[Qemu-devel] QEMU News

[Qemu-devel] QEMU News

[Fabrice Bellard]

Hi,

The last commits add the following:

- New feature: the complete state of the virtual machine can be saved on 
disk (and hopefully restored !) with the commands 'savevm' and 'loadvm'. 
Not all the drivers implement the feature, but it already works enough 
to be usable. The saved virtual machine state does not include a 
description of its configuration, so when restoring the virtual machine, 
you must ensure that you have the same virtual configuration.

- First debug commands in the monitor: 'stop' (stop the VM), 'cont' 
(resume execution of the VM), 'gdbserver' (enable the integrated 
gdbserver stub at any point).

- Timer API for the device drivers (similar to the Linux kernel timers). 
It will enable the use of a virtual cycle counter to have a fully 
deterministic emulation (useful for debugging).

Fabrice.

Re: [Qemu-devel] QEMU News

[Renzo Davoli]

[-- Attachment #1: Type: text/plain, Size: 619 bytes --]

On Wed, Mar 31, 2004 at 09:14:54PM +0200, Fabrice Bellard wrote:
> - Timer API for the device drivers (similar to the Linux kernel timers). 
> It will enable the use of a virtual cycle counter to have a fully 
> deterministic emulation (useful for debugging).

There is a bug. On a P4 I have got a SIGFPE.
The fix is here attached. 
The asm call inside cpu_get_real_ticks (file vl.c) must be defined
volatile.
It seems that otherwise somewhere (gcc?) a code optimization instead
of evaluating twice the same instruction takes the former output value,
thus the count of tick per seconds results to be zero.

ciao
	renzo

[-- Attachment #2: vl.c.patch --]
[-- Type: text/plain, Size: 266 bytes --]

--- vl.c.orig	2004-04-01 10:44:57.000000000 +0200
+++ vl.c	2004-04-01 10:38:25.000000000 +0200
@@ -341,7 +341,7 @@
 int64_t cpu_get_real_ticks(void)
 {
     int64_t val;
-    asm("rdtsc" : "=A" (val));
+    asm volatile ("rdtsc" : "=A" (val));
     return val;
 }
 

Re: [Qemu-devel] QEMU News

[Jean-Michel POURE]

Le jeudi 1 Avril 2004 10:55, Renzo Davoli a écrit :
> There is a bug. On a P4 I have got a SIGFPE.
> The fix is here attached.

I can confirm the same problem which was fixed by the patch.
Cheers, Jean-Michel