[Bug 216198] New: uvd page faults on Radeon

[bugzilla-daemon@kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=216198

            Bug ID: 216198
           Summary: uvd page faults on Radeon
           Product: Drivers
           Version: 2.5
    Kernel Version: 5.15
          Hardware: AMD
                OS: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: Video(DRI - non Intel)
          Assignee: drivers_video-dri@kernel-bugs.osdl.org
          Reporter: nutodafozo@freeweb.email
        Regression: No

Created attachment 301322
  --> https://bugzilla.kernel.org/attachment.cgi?id=301322&action=edit
oops

After upgrading ubuntu from 20.04 (had stock kernels 5.4, 5.8 and 5.13) to
22.04 with 5.15, I've started to experience page faults when hw acceleration on
Radeon RX550 is working in Firefox.


2 times already in 5 days, same stacktraces / addresses of faulting instruction
and the same zero $rbx register.
I've had the same configuration for over a year, never had such oopses before
on previous kernels.


> [X77.269660] BUG: unable to handle page fault for address: fffffffffffffff8
> 
> ffffffff8111d4f0 <swake_up_locked>:
> ffffffff8111d4f0:       e8 fb 90 f6 ff          call   ffffffff810865f0
> <__fentry__>
> ffffffff8111d4f5:       48 8b 57 08             mov    0x8(%rdi),%rdx
> ffffffff8111d4f9:       48 8d 47 08             lea    0x8(%rdi),%rax
> ffffffff8111d4fd:       48 39 c2                cmp    %rax,%rdx
> ffffffff8111d500:       74 2d                   je     ffffffff8111d52f
> <swake_up_locked+0x3f>
> ffffffff8111d502:       55                      push   %rbp
> ffffffff8111d503:       48 89 e5                mov    %rsp,%rbp
> ffffffff8111d506:       53                      push   %rbx
> ffffffff8111d507:       48 8b 5f 08             mov    0x8(%rdi),%rbx
> ffffffff8111d50b:       48 8b 7b f8             mov    -0x8(%rbx),%rdi  
> !<<<<<<<<<
> ffffffff8111d50f:       e8 4c d6 fd ff          call   ffffffff810fab60
> <wake_up_process>
> 
> void swake_up_locked(struct swait_queue_head *q)
> {
>       struct swait_queue *curr;
>       if (list_empty(&q->task_list))
>               return;
>       curr = list_first_entry(&q->task_list, typeof(*curr), task_list);
>       wake_up_process(curr->task);
>       list_del_init(&curr->task_list);
> }

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching the assignee of the bug.