vmware 3 build 1445 trouble with linux 2.4.17

Knut J Bjuland

unread,

Dec 31, 2001, 6:03:07 PM12/31/01

to

Are there somebody with success with vmware with linux 2.4.17. There
appear to be some trouble with apm since I have acpi enable in my
kernel. I use Redhat 7.2 with linux-2.4.17 with XFS.

Petr Vandrovec

unread,

Jan 2, 2002, 1:16:27 PM1/2/02

to

I have no problem with 2.4.17 and 2.4.18-pre1, of course both from Marcello,
and not from RedHat, and both without XFS.
Petr

P.S.: With build 1455...

Knut J Bjuland

unread,

Jan 4, 2002, 5:05:36 AM1/4/02

to

Petr Vandrovec wrote:

I forgot to write that my kernek is patched with preemptible kernel patch.
This may cause vmware to break.

Petr Vandrovec

unread,

Jan 4, 2002, 3:43:21 PM1/4/02

to

Replace may with will, or do 'swapoff -a' before running VMware.
Petr

Brian C. Huffman

unread,

Jan 8, 2002, 7:18:55 PM1/8/02

to

I also have the preemtible kernel patches (and XFS) and am having
problems with VMWare 3.0. I didn't see this until 2.4.17, though. What
broke? And what do the preemtible patches do that break things?

Thanks,
Brian

Petr Vandrovec

unread,

Jan 9, 2002, 12:01:33 PM1/9/02

to

"Brian C. Huffman" wrote:
>
> I also have the preemtible kernel patches (and XFS) and am having
> problems with VMWare 3.0. I didn't see this until 2.4.17, though. What
> broke? And what do the preemtible patches do that break things?

It can deadlock when accessing userspace, one thread acquires vmmon lock
and big kernel lock, and then it accesses userspace. Then code is
removed from CPU, as preemptiple copy_{to,from}_user can be blocking
with preemptible patches even if userspace is mlocked() into memory.
Another thread acquires big kernel lock, and when it tries to get
vmmon lock - oops - it finds that vmmon lock is already active. But as
it cannot reschedule at that point due to owning other locks, first
thread cannot continue, and it dies.

If you understand kernel programming, you can fix vmmon's common/vmx86
and linux/driver.c so that it calls HostIF_Copy*User without locks
only. As you must make sure that your pointers are still valid after
you'll release and reacquire big kernel lock and/or vmmon lock, it
is not trivial change.
Petr

Brian C. Huffman

unread,

Jan 10, 2002, 10:13:20 AM1/10/02

to

Petr,

I have recompiled my kernel w/o the preempt patches and am still having
the same problem with VMWare 3.0. :-( I even installed the
VMwareWorkstationKernelModules-3.0.0-1491.i386.rpm (although I have no
idea what was fixed in this). I have RedHat 7.2 w/ the 2.4.17 kernel
(using XFS), I can consistently reproduce this problem and it. Here's
the logfile output:

Jan 10 09:30:44: VMX|F(138) line=1408 0x8:0x800630c4 fault=13
Jan 10 09:30:44: VMX|CPL0 stack shutdown
Jan 10 09:30:44: VMX|VCPU: eflags=0x82 eip=0x8042c15b
Jan 10 09:30:44: VMX|VCPU: EAX=0xfd45d500 EBX=0x0 ECX=0x0 EDX=0x10002
Jan 10 09:30:44: VMX|VCPU: EDI=0xfd45d53d ESI=0xfd45d591 EBP=0xfd45d15d
ESP=0xfd45cde9
Jan 10 09:30:44: VMX|VCPU: ES=0x23 CS=0x8 SS=0x10 DS=0x23 FS=0x30 GS=0x0
Jan 10 09:30:44: VMX|VCPU: code=4 stack=4 task=4
Jan 10 09:30:44: VMX|VCPU: cr0=0x80010031 cr3=0x3176000 cr4=0x6d1
Jan 10 09:30:44: VMX|Dumping Ring Buffer -- not used
Jan 10 09:30:44: VMX|1473 Rec Stats: eip=0x80063b1e esp=0xfd45f3a1
count=1 maxCount=4 pfLastIndex=0 curIndex=0
Jan 10 09:30:44: VMX|PanicMsg: 'panic.cpl0'
Jan 10 09:30:44: VMX|Coredump with build $Name: build-1455 $
Jan 10 09:30:44: VMX|Writing monitor corefile
"/root/vmware/win2000Pro/vmware-core"
Jan 10 09:30:44: VMX|CoreDump: dumping core with superuser privileges
Jan 10 09:30:44: VMX|Setting vaddr to 0x0
Jan 10 09:30:44: VMX|Msg_Post Error: panic.cpl0
*** Virtual Machine kernel stack fault (hardware reset) ***
[...]

Any ideas?

Thanks,
Brian

Petr Vandrovec

unread,

Jan 11, 2002, 2:07:37 PM1/11/02

to

"Brian C. Huffman" wrote:
> I have recompiled my kernel w/o the preempt patches and am still having
> the same problem with VMWare 3.0. :-( I even installed the
> VMwareWorkstationKernelModules-3.0.0-1491.i386.rpm (although I have no
> idea what was fixed in this). I have RedHat 7.2 w/ the 2.4.17 kernel
> (using XFS), I can consistently reproduce this problem and it. Here's
> the logfile output:
>
> Jan 10 09:30:44: VMX|F(138) line=1408 0x8:0x800630c4 fault=13
> Jan 10 09:30:44: VMX|CPL0 stack shutdown
> Jan 10 09:30:44: VMX|VCPU: eflags=0x82 eip=0x8042c15b
> Jan 10 09:30:44: VMX|VCPU: EAX=0xfd45d500 EBX=0x0 ECX=0x0 EDX=0x10002
> Jan 10 09:30:44: VMX|VCPU: EDI=0xfd45d53d ESI=0xfd45d591 EBP=0xfd45d15d
> ESP=0xfd45cde9
> Jan 10 09:30:44: VMX|VCPU: ES=0x23 CS=0x8 SS=0x10 DS=0x23 FS=0x30 GS=0x0
> Jan 10 09:30:44: VMX|VCPU: code=4 stack=4 task=4
> Jan 10 09:30:44: VMX|VCPU: cr0=0x80010031 cr3=0x3176000 cr4=0x6d1

It looks to me that your guest commited suicide at address 0x8:0x800630c4,
while being in kernel mode with interrupts disabled.
Petr