mirror of
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git
synced 2024-09-05 10:29:37 +00:00
![Josef Bacik](/assets/img/avatar_default.png)
We have been hitting the following lockdep splat with btrfs/187 recently WARNING: possible circular locking dependency detected 5.19.0-rc8+ #775 Not tainted ------------------------------------------------------ btrfs/752500 is trying to acquire lock: ffff97e1875a97b8 (btrfs-treloc-02#2){+.+.}-{3:3}, at: __btrfs_tree_lock+0x24/0x110 but task is already holding lock: ffff97e1875a9278 (btrfs-tree-01/1){+.+.}-{3:3}, at: __btrfs_tree_lock+0x24/0x110 which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #2 (btrfs-tree-01/1){+.+.}-{3:3}: down_write_nested+0x41/0x80 __btrfs_tree_lock+0x24/0x110 btrfs_init_new_buffer+0x7d/0x2c0 btrfs_alloc_tree_block+0x120/0x3b0 __btrfs_cow_block+0x136/0x600 btrfs_cow_block+0x10b/0x230 btrfs_search_slot+0x53b/0xb70 btrfs_lookup_inode+0x2a/0xa0 __btrfs_update_delayed_inode+0x5f/0x280 btrfs_async_run_delayed_root+0x24c/0x290 btrfs_work_helper+0xf2/0x3e0 process_one_work+0x271/0x590 worker_thread+0x52/0x3b0 kthread+0xf0/0x120 ret_from_fork+0x1f/0x30 -> #1 (btrfs-tree-01){++++}-{3:3}: down_write_nested+0x41/0x80 __btrfs_tree_lock+0x24/0x110 btrfs_search_slot+0x3c3/0xb70 do_relocation+0x10c/0x6b0 relocate_tree_blocks+0x317/0x6d0 relocate_block_group+0x1f1/0x560 btrfs_relocate_block_group+0x23e/0x400 btrfs_relocate_chunk+0x4c/0x140 btrfs_balance+0x755/0xe40 btrfs_ioctl+0x1ea2/0x2c90 __x64_sys_ioctl+0x88/0xc0 do_syscall_64+0x38/0x90 entry_SYSCALL_64_after_hwframe+0x63/0xcd -> #0 (btrfs-treloc-02#2){+.+.}-{3:3}: __lock_acquire+0x1122/0x1e10 lock_acquire+0xc2/0x2d0 down_write_nested+0x41/0x80 __btrfs_tree_lock+0x24/0x110 btrfs_lock_root_node+0x31/0x50 btrfs_search_slot+0x1cb/0xb70 replace_path+0x541/0x9f0 merge_reloc_root+0x1d6/0x610 merge_reloc_roots+0xe2/0x260 relocate_block_group+0x2c8/0x560 btrfs_relocate_block_group+0x23e/0x400 btrfs_relocate_chunk+0x4c/0x140 btrfs_balance+0x755/0xe40 btrfs_ioctl+0x1ea2/0x2c90 __x64_sys_ioctl+0x88/0xc0 do_syscall_64+0x38/0x90 entry_SYSCALL_64_after_hwframe+0x63/0xcd other info that might help us debug this: Chain exists of: btrfs-treloc-02#2 --> btrfs-tree-01 --> btrfs-tree-01/1 Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(btrfs-tree-01/1); lock(btrfs-tree-01); lock(btrfs-tree-01/1); lock(btrfs-treloc-02#2); *** DEADLOCK *** 7 locks held by btrfs/752500: #0: ffff97e292fdf460 (sb_writers#12){.+.+}-{0:0}, at: btrfs_ioctl+0x208/0x2c90 #1: ffff97e284c02050 (&fs_info->reclaim_bgs_lock){+.+.}-{3:3}, at: btrfs_balance+0x55f/0xe40 #2: ffff97e284c00878 (&fs_info->cleaner_mutex){+.+.}-{3:3}, at: btrfs_relocate_block_group+0x236/0x400 #3: ffff97e292fdf650 (sb_internal#2){.+.+}-{0:0}, at: merge_reloc_root+0xef/0x610 #4: ffff97e284c02378 (btrfs_trans_num_writers){++++}-{0:0}, at: join_transaction+0x1a8/0x5a0 #5: ffff97e284c023a0 (btrfs_trans_num_extwriters){++++}-{0:0}, at: join_transaction+0x1a8/0x5a0 #6: ffff97e1875a9278 (btrfs-tree-01/1){+.+.}-{3:3}, at: __btrfs_tree_lock+0x24/0x110 stack backtrace: CPU: 1 PID: 752500 Comm: btrfs Not tainted 5.19.0-rc8+ #775 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.13.0-2.fc32 04/01/2014 Call Trace: dump_stack_lvl+0x56/0x73 check_noncircular+0xd6/0x100 ? lock_is_held_type+0xe2/0x140 __lock_acquire+0x1122/0x1e10 lock_acquire+0xc2/0x2d0 ? __btrfs_tree_lock+0x24/0x110 down_write_nested+0x41/0x80 ? __btrfs_tree_lock+0x24/0x110 __btrfs_tree_lock+0x24/0x110 btrfs_lock_root_node+0x31/0x50 btrfs_search_slot+0x1cb/0xb70 ? lock_release+0x137/0x2d0 ? _raw_spin_unlock+0x29/0x50 ? release_extent_buffer+0x128/0x180 replace_path+0x541/0x9f0 merge_reloc_root+0x1d6/0x610 merge_reloc_roots+0xe2/0x260 relocate_block_group+0x2c8/0x560 btrfs_relocate_block_group+0x23e/0x400 btrfs_relocate_chunk+0x4c/0x140 btrfs_balance+0x755/0xe40 btrfs_ioctl+0x1ea2/0x2c90 ? lock_is_held_type+0xe2/0x140 ? lock_is_held_type+0xe2/0x140 ? __x64_sys_ioctl+0x88/0xc0 __x64_sys_ioctl+0x88/0xc0 do_syscall_64+0x38/0x90 entry_SYSCALL_64_after_hwframe+0x63/0xcd This isn't necessarily new, it's just tricky to hit in practice. There are two competing things going on here. With relocation we create a snapshot of every fs tree with a reloc tree. Any extent buffers that get initialized here are initialized with the reloc root lockdep key. However since it is a snapshot, any blocks that are currently in cache that originally belonged to the fs tree will have the normal tree lockdep key set. This creates the lock dependency of reloc tree -> normal tree for the extent buffer locking during the first phase of the relocation as we walk down the reloc root to relocate blocks. However this is problematic because the final phase of the relocation is merging the reloc root into the original fs root. This involves searching down to any keys that exist in the original fs root and then swapping the relocated block and the original fs root block. We have to search down to the fs root first, and then go search the reloc root for the block we need to replace. This creates the dependency of normal tree -> reloc tree which is why lockdep complains. Additionally even if we were to fix this particular mismatch with a different nesting for the merge case, we're still slotting in a block that has a owner of the reloc root objectid into a normal tree, so that block will have its lockdep key set to the tree reloc root, and create a lockdep splat later on when we wander into that block from the fs root. Unfortunately the only solution here is to make sure we do not set the lockdep key to the reloc tree lockdep key normally, and then reset any blocks we wander into from the reloc root when we're doing the merged. This solves the problem of having mixed tree reloc keys intermixed with normal tree keys, and then allows us to make sure in the merge case we maintain the lock order of normal tree -> reloc tree We handle this by setting a bit on the reloc root when we do the search for the block we want to relocate, and any block we search into or COW at that point gets set to the reloc tree key. This works correctly because we only ever COW down to the parent node, so we aren't resetting the key for the block we're linking into the fs root. With this patch we no longer have the lockdep splat in btrfs/187. Signed-off-by: Josef Bacik <josef@toxicpanda.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
379 lines
9.7 KiB
C
379 lines
9.7 KiB
C
// SPDX-License-Identifier: GPL-2.0
|
|
/*
|
|
* Copyright (C) 2008 Oracle. All rights reserved.
|
|
*/
|
|
|
|
#include <linux/sched.h>
|
|
#include <linux/pagemap.h>
|
|
#include <linux/spinlock.h>
|
|
#include <linux/page-flags.h>
|
|
#include <asm/bug.h>
|
|
#include "misc.h"
|
|
#include "ctree.h"
|
|
#include "extent_io.h"
|
|
#include "locking.h"
|
|
|
|
/*
|
|
* Lockdep class keys for extent_buffer->lock's in this root. For a given
|
|
* eb, the lockdep key is determined by the btrfs_root it belongs to and
|
|
* the level the eb occupies in the tree.
|
|
*
|
|
* Different roots are used for different purposes and may nest inside each
|
|
* other and they require separate keysets. As lockdep keys should be
|
|
* static, assign keysets according to the purpose of the root as indicated
|
|
* by btrfs_root->root_key.objectid. This ensures that all special purpose
|
|
* roots have separate keysets.
|
|
*
|
|
* Lock-nesting across peer nodes is always done with the immediate parent
|
|
* node locked thus preventing deadlock. As lockdep doesn't know this, use
|
|
* subclass to avoid triggering lockdep warning in such cases.
