Re: [PATCH for-4.15] xen/vgic: Implement write to ISPENDR in vGICv{2, 3}

From: Julien Grall <julien@xen.org>
To: Stefano Stabellini <sstabellini@kernel.org>,
	Bertrand Marquis <Bertrand.Marquis@arm.com>
Cc: "xen-devel@lists.xenproject.org" <xen-devel@lists.xenproject.org>,
	"iwj@xenproject.org" <iwj@xenproject.org>,
	Julien Grall <jgrall@amazon.com>,
	Volodymyr Babchuk <Volodymyr_Babchuk@epam.com>,
	dfaggioli@suse.com, George.Dunlap@citrix.com
Subject: Re: [PATCH for-4.15] xen/vgic: Implement write to ISPENDR in vGICv{2, 3}
Date: Tue, 23 Feb 2021 10:31:18 +0000	[thread overview]
Message-ID: <767e2028-ca86-bd0f-e936-c386066c11c8@xen.org> (raw)
In-Reply-To: <alpine.DEB.2.21.2102221253390.3234@sstabellini-ThinkPad-T480s>

Hi Stefano,

On 23/02/2021 01:24, Stefano Stabellini wrote:
> On Mon, 22 Feb 2021, Bertrand Marquis wrote:
>>> On 20 Feb 2021, at 14:04, Julien Grall <julien@xen.org> wrote:
> The consequence of this patch is that a guest can cause vcpu_unblock(v),
> hence vcpu_wake(v), to be called for its own vcpu, which basically
> always changes v to RUNSTATE_runnable. However, that alone shouldn't
> allow v to always come up ahead of any other vcpus in the queue, right?
> It should be safe. I just wanted a second opinion on this :-)

vcpu_wake() only tells the scheduler that the vCPU can be run, it is 
then up to the scheduler to decide what to do. AFAIU, for credit{1, 2}, 
each vCPU will have some credit. If you run out of credit, then you vCPU 
will not be descheduled if there is work do it.

> 
> It was possible to trigger interrupts for your own vcpus even before, but
> now the code path is going to be direct for virtual interrupts.

You can already trigger "directly" virtual interrupts using the event 
channels.

Cheers,

-- 
Julien Grall