Bug 1732113 - Recent Rawhide kernels (rc0.git7 and rc1) have problems with qemu 'std' / 'VGA' driver
Summary: Recent Rawhide kernels (rc0.git7 and rc1) have problems with qemu 'std' / 'VG...
Keywords:
Status: NEW
Alias: None
Product: Fedora
Classification: Fedora
Component: kernel
Version: rawhide
Hardware: All
OS: Linux
unspecified
medium
Target Milestone: ---
Assignee: Kernel Maintainer List
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard: openqa
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2019-07-22 17:52 UTC by Adam Williamson
Modified: 2019-08-02 06:37 UTC (History)
18 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed:
Type: Bug
Embargoed:


Attachments (Terms of Use)

Description Adam Williamson 2019-07-22 17:52:08 UTC
Since kernel-5.3.0-0.rc0.git7.1.fc31 landed in Rawhide, in the Fedora-Rawhide-20190720.n.1 compose, almost all openQA tests are failing. Investigating this manually, it seems to be due to issues with the qemu graphics driver which qemu calls 'std' and virt-manager calls 'VGA'.

With rc0.git7, there seem to be issues with VTs. Booting to runlevel 3 gets stuck at a black screen. Similarly, trying to switch to a VT from a running graphical desktop does not work; openQA shows a black screen when it tries to do this, when I try it in a local VM, I see some of the GNOME desktop elements disappear but then the screen freezes (cursor not moveable). Sometimes I can successfully switch back to the running GNOME session with ctrl-alt-f2, sometimes I can't.

With rc1.git0.1, my VM won't even boot to GNOME any more if the graphics are set to 'VGA'. It just gets stuck at a boot splash screen.

If I change the VM to use qxl or virtio graphics, everything is fine, with both kernels.

This is a problem because openQA runs into other issues if I set it to use qxl or virtio graphics. The last one we had with qxl was https://bugzilla.redhat.com/show_bug.cgi?id=1638578 , the last one we had with virtio was https://bugzilla.redhat.com/show_bug.cgi?id=1638578 . I guess I'll have to try out qxl again and see what happens.

Comment 1 Adam Williamson 2019-07-22 17:53:03 UTC
huh, actually, rc1 doesn't seem to boot even with virtio graphics. But that's probably a different problem.

Comment 2 Cole Robinson 2019-07-23 20:39:14 UTC
I see Gerd had a patch to vga/bochs recently, but it should be in 5.3-rc1 which you also report issues with

commit dd2b5959785c75c23ab2295a4095700e47dabff3
Author: Gerd Hoffmann <kraxel>
Date:   Thu Jun 27 10:12:06 2019 +0200

    drm/bochs: fix framebuffer setup.


Gerd, any thoughts?

Comment 3 Adam Williamson 2019-07-23 20:44:42 UTC
The failure I saw when testing RC1 was probably this SCSI thing:

https://marc.info/?l=linux-scsi&m=156378725427719&w=2

so unrelated. I'll try and find or build a kernel with both things fixed, or see if I can hack up a VM to avoid the SCSI problem with RC1 so I can check if this video problem is gone.

Comment 4 Adam Williamson 2019-07-23 20:55:08 UTC
OK, nope, still seems broken with 5.3.0-0.rc1.git0.1.fc31.x86_64 (I got it booting in my test VM).

Comment 5 Adam Williamson 2019-07-23 22:13:57 UTC
5.3.0-0.rc1.git1.1.fc31.x86_64 also still broken.

Comment 7 Adam Williamson 2019-08-01 16:40:09 UTC
Thanks for that. virtio has been working quite well, so I might stick with it long term, but if someone gives me a heads-up when those fixes are in a Fedora kernel build I can kick staging back to std to test them.

Comment 8 Gerd Hoffmann 2019-08-02 06:37:08 UTC
Fixes are in drm-misc-fixes branch now.
Might be too late for -rc3, but -rc4 should be fixed.


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