Re: [Intel-gfx] [PATCH 2/2] drm/i915/selftests: add igt_vma_move_to_active_unlocked

[Andrzej Hajda <andrzej.hajda@intel.com>]

On 21.10.2022 17:39, Andi Shyti wrote:
> Hi Andrzej,
> 
> [...]
> 
>> +static inline int __must_check
>> +igt_vma_move_to_active_unlocked(struct i915_vma *vma, struct i915_request *rq,
>> +				unsigned int flags)
>> +{
>> +	int err;
>> +
>> +	i915_vma_lock(vma);
>> +	err = _i915_vma_move_to_active(vma, rq, &rq->fence, flags);
>> +	i915_vma_unlock(vma);
>> +	return err;
>> +}
>> +
> 
> there are calls to i915_vma_move_to_active also outside
> selftests, why not having a i915_move_to_active_unlocked() in
> i915_vma.h?

As I said before, Chris suggested real users of this call should use 
locking explicitly.

> 
> Besides here you break also the bisect, because between patch 1
> and 2 the i915_move_to_avtive would also call
> i915_request_await_object(). Right or am I getting confused?

Hmm, looking at v2, I do not see breakage. Patch 1 moves all occurrences 
of i915_request_await_object inside i915_vma_move_to_active.
Patch 2, just replaces sequence of calls with call to new helper.

Regards
Andrzej

> 
> Andi
> 
> [...]