Host crash when 3D Acceleration enabled: nouveau E [DRM] GPU lockup - switching to software fbcon

[Philip Seeger]

When logging into a KDE/Plasma desktop session in a Fedora vm, the host crashed (or rather X11). This happens after upgrading this vm to Fedora 22. The vm worked perfectly fine (using KDE) with Fedora 21.

After logging in, the login manager disappeared (guest) and a few seconds later, the whole screen turned black (host), the last few boot messages visible at the top of the screen. Followed by:

nouveau E [DRM] GPU lockup - switching to software fbcon
*** Error in '/usr/bin/X': malloc(): memory corruption: 0x... ***

dmesg:

[1266287.821949] nouveau E[  PGRAPH][0000:03:00.0] DATA_ERROR [INVALID_BITFIELD] ch 6 [0x003f9de000 VBoxCrWinCmd[5365]] subc 0 class 0x9297 mthd 0x0810 data 0x20010e0c
[1266287.821976] nouveau E[  PGRAPH][0000:03:00.0] DATA_ERROR [INVALID_BITFIELD] ch 6 [0x003f9de000 VBoxCrWinCmd[5365]] subc 0 class 0x9297 mthd 0x0818 data 0x80000573
[1266287.821999] nouveau E[  PGRAPH][0000:03:00.0] DATA_ERROR [INVALID_VALUE] ch 6 [0x003f9de000 VBoxCrWinCmd[5365]] subc 0 class 0x9297 mthd 0x081c data 0x20010e0a
[1266287.836233] VBoxCrWinCmd[5365]: segfault at 8 ip 00007f56101e067e sp 00007f5610672be0 error 4 in libdrm_nouveau.so.2.0.0[7f56101dd000+6000]
[1266287.836259] nouveau E[   PFIFO][0000:03:00.0] PBDMA0: (unknown bits 0x00040000)
[1266287.836265] nouveau E[   PFIFO][0000:03:00.0] PBDMA0: ch 6 [VBoxCrWinCmd[5365]] subc 0 mthd 0x0000 data 0x00000000
[1266287.836278] nouveau E[   PFIFO][0000:03:00.0] PBDMA0: (unknown bits 0x00040000)
[1266287.836280] nouveau E[   PFIFO][0000:03:00.0] PBDMA0: ch 6 [VBoxCrWinCmd[5365]] subc 0 mthd 0x0000 data 0x00000000
[1266287.836294] nouveau E[   PFIFO][0000:03:00.0] PBDMA0: (unknown bits 0x00040000)
[1266287.836297] nouveau E[   PFIFO][0000:03:00.0] PBDMA0: ch 6 [VBoxCrWinCmd[5365]] subc 0 mthd 0x0000 data 0x00000000
[1266303.224025] nouveau E[VBoxCrWinCmd[5365]] failed to idle channel 0xcccc0000 [VBoxCrWinCmd[5365]]
[1266318.224019] nouveau E[VBoxCrWinCmd[5365]] failed to idle channel 0xcccc0000 [VBoxCrWinCmd[5365]]
[1266348.040022] nouveau E[Xorg[2198]] failed to idle channel 0xcccc0000 [Xorg[2198]]
[1266363.040023] nouveau E[Xorg[2198]] failed to idle channel 0xcccc0000 [Xorg[2198]]
[1266363.040124] nouveau E[   PFIFO][0000:03:00.0] read fault at 0x000001b000 [PAGE_NOT_PRESENT] from PFIFO/BAR_READ on channel 0x003fc1e000 [unknown]
[1266364.868056] traps: VirtualBox[9472] general protection ip:7f711bb2fb2d sp:7ffc0e85d020 error:0 in libc-2.19.so[7f711baf8000+19f000]
[1266368.699587] traps: VirtualBox[14740] general protection ip:7fc31e192b2d sp:7ffca651c460 error:0 in libc-2.19.so[7fc31e15b000+19f000]
[1266368.821819] traps: SHCLIP[9412] general protection ip:7fbe37d06b2d sp:7fbdd8685aa0 error:0 in libc-2.19.so[7fbe37ccf000+19f000]
[1266380.248008] nouveau E[VBoxCrWinCmd[9523]] failed to idle channel 0xcccc0000 [VBoxCrWinCmd[9523]]
[1266395.248016] nouveau E[VBoxCrWinCmd[9523]] failed to idle channel 0xcccc0000 [VBoxCrWinCmd[9523]]
[1266395.529060] nouveau E[     DRM] GPU lockup - switching to software fbcon
[1266398.440393] TimerLR[877]: segfault at 41 ip 00007faca1de01fa sp 00007fac991227d8 error 4 in VBoxXPCOMIPCC.so[7faca1dcb000+1e000]
[1266398.573339] device eth0 left promiscuous mode
[1266398.593524] vboxnetflt: 352611 out of 69075285 packets were not sent (directed to host)
[1266404.729826] device vboxnet0 left promiscuous mode
[1266404.742346] vboxnetflt: 809 out of 11882 packets were not sent (directed to host)
[1266624.044009] nouveau E[Xorg[26442]] failed to idle channel 0xcccc0002 [Xorg[26442]]
[1266639.044006] nouveau E[Xorg[26442]] failed to idle channel 0xcccc0002 [Xorg[26442]]
[1266654.044007] nouveau E[Xorg[26442]] failed to idle channel 0xcccc0001 [Xorg[26442]]
[1266669.044008] nouveau E[Xorg[26442]] failed to idle channel 0xcccc0001 [Xorg[26442]]
[1266684.044007] nouveau E[Xorg[26442]] failed to idle channel 0xcccc0000 [Xorg[26442]]
[1266699.044009] nouveau E[Xorg[26442]] failed to idle channel 0xcccc0000 [Xorg[26442]]

After resetting the host and fixing some issues caused by the crash, I started this vm again (but nothing else). This time, only this vm process crashed, not affecting other processes on the host:

[ 1566.647048] ShCrOpenGL[6711]: segfault at 8 ip 00007efcfcce267e sp 00007efcfe974960 error 4 in libdrm_nouveau.so.2.0.0[7efcfccdf000+6000]

This vm had 3D acceleration enabled. After disabling 3D acceleration, the vm will load KDE without crashing.

Host: Debian 8

[Philip Seeger]

VM log