summaryrefslogtreecommitdiff
path: root/fs/bcachefs/btree_locking.c
AgeCommit message (Collapse)Author
2025-06-26bcachefs: Plumb correct ip to trans_relock_fail tracepointKent Overstreet
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-06-11bcachefs: Fix possible console lock involved deadlockAlan Huang
Link: https://lore.kernel.org/all/6822ab02.050a0220.f2294.00cb.GAE@google.com/T/ Reported-by: syzbot+2c3ef91c9523c3d1a25c@syzkaller.appspotmail.com Signed-off-by: Alan Huang <mmpgouride@gmail.com> Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-06-01bcachefs: Replace rcu_read_lock() with guardsKent Overstreet
The new guard(), scoped_guard() allow for more natural code. Some of the uses with creative flow control have been left. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-30bcachefs: factor out break_cycle_fail()Kent Overstreet
More stack usage work. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-23bcachefs: Path must be locked if trans->locked && should_be_lockedKent Overstreet
If path->should_be_locked is true, that means user code (of the btree API) has seen, in this transaction, something guarded by the node this path has locked, and we have to keep it locked until the end of the transaction. Assert that we're not violating this; should_be_locked should also be cleared only in _very_ special situations. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-23bcachefs: Plumb btree_trans for more locking assertsKent Overstreet
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-23bcachefs: Clear trans->locked before unlockKent Overstreet
We're adding new should_be_locked assertions: it's going to be illegal to unlock a should_be_locked path when trans->locked is true. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-23bcachefs: bch2_path_get() reuses paths if upgrade_fails & !should_be_lockedKent Overstreet
Small additional optimization over the previous patch, bringing us closer to the original behaviour, except when we need to clone to avoid a transaction restart. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-23bcachefs: Fix btree_path_get_locks when not doing trans restartKent Overstreet
btree_path_get_locks, on failure, shouldn't unlock if we're not issuing a transaction restart: we might drop locks we're not supposed to (if path->should_be_locked is set). Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-23bcachefs: btree_node_locked_type_nowrite()Kent Overstreet
Small helper to improve locking assertions. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-21bcachefs: Improve trace_trans_restart_upgradeKent Overstreet
- Convert to a 'fs_str' tracepoint that just emits as a string: this lets us build up the tracepoint with a printbuf, using our pretty printers, and they're much easier to manage - Include locks_held, before and after - Include the btree node pointer we failed on (error pointer, null, or real node) Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-21bcachefs: relock_fail tracepoint now includes btreeKent Overstreet
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-21bcachefs: debug_check_btree_locking modparamKent Overstreet
Don't put btree locking asserts behind CONFIG_BCACHEFS_DEBUG, put them behind a module parameter. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-21bcachefs: Slim down inlined part of bch2_btree_path_upgrade()Kent Overstreet
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-21bcachefs: print_str_as_lines() -> print_str()Kent Overstreet
bch2_print_string_as_lines() is a low level helper that allows messages longer than 1k to be printed without truncation. But we should always be printing with the helpers that take a filesystem object, if we're in fsck they direct output to the userspace process controlling fsck instead of the dmesg log. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-05-21bcachefs: Kill bch2_trans_unlock_noassertAlan Huang
Signed-off-by: Alan Huang <mmpgouride@gmail.com> Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-03-24bcachefs: Fix race in print_chain()Kent Overstreet
00636 Unable to handle kernel NULL pointer dereference at virtual address 00000000000000b0 00636 Mem abort info: 00636 ESR = 0x0000000096000005 00636 EC = 0x25: DABT (current EL), IL = 32 bits 00636 SET = 0, FnV = 0 00636 EA = 0, S1PTW = 0 00636 FSC = 0x05: level 1 translation fault 00636 Data abort info: 00636 ISV = 0, ISS = 0x00000005, ISS2 = 0x00000000 00636 CM = 0, WnR = 0, TnD = 0, TagAccess = 0 00636 GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0 00636 user pgtable: 4k pages, 39-bit VAs, pgdp=0000000101b10000 00636 [00000000000000b0] pgd=0000000000000000, p4d=0000000000000000, pud=0000000000000000 00636 Internal error: Oops: 0000000096000005 [#1] SMP 00636 Modules linked in: 00636 CPU: 12 UID: 0 PID: 79369 Comm: cat Not tainted 6.14.0-rc6-ktest-g3783b8973ab7 #17757 00636 Hardware name: linux,dummy-virt (DT) 00636 pstate: 20001005 (nzCv daif -PAN -UAO -TCO -DIT +SSBS BTYPE=--) 00636 pc : print_chain+0xb8/0x170 00636 lr : print_chain+0xa0/0x170 00636 sp : ffffff80d9c1bbb0 00636 x29: ffffff80d9c1bbb0 x28: 0000000000000002 x27: ffffff80c1be8250 00636 x26: