Re: [PATCH v2] ocfs2: stabilize the dentry name while tracing reads

[Joseph Qi <joseph.qi@linux.alibaba.com>]

On 10/16/26 1:36 AM, Cen Zhang wrote:
> An open file pins its dentry, but not the external allocation backing
> an individual dentry name. ocfs2_file_read_iter() passes d_name.name to
> its trace event without holding d_lock or a name snapshot. The event's
> __string field needs those bytes for both strlen() sizing and the
> subsequent copy.
> 
> With the read event enabled, a read of an open long-named file can be
> preempted after evaluating the trace argument. A successful concurrent
> long-to-short rename of the same dentry can then reclaim that name,
> provided no other name snapshot retains it:
> 
>   Reader                              Renamer
>   ocfs2_file_read_iter()
>     load old d_name.name
>     <preempted>
>                                       vfs_rename()
>                                         take_dentry_name_snapshot()
>                                         ocfs2_rename()
>                                           ocfs2_dentry_move()
>                                             d_move()/copy_name()
>                                         release_dentry_name_snapshot()
>                                       kfree_rcu() reclamation completes
>     trace callback calls strlen()
>       through the old name pointer
> 
> The pointer load precedes the rename, and sizing follows reclamation,
> so strlen() reads freed name storage. The tracepoint's SRCU section
> protects its probes, and the event callback disables preemption only
> after the caller has loaded the name pointer.
> 
> Take a dentry name snapshot before tracing and release it after the
> trace call. The snapshot copies inline names and pins external name
> storage, keeping both the string and its length stable while the event
> sizes and copies the name. This avoids holding d_lock across trace
> callbacks. Check whether the event is enabled before taking a snapshot
> to avoid adding snapshot work to reads when tracing is disabled.
> 
> KASAN report as below:
> 
>     BUG: KASAN: slab-use-after-free in strlen+0x85/0xa0
>     Read of size 1 at addr ffff88810a5b2b18 by task python3/496
> 
>     [CPU and host details omitted.]
>     Call Trace:
>      <TASK>
>      dump_stack_lvl+0x93/0xd0
>      print_report+0xce/0x630
>      ? strlen+0x85/0xa0
>      ? srso_alias_return_thunk+0x5/0xfbef5
>      ? __virt_addr_valid+0x20e/0x420
>      ? strlen+0x85/0xa0
>      kasan_report+0xe0/0x110
>      ? strlen+0x85/0xa0
>      strlen+0x85/0xa0
>      trace_event_raw_event_ocfs2__file_ops+0xf0/0x480
>      ? __pfx_trace_event_raw_event_ocfs2__file_ops+0x10/0x10
>      ? srso_alias_return_thunk+0x5/0xfbef5
>      ? pmbd_probe_hit_cookie+0xe7/0x1c0
>      ? srso_alias_return_thunk+0x5/0xfbef5
>      ? __pfx_pmbd_probe_hit_cookie+0x10/0x10
>      ocfs2_file_read_iter+0x6cd/0xb30
>      ? __pfx_ocfs2_file_read_iter+0x10/0x10
>      ? selinux_file_permission+0x11d/0x540
>      ? srso_alias_return_thunk+0x5/0xfbef5
>      ? security_file_permission+0x26/0x80
>      vfs_read+0x7d2/0xc20
>      ? __pfx_vfs_read+0x10/0x10
>      ? srso_alias_return_thunk+0x5/0xfbef5
>      ? __pfx_do_sys_openat2+0x10/0x10
>      ksys_read+0x111/0x200
>      ? __pfx_ksys_read+0x10/0x10
>      ? irqentry_exit+0xbe/0x820
>      ? srso_alias_return_thunk+0x5/0xfbef5
