Arch Linux

towerclimber

Member

Registered: 2015-05-01

Posts: 4

Kernel boot problems

I need some help, My laptop is a Toshiba Satellite C55Dt AMD A6 with a radeon 8400. My problem is kernel 4.8 and below boot properly but initram fails to boot on my self compiled 4.9 and 4.10 kernels using arch AUR linux-git and linuxpf. I cannot figure out how to get a log because journalctl is not running when the kernel panicks.
The error is to the effect "bad: scheduling from idle thread!" CPU: 0 PID: 0 Comm: swapper/0 Tainted: nothing really interesting before or after and it doesn't list anything specifically. I also tried enabling Southern Islands in the kernel config. In the kernel config I used the default linux 4.8.

I have torn apart the initramfs with lsinitcpio and then did the same to my 4.8 that boots and they are the same. I can boot if I disable acpi at boot

Last edited by towerclimber (2017-01-10 01:46:25)

loqs

Member

Registered: 2014-03-06

Posts: 17,321

Re: Kernel boot problems

Would recommend you bisect the kernel between 4.8 and 4.9 and then see if any already reported issue matches the bad commit as identified by the bisect.

fredbezies

Member

Registered: 2011-07-28

Posts: 353

Re: Kernel boot problems

I need some help, My laptop is a Toshiba Satellite C55Dt AMD A6 with a radeon 8400. My problem is kernel 4.8 and below boot properly but initram fails to boot on my self compiled 4.9 and 4.10 kernels using arch AUR linux-git and linuxpf. I cannot figure out how to get a log because journalctl is not running when the kernel panicks.
The error is to the effect "bad: scheduling from idle thread!" CPU: 0 PID: 0 Comm: swapper/0 Tainted: nothing really interesting before or after and it doesn't list anything specifically. I also tried enabling Southern Islands in the kernel config. In the kernel config I used the default linux 4.8.

I have torn apart the initramfs with lsinitcpio and then did the same to my 4.8 that boots and they are the same. I can boot if I disable acpi at boot

Welcome to the linux 4.9 kernel is busted with some CPUs. Just look at these bugs reports :

https://bugs.archlinux.org/task/52246
https://bugs.archlinux.org/task/52271
https://bugs.archlinux.org/task/52238

Looks like 4.9 will be one the of worst kernel I've seen for years

And upstream bugs :

https://bugzilla.kernel.org/show_bug.cgi?id=192111
https://bugzilla.kernel.org/show_bug.cgi?id=191801

loqs

Member

Registered: 2014-03-06

Posts: 17,321

Re: Kernel boot problems

@fredbezies

Can we revert this patch and see if it boots correctly?

https://git.kernel.org/cgit/linux/kerne … 22b6d083f8

That patch is from 4.9.2 so it would not be present in 4.9 for you to revert.
Also how is you own bisection between 4.8 and 4.9 progressing?

fredbezies

Member

Registered: 2011-07-28

Posts: 353

Re: Kernel boot problems

@fredbezies

Can we revert this patch and see if it boots correctly?

https://git.kernel.org/cgit/linux/kerne … 22b6d083f8

That patch is from 4.9.2 so it would not be present in 4.9 for you to revert.
Also how is you own bisection between 4.8 and 4.9 progressing?

Ok. Thanks for the info. Bisecting ? I don't have time to do it, neither computing power to do so. It takes me 2 hours to build and homemade kernel. So, I think I'll have to wait until a working kernel with ACPI fixed as it is an ACPI bug for me is available.

And if linux 4.9 is the new LTS version, I think linux distributions will be in a very bad shape

fredbezies

Member

Registered: 2011-07-28

Posts: 353

Re: Kernel boot problems

Want another 4.9.x booting bug ? This one with UEFI : https://bugzilla.kernel.org/show_bug.cgi?id=191121

Linux 4.9.x is - as of january 10th, 2017 - one of the worst kernel I've seen in years in my linux-only computing life.

Not an ACPI bug like mine in fact.

Kernel team is doing a wrong thing telling that 4.8.17 is the last 4.8.x kernel... Looks like they're not aware of booting bug with 4.9.x...

https://lkml.org/lkml/2017/1/9/99

Last edited by fredbezies (2017-01-10 15:01:06)

loqs

Member

Registered: 2014-03-06

Posts: 17,321

Re: Kernel boot problems

Ok. Thanks for the info. Bisecting ? I don't have time to do it, neither computing power to do so. It takes me 2 hours to build and homemade kernel.

Assuming you sleep for more than 2 hours per day would that not allow you to perform one bisection cycle per sleep cycle or is the computer fully utilized while you sleep as well?
Without a backtrace or bisection to work off there is not much for upstream to work with.

Another upstream bug closed as unreproducible :

https://bugzilla.kernel.org/show_bug.cgi?id=188221

Or apply / revert the patch found in this bug report? Applied it, build a kernel and still no luck

Can you try the attached patch? It prints out all the watchdog resources so that we can see the actual conflict. Once you have the patch applied please attach dmesg to this bug.

What did you expect the patch to do?
Edit:
grammar missing with

Last edited by loqs (2017-01-10 15:13:31)

fredbezies

Member

Registered: 2011-07-28

Posts: 353

Re: Kernel boot problems

Ok. Thanks for the info. Bisecting ? I don't have time to do it, neither computing power to do so. It takes me 2 hours to build and homemade kernel.

Assuming you sleep for more than 2 hours per day would that not allow you to perform one bisection cycle per sleep cycle or is the computer fully utilized while you sleep as well?

One problem : there is only 3 meters between my bed and my main computer, the only one which works with linux 4.9 and AMD powered. So I have to shut it down if I want to sleep.

