Thread (8 messages) 8 messages, 5 authors, 2021-03-23

Re: [igt-dev] [PATCH i-g-t] Revert "tests/kms_plane: Don't unset primary_fb when testing other planes"

From: Juha-Pekka Heikkila <hidden>
Date: 2021-03-23 08:01:19

On 23.3.2021 2.24, Lyude Paul wrote:
hm - I gave my reviewed-by: here before but taking a closer look at this, I am
seeing underruns being noted in the pixel format tests:

https://intel-gfx-ci.01.org/tree/drm-tip/index.html?testfilter=kms_plane@pixel-format&hosts=tgl%7Cicl

I haven't gotten a closer look yet, but I'm extremely doubtful this is an error
on the test's part unless those underruns were there before.
Just to note those underruns are not only problem there. I assume 
they're merely symptom of some misprogramming.

Here is just plain crc fail while running on sprite plane, ie while 
changes are in use:
https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_9879/shard-iclb7/igt@kms_plane@pixel-format-pipe-c-planes-source-clamping.html

Here is suddenly reported read from powered down well:
https://intel-gfx-ci.01.org/tree/drm-tip/IGT_6038/shard-iclb3/igt@kms_plane@pixel-format-pipe-a-planes-source-clamping.html

In pixel format test there's separate path for planar formats and all 
these problems start to show when use planar formats. I'd guess that's 
the place to look for.

/Juha-Pekka
On Mon, 2021-03-22 at 10:25 +0200, Martin Peres wrote:
quoted
The patch led to a regression in i915 results. The regression wasn't
caught pre-merge because the test was blacklisted for pre-merge by
yours truly due to its prohibitive execution time, but my creation
came back to bite me in the ass, Frakenstein-style.

It is currently unclear if the regression introduced by this change is
an i915 bug, a KMS spec violation introduced by the change, or
generally a loose-spec and just incompatible HW between NVIDIA and
Intel. While we figure this thing out, let's just revert the change
and put the cost of integration on the people proposing the
change (Lyude and I).

Sorry for the noise!

Signed-off-by: Martin Peres <redacted>
Cc: Jani Saarinen <redacted>
Cc: Lyude Paul <lyude@redhat.com>
Cc: juhapekka.heikkila@gmail.com
---
  tests/kms_plane.c | 9 ++++-----
  1 file changed, 4 insertions(+), 5 deletions(-)
diff --git a/tests/kms_plane.c b/tests/kms_plane.c
index c87983a4..298a9375 100644
--- a/tests/kms_plane.c
+++ b/tests/kms_plane.c
@@ -817,10 +817,9 @@ enum crc_set { PACKED_CRC_SET,
                MAX_CRC_SET };
  
  static bool test_format_plane(data_t *data, enum pipe pipe,
-                             igt_output_t *output, igt_plane_t *plane, igt_fb_t
*primary_fb)
+                             igt_output_t *output, igt_plane_t *plane)
  {
         struct igt_fb fb = {};
-       struct igt_fb *clear_fb = plane->type == DRM_PLANE_TYPE_PRIMARY ?
primary_fb : NULL;
         drmModeModeInfo *mode;
         uint32_t format, ref_format;
         uint64_t modifier, ref_modifier;
@@ -880,7 +879,7 @@ static bool test_format_plane(data_t *data, enum pipe pipe,
                         height = test_fb.height;
                 }
  
-               igt_plane_set_fb(plane, clear_fb);
+               igt_plane_set_fb(plane, NULL);
  
                 igt_remove_fb(data->drm_fd, &test_fb);
         }
@@ -938,7 +937,7 @@ static bool test_format_plane(data_t *data, enum pipe pipe,
  
         igt_pipe_crc_stop(data->pipe_crc);
  
-       igt_plane_set_fb(plane, clear_fb);
+       igt_plane_set_fb(plane, NULL);
         igt_remove_fb(data->drm_fd, &fb);
  
         return result;
@@ -1011,7 +1010,7 @@ test_pixel_formats(data_t *data, enum pipe pipe)
         for_each_plane_on_pipe(&data->display, pipe, plane) {
                 if (skip_plane(data, plane))
                         continue;
-               result &= test_format_plane(data, pipe, output, plane,
&primary_fb);
+               result &= test_format_plane(data, pipe, output, plane);
         }
  
         test_fini(data);
_______________________________________________
igt-dev mailing list
igt-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/igt-dev
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help