Probably better fixed by calling vcpu_sleep_nosync() a bit more
aggressively. I''ll sort out a fix based around that. Nice catch!
-- Keir
On 10/3/08 19:14, "Ben Guthro" <bguthro@virtualiron.com> wrote:
> domain_shutdown() needs to call raise_softirq(SCHEDULE_SOFTIRQ).
> This forces the current vcpu to be rescheduled and paused.
> Without this, during live migrate, if another vcpu has defer_shutdown==1
> then the current vcpu can return from domain_shutdown(),
> do a vmenter, and continue running guest code, "resume" code that
> should only be running on the receiving node!
>
> Signed-off-by: Robert Phillips <rphillips@virtualiron.com>
> Signed-off-by: Ben Guthro <bguthro@virtualiron.com>
> diff -r 4674efa621d4 xen/common/domain.c
> --- a/xen/common/domain.c
> +++ b/xen/common/domain.c
> @@ -415,6 +415,7 @@ void domain_shutdown(struct domain *d, u
> __domain_finalise_shutdown(d);
>
> spin_unlock(&d->shutdown_lock);
> + raise_softirq(SCHEDULE_SOFTIRQ);
> }
>
> void domain_resume(struct domain *d)
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel