_or_tabstops || out->suppress_indent_tabstop_handling) {
prt_bytes(out, str, count);
return;
}
while (str != end) {
switch (*str) {
case '\n':
prt_bytes(out, unprinted_start, str - unprinted_start);
unprinted_start = str + 1;
bch2_prt_newline(out);
break;
case '\t':
if (likely(cur_tabstop(out))) {
prt_bytes(out, unprinted_start, str - unprinted_start);
unprinted_start = str + 1;
__prt_tab(out);
}
break;
case '\r':
if (likely(cur_tabstop(out))) {
prt_bytes(out, unprinted_start, str - unprinted_start);
unprinted_start = str + 1;
__prt_tab_rjust(out);
}
break;
}
str++;
}
prt_bytes(out, unprinted_start, str - unprinted_start);
unsigned indent_pos = out->pos;
prt_bytes(out, str, count);
printbuf_do_indent(out, indent_pos);
}
/**
@ -348,9 +389,10 @@ void bch2_prt_bytes_indented(struct printbuf *out, const char *str, unsigned cou
void bch2_prt_human_readable_u64(struct printbuf *out, u64 v)
{
bch2_printbuf_make_room(out, 10);
out->pos += string_get_size(v, 1, !out->si_units,
out->buf + out->pos,
printbuf_remaining_size(out));
unsigned len = string_get_size(v, 1, !out->si_units,
out->buf + out->pos,
printbuf_remaining_size(out));
printbuf_advance_pos(out, len);
}
/**
@ -402,9 +444,7 @@ void bch2_prt_string_option(struct printbuf *out,
const char * const list[],
size_t selected)
{
size_t i;
for (i = 0; list[i]; i++)
for (size_t i = 0; list[i]; i++)
bch2_prt_printf(out, i == selected ? "[%s] " : "%s ", list[i]);
}

53
fs/bcachefs/printbuf.h
View File

@ -86,6 +86,7 @@ struct printbuf {
u8 atomic;
bool allocation_failure:1;
bool heap_allocated:1;
bool overflow:1;
enum printbuf_si si_units:1;
bool human_readable_units:1;
bool has_indent_or_tabstops:1;
@ -142,7 +143,9 @@ void bch2_prt_bitflags_vector(struct printbuf *, const char * const[],
*/
static inline unsigned printbuf_remaining_size(struct printbuf *out)
{
return out->pos < out->size ? out->size - out->pos : 0;
if (WARN_ON(out->size && out->pos >= out->size))
out->pos = out->size - 1;
return out->size - out->pos;
}
/*
@ -151,7 +154,7 @@ static inline unsigned printbuf_remaining_size(struct printbuf *out)
*/
static inline unsigned printbuf_remaining(struct printbuf *out)
{
return out->pos < out->size ? out->size - out->pos - 1 : 0;
return out->size ? printbuf_remaining_size(out) - 1 : 0;
}
static inline unsigned printbuf_written(struct printbuf *out)
@ -159,30 +162,25 @@ static inline unsigned printbuf_written(struct printbuf *out)
return out->size ? min(out->pos, out->size - 1) : 0;
}
/*
* Returns true if output was truncated:
*/
static inline bool printbuf_overflowed(struct printbuf *out)
static inline void printbuf_nul_terminate_reserved(struct printbuf *out)
{
return out->pos >= out->size;
if (WARN_ON(out->size && out->pos >= out->size))
out->pos = out->size - 1;
if (out->size)
out->buf[out->pos] = 0;
}
static inline void printbuf_nul_terminate(struct printbuf *out)
{
bch2_printbuf_make_room(out, 1);
if (out->pos < out->size)
out->buf[out->pos] = 0;
else if (out->size)
out->buf[out->size - 1] = 0;
printbuf_nul_terminate_reserved(out);
}
/* Doesn't call bch2_printbuf_make_room(), doesn't nul terminate: */
static inline void __prt_char_reserved(struct printbuf *out, char c)
{
if (printbuf_remaining(out))
out->buf[out->pos] = c;
out->pos++;
out->buf[out->pos++] = c;
}
/* Doesn't nul terminate: */
@ -194,37 +192,34 @@ static inline void __prt_char(struct printbuf *out, char c)
static inline void prt_char(struct printbuf *out, char c)
{
__prt_char(out, c);
printbuf_nul_terminate(out);
bch2_printbuf_make_room(out, 2);
__prt_char_reserved(out, c);
printbuf_nul_terminate_reserved(out);
}
static inline void __prt_chars_reserved(struct printbuf *out, char c, unsigned n)
{
unsigned i, can_print = min(n, printbuf_remaining(out));
unsigned can_print = min(n, printbuf_remaining(out));
for (i = 0; i < can_print; i++)
for (unsigned i = 0; i < can_print; i++)
out->buf[out->pos++] = c;
out->pos += n - can_print;
}
static inline void prt_chars(struct printbuf *out, char c, unsigned n)
{
bch2_printbuf_make_room(out, n);
__prt_chars_reserved(out, c, n);
printbuf_nul_terminate(out);
printbuf_nul_terminate_reserved(out);
}
static inline void prt_bytes(struct printbuf *out, const void *b, unsigned n)
{
unsigned i, can_print;
bch2_printbuf_make_room(out, n);
can_print = min(n, printbuf_remaining(out));
unsigned can_print = min(n, printbuf_remaining(out));
for (i = 0; i < can_print; i++)
for (unsigned i = 0; i < can_print; i++)
out->buf[out->pos++] = ((char *) b)[i];
out->pos += n - can_print;
printbuf_nul_terminate(out);
}
@ -241,18 +236,18 @@ static inline void prt_str_indented(struct printbuf *out, const char *str)
static inline void prt_hex_byte(struct printbuf *out, u8 byte)
{
bch2_printbuf_make_room(out, 2);
bch2_printbuf_make_room(out, 3);
__prt_char_reserved(out, hex_asc_hi(byte));
__prt_char_reserved(out, hex_asc_lo(byte));
printbuf_nul_terminate(out);
printbuf_nul_terminate_reserved(out);
}
static inline void prt_hex_byte_upper(struct printbuf *out, u8 byte)
{
bch2_printbuf_make_room(out, 2);
bch2_printbuf_make_room(out, 3);
__prt_char_reserved(out, hex_asc_upper_hi(byte));
__prt_char_reserved(out, hex_asc_upper_lo(byte));
printbuf_nul_terminate(out);
printbuf_nul_terminate_reserved(out);
}
/**

123
fs/bcachefs/quota.c
View File

@ -20,7 +20,7 @@ static const char * const bch2_quota_counters[] = {
};
static int bch2_sb_quota_validate(struct bch_sb *sb, struct bch_sb_field *f,
struct printbuf *err)
enum bch_validate_flags flags, struct printbuf *err)
{
struct bch_sb_field_quota *q = field_to_type(f, quota);
@ -60,8 +60,7 @@ const struct bch_sb_field_ops bch_sb_field_ops_quota = {
};
int bch2_quota_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
struct printbuf *err)
enum bch_validate_flags flags, struct printbuf *err)
{
int ret = 0;
@ -97,45 +96,14 @@ static void qc_info_to_text(struct printbuf *out, struct qc_info *i)
printbuf_tabstops_reset(out);
printbuf_tabstop_push(out, 20);
prt_str(out, "i_fieldmask");
prt_tab(out);
prt_printf(out, "%x", i->i_fieldmask);
prt_newline(out);
prt_str(out, "i_flags");
prt_tab(out);
prt_printf(out, "%u", i->i_flags);
prt_newline(out);
prt_str(out, "i_spc_timelimit");
prt_tab(out);
prt_printf(out, "%u", i->i_spc_timelimit);
prt_newline(out);
prt_str(out, "i_ino_timelimit");
prt_tab(out);
prt_printf(out, "%u", i->i_ino_timelimit);
prt_newline(out);
prt_str(out, "i_rt_spc_timelimit");
prt_tab(out);
prt_printf(out, "%u", i->i_rt_spc_timelimit);
prt_newline(out);
prt_str(out, "i_spc_warnlimit");
prt_tab(out);
prt_printf(out, "%u", i->i_spc_warnlimit);
prt_newline(out);
prt_str(out, "i_ino_warnlimit");
prt_tab(out);
prt_printf(out, "%u", i->i_ino_warnlimit);
prt_newline(out);
prt_str(out, "i_rt_spc_warnlimit");
prt_tab(out);
prt_printf(out, "%u", i->i_rt_spc_warnlimit);
prt_newline(out);
prt_printf(out, "i_fieldmask\t%x\n", i->i_fieldmask);
prt_printf(out, "i_flags\t%u\n", i->i_flags);
prt_printf(out, "i_spc_timelimit\t%u\n", i->i_spc_timelimit);
prt_printf(out, "i_ino_timelimit\t%u\n", i->i_ino_timelimit);
prt_printf(out, "i_rt_spc_timelimit\t%u\n", i->i_rt_spc_timelimit);
prt_printf(out, "i_spc_warnlimit\t%u\n", i->i_spc_warnlimit);
prt_printf(out, "i_ino_warnlimit\t%u\n", i->i_ino_warnlimit);
prt_printf(out, "i_rt_spc_warnlimit\t%u\n", i->i_rt_spc_warnlimit);
}
static void qc_dqblk_to_text(struct printbuf *out, struct qc_dqblk *q)
@ -143,60 +111,17 @@ static void qc_dqblk_to_text(struct printbuf *out, struct qc_dqblk *q)
printbuf_tabstops_reset(out);
printbuf_tabstop_push(out, 20);
prt_str(out, "d_fieldmask");
prt_tab(out);
prt_printf(out, "%x", q->d_fieldmask);
prt_newline(out);
prt_str(out, "d_spc_hardlimit");
prt_tab(out);
prt_printf(out, "%llu", q->d_spc_hardlimit);
prt_newline(out);
prt_str(out, "d_spc_softlimit");
prt_tab(out);
prt_printf(out, "%llu", q->d_spc_softlimit);
prt_newline(out);
prt_str(out, "d_ino_hardlimit");
prt_tab(out);
prt_printf(out, "%llu", q->d_ino_hardlimit);
prt_newline(out);
prt_str(out, "d_ino_softlimit");
prt_tab(out);
prt_printf(out, "%llu", q->d_ino_softlimit);
prt_newline(out);
prt_str(out, "d_space");
prt_tab(out);
prt_printf(out, "%llu", q->d_space);
prt_newline(out);
prt_str(out, "d_ino_count");
prt_tab(out);
prt_printf(out, "%llu", q->d_ino_count);
prt_newline(out);
prt_str(out, "d_ino_timer");
prt_tab(out);
prt_printf(out, "%llu", q->d_ino_timer);
prt_newline(out);
prt_str(out, "d_spc_timer");
prt_tab(out);
prt_printf(out, "%llu", q->d_spc_timer);
prt_newline(out);
prt_str(out, "d_ino_warns");
prt_tab(out);
prt_printf(out, "%i", q->d_ino_warns);
prt_newline(out);
prt_str(out, "d_spc_warns");
prt_tab(out);
prt_printf(out, "%i", q->d_spc_warns);
prt_newline(out);
prt_printf(out, "d_fieldmask\t%x\n", q->d_fieldmask);
prt_printf(out, "d_spc_hardlimit\t%llu\n", q->d_spc_hardlimit);
prt_printf(out, "d_spc_softlimit\t%llu\n", q->d_spc_softlimit);
prt_printf(out, "d_ino_hardlimit\%llu\n", q->d_ino_hardlimit);
prt_printf(out, "d_ino_softlimit\t%llu\n", q->d_ino_softlimit);
prt_printf(out, "d_space\t%llu\n", q->d_space);
prt_printf(out, "d_ino_count\t%llu\n", q->d_ino_count);
prt_printf(out, "d_ino_timer\t%llu\n", q->d_ino_timer);
prt_printf(out, "d_spc_timer\t%llu\n", q->d_spc_timer);
prt_printf(out, "d_ino_warns\t%i\n", q->d_ino_warns);
prt_printf(out, "d_spc_warns\t%i\n", q->d_spc_warns);
}
static inline unsigned __next_qtype(unsigned i, unsigned qtypes)
@ -610,10 +535,10 @@ int bch2_fs_quota_read(struct bch_fs *c)
int ret = bch2_trans_run(c,
for_each_btree_key(trans, iter, BTREE_ID_quotas, POS_MIN,
BTREE_ITER_PREFETCH, k,
BTREE_ITER_prefetch, k,
__bch2_quota_set(c, k, NULL)) ?:
for_each_btree_key(trans, iter, BTREE_ID_inodes, POS_MIN,
BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
BTREE_ITER_prefetch|BTREE_ITER_all_snapshots, k,
bch2_fs_quota_read_inode(trans, &iter, k)));
bch_err_fn(c, ret);
return ret;
@ -900,7 +825,7 @@ static int bch2_set_quota_trans(struct btree_trans *trans,
int ret;
k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_quotas, new_quota->k.p,
BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
BTREE_ITER_slots|BTREE_ITER_intent);
ret = bkey_err(k);
if (unlikely(ret))
return ret;

