[igt-dev] [PATCH i-g-t v3 3/4] tests/kms_frontbuffer_tracking: Convert test to use igt_display, v3.

[Maarten Lankhorst]

Op 07-03-18 om 16:26 schreef Arkadiusz Hiler:
> On Wed, Mar 07, 2018 at 05:17:37PM +0200, Arkadiusz Hiler wrote:
>> On Tue, Feb 27, 2018 at 09:52:31AM +0100, Maarten Lankhorst wrote:
>>> Straight conversion, no behavioral changes yet.
>>>
>>> Changes since v1:
>>> - Handle 2x outputs correctly.
>>> Changes since v2:
>>> - Set correct parameters in set_mode_for_params, so atomic commit for
>>>   2x outputs doesn't scale.
>>> - Fix accidental scaling in set_prim_plane_for_params.
>>>
>>> Signed-off-by: Maarten Lankhorst <maarten.lankhorst at linux.intel.com>
>> Generally the conversion looks good to me, but I can't slap r-b
>> yet because there are those new fails:
>>
>>
>> Test kms_frontbuffer_tracking:
>>         Subgroup fbc-2p-primscrn-pri-shrfb-draw-render:
>>                 pass       -> FAIL       (shard-hsw)
>>         Subgroup fbc-2p-scndscrn-pri-shrfb-draw-mmap-gtt:
>>                 pass       -> FAIL       (shard-hsw)
>>         Subgroup fbc-2p-scndscrn-shrfb-plflip-blt:
>>                 pass       -> FAIL       (shard-hsw)
>>
>>
>> There's even a bug for that (fdo#101623), but we never had hit the issue
>> with those particular 3 tests before, so they are not suppressed by the
>> cibuglog.
>>
>> I've done a CI rerun to make sure it's consistent (and it is, 2/2 :-))
>> and I've tried to figure out why we hit the issue with your changes
>> while I was doing the review, but I haven't spotted anything suspicious.
> On a second thought it may be the optimizations from the 4th patch
> making the issue more consistent.
>
> I'll send the first three patches to get a round of CI treatment.
>
Yeah, I'm also looking at it. Patch 4/4 needs enable_fbc to be switchable at runtime by touching the module parameter, which I've done now,

and also psr to be switchable at runtime in the same way. https://patchwork.freedesktop.org/series/39545/