Handle EWOULDBLOCK the same way as EPIPE. It seems the TCP socket layer

can return EWOULDBLOCK on some occasion when the connection is broken.
This commit is contained in:
bouyer 2009-11-05 08:11:24 +00:00
parent 96b6b6617b
commit 5dabc83466
1 changed files with 3 additions and 3 deletions

View File

@ -1,4 +1,4 @@
/* $NetBSD: nfs_socket.c,v 1.181 2009/10/16 23:36:05 pooka Exp $ */
/* $NetBSD: nfs_socket.c,v 1.182 2009/11/05 08:11:24 bouyer Exp $ */
/*
* Copyright (c) 1989, 1991, 1993, 1995
@ -39,7 +39,7 @@
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: nfs_socket.c,v 1.181 2009/10/16 23:36:05 pooka Exp $");
__KERNEL_RCSID(0, "$NetBSD: nfs_socket.c,v 1.182 2009/11/05 08:11:24 bouyer Exp $");
#ifdef _KERNEL_OPT
#include "fs_nfs.h"
@ -1145,7 +1145,7 @@ tryagain:
/*
* Wait for the reply from our send or the timer's.
*/
if (!error || error == EPIPE)
if (!error || error == EPIPE || error == EWOULDBLOCK)
error = nfs_reply(rep, lwp);
/*