Re: [PATCH i-g-t v3 4/4] kms_vblank: Switch from using crtc0 statically to explicitly setting mode.

[Robert Foss <robert.foss@collabora.com>]

>> Previously crtc0 was statically used for VBLANK tests, but
>> that assumption is not valid for the VC4 platform.
>> Instead we're now explicitly setting the mode.
>>
>> Also add support for testing all connected connectors during
>> the same test.
>
> Is there any reason why this cannot be done in a subsequent patch?
>
It could be done in a separate patch, but I think it would be
jumping through hoops.
Finding the first valid connector is pretty much the same work as
iterating through all of the connected connectors.

If the consensus is separating the two, I'll happily do so,
but the code for the intermediate step would be somewhat convoluted
I think.

>> Signed-off-by: Robert Foss <robert.foss@collabora.com>
>> ---
>>   tests/kms_vblank.c | 179 +++++++++++++++++++++++++++++++++++++++++------------
>>   1 file changed, 140 insertions(+), 39 deletions(-)
>>
>> diff --git a/tests/kms_vblank.c b/tests/kms_vblank.c
>> index 40ab6fd..970fefe 100644
>> --- a/tests/kms_vblank.c
>> +++ b/tests/kms_vblank.c
>> @@ -44,6 +44,14 @@
>>
>>   IGT_TEST_DESCRIPTION("Test speed of WaitVblank.");
>>
>> +typedef struct {
>> +       int drm_fd;
>> +       igt_display_t display;
>> +       struct igt_fb primary_fb;
>> +       igt_output_t *output;
>> +       enum pipe pipe;
>> +} data_t;
>> +
>>   static double elapsed(const struct timespec *start,
>>                        const struct timespec *end,
>>                        int loop)
>> @@ -51,75 +59,164 @@ static double elapsed(const struct timespec *start,
>>          return (1e6*(end->tv_sec - start->tv_sec) + (end->tv_nsec - start->tv_nsec)/1000)/loop;
>>   }
>>
>> -static bool crtc0_active(int fd)
>> +static uint32_t crtc_id_to_flag(uint32_t crtc_id)
>
> This seems to be pipe id, not the crtc id. Maybe this belongs to the library.
>
Agreed, fixing name and moving to igt_kms in v4.

>>   {
>> -       union drm_wait_vblank vbl;
>> +       if (crtc_id == 0)
>> +               return 0;
>> +       else if (crtc_id == 1)
>> +               return 1 | _DRM_VBLANK_SECONDARY;
>> +       else
>> +               return crtc_id << 1;
>> +}
>>
>> -       memset(&vbl, 0, sizeof(vbl));
>> -       vbl.request.type = DRM_VBLANK_RELATIVE;
>> -       return drmIoctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl) == 0;
>> +static bool prepare_crtc(data_t *data, igt_output_t *output)
>> +{
>> +       drmModeModeInfo *mode;
>> +       igt_display_t *display = &data->display;
>> +       igt_plane_t *primary;
>> +
>> +       /* select the pipe we want to use */
>> +       igt_output_set_pipe(output, data->pipe);
>> +       igt_display_commit(display);
>> +
>> +       if (!output->valid) {
>> +               igt_output_set_pipe(output, PIPE_ANY);
>> +               igt_display_commit(display);
>> +               return false;
>> +       }
>> +
>> +       /* create and set the primary plane fb */
>> +       mode = igt_output_get_mode(output);
>
> I thought the plan was to set a mode, instead of relying on one being
> already set (eg. by fbcon)? Otherwise, I don't see why this cannot be
> in the library?
>
I'm not sure I understand this comment entirely. The mode (as far as I 
understand it) is set during the kmstest_set_vt_graphics_mode() call.

>> +       igt_create_color_fb(data->drm_fd, mode->hdisplay, mode->vdisplay,
>> +                           DRM_FORMAT_XRGB8888,
>> +                           LOCAL_DRM_FORMAT_MOD_NONE,
>> +                           0.0, 0.0, 0.0,
>> +                           &data->primary_fb);
>> +
>> +       primary = igt_output_get_plane(output, IGT_PLANE_PRIMARY);
>> +       igt_plane_set_fb(primary, &data->primary_fb);
>> +
>> +       igt_display_commit(display);
>> +
>> +       igt_wait_for_vblank(data->drm_fd, data->pipe);
>> +
>> +       return true;
>> +}
>> +
>> +static void cleanup_crtc(data_t *data, igt_output_t *output)
>> +{
>> +       igt_display_t *display = &data->display;
>> +       igt_plane_t *primary;
>> +
>> +       igt_remove_fb(data->drm_fd, &data->primary_fb);
>> +
>> +       primary = igt_output_get_plane(output, IGT_PLANE_PRIMARY);
>> +       igt_plane_set_fb(primary, NULL);
>> +
>> +       igt_output_set_pipe(output, PIPE_ANY);
>> +       igt_display_commit(display);
>>   }
>>
>> -static void accuracy(int fd)
>> +static void run_test(data_t *data, void (*testfunc)(data_t *, bool), bool boolean)
>
> Probably not a big deal, but I think it's more customary to have a
> bitfield in data_t that tells what behavior the test should have,
> rather than passing a callback.
>
Agreed, moving to bitfield in v4.

