From mboxrd@z Thu Jan 1 00:00:00 1970 From: Julien Grall Subject: Re: [PATCH v4 12/17] xen/arm: ITS: Initialize LPI irq descriptors and route Date: Wed, 22 Jul 2015 15:17:32 +0100 Message-ID: <55AFA5FC.3080704@citrix.com> References: <1436514172-3263-1-git-send-email-vijay.kilari@gmail.com> <1436514172-3263-13-git-send-email-vijay.kilari@gmail.com> <1436542215.10074.100.camel@citrix.com> <55AF9CFE.8030408@citrix.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <55AF9CFE.8030408@citrix.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Vijay Kilari , Ian Campbell Cc: Stefano Stabellini , Prasun Kapoor , Vijaya Kumar K , Tim Deegan , "xen-devel@lists.xen.org" , Stefano Stabellini , manish.jaggi@caviumnetworks.com List-Id: xen-devel@lists.xenproject.org On 22/07/15 14:39, Julien Grall wrote: > On 22/07/15 14:31, Vijay Kilari wrote: >>>> + p->desc = desc; >>> >>> This should have happened during routing, not now. >> >> While routing we don't have vlpi to update p->desc. > > Why do you need the p->desc? You don't use at all the p->desc during the > injection because you replaced all the > if ( p->desc ) by if ( p->desc && !is_lpi(p->irq) ) Looking to the spec (ARM IHI 0069A section 5.4): "Because an LPI does not have an active state, it is not possible to associate a virtual LPI with a physical interrupt." So it looks like to me that we never need the IRQ desc for LPI. Furthermore, it may need some care in gic_update_one_lr given that the active bit is never set. Regards, -- Julien Grall