[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v2 2/5] vm_event: Implement ARM SMC events


> diff --git a/xen/common/vm_event.c b/xen/common/vm_event.c
> index 2906407..a29bda8 100644
> --- a/xen/common/vm_event.c
> +++ b/xen/common/vm_event.c
> @@ -818,7 +818,6 @@ int vm_event_monitor_traps(struct vcpu *v, uint8_t sync,
>          req->altp2m_idx = altp2m_vcpu_idx(v);
>      }
> -    vm_event_fill_regs(req);
>      vm_event_put_request(d, &d->vm_event->monitor, req);
>      return 1;

So now for x86 we only vm_fill_regs() for CR writes and breakpoints (and
EPT faults, but that's in p2m.c which hasn't been touched by this
patch)? That's a pretty big change, and one that's not explained in the
patch description (which makes no mention of any x86 changes).

Having that call in vm_event_monitor_traps() made sure that all
vm_events get a copy of the respective registers. In the x86 case, that
includes the guest request and MSR write events, which now no longer
seem to carry that information, unless I'm missing something.

That behaviour should not change for x86 events, please.

Yeap, good catch. It needs to be moved from the common path because the inputs to the function will differ on ARM and x86. I'll double-check that the x86 paths will remain functionally the same.

So for mem_access nothing changes in this patch, fill_regs was already called from p2m.c. For MSR's I just missed adding the extra call. As for vm_event_monitor_guest_request, it will needs to be moved to be arch-specific. I think I'll do it as a precursor patch where I move it to be in the arch-specific monitor code (where it should be actually). Will do these fixes in the next round.


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.