Gentoo Websites Logo
Go to: Gentoo Home Documentation Forums Lists Bugs Planet Store Wiki Get Gentoo!
Bug 866023 - sys-kernel/gentoo-sources-5.19: i915 hw conversion hangs on N3700/Airmont HD Graphics (Braswell)
Summary: sys-kernel/gentoo-sources-5.19: i915 hw conversion hangs on N3700/Airmont HD ...
Status: RESOLVED FIXED
Alias: None
Product: Gentoo Linux
Classification: Unclassified
Component: Current packages (show other bugs)
Hardware: All Linux
: Normal normal (vote)
Assignee: Gentoo Kernel Bug Wranglers and Kernel Maintainers
URL: https://gitlab.freedesktop.org/drm/in...
Whiteboard: 5.19.6
Keywords: InVCS
Depends on:
Blocks:
 
Reported: 2022-08-22 09:06 UTC by Luigi 'Comio' Mantellini
Modified: 2022-09-23 12:53 UTC (History)
0 users

See Also:
Package list:
Runtime testing required: ---


Attachments
Revert "drm/i915: Individualize fences before adding to dma_resv obj" upstream commit (0001-Revert-drm-i915-Individualize-fences-before-adding-t.patch,3.12 KB, patch)
2022-08-29 10:28 UTC, Luigi 'Comio' Mantellini
Details | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Luigi 'Comio' Mantellini 2022-08-22 09:06:24 UTC
I'm filling this bug in order to inform the last 5.19.x kernel doesn't permit to use GPU HW acceleration. The previous 5.18.x works perfectly.

This is the dmesg:

ago 18 14:16:54 abc kernel: i915 0000:00:02.0: [drm] Resetting vcs0 for preemption time out
ago 18 14:16:54 abc kernel: i915 0000:00:02.0: [drm] Plex Transcoder[77988] context reset due to GPU hang
ago 18 14:16:54 abc kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 8:4:a8ffdf3d, in Plex Transcoder [77988]
ago 18 14:17:09 abc kernel: i915 0000:00:02.0: [drm] GPU HANG: ecode 8:1:85d4fffb
ago 18 14:17:09 abc kernel: i915 0000:00:02.0: [drm] Resetting rcs0 for stopped heartbeat on rcs0

(tested on 5.19.0, 5.19.1 and 5.19.2)

ciao

luigi

Reproducible: Always

Steps to Reproduce:
1. Boot kernel 5.19.x
2. Start a GPU HW transcoding (via vaapi/Plex)
3. system hangs
Actual Results:  
System Hangs

Expected Results:  
GPU HW transcoding should works

Kernel 5.18.x works
Comment 1 Mike Pagano gentoo-dev 2022-08-22 18:24:20 UTC
Can you try by passing the argument enable_guc=2 to the i915 module ?
Comment 2 Mike Pagano gentoo-dev 2022-08-22 18:34:02 UTC
I see similar bugs around this upstream.

Can you please report this upstream?

There are two open relevant bugs I see:
https://gitlab.freedesktop.org/drm/intel/-/issues/3740
https://gitlab.freedesktop.org/drm/intel/-/issues/3398


There's also instructions on how to report i915 bugs:

https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs
Comment 3 Luigi 'Comio' Mantellini 2022-08-23 09:42:19 UTC
(In reply to Mike Pagano from comment #1)
> Can you try by passing the argument enable_guc=2 to the i915 module ?

I will give a try, but from my understanding guc is not available on my GPU.

ciao

luigi
Comment 4 Luigi 'Comio' Mantellini 2022-08-23 09:56:40 UTC
(In reply to Luigi 'Comio' Mantellini from comment #3)
> (In reply to Mike Pagano from comment #1)
> > Can you try by passing the argument enable_guc=2 to the i915 module ?
> 
> I will give a try, but from my understanding guc is not available on my GPU.
> 
> ciao
> 
> luigi

on gentoo-kernel-bin-5.19.3 even enabling guc the issue remains:

...
[    3.700478] i915 0000:00:02.0: [drm] Incompatible option enable_guc=2 - GuC is not supported!
[    3.700501] i915 0000:00:02.0: [drm] Incompatible option enable_guc=2 - HuC is not supported!
...
[  158.199612] i915 0000:00:02.0: [drm] GPU HANG: ecode 8:1:85defffb, in Plex Transcoder [5168]
[  158.200658] i915 0000:00:02.0: [drm] Resetting rcs0 for stopped heartbeat on rcs0
[  158.200726] i915 0000:00:02.0: [drm] Plex Transcoder[5168] context reset due to GPU hang
[  158.205685] i915 0000:00:02.0: [drm] Resetting vcs0 for CS error
[  158.205709] i915 0000:00:02.0: [drm] Plex Transcoder[5168] context reset due to GPU hang
[  158.228581] i915 0000:00:02.0: [drm] GPU HANG: ecode 8:4:40000ffe, in Plex Transcoder [5168]
...

I will fill a bug report on upstream.

ciao

luigi
Comment 5 Luigi 'Comio' Mantellini 2022-08-23 10:33:43 UTC
(In reply to Mike Pagano from comment #2)
> I see similar bugs around this upstream.
> 
> Can you please report this upstream?
> 
> There are two open relevant bugs I see:
> https://gitlab.freedesktop.org/drm/intel/-/issues/3740
> https://gitlab.freedesktop.org/drm/intel/-/issues/3398
> 
> 
> There's also instructions on how to report i915 bugs:
> 
> https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs


Upstream bug:

https://gitlab.freedesktop.org/drm/intel/-/issues/6665

ciao

luigi
Comment 6 Luigi 'Comio' Mantellini 2022-08-29 10:28:58 UTC
Created attachment 801835 [details, diff]
Revert "drm/i915: Individualize fences before adding to dma_resv obj" upstream commit

This revert seems to solve the issue.
Comment 7 Luigi 'Comio' Mantellini 2022-08-29 10:29:44 UTC
See upstream bug https://gitlab.freedesktop.org/drm/intel/-/issues/6665#note_1526565

ciao
Comment 8 Mike Pagano gentoo-dev 2022-08-31 13:33:50 UTC
Great work, thanks, queued for next release for gentoo-sources 5.19.6

commit cad5407b6e30efaf8a467c509b29ba5fb67f0cc8 (HEAD -> 5.19, origin/5.19)
Author: Mike Pagano <mpagano@gentoo.org>
Date:   Wed Aug 31 09:32:29 2022 -0400

    Revert fix from upstream for DRM/i915 thanks to Luigi 'Comio' Mantellini
    
    Bug: https://bugs.gentoo.org/866023
    
    Signed-off-by: Mike Pagano <mpagano@gentoo.org>
Comment 9 Larry the Git Cow gentoo-dev 2022-08-31 16:01:03 UTC
The bug has been closed via the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=b6ac0a98df573e702d624119d0fb4680fffeca1a

commit b6ac0a98df573e702d624119d0fb4680fffeca1a
Author:     Mike Pagano <mpagano@gentoo.org>
AuthorDate: 2022-08-31 15:58:55 +0000
Commit:     Mike Pagano <mpagano@gentoo.org>
CommitDate: 2022-08-31 15:58:55 +0000

    sys-kernel/gentoo-sources: add 5.19.6 and two fixes
    
    Revert fix from upstream for DRM/i915
    thanks to Luigi 'Comio' Mantellini
    Closes: https://bugs.gentoo.org/866023
    
    x86/sev: Don't use cc_platform_has() for early SEV-SNP calls
    Closes: https://bugs.gentoo.org/865831
    
    Signed-off-by: Mike Pagano <mpagano@gentoo.org>

 sys-kernel/gentoo-sources/Manifest                 |  3 +++
 .../gentoo-sources/gentoo-sources-5.19.6.ebuild    | 29 ++++++++++++++++++++++
 2 files changed, 32 insertions(+)
Comment 10 Luigi 'Comio' Mantellini 2022-09-23 12:52:23 UTC
Hi Mile Pagano,

The Linux Kernel 5.19.11 has the following patch to fix the issue:

commit dd52bde6767eba4f0800a2a60f91bc97cda3775c
Author: Nirmoy Das <nirmoy.das@intel.com>
Date:   Wed Sep 7 19:26:41 2022 +0200

    drm/i915: Set correct domains values at _i915_vma_move_to_active
    
    commit 08b812985996924c0ccf79d54a31fc9757c0a6ca upstream.

You can remove the fix from the gentoo-patchset.

ciao

luigi


(In reply to Mike Pagano from comment #8)
> Great work, thanks, queued for next release for gentoo-sources 5.19.6
> 
> commit cad5407b6e30efaf8a467c509b29ba5fb67f0cc8 (HEAD -> 5.19, origin/5.19)
> Author: Mike Pagano <mpagano@gentoo.org>
> Date:   Wed Aug 31 09:32:29 2022 -0400
> 
>     Revert fix from upstream for DRM/i915 thanks to Luigi 'Comio' Mantellini
>     
>     Bug: https://bugs.gentoo.org/866023
>     
>     Signed-off-by: Mike Pagano <mpagano@gentoo.org>
Comment 11 Luigi 'Comio' Mantellini 2022-09-23 12:53:01 UTC
s/Mile/Mike/g