[Intel-gfx] [PATCH 25/33] drm/i915/guc: Support request cancellation

[Tvrtko Ursulin]

On 05/10/2021 08:06, Sebastian Andrzej Siewior wrote:
> On 2021-07-27 12:15:59 [-0700], Daniele Ceraolo Spurio wrote:
>> On 7/26/2021 5:23 PM, Matthew Brost wrote:
>>> This adds GuC backend support for i915_request_cancel(), which in turn
>>> makes CONFIG_DRM_I915_REQUEST_TIMEOUT work.
>>>
>> Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio at intel.com>
> 
> I have a few instances of ODEBUG warnings since this commit
>     62eaf0ae217d4 ("drm/i915/guc: Support request cancellation")
> 
> like:
> 
> | ------------[ cut here ]------------
> | ODEBUG: init destroyed (active state 0) object type: i915_sw_fence hint: sw_fence_dummy_notify+0x0/0x10
> | WARNING: CPU: 0 PID: 987 at lib/debugobjects.c:505 debug_print_object+0x6e/0x90
> | Modules linked in:
> | CPU: 0 PID: 987 Comm: Xorg Not tainted 5.15.0-rc4+ #67
> | Hardware name: To Be Filled By O.E.M. To Be Filled By O.E.M./Z68 Pro3-M, BIOS P2.10 04/24/2012
> | RIP: 0010:debug_print_object+0x6e/0x90
> …
> | Call Trace:
> |  i915_sw_fence_reinit+0x10/0x40
> |  intel_context_init+0x185/0x1e0
> |  intel_context_create+0x2e/0x100
> |  default_engines+0x9d/0x120
> |  i915_gem_create_context+0x40a/0x5d0
> |  ? trace_kmalloc+0x29/0xd0
> |  ? kmem_cache_alloc_trace+0xdd/0x190
> |  i915_gem_context_open+0x140/0x1c0
> |  i915_gem_open+0x70/0xa0
> |  drm_file_alloc+0x1af/0x270
> |  drm_open+0xdc/0x270
> |  drm_stub_open+0xa6/0x130
> |  chrdev_open+0xbe/0x250
> |  ? cdev_device_add+0x80/0x80
> |  do_dentry_open+0x15e/0x390
> |  path_openat+0x76b/0xa60
> |  do_filp_open+0xa4/0x150
> |  ? lock_release+0x149/0x2f0
> |  ? _raw_spin_unlock+0x24/0x40
> |  do_sys_openat2+0x92/0x160
> |  __x64_sys_openat+0x4f/0x90
> |  do_syscall_64+0x3b/0xc0
> |  entry_SYSCALL_64_after_hwframe+0x44/0xae
> | RIP: 0033:0x7f91b5cfdf07
> 
> and:
> | ODEBUG: activate destroyed (active state 0) object type: i915_sw_fence hint: sw_fence_dummy_notify+0x0/0x10
> | WARNING: CPU: 0 PID: 987 at lib/debugobjects.c:505 debug_print_object+0x6e/0x90
> |
> | Call Trace:
> |  debug_object_activate+0x174/0x200
> |  i915_sw_fence_commit+0x10/0x20
> |  intel_context_init+0x18d/0x1e0
> |  intel_context_create+0x2e/0x100
> |  default_engines+0x9d/0x120
> 
> ---
> 
> | ODEBUG: active_state destroyed (active state 0) object type: i915_sw_fence hint: sw_fence_dummy_notify+0x0/0x10
> | WARNING: CPU: 0 PID: 987 at lib/debugobjects.c:505 debug_print_object+0x6e/0x90
> | Call Trace:
> |  __i915_sw_fence_complete+0x6f/0x280
> |  intel_context_init+0x18d/0x1e0
> |  intel_context_create+0x2e/0x100
> |  default_engines+0x9d/0x120
> 
> Is this known? This is yesterday's -rc4, I first noticed it in -rc3.

Needs this fix:

commit d576b31bdece7b5034047cbe21170e948198d32f
Author: Matthew Auld <matthew.auld at intel.com>
Date:   Fri Sep 24 15:46:46 2021 +0100

     drm/i915: remember to call i915_sw_fence_fini

But in the fix we forgot to add:

Fixes: 62eaf0ae217d ("drm/i915/guc: Support request cancellation")

So not sure if it will appear on it's own. Adding Joonas and Rodrigo for 
maintainer level help.

Regards,

Tvrtko