Return-Path: Received: from relay1.sgi.com ([192.48.179.29]:50526 "EHLO relay.sgi.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757299Ab0JUT0k (ORCPT ); Thu, 21 Oct 2010 15:26:40 -0400 Subject: [PATCH] xs_bind retry binding forever To: Trond.Myklebust@netapp.com From: Ben Myers Cc: linux-nfs@vger.kernel.org Date: Thu, 21 Oct 2010 13:33:32 -0500 Message-ID: <20101021183203.12776.28469.stgit@lady3jane.americas.sgi.com> Content-Type: text/plain; charset="utf-8" Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 Hey Trond, I've been working on a bug related to autofs where a nfs mount returns EIO in the middle of a job at which point things start to fail. It turns out the machine isn't running nscd and that is causing a large number of reserved source ports to be used over localhost and they crowd out the nfs client. Prior to commit 67a391d72ca7efb387c30ec761a487e50a3ff085 this was not a problem because we retried forever until we got a reserved source port. It's ok to say 'use nscd', but in many cases nscd is not turned on by default so I'd like to propose we revert to the previous behavior: retry binding forever. Thanks, Ben --- Ben Myers (1): xs_bind retry binding forever net/sunrpc/xprtsock.c | 22 ++++++++++++++++++---- 1 files changed, 18 insertions(+), 4 deletions(-) --