Re: Unexpected lockdep selftest failures

[Boqun Feng <boqun.feng@gmail.com>]

Hi,

On Fri, Nov 22, 2024 at 12:18:40PM -0800, Bart Van Assche wrote:
> Hi,
> 
> With the for-next branch of this tree: git://git.kernel.dk/linux-block
> (commit 12ab2c13ca77 ("Merge branch 'for-6.13/block' into for-next")) I
> see the following:
> 
> 
> [    0.887603][    T0]
> --------------------------------------------------------------------------
> [    0.888763][    T0]   | Wound/wait tests |
> [    0.889310][    T0]   ---------------------
> [    0.889867][    T0]                   ww api failures:  ok  |FAILED| ok
> |
> [    0.892597][    T0]                ww contexts mixing:  ok  |  ok  |
> [    0.894638][    T0]              finishing ww context:  ok  |  ok  | ok
> |  ok  |
> [    0.898020][    T0]                locking mismatches:  ok  |  ok  | ok
> |
> [    0.900689][    T0]                  EDEADLK handling:  ok  |  ok  | ok
> |  ok  |  ok  |  ok  |  ok  |  ok  |  ok  |  ok  |
> [    0.908172][    T0]            spinlock nest unlocked:  ok  |
> [    0.909484][    T0]                spinlock nest test:  ok  |
> [    0.910902][    T0] -----------------------------------------------------
> [    0.911824][    T0]                                  |block | try
> |context|
> [    0.912970][    T0] -----------------------------------------------------
> [    0.913890][    T0]                           context:  ok  |  ok  | ok
> |
> [    0.916613][    T0]                               try:  ok  |  ok  | ok
> |
> [    0.919235][    T0]                             block:  ok  |  ok  | ok
> |
> [    0.921852][    T0]                          spinlock:  ok  |  ok
> |FAILED|
> [    0.924666][    T0]
> --------------------------------------------------------------------------
> [    0.925852][    T0]   | queued read lock tests |
> [    0.926506][    T0]   ---------------------------
> [    0.927132][    T0]       hardirq read-lock/lock-read:  ok  |
> [    0.928496][    T0]       hardirq lock-read/read-lock:  ok  |
> [    0.929860][    T0]                 hardirq inversion:  ok  |
> [    0.931269][    T0]   --------------------
> [    0.931827][    T0]   | fs_reclaim tests |
> [    0.932383][    T0]   --------------------
> [    0.932932][    T0]                   correct nesting:  ok  |
> [    0.934252][    T0]                     wrong nesting:  ok  |
> [    0.935518][    T0]                 protected nesting:  ok  |
> [    0.936784][    T0]
> --------------------------------------------------------------------------
> [    0.937936][    T0]   | wait context tests |
> [    0.938516][    T0]
> --------------------------------------------------------------------------
> [    0.939661][    T0]                                  | rcu  | raw  | spin
> |mutex |
> [    0.940646][    T0]
> --------------------------------------------------------------------------
> [    0.941798][    T0]                in hardirq context:  ok  |  ok  | ok
> |  ok  |
> [    0.944946][    T0] in hardirq context (not threaded):  ok  |  ok  | ok
> |  ok  |
> [    0.948072][    T0]                in softirq context:  ok  |  ok  | ok
> |  ok  |
> [    0.951206][    T0]                    in RCU context:  ok  |  ok  | ok
> |  ok  |
> [    0.954345][    T0]                 in RCU-bh context:  ok  |  ok  | ok
> |  ok  |
> [    0.957477][    T0]              in RCU-sched context:  ok  |  ok  | ok
> |  ok  |
> [    0.960612][    T0]           in RAW_SPINLOCK context:  ok  |  ok  | ok
> |  ok  |
> [    0.963927][    T0]               in SPINLOCK context:  ok  |  ok  | ok
> |  ok  |
> [    0.967252][    T0]                  in MUTEX context:  ok  |  ok  | ok
> |  ok  |
> [    0.970571][    T0]
> --------------------------------------------------------------------------
> [    0.971702][    T0]   | local_lock tests |
> [    0.972422][    T0]   ---------------------
> [    0.972965][    T0]           local_lock inversion  2:  ok  |
> [    0.974319][    T0]           local_lock inversion 3A:  ok  |
> [    0.975708][    T0]           local_lock inversion 3B:  ok  |
> [    0.977106][    T0]       hardirq_unsafe_softirq_safe:  ok  |
> [    0.978595][    T0]
> --------------------------------------------------------------------------
> [    0.979723][    T0]   | lockdep_set_subclass() name test|
> [    0.980424][    T0]   -----------------------------------
> [    0.981123][    T0]     compare name before and after:  ok  |
> [    0.982423][    T0]
> -----------------------------------------------------------------
> [    0.983434][    T0] BUG:   2 unexpected failures (out of 395) - debugging
> disabled! |
> [    0.984441][    T0]
> -----------------------------------------------------------------
> 
> 
> Is this a known issue?
> 

Yes, these were reported at:

	https://lore.kernel.org/lkml/Zw19sMtnKdyOVQoh@boqun-archlinux/

and Thomas (Cced) send an updated version:

	https://lore.kernel.org/lkml/20241017151007.92215-1-thomas.hellstrom@linux.intel.com/

, which wasn't picked up. :(

Thomas, could you send a fix ontop? I'm happy to review and test it.
Thanks!

Regards,
Boqun

> Thanks,
> 
> Bart.