Bug 27526 – Vmware workstation 4.0.1 dies after virtual disk (SCSI:0:0) times out unexpectadly

• Product: Gentoo Linux
• Status: RESOLVED FIXED

[Description Steve 2003-08-29 05:37:51 UTC]

Vmware will die whenever a write of data is attempted to the scsi hardisk within
vmware (the virutal disk). The same hardware on a red-hat installation yeilds
correct writes to the virtual disk and normal functioning of vmware. This
behavior was discovered during a windows 2000 install (as the guest-os) and the
error was eventually localised when the guest-os was the gentoo-liveCD

Reproducible: Always
Steps to Reproduce:
1.Install vmware
2.Run vmware, all other steps from here are refering to actions taken within the
guest os
3. attempt an install of Windows 2000 pro, notice that VMware dies when the
virtual disk attempted to be formatted, this is a consistant error that can be
reproduced
4. Then take your live cd, and boot off of it (as the guest-os) notice even
though a partion for ntfs *should* have been written out, it was not.
5. Use fdisk /dev/sda to write out a ntfs partion. Notice that the partion table
can be written correctly. If you wish, you may reboot with the live cd and
notice the partion is still there - therefore we know the partion was written
out to the virtual disk in a persistant state.
6. attempt a dd if=/dev/sda of=/dev/sda notice that when data is moved from the
buffer to the virtual disk, a error is generated and the guest-os crashes
Actual Results:  
the following error is generated in <location of virtual disk files>/vmware.log,
the guest-os then proceedes to crash, vmware however remains running.


<backtrace after signal 7>

Aug 28 02:34:05: vmx| VMX IPC closed the connection with thread scsi0:1
(0x8224e8c)
Aug 28 02:34:05: vmx| AIO: Unexpected loss of IPC to channel scsi0:1 (thread
scsi0:1)
Aug 28 02:34:05: vmx| AIO: Unexpected loss of channel scsi0:1 (thread
 scsi0:1)

<more backtrace stuff from the stack>

Aug 28 02:34:05: vmx| Msg_Post: Error
Aug 28 02:34:05: vmx| [msg.log.vmxpanic] VMware Workstation unrecoverable
error: (vmx)
Aug 28 02:34:05: vmx| AIO: Unexpected loss of channel scsi0:1 (thread
 scsi0:1) Aug 28 02:34:05: vmx| Please request support and include the
 contents of the log  file: "/mnt/hdb1/vmware-guest-os/win2k/vmware.log".  We
 will respond on the basis of your support entitlement.
Aug 28 02:34:05: vmx| ----------------------------------------

This error log is generated when the live-cd was used to diagnose the problem.
The Windows 2000 install did not generate anything helpfull in the log file.

No errors were found in /var/log/everything during the time of running

Expected Results:  
dd should complete, Windows 2000 install should have completed, moreover, the
partion created by windows 2000 should have been written to disk, it was not, as
seen when a fdisk /dev/sda was run from the livecd after the windows 2000 setup
had crashed

root@StevesWorld steve # emerge info
Portage 2.0.48-r5 (default-x86-1.4, gcc-3.2.3, glibc-2.3.2-r1)
=================================================================
System uname: 2.4.20-gentoo-r6 i686 AMD Athlon(TM) XP 1700+
GENTOO_MIRRORS="http://www.fhh.opensource-mirror.de/gentoo.org/
rsync://rsync.fhh.opensource-mirror.de/gentoo"
CONFIG_PROTECT="/etc /var/qmail/control /usr/kde/2/share/config
/usr/kde/3/share/config /usr/X11R6/lib/X11/xkb /usr/kde/3.1/share/config
/usr/share/config"
CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d"
PORTDIR="/usr/portage"
DISTDIR="/usr/portage/distfiles"
PKGDIR="/usr/portage/packages"
PORTAGE_TMPDIR="/var/tmp"
PORTDIR_OVERLAY=""
USE="x86 oss avi crypt foomaticdb gif jpeg mad mikmod mpeg ncurses nls pdflib
png quicktime spell truetype xml2 xmms xv zlib gdbm readline arts svga tcltk
java gpm tcpd pam perl python imlib oggvorbis qt motif opengl cdr aalib dga dvd
esd maildir mbox md5sum xvid kde X acl ssl 3dnow mmx apm -cups -encode gtk
-gnome -libg++ -libwww -sdl -slang -berkdb"
COMPILER="gcc3"
CHOST="i686-pc-linux-gnu"
CFLAGS="-march=athlon-xp -O3 -pipe"
CXXFLAGS="-O2 -mcpu=i686 -pipe"
ACCEPT_KEYWORDS="x86"
MAKEOPTS="-j2"
AUTOCLEAN="yes"
SYNC="rsync://rsync.de.gentoo.org/gentoo-portage"
FEATURES="sandbox ccache distcc usersandbox regenworld"

steve@StevesWorld steve $ uname -a
Linux StevesWorld 2.4.20-gentoo-r6 #4 Thu Aug 28 01:20:26 ADT 2003 i686 AMD
Athlon(TM) XP 1700+ AuthenticAMD GNU/Linux

Kerenel configuration is attached

[Comment 1 Steve 2003-08-29 05:41:31 UTC]

Created attachment 16753 [details]
Current Kernel configuration used with VMware at the time of bug

[Comment 2 Steve 2003-08-29 05:42:08 UTC]

Created attachment 16754 [details]
Current Kernel configuration used with VMware at the time of bug

[Comment 3 Chris Gianelloni (RETIRED) 2003-08-29 11:07:03 UTC]

Sorry, I had told the user to assign this to games since I know we're doing app-emulation.

I'm going to grab this one myself.

[Comment 4 Chris Gianelloni (RETIRED) 2003-08-29 11:50:19 UTC]

First, I would suggest you turn off CONFIG_IDEDISK_MULTI_MODE.  Also, are you enabling low latency via the sysctl interface or are you leaving it turned off?

I am going to test a kernel using your exact .config and see what happens.  I WILL have to change one thing, the IDE controller, since I do not have a SIS controller.  Also, you should always test something you may think is kernel-related WITHOUT the gcc>31 optimizations, as they are known to break things.  Luckily for you, I also have a high-end AMD system, so I can even use your exact settings.

[Comment 5 Chris Gianelloni (RETIRED) 2003-08-29 11:58:07 UTC]

Also, GRSEC is known to break TONS of stuff, VMWare being a target.  I would suspect your use of grsecurity is causing the problem.

[Comment 6 Steve 2003-08-29 19:11:12 UTC]

The grsecurty options that are on should not affect vmware, however, I'm going to 
try compiling without grsecurity on, and without multimode on the ide controler, also 
I have left low latency schdualing off. 

[Comment 7 Chris Gianelloni (RETIRED) 2003-09-01 05:56:40 UTC]

I am going to close this bug, since we have a new version of VMWare in portage (which fixes a security bug).  If you still have a problem with the new version, then create a new bug and assign it to me.

The bug number for the new version is 27293.