system starts, but display seems to deactivate

seth · 2024-10-17 18:12:08

https://gitlab.freedesktop.org/drm/amd/-/issues/ and click "new ticket" but teh proper syntax would be "HDMI-A-1:edid/HDMIedid.bin", skip the "card0-" part.

sukolyn · 2024-10-17 20:53:48

not better.

on my way to "File a bug upstream".

Edit : https://gitlab.freedesktop.org/drm/amd/-/issues/3696
is it all right/enough ?

seth · 2024-10-18 05:58:29

Yes. As predicted Alex is now asking you to https://wiki.gentoo.org/wiki/Kernel_git-bisect the issue.

sukolyn · 2024-10-18 19:10:56

I don't know since when drm-tip doesn't work anymore, or if it even ever worked, in my case.
so, if I understood correctly, bisecting will be difficult.

Last edited by sukolyn (2024-10-18 19:11:32)

seth · 2024-10-18 19:13:41

You'd bisect between the last 6.10 and the first 6.11 kernel, drm-tip has nothing to do with this - the goal is to figure which of the maaaaaany commits broke this and bisection can do that in logarithmic time.

sukolyn · 2024-10-18 20:20:55

I understand the logic of bisecting (it's a kind of "more or less" guess a number), but I can't figure out how to apply what I read in the Gentoo page to kernels 6.10 to 6.11

what do I have to download? how?
does it mean that I'll have to compile every kernel I download? if so, then it may take weeks until I find the "bad" commit.

seth · 2024-10-18 20:40:34

The main kernel PKGBUILDs no longer use git, but eg. linux-drm-tip-git does and you can use it for bisection, using "makepkg -e" to avoid overriding your git changes to the kernel git tree.

Let's say you'd have to build and test 20 kernels: that will cover > 1.000.000 commits
There're 15122 commits in the 6.11 changelog, which can be covered in 14 bisection steps

sukolyn · 2024-10-18 20:54:50

sorry. it's not clear.

first, what do I have to download ?

seth · 2024-10-19 07:14:48

You already have the linux-drm-tip-git package, right?
The download location (where you built it) contains a kernel git tree.

sukolyn · 2024-10-19 12:35:03

$ cd linux-drm-tip-git
$ ls -p1 .git
branches/
hooks/
info/
logs/
objects/
refs/
config
description
FETCH_HEAD
HEAD
index
ORIG_HEAD
packed-refs
$

seth · 2024-10-19 12:43:13

This is either the AUR package or the kernel inside…

cat .git/config

sukolyn · 2024-10-19 13:47:04

it is the AUR package :

$ cat .git/config
[core]
        repositoryformatversion = 0
        filemode = true
        bare = false
        logallrefupdates = true
[remote "origin"]
        url = https://aur.archlinux.org/linux-drm-tip-git.git
        fetch = +refs/heads/*:refs/remotes/origin/*
[branch "master"]
        remote = origin
        merge = refs/heads/master

seth · 2024-10-19 14:18:13

There should be a linux-drm-tip-git directory inside? And that one is also a git structure, but this time for the kernel.
That's what you're going to bisect.

sukolyn · 2024-10-19 14:49:42

seth wrote:

There should be a linux-drm-tip-git directory inside?

yes, there is.

That's what you're going to bisect.

lost

$ cd linux-drm-tip-git
$ ls -1p
linux-drm-tip-git/
config
PKGBUILD
linux-drm-tip-docs-git-6.12.r1310331.87312f4bbb70-1-x86_64.pkg.tar.zst
linux-drm-tip-git-6.12.r1310331.87312f4bbb70-1-x86_64.pkg.tar.zst
linux-drm-tip-headers-git-6.12.r1310331.87312f4bbb70-1-x86_64.pkg.tar.zst

so

$ cd linux-drm-tip-git
branches/
hooks/
info/
objects/
refs/
config
description
HEAD
packed-refs

and then?

seth · 2024-10-19 18:30:32

"cat .git/config" there will show you that it's the linux source tree and then you start the bisection process there, ie. "git bisect start" and define a good and a bad version like in the gentoo wiki. (I guess 6.10.9 is good and 6.11.0 is bad?)
Then "cd .." back into the AUR package and there "makepkg -e" to build a kernel package w/o manipulating the linux git sources.
Install and test the kernel then head back to the linux source, "git bisect" good or bad, depending on the success, cd .., makepkg -e, rinse and repeat until your find the breaking commit.

sukolyn · 2024-10-20 05:34:22

"cat .git/config" there will show you that it's the linux source tree

what should I see ?

