[BUG] landlock: sleeping function called from invalid context in hook_sb_delete()

[Mickaël Salaün]

Thanks for the report.

This was indeed a false positive.  The fix (previously reported by
syzkaller) is merged in -next and soon in the master branche:
https://lore.kernel.org/r/20251105212025.807549-1-mjguzik@gmail.com


On Thu, Nov 20, 2025 at 09:48:43AM +0100, Günther Noack wrote:
> Hello!
> 
> Thanks for the report!
> 
> CC-ing Mickaël, who authored that code
> 
> On Wed, Nov 12, 2025 at 10:35:17AM +0800, 许佳凯 wrote:
> > The call trace indicates that hook_sb_delete() holds s_inode_list_lock (a spinlock) while invoking operations that may eventually call iput(), which can sleep.
> > This violates the locking context expectations and triggers __might_sleep() warnings.
> > The issue seems to be related to how Landlock handles superblock cleanup during security_sb_delete().
> > 
> > 
> > I’m currently only reporting this issue to the community; the exact fix will likely need to be confirmed and implemented by the Landlock and filesystem maintainers.
> 
> This looks like a false positive to me.
> 
> There are three places where iput() is being called in hook_sb_delete,
> two of them are in places where it is *not* holding the
> s_inode_list_lock.  The one that *is* holding the s_inode_list_lock
> has the following comment:
> 
> /*
>  * At this point, we own the ihold() reference that was
>  * originally set up by get_inode_object() and the
>  * __iget() reference that we just set in this loop
>  * walk.  Therefore the following call to iput() will
>  * not sleep nor drop the inode because there is now at
>  * least two references to it.
>  */
> 
> That seems to indicate that the sleepability concern was taken into
> consideration.  iput() only sleeps if the refcount reaches zero, and
> if you can exclude that, it won't sleep.
> 
> —Günther
> 
> -- 
>