Re: [PATCH] ocfs2: remove debugfs before shutting down recovery

On Mon 25 May 2026 at 15:06, Zhang Cen <[email protected]> 
wrote:

> ocfs2_osb_debug_open() builds the per-mount fs_state snapshot by 
> calling
> ocfs2_osb_dump(), which reads osb->recovery_map. During normal 
> unmount,
> ocfs2_dismount_volume() currently calls ocfs2_recovery_exit() 
> before it
> removes osb->osb_debug_root, so a concurrent fs_state open can 
> still
> enter ocfs2_osb_dump() after the recovery map has been freed.
>
> operations, so moving it ahead of ocfs2_recovery_exit() closes 
> the
>
"operations,"? Seems one line is missing.

> post-free/pre-remove window without changing the recovery-state 
> logic.
> This also makes the normal unmount path match the existing 
> mount-error
> state.
>
> The buggy scenario involves two paths, with each column showing 
> the
> order within that path:
>
> 1. Open the per-mount fs_state file      1. 
> ocfs2_dismount_volume() starts
> 2. ocfs2_osb_debug_open() calls          2. 
> ocfs2_recovery_exit() frees
>    ocfs2_osb_dump()                         osb->recovery_map
>    osb->recovery_map                        runs later
>
> Validation reproduced this kernel report:
>
Please paste the link of report or steps to reproduce.

The codes LGTM.

--
Su

> KASAN slab-use-after-free in ocfs2_osb_debug_open+0x478/0xaa0
> RIP: 0033:0x7f65fc97a001
> The buggy address belongs to the object at ffff8881049c3da0 
> which belongs
> to the cache kmalloc-8 of size 8
> The buggy address is located 0 bytes inside of freed 8-byte 
> region
> [ffff8881049c3da0, ffff8881049c3da8)
> Read of size 4
> Call trace:
>   dump_stack_lvl+0x66/0xa0 (?:?)
>   print_report+0xd0/0x630 (?:?)
>   ocfs2_osb_debug_open+0x478/0xaa0 (fs/ocfs2/super.c:343)
>   srso_alias_return_thunk+0x5/0xfbef5 (?:?)
>   __virt_addr_valid+0x188/0x2f0 (?:?)
>   kasan_report+0xe4/0x120 (?:?)
>   full_proxy_open_regular+0x113/0x170 (?:?)
>   do_dentry_open+0x233/0x7f0 (?:?)
>   vfs_open+0x5a/0x1b0 (?:?)
>   security_inode_permission+0x19/0x60 (?:?)
>   path_openat+0x679/0x1540 (?:?)
>   kmem_cache_alloc_noprof+0x1ea/0x5f0 (?:?)
>   do_getname+0x2e/0x1d0 (?:?)
>   do_sys_openat2+0xa4/0x150 (?:?)
>   __x64_sys_openat+0xd0/0x140 (?:?)
>   do_syscall_64+0x10c/0x640 (arch/x86/entry/syscall_64.c:87)
>   entry_SYSCALL_64_after_hwframe+0x77/0x7f (?:?)
>   do_file_open+0x190/0x2a0 (?:?)
>   __lock_acquire+0x42f/0x1a60 (?:?)
>   _raw_spin_unlock+0x23/0x40 (?:?)
>   alloc_fd+0x210/0x350 (?:?)
>   do_sys_openat2+0xce/0x150 (?:?)
>   irqentry_exit+0xac/0x6e0 (?:?)
> Freed by task stack:
>   kasan_save_stack+0x33/0x60 (?:?)
>   kasan_save_track+0x14/0x30 (?:?)
>   kasan_save_free_info+0x3b/0x60 (?:?)
>   __kasan_slab_free+0x5f/0x80 (?:?)
>   kfree+0x30f/0x580 (?:?)
>   ocfs2_dismount_volume+0x168/0x560 (fs/ocfs2/super.c:1868)
>   generic_shutdown_super+0xc3/0x220 (fs/ocfs2/super.c:?)
>   kill_block_super+0x29/0x60 (fs/ocfs2/super.c:?)
>   deactivate_locked_super+0x66/0xe0 (fs/ocfs2/super.c:?)
>   cleanup_mnt+0x13d/0x210 (?:?)
>   task_work_run+0xfa/0x170 (?:?)
>   exit_to_user_mode_loop+0xd6/0x430 (?:?)
>   do_syscall_64+0x3cb/0x640 (arch/x86/entry/syscall_64.c:87)
>   entry_SYSCALL_64_after_hwframe+0x77/0x7f (?:?)
>
> Fixes: 5e7a3ed9f1a6 ("ocfs2: further debugfs cleanups")
> Assisted-by: Codex:gpt-5.5
> Signed-off-by: Zhang Cen <[email protected]>
> ---
>  fs/ocfs2/super.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
> index b875f01c9756..357f210b6a43 100644
> --- a/fs/ocfs2/super.c
> +++ b/fs/ocfs2/super.c
> @@ -1826,6 +1826,9 @@ static void ocfs2_dismount_volume(struct 
> super_block *sb, int mnt_err)
>
>  	ocfs2_truncate_log_shutdown(osb);
>
> +	ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
> +	debugfs_remove_recursive(osb->osb_debug_root);
> +
>  	/* This will disable recovery and flush any recovery work. */
>  	ocfs2_recovery_exit(osb);
>
> @@ -1865,9 +1868,6 @@ static void ocfs2_dismount_volume(struct 
> super_block *sb, int mnt_err)
>
>  	ocfs2_dlm_shutdown(osb, hangup_needed);
>
> -	ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
> -	debugfs_remove_recursive(osb->osb_debug_root);
> -
>  	if (hangup_needed)
>  		ocfs2_cluster_hangup(osb->uuid_str, 
>  strlen(osb->uuid_str));