[Qemu-devel] sparc32 reset's regression

[Qemu-devel] sparc32 reset's regression

[Artyom Tarasenko]

reset on SS-10/SS-20 machines hangs on memory detection if
qemu-system-sparc is launched with OBP (not OpenBios).

git bisect shows that the regression comes with the patch
"Don't set IRQs on device reset and loadvm/savevm" on 2009-06-17 19:20:01.

Artyom

Re: [Qemu-devel] sparc32 reset's regression

[Blue Swirl]

[-- Attachment #1: Type: text/plain, Size: 410 bytes --]

On Wed, Aug 5, 2009 at 11:44 AM, Artyom
Tarasenko<atar4qemu@googlemail.com> wrote:
> reset on SS-10/SS-20 machines hangs on memory detection if
> qemu-system-sparc is launched with OBP (not OpenBios).
>
> git bisect shows that the regression comes with the patch
> "Don't set IRQs on device reset and loadvm/savevm" on 2009-06-17 19:20:01.

The commit looks fine to me. Would this patch happen to fix the bug?

[-- Attachment #2: 0001-Fix-reset.patch --]
[-- Type: application/mbox, Size: 749 bytes --]

Re: [Qemu-devel] sparc32 reset's regression

[Artyom Tarasenko]

> The commit looks fine to me. Would this patch happen to fix the bug?

Yes, this patch fixes the bug, thanks.

It can be though that the bug is not directly related to the initial commit.
There seems to be a race condition. Sometimes the bug didn't appear
even without the fix if I put a lot of DPRINTFs in irq handling in
sun4m.c and cpu-exec.c and turn the logging on. Unfortunately it's not
always reproducible, so I can't tell yet which printf hides the
problem.