Re: [Xen-devel] arm/vtimer: Physical timer emulation and the physical counter

[Julien Grall <julien@xxxxxxx>]

Hi Jeff,

On 21/11/2019 15:47, Jeff Kubascik wrote:
> On 11/21/2019 10:06 AM, Julien Grall wrote:> Hi,
> > On 21/11/2019 14:31, Jeff Kubascik wrote:
> > > On 11/19/2019 7:48 AM, Julien Grall wrote:
> > > > So I think we can handle the physical timer in similar fashion to the
> > > > virtual timer. This should likely improve performance for OS still using
> > > > the physcial timer (AFAIK linux arm64 will use the virt timer by default).
> > > One caveat is the virtual timer currently masks itself when it fires. This
> > > requires the guest to unmask it every time in the interrupt handler, which is a
> > > deviation from normal ARMv8 behavior. Every RTOS port to Xen I have worked with
> > > requires this modification for the system tick to work correctly. We may need to
> > > copy this behavior for the physical timer as well.
> > That's a good point. This should be solved by Stewart's series (I
> > haven't yet reviewed it).
> That would be great if we could fix this deviation in Xen - it would be one less
> problem to deal with when porting guests.
>
> > While waiting on the series, we should still fix the problem. But I
> > would like to avoid trapping the physical timer counter register. So I
> > would suggest to adapt the emulation of the CVAL & co.
> I would propose removing phys_timer_base.offset and just make the offset zero
> for the physical timer traps. I can put together a patch for this.
I think it should be fine to remove phys_timer_base.offset completely.

Looking forward to review the patch.

Cheers,

--
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel