Bug 1921523

Summary: sddm-greeter crashes in VMware Fusion or VMware Workstation when hypervisor 3D acceleration is enabled
Product: [Fedora] Fedora Reporter: Neal Gompa <ngompa13>
Component: mesaAssignee: Adam Jackson <ajax>
Status: CLOSED ERRATA QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: 33CC: ajax, aleixpol, bskeggs, caillon+fedoraproject, cpowell, igor.raits, jglisse, jgrulich, kde-sig, lyude, me, m, nixuser, pierluigi.fiorini, rclark, rdieter, rhughes, rstrode, sergio, tstellar
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: mesa-20.3.4-2.fc33 Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2021-02-26 01:08:53 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
sddm-greeter coredump
none
sddm-greeter coredump on Rawhide
none
gdb backtrace output from the coredump on Rawhide none

Description Neal Gompa 2021-01-28 05:56:12 UTC
Created attachment 1751574 [details]
sddm-greeter coredump

Description of problem:
After updating to sddm-0.19.0-3.fc33, I am no longer able to login to my Plasma session if I have VMware configured to have accelerated 3D support enabled (which is the default for VMware when creating Fedora VMs). If I disable accelerated 3D support, then it falls back to LLVMpipe and everything works.

Version-Release number of selected component (if applicable):
0.19.0-3.fc33

How reproducible:
Always

Steps to Reproduce:
1. Install Fedora 33 in VMware
2. Apply sddm-0.19.0-3.fc33 update
3. Reboot

Actual results:
System boots to a black screen (greeter crashes in the background)

Expected results:
System boots into SDDM, allowing login

Additional info:
ABRT refused to report this, saying server reported "invalid uReport" or some such, so I've attached a coredump to this for manual evaluation.

Comment 1 Neal Gompa 2021-01-29 03:53:42 UTC
Filed issue upstream, just in case.

Comment 2 Christopher Powell 2021-02-12 18:00:59 UTC
I can confirm the described behavior here.

Comment 3 Sergio Basto 2021-02-13 19:41:54 UTC
[   13.630969] QSGRenderThread[993]: segfault at 0 ip 0000000000000000 sp 00007f4203ffea68 error 14
in sddm-greeter[55a60778e000+14000]
[   13.630987] Code: Unable to access opcode bytes at RIP 0xffffffffffffffd6.

but if I start my vm with kdm , error appears every where , so I deduce the problem is not on sddm itself is in kde or Qt5

Comment 4 Neal Gompa 2021-02-16 01:05:50 UTC
Created attachment 1757181 [details]
sddm-greeter coredump on Rawhide

I've added a new coredump from a Rawhide run where I have all the debuginfo installed.

Comment 5 Neal Gompa 2021-02-16 01:07:50 UTC
Created attachment 1757182 [details]
gdb backtrace output from the coredump on Rawhide

I've attached the full output from gdb when attempting to get a backtrace from the coredump.

From the log, the backtrace seems to indicate that this *might* be a Qt problem?

(gdb) bt
#0  0x0000000000000000 in ?? ()
#1  0x00007ff559e5c9c6 in QGLXContext::makeCurrent (this=this@entry=0x55e8b69b80b0, surface=0x55e8b7145fb0) at qglxintegration.cpp:611
#2  0x00007ff56d584c8e in QOpenGLContext::makeCurrent (this=0x55e8b6becbf0, surface=0x55e8b69a1fc0) at kernel/qopenglcontext.cpp:992
#3  0x00007ff56e2d3a58 in QSGRenderThread::run (this=0x55e8b71b0ce0) at scenegraph/qsgthreadedrenderloop.cpp:1035
#4  0x00007ff56cf78751 in QThreadPrivate::start (arg=0x55e8b71b0ce0) at thread/qthread_unix.cpp:329
#5  0x00007ff56c608269 in start_thread (arg=0x7ff53ee28640) at pthread_create.c:473
#6  0x00007ff56cb89663 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

Comment 6 Neal Gompa 2021-02-16 01:59:48 UTC
This is with qt5-qtbase-5.15.2-13.fc34, for reference.

Comment 7 Neal Gompa 2021-02-17 00:36:41 UTC
Some further digging and step-by-step checking of the graphics stack has led me to identify that the culprit is Mesa.

Starting from Fedora 33 GA and upgrading everything *except* mesa results in a working system.

Then, stepping through the Bodhi updates, I pinpointed the update that broke everything: the update to mesa-20.3.3-3.fc33.

On mesa-20.2.6-1.fc33[1], it works correctly and SDDM loads, but once updated to mesa-20.3.3-3.fc33[2], it crashes as reported.

[1]: https://bodhi.fedoraproject.org/updates/FEDORA-2020-f04e9f1683
[2]: https://bodhi.fedoraproject.org/updates/FEDORA-2021-cd899639ab

Comment 8 Neal Gompa 2021-02-17 00:40:13 UTC
This is also still happening on Mesa 21~rc2 in F34 and Rawhide. :(

Comment 9 Christopher Powell 2021-02-17 00:42:51 UTC
(In reply to Neal Gompa from comment #7)
> Some further digging and step-by-step checking of the graphics stack has led
> me to identify that the culprit is Mesa.
> 
> Starting from Fedora 33 GA and upgrading everything *except* mesa results in
> a working system.
> 
> Then, stepping through the Bodhi updates, I pinpointed the update that broke
> everything: the update to mesa-20.3.3-3.fc33.
> 
> On mesa-20.2.6-1.fc33[1], it works correctly and SDDM loads, but once
> updated to mesa-20.3.3-3.fc33[2], it crashes as reported.
> 
> [1]: https://bodhi.fedoraproject.org/updates/FEDORA-2020-f04e9f1683
> [2]: https://bodhi.fedoraproject.org/updates/FEDORA-2021-cd899639ab

It seems worth exploring whether a downgrade from 20.3.3 -> 20.2.6 is even possible, and if the problem goes away after doing so. Any thoughts on such a process?

Comment 10 Neal Gompa 2021-02-17 01:18:23 UTC
(In reply to Christopher Powell from comment #9)
> (In reply to Neal Gompa from comment #7)
> > Some further digging and step-by-step checking of the graphics stack has led
> > me to identify that the culprit is Mesa.
> > 
> > Starting from Fedora 33 GA and upgrading everything *except* mesa results in
> > a working system.
> > 
> > Then, stepping through the Bodhi updates, I pinpointed the update that broke
> > everything: the update to mesa-20.3.3-3.fc33.
> > 
> > On mesa-20.2.6-1.fc33[1], it works correctly and SDDM loads, but once
> > updated to mesa-20.3.3-3.fc33[2], it crashes as reported.
> > 
> > [1]: https://bodhi.fedoraproject.org/updates/FEDORA-2020-f04e9f1683
> > [2]: https://bodhi.fedoraproject.org/updates/FEDORA-2021-cd899639ab
> 
> It seems worth exploring whether a downgrade from 20.3.3 -> 20.2.6 is even
> possible, and if the problem goes away after doing so. Any thoughts on such
> a process?

It worked in my VMs, so it's possible. But I'd rather have this properly fixed, since it also affects Mesa 21 in Fedora 34/Rawhide...

Comment 11 Sergio Basto 2021-02-17 01:55:34 UTC
Nice catch , I can confirm I downgrade to mesa-20.2.3-2.fc34 and 3D acceleration starts to work again

Comment 12 Neal Gompa 2021-02-20 13:18:49 UTC
This is still happening with mesa-21~rc5. :(

Comment 13 Fedora Update System 2021-02-22 06:35:07 UTC
FEDORA-2021-e7e5787903 has been submitted as an update to Fedora 33. https://bodhi.fedoraproject.org/updates/FEDORA-2021-e7e5787903

Comment 14 Fedora Update System 2021-02-24 21:54:56 UTC
FEDORA-2021-e7e5787903 has been pushed to the Fedora 33 testing repository.
Soon you'll be able to install the update with the following command:
`sudo dnf upgrade --enablerepo=updates-testing --advisory=FEDORA-2021-e7e5787903`
You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2021-e7e5787903

See also https://fedoraproject.org/wiki/QA:Updates_Testing for more information on how to test updates.

Comment 15 Fedora Update System 2021-02-26 01:08:53 UTC
FEDORA-2021-e7e5787903 has been pushed to the Fedora 33 stable repository.
If problem still persists, please make note of it in this bug report.