From: "Talpey, Thomas" Subject: Re: [PATCH REPOST2] svcrdma: Limit ORD based on client's advertised IRD Date: Tue, 20 May 2008 22:02:27 -0400 Message-ID: References: <1211242295.31725.95.camel@trinity.ogc.int> <20080520212434.GF8177@fieldses.org> <1211322408.12428.34.camel@trinity.ogc.int> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Cc: linux-nfs@vger.kernel.org To: Tom Tucker , "J. Bruce Fields" Return-path: Received: from mx2.netapp.com ([216.240.18.37]:57147 "EHLO mx2.netapp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932689AbYEUCC0 (ORCPT ); Tue, 20 May 2008 22:02:26 -0400 In-Reply-To: <1211322408.12428.34.camel-SMNkleLxa3ZimH42XvhXlA@public.gmane.org> References: <1211242295.31725.95.camel-SMNkleLxa3ZimH42XvhXlA@public.gmane.org> <20080520212434.GF8177@fieldses.org> <1211322408.12428.34.camel-SMNkleLxa3ZimH42XvhXlA@public.gmane.org> Sender: linux-nfs-owner@vger.kernel.org List-ID: At 06:26 PM 5/20/2008, Tom Tucker wrote: > >On Tue, 2008-05-20 at 17:24 -0400, J. Bruce Fields wrote: >> On Mon, May 19, 2008 at 07:11:35PM -0500, Tom Tucker wrote: >> > When adapters have differing IRD limits, the RDMA transport will fail to >> > connect properly. >> >> How will this look to the user? Are they going to get some sort of >> mysterious failure when they try to mount, and then have to go echo >> lower values into /proc/sys/sunrpc/svc_rdma/max_outbound_read_requests >> until it works? >> > >Yes, the mount will fail with a connection error. It's disgusting. The client error is ECONNREFUSED, which mount.nfs ever-so-helpfully reports as "mount: internal error". Chuck and I went over some options at cthon to make this more useful, I'm fishing around, it's a simple error code choice. Tom.