vmwgfx crash after Linux update to 6.10.4-arch2-1

gromit · 2024-08-14 21:25:19

$ git bisect good
af6441e6f3d41e95bfc5bfc11960c259bb4f0f11 is the first bad commit
commit af6441e6f3d41e95bfc5bfc11960c259bb4f0f11
Author: Zack Rusin <zack.rusin@broadcom.com>
Date:   Mon Jul 22 14:41:15 2024 -0400

    drm/vmwgfx: Fix handling of dumb buffers
    
    commit d6667f0ddf46c671d379cd5fe66ce0a54d2a743a upstream.
    
    Dumb buffers can be used in kms but also through prime with gallium's
    resource_from_handle. In the second case the dumb buffers can be
    rendered by the GPU where with the regular DRM kms interfaces they
    are mapped and written to by the CPU. Because the same buffer can
    be written to by the GPU and CPU vmwgfx needs to use vmw_surface (object
    which properly tracks dirty state of the guest and gpu memory)
    instead of vmw_bo (which is just guest side memory).
    
    Furthermore the dumb buffer handles are expected to be gem objects by
    a lot of userspace.
    
    Make vmwgfx accept gem handles in prime and kms but internally switch
    to vmw_surface's to properly track the dirty state of the objects between
    the GPU and CPU.
    
    Fixes new kwin and kde on wayland.
    
    Signed-off-by: Zack Rusin <zack.rusin@broadcom.com>
    Fixes: b32233acceff ("drm/vmwgfx: Fix prime import/export")
    Cc: Broadcom internal kernel review list <bcm-kernel-feedback-list@broadcom.com>
    Cc: dri-devel@lists.freedesktop.org
    Cc: <stable@vger.kernel.org> # v6.9+
    Reviewed-by: Maaz Mombasawala <maaz.mombasawala@broadcom.com>
    Reviewed-by: Martin Krastev <martin.krastev@broadcom.com>
    Link: https://patchwork.freedesktop.org/patch/msgid/20240722184313.181318-4-zack.rusin@broadcom.com
    Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Could you please verify that the following image that has the above commit reverted works?

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-6.10.4.arch2-1.1-x86_64.pkg.tar.zst

loqs · 2024-08-14 21:49:09

Is the issue also present in 6.11-rc3?

gromit · 2024-08-14 21:52:00

You can get the version loqs is asking about here:

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-mainline-6.11rc3-1-x86_64.pkg.tar.zst

rdkehn · 2024-08-15 00:23:57

Apologizes, my day job kept me away today.

Linux 6.10.4-arch2-1.1 #1 SMP PREEMPT_DYNAMIC Wed, 14 Aug 2024 21:15:21 +0000

works

gromit · 2024-08-15 08:51:10

Also could you please test this version of the mainline kernel with the patch reverted?

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-mainline-6.11rc3-1.1-x86_64.pkg.tar.zst

I would expect the mainline release with it to fail and the one without it to work.

How do we go about bug reporting? Should I write the report and credit you or do we do it the other way around?

pieska · 2024-08-15 10:55:25

https://pkgbuild.com/\~gromit/linux-bis … kg.tar.zst

works but uname reports

Linux bisect 6.10.4-arch2-1.1 #1 SMP PREEMPT_DYNAMIC Wed, 14 Aug 2024 21:15:21 +0000 x86_64 GNU/Linux

gromit · 2024-08-15 11:07:41

Yeah you booted the wrong thing, the packages with linux-mainline in the name install their kernel images under a different name, so you'll have to teach your bootloader how to boot them.

pieska · 2024-08-15 11:12:30

D'oh!

Linux bisect 6.11.0-rc3-1.1-mainline-dirty #1 SMP PREEMPT_DYNAMIC Thu, 15 Aug 2024 08:28:36 +0000 x86_64 GNU/Linux

works

pieska · 2024-08-15 11:18:10

gromit wrote:

You can get the version loqs is asking about here:
sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-mainline-6.11rc3-1-x86_64.pkg.tar.zst

Linux bisect 6.11.0-rc3-1-mainline #1 SMP PREEMPT_DYNAMIC Mon, 12 Aug 2024 07:01:21 +0000 x86_64 GNU/Linux

fails

so it's not fixed in rc3

gromit · 2024-08-15 11:52:17

Great, that is good to know! Again to my question above, do you want to write the bug report or should I take care of that?
If I should do it, could you give me an email adress and a name so I can CC you and credit you for reporting the bug first?

rdkehn · 2024-08-15 12:23:15

Linux 6.11.0-rc3-1.1-mainline-dirty #1 SMP PREEMPT_DYNAMIC Thu, 15 Aug 2024 08:28:36 +0000 x86_64 GNU/Linux

Works

Linux 6.11.0-rc3-1-mainline #1 SMP PREEMPT_DYNAMIC Mon, 12 Aug 2024 07:01:21 +0000 x86_64 GNU/Linux

Does not work.

Please write the bug report and take credit (you did all the work). You can CC rdkehn@gmail.com if you wish.

gromit · 2024-08-15 18:04:18

I have just sent out the report: https://lore.kernel.org/all/0d0330f3-2a … @heusel.eu

gromit · 2024-08-15 19:06:18

Did you see the answer from the maintainer already? they would like a full journal output, information on what runs on the machine and the vmware.log ..
If you want you can reply directly to the mail or post them here and have me forward them

Last edited by gromit (2024-08-15 19:06:38)

loqs · 2024-08-17 00:18:51

@gromit I assume you will be doing a build with the proposed fix applied?

gromit · 2024-08-17 08:29:12

This is the linux package with the fix from https://lore.kernel.org/all/20240816183 … oadcom.com applied:

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-6.10.5.arch1-1.1-x86_64.pkg.tar.zst

Could you please test it?

pieska · 2024-08-18 18:34:25

Linux bisect 6.10.5-arch1-1.1 #1 SMP PREEMPT_DYNAMIC Sat, 17 Aug 2024 08:13:20 +0000 x86_64 GNU/Linux

works, thanks

CharlesMeng · 2024-08-19 02:11:07

This also affected the ISO images built with latest kernel.

The next official ISO 2024-09-01 could have the problem.

pieska · 2024-08-21 14:33:09

6.10.6-arch1-1 is unpatched and doesn't work

loqs · 2024-08-25 21:34:33

CharlesMeng wrote:

This also affected the ISO images built with latest kernel.
The next official ISO 2024-09-01 could have the problem.

pieska wrote:

6.10.6-arch1-1 is unpatched and doesn't work

Please consider opening an issue on Arch's gitlab instance against the linux package asking for the fix to be be backported.
Edit:
https://git.kernel.org/pub/scm/linux/ke … d30290c5e4 not included in linux 6.10.7.arch1-1.

Last edited by loqs (2024-08-30 18:28:04)

pieska · 2024-09-01 09:39:35

will do, wait for my account creation request to be processed.

loqs · 2024-09-01 13:39:46

https://git.kernel.org/pub/scm/linux/ke … t-3d.patch is queued for 6.10.8.

Arch Linux

#26 2024-08-14 21:25:19

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#27 2024-08-14 21:49:09

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#28 2024-08-14 21:52:00

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#29 2024-08-15 00:23:57

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#30 2024-08-15 08:51:10

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#31 2024-08-15 10:55:25

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#32 2024-08-15 11:07:41

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#33 2024-08-15 11:12:30

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#34 2024-08-15 11:18:10

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#35 2024-08-15 11:52:17

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#36 2024-08-15 12:23:15

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#37 2024-08-15 18:04:18

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#38 2024-08-15 19:06:18

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#39 2024-08-17 00:18:51

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#40 2024-08-17 08:29:12

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#41 2024-08-18 18:34:25

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#42 2024-08-19 02:11:07

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#43 2024-08-21 14:33:09

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#44 2024-08-25 21:34:33

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#45 2024-09-01 09:39:35

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

#46 2024-09-01 13:39:46

Re: vmwgfx crash after Linux update to 6.10.4-arch2-1

Board footer