From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ian Campbell Subject: Re: [PATCH v3 4/4] libxl: add support for vscsi Date: Thu, 26 Mar 2015 13:46:55 +0000 Message-ID: <1427377615.13935.59.camel@citrix.com> References: <1425635156-2357-1-git-send-email-olaf@aepfle.de> <1425635156-2357-5-git-send-email-olaf@aepfle.de> <1426088006.21353.282.camel@citrix.com> <20150312162031.GB19536@aepfle.de> <1426178782.32572.55.camel@citrix.com> <20150313134420.GC25259@aepfle.de> <1426256282.32572.154.camel@citrix.com> <20150326125548.GA21507@aepfle.de> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20150326125548.GA21507@aepfle.de> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Olaf Hering Cc: Wei Liu , Stefano Stabellini , Ian Jackson , xen-devel@lists.xen.org List-Id: xen-devel@lists.xenproject.org On Thu, 2015-03-26 at 13:55 +0100, Olaf Hering wrote: > On Fri, Mar 13, Ian Campbell wrote: > > > > > > > On Fri, 2015-03-06 at 10:45 +0100, Olaf Hering wrote: > > > > > > > + ("feature_host", bool), > > > > > > What is this feature thing? What does !host imply? > > > > > This enables raw SCSI command passthrough in xenlinux. If the flag is > > > > > off then each command is checked. The check allows many commands to > > > > > passthrough, one command needs emulation and unhandled commands are > > > > > rejected. The flag is a noop in pvops because it doesnt look at such > > > > > flag. I was told its not required, dont know the details. > > > > Sounds like it at the very least needs a better name. > > > Ok, will rename it to feature_noemul or feature_xenlinux_raw_scsicmd or > > > whatever. > > Neither of those appear to be a good description of what you say quoted > > above. > > Ian, after some thought its not clear to me why the name "feature_host" > should be changed. The config option is "feature-host", the backend > expects a property named "feature-host". So it seems natural to also > name the variable that way. In the end it just passes a boolean from the > config to xenstore, nothing else is done with it. Regardless of the current internal details feature_host gives me no clue at all what the flag does, and this is a public facing API. If not a better name then some docs (or a pointer to them) would help. > > Also, should this be a defbool or not? > > This is a good idea. I remmeber xend had code to make sure that all > devices within a vhost had the same value for "feature-host". I think > the intention was that whole physical hosts are passed to a domU. A > defbool makes it easier to preserve the initial value based on the first > config item. In which case it sounds to me like this flag should be at a per-vhost level and not at a per device level. Ian.