>> +{
>> +       igt_display_t *display = &data->display;
>> +       igt_output_t *output;
>> +       enum pipe p;
>> +       int valid_tests = 0;
>
> unsigned
>
Agreed, fixing in v4.

>> +
>> +       for_each_connected_output(display, output) {
>> +               for_each_pipe(display, p) {
>> +                       data->pipe = p;
>> +
>> +                       if (!prepare_crtc(data, output))
>> +                               continue;
>> +
>> +                       valid_tests++;
>> +
>> +                       igt_info("Beginning %s on pipe %s, connector %s\n",
>> +                                igt_subtest_name(),
>> +                                kmstest_pipe_name(data->pipe),
>> +                                igt_output_name(output));
>> +
>> +                       testfunc(data, boolean);
>> +
>> +                       igt_info("\n%s on pipe %s, connector %s: PASSED\n\n",
>> +                                igt_subtest_name(),
>> +                                kmstest_pipe_name(data->pipe),
>> +                                igt_output_name(output));
>> +
>> +                       /* cleanup what prepare_crtc() has done */
>> +                       cleanup_crtc(data, output);
>> +               }
>> +       }
>> +
>> +       igt_require_f(valid_tests, "no valid crtc/connector combinations found\n");
>> +}
>> +
>> +static void accuracy(data_t *data, bool busy)
>>   {
>>          union drm_wait_vblank vbl;
>>          unsigned long target;
>> +       uint32_t crtc_id_flag;
>>          int n;
>>
>>          memset(&vbl, 0, sizeof(vbl));
>> +       crtc_id_flag = crtc_id_to_flag(data->pipe);
>>
>> -       vbl.request.type = DRM_VBLANK_RELATIVE;
>> +       vbl.request.type = DRM_VBLANK_RELATIVE | crtc_id_flag;
>>          vbl.request.sequence = 1;
>> -       do_ioctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>> +       do_ioctl(data->drm_fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>
> Don't see much point in moving the drm_fd into the data struct, I
> think I would just make fd a global and avoid these changes.
>
Agreed, moving drm_fd out of data struct in v4.

>>          target = vbl.reply.sequence + 60;
>>          for (n = 0; n < 60; n++) {
>> -               vbl.request.type = DRM_VBLANK_RELATIVE;
>> +               vbl.request.type = DRM_VBLANK_RELATIVE | crtc_id_flag;
>>                  vbl.request.sequence = 1;
>> -               do_ioctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>> +               do_ioctl(data->drm_fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>>
>> -               vbl.request.type = DRM_VBLANK_ABSOLUTE | DRM_VBLANK_EVENT;
>> +               vbl.request.type = DRM_VBLANK_ABSOLUTE | DRM_VBLANK_EVENT | crtc_id_flag;
>>                  vbl.request.sequence = target;
>> -               do_ioctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>> +               do_ioctl(data->drm_fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>>          }
>> -       vbl.request.type = DRM_VBLANK_RELATIVE;
>> +       vbl.request.type = DRM_VBLANK_RELATIVE | crtc_id_flag;
>>          vbl.request.sequence = 0;
>> -       do_ioctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>> +       do_ioctl(data->drm_fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>>          igt_assert_eq(vbl.reply.sequence, target);
>>
>>          for (n = 0; n < 60; n++) {
>>                  struct drm_event_vblank ev;
>> -               igt_assert_eq(read(fd, &ev, sizeof(ev)), sizeof(ev));
>> +               igt_assert_eq(read(data->drm_fd, &ev, sizeof(ev)), sizeof(ev));
>>                  igt_assert_eq(ev.sequence, target);
>>          }
>>   }
>>
>> -static void vblank_query(int fd, bool busy)
>> +static void vblank_query(data_t *data, bool busy)
>>   {
>>          union drm_wait_vblank vbl;
>>          struct timespec start, end;
>>          unsigned long sq, count = 0;
>>          struct drm_event_vblank buf;
>> +       uint32_t crtc_id_flag;
>>
>>          memset(&vbl, 0, sizeof(vbl));
>> +       crtc_id_flag = crtc_id_to_flag(data->pipe);
>>
>>          if (busy) {
>> -               vbl.request.type = DRM_VBLANK_RELATIVE | DRM_VBLANK_EVENT;
>> +               vbl.request.type = DRM_VBLANK_RELATIVE | DRM_VBLANK_EVENT | crtc_id_flag;
>>                  vbl.request.sequence = 72;
>> -               do_ioctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>> +               do_ioctl(data->drm_fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>>          }
>>
>> -       vbl.request.type = DRM_VBLANK_RELATIVE;
>> +       vbl.request.type = DRM_VBLANK_RELATIVE | crtc_id_flag;
>>          vbl.request.sequence = 0;
>> -       do_ioctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>> +       do_ioctl(data->drm_fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>>
>>          sq = vbl.reply.sequence;
>>
>>          clock_gettime(CLOCK_MONOTONIC, &start);
>>          do {
>> -               vbl.request.type = DRM_VBLANK_RELATIVE;
>> +               vbl.request.type = DRM_VBLANK_RELATIVE | crtc_id_flag;
>>                  vbl.request.sequence = 0;
>> -               do_ioctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>> +               do_ioctl(data->drm_fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>>                  count++;
>>          } while ((vbl.reply.sequence - sq) <= 60);
>>          clock_gettime(CLOCK_MONOTONIC, &end);
>> @@ -128,35 +225,37 @@ static void vblank_query(int fd, bool busy)
>>                   busy ? "busy" : "idle", elapsed(&start, &end, count));
>>
>>          if (busy)
>> -               igt_assert_eq(read(fd, &buf, sizeof(buf)), sizeof(buf));
>> +               igt_assert_eq(read(data->drm_fd, &buf, sizeof(buf)), sizeof(buf));
>>   }
>>
>> -static void vblank_wait(int fd, bool busy)
>> +static void vblank_wait(data_t *data, bool busy)
>>   {
>>          union drm_wait_vblank vbl;
>>          struct timespec start, end;
>>          unsigned long sq, count = 0;
>>          struct drm_event_vblank buf;
>> +       uint32_t crtc_id_flag;
>>
>>          memset(&vbl, 0, sizeof(vbl));
>> +       crtc_id_flag = crtc_id_to_flag(data->pipe);
>>
>>          if (busy) {
>> -               vbl.request.type = DRM_VBLANK_RELATIVE | DRM_VBLANK_EVENT;
>> +               vbl.request.type |= DRM_VBLANK_RELATIVE | DRM_VBLANK_EVENT | crtc_id_flag;
>
> I would personally just add such a line afterwards, find it more
> readable (here and elsewhere):
>
> vbl.request.type |= crtc_id_flag;
>
Agreed, changing in v4.

