From: Qing He Subject: Re: [PATCH] ioemu: passthough: add no_wb option for pci conf write Date: Tue, 10 Nov 2009 09:36:07 +0800 Message-ID: <20091110013607.GA26056@ub-qhe2> References: <20091106091756.GA29376@ub-qhe2> <19188.26651.778508.710707@mariner.uk.xensource.com> <20091109030345.GA16917@ub-qhe2> <19192.20266.894840.233915@mariner.uk.xensource.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <19192.20266.894840.233915@mariner.uk.xensource.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Ian Jackson Cc: "xen-devel@lists.xensource.com" List-Id: xen-devel@lists.xenproject.org On Tue, 2009-11-10 at 01:19 +0800, Ian Jackson wrote: > The problem occurs because of the previous assumption that guest PCI > config space writes are all write-through, possibly with some > modification to the written value. The new flag prevents the > write-through (not a write _back_). Well, if you consider it as the register write back stage of an instruction pipeline, the term becomes natural. It's the write of a read-execute-write pattern, that's why `back' is used. > > In which case I think it's fine if somewhat misnamed. But it would be > better to consider whether the assumpion is actually valid; perhaps it > would be better for the write handlers to explicitly do the write to > real config space themselves if they need it ? I'd like to make the change as small. If the write is moved to the handler, all the handlers have to change. And for the current generalized pci config space algorithm, I think its logics is quite clear. There was even discussion that long run letting QEmu write physical pci config space is not desirable, either it be moved to some place like pcistub or hypervisor audit every write. For now, I think a few line change is fine.