Re: [RFC PATCH 07/16] x86/xen: make vcpu_info part of xenhost_t

From: Ankur Arora
Date: Mon Jun 17 2019 - 02:34:21 EST


On 2019-06-14 4:53 a.m., Juergen Gross wrote:
On 09.05.19 19:25, Ankur Arora wrote:
Abstract out xen_vcpu_id probing via (*probe_vcpu_id)(). Once that is
availab,e the vcpu_info registration happens via the VCPUOP hypercall.

Note that for the nested case, there are two vcpu_ids, and two vcpu_info
areas, one each for the default xenhost and the remote xenhost.
The vcpu_info is used via pv_irq_ops, and evtchn signaling.

The other VCPUOP hypercalls are used for management (and scheduling)
which is expected to be done purely in the default hypervisor.
However, scheduling of L1-guest does imply L0-Xen-vcpu_info switching,
which might mean that the remote hypervisor needs some visibility
into related events/hypercalls in the default hypervisor.

Another candidate for dropping due to layering violation, I guess.
Yeah, a more narrowly tailored interface, where perhaps the L1-Xen
maps events for L0-Xen makes sense.
Also, just realized that given that L0-Xen has no control over
scheduling of L1-Xen's guests (some of which it might want to
send events to), it makes sense for L1-Xen to have some state
for guest evtchns which pertain to L0-Xen.


Ankur



Juergen