Bug 1274403 - GNOME Shell crashes when booted in a KVM with qemu32 CPU
GNOME Shell crashes when booted in a KVM with qemu32 CPU
Product: Fedora
Classification: Fedora
Component: llvm (Show other bugs)
i686 Linux
unspecified Severity medium
: ---
: ---
Assigned To: Adam Jackson
Fedora Extras Quality Assurance
: 1192017 (view as bug list)
Depends On:
  Show dependency treegraph
Reported: 2015-10-22 12:14 EDT by Adam Williamson
Modified: 2016-12-20 10:06 EST (History)
19 users (show)

See Also:
Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2016-12-20 10:06:48 EST
Type: Bug
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)

External Trackers
Tracker ID Priority Status Summary Last Updated
Debian BTS 775235 None None None 2016-01-04 11:54 EST
CentOS 8748 None None None 2016-01-04 12:30 EST

  None (edit)
Description Adam Williamson 2015-10-22 12:14:50 EDT
Note: the component is a guess for now, I just want to get this bug filed for the Go/No-Go meeting, will investigate more soon.

If you boot a 32-bit Fedora 23 Workstation live image - this has been happening ever since we started testing 32-bit, so might well be from F22 or earlier - on a KVM with the CPU model set to 'qemu32' and all other settings default (i.e. QXL/SPICE graphics), it boots to Shell's crash screen ("Oh no! Something has gone wrong.") This doesn't happen if the CPU is set to a 64-bit model like Westmere. We don't know yet if anything like this affects real 32-bit hardware, I've asked jsedlak to test.
Comment 1 Mike Ruckman 2015-10-22 12:18:57 EDT
I can confirm that it installed on bare metal for me.
Comment 2 Adam Williamson 2015-10-22 12:20:08 EDT
OK, not gonna bother proposing as a blocker/FE then, qemu32 is a fairly artificial scenario. Still, I'll look into it.
Comment 3 Adam Williamson 2015-10-22 12:21:49 EDT
Happens with VGA/VNC too, so it ain't qxl.
Comment 4 Adam Williamson 2015-10-22 12:27:29 EDT
OK, another guess - I see this in the log:

gnome-session[1697]: LLVM ERROR: Do not know how to split the result of this operator!

If I go to a VT and run gnome-shell from there I get this:

current session already has an ibus-daemon.

(gnome-shell:3298): mutter-WARNING **: STACK_OP_RAISE_ABOVE: window 0x5a01e00015 not in stack

(gnome-shell:3298): Gjs-WARNING **: JS ERROR: could not get remote objects for service org.gnome.SettingsDaemon.Smartcard path /org/gnome/SettingsDaemon/Smartcard: Gio.DBusError: GDBus.Error:org.freedesktop.DBus.Error.ServiceUnknown: The name org.gnome.SettingsDaemon.Smartcard was not provided by any .service files

(gnome-shell:3298): mutter-WARNING **: STACK_OP_RAISE_ABOVE: window 0x5a01e00015 not in stack
LLVM ERROR: Do not know how to split the result of this operator!
Comment 5 Tim Flink 2015-10-22 12:42:00 EDT
The F23 RC2 workstation i686 live image booted fine from litd-created USB media on my 32-bit P4.

FWIW, the graphics adapter is a GeForce 6200 AGP card.
Comment 6 Adam Williamson 2015-10-22 13:02:08 EDT
Here's the /proc/cpuinfo for the qemu32 CPU:

processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 6
model name	: QEMU Virtual CPU version
stepping	: 3
microcode	: 0x1
cpu MHz		: 3502.620
cache size	: 4096 KB
physical id	: 0
siblings	: 1
core id		: 0
cpu cores	: 1
apicid		: 0
initial apicid	: 0
fdiv_bug	: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 4
wp		: yes
flags		: fpu de pse tsc msr pae mce cx8 apic sep pge cmov mmx fxsr sse sse2 pni x2apic popcnt hypervisor
bugs		:
bogomips	: 7005.24
clflush size	: 32
cache_alignment	: 32
address sizes	: 36 bits physical, 32 bits virtual
power management:
Comment 7 Tim Flink 2015-10-22 13:07:00 EDT
I tried again, both using "Basic Graphics Mode" and adding nomodeset to the boot args. Both times, I was able to get to GNOME shell without issue and the system seems to work fine under light usage.
Comment 8 Tim Flink 2015-10-22 13:10:02 EDT
/proc/cpuinfo from my test machine (omitted the first entry as it's pretty much identical)

processor	: 1
vendor_id	: GenuineIntel
cpu family	: 15
model		: 2
model name	: Intel(R) Pentium(R) 4 CPU 2.60GHz
stepping	: 9
microcode	: 0x2e
cpu MHz		: 2605.840
cache size	: 512 KB
physical id	: 0
siblings	: 2
core id		: 0
cpu cores	: 1
apicid		: 1
initial apicid	: 1
fdiv_bug	: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 2
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe pebs bts cid xtpr
bugs		:
bogomips	: 5210.88
clflush size	: 64
cache_alignment	: 128
address sizes	: 36 bits physical, 32 bits virtual
power management:
Comment 9 Kevin Kofler 2016-01-04 11:53:38 EST
*** Bug 1192017 has been marked as a duplicate of this bug. ***
Comment 10 Kevin Kofler 2016-01-04 11:59:14 EST
So I have bad news for you:
* This is a core LLVM/llvmpipe issue, which probably affects much more than just gnome-shell.
* This probably also affects real hardware without SSE2, and there there is no easy workaround. (For QEMU, you can just use --copy-host-cpu and it should work, assuming of course that the host CPU does SSE2 to begin with.)
* This is a regression in LLVM 3.5, it affects Fedora ≥ 21 and RHEL ≥ 7. (According to the Debian bug, this worked fine with LLVM 3.4, and rebuilding mesa against a downgraded LLVM "fixed" it for them.)
* The only workaround (other than downgrading LLVM) that Debian has to offer is to change mesa to trust the CPU flags over the CPU model, which fixes things for QEMU, but is of course no help at all for real non-SSE2 hardware.
Comment 11 Kevin Kofler 2016-01-04 12:33:04 EST
I'm removing https://bugs.debian.org/776911 which is actually a mixed duplicate of https://bugs.debian.org/775235 (i.e. this bug) and https://bugs.debian.org/770130 (an unrelated issue with Intel GPUs).
Comment 12 Adam Williamson 2016-01-04 12:48:07 EST
Thanks for the info - do we have an actual *upstream* llvm bug report for this, as opposed to just a report from another distro (which is a different downstream)?
Comment 13 Kevin Kofler 2016-01-04 14:09:30 EST
There's this one:
which sounds like the same issue, but what's interesting is that that CPU supports sse2 according to the cpuinfo. But since Mesa takes the CPU family rather than the full list of flags, it might be that it picks a non-sse2 CPU family (something too generic like "athlon"), not sure.

(There are really 2 issues at play here: The one that Mesa/llvmpipe does not always recognize SSE2 when available, and the one that LLVM errors when generating the non-SSE2 code.)
Comment 14 Kevin Kofler 2016-01-04 14:11:20 EST
Also of interest would be whether this also affects other OpenGL applications (plasma-desktop and other QML 2 applications, OpenGL games, etc.) or whether it's something that gnome-shell specifically does that triggers the bug.
Comment 15 Fedora End Of Life 2016-11-24 07:52:44 EST
This message is a reminder that Fedora 23 is nearing its end of life.
Approximately 4 (four) weeks from now Fedora will stop maintaining
and issuing updates for Fedora 23. It is Fedora's policy to close all
bug reports from releases that are no longer maintained. At that time
this bug will be closed as EOL if it remains open with a Fedora  'version'
of '23'.

Package Maintainer: If you wish for this bug to remain open because you
plan to fix it in a currently maintained version, simply change the 'version' 
to a later Fedora version.

Thank you for reporting this issue and we are sorry that we were not 
able to fix it before Fedora 23 is end of life. If you would still like 
to see this bug fixed and are able to reproduce it against a later version 
of Fedora, you are encouraged  change the 'version' to a later Fedora 
version prior this bug is closed as described in the policy above.

Although we aim to fix as many bugs as possible during every release's 
lifetime, sometimes those efforts are overtaken by events. Often a 
more recent Fedora release includes newer upstream software that fixes 
bugs or makes them obsolete.
Comment 16 Fedora End Of Life 2016-12-20 10:06:48 EST
Fedora 23 changed to end-of-life (EOL) status on 2016-12-20. Fedora 23 is
no longer maintained, which means that it will not receive any further
security or bug fix updates. As a result we are closing this bug.

If you can reproduce this bug against a currently maintained version of
Fedora please feel free to reopen this bug against that version. If you
are unable to reopen this bug, please file a new report against the
current release. If you experience problems, please add a comment to this

Thank you for reporting this bug and we are sorry it could not be fixed.

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