From: Wei Liu Subject: Re: [RFC PATCH V3 21/22] Only allow extended event channel on Dom0 and driver domains Date: Thu, 28 Feb 2013 19:29:35 +0000 Message-ID: <1362079775.2109.141.camel@zion.uk.xensource.com> References: <1361975655-22295-1-git-send-email-wei.liu2@citrix.com> <1361975655-22295-22-git-send-email-wei.liu2@citrix.com> <512F50E4.6020201@citrix.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <512F50E4.6020201@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: David Vrabel Cc: "Keir (Xen.org)" , wei.liu2@citrix.com, Ian Campbell , "jbeulich@suse.com" , "xen-devel@lists.xen.org" List-Id: xen-devel@lists.xenproject.org On Thu, 2013-02-28 at 12:43 +0000, David Vrabel wrote: > On 27/02/13 14:34, Wei Liu wrote: > > For non-Dom0 domains, add a flag to indicate whether it can use extended event > > channel, admins can specify this flag when creating a driver domain. > > > > The rationale behide this option is, a) extended event channel will > > consume global mapping space in Xen, b) likely that only Dom0 and driver > > domains need this. > > I would prefer the toolstack to limit the max port that the guest can > use (see the EVTCHNOP_set_limit sub-op I proposed), than this single bit > (which isn't useful for the FIFO-based proposal). > I'm not very keen on adding that EVTCHNOP_set_limit now, because: * limiting number of event channels in 2/3-level brings no significant improvement. * you would need to way to notify guest kernel its limit (be another OP or via Xenstore), because in any case a guest uses user space evtchn driver it would need to know the limit. In any case, the purpose of this option is to indicate whether the guest can use *any* of the extended ABIs. A better naming should be evtchn_extended_allowed, which should default to 1 in FIFO-based ABIs. Wei. > David