Re: [Qemu-discuss] Stopping and snapshoting at arbitrary point

2015-06-04 Thread Peter Maydell
On 4 June 2015 at 06:53, Igor R  wrote:
> Is it possible to stop QEMU at an arbitrary point and save a snapshot?
> I'm using PANDA framework built upon QEMU, and I'm attempting to stop
> and make a snapshot within a PANDA callback, as I described here:
> http://mailman.mit.edu/pipermail/panda-users/2015-April/000359.html
>
> I was told it wouldn't work due to QEMU limitations. So, my question
> is whether it's possible to force QEMU to update the registers and to
> be prepared for pausing and snapshoting at an arbitrary point?

The best you can do is to force the basic blocks to all be
one instruction long, with the -singlestep option. This will
obviously have a detrimental effect on performance.

-- PMM



Re: [Qemu-discuss] Stopping and snapshoting at arbitrary point

2015-06-04 Thread Igor R
Thanks for your response.
Maybe it's possible to switch singlestep mode on and off at runtime? My
PANDA plugin runs within QEMU context, so if it's possible to change bb
length programmatically, it would solve the problem.
4 Июн 2015 г. 10:51 пользователь "Peter Maydell" 
написал:

> On 4 June 2015 at 06:53, Igor R  wrote:
> > Is it possible to stop QEMU at an arbitrary point and save a snapshot?
> > I'm using PANDA framework built upon QEMU, and I'm attempting to stop
> > and make a snapshot within a PANDA callback, as I described here:
> > http://mailman.mit.edu/pipermail/panda-users/2015-April/000359.html
> >
> > I was told it wouldn't work due to QEMU limitations. So, my question
> > is whether it's possible to force QEMU to update the registers and to
> > be prepared for pausing and snapshoting at an arbitrary point?
>
> The best you can do is to force the basic blocks to all be
> one instruction long, with the -singlestep option. This will
> obviously have a detrimental effect on performance.
>
> -- PMM
>