Re: [Qemu-devel] [RFC PATCH v10 18/24] replay: replay aio requests

[Pavel Dovgaluk]

> From: Paolo Bonzini [mailto:address@hidden
> On 27/02/2015 14:11, Pavel Dovgalyuk wrote:
> > This patch adds identifier to aio requests. ID is used for creating bottom
> > halves and identifying them while replaying.
> > The patch also introduces several functions that make possible replaying
> > of the aio requests.
> 
> Out of curiosity, why did you use this approach instead of using a
> RR-specific block device backend (as you did for network, I think)?  The
> backend could just store the data that is read in the RR file (or in a
> separate file that can be easily mmap-ed), and use a timer to trigger it
> at the right QEMU_CLOCK_VIRTUAL tick.
> 
> I'm sure you considered something like this.  Did you still get
> non-determinism?

We considered this approach. But it requires too much data to be written.
E.g. when loading an OS guest VM reads gigabytes of data. Writing all 
this data to file incurs significant slowdown.

Pavel Dovgalyuk