Re: [Intel-gfx] [RFT i-g-t 1/2] tests/gem_shrink: Background, direct and OOM shrinker plus userptr tests

[Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>]

On 07/12/2018 14:06, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2018-12-07 14:04:05)
>> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>>
>> ...
>>
>> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
>> ---
>>   lib/igt_core.c          |  18 +++
>>   lib/igt_core.h          |   1 +
>>   tests/i915/gem_shrink.c | 299 ++++++++++++++++++++++++++++++++++++++++
>>   3 files changed, 318 insertions(+)
>>
>> diff --git a/lib/igt_core.c b/lib/igt_core.c
>> index 64883d6402af..d8fa0c83e279 100644
>> --- a/lib/igt_core.c
>> +++ b/lib/igt_core.c
>> @@ -1680,6 +1680,24 @@ void igt_stop_helper(struct igt_helper_process *proc)
>>          assert(helper_was_alive(proc, status));
>>   }
>>   
>> +/**
>> + * igt_try_stop_helper:
>> + * @proc: #igt_helper_process structure
>> + *
>> + * Terminates a helper process if it is still running.
>> + */
>> +void igt_try_stop_helper(struct igt_helper_process *proc)
> 
> General thoughtless comment about try_func is that usually report a
> bool.

Okay, another TODO item. First I wanted to call them __igt_stop_helper 
which would have avoided it. :)

But in general no need to pay too much attention for now, I am only 
using it for CI access.

Hitting the nested lock path is proving to be tricky, locally it only 
manages a handful of times per run. And since it is nested we cannot 
count on lockdep to stop things. Userptr half also regressed in so it's 
not triggering much shrinking any longer. :I

So perhaps strength in (CI) numbers shows something new. Or not..

Regards,

Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx