Bug 1467037

Summary: Kernel module crash (amdgpu) on attempting to unbind AMD graphics in kernel 4.11 (regression from 4.10)
Product: [Fedora] Fedora Reporter: York Possemiers <ypossem>
Component: xorg-x11-drv-atiAssignee: X/OpenGL Maintenance List <xgl-maint>
Status: CLOSED CURRENTRELEASE QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: unspecified    
Version: 25CC: gansalmon, ichavero, itamar, jonathan, kernel-maint, labbott, madhu.chinakonda, mchehab, xgl-maint
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2017-08-27 07:13:33 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
Journalctl -b -1 of the event none

Description York Possemiers 2017-07-02 05:23:34 UTC
Created attachment 1293572 [details]
Journalctl -b -1 of the event

Description of problem:
With the released of the 4.11 series kernels into the Fedora updates system, I have no longer been able to remove the binding of my graphics card in order to have it participate in pci passthrough. The last working kernel for this was 4.10.17-200.fc25.x86_64.

This bug is highly consistent.

In order to do the unbinding, I use the following script, executed via SSH

#!/bin/bash
echo `date "+%H:%M:%S.%N"`"Leaving graphical mode"
time systemctl isolate multi-user.target
sleep 5
echo `date "+%H:%M:%S.%N"`"removing graphics binding"
time echo "0000:01:00.0" > /sys/bus/pci/drivers/amdgpu/unbind
sleep 5
echo `date "+%H:%M:%S.%N"`"Removing amdgpu module"
time modprobe -r amdgpu
sleep 5
echo `date "+%H:%M:%S.%N"`"Done"

Here is the latest attempt I've had using the 4.11.7-200.fc25.x86_64 kernel:

14:39:52.408096500Leaving graphical mode

real	0m0.253s
user	0m0.005s
sys	0m0.008s
14:39:57.666116884removing graphics binding
Segmentation fault

Remote interactions with the computer are still possible, however rebinding doesn't seem to work, and shutdown/reboots do not complete. The timestamps of the above and the attached output of journalctl do match.

Comment 1 York Possemiers 2017-07-28 07:50:59 UTC
Bug persists in 4.11.12

Comment 2 York Possemiers 2017-08-27 07:13:33 UTC
Bug seems absent from kernel-4.12.8-200.fc25.x86_64