Message-ID: <Yj1/Y5ql0d0h39rX@google.com>
Date:   Fri, 25 Mar 2022 08:37:55 +0000
From:   Oliver Upton <oupton@...gle.com>
To:     Gavin Shan <gshan@...hat.com>
Cc:     kvmarm@...ts.cs.columbia.edu, maz@...nel.org,
        linux-kernel@...r.kernel.org, eauger@...hat.com,
        shan.gavin@...il.com, Jonathan.Cameron@...wei.com,
        pbonzini@...hat.com, vkuznets@...hat.com, will@...nel.org
Subject: Re: [PATCH v5 19/22] KVM: arm64: Support SDEI ioctl commands on vCPU

On Fri, Mar 25, 2022 at 03:59:50PM +0800, Gavin Shan wrote:
> Hi Oliver,
> 
> On 3/24/22 1:55 AM, Oliver Upton wrote:
> > On Tue, Mar 22, 2022 at 04:07:07PM +0800, Gavin Shan wrote:
> > > This supports ioctl commands on vCPU to manage the various object.
> > > It's primarily used by VMM to accomplish migration. The ioctl
> > > commands introduced by this are highlighted as below:
> > > 
> > >     * KVM_SDEI_CMD_GET_VCPU_EVENT_COUNT
> > >       Return the total count of vCPU events, which have been queued
> > >       on the target vCPU.
> > > 
> > >     * KVM_SDEI_CMD_GET_VCPU_EVENT
> > >     * KVM_SDEI_CMD_SET_VCPU_EVENT
> > >       Get or set vCPU events.
> > > 
> > >     * KVM_SDEI_CMD_GET_VCPU_STATE
> > >     * KVM_SDEI_CMD_SET_VCPU_STATE
> > >       Get or set vCPU state.
> > 
> > All of this GET/SET stuff can probably be added to KVM_{GET,SET}_ONE_REG
> > immediately. Just introduce new registers any time a new event comes
> > along. The only event we have at the end of this series is the
> > software-signaled event, with async PF coming later right?
> > 
> > Some special consideration is likely necessary to avoid adding a
> > register for every u64 chunk of data. I don't think we need to afford
> > userspace any illusion of granularity with these, and can probably lump
> > it all under one giant pseudoregister.
> > 
> 
> Yes, KVM_{GET,SET}_ONE_REG is the ideal interface for migration. You're
> correct we're only concerned by software signaled event and the one for
> Async PF.
> 
> I didn't look into Raghavendra's series deeply. Actually, a lump of
> registers can be avoid after 2048 byte size is specified in its
> encoding. I think 2048 bytes are enough for now since there are
> only two supported events.

When I had said 'one giant pseudoregister' I actually meant 'one
pseudoregister per event', not all of SDEI into a single
structure. Since most of this is in flux now, it is hard to point
out what/how we should migrate from conversation alone.

And on the topic of Raghavendra's series, I do not believe it is
required anymore here w/ the removal of shared events, which I'm
strongly in favor of.

Let's delve deeper into migration on the next series :)

> In the future, we probably have varied number of SDEI events to
> be migrated. In that case, we need to add a new bit to the encoding
> of the pseudo system register, so that VMM (QEMU) can support
> variable sized system register and keep reading and writing on
> these registers on migration:
> 
>     PSEUDO_SDEI_ADDR: 64-bits in width
>     PSEUDO_SDEI_DATA: has varied size
> 
> PSEUDO_SDEI_ADDR is used to (1) Indicate the size of PSEUDO_SDEI_DATA
> (2) The information to be read/written, for example the (shared/private)
> registered events on VM and vCPU, VCPU state.
> 
> PSEUDO_SDEI_DATA is used to (1) Retrieved information or that to be
> written. (2) Flags to indicate current block of information is the
> last one or not.

I don't think we have sufficient encoding space in the register ID
to allow for arbitrary length registers. Any new registers for SDEI will
need to fit into one of the predefined sizes. Note that we've already
conditioned userspace to handle registers this way and anything else is
an ABI change.

--
Thanks,
Oliver

Powered by blists - more mailing lists