Re: [Xen-devel] [PATCH v3 2/6] VMX: Properly handle pi when all the assigned devices are removed

["Wu, Feng" <feng.wu@xxxxxxxxx>]

> -----Original Message-----
> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> Sent: Thursday, September 1, 2016 6:24 PM
> To: Wu, Feng <feng.wu@xxxxxxxxx>
> Cc: andrew.cooper3@xxxxxxxxxx; dario.faggioli@xxxxxxxxxx;
> george.dunlap@xxxxxxxxxxxxx; Tian, Kevin <kevin.tian@xxxxxxxxx>; xen-
> devel@xxxxxxxxxxxxx
> Subject: RE: [PATCH v3 2/6] VMX: Properly handle pi when all the assigned
> devices are removed
> 
> >>> On 01.09.16 at 11:22, <feng.wu@xxxxxxxxx> wrote:
> >> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> >> Sent: Thursday, September 1, 2016 4:21 PM
> >> >>> On 31.08.16 at 05:56, <feng.wu@xxxxxxxxx> wrote:
> >> > This patch handles some concern cases when the last assigned device
> >> > is removed from the domain. In this case we should carefully handle
> >> > pi descriptor and the per-cpu blocking list, to make sure:
> >> > - all the PI descriptor are in the right state when next time a
> >> > devices is assigned to the domain again.
> >> > - No remaining vcpus of the domain in the per-cpu blocking list.
> >> >
> >> > Basically, we pause the domain before zapping the PI hooks and
> >> > removing the vCPU from the blocking list, then unpause it after
> >> > that.
> >> >
> >> > Signed-off-by: Feng Wu <feng.wu@xxxxxxxxx>
> >>
> >> Looks plausible, but
> >> a) as already for patch 1 I'm missing information on what changed
> >>    since v2 and
> >
> > The biggest changes since v2 is that we use domain pause/unpause
> > (suggested by George) to handle the concern case, while v2 was using
> > some ugly and tricky method to do it, which was considered as hard
> > to maintain.
> >
> >> b) doesn't this make unnecessary patch 1?
> >
> > The purpose of patch 1 is to make sure the two hooks are installed
> > while CPU side PI is available event VT-d PI is not supported, I cannot
> > see why this patch will make it unnecessary.
> 
> So I guess this doesn't hold anymore with your subsequent reply
> to my comments on patch 1?

I think we still need patch 1, since we may pause a runnable vcpu
which is in the run queue ('SN' bit is set), and 'SN' continues to be set
when the vCPU is paused. If we zap these two hooks, the 'SN' can still
be set after the vCPU is unpaused

Thanks,
Feng

> 
> Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel