[PATCH v3] hugetlbfs: move hugetlbfs_free_inode tracepoint to ->destroy_inode()

[PATCH v3] hugetlbfs: move hugetlbfs_free_inode tracepoint to ->destroy_inode()

[Joas Antonio dos Santos via B4 Relay]

From: Joas Antonio dos Santos <joasantonio108@gmail.com>

hugetlbfs_free_inode() calls trace_hugetlbfs_free_inode(), and the
hugetlbfs__inode event class records inode->i_sb->s_dev.  ->free_inode()
runs from an RCU callback (destroy_inode() queues i_callback() with
call_rcu()), and Documentation/filesystems/porting.rst says that this
callback "is *NOT* ordered wrt superblock destruction; as the matter of
fact, the superblock and all associated structures might be already
gone".  put_super() frees the superblock with call_rcu() followed by a
workqueue item (destroy_super_rcu() -> destroy_super_work()).  syzbot
hit this with the event enabled:

  BUG: KASAN: slab-use-after-free in trace_event_raw_event_hugetlbfs__inode+0x45b/0x4c0 include/trace/events/hugetlbfs.h:36
  Read of size 4 at addr ffff888038850010 by task syz-executor952/6134
   hugetlbfs_free_inode+0x15e/0x1f0 fs/hugetlbfs/inode.c:1182
   i_callback+0x46/0x80 fs/inode.c:317
   rcu_do_batch kernel/rcu/tree.c:2650 [inline]
   rcu_core+0x5f0/0x1220 kernel/rcu/tree.c:2919
  Allocated by task 6134:
   alloc_super fs/super.c:326 [inline]
   sget_fc+0x1ba/0x1f60 fs/super.c:874
   hugetlbfs_get_tree+0x1e7/0x2b0 fs/hugetlbfs/inode.c:1431
  Freed by task 855:
   kfree+0x22b/0x6d0 mm/slub.c:6923
   process_one_work+0xac7/0x1b10 kernel/workqueue.c:3495

Emit the event from hugetlbfs_destroy_inode() instead.  destroy_inode()
calls ->destroy_inode() synchronously, before it queues the RCU
callback, while the superblock is still alive; hugetlbfs_destroy_inode()
already dereferences inode->i_sb there.  ->free_inode() is left with
only the kmem_cache_free().

Since the event no longer fires from ->free_inode(), rename it to
hugetlbfs_destroy_inode so that its name matches where it is emitted,
as suggested by David and Muchun.

Found by syzbot; analysed and tested with the syzbot C reproducer on an
arm64 KASAN kernel.

Fixes: 014ad7c42a69 ("hugetlbfs: use tracepoints in hugetlbfs functions.")
Reported-by: syzbot+91e1fc35f51774c479b2@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/6ae27525.5451848c.f7707.0001.GAE@google.com/
Suggested-by: David Hildenbrand <david@kernel.org>
Suggested-by: Muchun Song <muchun.song@linux.dev>
Cc: stable@vger.kernel.org
Signed-off-by: Joas Antonio dos Santos <joasantonio108@gmail.com>
---
Andrew, this replaces
hugetlbfs-emit-hugetlbfs_free_inode-from-destroy_inode.patch in
mm-hotfixes-unstable.

Tested with the syzbot C reproducer (adjusted for the new event name)
on an arm64 KASAN kernel: no splat in 3 of 3 boots, and
hugetlbfs_destroy_inode events are recorded.

Tool-use note: the analysis and testing were done with help from an
LLM-based assistant; I reviewed the change and ran the tests myself.
---
Changes in v3:
- Rename the tracepoint to hugetlbfs_destroy_inode in the same patch,
  so its name matches where it now fires (David, Muchun).
- v2 was the resend with the Assisted-by trailer dropped; it should
  have been marked v2 in the subject, sorry about that.
- Link to v2: https://lore.kernel.org/r/20261031-b4-hugetlbfs-trace-v1-1-4ec8d02a30cc@gmail.com
---
 fs/hugetlbfs/inode.c             | 6 +++++-
 include/trace/events/hugetlbfs.h | 2 +-
 2 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
index 01c4d6f44e..095b7fceee 100644
--- a/fs/hugetlbfs/inode.c
+++ b/fs/hugetlbfs/inode.c
@@ -1179,12 +1179,16 @@ static struct inode *hugetlbfs_alloc_inode(struct super_block *sb)
 
 static void hugetlbfs_free_inode(struct inode *inode)
 {
-	trace_hugetlbfs_free_inode(inode);
 	kmem_cache_free(hugetlbfs_inode_cachep, HUGETLBFS_I(inode));
 }
 
 static void hugetlbfs_destroy_inode(struct inode *inode)
 {
+	/*
+	 * Not in ->free_inode(): that runs from an RCU callback, when the
+	 * superblock the event reads may already be freed.
+	 */
+	trace_hugetlbfs_destroy_inode(inode);
 	hugetlbfs_inc_free_inodes(HUGETLBFS_SB(inode->i_sb));
 }
 
diff --git a/include/trace/events/hugetlbfs.h b/include/trace/events/hugetlbfs.h
index 8ba72c1d4f..a7932e0db0 100644
--- a/include/trace/events/hugetlbfs.h
+++ b/include/trace/events/hugetlbfs.h
@@ -72,7 +72,7 @@ DEFINE_EVENT(hugetlbfs__inode, hugetlbfs_evict_inode,
 	TP_ARGS(inode)
 );
 
-DEFINE_EVENT(hugetlbfs__inode, hugetlbfs_free_inode,
+DEFINE_EVENT(hugetlbfs__inode, hugetlbfs_destroy_inode,
 
 	TP_PROTO(struct inode *inode),
 

---
base-commit: c339f4010dea8b0d754f84ace70cd1b089878ef8
change-id: 20261031-b4-hugetlbfs-trace-341b338ab57b

Best regards,
-- 
Joas Antonio dos Santos <joasantonio108@gmail.com>

Re: [PATCH v3] hugetlbfs: move hugetlbfs_free_inode tracepoint to ->destroy_inode()

[Muchun Song]

> On Nov 1, 2026, at 21:48, Joas Antonio dos Santos via B4 Relay <devnull+joasantonio108.gmail.com@kernel.org> wrote:
> 
> From: Joas Antonio dos Santos <joasantonio108@gmail.com>
> 
> hugetlbfs_free_inode() calls trace_hugetlbfs_free_inode(), and the
> hugetlbfs__inode event class records inode->i_sb->s_dev.  ->free_inode()
> runs from an RCU callback (destroy_inode() queues i_callback() with
> call_rcu()), and Documentation/filesystems/porting.rst says that this
> callback "is *NOT* ordered wrt superblock destruction; as the matter of
> fact, the superblock and all associated structures might be already
> gone".  put_super() frees the superblock with call_rcu() followed by a
> workqueue item (destroy_super_rcu() -> destroy_super_work()).  syzbot
> hit this with the event enabled:
> 
>  BUG: KASAN: slab-use-after-free in trace_event_raw_event_hugetlbfs__inode+0x45b/0x4c0 include/trace/events/hugetlbfs.h:36
>  Read of size 4 at addr ffff888038850010 by task syz-executor952/6134
>   hugetlbfs_free_inode+0x15e/0x1f0 fs/hugetlbfs/inode.c:1182
>   i_callback+0x46/0x80 fs/inode.c:317
>   rcu_do_batch kernel/rcu/tree.c:2650 [inline]
>   rcu_core+0x5f0/0x1220 kernel/rcu/tree.c:2919
>  Allocated by task 6134:
>   alloc_super fs/super.c:326 [inline]
>   sget_fc+0x1ba/0x1f60 fs/super.c:874
>   hugetlbfs_get_tree+0x1e7/0x2b0 fs/hugetlbfs/inode.c:1431
>  Freed by task 855:
>   kfree+0x22b/0x6d0 mm/slub.c:6923
>   process_one_work+0xac7/0x1b10 kernel/workqueue.c:3495
> 
> Emit the event from hugetlbfs_destroy_inode() instead.  destroy_inode()
> calls ->destroy_inode() synchronously, before it queues the RCU
> callback, while the superblock is still alive; hugetlbfs_destroy_inode()
> already dereferences inode->i_sb there.  ->free_inode() is left with
> only the kmem_cache_free().
> 
> Since the event no longer fires from ->free_inode(), rename it to
> hugetlbfs_destroy_inode so that its name matches where it is emitted,
> as suggested by David and Muchun.
> 
> Found by syzbot; analysed and tested with the syzbot C reproducer on an
> arm64 KASAN kernel.
> 
> Fixes: 014ad7c42a69 ("hugetlbfs: use tracepoints in hugetlbfs functions.")
> Reported-by: syzbot+91e1fc35f51774c479b2@syzkaller.appspotmail.com
> Closes: https://lore.kernel.org/all/6ae27525.5451848c.f7707.0001.GAE@google.com/
> Suggested-by: David Hildenbrand <david@kernel.org>
> Suggested-by: Muchun Song <muchun.song@linux.dev>
> Cc: stable@vger.kernel.org
> Signed-off-by: Joas Antonio dos Santos <joasantonio108@gmail.com>
> ---
> Andrew, this replaces
> hugetlbfs-emit-hugetlbfs_free_inode-from-destroy_inode.patch in
> mm-hotfixes-unstable.
> 
> Tested with the syzbot C reproducer (adjusted for the new event name)
> on an arm64 KASAN kernel: no splat in 3 of 3 boots, and
> hugetlbfs_destroy_inode events are recorded.
> 
> Tool-use note: the analysis and testing were done with help from an
> LLM-based assistant; I reviewed the change and ran the tests myself.
> ---
> Changes in v3:
> - Rename the tracepoint to hugetlbfs_destroy_inode in the same patch,
>  so its name matches where it now fires (David, Muchun).
> - v2 was the resend with the Assisted-by trailer dropped; it should
>  have been marked v2 in the subject, sorry about that.
> - Link to v2: https://lore.kernel.org/r/20261031-b4-hugetlbfs-trace-v1-1-4ec8d02a30cc@gmail.com
> ---
> fs/hugetlbfs/inode.c             | 6 +++++-
> include/trace/events/hugetlbfs.h | 2 +-
> 2 files changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
> index 01c4d6f44e..095b7fceee 100644
> --- a/fs/hugetlbfs/inode.c
> +++ b/fs/hugetlbfs/inode.c
> @@ -1179,12 +1179,16 @@ static struct inode *hugetlbfs_alloc_inode(struct super_block *sb)
> 
> static void hugetlbfs_free_inode(struct inode *inode)
> {
> -	trace_hugetlbfs_free_inode(inode);
> 	kmem_cache_free(hugetlbfs_inode_cachep, HUGETLBFS_I(inode));
> }
> 
> static void hugetlbfs_destroy_inode(struct inode *inode)
> {
> +	/*
> +	 * Not in ->free_inode(): that runs from an RCU callback, when the
> +	 * superblock the event reads may already be freed.
> +	 */

Since the trace point name has already been changed, the comment here
doesn't seem to serve much purpose anymore, as there isn't any ambiguity.

With the comment removed:

Reviewed-by: Muchun Song <muchun.song@linux.dev>

Thanks.

> +	trace_hugetlbfs_destroy_inode(inode);
> 	hugetlbfs_inc_free_inodes(HUGETLBFS_SB(inode->i_sb));
> }
> 
> diff --git a/include/trace/events/hugetlbfs.h b/include/trace/events/hugetlbfs.h
> index 8ba72c1d4f..a7932e0db0 100644
> --- a/include/trace/events/hugetlbfs.h
> +++ b/include/trace/events/hugetlbfs.h
> @@ -72,7 +72,7 @@ DEFINE_EVENT(hugetlbfs__inode, hugetlbfs_evict_inode,
> 	TP_ARGS(inode)
> );
> 
> -DEFINE_EVENT(hugetlbfs__inode, hugetlbfs_free_inode,
> +DEFINE_EVENT(hugetlbfs__inode, hugetlbfs_destroy_inode,
> 
> 	TP_PROTO(struct inode *inode),
> 
> 
> ---
> base-commit: c339f4010dea8b0d754f84ace70cd1b089878ef8
> change-id: 20261031-b4-hugetlbfs-trace-341b338ab57b
> 
> Best regards,
> -- 
> Joas Antonio dos Santos <joasantonio108@gmail.com>
> 
> 

Re: [PATCH v3] hugetlbfs: move hugetlbfs_free_inode tracepoint to ->destroy_inode()

[Andrew Morton]

On Mon, 2 Nov 2026 10:24:38 +0800 Muchun Song <muchun.song@linux.dev> wrote:

> > static void hugetlbfs_destroy_inode(struct inode *inode)
> > {
> > +	/*
> > +	 * Not in ->free_inode(): that runs from an RCU callback, when the
> > +	 * superblock the event reads may already be freed.
> > +	 */
> 
> Since the trace point name has already been changed, the comment here
> doesn't seem to serve much purpose anymore, as there isn't any ambiguity.
> 
> With the comment removed:
> 
> Reviewed-by: Muchun Song <muchun.song@linux.dev>

Thanks.  I've added a note-to-self to track this, but I've left the
patch in-place as your observation doesn't invalidate ongoing testing.

I always do this!  But this doesn't appear to to be widely known :(. 
My fault, really.

My notes-to-self are mastered in
https://git.kernel.org/pub/scm/linux/kernel/git/akpm/25-new.git/tree/pc/devel-series,
so the overly-curious can always look up status in there.

If a patch has a note, I won't progress it until that note has been
cleared away.  This often involves sending annoying private emails to
people, or fixing stuff myself.