Skip to contentSkip to conversation

Question about CCS modifier on GLK

DISCUSSION Gabriel Krisman Bertazi 3 participants 3 messages Dec 18, 2017 – Dec 20, 2017

lore
Share & tools
lore
#0 lore raw

Hi Ben and list folks,

I've been investigating some CI failures with the kms_ccs testcase in the GLK hardware. The original bug is linked below, but there are other more basic tests failing when trying CCS on pipe C.

https://bugs.freedesktop.org/show_bug.cgi?id=104096

The reason for the failure is very straightforward, the function skl_check_ccs_aux_surface() in intel_display.c rejects CCS modifiers for PIPE_C, since that is not supported for SKL, KBL and as far as I can tell APL as well.

But, since I don't have the specs for GLK yet (it's not available in 01.org), I can't tell if GLK will support it and the kernel must be changed or if the testcase itself is bogus. My money is on the later, but I could use a confirmation before writing down the patch.

In short, does pipe C support render compression (ccs modifier) in glk?

Thanks,

Signature · 5 lines
Gabriel Krisman Bertazi
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw
Quoted 14 lines · Gabriel Krisman Bertazi

Hi Ben and list folks,

I've been investigating some CI failures with the kms_ccs testcase in the GLK hardware. The original bug is linked below, but there are other more basic tests failing when trying CCS on pipe C.

https://bugs.freedesktop.org/show_bug.cgi?id=104096

The reason for the failure is very straightforward, the function skl_check_ccs_aux_surface() in intel_display.c rejects CCS modifiers for PIPE_C, since that is not supported for SKL, KBL and as far as I can tell APL as well.

Quoted 6 lines

But, since I don't have the specs for GLK yet (it's not available in 01.org), I can't tell if GLK will support it and the kernel must be changed or if the testcase itself is bogus. My money is on the later, but I could use a confirmation before writing down the patch.

In short, does pipe C support render compression (ccs modifier) in glk?

No. But the current code is somewhat busted anyway. See
https://patchwork.freedesktop.org/series/29308/

Signature · 6 lines
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#2 lore raw
In reply to Gabriel Krisman Bertazi
Quoted 14 lines · Gabriel Krisman Bertazi

Hi Ben and list folks,

I've been investigating some CI failures with the kms_ccs testcase in the GLK hardware. The original bug is linked below, but there are other more basic tests failing when trying CCS on pipe C.

https://bugs.freedesktop.org/show_bug.cgi?id=104096

The reason for the failure is very straightforward, the function skl_check_ccs_aux_surface() in intel_display.c rejects CCS modifiers for PIPE_C, since that is not supported for SKL, KBL and as far as I can tell APL as well.

Quoted 9 lines

But, since I don't have the specs for GLK yet (it's not available in 01.org), I can't tell if GLK will support it and the kernel must be changed or if the testcase itself is bogus. My money is on the later, but I could use a confirmation before writing down the patch.

In short, does pipe C support render compression (ccs modifier) in glk?

Thanks,

The hardware almost certainly works, but it wasn't validated. So we should probably just disable it, and I've reviewed your patch. As Ville pointed out, the series which was merged wasn't quite right to begin with, and it'd be helpful if you could review his patches to fix it up.

Quoted 2 lines
Signature · 1 line
Gabriel Krisman Bertazi

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx