From 9e7cc91a6d18a4973c6d2cc104871439c9e94f3d Mon Sep 17 00:00:00 2001 From: Wang Xiaoguang Date: Mon, 1 Aug 2016 13:28:08 +0800 Subject: [PATCH] btrfs: fix fsfreeze hang caused by delayed iputs deal When running fstests generic/068, sometimes we got below deadlock: xfs_io D ffff8800331dbb20 0 6697 6693 0x00000080 ffff8800331dbb20 ffff88007acfc140 ffff880034d895c0 ffff8800331dc000 ffff880032d243e8 fffffffeffffffff ffff880032d24400 0000000000000001 ffff8800331dbb38 ffffffff816a9045 ffff880034d895c0 ffff8800331dbba8 Call Trace: [] schedule+0x35/0x80 [] rwsem_down_read_failed+0xf2/0x140 [] ? __filemap_fdatawrite_range+0xd1/0x100 [] call_rwsem_down_read_failed+0x18/0x30 [] ? btrfs_alloc_block_rsv+0x2c/0xb0 [btrfs] [] percpu_down_read+0x35/0x50 [] __sb_start_write+0x2c/0x40 [] start_transaction+0x2a5/0x4d0 [btrfs] [] btrfs_join_transaction+0x17/0x20 [btrfs] [] btrfs_evict_inode+0x3c4/0x5d0 [btrfs] [] evict+0xba/0x1a0 [] iput+0x196/0x200 [] btrfs_run_delayed_iputs+0x70/0xc0 [btrfs] [] btrfs_commit_transaction+0x928/0xa80 [btrfs] [] btrfs_freeze+0x30/0x40 [btrfs] [] freeze_super+0xf0/0x190 [] do_vfs_ioctl+0x4a5/0x5c0 [] ? do_audit_syscall_entry+0x66/0x70 [] ? syscall_trace_enter_phase1+0x11f/0x140 [] SyS_ioctl+0x79/0x90 [] do_syscall_64+0x62/0x110 [] entry_SYSCALL64_slow_path+0x25/0x25 >From this warning, freeze_super() already holds SB_FREEZE_FS, but btrfs_freeze() will call btrfs_commit_transaction() again, if btrfs_commit_transaction() finds that it has delayed iputs to handle, it'll start_transaction(), which will try to get SB_FREEZE_FS lock again, then deadlock occurs. The root cause is that in btrfs, sync_filesystem(sb) does not make sure all metadata is updated. There still maybe some codes adding delayed iputs, see below sample race window: CPU1 | CPU2 |-> freeze_super() | |-> sync_filesystem(sb); | | |-> cleaner_kthread() | | |-> btrfs_delete_unused_bgs() | | |-> btrfs_remove_chunk() | | |-> btrfs_remove_block_group() | | |-> btrfs_add_delayed_iput() | | |-> sb->s_writers.frozen = SB_FREEZE_FS; | |-> sb_wait_write(sb, SB_FREEZE_FS); | | acquire SB_FREEZE_FS lock. | | | |-> btrfs_freeze() | |-> btrfs_commit_transaction() | |-> btrfs_run_delayed_iputs() | | will handle delayed iputs, | | that means start_transaction() | | will be called, which will try | | to get SB_FREEZE_FS lock. | To fix this issue, introduce a "int fs_frozen" to record internally whether fs has been frozen. If fs has been frozen, we can not handle delayed iputs. Signed-off-by: Wang Xiaoguang Reviewed-by: David Sterba [ add comment to btrfs_freeze ] Signed-off-by: David Sterba Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 2 ++ fs/btrfs/disk-io.c | 1 + fs/btrfs/super.c | 16 ++++++++++++++++ fs/btrfs/transaction.c | 7 ++++++- 4 files changed, 25 insertions(+), 1 deletion(-) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 09cdff0d58e8..ec4154faab61 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1080,6 +1080,8 @@ struct btrfs_fs_info { struct list_head pinned_chunks; int creating_free_space_tree; + /* Used to record internally whether fs has been frozen */ + int fs_frozen; }; struct btrfs_subvolume_writers { diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 7857f64e1cae..17062223fac3 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -2631,6 +2631,7 @@ int open_ctree(struct super_block *sb, atomic_set(&fs_info->qgroup_op_seq, 0); atomic_set(&fs_info->reada_works_cnt, 0); atomic64_set(&fs_info->tree_mod_seq, 0); + fs_info->fs_frozen = 0; fs_info->sb = sb; fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; fs_info->metadata_ratio = 0; diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 864ce334f696..4071fe2bd098 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -2241,6 +2241,13 @@ static int btrfs_freeze(struct super_block *sb) struct btrfs_trans_handle *trans; struct btrfs_root *root = btrfs_sb(sb)->tree_root; + root->fs_info->fs_frozen = 1; + /* + * We don't need a barrier here, we'll wait for any transaction that + * could be in progress on other threads (and do delayed iputs that + * we want to avoid on a frozen filesystem), or do the commit + * ourselves. + */ trans = btrfs_attach_transaction_barrier(root); if (IS_ERR(trans)) { /* no transaction, don't bother */ @@ -2251,6 +2258,14 @@ static int btrfs_freeze(struct super_block *sb) return btrfs_commit_transaction(trans, root); } +static int btrfs_unfreeze(struct super_block *sb) +{ + struct btrfs_root *root = btrfs_sb(sb)->tree_root; + + root->fs_info->fs_frozen = 0; + return 0; +} + static int btrfs_show_devname(struct seq_file *m, struct dentry *root) { struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb); @@ -2299,6 +2314,7 @@ static const struct super_operations btrfs_super_ops = { .statfs = btrfs_statfs, .remount_fs = btrfs_remount, .freeze_fs = btrfs_freeze, + .unfreeze_fs = btrfs_unfreeze, }; static const struct file_operations btrfs_ctl_fops = { diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 9cca0a721961..95d41919d034 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -2278,8 +2278,13 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, kmem_cache_free(btrfs_trans_handle_cachep, trans); + /* + * If fs has been frozen, we can not handle delayed iputs, otherwise + * it'll result in deadlock about SB_FREEZE_FS. + */ if (current != root->fs_info->transaction_kthread && - current != root->fs_info->cleaner_kthread) + current != root->fs_info->cleaner_kthread && + !root->fs_info->fs_frozen) btrfs_run_delayed_iputs(root); return ret; -- 2.20.1