ffffff80dd9b0000 x25: 0000000000000020 x24: 000000000000002d 00636 x23: 000000000000003c x22: ffffffc080a54518 x21: ffffff80da6e00d0 00636 x20: ffffff80da6e0170 x19: ffffff80c1a1d240 x18: 00000000ffffffff 00636 x17: 3535303937202d3c x16: 203139202d3c2035 x15: 00000000ffffffff 00636 x14: 0000000000000000 x13: ffffff80d71b63f1 x12: 0000000000000006 00636 x11: ffffffc080beb1c0 x10: 0000000000000020 x9 : 00000000000134cc 00636 x8 : 0000000000000020 x7 : 0000000000000004 x6 : 0000000000000020 00636 x5 : ffffff80d71b63f7 x4 : ffffffc080a5451b x3 : 0000000000000000 00636 x2 : 0000000000000000 x1 : 0000000000000000 x0 : 0000000000000000 00636 Call trace: 00636 print_chain+0xb8/0x170 (P) 00636 bch2_check_for_deadlock+0x444/0x5a0 00636 bch2_btree_deadlock_read+0xb4/0x1c8 00636 full_proxy_read+0x74/0xd8 00636 vfs_read+0x90/0x300 Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-03-24bcachefs: btree_trans_restart_foreign_task()Kent Overstreet
In debug mode, we save the call stack on transaction restart - but there's no locking, so we can't touch it if we're issuing the restart from another thread. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-02-26bcachefs: Fix deadlockAlan Huang
This fixes two deadlocks: 1.pcpu_alloc_mutex involved one as pointed by syzbot[1] 2.recursion deadlock. The root cause is that we hold the bc lock during alloc_percpu, fix it by following the pattern used by __btree_node_mem_alloc(). [1] https://lore.kernel.org/all/66f97d9a.050a0220.6bad9.001d.GAE@google.com/T/ Reported-by: syzbot+fe63f377148a6371a9db@syzkaller.appspotmail.com Tested-by: syzbot+fe63f377148a6371a9db@syzkaller.appspotmail.com Signed-off-by: Alan Huang <mmpgouride@gmail.com> Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-01-14bcachefs: Pop all the transactions from the abort oneAlan Huang
The transaction is going to abort, so there will be no cycle involving this transaction anymore. Signed-off-by: Alan Huang <mmpgouride@gmail.com> Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-01-14bcachefs: Only abort the transactions in the cycleAlan Huang
When the cycle doesn't involve the initiator of the cycle detection, we might choose a transaction that is not involved in the cycle to abort. It shouldn't be that since it won't break the cycle, this patch therefore chooses the transaction in the cycle to abort. Signed-off-by: Alan Huang <mmpgouride@gmail.com> Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-01-14bcachefs: Introduce lock_graph_pop_fromAlan Huang
This patch introduces a helper function called lock_graph_pop_from, it pops the graph from i. Signed-off-by: Alan Huang <mmpgouride@gmail.com> Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-01-14bcachefs: Convert open-coded lock_graph_pop_all to helperAlan Huang
Signed-off-by: Alan Huang <mmpgouride@gmail.com> Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-01-14bcachefs: Do not allow no fail lock request to failAlan Huang
If the transaction chose itself as a victim before and restarted, it might request a no fail lock request this time. But it might be added to others' lock graph and be chose as the victim again, it's no longer safe without additional check. We can also convert the cycle detector to be fully RCU-based to solve that unsoundness, but the latency added to trans_put and additional memory required may not worth it. Signed-off-by: Alan Huang <mmpgouride@gmail.com> Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-01-14Revert "bcachefs: Fix bch2_btree_node_upgrade()"Alan Huang
This reverts commit 62448afee714354a26db8a0f3c644f58628f0792. six_lock_tryupgrade fails only if there is an intent lock held, it won't fail no matter how many read locks are held. Signed-off-by: Alan Huang <mmpgouride@gmail.com> Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-01-09bcachefs: bch2_trans_unlock_write()Kent Overstreet
New helper for dropping all write locks; which is distinct from the helper the transaction commit path uses, which is faster and only touches updates. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2025-01-09bcachefs: btree_path_very_locks(): verify lock seqKent Overstreet
If the btree_path's lock seq is wrong, the next bch2_trans_relock() operation is guaranteed to fail and we take an unnecessary transaction restart. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-12-21bcachefs: bch2_trans_relock() is trylock for lockdepKent Overstreet
fix some spurious lockdep splats Reported-by: syzbot+e088be3c2d5c05aaac35@syzkaller.appspotmail.com Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-07-14bcachefs: Kill bch2_assert_btree_nodes_not_locked()Kent Overstreet
We no longer track individual btree node locks with lockdep, so this will never be enabled. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-07-14lockdep: lockdep_set_notrack_class()Kent Overstreet
Add a new helper to disable lockdep tracking entirely for a given class. This is needed for bcachefs, which takes too many btree node locks for lockdep to track. Instead, we have a single lockdep_map for "btree_trans has any btree nodes locked", which makes more since given that we have centralized lock management and a cycle detector. Cc: Peter Zijlstra <peterz@infradead.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Will Deacon <will@kernel.org> Cc: Waiman Long <longman@redhat.com> Cc: Boqun Feng <boqun.feng@gmail.com> Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-07-11bcachefs: Set PF_MEMALLOC_NOFS when trans->lockedKent Overstreet
proper lock ordering is: fs_reclaim -> btree node locks Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-07-10bcachefs: fix scheduling while atomic in break_cycle()Kent Overstreet
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-06-05bcachefs: Fix GFP_KERNEL allocation in break_cycle()Kent Overstreet
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-05-08bcachefs: bch2_trans_verify_not_unlocked()Kent Overstreet
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-05-08bcachefs: trans->lockedKent Overstreet
Add a field for tracking whether a transaction object holds btree locks, and assertions to verify state. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-05-08bcachefs: bch2_btree_path_upgrade() checks nodes_locked, not uptodateKent Overstreet
In the key cache fill path, we use path_upgrade() on a path that isn't uptodate yet but should be locked. This change makes bch2_btree_path_upgrade() slightly looser so we can use it in key cache upgrade, instead of the __ version. Also, make the related assert - that path->uptodate implies nodes_locked - slightly clearer. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-05-08bcachefs: bch2_trans_relock_fail() - factor out slowpathKent Overstreet
Factor out slowpath into a separate helper Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-05-08bcachefs: bch2_btree_path_to_text()Kent Overstreet
Long form version of bch2_btree_path_to_text() - useful in error messages and tracepoints. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-05-08bcachefs: prt_printf() now respects \r\n\tKent Overstreet
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-04-10bcachefs: Kill read lock dropping in bch2_btree_node_lock_write_nofail()Kent Overstreet
dropping read locks in bch2_btree_node_lock_write_nofail() dates from before we had the cycle detector; we can now tell the cycle detector directly when taking a lock may not fail because we can't handle transaction restarts. This is needed for adding should_be_locked asserts. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-03-13bcachefs: Drop redundant btree_path_downgrade()sKent Overstreet
If a path doesn't have any active references, we shouldn't downgrade it; it'll either be reused, possibly with intent refs again, or dropped at bch2_trans_begin() time. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-01-22bcachefs: Add gfp flags param to bch2_prt_task_backtrace()Kent Overstreet
Fixes: e6a2566f7a00 ("bcachefs: Better journal tracepoints") Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev> Reported-by: smatch
2024-01-21bcachefs: Improve trace_trans_restart_relockKent Overstreet
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-01-05bcachefs: Improve would_deadlock trace eventKent Overstreet
We now include backtraces for every thread involved in the cycle. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-01-01bcachefs: optimize __bch2_trans_get(), kill DEBUG_TRANSACTIONSKent Overstreet
- Some tweaks to greatly reduce locking overhead for the list of btree transactions, so that it can always be enabled: leave btree_trans objects on the list when they're on the percpu single item freelist, and only check for duplicates in the same process when CONFIG_BCACHEFS_DEBUG is enabled - don't zero out the full btree_trans() unless we allocated it from the mempool Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-01-01bcachefs: rcu protect trans->pathsKent Overstreet
Upcoming patches are going to be changing trans->paths to a reallocatable buffer. We need to guard against use after free when it's used by other threads; this introduces RCU protection to those paths and changes them to check for trans->paths == NULL Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-01-01bcachefs: kill btree_path.idxKent Overstreet
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-01-01bcachefs: trans_for_each_path() no longer uses path->idxKent Overstreet
path->idx is now a code smell: we should be using path_idx_t, since it's stable across btree path reallocation. This is also a bit faster, using the same loop counter vs. fetching path->idx from each path we iterate over. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-01-01bcachefs: kill btree_path->(alloc_seq|downgrade_seq)Kent Overstreet
These were for extra info in tracepoints for debugging a specialized issue - we do not want to bloat btree_path for this, at least in release builds. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2024-01-01bcachefs: Improve trace_trans_restart_would_deadlockKent Overstreet
In the CI, we're seeing tests failing due to excessive would_deadlock transaction restarts - the tracepoint now includes the lock cycle that occured. Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
generated by cgit (git 2.34.1) at 2025-08-04 06:27:12 +0000