Merge branch 'linus' into cont_syslog
[safe/jmp/linux-2.6] / net / ipv6 / exthdrs.c
index ffcda45..8a659f9 100644 (file)
@@ -7,8 +7,6 @@
  *     Andi Kleen              <ak@muc.de>
  *     Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
  *
- *     $Id: exthdrs.c,v 1.13 2001/06/19 15:58:56 davem Exp $
- *
  *     This program is free software; you can redistribute it and/or
  *      modify it under the terms of the GNU General Public License
  *      as published by the Free Software Foundation; either version
@@ -16,7 +14,7 @@
  */
 
 /* Changes:
- *     yoshfuji                : ensure not to overrun while parsing 
+ *     yoshfuji                : ensure not to overrun while parsing
  *                               tlv options.
  *     Mitsuru KANDA @USAGI and: Remove ipv6_parse_exthdrs().
  *     YOSHIFUJI Hideaki @USAGI  Register inbound extension header
 #include <linux/types.h>
 #include <linux/socket.h>
 #include <linux/sockios.h>
-#include <linux/sched.h>
 #include <linux/net.h>
 #include <linux/netdevice.h>
 #include <linux/in6.h>
 #include <linux/icmpv6.h>
+#include <linux/slab.h>
 
+#include <net/dst.h>
 #include <net/sock.h>
 #include <net/snmp.h>
 
 #include <net/ndisc.h>
 #include <net/ip6_route.h>
 #include <net/addrconf.h>
+#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
+#include <net/xfrm.h>
+#endif
 
 #include <asm/uaccess.h>
 
+int ipv6_find_tlv(struct sk_buff *skb, int offset, int type)
+{
+       const unsigned char *nh = skb_network_header(skb);
+       int packet_len = skb->tail - skb->network_header;
+       struct ipv6_opt_hdr *hdr;
+       int len;
+
+       if (offset + 2 > packet_len)
+               goto bad;
+       hdr = (struct ipv6_opt_hdr *)(nh + offset);
+       len = ((hdr->hdrlen + 1) << 3);
+
+       if (offset + len > packet_len)
+               goto bad;
+
+       offset += 2;
+       len -= 2;
+
+       while (len > 0) {
+               int opttype = nh[offset];
+               int optlen;
+
+               if (opttype == type)
+                       return offset;
+
+               switch (opttype) {
+               case IPV6_TLV_PAD0:
+                       optlen = 1;
+                       break;
+               default:
+                       optlen = nh[offset + 1] + 2;
+                       if (optlen > len)
+                               goto bad;
+                       break;
+               }
+               offset += optlen;
+               len -= optlen;
+       }
+       /* not_found */
+ bad:
+       return -1;
+}
+EXPORT_SYMBOL_GPL(ipv6_find_tlv);
+
 /*
  *     Parsing tlv encoded headers.
  *
@@ -67,7 +113,7 @@ struct tlvtype_proc {
 
 static int ip6_tlvopt_unknown(struct sk_buff *skb, int optoff)
 {
-       switch ((skb->nh.raw[optoff] & 0xC0) >> 6) {
+       switch ((skb_network_header(skb)[optoff] & 0xC0) >> 6) {
        case 0: /* ignore */
                return 1;
 
@@ -78,12 +124,12 @@ static int ip6_tlvopt_unknown(struct sk_buff *skb, int optoff)
                /* Actually, it is redundant check. icmp_send
                   will recheck in any case.
                 */
-               if (ipv6_addr_is_multicast(&skb->nh.ipv6h->daddr))
+               if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr))
                        break;
        case 2: /* send ICMP PARM PROB regardless and drop packet */
                icmpv6_param_prob(skb, ICMPV6_UNK_OPTION, optoff);
                return 0;
-       };
+       }
 
        kfree_skb(skb);
        return 0;
@@ -94,19 +140,20 @@ static int ip6_tlvopt_unknown(struct sk_buff *skb, int optoff)
 static int ip6_parse_tlv(struct tlvtype_proc *procs, struct sk_buff *skb)
 {
        struct tlvtype_proc *curr;
-       int off = skb->h.raw - skb->nh.raw;
-       int len = ((skb->h.raw[1]+1)<<3);
+       const unsigned char *nh = skb_network_header(skb);
+       int off = skb_network_header_len(skb);
+       int len = (skb_transport_header(skb)[1] + 1) << 3;
 
-       if ((skb->h.raw + len) - skb->data > skb_headlen(skb))
+       if (skb_transport_offset(skb) + len > skb_headlen(skb))
                goto bad;
 
        off += 2;
        len -= 2;
 
        while (len > 0) {
-               int optlen = skb->nh.raw[off+1]+2;
+               int optlen = nh[off + 1] + 2;
 
-               switch (skb->nh.raw[off]) {
+               switch (nh[off]) {
                case IPV6_TLV_PAD0:
                        optlen = 1;
                        break;
@@ -118,9 +165,9 @@ static int ip6_parse_tlv(struct tlvtype_proc *procs, struct sk_buff *skb)
                        if (optlen > len)
                                goto bad;
                        for (curr=procs; curr->type >= 0; curr++) {
-                               if (curr->type == skb->nh.raw[off]) {
-                                       /* type specific length/alignment 
-                                          checks will be performed in the 
+                               if (curr->type == nh[off]) {
+                                       /* type specific length/alignment
+                                          checks will be performed in the
                                           func(). */
                                        if (curr->func(skb, off) == 0)
                                                return 0;
@@ -147,122 +194,206 @@ bad:
   Destination options header.
  *****************************/
 
-static struct tlvtype_proc tlvprocdestopt_lst[] = {
-       /* No destination options are defined now */
-       {-1,                    NULL}
-};
-
-static int ipv6_destopt_rcv(struct sk_buff **skbp, unsigned int *nhoffp)
+#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
+static int ipv6_dest_hao(struct sk_buff *skb, int optoff)
 {
-       struct sk_buff *skb = *skbp;
+       struct ipv6_destopt_hao *hao;
        struct inet6_skb_parm *opt = IP6CB(skb);
+       struct ipv6hdr *ipv6h = ipv6_hdr(skb);
+       struct in6_addr tmp_addr;
+       int ret;
 
-       if (!pskb_may_pull(skb, (skb->h.raw-skb->data)+8) ||
-           !pskb_may_pull(skb, (skb->h.raw-skb->data)+((skb->h.raw[1]+1)<<3))) {
-               IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
-               kfree_skb(skb);
-               return -1;
+       if (opt->dsthao) {
+               LIMIT_NETDEBUG(KERN_DEBUG "hao duplicated\n");
+               goto discard;
        }
+       opt->dsthao = opt->dst1;
+       opt->dst1 = 0;
 
-       opt->lastopt = skb->h.raw - skb->nh.raw;
-       opt->dst1 = skb->h.raw - skb->nh.raw;
+       hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) + optoff);
 
-       if (ip6_parse_tlv(tlvprocdestopt_lst, skb)) {
-               skb->h.raw += ((skb->h.raw[1]+1)<<3);
-               *nhoffp = opt->dst1;
-               return 1;
+       if (hao->length != 16) {
+               LIMIT_NETDEBUG(
+                       KERN_DEBUG "hao invalid option length = %d\n", hao->length);
+               goto discard;
        }
 
-       IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
-       return -1;
-}
+       if (!(ipv6_addr_type(&hao->addr) & IPV6_ADDR_UNICAST)) {
+               LIMIT_NETDEBUG(
+                       KERN_DEBUG "hao is not an unicast addr: %pI6\n", &hao->addr);
+               goto discard;
+       }
 
-static struct inet6_protocol destopt_protocol = {
-       .handler        =       ipv6_destopt_rcv,
-       .flags          =       INET6_PROTO_NOPOLICY,
-};
+       ret = xfrm6_input_addr(skb, (xfrm_address_t *)&ipv6h->daddr,
+                              (xfrm_address_t *)&hao->addr, IPPROTO_DSTOPTS);
+       if (unlikely(ret < 0))
+               goto discard;
 
-void __init ipv6_destopt_init(void)
-{
-       if (inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS) < 0)
-               printk(KERN_ERR "ipv6_destopt_init: Could not register protocol\n");
-}
+       if (skb_cloned(skb)) {
+               if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
+                       goto discard;
 
-/********************************
-  NONE header. No data in packet.
- ********************************/
+               /* update all variable using below by copied skbuff */
+               hao = (struct ipv6_destopt_hao *)(skb_network_header(skb) +
+                                                 optoff);
+               ipv6h = ipv6_hdr(skb);
+       }
 
-static int ipv6_nodata_rcv(struct sk_buff **skbp, unsigned int *nhoffp)
-{
-       struct sk_buff *skb = *skbp;
+       if (skb->ip_summed == CHECKSUM_COMPLETE)
+               skb->ip_summed = CHECKSUM_NONE;
+
+       ipv6_addr_copy(&tmp_addr, &ipv6h->saddr);
+       ipv6_addr_copy(&ipv6h->saddr, &hao->addr);
+       ipv6_addr_copy(&hao->addr, &tmp_addr);
 
+       if (skb->tstamp.tv64 == 0)
+               __net_timestamp(skb);
+
+       return 1;
+
+ discard:
        kfree_skb(skb);
        return 0;
 }
+#endif
 
-static struct inet6_protocol nodata_protocol = {
-       .handler        =       ipv6_nodata_rcv,
-       .flags          =       INET6_PROTO_NOPOLICY,
+static struct tlvtype_proc tlvprocdestopt_lst[] = {
+#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
+       {
+               .type   = IPV6_TLV_HAO,
+               .func   = ipv6_dest_hao,
+       },
+#endif
+       {-1,                    NULL}
 };
 
-void __init ipv6_nodata_init(void)
+static int ipv6_destopt_rcv(struct sk_buff *skb)
 {
-       if (inet6_add_protocol(&nodata_protocol, IPPROTO_NONE) < 0)
-               printk(KERN_ERR "ipv6_nodata_init: Could not register protocol\n");
+       struct inet6_skb_parm *opt = IP6CB(skb);
+#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
+       __u16 dstbuf;
+#endif
+       struct dst_entry *dst;
+
+       if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
+           !pskb_may_pull(skb, (skb_transport_offset(skb) +
+                                ((skb_transport_header(skb)[1] + 1) << 3)))) {
+               IP6_INC_STATS_BH(dev_net(skb_dst(skb)->dev), ip6_dst_idev(skb_dst(skb)),
+                                IPSTATS_MIB_INHDRERRORS);
+               kfree_skb(skb);
+               return -1;
+       }
+
+       opt->lastopt = opt->dst1 = skb_network_header_len(skb);
+#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
+       dstbuf = opt->dst1;
+#endif
+
+       dst = dst_clone(skb_dst(skb));
+       if (ip6_parse_tlv(tlvprocdestopt_lst, skb)) {
+               dst_release(dst);
+               skb->transport_header += (skb_transport_header(skb)[1] + 1) << 3;
+               opt = IP6CB(skb);
+#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
+               opt->nhoff = dstbuf;
+#else
+               opt->nhoff = opt->dst1;
+#endif
+               return 1;
+       }
+
+       IP6_INC_STATS_BH(dev_net(dst->dev),
+                        ip6_dst_idev(dst), IPSTATS_MIB_INHDRERRORS);
+       dst_release(dst);
+       return -1;
 }
 
 /********************************
   Routing header.
  ********************************/
 
-static int ipv6_rthdr_rcv(struct sk_buff **skbp, unsigned int *nhoffp)
+static int ipv6_rthdr_rcv(struct sk_buff *skb)
 {
-       struct sk_buff *skb = *skbp;
        struct inet6_skb_parm *opt = IP6CB(skb);
-       struct in6_addr *addr;
+       struct in6_addr *addr = NULL;
        struct in6_addr daddr;
+       struct inet6_dev *idev;
        int n, i;
-
        struct ipv6_rt_hdr *hdr;
        struct rt0_hdr *rthdr;
+       struct net *net = dev_net(skb->dev);
+       int accept_source_route = net->ipv6.devconf_all->accept_source_route;
+
+       idev = in6_dev_get(skb->dev);
+       if (idev) {
+               if (accept_source_route > idev->cnf.accept_source_route)
+                       accept_source_route = idev->cnf.accept_source_route;
+               in6_dev_put(idev);
+       }
 
-       if (!pskb_may_pull(skb, (skb->h.raw-skb->data)+8) ||
-           !pskb_may_pull(skb, (skb->h.raw-skb->data)+((skb->h.raw[1]+1)<<3))) {
-               IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
+       if (!pskb_may_pull(skb, skb_transport_offset(skb) + 8) ||
+           !pskb_may_pull(skb, (skb_transport_offset(skb) +
+                                ((skb_transport_header(skb)[1] + 1) << 3)))) {
+               IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
+                                IPSTATS_MIB_INHDRERRORS);
                kfree_skb(skb);
                return -1;
        }
 
-       hdr = (struct ipv6_rt_hdr *) skb->h.raw;
+       hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
 
-       if (ipv6_addr_is_multicast(&skb->nh.ipv6h->daddr) ||
+       if (ipv6_addr_is_multicast(&ipv6_hdr(skb)->daddr) ||
            skb->pkt_type != PACKET_HOST) {
-               IP6_INC_STATS_BH(IPSTATS_MIB_INADDRERRORS);
+               IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
+                                IPSTATS_MIB_INADDRERRORS);
                kfree_skb(skb);
                return -1;
        }
 
 looped_back:
        if (hdr->segments_left == 0) {
-               opt->lastopt = skb->h.raw - skb->nh.raw;
-               opt->srcrt = skb->h.raw - skb->nh.raw;
-               skb->h.raw += (hdr->hdrlen + 1) << 3;
+               switch (hdr->type) {
+#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
+               case IPV6_SRCRT_TYPE_2:
+                       /* Silently discard type 2 header unless it was
+                        * processed by own
+                        */
+                       if (!addr) {
+                               IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
+                                                IPSTATS_MIB_INADDRERRORS);
+                               kfree_skb(skb);
+                               return -1;
+                       }
+                       break;
+#endif
+               default:
+                       break;
+               }
+
+               opt->lastopt = opt->srcrt = skb_network_header_len(skb);
+               skb->transport_header += (hdr->hdrlen + 1) << 3;
                opt->dst0 = opt->dst1;
                opt->dst1 = 0;
-               *nhoffp = (&hdr->nexthdr) - skb->nh.raw;
+               opt->nhoff = (&hdr->nexthdr) - skb_network_header(skb);
                return 1;
        }
 
-       if (hdr->type != IPV6_SRCRT_TYPE_0) {
-               IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
-               icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, (&hdr->type) - skb->nh.raw);
-               return -1;
-       }
-       
-       if (hdr->hdrlen & 0x01) {
-               IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
-               icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, (&hdr->hdrlen) - skb->nh.raw);
-               return -1;
+       switch (hdr->type) {
+#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
+       case IPV6_SRCRT_TYPE_2:
+               if (accept_source_route < 0)
+                       goto unknown_rh;
+               /* Silently discard invalid RTH type 2 */
+               if (hdr->hdrlen != 2 || hdr->segments_left != 1) {
+                       IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
+                                        IPSTATS_MIB_INHDRERRORS);
+                       kfree_skb(skb);
+                       return -1;
+               }
+               break;
+#endif
+       default:
+               goto unknown_rh;
        }
 
        /*
@@ -273,8 +404,11 @@ looped_back:
        n = hdr->hdrlen >> 1;
 
        if (hdr->segments_left > n) {
-               IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
-               icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, (&hdr->segments_left) - skb->nh.raw);
+               IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
+                                IPSTATS_MIB_INHDRERRORS);
+               icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
+                                 ((&hdr->segments_left) -
+                                  skb_network_header(skb)));
                return -1;
        }
 
@@ -282,19 +416,17 @@ looped_back:
           Do not damage packets queued somewhere.
         */
        if (skb_cloned(skb)) {
-               struct sk_buff *skb2 = skb_copy(skb, GFP_ATOMIC);
-               kfree_skb(skb);
                /* the copy is a forwarded packet */
-               if (skb2 == NULL) {
-                       IP6_INC_STATS_BH(IPSTATS_MIB_OUTDISCARDS);      
+               if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
+                       IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
+                                        IPSTATS_MIB_OUTDISCARDS);
+                       kfree_skb(skb);
                        return -1;
                }
-               *skbp = skb = skb2;
-               opt = IP6CB(skb2);
-               hdr = (struct ipv6_rt_hdr *) skb2->h.raw;
+               hdr = (struct ipv6_rt_hdr *)skb_transport_header(skb);
        }
 
-       if (skb->ip_summed == CHECKSUM_HW)
+       if (skb->ip_summed == CHECKSUM_COMPLETE)
                skb->ip_summed = CHECKSUM_NONE;
 
        i = n - --hdr->segments_left;
@@ -303,131 +435,148 @@ looped_back:
        addr = rthdr->addr;
        addr += i - 1;
 
+       switch (hdr->type) {
+#if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
+       case IPV6_SRCRT_TYPE_2:
+               if (xfrm6_input_addr(skb, (xfrm_address_t *)addr,
+                                    (xfrm_address_t *)&ipv6_hdr(skb)->saddr,
+                                    IPPROTO_ROUTING) < 0) {
+                       IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
+                                        IPSTATS_MIB_INADDRERRORS);
+                       kfree_skb(skb);
+                       return -1;
+               }
+               if (!ipv6_chk_home_addr(dev_net(skb_dst(skb)->dev), addr)) {
+                       IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
+                                        IPSTATS_MIB_INADDRERRORS);
+                       kfree_skb(skb);
+                       return -1;
+               }
+               break;
+#endif
+       default:
+               break;
+       }
+
        if (ipv6_addr_is_multicast(addr)) {
-               IP6_INC_STATS_BH(IPSTATS_MIB_INADDRERRORS);
+               IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
+                                IPSTATS_MIB_INADDRERRORS);
                kfree_skb(skb);
                return -1;
        }
 
        ipv6_addr_copy(&daddr, addr);
-       ipv6_addr_copy(addr, &skb->nh.ipv6h->daddr);
-       ipv6_addr_copy(&skb->nh.ipv6h->daddr, &daddr);
+       ipv6_addr_copy(addr, &ipv6_hdr(skb)->daddr);
+       ipv6_addr_copy(&ipv6_hdr(skb)->daddr, &daddr);
 
-       dst_release(xchg(&skb->dst, NULL));
+       skb_dst_drop(skb);
        ip6_route_input(skb);
-       if (skb->dst->error) {
-               skb_push(skb, skb->data - skb->nh.raw);
+       if (skb_dst(skb)->error) {
+               skb_push(skb, skb->data - skb_network_header(skb));
                dst_input(skb);
                return -1;
        }
 
-       if (skb->dst->dev->flags&IFF_LOOPBACK) {
-               if (skb->nh.ipv6h->hop_limit <= 1) {
-                       IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
+       if (skb_dst(skb)->dev->flags&IFF_LOOPBACK) {
+               if (ipv6_hdr(skb)->hop_limit <= 1) {
+                       IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)),
+                                        IPSTATS_MIB_INHDRERRORS);
                        icmpv6_send(skb, ICMPV6_TIME_EXCEED, ICMPV6_EXC_HOPLIMIT,
-                                   0, skb->dev);
+                                   0);
                        kfree_skb(skb);
                        return -1;
                }
-               skb->nh.ipv6h->hop_limit--;
+               ipv6_hdr(skb)->hop_limit--;
                goto looped_back;
        }
 
-       skb_push(skb, skb->data - skb->nh.raw);
+       skb_push(skb, skb->data - skb_network_header(skb));
        dst_input(skb);
        return -1;
+
+unknown_rh:
+       IP6_INC_STATS_BH(net, ip6_dst_idev(skb_dst(skb)), IPSTATS_MIB_INHDRERRORS);
+       icmpv6_param_prob(skb, ICMPV6_HDR_FIELD,
+                         (&hdr->type) - skb_network_header(skb));
+       return -1;
 }
 
-static struct inet6_protocol rthdr_protocol = {
+static const struct inet6_protocol rthdr_protocol = {
        .handler        =       ipv6_rthdr_rcv,
-       .flags          =       INET6_PROTO_NOPOLICY,
+       .flags          =       INET6_PROTO_NOPOLICY | INET6_PROTO_GSO_EXTHDR,
 };
 
-void __init ipv6_rthdr_init(void)
-{
-       if (inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING) < 0)
-               printk(KERN_ERR "ipv6_rthdr_init: Could not register protocol\n");
+static const struct inet6_protocol destopt_protocol = {
+       .handler        =       ipv6_destopt_rcv,
+       .flags          =       INET6_PROTO_NOPOLICY | INET6_PROTO_GSO_EXTHDR,
 };
 
-/*
-   This function inverts received rthdr.
-   NOTE: specs allow to make it automatically only if
-   packet authenticated.
-
-   I will not discuss it here (though, I am really pissed off at
-   this stupid requirement making rthdr idea useless)
-
-   Actually, it creates severe problems  for us.
-   Embryonic requests has no associated sockets,
-   so that user have no control over it and
-   cannot not only to set reply options, but
-   even to know, that someone wants to connect
-   without success. :-(
-
-   For now we need to test the engine, so that I created
-   temporary (or permanent) backdoor.
-   If listening socket set IPV6_RTHDR to 2, then we invert header.
-                                                   --ANK (980729)
- */
+static const struct inet6_protocol nodata_protocol = {
+       .handler        =       dst_discard,
+       .flags          =       INET6_PROTO_NOPOLICY,
+};
 
-struct ipv6_txoptions *
-ipv6_invert_rthdr(struct sock *sk, struct ipv6_rt_hdr *hdr)
+int __init ipv6_exthdrs_init(void)
 {
-       /* Received rthdr:
+       int ret;
 
-          [ H1 -> H2 -> ... H_prev ]  daddr=ME
+       ret = inet6_add_protocol(&rthdr_protocol, IPPROTO_ROUTING);
+       if (ret)
+               goto out;
 
-          Inverted result:
-          [ H_prev -> ... -> H1 ] daddr =sender
+       ret = inet6_add_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
+       if (ret)
+               goto out_rthdr;
 
-          Note, that IP output engine will rewrite this rthdr
-          by rotating it left by one addr.
-        */
+       ret = inet6_add_protocol(&nodata_protocol, IPPROTO_NONE);
+       if (ret)
+               goto out_destopt;
 
-       int n, i;
-       struct rt0_hdr *rthdr = (struct rt0_hdr*)hdr;
-       struct rt0_hdr *irthdr;
-       struct ipv6_txoptions *opt;
-       int hdrlen = ipv6_optlen(hdr);
-
-       if (hdr->segments_left ||
-           hdr->type != IPV6_SRCRT_TYPE_0 ||
-           hdr->hdrlen & 0x01)
-               return NULL;
+out:
+       return ret;
+out_rthdr:
+       inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
+out_destopt:
+       inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
+       goto out;
+};
 
-       n = hdr->hdrlen >> 1;
-       opt = sock_kmalloc(sk, sizeof(*opt) + hdrlen, GFP_ATOMIC);
-       if (opt == NULL)
-               return NULL;
-       memset(opt, 0, sizeof(*opt));
-       opt->tot_len = sizeof(*opt) + hdrlen;
-       opt->srcrt = (void*)(opt+1);
-       opt->opt_nflen = hdrlen;
-
-       memcpy(opt->srcrt, hdr, sizeof(*hdr));
-       irthdr = (struct rt0_hdr*)opt->srcrt;
-       /* Obsolete field, MBZ, when originated by us */
-       irthdr->bitmap = 0;
-       opt->srcrt->segments_left = n;
-       for (i=0; i<n; i++)
-               memcpy(irthdr->addr+i, rthdr->addr+(n-1-i), 16);
-       return opt;
+void ipv6_exthdrs_exit(void)
+{
+       inet6_del_protocol(&nodata_protocol, IPPROTO_NONE);
+       inet6_del_protocol(&destopt_protocol, IPPROTO_DSTOPTS);
+       inet6_del_protocol(&rthdr_protocol, IPPROTO_ROUTING);
 }
 
 /**********************************
   Hop-by-hop options.
  **********************************/
 
+/*
+ * Note: we cannot rely on skb_dst(skb) before we assign it in ip6_route_input().
+ */
+static inline struct inet6_dev *ipv6_skb_idev(struct sk_buff *skb)
+{
+       return skb_dst(skb) ? ip6_dst_idev(skb_dst(skb)) : __in6_dev_get(skb->dev);
+}
+
+static inline struct net *ipv6_skb_net(struct sk_buff *skb)
+{
+       return skb_dst(skb) ? dev_net(skb_dst(skb)->dev) : dev_net(skb->dev);
+}
+
 /* Router Alert as of RFC 2711 */
 
 static int ipv6_hop_ra(struct sk_buff *skb, int optoff)
 {
-       if (skb->nh.raw[optoff+1] == 2) {
+       const unsigned char *nh = skb_network_header(skb);
+
+       if (nh[optoff + 1] == 2) {
                IP6CB(skb)->ra = optoff;
                return 1;
        }
        LIMIT_NETDEBUG(KERN_DEBUG "ipv6_hop_ra: wrong RA length %d\n",
-                      skb->nh.raw[optoff+1]);
+                      nh[optoff + 1]);
        kfree_skb(skb);
        return 0;
 }
@@ -436,36 +585,41 @@ static int ipv6_hop_ra(struct sk_buff *skb, int optoff)
 
 static int ipv6_hop_jumbo(struct sk_buff *skb, int optoff)
 {
+       const unsigned char *nh = skb_network_header(skb);
+       struct net *net = ipv6_skb_net(skb);
        u32 pkt_len;
 
-       if (skb->nh.raw[optoff+1] != 4 || (optoff&3) != 2) {
+       if (nh[optoff + 1] != 4 || (optoff & 3) != 2) {
                LIMIT_NETDEBUG(KERN_DEBUG "ipv6_hop_jumbo: wrong jumbo opt length/alignment %d\n",
-                              skb->nh.raw[optoff+1]);
-               IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
+                              nh[optoff+1]);
+               IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
+                                IPSTATS_MIB_INHDRERRORS);
                goto drop;
        }
 
-       pkt_len = ntohl(*(u32*)(skb->nh.raw+optoff+2));
+       pkt_len = ntohl(*(__be32 *)(nh + optoff + 2));
        if (pkt_len <= IPV6_MAXPLEN) {
-               IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
+               IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
+                                IPSTATS_MIB_INHDRERRORS);
                icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff+2);
                return 0;
        }
-       if (skb->nh.ipv6h->payload_len) {
-               IP6_INC_STATS_BH(IPSTATS_MIB_INHDRERRORS);
+       if (ipv6_hdr(skb)->payload_len) {
+               IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
+                                IPSTATS_MIB_INHDRERRORS);
                icmpv6_param_prob(skb, ICMPV6_HDR_FIELD, optoff);
                return 0;
        }
 
        if (pkt_len > skb->len - sizeof(struct ipv6hdr)) {
-               IP6_INC_STATS_BH(IPSTATS_MIB_INTRUNCATEDPKTS);
+               IP6_INC_STATS_BH(net, ipv6_skb_idev(skb),
+                                IPSTATS_MIB_INTRUNCATEDPKTS);
                goto drop;
        }
-       if (pkt_len + sizeof(struct ipv6hdr) < skb->len) {
-               __pskb_trim(skb, pkt_len + sizeof(struct ipv6hdr));
-               if (skb->ip_summed == CHECKSUM_HW)
-                       skb->ip_summed = CHECKSUM_NONE;
-       }
+
+       if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
+               goto drop;
+
        return 1;
 
 drop:
@@ -485,11 +639,30 @@ static struct tlvtype_proc tlvprochopopt_lst[] = {
        { -1, }
 };
 
-int ipv6_parse_hopopts(struct sk_buff *skb, int nhoff)
+int ipv6_parse_hopopts(struct sk_buff *skb)
 {
-       IP6CB(skb)->hop = sizeof(struct ipv6hdr);
-       if (ip6_parse_tlv(tlvprochopopt_lst, skb))
-               return sizeof(struct ipv6hdr);
+       struct inet6_skb_parm *opt = IP6CB(skb);
+
+       /*
+        * skb_network_header(skb) is equal to skb->data, and
+        * skb_network_header_len(skb) is always equal to
+        * sizeof(struct ipv6hdr) by definition of
+        * hop-by-hop options.
+        */
+       if (!pskb_may_pull(skb, sizeof(struct ipv6hdr) + 8) ||
+           !pskb_may_pull(skb, (sizeof(struct ipv6hdr) +
+                                ((skb_transport_header(skb)[1] + 1) << 3)))) {
+               kfree_skb(skb);
+               return -1;
+       }
+
+       opt->hop = sizeof(struct ipv6hdr);
+       if (ip6_parse_tlv(tlvprochopopt_lst, skb)) {
+               skb->transport_header += (skb_transport_header(skb)[1] + 1) << 3;
+               opt = IP6CB(skb);
+               opt->nhoff = sizeof(struct ipv6hdr);
+               return 1;
+       }
        return -1;
 }
 
@@ -511,7 +684,7 @@ static void ipv6_push_rthdr(struct sk_buff *skb, u8 *proto,
        int hops;
 
        ihdr = (struct rt0_hdr *) opt;
-       
+
        phdr = (struct rt0_hdr *) skb_push(skb, (ihdr->rt_hdr.hdrlen + 1) << 3);
        memcpy(phdr, ihdr, sizeof(struct rt0_hdr));
 
@@ -554,6 +727,8 @@ void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
                ipv6_push_exthdr(skb, proto, NEXTHDR_HOP, opt->hopopt);
 }
 
+EXPORT_SYMBOL(ipv6_push_nfrag_opts);
+
 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, u8 *proto)
 {
        if (opt->dst1opt)
@@ -581,6 +756,8 @@ ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
        return opt2;
 }
 
+EXPORT_SYMBOL_GPL(ipv6_dup_options);
+
 static int ipv6_renew_option(void *ohdr,
                             struct ipv6_opt_hdr __user *newopt, int newoptlen,
                             int inherit,
@@ -616,20 +793,24 @@ ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
        struct ipv6_txoptions *opt2;
        int err;
 
-       if (newtype != IPV6_HOPOPTS && opt->hopopt)
-               tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
-       if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
-               tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
-       if (newtype != IPV6_RTHDR && opt->srcrt)
-               tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
-       if (newtype != IPV6_DSTOPTS && opt->dst1opt)
-               tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
+       if (opt) {
+               if (newtype != IPV6_HOPOPTS && opt->hopopt)
+                       tot_len += CMSG_ALIGN(ipv6_optlen(opt->hopopt));
+               if (newtype != IPV6_RTHDRDSTOPTS && opt->dst0opt)
+                       tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst0opt));
+               if (newtype != IPV6_RTHDR && opt->srcrt)
+                       tot_len += CMSG_ALIGN(ipv6_optlen(opt->srcrt));
+               if (newtype != IPV6_DSTOPTS && opt->dst1opt)
+                       tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
+       }
+
        if (newopt && newoptlen)
                tot_len += CMSG_ALIGN(newoptlen);
 
        if (!tot_len)
                return NULL;
 
+       tot_len += sizeof(*opt2);
        opt2 = sock_kmalloc(sk, tot_len, GFP_ATOMIC);
        if (!opt2)
                return ERR_PTR(-ENOBUFS);
@@ -639,25 +820,25 @@ ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
        opt2->tot_len = tot_len;
        p = (char *)(opt2 + 1);
 
-       err = ipv6_renew_option(opt->hopopt, newopt, newoptlen,
+       err = ipv6_renew_option(opt ? opt->hopopt : NULL, newopt, newoptlen,
                                newtype != IPV6_HOPOPTS,
                                &opt2->hopopt, &p);
        if (err)
                goto out;
 
-       err = ipv6_renew_option(opt->dst0opt, newopt, newoptlen,
+       err = ipv6_renew_option(opt ? opt->dst0opt : NULL, newopt, newoptlen,
                                newtype != IPV6_RTHDRDSTOPTS,
                                &opt2->dst0opt, &p);
        if (err)
                goto out;
 
-       err = ipv6_renew_option(opt->srcrt, newopt, newoptlen,
+       err = ipv6_renew_option(opt ? opt->srcrt : NULL, newopt, newoptlen,
                                newtype != IPV6_RTHDR,
-                               (struct ipv6_opt_hdr **)opt2->srcrt, &p);
+                               (struct ipv6_opt_hdr **)&opt2->srcrt, &p);
        if (err)
                goto out;
 
-       err = ipv6_renew_option(opt->dst1opt, newopt, newoptlen,
+       err = ipv6_renew_option(opt ? opt->dst1opt : NULL, newopt, newoptlen,
                                newtype != IPV6_DSTOPTS,
                                &opt2->dst1opt, &p);
        if (err)
@@ -670,7 +851,26 @@ ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
 
        return opt2;
 out:
-       sock_kfree_s(sk, p, tot_len);
+       sock_kfree_s(sk, opt2, opt2->tot_len);
        return ERR_PTR(err);
 }
 
+struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
+                                         struct ipv6_txoptions *opt)
+{
+       /*
+        * ignore the dest before srcrt unless srcrt is being included.
+        * --yoshfuji
+        */
+       if (opt && opt->dst0opt && !opt->srcrt) {
+               if (opt_space != opt) {
+                       memcpy(opt_space, opt, sizeof(*opt_space));
+                       opt = opt_space;
+               }
+               opt->opt_nflen -= ipv6_optlen(opt->dst0opt);
+               opt->dst0opt = NULL;
+       }
+
+       return opt;
+}
+