From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sheng Yang Subject: Re: [PATCH 6/6] xen/hybrid: Enable grant table and xenbus Date: Wed, 3 Feb 2010 00:46:41 +0800 Message-ID: <201002030046.41799.sheng@linux.intel.com> References: <1265098747-10117-1-git-send-email-sheng@linux.intel.com> <201002022124.41019.sheng@linux.intel.com> <1265127844.24394.613.camel@zakaz.uk.xensource.com> Mime-Version: 1.0 Content-Type: Text/Plain; charset="utf-8" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1265127844.24394.613.camel@zakaz.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 Campbell Cc: xen-devel , Jeremy Fitzhardinge , Keir Fraser , "linux-kernel@vger.kernel.org" List-Id: xen-devel@lists.xenproject.org On Wednesday 03 February 2010 00:24:04 Ian Campbell wrote: > On Tue, 2010-02-02 at 13:24 +0000, Sheng Yang wrote: > > I am not sure if I understand you right, but I think the issue is, > > there is no PVonHVM drivers in Linux upstream. The drivers are > > currently maintained by OSVs, and the one in Xen upstream code only > > support 2.6.18. So I didn't take them into consideration at the time. > > True, but this is something which should be taken care of by the core > Xen-aware code not something which should be pushed down into each > driver. > > Someone who wants to add PVonHVM functionality shouldn't have to go and > remove a bunch of conditionals from each driver (or worse add > alternative clauses to each check!). > > > I think the "xen_evtchn_enable()" looks much better. Would replace > > these ugly lines in the next version. > > I think it would be cleaner to encapsulate this in the evtchn code > rather than leaking platform knowledge into each driver. IOW the evtchn > functions should return failure if event channels are not enabled and > the driver should cope with this gracefully. Agree. That what I suppose to do. What the drivers should only know is, if event channel is enabled. > Or perhaps at the xenbus driver level we should be deciding whether or > not we have enough paravirtualisation to be worth probing the drivers at > all? I think current scheme is direct enough for now. We can improve it later. -- regards Yang, Sheng