>>                  vbl.request.sequence = 72;
>> -               do_ioctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>> +               do_ioctl(data->drm_fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>>          }
>>
>> -       vbl.request.type = DRM_VBLANK_RELATIVE;
>> +       vbl.request.type |= DRM_VBLANK_RELATIVE | crtc_id_flag;
>>          vbl.request.sequence = 0;
>> -       do_ioctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>> +       do_ioctl(data->drm_fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>>
>>          sq = vbl.reply.sequence;
>>
>>          clock_gettime(CLOCK_MONOTONIC, &start);
>>          do {
>> -               vbl.request.type = DRM_VBLANK_RELATIVE;
>> +               vbl.request.type = DRM_VBLANK_RELATIVE | crtc_id_flag;
>>                  vbl.request.sequence = 1;
>> -               do_ioctl(fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>> +               do_ioctl(data->drm_fd, DRM_IOCTL_WAIT_VBLANK, &vbl);
>>                  count++;
>>          } while ((vbl.reply.sequence - sq) <= 60);
>>          clock_gettime(CLOCK_MONOTONIC, &end);
>> @@ -167,32 +266,34 @@ static void vblank_wait(int fd, bool busy)
>>                   elapsed(&start, &end, count));
>>
>>          if (busy)
>> -               igt_assert_eq(read(fd, &buf, sizeof(buf)), sizeof(buf));
>> +               igt_assert_eq(read(data->drm_fd, &buf, sizeof(buf)), sizeof(buf));
>>   }
>>
>>   igt_main
>>   {
>> -       int fd;
>> +       data_t data;
>>
>>          igt_skip_on_simulation();
>>
>>          igt_fixture {
>> -               fd = drm_open_driver(DRIVER_ANY);
>> -               igt_require(crtc0_active(fd));
>> +               data.drm_fd = drm_open_driver(DRIVER_ANY);
>> +               kmstest_set_vt_graphics_mode();
>> +               igt_display_init(&data.display, data.drm_fd);
>>          }
>>
>>          igt_subtest("accuracy")
>> -               accuracy(fd);
>> +               run_test(&data, accuracy, false);
>>
>>          igt_subtest("query-idle")
>> -               vblank_query(fd, false);
>> +               run_test(&data, vblank_query, false);
>>
>>          igt_subtest("query-busy")
>> -               vblank_query(fd, true);
>> +               run_test(&data, vblank_query, true);
>>
>>          igt_subtest("wait-idle")
>> -               vblank_wait(fd, false);
>> +               run_test(&data, vblank_wait, false);
>>
>>          igt_subtest("wait-busy")
>> -               vblank_wait(fd, true);
>> +               run_test(&data, vblank_wait, true);
>>   }
>> +
>
> Thanks!
>
> Tomeu
>
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx