6216419 boot-net stalls while waiting for ARP reply generated for IP outside the client's subnet....

5 files changed, 99 insertions, 23 deletions

diff --git a/usr/src/stand/lib/fs/nfs/clnt_budp.c b/usr/src/stand/lib/fs/nfs/clnt_budp.c
index bf2e895373..cfd14574ff 100644
--- a/usr/src/stand/lib/fs/nfs/clnt_budp.c
+++ b/usr/src/stand/lib/fs/nfs/clnt_budp.c
@@ -2,9 +2,8 @@
  * CDDL HEADER START
  *
  * The contents of this file are subject to the terms of the
- * Common Development and Distribution License, Version 1.0 only
- * (the "License").  You may not use this file except in compliance
- * with the License.
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
  *
  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
  * or http://www.opensolaris.org/os/licensing.
@@ -19,8 +18,9 @@
  *
  * CDDL HEADER END
  */
+
 /*
- * Copyright 2004 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -59,6 +59,9 @@
 
 #define	dprintf if (boothowto & RB_DEBUG) printf
 
+/* retries to send RPC message when sendto fails */
+#define	RPC_UDP_SEND_RETRIES	3
+
 extern int errno;
 
 /*
@@ -244,6 +247,7 @@ clntbudp_call(cl, proc, xargs, argsp, xresults, resultsp, utimeout)
 	int nrefreshes = 2;	/* number of times to refresh cred */
 	struct timeval timeout;
 	int errors;
+	short send_retries = RPC_UDP_SEND_RETRIES;
 
 	cu = (struct cu_data *)cl->cl_private;
 	if (cu->cu_total.tv_usec == -1)
@@ -283,8 +287,24 @@ send_again:
 	if (sendto(cu->cu_sock, cu->cu_outbuf, outlen, 0,
 	    (struct sockaddr *)&(cu->cu_raddr), cu->cu_rlen)
 	    != outlen) {
+		if (errno == ETIMEDOUT) {
+			/*
+			 * sendto() times out probably because
+			 * ARP times out while waiting for reply.
+			 * We retry sending RPC message again.
+			 */
+			if (send_retries-- > 0) {
+				dprintf("clntbudp_call: timedout, try sending"
+				    "RPC again\n");
+				errno = 0;
+				goto send_again;
+			}
+			cu->cu_error.re_status = RPC_TIMEDOUT;
+		} else {
+			cu->cu_error.re_status = RPC_CANTSEND;
+		}
 		cu->cu_error.re_errno = errno;
-		return (cu->cu_error.re_status = RPC_CANTSEND);
+		return (cu->cu_error.re_status);
 	}
 
 	/*
diff --git a/usr/src/stand/lib/fs/nfs/mount.c b/usr/src/stand/lib/fs/nfs/mount.c
index 89d5c2f71a..760aac3414 100644
--- a/usr/src/stand/lib/fs/nfs/mount.c
+++ b/usr/src/stand/lib/fs/nfs/mount.c
@@ -2,9 +2,8 @@
  * CDDL HEADER START
  *
  * The contents of this file are subject to the terms of the
- * Common Development and Distribution License, Version 1.0 only
- * (the "License").  You may not use this file except in compliance
- * with the License.
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
  *
  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
  * or http://www.opensolaris.org/os/licensing.
@@ -19,8 +18,9 @@
  *
  * CDDL HEADER END
  */
+
 /*
- * Copyright 2004 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -60,6 +60,9 @@
 #include "dhcpv4.h"
 #include <sys/mntent.h>
 
+/* ARP timeout in milliseconds for BOOTP/RARP */
+#define	ARP_INETBOOT_TIMEOUT 1000
+
 struct nfs_file		roothandle;			/* root file handle */
 static char		root_hostname[SYS_NMLN];	/* server hostname */
 static char		my_hostname[MAXHOSTNAMELEN];
@@ -73,6 +76,7 @@ char			rootopts[MAX_PATH_LEN];
 static gid_t		fake_gids = 1;	/* fake gids list for auth_unix */
 
 extern void set_default_filename(char *);	/* boot.c */
+extern void mac_set_arp_timeout(unsigned int);	/* mac.c */
 
 /*
  * xdr routines used by mount.
@@ -604,6 +608,23 @@ boot_nfs_mountroot(char *str)
 		 * services are local. Use DHCP if this bothers you.
 		 */
 		dontroute = TRUE;
+		/*
+		 * We are trying to keep the ARP response
+		 * timeout on the lower side with BOOTP/RARP.
+		 * We are doing this for BOOTP/RARP where policy
+		 * doesn't allow to route the packets outside
+		 * the subnet as it has no idea about the
+		 * netmask. By doing so, we are reducing
+		 * ARP response timeout for any packet destined
+		 * for outside booting clients subnet. Client can
+		 * not expect such ARP replies and will finally
+		 * timeout after a long delay. This would cause
+		 * booting client to get stalled for a longer
+		 * time. We can not avoid accepting any outside
+		 * subnet packets accidentally destined for the
+		 * booting client.
+		 */
+		mac_set_arp_timeout(ARP_INETBOOT_TIMEOUT);
 
 		/* now that we have an IP address, turn off promiscuous mode */
 		(void) ipv4_setpromiscuous(FALSE);
diff --git a/usr/src/stand/lib/fs/nfs/rpc.c b/usr/src/stand/lib/fs/nfs/rpc.c
index 55b76fe154..db813a2d1d 100644
--- a/usr/src/stand/lib/fs/nfs/rpc.c
+++ b/usr/src/stand/lib/fs/nfs/rpc.c
@@ -2,9 +2,8 @@
  * CDDL HEADER START
  *
  * The contents of this file are subject to the terms of the
- * Common Development and Distribution License, Version 1.0 only
- * (the "License").  You may not use this file except in compliance
- * with the License.
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
  *
  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
  * or http://www.opensolaris.org/os/licensing.
@@ -19,8 +18,9 @@
  *
  * CDDL HEADER END
  */
+
 /*
- * Copyright 2004 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  *
  * This file contains a simple implementation of RPC. Standard XDR is
@@ -348,7 +348,17 @@ brpc_call(
 	do {
 		if (sendto(s, trm_msg, xmit_len, flags, (struct sockaddr *)to,
 		    sizeof (struct sockaddr_in)) < 0) {
-			rpc_error.re_status = RPC_CANTSEND;
+			/*
+			 * If errno is set to ETIMEDOUT, return
+			 * with RPC status as RPC_TIMEDOUT. Calling
+			 * funciton will take care of this error by
+			 * retrying the RPC call.
+			 */
+			if (errno == ETIMEDOUT) {
+				rpc_error.re_status = RPC_TIMEDOUT;
+			} else {
+				rpc_error.re_status = RPC_CANTSEND;
+			}
 			goto gt_error;
 		}
 
diff --git a/usr/src/stand/lib/inet/mac.c b/usr/src/stand/lib/inet/mac.c
index 009523b5f9..a1a9f2dd59 100644
--- a/usr/src/stand/lib/inet/mac.c
+++ b/usr/src/stand/lib/inet/mac.c
@@ -2,9 +2,8 @@
  * CDDL HEADER START
  *
  * The contents of this file are subject to the terms of the
- * Common Development and Distribution License, Version 1.0 only
- * (the "License").  You may not use this file except in compliance
- * with the License.
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
  *
  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
  * or http://www.opensolaris.org/os/licensing.
@@ -19,8 +18,9 @@
  *
  * CDDL HEADER END
  */
+
 /*
- * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -71,6 +71,8 @@ struct ofw_net_types {
 };
 #endif	/* !__i386 */
 
+void mac_set_arp_timeout(unsigned int);
+
 /*
  * given the mac type, initialize the mac interface state.
  */
@@ -395,6 +397,13 @@ mac_get_type(void)
 		return (mac_state.mac_type);
 }
 
+void
+mac_set_arp_timeout(unsigned int timeout)
+{
+	mac_state.mac_arp_timeout =
+	    timeout;
+}
+
 int
 mac_get_arp_timeout(void)
 {
diff --git a/usr/src/stand/lib/tcp/tcp.c b/usr/src/stand/lib/tcp/tcp.c
index c6f8afe0e8..7525ce3f83 100644
--- a/usr/src/stand/lib/tcp/tcp.c
+++ b/usr/src/stand/lib/tcp/tcp.c
@@ -2,9 +2,8 @@
  * CDDL HEADER START
  *
  * The contents of this file are subject to the terms of the
- * Common Development and Distribution License, Version 1.0 only
- * (the "License").  You may not use this file except in compliance
- * with the License.
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
  *
  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
  * or http://www.opensolaris.org/os/licensing.
@@ -19,8 +18,9 @@
  *
  * CDDL HEADER END
  */
+
 /*
- * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  *
  * tcp.c, Code implementing the TCP protocol.
@@ -678,6 +678,16 @@ tcp_send(int sock_id, tcp_t *tcp, const void *msg, int len)
 			return (-1);
 
 	tcp_wput_data(tcp, head, sock_id);
+	/*
+	 * errno should be reset here as it may be
+	 * set to ETIMEDOUT. This may be set by
+	 * the MAC driver in case it has timed out
+	 * waiting for ARP reply. Any segment which
+	 * was not transmitted because of ARP timeout
+	 * will be retransmitted by TCP.
+	 */
+	if (errno == ETIMEDOUT)
+		errno = 0;
 	return (cnt);
 }
 
@@ -1398,7 +1408,13 @@ tcp_connect(int sock_id)
 		/* Send out the SYN packet. */
 		ret = ipv4_tcp_output(sock_id, syn_mp);
 		freeb(syn_mp);
-		if (ret < 0) {
+		/*
+		 * errno ETIMEDOUT is set by the mac driver
+		 * in case it is not able to receive ARP reply.
+		 * TCP will retransmit this segment so we can
+		 * ignore the ARP timeout.
+		 */
+		if ((ret < 0) && (errno != ETIMEDOUT)) {
 			return (-1);
 		}
 		/* tcp_state_wait() will finish the 3 way handshake. */