From mboxrd@z Thu Jan 1 00:00:00 1970 From: Joerg Roedel Subject: Re: [v4 2/8] iommu, x86: Define new irte structure for VT-d Posted-Interrupts Date: Tue, 31 Mar 2015 17:16:43 +0200 Message-ID: <20150331151643.GF22683@8bytes.org> References: <1422864424-19411-1-git-send-email-feng.wu@intel.com> <1422864424-19411-3-git-send-email-feng.wu@intel.com> <20150323115754.GN4441@8bytes.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org To: "Wu, Feng" Cc: "dwmw2@infradead.org" , "jiang.liu@linux.intel.com" , "iommu@lists.linux-foundation.org" , "linux-kernel@vger.kernel.org" List-Id: iommu@lists.linux-foundation.org On Tue, Mar 24, 2015 at 02:32:01AM +0000, Wu, Feng wrote: > > I think it is better to put this as a union into struct irte. It saves > > memory and unnecessary casting in later patches. > > Thanks for the comments! Thinking more about this, I think its probably fine to keep the two versions of the irte seperate like in this patch-set. It allows to update the non-posted irte when the posted irte is active at the moment and makes the transition between both irte variants easier. But what I still don't like is the type casting necessary when calling modify_irte(). Can you abstract this and put the decission whether irte or irte_pi is set active into modify_irte? It required to change the interface of modify_irte, but that should be easy. Joerg