* [Qemu-devel][BUG][PATCH] Fix crash in kvm.c
@ 2008-12-05 22:15 Stefan Weil
2008-12-05 22:37 ` Anthony Liguori
0 siblings, 1 reply; 4+ messages in thread
From: Stefan Weil @ 2008-12-05 22:15 UTC (permalink / raw)
To: QEMU Developers
[-- Attachment #1: Type: text/plain, Size: 485 bytes --]
I got a crash (array access out of bounds results in access fault)
with the current Qemu trunk when kvm is enabled:
qemu -fda fd.img -cdrom cdrom.img -hda hda.img -hdb raw.img -m 256 -boot
c -enable-kvm
Host is Debian x86_64, the crash occurs before any code is emulated.
With the patch, the emulation (Win 98) boots, but has problems with the
display of
icons and the mouse cursor. Qemu displays lots of
"BUG: kvm_physical_sync_dirty_bitmap: invalid parameters" messages.
Stefan
[-- Attachment #2: kvm.patch --]
[-- Type: text/x-diff, Size: 1125 bytes --]
Fix crash with kvm enabled.
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Index: target-i386/kvm.c
===================================================================
--- target-i386/kvm.c (Revision 5889)
+++ target-i386/kvm.c (Arbeitskopie)
@@ -12,6 +12,7 @@
*
*/
+#include <assert.h>
#include <sys/types.h>
#include <sys/ioctl.h>
#include <sys/mman.h>
@@ -39,7 +40,8 @@
struct kvm_cpuid cpuid;
struct kvm_cpuid_entry entries[100];
} __attribute__((packed)) cpuid_data;
- int limit, i, cpuid_i;
+ int limit, cpuid_i;
+ unsigned i;
uint32_t eax, ebx, ecx, edx;
cpuid_i = 0;
@@ -49,6 +51,7 @@
for (i = 0; i <= limit; i++) {
struct kvm_cpuid_entry *c = &cpuid_data.entries[cpuid_i++];
+ assert(cpuid_i < 100);
cpu_x86_cpuid(env, i, &eax, &ebx, &ecx, &edx);
c->function = i;
@@ -63,6 +66,7 @@
for (i = 0x80000000; i <= limit; i++) {
struct kvm_cpuid_entry *c = &cpuid_data.entries[cpuid_i++];
+ assert(cpuid_i < 100);
cpu_x86_cpuid(env, i, &eax, &ebx, &ecx, &edx);
c->function = i;
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Qemu-devel][BUG][PATCH] Fix crash in kvm.c
2008-12-05 22:15 [Qemu-devel][BUG][PATCH] Fix crash in kvm.c Stefan Weil
@ 2008-12-05 22:37 ` Anthony Liguori
[not found] ` <4939B25F.8030203@mail.berlios.de>
0 siblings, 1 reply; 4+ messages in thread
From: Anthony Liguori @ 2008-12-05 22:37 UTC (permalink / raw)
To: qemu-devel, Glauber de Oliveira Costa, Stefan Weil
Stefan Weil wrote:
> I got a crash (array access out of bounds results in access fault)
> with the current Qemu trunk when kvm is enabled:
>
> qemu -fda fd.img -cdrom cdrom.img -hda hda.img -hdb raw.img -m 256 -boot
> c -enable-kvm
>
> Host is Debian x86_64, the crash occurs before any code is emulated.
>
Is the patch incomplete, perhaps? It seems to just add asserts which
shouldn't fix anything.
I don't think any leaf should return a max leaf greater than 100
elements so I'd be pretty surprised to see this happen. I'd really like
to see the back trace to see which leaf is the problematic one and what
the greatest leaf being reported is.
> With the patch, the emulation (Win 98) boots, but has problems with the
> display of
> icons and the mouse cursor. Qemu displays lots of
> "BUG: kvm_physical_sync_dirty_bitmap: invalid parameters" messages.
>
It looks like Avi and Glauber found some bugs in this code. There are
patches for kvm-userspace right now that need porting to QEMU. I'll
take a look at that this weekend unless Glauber was already planning too.
Regards,
Anthony Liguori
> Stefan
>
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Qemu-devel][BUG][PATCH] Fix crash in kvm.c
[not found] ` <4939B2F3.5060307@codemonkey.ws>
@ 2008-12-06 10:00 ` Stefan Weil
2008-12-11 21:03 ` Anthony Liguori
0 siblings, 1 reply; 4+ messages in thread
From: Stefan Weil @ 2008-12-06 10:00 UTC (permalink / raw)
To: Anthony Liguori, QEMU Developers
[-- Attachment #1: Type: text/plain, Size: 951 bytes --]
Anthony Liguori schrieb:
> Stefan Weil wrote:
>> Anthony Liguori schrieb:
>>
>>> Stefan Weil wrote:
>>>
>>>> I got a crash (array access out of bounds results in access fault)
>>>> with the current Qemu trunk when kvm is enabled:
>>>>
>>>> qemu -fda fd.img -cdrom cdrom.img -hda hda.img -hdb raw.img -m 256
>>>> -boot
>>>> c -enable-kvm
>>>>
>>>> Host is Debian x86_64, the crash occurs before any code is emulated.
>>>>
>>> Is the patch incomplete, perhaps? It seems to just add asserts which
>>> shouldn't fix anything.
>>>
>>
>> The essential change was to replace "int i" by "unsigned i".
>> i = 0x80000000 is negative, so the for loop went until there was an
>> access fault.
>>
>
> We should change limit too, right?
>
> Regards,
>
> Anthony Liguori
>
Right, and cpuid_i, too. "uint32_t" is even better than "unsigned".
Here is a new patch, now without assertions, so it can be applied to
Qemu trunk.
Regards
Stefan
[-- Attachment #2: kvm.patch --]
[-- Type: text/x-diff, Size: 504 bytes --]
Fix crash with kvm enabled.
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Index: target-i386/kvm.c
===================================================================
--- target-i386/kvm.c (revision 5889)
+++ target-i386/kvm.c (working copy)
@@ -39,7 +39,7 @@
struct kvm_cpuid cpuid;
struct kvm_cpuid_entry entries[100];
} __attribute__((packed)) cpuid_data;
- int limit, i, cpuid_i;
+ uint32_t limit, i, cpuid_i;
uint32_t eax, ebx, ecx, edx;
cpuid_i = 0;
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [Qemu-devel][BUG][PATCH] Fix crash in kvm.c
2008-12-06 10:00 ` Stefan Weil
@ 2008-12-11 21:03 ` Anthony Liguori
0 siblings, 0 replies; 4+ messages in thread
From: Anthony Liguori @ 2008-12-11 21:03 UTC (permalink / raw)
To: Stefan Weil; +Cc: QEMU Developers
Stefan Weil wrote:
> Right, and cpuid_i, too. "uint32_t" is even better than "unsigned".
> Here is a new patch, now without assertions, so it can be applied to
> Qemu trunk.
>
Applied. Thanks.
Regards,
Anthony Liguori
> Regards
> Stefan
>
>
>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2008-12-11 21:03 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-12-05 22:15 [Qemu-devel][BUG][PATCH] Fix crash in kvm.c Stefan Weil
2008-12-05 22:37 ` Anthony Liguori
[not found] ` <4939B25F.8030203@mail.berlios.de>
[not found] ` <4939B2F3.5060307@codemonkey.ws>
2008-12-06 10:00 ` Stefan Weil
2008-12-11 21:03 ` Anthony Liguori
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).