Without a backtrace or bisection to work off there is not much for upstream to work with.

Looks like there is already bugs related to acpi and linux 4.9 existing bugs. I'm ready to bet it is a bug which was added back in 4.9-rc1 kernel. As stated in this bug :

https://bugzilla.kernel.org/show_bug.cgi?id=191801

So, as I've not the possibility to bisect it by myself, I will wait until somebody else does it. I know it's wrong, but I prefer not to see my electricity bill exploding because of nightly bisecting.

Another upstream bug closed as unreproducible :

https://bugzilla.kernel.org/show_bug.cgi?id=188221

Or apply / revert the patch found in this bug report? Applied it, build a kernel and still no luck

Can you try the attached patch? It prints out all the watchdog resources so that we can see the actual conflict. Once you have the patch applied please attach dmesg to this bug.

What did you expect the patch to do?
Edit:
grammar missing with

I misread description. Oops. My bad. As I'm getting problems with ACPI, I wrongly thought it can help

loqs

Member

Registered: 2014-03-06

Posts: 17,321

Re: Kernel boot problems

Looks like there is already bugs related to acpi and linux 4.9 existing bugs. I'm ready to bet it is a bug which was added back in 4.9-rc1 kernel. As stated in this bug :

https://bugzilla.kernel.org/show_bug.cgi?id=191801

If the fix from https://bugzilla.kernel.org/show_bug.cgi?id=191121#c15 also fixes 191801 then that would indicate to me it is a duplicate and an EFI issue which would be different to your issue 192111.

fredbezies

Member

Registered: 2011-07-28

Posts: 353

Re: Kernel boot problems

Looks like there is already bugs related to acpi and linux 4.9 existing bugs. I'm ready to bet it is a bug which was added back in 4.9-rc1 kernel. As stated in this bug :

https://bugzilla.kernel.org/show_bug.cgi?id=191801

If the fix from https://bugzilla.kernel.org/show_bug.cgi?id=191121#c15 also fixes 191801 then that would indicate to me it is a duplicate and an EFI issue which would be different to your issue 192111.

Indeed. Releasing a kernel that may not boot with a lot of CPUs is not the greatest thing you can provide. Choosing it as a LTS version, another big mistake. Is free software world going mad ?

Edit : well, looks like my laptop will wait until linux 4.10 is released to go back to short time support kernel.

Just look at this other linux 4.9 related bug : https://bugs.freedesktop.org/show_bug.cgi?id=98690

Last edited by fredbezies (2017-01-10 17:46:07)

towerclimber

Member

Registered: 2015-05-01

Posts: 4

Re: Kernel boot problems

Just wanted to update this post with some new info.
I tried the testing kernel 4.9 and it also did not boot. Reminded me that Opensuse latest rolling release kernel also did not boot. Fedora 25 has a 4.8.6 kernel that boots fine, so it seems that I need to learn how to use git properly in order to check out certain commits?

Can anybody walk me through a quick setup? I say bypass the AUR and if you could show me how to check out a certain build that might work. I could learn so much just from that alone, as it would explain the current problem I am having.  I know I don't need to update but this is just a learning experience for me.

fredbezies

Member

Registered: 2011-07-28

Posts: 353

Re: Kernel boot problems

Looks like things are moving. Somebody bisected the kernel code and found a commit which can explain booting problems on a bug I opened upstream.

https://bugzilla.kernel.org/show_bug.cgi?id=192111

See this comment : https://bugzilla.kernel.org/show_bug.cgi?id=192111#c9

I'll try to apply it on linux 4.9.2 source code and build a new kernel. With a little luck, it'll boot.

loqs

Member

Registered: 2014-03-06

Posts: 17,321

Re: Kernel boot problems

I know you asked for a quick setup without AUR but this one is based on the linux-git package but optionally uses the configs from linux package in the testing repo which is also 4.9 based

$ git clone https://aur.archlinux.org/linux-git.git
# abs testing/linux    #for the configs from testing if you do not want them skip to makepkg -o
$ cp -r /var/abs/testing/linux .
$ cp ../linux/config.x86_64 .
$ cp ../linux/config .
$ cp ../linux/linux.preset linux-git.preset #the existing preset would also work
$ updpkgsums
$ makepkg -o #may want s option if you have missing dependencies 
$ cd src/linux/
$ git bisect start
$ git bisect bad v4.9
$ git bisect good v4.8
Bisecting: 8695 revisions left to test after this (roughly 13 steps)
[a5af7e1fc69a46f29b977fd4b570e0ac414c2338] rxrpc: Fix loss of PING RESPONSE ACK production due to PING ACKs
$ cd ../..
$ makepkg -e #should produce Finished making: linux-git 4.8.r8696.ga5af7e1fc69a-1

Install kernel add option to boot loader if needed and test
Then repeat the following until git bisect has found the first bad commit

$ cd linux-git/src/linux
$ git bisect $result #replace $result with good or bad
$ cd ../..
$ makepkg -e

Edit:
Would also recomend setting Makepkg#MAKEFLAGS to speedup builds.
@fredbezies I hope it works

Last edited by loqs (2017-01-11 09:42:13)

fredbezies

Member

Registered: 2011-07-28

Posts: 353

Re: Kernel boot problems

Report done after nearly a 3 hours building process : it boots! At least, a freeze is kinda "fixed" by removing this commit : https://git.kernel.org/cgit/linux/kerne … ae502cae8e

Let's hope there will be a fix for this bug as linux 4.9 is set to be the next LTS kernel...

Edit : I'm writing the "modified" 4.9.2 linux kernel I built this morning installed on my main computer, an AthlonX2-215 based.. So far, so good.

Last edited by fredbezies (2017-01-11 13:57:23)