Re: [Qemu-devel] [RISU PATCH v3 00/10] Record/payback patches

[Peter Maydell]

On 9 December 2016 at 11:48, Alex Bennée <address@hidden> wrote:
> I did a bunch of tweaking to see if I could abstract the hackage for
> record/replay a bit more. With a little bit of light re-factoring of
> the send and recv functions I can move all the specific send/recv
> logic into the main risu.c file and avoid having copy and pasted
> functions for each architecture.
>
> I've also introduced a header which makes tracking where you are in
> the recorded stream a little easier when debugging.
>
> It should be noted the functions:
>
>   int send_register_info(write_fn write_fn, void *uc);
>   int recv_and_compare_register_info(read_fn read_fn, respond_fn respond, 
> void *uc);
>
> are pretty identical for each arch. I did consider folding even more
> of the logic into the common code but I wasn't sure if that would
> loose flexibility in the future for other architectures so I left it
> as is.
>
> I've make the same changes to ppc64 although I've been compiling and
> testing under a qemu linux-user environment which might explain why
> I'm seeing crashes when I try and run it. I would appreciate the PPC
> guys trying it on real hardware and debugging if needed.
>
> For reference I'm using a linux-user powered docker image (built with
> the debian-bootstrap recipe in the qemu.git tree) for cross compiling:
>
>   docker run --rm -it -v /home/alex/lsrc/qemu/risu.git:/src --user=alex:alex 
> -w /src debian:ppc64el make
>
> There have been some minor clean-ups to the helper scripts mainly to
> aid testing by allowing the override of the RISU binary (I have
> risu-arm, risu-arm64 and risu-ppc64 in my development directory).
> Notes in the --- comments on each patch.

This generally looks good -- I've applied the two trivial typo
fixes, and commented on the other patches except for the new
scripts. I don't think there's much more work to do to get these
applied.

Could we have a patch which updates README to document the
new functionality?

I won't have a chance to look at the scripts until next year, I
expect, but from a quick glance they seem to be reasonably
separate from the actual record/replay implementation.

thanks
-- PMM