Segmentation fault after creating a snapshot

[pabad]

I'm using Gentoo linux 2007.0 on an AMD64. I've tried this both with the open source and closed source versions I've created a virtual machine were I've successfully installed Windows XP. I've decided to create a snapshot, only to find that virtualbox will no longer run.

So far I've manually edited the virtual computer configuration file and the virtualbox.xml file to delete the snapshot and then virtualbox starts again, until I try again to create a snaphot.

I've also run VBoxSVC on the console and virtual box on another terminal under these circunstances, and found the following output: * innotek VirtualBox XPCOM Server Version 1.4.0 (C) 2004-2007 innotek GmbH All rights reserved.

Starting event loop.... [press Ctrl-C to quit] Informational: VirtualBox object created (rc=00000000). Segmentation fault

I'm attaching the computer definition file, and the virtualbox configuration file hoping it may be of any help. I'm keeping the files and configurations in case someone needs me to try anything or test a fix.

[pabad]

configuration file on home .VirtualBox folder

[pabad]

Virtual Computer definition

[pabad]

Virtual computer file AFTER creaating the attachment

[pabad]

Computer definition file AFTER creating the snapshot

[pabad]

configuration file on home .VirtualBox folder AFTER creating the snapshot

[pabad]

I've further noticed that if I remove the <DiffHardDisk> tag created for the disk differencing, gui is able to start up, though of course I can't access the virtual machine due to the missing drive file.

[pabad]

More information: I'm able to execute the virtual computer through the command line via the following command: VBoxManage startvm WinXP

and it recognizes the snapshot. I've also been able to create additional snapshots and rollback them through the command line without any problem.

So it seems to be related to the user interface loading the configuration file

[pabad]

I'm getting the problem on the latest snvsource as well. Running a debug build yields the following messages:

Type Manifest File: /root/.VirtualBox/xpti.dat
nsNativeComponentLoader: autoregistering begins.
nsNativeComponentLoader: autoregistering succeeded
nNCL: registering deferred (0)
###!!! ASSERTION: refcnt levels are still left: 'mRefCntLevels.GetSize() == 0', file /root/vbox/vbox/src/libs/xpcom18a4/ipc/ipcd/extensions/dconnect/src/ipcDConnectService.cpp, line 1627
Break: at file /root/vbox/vbox/src/libs/xpcom18a4/ipc/ipcd/extensions/dconnect/src/ipcDConnectService.cpp, line 1627
nsStringStats
 => mAllocCount:             51
 => mReallocCount:           15
 => mFreeCount:              42  --  LEAKED 9 !!!
 => mShareCount:             89
 => mAdoptCount:              9
 => mAdoptFreeCount:          9

The messages reported by VBoxSVC are:

Type Manifest File: /root/.VirtualBox/xpti.dat
###!!! ASSERTION: bad call: '!gMemory', file /root/vbox/vbox/src/libs/xpcom18a4/xpcom/glue/nsMemory.cpp, line 66
Break: at file /root/vbox/vbox/src/libs/xpcom18a4/xpcom/glue/nsMemory.cpp, line 66
nsNativeComponentLoader: autoregistering begins.
*** Registering VirtualBox_Client_Module components (all right -- a generic module!)
*** Registering VirtualBox_Server_Module components (all right -- a generic module!)
nsNativeComponentLoader: autoregistering succeeded
nNCL: registering deferred (0)
nsNativeComponentLoader: autoregistering begins.
nsNativeComponentLoader: autoregistering succeeded
nNCL: registering deferred (0)
*************************************************
innotek VirtualBox XPCOM Server Version 1.4.1_OSE
(C) 2004-2007 innotek GmbH
All rights reserved.
Debug version.

Starting event loop....
[press Ctrl-C to quit]
Informational: VirtualBox object created (rc=00000000).
WARNING: failed to send RELEASE event, file /root/vbox/vbox/src/libs/xpcom18a4/ipc/ipcd/extensions/dconnect/src/ipcDConnectService.cpp, line 1682
Informational: VirtualBox object deleted.

Also, the problem doesn't happen always, and in the debug build it happens less than on the release build. Can it be a buffer contention problem?

[pabad]

Trace obtained on a segmentation fault with a debug build

[Frank Mehnert]

Please try to generate a backtrace with gdb by starting a debug version of VBoxSVC and VirtualBox with gdb:

LD_LIBRARY_PATH=. gdb ./VBoxSVC
LD_LIBRARY_PATH=. gdb ./VirtualBox

(I assume that the current directory is the output/bin directory of the OSE version.

[Frank Mehnert]

Please run both commands in different windows and enter 'bt' as command for gdb after the segfault occurred.

[Frank Mehnert]

Some guy on #vbox told he has similar problems. His problem is related to #490: An endless loop until the stack overflows. We are about to fix this.

[dec0de]

Ok.. this will sound really weird, but I've got the same segment fault with the 1.4.0 running SuSE 10.2 and by accident I put the parameter -m to it in a terminal window and lo and behold it starts! I have no clue what 'VirtualBox -m' does, or if there are any parameters at all, but if not putting '-m' it will always segment fault, and if having it there it will start most of the times. I have tried with -h and -v for fun, but that wont work. It seems to like -m though.

Weird...

[Frank Mehnert]

We believe we fixed this problem in svn.

[pabad]

I can confirm that the bug is fixed in svn. (rev 3918)

Thank you!

[Frank Mehnert]

Please reopen if still happens with 1.5.0.