[igt-dev] [PATCH i-g-t] tests/kms_plane_alpha_blend: Fix a check in is_6bpc function

Thu Jul 22 13:40:15 UTC 2021

> -----Original Message-----
> From: Srinivas, Vidya
> Sent: Thursday, July 15, 2021 9:21 AM
> To: 'Ville Syrjälä' <ville.syrjala at linux.intel.com>
> Cc: igt-dev at lists.freedesktop.org
> Subject: RE: [igt-dev] [PATCH i-g-t] tests/kms_plane_alpha_blend: Fix a check
> in is_6bpc function
> 
> 
> 
> > -----Original Message-----
> > From: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > Sent: Thursday, July 15, 2021 8:36 AM
> > To: Srinivas, Vidya <vidya.srinivas at intel.com>
> > Cc: igt-dev at lists.freedesktop.org
> > Subject: Re: [igt-dev] [PATCH i-g-t] tests/kms_plane_alpha_blend: Fix
> > a check in is_6bpc function
> >
> > On Wed, Jul 14, 2021 at 03:01:36PM +0000, Srinivas, Vidya wrote:
> > > Hello Ville,
> > >
> > > Thank you so much. I limited the bpp to 18 in compute_sink_pipe_bpp.
> > > I tested on external monitor both HDMI and DP. Tests did not fail.
> > > It passed
> > without CRC mismatch.
> > > I checked display_info and ensure bpp shows 18 on pipe B for both
> > > HDMI
> > and DP case.
> > > Dithering was set to ON.
> > >
> > > [CRTC:152:pipe B]:
> > >         uapi: enable=yes, active=yes, mode="3840x2160": 30 266640
> > > 3840
> > 3848 3992 4000 2160 2214 2219 2222 0x40 0x9
> > >         hw: active=yes, adjusted_mode="3840x2160": 30 266640 3840
> > > 3848
> > 3992 4000 2160 2214 2219 2222 0x40 0x9
> > >         pipe src size=3840x2160, dither=yes, bpp=18
> > >
> > > I am really sorry. Should I revert this patch? Kindly guide me
> > > further. Thank
> > you so much.
> >
> > Well, before we go making too many changes would be nice to see if we
> > can figure out what is really causing the crc mismatches on eDP. I
> > can't immediately think of anything eDP specific it could be really,
> > apart from maybe some PSR/etc. related bug. Have we ruled those sort if
> things out?
> 
> Hello Ville, thank you so much. The JSL chromebook we are using is a non-
> PSR panel.
> I will try to see if I can figure out anything else and update.
> 
> Regards
> Vidya

Hello Ville,

Tejas helped find the root cause. The issue was not with respect to BPC or dithering as you said.
The JSL panel I have on this chromebook is 1366x768. If we align the width to 64 for the buffers
being created, it works fine. Many thanks to you for pointing out the wrong fix and thanks to Tejas
for helping with the fix. He will submit the patch. I will request a revert of the wrong patch.
Really sorry for that.

Regards
Vidya
> 
> >
> > >
> > > Regards
> > > Vidya
> > >
> > > -----Original Message-----
> > > From: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > Sent: Wednesday, July 14, 2021 7:57 PM
> > > To: Srinivas, Vidya <vidya.srinivas at intel.com>
> > > Cc: igt-dev at lists.freedesktop.org
> > > Subject: Re: [igt-dev] [PATCH i-g-t] tests/kms_plane_alpha_blend:
> > > Fix a check in is_6bpc function
> > >
> > > On Wed, Jul 14, 2021 at 01:22:26PM +0000, Srinivas, Vidya wrote:
> > > > Hello Ville,
> > > >
> > > > I am sorry, I don’t have a 6 bpc external panel.
> > >
> > > You don't need one. Just limit the bpp to 18 when computing the link
> > params.
> > >
> > > > Juha-Pekka had suggested we use the set_lut table as being used in
> > > > kms_flip_scaled_crc.c But when I tried that, max upto 0x7F it
> > > > would pass
> > (not with original 0xFF00). After that it would give back CRC mismatch.
> > > >
> > > > Set_lut:
> > > > uint16_t v  = (i * 0xffff / (lut_size - 1)) & 0x7f; lut[i].red =
> > > > v; lut[i].green = v; lut[i].blue = v;
> > > >
> > > > Other solutions which worked were:
> > > > 1. Use alpha for primary plane
> > > >
> https://patchwork.freedesktop.org/patch/440263/?series=90828&rev=6
> > > >
> > > > 2. Adding this line before setting the primary gray
> > > >
> > > > igt_plane_set_prop_value(igt_pipe_get_plane_type(&display->pipes[p
> > > > ip e] , DRM_PLANE_TYPE_PRIMARY), IGT_PLANE_ALPHA, 0);
> > > >
> > > > 2. Remove primary plane
> > > >
> https://patchwork.freedesktop.org/patch/436178/?series=90828&rev=1
> > > >
> > > > 3. Use lower alpha values 0x7e for coverage, pre-multiplied and
> > > > constant
> > > >
> https://patchwork.freedesktop.org/patch/438831/?series=90828&rev=3
> > > >
> > > > 4. Use an alpha buffer for primary commit
> > > >
> https://patchwork.freedesktop.org/patch/440157/?series=90828&rev=5
> > > >
> > > > 5. Use black_fb instead of gray_fb in
> > > > coverage-vs-premult-vs-constant for primary
> > > >
> https://patchwork.freedesktop.org/patch/441554/?series=90828&rev=7
> > > >
> > > > Apart from these, I disabled dithering in kernel and that too did not
> help.
> > Screen shows content correctly but CRC mismatch is seen.
> > > >
> > > > Regards
> > > > Vidya
> > > >
> > > >
> > > > -----Original Message-----
> > > > From: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > > Sent: Wednesday, July 14, 2021 6:34 PM
> > > > To: Srinivas, Vidya <vidya.srinivas at intel.com>
> > > > Cc: igt-dev at lists.freedesktop.org
> > > > Subject: Re: [igt-dev] [PATCH i-g-t] tests/kms_plane_alpha_blend:
> > > > Fix a check in is_6bpc function
> > > >
> > > > On Wed, Jul 14, 2021 at 11:21:47AM +0000, Srinivas, Vidya wrote:
> > > > > Hello Ville,
> > > > >
> > > > > Apologies. I tried to turn off dithering from kernel itself. But
> > > > > still CRC
> > mismatch came up.
> > > > > This is on Jasperlake chromebook which is using 6bpc panel.
> > > > >
> > > > > Same jasperlake with 8bpc panel is working fine.
> > > > >
> > > > > So, as last resort we did this ☹.
> > > >
> > > > Well that's a bit unfortunate because now the commit messages are
> > > > lying,
> > and we still don't know what is going on.
> > > >
> > > > Do the same failures happen with external DP @ 6bpc? I wouldn't
> > > > really
> > expect any crc differences since AFAIK nothing in the pipe should
> > change before the dithering block, and with dithering disabled the
> > port should just chop off the extra bits at the very end.
> > > >
> > > > >
> > > > > Regards
> > > > > Vidya
> > > > >
> > > > > -----Original Message-----
> > > > > From: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > > > Sent: Wednesday, July 14, 2021 3:55 PM
> > > > > To: Srinivas, Vidya <vidya.srinivas at intel.com>
> > > > > Cc: igt-dev at lists.freedesktop.org
> > > > > Subject: Re: [igt-dev] [PATCH i-g-t] tests/kms_plane_alpha_blend:
> > > > > Fix a check in is_6bpc function
> > > > >
> > > > > On Wed, Jul 14, 2021 at 01:30:02PM +0530, Vidya Srinivas wrote:
> > > > > > Fixes commit 3bf28f9dffd41b85c262d4e6664ffbdf5b7d9a93.
> > > > > >
> > > > > > is_6bpc is supposed to return false if connector is not eDP or DSI.
> > > > > > The right check will be DRM_MODE_CONNECTOR_eDP &&
> > > > > > DRM_MODE_CONNECTOR_DSI By mistake
> > DRM_MODE_CONNECTOR_eDP ||
> > > > > > DRM_MODE_CONNECTOR_DSI was being used.
> > > > >
> > > > > Instead of these incomplete hacks why aren't we just turning off
> > dithering when doing crc capture?
> > > > >
> > > > > >
> > > > > > Credits-to: Ashutosh Dixit <ashutosh.dixit at intel.com>
> > > > > > Signed-off-by: Vidya Srinivas <vidya.srinivas at intel.com>
> > > > > > ---
> > > > > >  tests/kms_plane_alpha_blend.c | 2 +-
> > > > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > > >
> > > > > > diff --git a/tests/kms_plane_alpha_blend.c
> > > > > > b/tests/kms_plane_alpha_blend.c index
> > > > > > a3529dc2190f..8020216c1594
> > > > > > 100644
> > > > > > --- a/tests/kms_plane_alpha_blend.c
> > > > > > +++ b/tests/kms_plane_alpha_blend.c
> > > > > > @@ -454,7 +454,7 @@ static bool is_6bpc(igt_display_t
> > > > > > *display,
> > enum pipe pipe) {
> > > > > >  		return false;
> > > > > >
> > > > > >  	c = output->config.connector;
> > > > > > -	if (c->connector_type != DRM_MODE_CONNECTOR_eDP ||
> > > > > > +	if (c->connector_type != DRM_MODE_CONNECTOR_eDP &&
> > > > > >  		c->connector_type != DRM_MODE_CONNECTOR_DSI)
> > > > > >  		return false;
> > > > > >
> > > > > > --
> > > > > > 2.32.0
> > > > > >
> > > > > > _______________________________________________
> > > > > > igt-dev mailing list
> > > > > > igt-dev at lists.freedesktop.org
> > > > > > https://lists.freedesktop.org/mailman/listinfo/igt-dev
> > > > >
> > > > > --
> > > > > Ville Syrjälä
> > > > > Intel
> > > >
> > > > --
> > > > Ville Syrjälä
> > > > Intel
> > >
> > > --
> > > Ville Syrjälä
> > > Intel
> >
> > --
> > Ville Syrjälä
> > Intel