|
|
*
|
|
* The key is set by the readpage_end_io_hook after the buffer has passed
|
|
* csum validation but before the pages are unlocked. It is also set by
|
|
* btrfs_init_new_buffer on freshly allocated blocks.
|
|
*
|
|
* We also add a check to make sure the highest level of the tree is the
|
|
* same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
|
|
* needs update as well.
|
|
*/
|
|
#ifdef CONFIG_DEBUG_LOCK_ALLOC
|
|
#if BTRFS_MAX_LEVEL != 8
|
|
#error
|
|
#endif
|
|
|
|
#define DEFINE_LEVEL(stem, level) \
|
|
.names[level] = "btrfs-" stem "-0" #level,
|
|
|
|
#define DEFINE_NAME(stem) \
|
|
DEFINE_LEVEL(stem, 0) \
|
|
DEFINE_LEVEL(stem, 1) \
|
|
DEFINE_LEVEL(stem, 2) \
|
|
DEFINE_LEVEL(stem, 3) \
|
|
DEFINE_LEVEL(stem, 4) \
|
|
DEFINE_LEVEL(stem, 5) \
|
|
DEFINE_LEVEL(stem, 6) \
|
|
DEFINE_LEVEL(stem, 7)
|
|
|
|
static struct btrfs_lockdep_keyset {
|
|
u64 id; /* root objectid */
|
|
/* Longest entry: btrfs-free-space-00 */
|
|
char names[BTRFS_MAX_LEVEL][20];
|
|
struct lock_class_key keys[BTRFS_MAX_LEVEL];
|
|
} btrfs_lockdep_keysets[] = {
|
|
{ .id = BTRFS_ROOT_TREE_OBJECTID, DEFINE_NAME("root") },
|
|
{ .id = BTRFS_EXTENT_TREE_OBJECTID, DEFINE_NAME("extent") },
|
|
{ .id = BTRFS_CHUNK_TREE_OBJECTID, DEFINE_NAME("chunk") },
|
|
{ .id = BTRFS_DEV_TREE_OBJECTID, DEFINE_NAME("dev") },
|
|
{ .id = BTRFS_CSUM_TREE_OBJECTID, DEFINE_NAME("csum") },
|
|
{ .id = BTRFS_QUOTA_TREE_OBJECTID, DEFINE_NAME("quota") },
|
|
{ .id = BTRFS_TREE_LOG_OBJECTID, DEFINE_NAME("log") },
|
|
{ .id = BTRFS_TREE_RELOC_OBJECTID, DEFINE_NAME("treloc") },
|
|
{ .id = BTRFS_DATA_RELOC_TREE_OBJECTID, DEFINE_NAME("dreloc") },
|
|
{ .id = BTRFS_UUID_TREE_OBJECTID, DEFINE_NAME("uuid") },
|
|
{ .id = BTRFS_FREE_SPACE_TREE_OBJECTID, DEFINE_NAME("free-space") },
|
|
{ .id = 0, DEFINE_NAME("tree") },
|
|
};
|
|
|
|
#undef DEFINE_LEVEL
|
|
#undef DEFINE_NAME
|
|
|
|
void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, int level)
|
|
{
|
|
struct btrfs_lockdep_keyset *ks;
|
|
|
|
BUG_ON(level >= ARRAY_SIZE(ks->keys));
|
|
|
|
/* Find the matching keyset, id 0 is the default entry */
|
|
for (ks = btrfs_lockdep_keysets; ks->id; ks++)
|
|
if (ks->id == objectid)
|
|
break;
|
|
|
|
lockdep_set_class_and_name(&eb->lock, &ks->keys[level], ks->names[level]);
|
|
}
|
|
|
|
void btrfs_maybe_reset_lockdep_class(struct btrfs_root *root, struct extent_buffer *eb)
|
|
{
|
|
if (test_bit(BTRFS_ROOT_RESET_LOCKDEP_CLASS, &root->state))
|
|
btrfs_set_buffer_lockdep_class(root->root_key.objectid,
|
|
eb, btrfs_header_level(eb));
|
|
}
|
|
|
|
#endif
|
|
|
|
/*
|
|
* Extent buffer locking
|
|
* =====================
|
|
*
|
|
* We use a rw_semaphore for tree locking, and the semantics are exactly the
|
|
* same:
|
|
*
|
|
* - reader/writer exclusion
|
|
* - writer/writer exclusion
|
|
* - reader/reader sharing
|
|
* - try-lock semantics for readers and writers
|
|
*
|
|
* The rwsem implementation does opportunistic spinning which reduces number of
|
|
* times the locking task needs to sleep.