4
fs/bcachefs/quota.h
View File

@ -5,11 +5,11 @@
#include "inode.h"
#include "quota_types.h"
enum bkey_invalid_flags;
enum bch_validate_flags;
extern const struct bch_sb_field_ops bch_sb_field_ops_quota;
int bch2_quota_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_quota_to_text(struct printbuf *, struct bch_fs *, struct bkey_s_c);
#define bch2_bkey_ops_quota ((struct bkey_ops) { \

10
fs/bcachefs/rebalance.c
View File

@ -42,7 +42,7 @@ static int __bch2_set_rebalance_needs_scan(struct btree_trans *trans, u64 inum)
bch2_trans_iter_init(trans, &iter, BTREE_ID_rebalance_work,
SPOS(inum, REBALANCE_WORK_SCAN_OFFSET, U32_MAX),
BTREE_ITER_INTENT);
BTREE_ITER_intent);
k = bch2_btree_iter_peek_slot(&iter);
ret = bkey_err(k);
if (ret)
@ -89,7 +89,7 @@ static int bch2_clear_rebalance_needs_scan(struct btree_trans *trans, u64 inum,
bch2_trans_iter_init(trans, &iter, BTREE_ID_rebalance_work,
SPOS(inum, REBALANCE_WORK_SCAN_OFFSET, U32_MAX),
BTREE_ITER_INTENT);
BTREE_ITER_intent);
k = bch2_btree_iter_peek_slot(&iter);
ret = bkey_err(k);
if (ret)
@ -140,7 +140,7 @@ static struct bkey_s_c next_rebalance_extent(struct btree_trans *trans,
bch2_trans_iter_init(trans, extent_iter,
work_pos.inode ? BTREE_ID_extents : BTREE_ID_reflink,
work_pos,
BTREE_ITER_ALL_SNAPSHOTS);
BTREE_ITER_all_snapshots);
k = bch2_btree_iter_peek_slot(extent_iter);
if (bkey_err(k))
return k;
@ -323,12 +323,14 @@ static int do_rebalance(struct moving_context *ctxt)
struct bkey_s_c k;
int ret = 0;
bch2_trans_begin(trans);
bch2_move_stats_init(&r->work_stats, "rebalance_work");
bch2_move_stats_init(&r->scan_stats, "rebalance_scan");
bch2_trans_iter_init(trans, &rebalance_work_iter,
BTREE_ID_rebalance_work, POS_MIN,
BTREE_ITER_ALL_SNAPSHOTS);
BTREE_ITER_all_snapshots);
while (!bch2_move_ratelimit(ctxt)) {
if (!r->enabled) {

