On Wed, 18 Dec 2013 16:17:39 -0500
Konrad Rzeszutek Wilk <[email protected]> wrote:
> On Wed, Dec 18, 2013 at 06:31:43PM +0000, Stefano Stabellini wrote:
> > On Tue, 17 Dec 2013, Konrad Rzeszutek Wilk wrote:
> > > From: Mukesh Rathor <[email protected]>
> > >
> > > PVH is a PV guest with a twist - there are certain things
> > > that work in it like HVM and some like PV. There is
> > > a similar mode - PVHVM where we run in HVM mode with
> > > PV code enabled - and this patch explores that.
> > >
> > > The most notable PV interfaces are the XenBus and event channels.
> > > For PVH, we will use XenBus and event channels.
> > >
> > > For the XenBus mechanism we piggyback on how it is done for
> > > PVHVM guests.
> > >
> > > Ditto for the event channel mechanism - we piggyback on PVHVM -
> > > by setting up a specific vector callback and that
> > > vector ends up calling the event channel mechanism to
> > > dispatch the events as needed.
> > >
> > > This means that from a pvops perspective, we can use
> > > native_irq_ops instead of the Xen PV specific. Albeit in the
> > > future we could support pirq_eoi_map. But that is
> > > a feature request that can be shared with PVHVM.
> > >
> > > Signed-off-by: Mukesh Rathor <[email protected]>
> > > Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
> > > ---
> > > arch/x86/xen/enlighten.c | 6 ++++++
> > > arch/x86/xen/irq.c | 5 ++++-
> > > drivers/xen/events.c | 5 +++++
> > > drivers/xen/xenbus/xenbus_client.c | 3 ++-
> > > 4 files changed, 17 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
> > > index e420613..7fceb51 100644
> > > --- a/arch/x86/xen/enlighten.c
> > > +++ b/arch/x86/xen/enlighten.c
> > > @@ -1134,6 +1134,8 @@ void xen_setup_shared_info(void)
> > > /* In UP this is as good a place as any to set up shared
> > > info */ xen_setup_vcpu_info_placement();
> > > #endif
> > > + if (xen_pvh_domain())
> > > + return;
> > >
> > > xen_setup_mfn_list_list();
> > > }
> >
> > This is another one of those cases where I think we would benefit
> > from introducing xen_setup_shared_info_pvh instead of adding more
> > ifs here.
>
> Actually this one can be removed.
>
> >
> >
> > > @@ -1146,6 +1148,10 @@ void xen_setup_vcpu_info_placement(void)
> > > for_each_possible_cpu(cpu)
> > > xen_vcpu_setup(cpu);
> > >
> > > + /* PVH always uses native IRQ ops */
> > > + if (xen_pvh_domain())
> > > + return;
> > > +
> > > /* xen_vcpu_setup managed to place the vcpu_info within
> > > the percpu area for all cpus, so make use of it */
> > > if (have_vcpu_info_placement) {
> >
> > Same here?
>
> Hmmm, I wonder if the vcpu info placement could work with PVH.
It should now (after a patch I sent while ago)... the comment implies
that PVH uses native IRQs even case of vcpu info placlement...
perhaps it would be more clear to do:
for_each_possible_cpu(cpu)
xen_vcpu_setup(cpu);
/* PVH always uses native IRQ ops */
if (have_vcpu_info_placement && !xen_pvh_domain) {
pv_irq_ops.save_fl = __PV_IS_CALLEE_SAVE(xen_save_fl_direct);
.........
On Fri, 3 Jan 2014, Mukesh Rathor wrote:
> On Wed, 18 Dec 2013 16:17:39 -0500
> Konrad Rzeszutek Wilk <[email protected]> wrote:
>
> > On Wed, Dec 18, 2013 at 06:31:43PM +0000, Stefano Stabellini wrote:
> > > On Tue, 17 Dec 2013, Konrad Rzeszutek Wilk wrote:
> > > > From: Mukesh Rathor <[email protected]>
> > > >
> > > > PVH is a PV guest with a twist - there are certain things
> > > > that work in it like HVM and some like PV. There is
> > > > a similar mode - PVHVM where we run in HVM mode with
> > > > PV code enabled - and this patch explores that.
> > > >
> > > > The most notable PV interfaces are the XenBus and event channels.
> > > > For PVH, we will use XenBus and event channels.
> > > >
> > > > For the XenBus mechanism we piggyback on how it is done for
> > > > PVHVM guests.
> > > >
> > > > Ditto for the event channel mechanism - we piggyback on PVHVM -
> > > > by setting up a specific vector callback and that
> > > > vector ends up calling the event channel mechanism to
> > > > dispatch the events as needed.
> > > >
> > > > This means that from a pvops perspective, we can use
> > > > native_irq_ops instead of the Xen PV specific. Albeit in the
> > > > future we could support pirq_eoi_map. But that is
> > > > a feature request that can be shared with PVHVM.
> > > >
> > > > Signed-off-by: Mukesh Rathor <[email protected]>
> > > > Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
> > > > ---
> > > > arch/x86/xen/enlighten.c | 6 ++++++
> > > > arch/x86/xen/irq.c | 5 ++++-
> > > > drivers/xen/events.c | 5 +++++
> > > > drivers/xen/xenbus/xenbus_client.c | 3 ++-
> > > > 4 files changed, 17 insertions(+), 2 deletions(-)
> > > >
> > > > diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
> > > > index e420613..7fceb51 100644
> > > > --- a/arch/x86/xen/enlighten.c
> > > > +++ b/arch/x86/xen/enlighten.c
> > > > @@ -1134,6 +1134,8 @@ void xen_setup_shared_info(void)
> > > > /* In UP this is as good a place as any to set up shared
> > > > info */ xen_setup_vcpu_info_placement();
> > > > #endif
> > > > + if (xen_pvh_domain())
> > > > + return;
> > > >
> > > > xen_setup_mfn_list_list();
> > > > }
> > >
> > > This is another one of those cases where I think we would benefit
> > > from introducing xen_setup_shared_info_pvh instead of adding more
> > > ifs here.
> >
> > Actually this one can be removed.
> >
> > >
> > >
> > > > @@ -1146,6 +1148,10 @@ void xen_setup_vcpu_info_placement(void)
> > > > for_each_possible_cpu(cpu)
> > > > xen_vcpu_setup(cpu);
> > > >
> > > > + /* PVH always uses native IRQ ops */
> > > > + if (xen_pvh_domain())
> > > > + return;
> > > > +
> > > > /* xen_vcpu_setup managed to place the vcpu_info within
> > > > the percpu area for all cpus, so make use of it */
> > > > if (have_vcpu_info_placement) {
> > >
> > > Same here?
> >
> > Hmmm, I wonder if the vcpu info placement could work with PVH.
>
> It should now (after a patch I sent while ago)... the comment implies
> that PVH uses native IRQs even case of vcpu info placlement...
>
> perhaps it would be more clear to do:
>
> for_each_possible_cpu(cpu)
> xen_vcpu_setup(cpu);
> /* PVH always uses native IRQ ops */
> if (have_vcpu_info_placement && !xen_pvh_domain) {
> pv_irq_ops.save_fl = __PV_IS_CALLEE_SAVE(xen_save_fl_direct);
> .........
Yeah, this looks better