Re: [Qemu-devel] [PATCH v5 0/7] Deterministic replay extensions

["Pavel Dovgalyuk" <dovgaluk@ispras.ru>]

> From: igor.rubinov@gmail.com [mailto:igor.rubinov@gmail.com] On Behalf Of Igor R
> > This set of patches is related to the reverse execution and deterministic
> > replay of qemu execution. It includes recording and replaying of serial devices
> > and block devices operations.
> 
> Thanks for the great feature.
> 
> A couple of questions:
> * Is it possible to enable/disable recording when QEMU is already
> running? I.e. is it ok to call replay_configure() or replay_enable()
> from the monitor? Would it be possible to replay such a log?

Our internal code base does not include this feature.
But we plan to implement this after submitting all already existing patches.

> * Currently, when running QEMU with the above drive configuration
> (even w/o icount), savevm command doesn't seem to work correctly, and
> subsequent loavm crashes QEMU. Is it just a bug or this feature
> doesn't support snapshots?

Supporting snapshots includes much more work in addition to correct interaction
with replay. Therefore this feature is not supported yet, but we will add it later.

> * The replaying is much slower than the original scenario. Is it
> affected by icount shift? What shift values would allow "real time"
> speed?

icount shows how many guest instructions correspond to one guest clock tick.
"Real speed" may differ from machine to machine.
However, replaying speed was not the subject of our investigations yet.
Recording speed is more important, because it may affect the behavior of the system.

Pavel Dovgalyuk