linux_dsm_epyc7002/fs/btrfs/disk-io.h
Josef Bacik 49d11bead7 btrfs: add a helper to read the tree_root commit root for backref lookup
I got the following lockdep splat with tree locks converted to rwsem
patches on btrfs/104:

  ======================================================
  WARNING: possible circular locking dependency detected
  5.9.0+ #102 Not tainted
  ------------------------------------------------------
  btrfs-cleaner/903 is trying to acquire lock:
  ffff8e7fab6ffe30 (btrfs-root-00){++++}-{3:3}, at: __btrfs_tree_read_lock+0x32/0x170

  but task is already holding lock:
  ffff8e7fab628a88 (&fs_info->commit_root_sem){++++}-{3:3}, at: btrfs_find_all_roots+0x41/0x80

  which lock already depends on the new lock.

  the existing dependency chain (in reverse order) is:

  -> #3 (&fs_info->commit_root_sem){++++}-{3:3}:
	 down_read+0x40/0x130
	 caching_thread+0x53/0x5a0
	 btrfs_work_helper+0xfa/0x520
	 process_one_work+0x238/0x540
	 worker_thread+0x55/0x3c0
	 kthread+0x13a/0x150
	 ret_from_fork+0x1f/0x30

  -> #2 (&caching_ctl->mutex){+.+.}-{3:3}:
	 __mutex_lock+0x7e/0x7b0
	 btrfs_cache_block_group+0x1e0/0x510
	 find_free_extent+0xb6e/0x12f0
	 btrfs_reserve_extent+0xb3/0x1b0
	 btrfs_alloc_tree_block+0xb1/0x330
	 alloc_tree_block_no_bg_flush+0x4f/0x60
	 __btrfs_cow_block+0x11d/0x580
	 btrfs_cow_block+0x10c/0x220
	 commit_cowonly_roots+0x47/0x2e0
	 btrfs_commit_transaction+0x595/0xbd0
	 sync_filesystem+0x74/0x90
	 generic_shutdown_super+0x22/0x100
	 kill_anon_super+0x14/0x30
	 btrfs_kill_super+0x12/0x20
	 deactivate_locked_super+0x36/0xa0
	 cleanup_mnt+0x12d/0x190
	 task_work_run+0x5c/0xa0
	 exit_to_user_mode_prepare+0x1df/0x200
	 syscall_exit_to_user_mode+0x54/0x280
	 entry_SYSCALL_64_after_hwframe+0x44/0xa9

  -> #1 (&space_info->groups_sem){++++}-{3:3}:
	 down_read+0x40/0x130
	 find_free_extent+0x2ed/0x12f0
	 btrfs_reserve_extent+0xb3/0x1b0
	 btrfs_alloc_tree_block+0xb1/0x330
	 alloc_tree_block_no_bg_flush+0x4f/0x60
	 __btrfs_cow_block+0x11d/0x580
	 btrfs_cow_block+0x10c/0x220
	 commit_cowonly_roots+0x47/0x2e0
	 btrfs_commit_transaction+0x595/0xbd0
	 sync_filesystem+0x74/0x90
	 generic_shutdown_super+0x22/0x100
	 kill_anon_super+0x14/0x30
	 btrfs_kill_super+0x12/0x20
	 deactivate_locked_super+0x36/0xa0
	 cleanup_mnt+0x12d/0x190
	 task_work_run+0x5c/0xa0
	 exit_to_user_mode_prepare+0x1df/0x200
	 syscall_exit_to_user_mode+0x54/0x280
	 entry_SYSCALL_64_after_hwframe+0x44/0xa9

  -> #0 (btrfs-root-00){++++}-{3:3}:
	 __lock_acquire+0x1167/0x2150
	 lock_acquire+0xb9/0x3d0
	 down_read_nested+0x43/0x130
	 __btrfs_tree_read_lock+0x32/0x170
	 __btrfs_read_lock_root_node+0x3a/0x50
	 btrfs_search_slot+0x614/0x9d0
	 btrfs_find_root+0x35/0x1b0
	 btrfs_read_tree_root+0x61/0x120
	 btrfs_get_root_ref+0x14b/0x600
	 find_parent_nodes+0x3e6/0x1b30
	 btrfs_find_all_roots_safe+0xb4/0x130
	 btrfs_find_all_roots+0x60/0x80
	 btrfs_qgroup_trace_extent_post+0x27/0x40
	 btrfs_add_delayed_data_ref+0x3fd/0x460
	 btrfs_free_extent+0x42/0x100
	 __btrfs_mod_ref+0x1d7/0x2f0
	 walk_up_proc+0x11c/0x400
	 walk_up_tree+0xf0/0x180
	 btrfs_drop_snapshot+0x1c7/0x780
	 btrfs_clean_one_deleted_snapshot+0xfb/0x110
	 cleaner_kthread+0xd4/0x140
	 kthread+0x13a/0x150
	 ret_from_fork+0x1f/0x30

  other info that might help us debug this:

  Chain exists of:
    btrfs-root-00 --> &caching_ctl->mutex --> &fs_info->commit_root_sem

   Possible unsafe locking scenario:

	 CPU0                    CPU1
	 ----                    ----
    lock(&fs_info->commit_root_sem);
				 lock(&caching_ctl->mutex);
				 lock(&fs_info->commit_root_sem);
    lock(btrfs-root-00);

   *** DEADLOCK ***

  3 locks held by btrfs-cleaner/903:
   #0: ffff8e7fab628838 (&fs_info->cleaner_mutex){+.+.}-{3:3}, at: cleaner_kthread+0x6e/0x140
   #1: ffff8e7faadac640 (sb_internal){.+.+}-{0:0}, at: start_transaction+0x40b/0x5c0
   #2: ffff8e7fab628a88 (&fs_info->commit_root_sem){++++}-{3:3}, at: btrfs_find_all_roots+0x41/0x80

  stack backtrace:
  CPU: 0 PID: 903 Comm: btrfs-cleaner Not tainted 5.9.0+ #102
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.13.0-2.fc32 04/01/2014
  Call Trace:
   dump_stack+0x8b/0xb0
   check_noncircular+0xcf/0xf0
   __lock_acquire+0x1167/0x2150
   ? __bfs+0x42/0x210
   lock_acquire+0xb9/0x3d0
   ? __btrfs_tree_read_lock+0x32/0x170
   down_read_nested+0x43/0x130
   ? __btrfs_tree_read_lock+0x32/0x170
   __btrfs_tree_read_lock+0x32/0x170
   __btrfs_read_lock_root_node+0x3a/0x50
   btrfs_search_slot+0x614/0x9d0
   ? find_held_lock+0x2b/0x80
   btrfs_find_root+0x35/0x1b0
   ? do_raw_spin_unlock+0x4b/0xa0
   btrfs_read_tree_root+0x61/0x120
   btrfs_get_root_ref+0x14b/0x600
   find_parent_nodes+0x3e6/0x1b30
   btrfs_find_all_roots_safe+0xb4/0x130
   btrfs_find_all_roots+0x60/0x80
   btrfs_qgroup_trace_extent_post+0x27/0x40
   btrfs_add_delayed_data_ref+0x3fd/0x460
   btrfs_free_extent+0x42/0x100
   __btrfs_mod_ref+0x1d7/0x2f0
   walk_up_proc+0x11c/0x400
   walk_up_tree+0xf0/0x180
   btrfs_drop_snapshot+0x1c7/0x780
   ? btrfs_clean_one_deleted_snapshot+0x73/0x110
   btrfs_clean_one_deleted_snapshot+0xfb/0x110
   cleaner_kthread+0xd4/0x140
   ? btrfs_alloc_root+0x50/0x50
   kthread+0x13a/0x150
   ? kthread_create_worker_on_cpu+0x40/0x40
   ret_from_fork+0x1f/0x30
  BTRFS info (device sdb): disk space caching is enabled
  BTRFS info (device sdb): has skinny extents

This happens because qgroups does a backref lookup when we create a
delayed ref.  From here it may have to look up a root from an indirect
ref, which does a normal lookup on the tree_root, which takes the read
lock on the tree_root nodes.

To fix this we need to add a variant for looking up roots that searches
the commit root of the tree_root.  Then when we do the backref search
using the commit root we are sure to not take any locks on the tree_root
nodes.  This gets rid of the lockdep splat when running btrfs/104.

Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2020-10-26 15:04:57 +01:00

151 lines
5.4 KiB
C

/* SPDX-License-Identifier: GPL-2.0 */
/*
* Copyright (C) 2007 Oracle. All rights reserved.
*/
#ifndef BTRFS_DISK_IO_H
#define BTRFS_DISK_IO_H
#define BTRFS_SUPER_INFO_OFFSET SZ_64K
#define BTRFS_SUPER_INFO_SIZE 4096
#define BTRFS_SUPER_MIRROR_MAX 3
#define BTRFS_SUPER_MIRROR_SHIFT 12
/*
* Fixed blocksize for all devices, applies to specific ways of reading
* metadata like superblock. Must meet the set_blocksize requirements.
*
* Do not change.
*/
#define BTRFS_BDEV_BLOCKSIZE (4096)
enum btrfs_wq_endio_type {
BTRFS_WQ_ENDIO_DATA,
BTRFS_WQ_ENDIO_METADATA,
BTRFS_WQ_ENDIO_FREE_SPACE,
BTRFS_WQ_ENDIO_RAID56,
};
static inline u64 btrfs_sb_offset(int mirror)
{
u64 start = SZ_16K;
if (mirror)
return start << (BTRFS_SUPER_MIRROR_SHIFT * mirror);
return BTRFS_SUPER_INFO_OFFSET;
}
struct btrfs_device;
struct btrfs_fs_devices;
void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info);
void btrfs_init_fs_info(struct btrfs_fs_info *fs_info);
int btrfs_verify_level_key(struct extent_buffer *eb, int level,
struct btrfs_key *first_key, u64 parent_transid);
struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr,
u64 parent_transid, int level,
struct btrfs_key *first_key);
void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr);
struct extent_buffer *btrfs_find_create_tree_block(
struct btrfs_fs_info *fs_info,
u64 bytenr);
void btrfs_clean_tree_block(struct extent_buffer *buf);
int __cold open_ctree(struct super_block *sb,
struct btrfs_fs_devices *fs_devices,
char *options);
void __cold close_ctree(struct btrfs_fs_info *fs_info);
int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors);
struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev);
struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev,
int copy_num);
int btrfs_commit_super(struct btrfs_fs_info *fs_info);
struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
struct btrfs_key *key);
int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
struct btrfs_root *root);
void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info);
struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
u64 objectid, bool check_ref);
struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info,
u64 objectid, dev_t anon_dev);
struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
u64 objectid);
void btrfs_free_fs_info(struct btrfs_fs_info *fs_info);
int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info);
void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info);
void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info);
void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
struct btrfs_root *root);
int btrfs_validate_metadata_buffer(struct btrfs_io_bio *io_bio, u64 phy_offset,
struct page *page, u64 start, u64 end,
int mirror);
blk_status_t btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio,
int mirror_num, unsigned long bio_flags);
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info);
#endif
/*
* This function is used to grab the root, and avoid it is freed when we
* access it. But it doesn't ensure that the tree is not dropped.
*
* If you want to ensure the whole tree is safe, you should use
* fs_info->subvol_srcu
*/
static inline struct btrfs_root *btrfs_grab_root(struct btrfs_root *root)
{
if (!root)
return NULL;
if (refcount_inc_not_zero(&root->refs))
return root;
return NULL;
}
void btrfs_put_root(struct btrfs_root *root);
void btrfs_mark_buffer_dirty(struct extent_buffer *buf);
int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
int atomic);
int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level,
struct btrfs_key *first_key);
blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
enum btrfs_wq_endio_type metadata);
blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
int mirror_num, unsigned long bio_flags,
u64 bio_offset, void *private_data,
extent_submit_bio_start_t *submit_bio_start);
blk_status_t btrfs_submit_bio_done(void *private_data, struct bio *bio,
int mirror_num);
int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info);
int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
struct btrfs_root *root);
void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *trans,
struct btrfs_fs_info *fs_info);
void btrfs_cleanup_one_transaction(struct btrfs_transaction *trans,
struct btrfs_fs_info *fs_info);
struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
u64 objectid);
int btree_lock_page_hook(struct page *page, void *data,
void (*flush_fn)(void *));
int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags);
int __init btrfs_end_io_wq_init(void);
void __cold btrfs_end_io_wq_exit(void);
#ifdef CONFIG_DEBUG_LOCK_ALLOC
void btrfs_init_lockdep(void);
void btrfs_set_buffer_lockdep_class(u64 objectid,
struct extent_buffer *eb, int level);
#else
static inline void btrfs_init_lockdep(void)
{ }
static inline void btrfs_set_buffer_lockdep_class(u64 objectid,
struct extent_buffer *eb, int level)
{
}
#endif
#endif