[openib-general] [PATCH] IPv6: Add correct padding to IPoIB link addr option

[Roland Dreier]

Sorry, somehow I left out the description of the patch saying why you
might want to apply it....


Anand Parthasarathy pointed out that draft-ietf-ipoib-ip-over-infiniband-09.txt says:

            [DISC] specifies the length of source/target option in
            number of 8-octets as indicated by a length of '3' above.
            Since the IPoIB link-layer address is only 20-octet long,
            two octets of zero MUST be prepended to fill the total
            option length of 24 octets.

The current Linux neighbour discovery code puts the padding after the
link address.  This patch fixes up ndisc.c to put the padding in the
correct place by adding a general ndisc_addr_option_pad() function
(which could be used in the future if someone ever implements RFC 3831
IPv6-over-FC or some other encapsulation that requires padding).

Signed-off-by: Roland Dreier <roland at topspin.com>

--- linux-bk.orig/net/ipv6/ndisc.c	2005-01-12 13:27:52.000000000 -0800
+++ linux-bk/net/ipv6/ndisc.c	2005-01-14 21:20:55.736745091 -0800
@@ -169,12 +169,33 @@
 
 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
 
-static u8 *ndisc_fill_option(u8 *opt, int type, void *data, int data_len)
+/*
+ * Return the padding between the option length and the start of the
+ * link addr.  Currently only IP-over-InfiniBand needs this, although
+ * if RFC 3831 IPv6-over-Fibre Channel is ever implemented it may
+ * also need a pad of 2.
+ */
+static int ndisc_addr_option_pad(unsigned short type)
+{
+	switch (type) {
+	case ARPHRD_INFINIBAND: return 2;
+	default:                return 0;
+	}
+}
+
+static u8 *ndisc_fill_addr_option(u8 *opt, int type, void *data, int data_len,
+				  unsigned short addr_type)
 {
 	int space = NDISC_OPT_SPACE(data_len);
+	int pad   = ndisc_addr_option_pad(addr_type);
 
 	opt[0] = type;
 	opt[1] = space>>3;
+
+	memset(opt + 2, 0, pad);
+	opt   += pad;
+	space -= pad;
+
 	memcpy(opt+2, data, data_len);
 	data_len += 2;
 	opt += data_len;
@@ -453,7 +474,8 @@
 	ipv6_addr_copy(&msg->target, solicited_addr);
 
 	if (inc_opt)
-		ndisc_fill_option(msg->opt, ND_OPT_TARGET_LL_ADDR, dev->dev_addr, dev->addr_len);
+		ndisc_fill_addr_option(msg->opt, ND_OPT_TARGET_LL_ADDR, dev->dev_addr,
+				       dev->addr_len, dev->type);
 
 	/* checksum */
 	msg->icmph.icmp6_cksum = csum_ipv6_magic(src_addr, daddr, len, 
@@ -536,7 +558,8 @@
 	ipv6_addr_copy(&msg->target, solicit);
 
 	if (send_llinfo)
-		ndisc_fill_option(msg->opt, ND_OPT_SOURCE_LL_ADDR, dev->dev_addr, dev->addr_len);
+		ndisc_fill_addr_option(msg->opt, ND_OPT_SOURCE_LL_ADDR, dev->dev_addr,
+				       dev->addr_len, dev->type);
 
 	/* checksum */
 	msg->icmph.icmp6_cksum = csum_ipv6_magic(&skb->nh.ipv6h->saddr,
@@ -610,7 +633,8 @@
 	opt = (u8*) (hdr + 1);
 
 	if (dev->addr_len)
-		ndisc_fill_option(opt, ND_OPT_SOURCE_LL_ADDR, dev->dev_addr, dev->addr_len);
+		ndisc_fill_addr_option(opt, ND_OPT_SOURCE_LL_ADDR, dev->dev_addr,
+				       dev->addr_len, dev->type);
 
 	/* checksum */
 	hdr->icmp6_cksum = csum_ipv6_magic(&skb->nh.ipv6h->saddr, daddr, len,
@@ -717,7 +741,8 @@
 	}
 
 	if (ndopts.nd_opts_src_lladdr) {
-		lladdr = (u8*)(ndopts.nd_opts_src_lladdr + 1);
+		lladdr = (u8*)(ndopts.nd_opts_src_lladdr + 1) +
+			ndisc_addr_option_pad(dev->type);
 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
 		if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len)) {
 			ND_PRINTK2(KERN_WARNING
@@ -874,7 +899,8 @@
 		return;
 	}
 	if (ndopts.nd_opts_tgt_lladdr) {
-		lladdr = (u8*)(ndopts.nd_opts_tgt_lladdr + 1);
+		lladdr = (u8*)(ndopts.nd_opts_tgt_lladdr + 1) +
+			ndisc_addr_option_pad(dev->type);
 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
 		if (lladdrlen != NDISC_OPT_SPACE(dev->addr_len)) {
 			ND_PRINTK2(KERN_WARNING
@@ -964,7 +990,8 @@
 	}
 
 	if (ndopts.nd_opts_src_lladdr) {
-		lladdr = (u8 *)(ndopts.nd_opts_src_lladdr + 1);
+		lladdr = (u8 *)(ndopts.nd_opts_src_lladdr + 1) +
+			ndisc_addr_option_pad(skb->dev->type);
 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
 		if (lladdrlen != NDISC_OPT_SPACE(skb->dev->addr_len))
 			goto out;
@@ -1130,7 +1157,8 @@
 		u8 *lladdr = NULL;
 		int lladdrlen;
 		if (ndopts.nd_opts_src_lladdr) {
-			lladdr = (u8*)((ndopts.nd_opts_src_lladdr)+1);
+			lladdr = (u8*)((ndopts.nd_opts_src_lladdr)+1) +
+				ndisc_addr_option_pad(skb->dev->type);
 			lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
 			if (lladdrlen != NDISC_OPT_SPACE(skb->dev->addr_len)) {
 				ND_PRINTK2(KERN_WARNING
@@ -1250,7 +1278,8 @@
 		return;
 	}
 	if (ndopts.nd_opts_tgt_lladdr) {
-		lladdr = (u8*)(ndopts.nd_opts_tgt_lladdr + 1);
+		lladdr = (u8*)(ndopts.nd_opts_tgt_lladdr + 1) +
+			ndisc_addr_option_pad(skb->dev->type);
 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
 		if (lladdrlen != NDISC_OPT_SPACE(skb->dev->addr_len)) {
 			ND_PRINTK2(KERN_WARNING
@@ -1379,7 +1408,8 @@
 	 */
 
 	if (dev->addr_len)
-		opt = ndisc_fill_option(opt, ND_OPT_TARGET_LL_ADDR, neigh->ha, dev->addr_len);
+		opt = ndisc_fill_addr_option(opt, ND_OPT_TARGET_LL_ADDR, neigh->ha,
+					     dev->addr_len, dev->type);
 
 	/*
 	 *	build redirect option and copy skb over to the new packet.