[drm/amdgpu] Driver crashes on 5.13.9 kernel

[drm/amdgpu] Driver crashes on 5.13.9 kernel

[Skyler Mäntysaari]

Hello everyone on the list,

I thought that this should probably be discussed here,  so I came across weird issue to me which is driver crashing while trying to get one of my monitors working on Gentoo.  I would like to ask here how that would happen that the Display appears to jump from DisplayPort-6 (physical port) to DisplayPort-7 (which doesn't exist physically)? Has anyone else experienced this?

It seems that the driver sees a rather large amount of inputs for the GPU, even though I only have 4, 3 of which are DisplayPort, and the issue monitor is also on DisplayPort. 

Hardware:
CPU: AMD Ryzen 5800X
GPU: AMD Radeon RX 6800
System Memory: 32GB of DDR4 3200Mhz
Display(s): BenQ Zowie XL2430 (1080p), DELL U2414H (1080p), DELL U2415 (1920x1200)
Type of Diplay Connection: All are connected via Display-Port

Related DRM issue: https://gitlab.freedesktop.org/drm/amd/-/issues/1621 which includes logs too.


Best regards,
Skyler Mäntysaari

Re: [drm/amdgpu] Driver crashes on 5.13.9 kernel

[Kari Argillander]

On Sun, Aug 29, 2021 at 06:38:39PM +0300, Skyler Mäntysaari wrote:
> Hello everyone on the list,

This is universal kernel list and it is not read by many. I have added
hopefully right list (amd-gfx@lists.freedesktop.org).

> Subject: Re: [drm/amdgpu] Driver crashes on 5.13.9 kernel

I have no influence or knowledge about this driver, but I still try to
help because it seems good bug report. Have you test with 5.13.13 or
5.14-rc7. Does this work with some other kernel? If needed can you git
bisect if needed? You will probably get some support for it if needed.

	Argillander

> I thought that this should probably be discussed here,  so I came
> across weird issue to me which is driver crashing while trying to get
> one of my monitors working on Gentoo.  I would like to ask here how
> that would happen that the Display appears to jump from DisplayPort-6
> (physical port) to DisplayPort-7 (which doesn't exist physically)? Has
> anyone else experienced this?
> 
> It seems that the driver sees a rather large amount of inputs for the
> GPU, even though I only have 4, 3 of which are DisplayPort, and the
> issue monitor is also on DisplayPort. 
> 
> Hardware:
> CPU: AMD Ryzen 5800X
> GPU: AMD Radeon RX 6800
> System Memory: 32GB of DDR4 3200Mhz
> Display(s): BenQ Zowie XL2430 (1080p), DELL U2414H (1080p), DELL U2415 (1920x1200)
> Type of Diplay Connection: All are connected via Display-Port
> 
> Related DRM issue:
> https://gitlab.freedesktop.org/drm/amd/-/issues/1621 which includes
> logs too.
> 
> 
> Best regards,
> Skyler Mäntysaari

Re: [drm/amdgpu] Driver crashes on 5.13.9 kernel

[Skyler Mäntysaari]

I have tried kernel 5.13.13, without any difference and I haven't tried with an older kernel, as this hardware is that new that I have very little faith in less than 5.x kernel would even have support for the needed GPU.

What do you mean with git bisect? I have checked that the crash happens somewhere in the monitor connection code:

[ 9605.269927] Call Trace:
[ 9605.269931]  core_link_enable_stream+0x746/0x870 [amdgpu]
[ 9605.270038]  dce110_apply_ctx_to_hw+0x519/0x560 [amdgpu]
[ 9605.270146]  dc_commit_state+0x2f6/0xa50 [amdgpu]
[ 9605.270249]  amdgpu_dm_atomic_commit_tail+0x569/0x26a0 [amdgpu]
[ 9605.270326]  ? kfree+0xc3/0x460
[ 9605.270329]  ? dcn30_validate_bandwidth+0x11f/0x270 [amdgpu]
[ 9605.270402]  ? dcn30_validate_bandwidth+0x11f/0x270 [amdgpu]
[ 9605.270469]  ? dm_plane_helper_prepare_fb+0x19c/0x250 [amdgpu]
[ 9605.270542]  ? __cond_resched+0x16/0x40
[ 9605.270544]  ? __wait_for_common+0x3b/0x160
[ 9605.270545]  ? __raw_callee_save___native_queued_spin_unlock+0x11/0x1e
[ 9605.270548]  commit_tail+0x94/0x130 [drm_kms_helper]
[ 9605.270557]  drm_atomic_helper_commit+0x113/0x140 [drm_kms_helper]
[ 9605.270562]  drm_atomic_helper_set_config+0x70/0xb0 [drm_kms_helper]
[ 9605.270568]  drm_mode_setcrtc+0x1d3/0x6d0 [drm]
[ 9605.270582]  ? drm_mode_getcrtc+0x180/0x180 [drm]
[ 9605.270590]  drm_ioctl_kernel+0xaa/0xf0 [drm]
[ 9605.270600]  drm_ioctl+0x220/0x3c0 [drm]
[ 9605.270609]  ? drm_mode_getcrtc+0x180/0x180 [drm]
[ 9605.270618]  amdgpu_drm_ioctl+0x49/0x80 [amdgpu]
[ 9605.270673]  __x64_sys_ioctl+0x83/0xb0
[ 9605.270675]  do_syscall_64+0x40/0xb0
[ 9605.270677]  entry_SYSCALL_64_after_hwframe+0x44/0xae


On Sun, Aug 29, 2021, at 20:34, Kari Argillander wrote:
> On Sun, Aug 29, 2021 at 06:38:39PM +0300, Skyler Mäntysaari wrote:
> > Hello everyone on the list,
> 
> This is universal kernel list and it is not read by many. I have added
> hopefully right list (amd-gfx@lists.freedesktop.org).
> 
> > Subject: Re: [drm/amdgpu] Driver crashes on 5.13.9 kernel
> 
> I have no influence or knowledge about this driver, but I still try to
> help because it seems good bug report. Have you test with 5.13.13 or
> 5.14-rc7. Does this work with some other kernel? If needed can you git
> bisect if needed? You will probably get some support for it if needed.
> 
> Argillander
> 
> > I thought that this should probably be discussed here,  so I came
> > across weird issue to me which is driver crashing while trying to get
> > one of my monitors working on Gentoo.  I would like to ask here how
> > that would happen that the Display appears to jump from DisplayPort-6
> > (physical port) to DisplayPort-7 (which doesn't exist physically)? Has
> > anyone else experienced this?
> > 
> > It seems that the driver sees a rather large amount of inputs for the
> > GPU, even though I only have 4, 3 of which are DisplayPort, and the
> > issue monitor is also on DisplayPort. 
> > 
> > Hardware:
> > CPU: AMD Ryzen 5800X
> > GPU: AMD Radeon RX 6800
> > System Memory: 32GB of DDR4 3200Mhz
> > Display(s): BenQ Zowie XL2430 (1080p), DELL U2414H (1080p), DELL U2415 (1920x1200)
> > Type of Diplay Connection: All are connected via Display-Port
> > 
> > Related DRM issue:
> > https://gitlab.freedesktop.org/drm/amd/-/issues/1621 which includes
> > logs too.
> > 
> > 
> > Best regards,
> > Skyler Mäntysaari

Re: [drm/amdgpu] Driver crashes on 5.13.9 kernel

[Kari Argillander]

On Mon, Aug 30, 2021 at 07:15:29PM +0300, Skyler Mäntysaari wrote:
> I have tried kernel 5.13.13, without any difference and I haven't
> tried with an older kernel, as this hardware is that new that I have
> very little faith in less than 5.x kernel would even have support for
> the needed GPU.

Yeah might be. 

> What do you mean with git bisect? I have checked that the crash
> happens somewhere in the monitor connection code:

If we found some older version which work. Then with git bisect we can
track which patch made this bug. But if it never work then git bisect
won't help.

You may still want to test with 5.14 as it was released today.

If that does not work then I cannot help more. It is still very
important to know if it is also broken in 5.14. Hopefully someone will
able to help you. 

  Argillander

> [ 9605.269927] Call Trace:
> [ 9605.269931]  core_link_enable_stream+0x746/0x870 [amdgpu]
> [ 9605.270038]  dce110_apply_ctx_to_hw+0x519/0x560 [amdgpu]
> [ 9605.270146]  dc_commit_state+0x2f6/0xa50 [amdgpu]
> [ 9605.270249]  amdgpu_dm_atomic_commit_tail+0x569/0x26a0 [amdgpu]
> [ 9605.270326]  ? kfree+0xc3/0x460
> [ 9605.270329]  ? dcn30_validate_bandwidth+0x11f/0x270 [amdgpu]
> [ 9605.270402]  ? dcn30_validate_bandwidth+0x11f/0x270 [amdgpu]
> [ 9605.270469]  ? dm_plane_helper_prepare_fb+0x19c/0x250 [amdgpu]
> [ 9605.270542]  ? __cond_resched+0x16/0x40
> [ 9605.270544]  ? __wait_for_common+0x3b/0x160
> [ 9605.270545]  ? __raw_callee_save___native_queued_spin_unlock+0x11/0x1e
> [ 9605.270548]  commit_tail+0x94/0x130 [drm_kms_helper]
> [ 9605.270557]  drm_atomic_helper_commit+0x113/0x140 [drm_kms_helper]
> [ 9605.270562]  drm_atomic_helper_set_config+0x70/0xb0 [drm_kms_helper]
> [ 9605.270568]  drm_mode_setcrtc+0x1d3/0x6d0 [drm]
> [ 9605.270582]  ? drm_mode_getcrtc+0x180/0x180 [drm]
> [ 9605.270590]  drm_ioctl_kernel+0xaa/0xf0 [drm]
> [ 9605.270600]  drm_ioctl+0x220/0x3c0 [drm]
> [ 9605.270609]  ? drm_mode_getcrtc+0x180/0x180 [drm]
> [ 9605.270618]  amdgpu_drm_ioctl+0x49/0x80 [amdgpu]
> [ 9605.270673]  __x64_sys_ioctl+0x83/0xb0
> [ 9605.270675]  do_syscall_64+0x40/0xb0
> [ 9605.270677]  entry_SYSCALL_64_after_hwframe+0x44/0xae
> 
> 
> On Sun, Aug 29, 2021, at 20:34, Kari Argillander wrote:
> > On Sun, Aug 29, 2021 at 06:38:39PM +0300, Skyler Mäntysaari wrote:
> > > Hello everyone on the list,
> > 
> > This is universal kernel list and it is not read by many. I have added
> > hopefully right list (amd-gfx@lists.freedesktop.org).
> > 
> > > Subject: Re: [drm/amdgpu] Driver crashes on 5.13.9 kernel
> > 
> > I have no influence or knowledge about this driver, but I still try to
> > help because it seems good bug report. Have you test with 5.13.13 or
> > 5.14-rc7. Does this work with some other kernel? If needed can you git
> > bisect if needed? You will probably get some support for it if needed.
> > 
> > Argillander
> > 
> > > I thought that this should probably be discussed here,  so I came
> > > across weird issue to me which is driver crashing while trying to get
> > > one of my monitors working on Gentoo.  I would like to ask here how
> > > that would happen that the Display appears to jump from DisplayPort-6
> > > (physical port) to DisplayPort-7 (which doesn't exist physically)? Has
> > > anyone else experienced this?
> > > 
> > > It seems that the driver sees a rather large amount of inputs for the
> > > GPU, even though I only have 4, 3 of which are DisplayPort, and the
> > > issue monitor is also on DisplayPort. 
> > > 
> > > Hardware:
> > > CPU: AMD Ryzen 5800X
> > > GPU: AMD Radeon RX 6800
> > > System Memory: 32GB of DDR4 3200Mhz
> > > Display(s): BenQ Zowie XL2430 (1080p), DELL U2414H (1080p), DELL U2415 (1920x1200)
> > > Type of Diplay Connection: All are connected via Display-Port
> > > 
> > > Related DRM issue:
> > > https://gitlab.freedesktop.org/drm/amd/-/issues/1621 which includes
> > > logs too.
> > > 
> > > 
> > > Best regards,
> > > Skyler Mäntysaari