I get into linux-drm-tip-git/linux-drm-tip-git/
do `git bisect stat'
when I `cd ..' it tells me 'You need to start by "git bisect start"'

so I start bisect under upper Linux-drm-tip-git, and it doesn't complain anymore,
then I `makepkg -ef' (-f because "a group of package has already been compiled...")

for the moment, it's compiling... linux-drm-tip-git 6.12.r1310445.38b97dc95877-1

so I think there's a problem because I told that 6.10.9 is good and 6.11.0 is bad
why is 6.12 compiling ?

seth · 2024-10-20 06:23:00

why is 6.12 compiling

do `git bisect stat' … when I `cd ..' it tells me 'You need to start by "git bisect start"'

Because you ran whatever in the linux source dir, then moved up to the AUR package git and started to bisect that, what's not gonna help anything because it only tracks changes in the AUR package of which there're few.
You're not running anything "git" in the dir w/ the PKGBUILD, you're running all the git bisect stuff in the "inner" linux-drm-tip-git directory. THEN move back into the upper AUR dir and ONLY "makepkg -e" there.

sukolyn · 2024-10-20 22:09:36

grrr

$ cd linux-drm-tip-git/linux-drm-tip-git
$ git bisect start
status: waiting for both good and bad commits
$ git bisect good 6.10.9
error: Bad rev input: 6.10.9
$ git bisect bad 6.11.0
error: Bad rev input: 6.11.0

Last edited by sukolyn (2024-10-20 22:10:01)

seth · 2024-10-21 06:00:13

"v6.10.9"
"v6.11.0"

git tag -l

sukolyn · 2024-10-21 09:07:11

same error with v6.10.9 and v6.11.0

`git tag -l' prints nothing.

seth · 2024-10-21 12:15:01

cgit.fdo is currently down, if you don't mind, bisect https://github.com/archlinux/linux but there's no PKGBUILD for that, so you'll have to follow the manual kernel installation process.
Otherwise you'll have to wait until I can lookup the relevant commit ids at cgit.freedesktop.org/drm-tip (assuming they don't tag the releases)

sukolyn · 2024-10-21 15:44:15

I already built a kernel (you gave me a link for it) using pkgctl for downloading, adapting PKGBUILD, updpkgsums, and makepkg -s

is that you you want me to bisect ?

seth · 2024-10-21 18:49:45

The regular arch kernels aren't built out of a git tree - drm-tip doesn't track the subversions, it's just the mainline tree.

Bad: 98f7e32f20d28ec452afb208f9cffc08448a2652 (6.11)
Good: 0c3836482481200ead7b416ca80c68a29cfdaabd (6.10)

To bisect the arch kernels, you'll have to install the kernel manually, adapting the PKGBUILD will be more involved than chaning some hashes.

sukolyn · 2024-10-21 19:21:47

The regular arch kernels aren't built out of a git tree

so, no `git bisect' then, I suppose.

Bad: 98f7e32f20d28ec452afb208f9cffc08448a2652 (6.11)
Good: 0c3836482481200ead7b416ca80c68a29cfdaabd (6.10)

where/how did you get these ?

about 13 steps expected
...

seth · 2024-10-21 19:25:29

I looked up the git tree of drm-tip, the commit IDs are however the same as in the mainline kernel.
So happy bisection.

Arch Linux

#51 2024-10-17 18:12:08

Re: system starts, but display seems to deactivate

#52 2024-10-17 20:53:48

Re: system starts, but display seems to deactivate

#53 2024-10-18 05:58:29

Re: system starts, but display seems to deactivate

#54 2024-10-18 19:10:56

Re: system starts, but display seems to deactivate

#55 2024-10-18 19:13:41

Re: system starts, but display seems to deactivate

#56 2024-10-18 20:20:55

Re: system starts, but display seems to deactivate

#57 2024-10-18 20:40:34

Re: system starts, but display seems to deactivate

#58 2024-10-18 20:54:50

Re: system starts, but display seems to deactivate

#59 2024-10-19 07:14:48

Re: system starts, but display seems to deactivate

#60 2024-10-19 12:35:03

Re: system starts, but display seems to deactivate

#61 2024-10-19 12:43:13

Re: system starts, but display seems to deactivate

#62 2024-10-19 13:47:04

Re: system starts, but display seems to deactivate

#63 2024-10-19 14:18:13

Re: system starts, but display seems to deactivate

#64 2024-10-19 14:49:42

Re: system starts, but display seems to deactivate

#65 2024-10-19 18:30:32

Re: system starts, but display seems to deactivate

#66 2024-10-20 05:34:22

Re: system starts, but display seems to deactivate

#67 2024-10-20 06:23:00

Re: system starts, but display seems to deactivate

#68 2024-10-20 22:09:36

Re: system starts, but display seems to deactivate

#69 2024-10-21 06:00:13

Re: system starts, but display seems to deactivate

#70 2024-10-21 09:07:11

Re: system starts, but display seems to deactivate

#71 2024-10-21 12:15:01

Re: system starts, but display seems to deactivate

#72 2024-10-21 15:44:15

Re: system starts, but display seems to deactivate

#73 2024-10-21 18:49:45

Re: system starts, but display seems to deactivate

#74 2024-10-21 19:21:47

Re: system starts, but display seems to deactivate

#75 2024-10-21 19:25:29

Re: system starts, but display seems to deactivate

Board footer