788847 – sys-kernel/gentoo-kernel-5.10.33: stablereq

Bug 788847 - sys-kernel/gentoo-kernel-5.10.33: stablereq

Summary: sys-kernel/gentoo-kernel-5.10.33: stablereq

Status:	RESOLVED FIXED

Alias:	None

Product:	Gentoo Linux
Classification:	Unclassified
Component:	Stabilization (show other bugs)
Hardware:	All Linux

Importance:	Normal enhancement (vote)
Assignee:	Distribution Kernel Project

URL:
Whiteboard:
Keywords:

Depends on:
Blocks:

Reported:	2021-05-07 22:25 UTC by Michał Górny
Modified:	2021-05-12 19:46 UTC (History)
CC List:	0 users

See Also:
Package list:	sys-kernel/gentoo-kernel-5.10.33 sys-kernel/gentoo-kernel-bin-5.10.33 amd64 x86 sys-kernel/gentoo-kernel-bin-5.10.33-r1 arm64 virtual/dist-kernel-5.10.33
Runtime testing required:	---

Flags:	nattka: sanity-check+

Attachments
amdgpu ppc64 64K pagesz patch (amdgpu-pagesz.patch,2.28 KB, patch) 2021-05-12 16:40 UTC, Georgy Yakovlev	no flags	Details \| Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.

Description Michał Górny archtester

2021-05-07 22:25:50 UTC

Please stabilize.

Comment 1 Georgy Yakovlev archtester

2021-05-08 02:44:26 UTC

amdgpu is busted on ppc64 in this release.
will look for workarounds/patches or solutions. can't stabilize as is, since amd cards is what 90% of ppc64le users have in their machines.

Comment 2 Michał Górny archtester

2021-05-08 06:44:33 UTC

Is a newer/older release better?

Comment 3 NATTkA bot gentoo-dev

2021-05-08 06:48:21 UTC Comment hidden (obsolete)

Resetting sanity check; keywords are not fully specified and arches are not CC-ed.

Comment 4 Georgy Yakovlev archtester

2021-05-08 11:11:11 UTC

.27 is current stable and it's the most stable amdgpu-wise fo ppc64.
I see amdgpu changes landed in .28 and nothing else till .35 as I grepped.

I haven't actually tested anything  besides .27 and .33, will try to bisect, but I suspect it's .28 that introduced the regression.

Don't wanna block things for other arches, so as last resort we could leave .27 in place and maybe even remove other keywords if situation does not change soon enough, and stabilize ppc64 later as problem fixed.
all that said assuming it's ppc64-only problem.

Comment 5 Michał Górny archtester

2021-05-09 10:06:31 UTC

skipping ppc64 then.

Comment 6 Sam James archtester

2021-05-10 13:15:04 UTC

amd64 done

Comment 7 Sam James archtester

2021-05-10 14:00:43 UTC

arm64 done

Comment 8 Georgy Yakovlev archtester

2021-05-12 16:40:39 UTC

Created attachment 707637 [details, diff]
amdgpu ppc64 64K pagesz patch

I've found and backported ppc64 fix.
Proposing adding single PATCH to PATCHES=( ) as one-time exception for ppc64, without revbump and I can stabilize it then.

problem is only observed on 64K pagesz kernels ( the one we ship in default configuration ), details in patch header.

Comment 9 Michał Górny archtester

2021-05-12 17:42:02 UTC

Your call.

Comment 10 Larry the Git Cow gentoo-dev

2021-05-12 18:43:11 UTC

The bug has been referenced in the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=ca82103ffc4fb31a981fb7671a8251796df0f3fa

commit ca82103ffc4fb31a981fb7671a8251796df0f3fa
Author:     Georgy Yakovlev <gyakovlev@gentoo.org>
AuthorDate: 2021-05-12 18:40:31 +0000
Commit:     Georgy Yakovlev <gyakovlev@gentoo.org>
CommitDate: 2021-05-12 18:41:05 +0000

    sys-kernel/gentoo-kernel: add amggpu pagesz patch to 5.10.33
    
    Bug: https://bugs.gentoo.org/788847
    Package-Manager: Portage-3.0.18, Repoman-3.0.3
    Signed-off-by: Georgy Yakovlev <gyakovlev@gentoo.org>

 .../files/5.10.33-amdgpu-pagesz.patch              | 50 ++++++++++++++++++++++
 .../gentoo-kernel/gentoo-kernel-5.10.33.ebuild     |  1 +
 2 files changed, 51 insertions(+)

Comment 11 Georgy Yakovlev archtester

2021-05-12 18:43:27 UTC

ppc64 done (without CC)

Comment 12 Michał Górny archtester

2021-05-12 19:46:09 UTC

x86 done

all arches done