|
|
*/
|
|
|
|
/*
|
|
* __btrfs_tree_read_lock - lock extent buffer for read
|
|
* @eb: the eb to be locked
|
|
* @nest: the nesting level to be used for lockdep
|
|
*
|
|
* This takes the read lock on the extent buffer, using the specified nesting
|
|
* level for lockdep purposes.
|
|
*/
|
|
void __btrfs_tree_read_lock(struct extent_buffer *eb, enum btrfs_lock_nesting nest)
|
|
{
|
|
u64 start_ns = 0;
|
|
|
|
if (trace_btrfs_tree_read_lock_enabled())
|
|
start_ns = ktime_get_ns();
|
|
|
|
down_read_nested(&eb->lock, nest);
|
|
trace_btrfs_tree_read_lock(eb, start_ns);
|
|
}
|
|
|
|
void btrfs_tree_read_lock(struct extent_buffer *eb)
|
|
{
|
|
__btrfs_tree_read_lock(eb, BTRFS_NESTING_NORMAL);
|
|
}
|
|
|
|
/*
|
|
* Try-lock for read.
|
|
*
|
|
* Return 1 if the rwlock has been taken, 0 otherwise
|
|
*/
|
|
int btrfs_try_tree_read_lock(struct extent_buffer *eb)
|
|
{
|
|
if (down_read_trylock(&eb->lock)) {
|
|
trace_btrfs_try_tree_read_lock(eb);
|
|
return 1;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Try-lock for write.
|
|
*
|
|
* Return 1 if the rwlock has been taken, 0 otherwise
|
|
*/
|
|
int btrfs_try_tree_write_lock(struct extent_buffer *eb)
|
|
{
|
|
if (down_write_trylock(&eb->lock)) {
|
|
eb->lock_owner = current->pid;
|
|
trace_btrfs_try_tree_write_lock(eb);
|
|
return 1;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Release read lock.
|
|
*/
|
|
void btrfs_tree_read_unlock(struct extent_buffer *eb)
|
|
{
|
|
trace_btrfs_tree_read_unlock(eb);
|
|
up_read(&eb->lock);
|
|
}
|
|
|
|
/*
|
|
* __btrfs_tree_lock - lock eb for write
|
|
* @eb: the eb to lock
|
|
* @nest: the nesting to use for the lock
|
|
*
|
|
* Returns with the eb->lock write locked.
|
|
*/
|
|
void __btrfs_tree_lock(struct extent_buffer *eb, enum btrfs_lock_nesting nest)
|
|
__acquires(&eb->lock)
|
|
{
|
|
u64 start_ns = 0;
|
|
|
|
if (trace_btrfs_tree_lock_enabled())
|
|
start_ns = ktime_get_ns();
|
|
|
|
down_write_nested(&eb->lock, nest);
|
|
eb->lock_owner = current->pid;
|
|
trace_btrfs_tree_lock(eb, start_ns);
|
|
}
|
|
|
|
void btrfs_tree_lock(struct extent_buffer *eb)
|
|
{
|
|
__btrfs_tree_lock(eb, BTRFS_NESTING_NORMAL);
|
|
}
|
|
|
|
/*
|
|
* Release the write lock.
|
|
*/
|
|
void btrfs_tree_unlock(struct extent_buffer *eb)
|
|
{
|
|
trace_btrfs_tree_unlock(eb);
|
|
eb->lock_owner = 0;
|
|
up_write(&eb->lock);
|
|
}
|
|
|
|
/*
|
|
* This releases any locks held in the path starting at level and going all the
|
|
* way up to the root.
|
|
*
|
|
* btrfs_search_slot will keep the lock held on higher nodes in a few corner
|
|
* cases, such as COW of the block at slot zero in the node. This ignores
|
|
* those rules, and it should only be called when there are no more updates to
|
|
* be done higher up in the tree.
|
|
*/
|
|
void btrfs_unlock_up_safe(struct btrfs_path *path, int level)
|
|
{
|
|
int i;
|
|
|
|
if (path->keep_locks)
|
|
return;
|
|
|
|
for (i = level; i < BTRFS_MAX_LEVEL; i++) {
|
|
if (!path->nodes[i])
|
|
continue;
|
|
if (!path->locks[i])
|
|
continue;
|
|
btrfs_tree_unlock_rw(path->nodes[i], path->locks[i]);
|
|
path->locks[i] = 0;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Loop around taking references on and locking the root node of the tree until
|
|
* we end up with a lock on the root node.
|
|
*
|
|
* Return: root extent buffer with write lock held
|
|
*/
|
|
struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root)
|
|
{
|
|
struct extent_buffer *eb;
|
|
|
|
while (1) {
|
|
eb = btrfs_root_node(root);
|
|
|
|
btrfs_maybe_reset_lockdep_class(root, eb);
|
|
btrfs_tree_lock(eb);
|
|
if (eb == root->node)
|
|
break;
|
|
btrfs_tree_unlock(eb);
|
|
free_extent_buffer(eb);
|
|
}
|
|
return eb;
|
|
}
|
|
|
|
/*
|
|
* Loop around taking references on and locking the root node of the tree until
|
|
* we end up with a lock on the root node.
|
|
*
|
|
* Return: root extent buffer with read lock held
|
|
*/
|
|
struct extent_buffer *btrfs_read_lock_root_node(struct btrfs_root *root)
|
|
{
|
|
struct extent_buffer *eb;
|
|
|
|
while (1) {
|
|
eb = btrfs_root_node(root);
|
|
|
|
btrfs_maybe_reset_lockdep_class(root, eb);
|
|
btrfs_tree_read_lock(eb);
|
|
if (eb == root->node)
|
|
break;
|
|
btrfs_tree_read_unlock(eb);
|
|
free_extent_buffer(eb);
|
|
}
|
|
return eb;
|
|
}
|
|
|
|
/*
|
|
* DREW locks
|
|
* ==========
|
|
*
|
|
* DREW stands for double-reader-writer-exclusion lock. It's used in situation
|
|
* where you want to provide A-B exclusion but not AA or BB.
|
|
*
|
|
* Currently implementation gives more priority to reader. If a reader and a
|
|
* writer both race to acquire their respective sides of the lock the writer
|
|
* would yield its lock as soon as it detects a concurrent reader. Additionally
|
|
* if there are pending readers no new writers would be allowed to come in and
|
|
* acquire the lock.
|
|
*/
|
|
|
|
int btrfs_drew_lock_init(struct btrfs_drew_lock *lock)
|
|
{
|
|
int ret;
|
|
|
|
ret = percpu_counter_init(&lock->writers, 0, GFP_KERNEL);
|
|
if (ret)
|
|
return ret;
|
|
|
|
atomic_set(&lock->readers, 0);
|
|
init_waitqueue_head(&lock->pending_readers);
|
|
init_waitqueue_head(&lock->pending_writers);
|
|
|
|
return 0;
|
|
}
|
|
|
|
void btrfs_drew_lock_destroy(struct btrfs_drew_lock *lock)
|
|
{
|
|
percpu_counter_destroy(&lock->writers);
|
|
}
|
|
|
|
/* Return true if acquisition is successful, false otherwise */
|
|
bool btrfs_drew_try_write_lock(struct btrfs_drew_lock *lock)
|
|
{
|
|
if (atomic_read(&lock->readers))
|
|
return false;
|
|
|
|
percpu_counter_inc(&lock->writers);
|
|
|
|
/* Ensure writers count is updated before we check for pending readers */
|
|
smp_mb();
|
|
if (atomic_read(&lock->readers)) {
|
|
btrfs_drew_write_unlock(lock);
|
|
return false;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
void btrfs_drew_write_lock(struct btrfs_drew_lock *lock)
|
|
{
|
|
while (true) {
|
|
if (btrfs_drew_try_write_lock(lock))
|
|
return;
|
|
wait_event(lock->pending_writers, !atomic_read(&lock->readers));
|
|
}
|
|
}
|
|
|
|
void btrfs_drew_write_unlock(struct btrfs_drew_lock *lock)
|
|
{
|
|
percpu_counter_dec(&lock->writers);
|
|
cond_wake_up(&lock->pending_readers);
|
|
}
|
|
|
|
void btrfs_drew_read_lock(struct btrfs_drew_lock *lock)
|
|
{
|
|
atomic_inc(&lock->readers);
|
|
|
|
/*
|
|
* Ensure the pending reader count is perceieved BEFORE this reader
|
|
* goes to sleep in case of active writers. This guarantees new writers
|
|
* won't be allowed and that the current reader will be woken up when
|
|
* the last active writer finishes its jobs.
|
|
*/
|
|
smp_mb__after_atomic();
|
|
|
|
wait_event(lock->pending_readers,
|
|
percpu_counter_sum(&lock->writers) == 0);
|
|
}
|
|
|
|
void btrfs_drew_read_unlock(struct btrfs_drew_lock *lock)
|
|
{
|
|
/*
|
|
* atomic_dec_and_test implies a full barrier, so woken up writers
|
|
* are guaranteed to see the decrement
|
|
*/
|
|
if (atomic_dec_and_test(&lock->readers))
|
|
wake_up(&lock->pending_writers);
|
|
}
|