VirtualBox

Ticket #5788 (closed defect: fixed)

Opened 4 years ago

Last modified 4 years ago

vboxvideo_drv.so did not work with kernel 2.6.32-trunk-686 in Debian Sidip

Reported by: sharuzzaman Owned by:
Priority: major Component: other
Version: VirtualBox 3.1.2 Keywords: vboxvideo
Cc: Guest type: Linux
Host type: Windows

Description

I'm running Debian Sid guest in VirtualBox 3.1.2 on Windows Vista host. I upgraded the kernel to latest package provided by Debian, 2.6.32-trunk-686. This causes X to abort and cannot start.

vboxvideo is working if using previous kernel 2.6.31-1-686.

The Xorg log file is attached.

Attachments

Xorg.0.log Download (22.1 KB) - added by sharuzzaman 4 years ago.
Xorg log file with error information for vboxvideo
virtualbox_maps.zip Download (16.1 KB) - added by sharuzzaman 4 years ago.
maps file for X server

Change History

Changed 4 years ago by sharuzzaman

Xorg log file with error information for vboxvideo

comment:1 Changed 4 years ago by sharuzzaman

Its Debian Sid, not Sidip. Some typo there.

comment:2 Changed 4 years ago by michael

Note to myself: the crash address is immediately after the call to VBoxMapVidMem() made from VBoxScreenInit(). (Really immediately after - the crash address, 0xc067 into the module, is the address of the assembly instruction straight after the assembly "call").

Another note, that I don't get this locally (Mac host, Debian Testing guest upgraded to sid immediately after the netinst install by changing the repository lines in sources.lst to "sid", commenting out the security updates lines and updating. Followed by apt-get install sudo, dkms, gnome, linux-image-2.6-686, rebooting, installing the Additions, rebooting again).

comment:3 Changed 4 years ago by michael

Sigh. Of course it is immediately after the assembly call, it is the return address on the stack. The crash is one level further down. The following address is about five megs above the module load address, so definitely not in the modules text segment as I assume it ought to be.

comment:4 Changed 4 years ago by michael

It would be very nice if you were somehow able to get /proc/<pid>/maps for the X server process just before the crash.

comment:5 Changed 4 years ago by sharuzzaman

Michael,

My Debian SID guest is booting up into KDM, then I will put my username and password, and upon pressing enter, X will crash.

I created a script that will run every 2 second since the guest boot up. The file attached is the /proc/<pid>/maps for X every 2 second from boot up (after rc.local executed) and leading to the crash.

Xorg.0.log for this session is also attached.

Changed 4 years ago by sharuzzaman

maps file for X server

comment:6 Changed 4 years ago by michael

Thanks for the map files (and for having the presence of mind to attach a new, matching X.Org logfile!) Unfortunately that still leaves me rather mystified. I can now see that the faulting instruction is the first of the virtual _ _kernel_vsyscall function in the VDSO (a virtual shared library in kernel space that is mapped into each user space process). This does not show on the stack as the segfault occurs before the function can set up the frame pointer. I will have to look at the code of vboxvideo to see where and why it is calling _ _kernel_vsyscall directly. And more to the point, I have no idea why the code there should segfault (is it missing?). This may need to wait for someone to get back from their Christmas break who knows more about this.

comment:7 Changed 4 years ago by sharuzzaman

Hi,

This issue has been resolved. The issue reported is not happening anymore. I believe there are issue in Xorg server in Debian, and have been fixed by the new release xorg-server (2:1.7.4-1)

See the changelog at  http://packages.debian.org/changelogs/pool/main/x/xorg-server/current/changelog especially for the part:

[ Julien Cristau ]

  • Don't call config_init() until after InitInput() has initialized the event queue, so that devices don't get enabled too early (closes: #564256, #564324).

Thank you for Michael assistance and kudos for a great product.

comment:8 Changed 4 years ago by frank

  • Status changed from new to closed
  • Resolution set to fixed

Thanks for the feedback!

Note: See TracTickets for help on using tickets.

www.oracle.com
ContactPrivacy policyTerms of Use