[xorg-bugzilla-noise] [Bug 516] New: vmware hangs the X server *hard*

[bugzilla-daemon at pdx.freedesktop.org]

http://pdx.freedesktop.org/cgi-bin/bugzilla/show_bug.cgi?id=516

           Summary: vmware hangs the X server *hard*
           Product: xorg
           Version: 6.7.0
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: major
          Priority: P2
         Component: Server/general
        AssignedTo: xorg-bugzilla-noise at freedesktop.org
        ReportedBy: J.Breuer at gmx-topmail.de


Running vmware-workstation-4.5.1.7568 under X.org leads - after a few minutes -
to the X server hanging.

Mouse cursor is still movable, but no longer changes shape according to context;
clicks are ignored, also all keyboard input is ignored. (This happens the same
with the 'ungrabbed' X mouse cursor as well as with the 'grabbed' VM cursor,
depending on  whether vmware had grabbed the cursor when the problem started).

The key sequences Ctrl-Alt-Fn (to switch to a text console) and
Ctrl-Alt-Backspace (to kill the server) have no effect; the graphical display
with its last displayed content stays on screen. Also the '*-Lock' keys no
longer toggle their respective feedback LEDs.

It is not possible to log into the machine in that state using ssh, the TCP
connection is correctly established (!) but identity exchange never takes place
- apparently the ssh daemon is not being given any CPU cycles.

*Wild guess*: The X server (or whichever part is responsible for moving the
mouse cursor) is the only process still executing, no other processes "get on
the CPU".

In this state, the machine can usually be sort-of-cleanly brought down using
Alt-SysRq-[sub].

If the machine is left in the above states for a few minutes, the machine also
stops responding to ping requests or Magic SysRq, and needs to be power-cycled.

Due to the nature of the bug I find it difficult to collect useful information
on that matter; any tips for doing so are appreciated. Unfortunately, setting up
something like a serial console is a bit of a problem because of the physical
layout here.

The bug appears to hit earlier when the VM is configured to use "Acceleration"
(in its "Advanced Settings" screen); it seems to occur later when "Acceleration"
is disabled. Unfortunately, I cannot seem to find any useful documentation on
what vmware's "Acceleration" option really does.

Information about the machines which exhibit the problem:

#1:
- PIII 800
- GigaByte-Mainboard with i440BX
- nVidia NV28 (Ti 4800) AGP
- NetGear FA310TX (Lite-On/Tulip smurfalike)

#2:
- Celeron 900
- GigaByte-Mainboard with i815
- Matrox G450 16MB
- NetGear FA310TX (Lite-On/Tulip smurfalike)

Both running the same Gentoo 1.4.9 (stable as of something like 4/18), glibc
2.3.2, gcc 3.3.2, Kernel 2.6.4 with identical config (#2 has the bd-claim patch).

The problem does not occur when using Gentoo's xfree-4.3.0-r5 (only tested on #1
as it used to work there flawlessly before moving to X.org; after seeing the bug
on #1 I reproduced it on #2 to eliminate the nVidia drivers as a potential problem).

On request, I'll provide further information as far as I'm able to.



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.