Re: [RFC v2 11/14] linux-headers/kvm.h: add capability to forward hypercall

From: Cornelia Huck <cohuck@redhat.com>
To: Guoheyi <guoheyi@huawei.com>
Cc: Mark Rutland <mark.rutland@arm.com>,
	Peter Maydell <peter.maydell@linaro.org>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	Marc Zyngier <marc.zyngier@arm.com>,
	qemu-devel@nongnu.org, qemu-arm@nongnu.org,
	James Morse <james.morse@arm.com>,
	Paolo Bonzini <pbonzini@redhat.com>,
	wanghaibin.wang@huawei.com, Dave Martin <Dave.Martin@arm.com>
Subject: Re: [RFC v2 11/14] linux-headers/kvm.h: add capability to forward hypercall
Date: Thu, 7 Nov 2019 13:12:10 +0100	[thread overview]
Message-ID: <20191107131210.26acf011.cohuck@redhat.com> (raw)
In-Reply-To: <da720295-2a84-e38b-9828-5287daefcfa2@huawei.com>

On Thu, 7 Nov 2019 19:57:22 +0800
Guoheyi <guoheyi@huawei.com> wrote:

> On 2019/11/7 16:57, Michael S. Tsirkin wrote:
> > On Thu, Nov 07, 2019 at 09:44:36AM +0800, Guoheyi wrote:  
> >>
> >> On 2019/11/7 1:55, Cornelia Huck wrote:  
> >>> On Tue, 5 Nov 2019 17:10:53 +0800
> >>> Heyi Guo <guoheyi@huawei.com> wrote:
> >>>  
> >>>> To keep backward compatibility, we add new KVM capability
> >>>> "KVM_CAP_FORWARD_HYPERCALL" to probe whether KVM supports forwarding
> >>>> hypercall to userspace.
> >>>>
> >>>> The capability should be enabled explicitly, for we don't want user
> >>>> space application to deal with unexpected hypercall exits. After
> >>>> enabling this cap, all HVC calls unhandled by kvm will be forwarded to
> >>>> user space.
> >>>>
> >>>> Signed-off-by: Heyi Guo <guoheyi@huawei.com>
> >>>> Cc: Peter Maydell <peter.maydell@linaro.org>
> >>>> Cc: "Michael S. Tsirkin" <mst@redhat.com>
> >>>> Cc: Cornelia Huck <cohuck@redhat.com>
> >>>> Cc: Paolo Bonzini <pbonzini@redhat.com>
> >>>> Cc: Dave Martin <Dave.Martin@arm.com>
> >>>> Cc: Marc Zyngier <marc.zyngier@arm.com>
> >>>> Cc: Mark Rutland <mark.rutland@arm.com>
> >>>> Cc: James Morse <james.morse@arm.com>
> >>>> ---
> >>>>    linux-headers/linux/kvm.h |  1 +
> >>>>    target/arm/sdei.c         | 16 ++++++++++++++++
> >>>>    target/arm/sdei.h         |  2 ++
> >>>>    3 files changed, 19 insertions(+)
> >>>>
> >>>> diff --git a/linux-headers/linux/kvm.h b/linux-headers/linux/kvm.h
> >>>> index 3d9b18f7f8..36c9b3859f 100644
> >>>> --- a/linux-headers/linux/kvm.h
> >>>> +++ b/linux-headers/linux/kvm.h
> >>>> @@ -1000,6 +1000,7 @@ struct kvm_ppc_resize_hpt {
> >>>>    #define KVM_CAP_PMU_EVENT_FILTER 173
> >>>>    #define KVM_CAP_ARM_IRQ_LINE_LAYOUT_2 174
> >>>>    #define KVM_CAP_HYPERV_DIRECT_TLBFLUSH 175
> >>>> +#define KVM_CAP_FORWARD_HYPERCALL 176
> >>>>    #ifdef KVM_CAP_IRQ_ROUTING  
> >>> Is this cap upstream already? I would have thought your header sync
> >>> would have brought it in, then. (Saying this, that header sync looks
> >>> awfully small.)
> >>>
> >>> If it is not upstream yet, please split off this hunk into a separate
> >>> patch -- it's a bit annoying, but makes life easier for merging.  
> >> No, it is not upstream yet. The whole framework and interfaces between KVM
> >> and qemu are still under discussion. I'll keep in mind of this when moving
> >> forward to next steps...
> >>
> >> Thanks,
> >> HG  
> > It's best to add it in some other place meanwhile.  
> Do you mean to split this patch from the whole patch set and send it 
> separately? Sorry I'm not clear about maintainers' work and may bring 
> you some trouble...

My preferred approach:

- add a commit entitled "placeholder for headers update" that contains
  the not-yet-upstream changes in the header files you need
- base the rest of your work on that
...
<review happens, series looks good>
...
- if kernel changes are upstream: replace the placeholder patch with a
  real update (may include separate patches, if you need an additional
  header); maintainer merges
- if kernel changes are not yet upstream: maintainer merges with
  placeholder to a feature branch, replaces with real update and merges
  once kernel patches hit upstream
(not every maintainer does the second approach; they may ask you
instead to resend with a proper headers update once the kernel changes
are upstream)