Segfault with custom vnc client

[Uli Stärk]

Hey,

I'm using Citrix XenServer 6.5. I worte a script that uses noVNC to connect to the rfb console via xapi. When I use GRML and try to boot it, the QEMU process segfaults and kills my VM. This happens when the screen resizes and the kernel is loading:

recvfrom(3, "\3\1\0\0\0\0\2\200\1\220\3\0\2\200\0\0\0P\1\220", 4096, 0, NULL, NULL) = 20
--- SIGSEGV {si_signo=SIGSEGV, si_code=SEGV_MAPERR, si_addr=0xb28000} ---

I can see in the child process the following message, right before the parent Segfaults:
read(4, "cirrus: blanking the screen line_offset=0 height=480\n", 53) = 53

This issue only happens, when I have my custom php/novnc-client connected. I also tried the nodejs/novnc package from xen-orchestra - same result. Using the stock client from Citrix XenCenter it works just fine. So I think it is related to noVNC. I hope this is just a bug and not exploitable to force a VM to crash or execute code.

XenServer launches the qemu with the following command line:

qemu-dm-25 --syslog -d 25 -m 2048 -boot dc -serial pty -vcpus 1 -videoram 4 -vncunused -k en-us -vnc 127.0.0.1:1 -usb -usbdevice tablet -net nic,vlan=0,macaddr=8a:43:e2:b1:57:df,model=rtl8139 -net tap,vlan=0,bridge=xenbr0,ifname=tap25.0 -acpi -monitor pty

XenServer 6.5 is using the following version:
# /usr/lib64/xen/bin/qemu-dm -help
QEMU PC emulator version 0.10.2, Copyright (c) 2003-2008 Fabrice Bellard

Greetings
Uli Stärk

[Daniel Berrange (berrange)]

Can you attach GDB to your qemu-dm process and attempt to capture a full stack trace when it crashes (ie thread apply all backtrace)

[Dubravko (dsever)]

Hi,

Did you resolve your problem? Because I have the same issus..

Dubravko

[Uli Stärk (ulistaerk)]

No, sorry. I've stopped researching this issue.

[Dr. David Alan Gilbert (dgilbert-h)]

Dubravko: can you get a good backtrace from your crash?

[Plam (lambert-olivier)]

Hi,

Xen Orchestra project leader here. Exactly the same problem with noVNC + qemu-dm.

I assume there is maybe an invalid param from noVNC somewhere, but crashing the process is not really expected.

The bug is also reported at Citrix (see https://bugs.xenserver.org/browse/XSO-381)

I don't know how to give you more info, if you have some commands to use to help you, tell me :)

[Thomas Huth (th-huth)]

QEMU 0.10 is pretty much outdated nowadays... can you reproduce this issue with the latest version of QEMU, and if so, provide a backtrace of the crash?

[Launchpad Janitor (janitor)]

[Expired for QEMU because there has been no activity for 60 days.]