154
fs/bcachefs/recovery.c
View File

@ -65,9 +65,20 @@ static void bch2_reconstruct_alloc(struct bch_fs *c)
__set_bit_le64(BCH_FSCK_ERR_ptr_to_missing_alloc_key, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_ptr_gen_newer_than_bucket_gen, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_stale_dirty_ptr, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_dev_usage_buckets_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_dev_usage_sectors_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_dev_usage_fragmented_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_fs_usage_btree_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_fs_usage_cached_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_fs_usage_persistent_reserved_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_fs_usage_replicas_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_alloc_key_data_type_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_alloc_key_gen_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_alloc_key_dirty_sectors_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_alloc_key_cached_sectors_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_alloc_key_stripe_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_alloc_key_stripe_redundancy_wrong, ext->errors_silent);
__set_bit_le64(BCH_FSCK_ERR_need_discard_key_wrong, ext->errors_silent);
@ -125,9 +136,9 @@ static int bch2_journal_replay_key(struct btree_trans *trans,
{
struct btree_iter iter;
unsigned iter_flags =
BTREE_ITER_INTENT|
BTREE_ITER_NOT_EXTENTS;
unsigned update_flags = BTREE_TRIGGER_NORUN;
BTREE_ITER_intent|
BTREE_ITER_not_extents;
unsigned update_flags = BTREE_TRIGGER_norun;
int ret;
if (k->overwritten)
@ -136,17 +147,17 @@ static int bch2_journal_replay_key(struct btree_trans *trans,
trans->journal_res.seq = k->journal_seq;
/*
* BTREE_UPDATE_KEY_CACHE_RECLAIM disables key cache lookup/update to
* BTREE_UPDATE_key_cache_reclaim disables key cache lookup/update to
* keep the key cache coherent with the underlying btree. Nothing
* besides the allocator is doing updates yet so we don't need key cache
* coherency for non-alloc btrees, and key cache fills for snapshots
* btrees use BTREE_ITER_FILTER_SNAPSHOTS, which isn't available until
* btrees use BTREE_ITER_filter_snapshots, which isn't available until
* the snapshots recovery pass runs.
*/
if (!k->level && k->btree_id == BTREE_ID_alloc)
iter_flags |= BTREE_ITER_CACHED;
iter_flags |= BTREE_ITER_cached;
else
update_flags |= BTREE_UPDATE_KEY_CACHE_RECLAIM;
update_flags |= BTREE_UPDATE_key_cache_reclaim;
bch2_trans_node_iter_init(trans, &iter, k->btree_id, k->k->k.p,
BTREE_MAX_DEPTH, k->level,
@ -191,7 +202,7 @@ int bch2_journal_replay(struct bch_fs *c)
struct journal *j = &c->journal;
u64 start_seq = c->journal_replay_seq_start;
u64 end_seq = c->journal_replay_seq_start;
struct btree_trans *trans = bch2_trans_get(c);
struct btree_trans *trans = NULL;
bool immediate_flush = false;
int ret = 0;
@ -205,6 +216,7 @@ int bch2_journal_replay(struct bch_fs *c)
BUG_ON(!atomic_read(&keys->ref));
move_gap(keys, keys->nr);
trans = bch2_trans_get(c);
/*
* First, attempt to replay keys in sorted order. This is more
@ -361,14 +373,17 @@ static int journal_replay_entry_early(struct bch_fs *c,
case BCH_JSET_ENTRY_dev_usage: {
struct jset_entry_dev_usage *u =
container_of(entry, struct jset_entry_dev_usage, entry);
struct bch_dev *ca = bch_dev_bkey_exists(c, le32_to_cpu(u->dev));
unsigned i, nr_types = jset_entry_dev_usage_nr_types(u);
unsigned nr_types = jset_entry_dev_usage_nr_types(u);
for (i = 0; i < min_t(unsigned, nr_types, BCH_DATA_NR); i++) {
ca->usage_base->d[i].buckets = le64_to_cpu(u->d[i].buckets);
ca->usage_base->d[i].sectors = le64_to_cpu(u->d[i].sectors);
ca->usage_base->d[i].fragmented = le64_to_cpu(u->d[i].fragmented);
}
rcu_read_lock();
struct bch_dev *ca = bch2_dev_rcu(c, le32_to_cpu(u->dev));
if (ca)
for (unsigned i = 0; i < min_t(unsigned, nr_types, BCH_DATA_NR); i++) {
ca->usage_base->d[i].buckets = le64_to_cpu(u->d[i].buckets);
ca->usage_base->d[i].sectors = le64_to_cpu(u->d[i].sectors);
ca->usage_base->d[i].fragmented = le64_to_cpu(u->d[i].fragmented);
}
rcu_read_unlock();
break;
}
@ -597,57 +612,55 @@ int bch2_fs_recovery(struct bch_fs *c)
if (c->opts.norecovery)
c->opts.recovery_pass_last = BCH_RECOVERY_PASS_journal_replay - 1;
if (!c->opts.nochanges) {
mutex_lock(&c->sb_lock);
struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
bool write_sb = false;
mutex_lock(&c->sb_lock);
struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
bool write_sb = false;
if (BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb)) {
ext->recovery_passes_required[0] |=
cpu_to_le64(bch2_recovery_passes_to_stable(BIT_ULL(BCH_RECOVERY_PASS_check_topology)));
write_sb = true;
}
u64 sb_passes = bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0]));
if (sb_passes) {
struct printbuf buf = PRINTBUF;
prt_str(&buf, "superblock requires following recovery passes to be run:\n ");
prt_bitflags(&buf, bch2_recovery_passes, sb_passes);
bch_info(c, "%s", buf.buf);
printbuf_exit(&buf);
}
if (bch2_check_version_downgrade(c)) {
struct printbuf buf = PRINTBUF;
prt_str(&buf, "Version downgrade required:");
__le64 passes = ext->recovery_passes_required[0];
bch2_sb_set_downgrade(c,
BCH_VERSION_MINOR(bcachefs_metadata_version_current),
BCH_VERSION_MINOR(c->sb.version));
passes = ext->recovery_passes_required[0] & ~passes;
if (passes) {
prt_str(&buf, "\n running recovery passes: ");
prt_bitflags(&buf, bch2_recovery_passes,
bch2_recovery_passes_from_stable(le64_to_cpu(passes)));
}
bch_info(c, "%s", buf.buf);
printbuf_exit(&buf);
write_sb = true;
}
if (check_version_upgrade(c))
write_sb = true;
if (write_sb)
bch2_write_super(c);
c->recovery_passes_explicit |= bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0]));
mutex_unlock(&c->sb_lock);
if (BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb)) {
ext->recovery_passes_required[0] |=
cpu_to_le64(bch2_recovery_passes_to_stable(BIT_ULL(BCH_RECOVERY_PASS_check_topology)));
write_sb = true;
}
u64 sb_passes = bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0]));
if (sb_passes) {
struct printbuf buf = PRINTBUF;
prt_str(&buf, "superblock requires following recovery passes to be run:\n ");
prt_bitflags(&buf, bch2_recovery_passes, sb_passes);
bch_info(c, "%s", buf.buf);
printbuf_exit(&buf);
}
if (bch2_check_version_downgrade(c)) {
struct printbuf buf = PRINTBUF;
prt_str(&buf, "Version downgrade required:");
__le64 passes = ext->recovery_passes_required[0];
bch2_sb_set_downgrade(c,
BCH_VERSION_MINOR(bcachefs_metadata_version_current),
BCH_VERSION_MINOR(c->sb.version));
passes = ext->recovery_passes_required[0] & ~passes;
if (passes) {
prt_str(&buf, "\n running recovery passes: ");
prt_bitflags(&buf, bch2_recovery_passes,
bch2_recovery_passes_from_stable(le64_to_cpu(passes)));
}
bch_info(c, "%s", buf.buf);
printbuf_exit(&buf);
write_sb = true;
}
if (check_version_upgrade(c))
write_sb = true;
if (write_sb)
bch2_write_super(c);
c->recovery_passes_explicit |= bch2_recovery_passes_from_stable(le64_to_cpu(ext->recovery_passes_required[0]));
mutex_unlock(&c->sb_lock);
if (c->opts.fsck && IS_ENABLED(CONFIG_BCACHEFS_DEBUG))
c->recovery_passes_explicit |= BIT_ULL(BCH_RECOVERY_PASS_check_topology);
@ -660,7 +673,9 @@ int bch2_fs_recovery(struct bch_fs *c)
goto err;
}
if (!c->sb.clean || c->opts.fsck || c->opts.retain_recovery_info) {
bch2_journal_pos_from_member_info_resume(c);
if (!c->sb.clean || c->opts.retain_recovery_info) {
struct genradix_iter iter;
struct journal_replay **i;
@ -832,8 +847,8 @@ use_clean:
}
mutex_lock(&c->sb_lock);
struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
bool write_sb = false;
ext = bch2_sb_field_get(c->disk_sb.sb, ext);
write_sb = false;
if (BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb) != le16_to_cpu(c->disk_sb.sb->version)) {
SET_BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb, le16_to_cpu(c->disk_sb.sb->version));
@ -868,6 +883,9 @@ use_clean:
write_sb = true;
}
if (bch2_blacklist_entries_gc(c))
write_sb = true;
if (write_sb)
bch2_write_super(c);
mutex_unlock(&c->sb_lock);
@ -890,10 +908,6 @@ use_clean:
bch_info(c, "scanning for old btree nodes done");
}
if (c->journal_seq_blacklist_table &&
c->journal_seq_blacklist_table->nr > 128)
queue_work(system_long_wq, &c->journal_seq_blacklist_gc_work);
ret = 0;
out:
bch2_flush_fsck_errs(c);

