Re: [PATCH 02/15] hyperv: uapi-fy synic event flags definitions

From: Roman Kagan <rkagan@virtuozzo.com>
To: Stephen Hemminger <stephen@networkplumber.org>
Cc: "Paolo Bonzini" <pbonzini@redhat.com>,
	"Radim Krčmář" <rkrcmar@redhat.com>,
	"K. Y. Srinivasan" <kys@microsoft.com>,
	"Vitaly Kuznetsov" <vkuznets@redhat.com>,
	kvm@vger.kernel.org, "Denis V . Lunev" <den@openvz.org>,
	"Haiyang Zhang" <haiyangz@microsoft.com>,
	x86@kernel.org, linux-kernel@vger.kernel.org,
	"Ingo Molnar" <mingo@redhat.com>,
	"H. Peter Anvin" <hpa@zytor.com>,
	devel@linuxdriverproject.org,
	"Thomas Gleixner" <tglx@linutronix.de>
Subject: Re: [PATCH 02/15] hyperv: uapi-fy synic event flags definitions
Date: Wed, 21 Dec 2016 09:33:05 +0300	[thread overview]
Message-ID: <20161221063305.GC15917@rkaganb.sw.ru> (raw)
In-Reply-To: <20161220092453.20d3a640@xeon-e3>

On Tue, Dec 20, 2016 at 09:24:53AM -0800, Stephen Hemminger wrote:
> On Tue, 20 Dec 2016 18:55:49 +0300
> Roman Kagan <rkagan@virtuozzo.com> wrote:
> 
> > Move definitions related to the Hyper-V SynIC event flags to a header
> > where they can be consumed by userspace.
> > 
> > While doing so, also clean up their use by switching to standard bitops
> > and struct-based dereferencing.  The latter is also done for message
> > pages.
> > 
> > Signed-off-by: Roman Kagan <rkagan@virtuozzo.com>
> > ---
> >  arch/x86/include/uapi/asm/hyperv.h | 13 +++++++++
> >  drivers/hv/hyperv_vmbus.h          | 24 ++--------------
> >  drivers/hv/channel_mgmt.c          | 10 +++----
> >  drivers/hv/connection.c            | 47 ++++++++++---------------------
> >  drivers/hv/vmbus_drv.c             | 57 ++++++++++++++------------------------
> >  5 files changed, 54 insertions(+), 97 deletions(-)
> > 
> > diff --git a/arch/x86/include/uapi/asm/hyperv.h b/arch/x86/include/uapi/asm/hyperv.h
> > index 6098ab5..af542a3 100644
> > --- a/arch/x86/include/uapi/asm/hyperv.h
> > +++ b/arch/x86/include/uapi/asm/hyperv.h
> > @@ -363,4 +363,17 @@ struct hv_timer_message_payload {
> >  #define HV_STIMER_AUTOENABLE		(1ULL << 3)
> >  #define HV_STIMER_SINT(config)		(__u8)(((config) >> 16) & 0x0F)
> >  
> > +/* Define synthetic interrupt controller flag constants. */
> > +#define HV_EVENT_FLAGS_COUNT		(256 * 8)
> > +
> > +/* Define the synthetic interrupt controller event flags format. */
> > +struct hv_synic_event_flags {
> > +	__u64 flags[HV_EVENT_FLAGS_COUNT / 64];
> > +};
> > +
> > +/* Define the synthetic interrupt flags page layout. */
> > +struct hv_synic_event_flags_page {
> > +	struct hv_synic_event_flags sintevent_flags[HV_SYNIC_SINT_COUNT];
> > +};
> > +
> >  #endif
> 
> How are these going to be exposed to user space?
> 
> They should really be unsigned long since there is no guarantee of atomic operation
> on 64 bit values on 32 bit architectures.

Indeed, absolutely.  These are bitmaps and should be unsigned long[], of
course.  I'll fix it in the next iteration.

Thanks,
Roman.