>      ? trace_hardirqs_on_prepare+0x11d/0x150
>      ? srso_alias_return_thunk+0x5/0xfbef5
>      do_syscall_64+0x114/0x620
>      entry_SYSCALL_64_after_hwframe+0x77/0x7f
>     [User register dump omitted.]
>      </TASK>
> 
>     Allocated by task 489:
>      kasan_save_stack+0x33/0x60
>      kasan_save_track+0x14/0x30
>      __kasan_kmalloc+0xaa/0xb0
>      __kmalloc_noprof+0x2de/0x780
>      __d_alloc+0x5fc/0x8c0
>      d_alloc_parallel+0x106/0x11d0
>      lookup_open+0x3ff/0x13e0
>      path_openat+0x95e/0x2020
>      do_file_open+0x21e/0x470
>      do_sys_openat2+0xf4/0x1b0
>      __x64_sys_openat+0x136/0x1e0
>      do_syscall_64+0x114/0x620
>      entry_SYSCALL_64_after_hwframe+0x77/0x7f
> 
>     Freed by task 0:
>      kasan_save_stack+0x33/0x60
>      kasan_save_track+0x14/0x30
>      kasan_save_free_info+0x3b/0x60
>      __kasan_slab_free+0x5f/0x80
>      __rcu_free_sheaf_prepare+0x65/0x2a0
>      rcu_free_sheaf_nobarn+0x1d/0x50
>      rcu_core+0x6be/0x1ed0
>      handle_softirqs+0x1fa/0x8d0
>      __irq_exit_rcu+0x113/0x1e0
>      irq_exit_rcu+0xe/0x20
>      sysvec_apic_timer_interrupt+0x6c/0x80
>      asm_sysvec_apic_timer_interrupt+0x1a/0x20
> 
>     The buggy address belongs to the object at ffff88810a5b2b00
>      which belongs to the cache kmalloc-rcl-192 of size 192
>     The buggy address is located 24 bytes inside of
>      freed 192-byte region [ffff88810a5b2b00, ffff88810a5b2bc0)
> 
>     [Page and shadow-memory dumps omitted.]
> 
> Fixes: 468eedde23d6 ("ocfs2: Remove mlog(0) from fs/ocfs2/file.c")
> Cc: <stable@vger.kernel.org>
> Signed-off-by: Cen Zhang <zzzccc427@gmail.com>
> Assisted-by: LLM
> 
> ---
> Changes in v2:
> - Replace d_lock around the tracepoint with a dentry name snapshot so trace
>   callbacks run without the caller holding the dentry lock.
> - Pass both the name and its length from the snapshot; keep the event
>   enabled check and release the snapshot after the trace call.
> 
> Link to v1: https://lore.kernel.org/r/pm-ocfs2-objects-candidate-0088-v3-d628f98bfcb6784a0bfb@gmail.com
> 
> diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
> index d6e977ba6565..fc357c227ad8 100644
> --- a/fs/ocfs2/file.c
> +++ b/fs/ocfs2/file.c
> @@ -2542,11 +2542,17 @@ static ssize_t ocfs2_file_read_iter(struct kiocb *iocb,
>  	int direct_io = iocb->ki_flags & IOCB_DIRECT ? 1 : 0;
>  	int nowait = iocb->ki_flags & IOCB_NOWAIT ? 1 : 0;
>  
> -	trace_ocfs2_file_read_iter(inode, filp, filp->f_path.dentry,
> -			(unsigned long long)OCFS2_I(inode)->ip_blkno,
> -			filp->f_path.dentry->d_name.len,
> -			filp->f_path.dentry->d_name.name,
> -			to->nr_segs);	/* GRRRRR */
> +	if (trace_ocfs2_file_read_iter_enabled()) {
> +		struct dentry *dentry = filp->f_path.dentry;
> +		struct name_snapshot name;
> +
> +		take_dentry_name_snapshot(&name, dentry);
> +		trace_ocfs2_file_read_iter(inode, filp, dentry,
> +					   OCFS2_I(inode)->ip_blkno,
> +					   name.name.len, name.name.name,
> +					   to->nr_segs); /* GRRRRR */
> +		release_dentry_name_snapshot(&name);
> +	}
>  
>  
>  	if (!inode) {

The change looks fine.
Six call sites feed the same ocfs2__file_ops class, which has the single
__string(d_name, d_name), so it seems we'd better send a series to fix
all of them:
  trace_ocfs2_file_open
  trace_ocfs2_file_release
  trace_ocfs2_sync_file
  trace_ocfs2_file_read_iter
  trace_ocfs2_file_write_iter
  trace_ocfs2_file_splice_read

Thanks,
Joseph