8
fs/bcachefs/recovery_passes.c
View File

@ -26,11 +26,6 @@ const char * const bch2_recovery_passes[] = {
NULL
};
static int bch2_check_allocations(struct bch_fs *c)
{
return bch2_gc(c, true, false);
}
static int bch2_set_may_go_rw(struct bch_fs *c)
{
struct journal_keys *keys = &c->journal_keys;
@ -227,7 +222,8 @@ int bch2_run_recovery_passes(struct bch_fs *c)
if (should_run_recovery_pass(c, c->curr_recovery_pass)) {
unsigned pass = c->curr_recovery_pass;
ret = bch2_run_recovery_pass(c, c->curr_recovery_pass);
ret = bch2_run_recovery_pass(c, c->curr_recovery_pass) ?:
bch2_journal_flush(&c->journal);
if (bch2_err_matches(ret, BCH_ERR_restart_recovery) ||
(ret && c->curr_recovery_pass < pass))
continue;

72
fs/bcachefs/reflink.c
View File

@ -30,7 +30,7 @@ static inline unsigned bkey_type_to_indirect(const struct bkey *k)
/* reflink pointers */
int bch2_reflink_p_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
@ -74,20 +74,20 @@ bool bch2_reflink_p_merge(struct bch_fs *c, struct bkey_s _l, struct bkey_s_c _r
}
static int trans_trigger_reflink_p_segment(struct btree_trans *trans,
struct bkey_s_c_reflink_p p,
u64 *idx, unsigned flags)
struct bkey_s_c_reflink_p p, u64 *idx,
enum btree_iter_update_trigger_flags flags)
{
struct bch_fs *c = trans->c;
struct btree_iter iter;
struct bkey_i *k;
__le64 *refcount;
int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1;
int add = !(flags & BTREE_TRIGGER_overwrite) ? 1 : -1;
struct printbuf buf = PRINTBUF;
int ret;
k = bch2_bkey_get_mut_noupdate(trans, &iter,
BTREE_ID_reflink, POS(0, *idx),
BTREE_ITER_WITH_UPDATES);
BTREE_ITER_with_updates);
ret = PTR_ERR_OR_ZERO(k);
if (ret)
goto err;
@ -102,7 +102,7 @@ static int trans_trigger_reflink_p_segment(struct btree_trans *trans,
goto err;
}
if (!*refcount && (flags & BTREE_TRIGGER_OVERWRITE)) {
if (!*refcount && (flags & BTREE_TRIGGER_overwrite)) {
bch2_bkey_val_to_text(&buf, c, p.s_c);
bch2_trans_inconsistent(trans,
"indirect extent refcount underflow at %llu while marking\n %s",
@ -111,7 +111,7 @@ static int trans_trigger_reflink_p_segment(struct btree_trans *trans,
goto err;
}
if (flags & BTREE_TRIGGER_INSERT) {
if (flags & BTREE_TRIGGER_insert) {
struct bch_reflink_p *v = (struct bch_reflink_p *) p.v;
u64 pad;
@ -141,12 +141,13 @@ err:
}
static s64 gc_trigger_reflink_p_segment(struct btree_trans *trans,
struct bkey_s_c_reflink_p p,
u64 *idx, unsigned flags, size_t r_idx)
struct bkey_s_c_reflink_p p, u64 *idx,
enum btree_iter_update_trigger_flags flags,
size_t r_idx)
{
struct bch_fs *c = trans->c;
struct reflink_gc *r;
int add = !(flags & BTREE_TRIGGER_OVERWRITE) ? 1 : -1;
int add = !(flags & BTREE_TRIGGER_overwrite) ? 1 : -1;
u64 start = le64_to_cpu(p.v->idx);
u64 end = le64_to_cpu(p.v->idx) + p.k->size;
u64 next_idx = end + le32_to_cpu(p.v->back_pad);
@ -163,10 +164,13 @@ static s64 gc_trigger_reflink_p_segment(struct btree_trans *trans,
BUG_ON((s64) r->refcount + add < 0);
r->refcount += add;
if (flags & BTREE_TRIGGER_gc)
r->refcount += add;
*idx = r->offset;
return 0;
not_found:
BUG_ON(!(flags & BTREE_TRIGGER_check_repair));
if (fsck_err(c, reflink_p_to_missing_reflink_v,
"pointer to missing indirect extent\n"
" %s\n"
@ -189,7 +193,7 @@ not_found:
set_bkey_val_u64s(&update->k, 0);
}
ret = bch2_btree_insert_trans(trans, BTREE_ID_extents, update, BTREE_TRIGGER_NORUN);
ret = bch2_btree_insert_trans(trans, BTREE_ID_extents, update, BTREE_TRIGGER_norun);
}
*idx = next_idx;
@ -200,8 +204,8 @@ fsck_err:
}
static int __trigger_reflink_p(struct btree_trans *trans,
enum btree_id btree_id, unsigned level,
struct bkey_s_c k, unsigned flags)
enum btree_id btree_id, unsigned level, struct bkey_s_c k,
enum btree_iter_update_trigger_flags flags)
{
struct bch_fs *c = trans->c;
struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
@ -210,12 +214,12 @@ static int __trigger_reflink_p(struct btree_trans *trans,
u64 idx = le64_to_cpu(p.v->idx) - le32_to_cpu(p.v->front_pad);
u64 end = le64_to_cpu(p.v->idx) + p.k->size + le32_to_cpu(p.v->back_pad);
if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
if (flags & BTREE_TRIGGER_transactional) {
while (idx < end && !ret)
ret = trans_trigger_reflink_p_segment(trans, p, &idx, flags);
}
if (flags & BTREE_TRIGGER_GC) {
if (flags & (BTREE_TRIGGER_check_repair|BTREE_TRIGGER_gc)) {
size_t l = 0, r = c->reflink_gc_nr;
while (l < r) {
@ -238,10 +242,10 @@ int bch2_trigger_reflink_p(struct btree_trans *trans,
enum btree_id btree_id, unsigned level,
struct bkey_s_c old,
struct bkey_s new,
unsigned flags)
enum btree_iter_update_trigger_flags flags)
{
if ((flags & BTREE_TRIGGER_TRANSACTIONAL) &&
(flags & BTREE_TRIGGER_INSERT)) {
if ((flags & BTREE_TRIGGER_transactional) &&
(flags & BTREE_TRIGGER_insert)) {
struct bch_reflink_p *v = bkey_s_to_reflink_p(new).v;
v->front_pad = v->back_pad = 0;
@ -253,7 +257,7 @@ int bch2_trigger_reflink_p(struct btree_trans *trans,
/* indirect extents */
int bch2_reflink_v_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
return bch2_bkey_ptrs_invalid(c, k, flags, err);
@ -281,23 +285,25 @@ bool bch2_reflink_v_merge(struct bch_fs *c, struct bkey_s _l, struct bkey_s_c _r
}
#endif
static inline void check_indirect_extent_deleting(struct bkey_s new, unsigned *flags)
static inline void
check_indirect_extent_deleting(struct bkey_s new,
enum btree_iter_update_trigger_flags *flags)
{
if ((*flags & BTREE_TRIGGER_INSERT) && !*bkey_refcount(new)) {
if ((*flags & BTREE_TRIGGER_insert) && !*bkey_refcount(new)) {
new.k->type = KEY_TYPE_deleted;
new.k->size = 0;
set_bkey_val_u64s(new.k, 0);
*flags &= ~BTREE_TRIGGER_INSERT;
*flags &= ~BTREE_TRIGGER_insert;
}
}
int bch2_trigger_reflink_v(struct btree_trans *trans,
enum btree_id btree_id, unsigned level,
struct bkey_s_c old, struct bkey_s new,
unsigned flags)
enum btree_iter_update_trigger_flags flags)
{
if ((flags & BTREE_TRIGGER_TRANSACTIONAL) &&
(flags & BTREE_TRIGGER_INSERT))
if ((flags & BTREE_TRIGGER_transactional) &&
(flags & BTREE_TRIGGER_insert))
check_indirect_extent_deleting(new, &flags);
return bch2_trigger_extent(trans, btree_id, level, old, new, flags);
@ -306,7 +312,7 @@ int bch2_trigger_reflink_v(struct btree_trans *trans,
/* indirect inline data */
int bch2_indirect_inline_data_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
enum bch_validate_flags flags,
struct printbuf *err)
{
return 0;
@ -326,7 +332,7 @@ void bch2_indirect_inline_data_to_text(struct printbuf *out,
int bch2_trigger_indirect_inline_data(struct btree_trans *trans,
enum btree_id btree_id, unsigned level,
struct bkey_s_c old, struct bkey_s new,
unsigned flags)
enum btree_iter_update_trigger_flags flags)
{
check_indirect_extent_deleting(new, &flags);
@ -349,7 +355,7 @@ static int bch2_make_extent_indirect(struct btree_trans *trans,
bch2_check_set_feature(c, BCH_FEATURE_reflink_inline_data);
bch2_trans_iter_init(trans, &reflink_iter, BTREE_ID_reflink, POS_MAX,
BTREE_ITER_INTENT);
BTREE_ITER_intent);
k = bch2_btree_iter_peek_prev(&reflink_iter);
ret = bkey_err(k);
if (ret)
@ -394,7 +400,7 @@ static int bch2_make_extent_indirect(struct btree_trans *trans,
r_p->v.idx = cpu_to_le64(bkey_start_offset(&r_v->k));
ret = bch2_trans_update(trans, extent_iter, &r_p->k_i,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
BTREE_UPDATE_internal_snapshot_node);
err:
bch2_trans_iter_exit(trans, &reflink_iter);
@ -455,9 +461,9 @@ s64 bch2_remap_range(struct bch_fs *c,
goto err;
bch2_trans_iter_init(trans, &src_iter, BTREE_ID_extents, src_start,
BTREE_ITER_INTENT);
BTREE_ITER_intent);
bch2_trans_iter_init(trans, &dst_iter, BTREE_ID_extents, dst_start,
BTREE_ITER_INTENT);
BTREE_ITER_intent);
while ((ret == 0 ||
bch2_err_matches(ret, BCH_ERR_transaction_restart)) &&
@ -567,7 +573,7 @@ s64 bch2_remap_range(struct bch_fs *c,
bch2_trans_begin(trans);
ret2 = bch2_inode_peek(trans, &inode_iter, &inode_u,
dst_inum, BTREE_ITER_INTENT);
dst_inum, BTREE_ITER_intent);
if (!ret2 &&
inode_u.bi_size < new_i_size) {

16
fs/bcachefs/reflink.h
View File

@ -2,15 +2,16 @@
#ifndef _BCACHEFS_REFLINK_H
#define _BCACHEFS_REFLINK_H
enum bkey_invalid_flags;
enum bch_validate_flags;
int bch2_reflink_p_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_reflink_p_to_text(struct printbuf *, struct bch_fs *,
struct bkey_s_c);
bool bch2_reflink_p_merge(struct bch_fs *, struct bkey_s, struct bkey_s_c);
int bch2_trigger_reflink_p(struct btree_trans *, enum btree_id, unsigned,
struct bkey_s_c, struct bkey_s, unsigned);
struct bkey_s_c, struct bkey_s,
enum btree_iter_update_trigger_flags);
#define bch2_bkey_ops_reflink_p ((struct bkey_ops) { \
.key_invalid = bch2_reflink_p_invalid, \
@ -21,11 +22,12 @@ int bch2_trigger_reflink_p(struct btree_trans *, enum btree_id, unsigned,
})
int bch2_reflink_v_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_reflink_v_to_text(struct printbuf *, struct bch_fs *,
struct bkey_s_c);
int bch2_trigger_reflink_v(struct btree_trans *, enum btree_id, unsigned,
struct bkey_s_c, struct bkey_s, unsigned);
struct bkey_s_c, struct bkey_s,
enum btree_iter_update_trigger_flags);
#define bch2_bkey_ops_reflink_v ((struct bkey_ops) { \
.key_invalid = bch2_reflink_v_invalid, \
@ -36,13 +38,13 @@ int bch2_trigger_reflink_v(struct btree_trans *, enum btree_id, unsigned,
})
int bch2_indirect_inline_data_invalid(struct bch_fs *, struct bkey_s_c,
enum bkey_invalid_flags, struct printbuf *);
enum bch_validate_flags, struct printbuf *);
void bch2_indirect_inline_data_to_text(struct printbuf *,
struct bch_fs *, struct bkey_s_c);
int bch2_trigger_indirect_inline_data(struct btree_trans *,
enum btree_id, unsigned,
struct bkey_s_c, struct bkey_s,
unsigned);
enum btree_iter_update_trigger_flags);
#define bch2_bkey_ops_indirect_inline_data ((struct bkey_ops) { \
.key_invalid = bch2_indirect_inline_data_invalid, \

20
fs/bcachefs/replicas.c
View File

@ -84,7 +84,7 @@ int bch2_replicas_entry_validate(struct bch_replicas_entry_v1 *r,
}
for (unsigned i = 0; i < r->nr_devs; i++)
if (!bch2_dev_exists(sb, r->devs[i])) {
if (!bch2_member_exists(sb, r->devs[i])) {
prt_printf(err, "invalid device %u in entry ", r->devs[i]);
goto bad;
}
@ -200,7 +200,7 @@ cpu_replicas_add_entry(struct bch_fs *c,
};
for (i = 0; i < new_entry->nr_devs; i++)
BUG_ON(!bch2_dev_exists2(c, new_entry->devs[i]));
BUG_ON(!bch2_dev_exists(c, new_entry->devs[i]));
BUG_ON(!new_entry->data_type);
verify_replicas_entry(new_entry);
@ -860,7 +860,7 @@ static int bch2_cpu_replicas_validate(struct bch_replicas_cpu *cpu_r,
}
static int bch2_sb_replicas_validate(struct bch_sb *sb, struct bch_sb_field *f,
struct printbuf *err)
enum bch_validate_flags flags, struct printbuf *err)
{
struct bch_sb_field_replicas *sb_r = field_to_type(f, replicas);
struct bch_replicas_cpu cpu_r;
@ -899,7 +899,7 @@ const struct bch_sb_field_ops bch_sb_field_ops_replicas = {
};
static int bch2_sb_replicas_v0_validate(struct bch_sb *sb, struct bch_sb_field *f,
struct printbuf *err)
enum bch_validate_flags flags, struct printbuf *err)
{
struct bch_sb_field_replicas_v0 *sb_r = field_to_type(f, replicas_v0);
struct bch_replicas_cpu cpu_r;
@ -947,18 +947,20 @@ bool bch2_have_enough_devs(struct bch_fs *c, struct bch_devs_mask devs,
percpu_down_read(&c->mark_lock);
for_each_cpu_replicas_entry(&c->replicas, e) {
unsigned i, nr_online = 0, nr_failed = 0, dflags = 0;
unsigned nr_online = 0, nr_failed = 0, dflags = 0;
bool metadata = e->data_type < BCH_DATA_user;
if (e->data_type == BCH_DATA_cached)
continue;
for (i = 0; i < e->nr_devs; i++) {
struct bch_dev *ca = bch_dev_bkey_exists(c, e->devs[i]);
rcu_read_lock();
for (unsigned i = 0; i < e->nr_devs; i++) {
nr_online += test_bit(e->devs[i], devs.d);
nr_failed += ca->mi.state == BCH_MEMBER_STATE_failed;
struct bch_dev *ca = bch2_dev_rcu(c, e->devs[i]);
nr_failed += ca && ca->mi.state == BCH_MEMBER_STATE_failed;
}
rcu_read_unlock();
if (nr_failed == e->nr_devs)
continue;

15
fs/bcachefs/sb-clean.c
View File

@ -266,9 +266,8 @@ void bch2_journal_super_entries_add_common(struct bch_fs *c,
}
}
static int bch2_sb_clean_validate(struct bch_sb *sb,
struct bch_sb_field *f,
struct printbuf *err)
static int bch2_sb_clean_validate(struct bch_sb *sb, struct bch_sb_field *f,
enum bch_validate_flags flags, struct printbuf *err)
{
struct bch_sb_field_clean *clean = field_to_type(f, clean);
@ -283,7 +282,7 @@ static int bch2_sb_clean_validate(struct bch_sb *sb,
entry = vstruct_next(entry)) {
if ((void *) vstruct_next(entry) > vstruct_end(&clean->field)) {
prt_str(err, "entry type ");
bch2_prt_jset_entry_type(err, le16_to_cpu(entry->type));
bch2_prt_jset_entry_type(err, entry->type);
prt_str(err, " overruns end of section");
return -BCH_ERR_invalid_sb_clean;
}
@ -298,10 +297,8 @@ static void bch2_sb_clean_to_text(struct printbuf *out, struct bch_sb *sb,
struct bch_sb_field_clean *clean = field_to_type(f, clean);
struct jset_entry *entry;
prt_printf(out, "flags: %x", le32_to_cpu(clean->flags));
prt_newline(out);
prt_printf(out, "journal_seq: %llu", le64_to_cpu(clean->journal_seq));
prt_newline(out);
prt_printf(out, "flags: %x\n", le32_to_cpu(clean->flags));
prt_printf(out, "journal_seq: %llu\n", le64_to_cpu(clean->journal_seq));
for (entry = clean->start;
entry != vstruct_end(&clean->field);
@ -392,6 +389,8 @@ void bch2_fs_mark_clean(struct bch_fs *c)
goto out;
}
bch2_journal_pos_from_member_info_set(c);
bch2_write_super(c);
out:
mutex_unlock(&c->sb_lock);

20
fs/bcachefs/sb-counters.c
View File

@ -20,9 +20,8 @@ static size_t bch2_sb_counter_nr_entries(struct bch_sb_field_counters *ctrs)
return (__le64 *) vstruct_end(&ctrs->field) - &ctrs->d[0];
};
static int bch2_sb_counters_validate(struct bch_sb *sb,
struct bch_sb_field *f,
struct printbuf *err)
static int bch2_sb_counters_validate(struct bch_sb *sb, struct bch_sb_field *f,
enum bch_validate_flags flags, struct printbuf *err)
{
return 0;
};
@ -31,19 +30,12 @@ static void bch2_sb_counters_to_text(struct printbuf *out, struct bch_sb *sb,
struct bch_sb_field *f)
{
struct bch_sb_field_counters *ctrs = field_to_type(f, counters);
unsigned int i;
unsigned int nr = bch2_sb_counter_nr_entries(ctrs);
for (i = 0; i < nr; i++) {
if (i < BCH_COUNTER_NR)
prt_printf(out, "%s ", bch2_counter_names[i]);
else
prt_printf(out, "(unknown)");
prt_tab(out);
prt_printf(out, "%llu", le64_to_cpu(ctrs->d[i]));
prt_newline(out);
}
for (unsigned i = 0; i < nr; i++)
prt_printf(out, "%s \t%llu\n",
i < BCH_COUNTER_NR ? bch2_counter_names[i] : "(unknown)",
le64_to_cpu(ctrs->d[i]));
};
int bch2_sb_counters_to_cpu(struct bch_fs *c)

25
fs/bcachefs/sb-downgrade.c
View File

@ -134,15 +134,25 @@ downgrade_entry_next_c(const struct bch_sb_field_downgrade_entry *e)
#define for_each_downgrade_entry(_d, _i) \
for (const struct bch_sb_field_downgrade_entry *_i = (_d)->entries; \
(void *) _i < vstruct_end(&(_d)->field) && \
(void *) &_i->errors[0] < vstruct_end(&(_d)->field); \
(void *) &_i->errors[0] <= vstruct_end(&(_d)->field) && \
(void *) downgrade_entry_next_c(_i) <= vstruct_end(&(_d)->field); \
_i = downgrade_entry_next_c(_i))
static int bch2_sb_downgrade_validate(struct bch_sb *sb, struct bch_sb_field *f,
struct printbuf *err)
enum bch_validate_flags flags, struct printbuf *err)
{
struct bch_sb_field_downgrade *e = field_to_type(f, downgrade);
for_each_downgrade_entry(e, i) {
for (const struct bch_sb_field_downgrade_entry *i = e->entries;
(void *) i < vstruct_end(&e->field);
i = downgrade_entry_next_c(i)) {
if (flags & BCH_VALIDATE_write &&
((void *) &i->errors[0] > vstruct_end(&e->field) ||
(void *) downgrade_entry_next_c(i) > vstruct_end(&e->field))) {
prt_printf(err, "downgrade entry overruns end of superblock section)");
return -BCH_ERR_invalid_sb_downgrade;
}
if (BCH_VERSION_MAJOR(le16_to_cpu(i->version)) !=
BCH_VERSION_MAJOR(le16_to_cpu(sb->version))) {
prt_printf(err, "downgrade entry with mismatched major version (%u != %u)",
@ -164,19 +174,16 @@ static void bch2_sb_downgrade_to_text(struct printbuf *out, struct bch_sb *sb,
printbuf_tabstop_push(out, 16);
for_each_downgrade_entry(e, i) {
prt_str(out, "version:");
prt_tab(out);
prt_str(out, "version:\t");
bch2_version_to_text(out, le16_to_cpu(i->version));
prt_newline(out);
prt_str(out, "recovery passes:");
prt_tab(out);
prt_str(out, "recovery passes:\t");
prt_bitflags(out, bch2_recovery_passes,
bch2_recovery_passes_from_stable(le64_to_cpu(i->recovery_passes[0])));
prt_newline(out);
prt_str(out, "errors:");
prt_tab(out);
prt_str(out, "errors:\t");
bool first = true;
for (unsigned j = 0; j < le16_to_cpu(i->nr_errors); j++) {
if (!first)

Some files were not shown because too many files have changed in this diff Show More