From mboxrd@z Thu Jan 1 00:00:00 1970 From: Keir Fraser Subject: Re: [PATCH] Re: cs:21768 causes guest spend more time on boot up Date: Mon, 19 Jul 2010 13:24:30 +0100 Message-ID: References: <20100719121905.GL13291@whitby.uk.xensource.com> Mime-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20100719121905.GL13291@whitby.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: Tim Deegan Cc: "Zhang, Yang Z" , "Zhang, Jianwu" , "xen-devel@lists.xensource.com" , "Xu, Jiajun" List-Id: xen-devel@lists.xenproject.org On 19/07/2010 13:19, "Tim Deegan" wrote: >>> The hang turned out to be entirely unrelated to the SMBIOS tables; the >>> xenbus client zeroes out teh xenstore ring entirely, and it looks like >>> newer dom0 xenbus backends can't handle that, so: >> >> What would it have to do with an in-kernel driver? Doesn't the comms page >> only get looked at by [o]xenstored? In which case we could fix them. > > Ah, so it does. I assumed it was the kernel because that's all that > changed on my test box since I last tested this stuff. I'm using the C > xenstored and its code looks like it should work fine with the page > getting zeroed under its feet. I'll dig further. Thanks. The revised patch might be acceptable, but if possible we're better off relying on undocumented xenstored behaviour than domU frontend behaviour since the former we have full control over (albeit we now have two daemons to consider). Really nice would be some kind of explicit reset command or protocol, but in this context since there are no watches or pending requests or anything, I guess whacking the xenstore page is sufficient engineering effort. -- Keir