KMS seems broken in vc4.ko

[Ryutaroh Matsumoto]

Hi Maxime,

Thank you for your response.

> We don't have v3d enabled in the mainline kernel yet, so this looks
> right to me. I'm not quite sure why kmscube would run some V3D ioctl
> when using software rendering though

I have three questions:

(1) It seems suggesting VC4_GET_PARAM should succeed in Raspberry Pi OS
with v3d loaded. But in the Rapsberry Pi OS kernel 5.10.17 on my RPi4B 8GB,
it fails as

[  245.308362] [drm:drm_ioctl [drm]] comm="kmscube" pid=1216, dev=0xe201, auth=1, VC4_GET_PARAM
[  245.308481] [drm:drm_ioctl [drm]] comm="kmscube", pid=1216, ret=-19

after that V3D_GET_PARAM and DRM_IOCTL_MODE_SETCRTC succeeded
as shown in my first email on 18 March 2021 (08:40:12 +0900),
which seems contradicting to what you wrote.

(2) Should I report (1) as a bug in Raspberry Pi OS kernel to RPi github?

(3) Your email seems suggesting libdrm or kmscube has an issue (bug) that
failure of  VC4_GET_PARAM causes
[  148.887116] [drm:drm_ioctl [drm]] comm="kmscube" pid=1926, dev=0xe200, auth=1, DRM_IOCTL_MODE_SETCRTC
[  148.887142] [drm:drm_mode_setcrtc [drm]] [CRTC:64:crtc-3]
[  148.887192] [drm:drm_mode_setcrtc [drm]] Invalid pixel format XR24 little-endian (0x34325258), modifier 0xffffffffffffff

Should I report it to kmscube and/or libdrm.
Libdrm seems more suspicious to me.

Best regards, Ryutaroh