Arch Linux

MALDATA

Member

Registered: 2012-08-06

Posts: 7

Kernel 5.7.2.arch1-1 appears to break web browsers?

I can't seem to find any other symptoms, but it looks like 5.7.2.arch1-1 makes my browsers go nuts. 5.6.15.arch1-1 works fine, but once I update to 5.7.2, chrome freezes up immediately when starting. Falkon does as well. The whole system becomes unstable when either browser starts. Dillo is the only one that worked ok. Other traffic, like through wget and curl, seems fine.

I'm running Arch on an Acer C720 chromebook. Anyone else experiencing this, or something like it?

hhg

Member

Registered: 2020-06-16

Posts: 3

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

Hello! Can you confirm the issue is not with Chromium based browsers? Could you try FF/WaterFox or terminal based browsers?

I ask you this, since the two broken browsers you mention are both Chormium-based, while Dillo does not seem to be (AFAIK).

Cheers!

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 21,740

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

The kernel doesn't inherently break browsers no... What could be a difference is that graphics drivers or cpu microcodes have issues, to check that you provide too little information on your system state, you might want to post/look through your journal on issue reproduction e.g.

sudo journalctl -b

hhg

Member

Registered: 2020-06-16

Posts: 3

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

The kernel doesn't inherently break browsers no...

Is it possible that the problem is rather with Chromium? For example, the kernel made some changes to APIs used by Chromium, now it breaks. Asking this since not all browsers are broken.

It could also be that when upgrading the system, a package, different from the kernel, which had been updated in the same cycle is the one causing the errors. This of course, only if OP did a system upgrade, rather than just a kernel one.

Edit: Formatting.

Last edited by hhg (2020-06-16 08:01:54)

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 21,740

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

Technically yes, practically that the kernel breaks userspace would be a grave bug that would have to be fixed for the kernel regardless.  It "should" never happen that the kernel breaks userspace, if it does, even if userspace did use something "incorrectly", the kernel devs should/will fix it since it used to work.

Other libraries leading to issues could also be the cause (a potential recent and often occurring candidate are new mesa versions for the OpenGL drivers) , but it's indeed quite likely that the new kernel has a bug that leads to issues (there are few pieces of software that have that invasive changes like a new major kernel releases so on big kernel changes it can often be inferred that they are a potential cause), but again we do have too little information from MALDATA that would make a proper assessment of the cause possible.

FWIW I do have no issues with chromium and the 5.7.2 kernel. So at the very least it isn't a systematic issue.

hhg

Member

Registered: 2020-06-16

Posts: 3

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

Technically yes, practically that the kernel breaks userspace would be a grave bug that would have to be fixed for the kernel regardless.  It "should" never happen that the kernel breaks userspace, if it does, even if userspace did use something "incorrectly", the kernel devs should/will fix it since it used to work...

Many thanks for the detailed explanation, really nice to learn something new and to have a bit more insight into how the kernel is made and how programs use it.

FWIW I do have no issues with chromium and the 5.7.2 kernel. So at the very least it isn't a systematic issue.

Well... I am out of ideas then

zpg443

Member

Registered: 2016-12-03

Posts: 271

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

The 5.7.2 kernel is not currently compatible with Intel systems that have the i915 GPU (which is a lot of systems).

https://gitlab.freedesktop.org/drm/intel/-/issues/2024

MALDATA

Member

Registered: 2012-08-06

Posts: 7

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

Hello! Can you confirm the issue is not with Chromium based browsers? Could you try FF/WaterFox or terminal based browsers?

I can confirm that Firefox also behaves fine (in addition to Dillo). Right now, I've got Firefox and Chrome both running at the same time. CPU load is low, but Chrome is unresponsive. I can't even type in a URL. Firefox is working fine, though.

The kernel doesn't inherently break browsers no... What could be a difference is that graphics drivers or cpu microcodes have issues, to check that you provide too little information on your system state, you might want to post/look through your journal

Yeah, I didn't really know where to start with this, so you're right, I didn't give you much to work with. I don't really associate journalctl with driver issues, but you nailed it. From journalctl:

Jun 15 20:55:59 aquila kernel: audit: type=1130 audit(1592272559.141:83): pid=1 uid=0 auid=4294967295 ses=4294967295 msg='unit=upower comm="systemd" exe="/usr/lib/systemd/systemd" hostname=? addr=? t>
Jun 15 20:56:04 aquila kernel: i915 0000:00:02.0: GPU HANG: ecode 7:1:8edcfc79, in Xorg [505]
Jun 15 20:56:04 aquila kernel: GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Jun 15 20:56:04 aquila kernel: Please file a _new_ bug report at https://gitlab.freedesktop.org/drm/intel/issues/new.
Jun 15 20:56:04 aquila kernel: Please see https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs for details.
Jun 15 20:56:04 aquila kernel: drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Jun 15 20:56:04 aquila kernel: The GPU crash dump is required to analyze GPU hangs, so please always attach it.
Jun 15 20:56:04 aquila kernel: GPU crash dump saved to /sys/class/drm/card0/error
Jun 15 20:56:04 aquila kernel: i915 0000:00:02.0: Resetting chip for stopped heartbeat on rcs0
Jun 15 20:56:05 aquila kernel: i915 0000:00:02.0: Xorg[505] context reset due to GPU hang
Jun 15 20:56:07 aquila kernel: i915 0000:00:02.0: GPU HANG: ecode 7:1:8edcfc7b, in chrome [9729]
Jun 15 20:56:07 aquila kernel: i915 0000:00:02.0: Resetting chip for stopped heartbeat on rcs0
Jun 15 20:56:08 aquila kernel: i915 0000:00:02.0: chrome[9729] context reset due to GPU hang
Jun 15 20:56:13 aquila kernel: i915 0000:00:02.0: GPU HANG: ecode 7:1:85ddfffd
Jun 15 20:56:13 aquila kernel: i915 0000:00:02.0: GPU HANG: ecode 7:1:85ddfffd
Jun 15 20:56:13 aquila kernel: i915 0000:00:02.0: Resetting chip for stopped heartbeat on bcs0

So... I guess that's it. GPU bug.

It could also be that when upgrading the system, a package, different from the kernel, which had been updated in the same cycle is the one causing the errors.

Yeah, I rolled back all of the packages that got updated and tested them individually. At first I thought it might be mesa, but nope. It's definitely just the kernel.

Thanks, all!

Trilby

Inspector Parrot

Registered: 2011-11-29

Posts: 29,532

Website

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

The 5.7.2 kernel is not currently compatible with Intel systems that have the i915 GPU (which is a lot of systems).

I assume this only applies to discrete intel GPUs?  I have a integrated intel which uses i915, am running 5.7.2, and have no issues with webengine / chrome browsers.

---

"UNIX is simple and coherent..." - Dennis Ritchie, "GNU's Not UNIX" -  Richard Stallman

MALDATA

Member

Registered: 2012-08-06

Posts: 7

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

The 5.7.2 kernel is not currently compatible with Intel systems that have the i915 GPU (which is a lot of systems).

Whoa... did not know that. Thanks for the heads-up. I use this little laptop pretty much exclusively for SSHing into a another machine, so it doesn't do much that needs the GPU (except for browsers). I was having some trouble pinning down why that was the only thing affected, but this definitely makes sense now.

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 21,740

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

It's definitely not all i915 users, because that would be everyone, certain CPU generations seem affected, from what I've seen/read so far most of the failing systems were Haswell based.

Last edited by V1del (2020-06-17 13:41:46)

zpg443

Member

Registered: 2016-12-03

Posts: 271

Re: Kernel 5.7.2.arch1-1 appears to break web browsers?

Thanks for pointing out that the issue is specific to i915 GPUs on Haswell-based CPUs, ones with HD Graphics, like the Pentium G3258. Yes, there are a lot.