Re: [f2fs-dev] [PATCH] f2fs: avoid false alarm of circular locking

[Chao Yu]

On 2023/8/19 8:30, Jaegeuk Kim wrote:
> ======================================================
> WARNING: possible circular locking dependency detected
> 6.5.0-rc5-syzkaller-00353-gae545c3283dc #0 Not tainted
> ------------------------------------------------------
> syz-executor273/5027 is trying to acquire lock:
> ffff888077fe1fb0 (&fi->i_sem){+.+.}-{3:3}, at: f2fs_down_write fs/f2fs/f2fs.h:2133 [inline]
> ffff888077fe1fb0 (&fi->i_sem){+.+.}-{3:3}, at: f2fs_add_inline_entry+0x300/0x6f0 fs/f2fs/inline.c:644
>
> but task is already holding lock:
> ffff888077fe07c8 (&fi->i_xattr_sem){.+.+}-{3:3}, at: f2fs_down_read fs/f2fs/f2fs.h:2108 [inline]
> ffff888077fe07c8 (&fi->i_xattr_sem){.+.+}-{3:3}, at: f2fs_add_dentry+0x92/0x230 fs/f2fs/dir.c:783
>
> which lock already depends on the new lock.
>
> the existing dependency chain (in reverse order) is:
>
> -> #1 (&fi->i_xattr_sem){.+.+}-{3:3}:
>         down_read+0x9c/0x470 kernel/locking/rwsem.c:1520
>         f2fs_down_read fs/f2fs/f2fs.h:2108 [inline]
>         f2fs_getxattr+0xb1e/0x12c0 fs/f2fs/xattr.c:532
>         __f2fs_get_acl+0x5a/0x900 fs/f2fs/acl.c:179
>         f2fs_acl_create fs/f2fs/acl.c:377 [inline]
>         f2fs_init_acl+0x15c/0xb30 fs/f2fs/acl.c:420
>         f2fs_init_inode_metadata+0x159/0x1290 fs/f2fs/dir.c:558
>         f2fs_add_regular_entry+0x79e/0xb90 fs/f2fs/dir.c:740
>         f2fs_add_dentry+0x1de/0x230 fs/f2fs/dir.c:788
>         f2fs_do_add_link+0x190/0x280 fs/f2fs/dir.c:827
>         f2fs_add_link fs/f2fs/f2fs.h:3554 [inline]
>         f2fs_mkdir+0x377/0x620 fs/f2fs/namei.c:781
>         vfs_mkdir+0x532/0x7e0 fs/namei.c:4117
>         do_mkdirat+0x2a9/0x330 fs/namei.c:4140
>         __do_sys_mkdir fs/namei.c:4160 [inline]
>         __se_sys_mkdir fs/namei.c:4158 [inline]
>         __x64_sys_mkdir+0xf2/0x140 fs/namei.c:4158
>         do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>         do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80
>         entry_SYSCALL_64_after_hwframe+0x63/0xcd
>
> -> #0 (&fi->i_sem){+.+.}-{3:3}:
>         check_prev_add kernel/locking/lockdep.c:3142 [inline]
>         check_prevs_add kernel/locking/lockdep.c:3261 [inline]
>         validate_chain kernel/locking/lockdep.c:3876 [inline]
>         __lock_acquire+0x2e3d/0x5de0 kernel/locking/lockdep.c:5144
>         lock_acquire kernel/locking/lockdep.c:5761 [inline]
>         lock_acquire+0x1ae/0x510 kernel/locking/lockdep.c:5726
>         down_write+0x93/0x200 kernel/locking/rwsem.c:1573
>         f2fs_down_write fs/f2fs/f2fs.h:2133 [inline]
>         f2fs_add_inline_entry+0x300/0x6f0 fs/f2fs/inline.c:644
>         f2fs_add_dentry+0xa6/0x230 fs/f2fs/dir.c:784
>         f2fs_do_add_link+0x190/0x280 fs/f2fs/dir.c:827
>         f2fs_add_link fs/f2fs/f2fs.h:3554 [inline]
>         f2fs_mkdir+0x377/0x620 fs/f2fs/namei.c:781
>         vfs_mkdir+0x532/0x7e0 fs/namei.c:4117
>         ovl_do_mkdir fs/overlayfs/overlayfs.h:196 [inline]
>         ovl_mkdir_real+0xb5/0x370 fs/overlayfs/dir.c:146
>         ovl_workdir_create+0x3de/0x820 fs/overlayfs/super.c:309
>         ovl_make_workdir fs/overlayfs/super.c:711 [inline]
>         ovl_get_workdir fs/overlayfs/super.c:864 [inline]
>         ovl_fill_super+0xdab/0x6180 fs/overlayfs/super.c:1400
>         vfs_get_super+0xf9/0x290 fs/super.c:1152
>         vfs_get_tree+0x88/0x350 fs/super.c:1519
>         do_new_mount fs/namespace.c:3335 [inline]
>         path_mount+0x1492/0x1ed0 fs/namespace.c:3662
>         do_mount fs/namespace.c:3675 [inline]
>         __do_sys_mount fs/namespace.c:3884 [inline]
>         __se_sys_mount fs/namespace.c:3861 [inline]
>         __x64_sys_mount+0x293/0x310 fs/namespace.c:3861
>         do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>         do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80
>         entry_SYSCALL_64_after_hwframe+0x63/0xcd
>
> other info that might help us debug this:
>
>   Possible unsafe locking scenario:
>
>         CPU0                    CPU1
>         ----                    ----
>    rlock(&fi->i_xattr_sem);
>                                 lock(&fi->i_sem);
>                                 lock(&fi->i_xattr_sem);
>    lock(&fi->i_sem);
>
> Reported-and-tested-by: syzbot+e5600587fa9cbf8e3826@xxxxxxxxxxxxxxxxxxxxxxxxx
> Fixes: 5eda1ad1aaff "f2fs: fix deadlock in i_xattr_sem and inode page lock"
> Signed-off-by: Jaegeuk Kim <jaegeuk@xxxxxxxxxx>

Reviewed-by: Chao Yu <chao@xxxxxxxxxx>

Thanks,