[Intel-gfx] [PATCH] drm/i915: Add process identifier to requests

[John Harrison]

On 11/02/2015 15:29, Chris Wilson wrote:
> On Wed, Feb 11, 2015 at 04:50:14PM +0200, Mika Kuoppala wrote:
>> We use the pid of the process which opened our device when
>> we track which was the culprit of the gpu hang. But as that
>> file descriptor might get inherited, we might blame the
>> wrong process when we record the error state.
>>
>> Track process identifiers in requests to always find
>> the correct offender.
>>
>> Cc: Kenneth Graunke <kenneth at whitecape.org>
>> Cc: Chris Wilson <chris at chris-wilson.co.uk>
>> Signed-off-by: Mika Kuoppala <mika.kuoppala at intel.com>
>> ---
>>   drivers/gpu/drm/i915/i915_drv.h       | 3 +++
>>   drivers/gpu/drm/i915/i915_gem.c       | 3 +++
>>   drivers/gpu/drm/i915/i915_gpu_error.c | 5 ++---
>>   3 files changed, 8 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
>> index c0b8644..9093654 100644
>> --- a/drivers/gpu/drm/i915/i915_drv.h
>> +++ b/drivers/gpu/drm/i915/i915_drv.h
>> @@ -2153,6 +2153,9 @@ struct drm_i915_gem_request {
>>   	/** file_priv list entry for this request */
>>   	struct list_head client_list;
>>   
>> +	/** process identifier submitting this request */
>> +	struct pid *pid;
>> +
>>   	uint32_t uniq;
>>   
>>   	/**
>> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
>> index c26d36c..47affaf 100644
>> --- a/drivers/gpu/drm/i915/i915_gem.c
>> +++ b/drivers/gpu/drm/i915/i915_gem.c
>> @@ -2483,6 +2483,7 @@ int __i915_add_request(struct intel_engine_cs *ring,
>>   	request->emitted_jiffies = jiffies;
>>   	list_add_tail(&request->list, &ring->request_list);
>>   	request->file_priv = NULL;
>> +	request->pid = get_pid(task_pid(current));
>>   
>>   	if (file) {
> I would suggest you only track processes for requests submitted by
> userspace. Then if there is no associated pid, we know that the kernel
> was in control (and not stuck figuring out if kworker was acting on
> behalf of the user or the kernel).
> -Chris
>

With the GPU scheduler, the actual batch buffer submission via 
i915_add_request() could be disconnected from the original IOCTL call 
into the driver. Thus the recorded pid would be the kernel worker thread 
not the user land application. Is there any particular reason why the 
pid could not be recorded when the request is first created rather than 
when it is submitted?