Back to home page

OSCL-LXR
 
 

     

0001 // SPDX-License-Identifier: GPL-2.0-or-later
0002 /*
0003  *  IPv6 tunneling device
0004  *  Linux INET6 implementation
0005  *
0006  *  Authors:
0007  *  Ville Nuorvala      <vnuorval@tcs.hut.fi>
0008  *  Yasuyuki Kozakai    <kozakai@linux-ipv6.org>
0009  *
0010  *      Based on:
0011  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
0012  *
0013  *      RFC 2473
0014  */
0015 
0016 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
0017 
0018 #include <linux/module.h>
0019 #include <linux/capability.h>
0020 #include <linux/errno.h>
0021 #include <linux/types.h>
0022 #include <linux/sockios.h>
0023 #include <linux/icmp.h>
0024 #include <linux/if.h>
0025 #include <linux/in.h>
0026 #include <linux/ip.h>
0027 #include <linux/net.h>
0028 #include <linux/in6.h>
0029 #include <linux/netdevice.h>
0030 #include <linux/if_arp.h>
0031 #include <linux/icmpv6.h>
0032 #include <linux/init.h>
0033 #include <linux/route.h>
0034 #include <linux/rtnetlink.h>
0035 #include <linux/netfilter_ipv6.h>
0036 #include <linux/slab.h>
0037 #include <linux/hash.h>
0038 #include <linux/etherdevice.h>
0039 
0040 #include <linux/uaccess.h>
0041 #include <linux/atomic.h>
0042 
0043 #include <net/icmp.h>
0044 #include <net/ip.h>
0045 #include <net/ip_tunnels.h>
0046 #include <net/ipv6.h>
0047 #include <net/ip6_route.h>
0048 #include <net/addrconf.h>
0049 #include <net/ip6_tunnel.h>
0050 #include <net/xfrm.h>
0051 #include <net/dsfield.h>
0052 #include <net/inet_ecn.h>
0053 #include <net/net_namespace.h>
0054 #include <net/netns/generic.h>
0055 #include <net/dst_metadata.h>
0056 
0057 MODULE_AUTHOR("Ville Nuorvala");
0058 MODULE_DESCRIPTION("IPv6 tunneling device");
0059 MODULE_LICENSE("GPL");
0060 MODULE_ALIAS_RTNL_LINK("ip6tnl");
0061 MODULE_ALIAS_NETDEV("ip6tnl0");
0062 
0063 #define IP6_TUNNEL_HASH_SIZE_SHIFT  5
0064 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
0065 
0066 static bool log_ecn_error = true;
0067 module_param(log_ecn_error, bool, 0644);
0068 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
0069 
0070 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
0071 {
0072     u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
0073 
0074     return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
0075 }
0076 
0077 static int ip6_tnl_dev_init(struct net_device *dev);
0078 static void ip6_tnl_dev_setup(struct net_device *dev);
0079 static struct rtnl_link_ops ip6_link_ops __read_mostly;
0080 
0081 static unsigned int ip6_tnl_net_id __read_mostly;
0082 struct ip6_tnl_net {
0083     /* the IPv6 tunnel fallback device */
0084     struct net_device *fb_tnl_dev;
0085     /* lists for storing tunnels in use */
0086     struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
0087     struct ip6_tnl __rcu *tnls_wc[1];
0088     struct ip6_tnl __rcu **tnls[2];
0089     struct ip6_tnl __rcu *collect_md_tun;
0090 };
0091 
0092 static inline int ip6_tnl_mpls_supported(void)
0093 {
0094     return IS_ENABLED(CONFIG_MPLS);
0095 }
0096 
0097 #define for_each_ip6_tunnel_rcu(start) \
0098     for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
0099 
0100 /**
0101  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
0102  *   @net: network namespace
0103  *   @link: ifindex of underlying interface
0104  *   @remote: the address of the tunnel exit-point
0105  *   @local: the address of the tunnel entry-point
0106  *
0107  * Return:
0108  *   tunnel matching given end-points if found,
0109  *   else fallback tunnel if its device is up,
0110  *   else %NULL
0111  **/
0112 
0113 static struct ip6_tnl *
0114 ip6_tnl_lookup(struct net *net, int link,
0115            const struct in6_addr *remote, const struct in6_addr *local)
0116 {
0117     unsigned int hash = HASH(remote, local);
0118     struct ip6_tnl *t, *cand = NULL;
0119     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
0120     struct in6_addr any;
0121 
0122     for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
0123         if (!ipv6_addr_equal(local, &t->parms.laddr) ||
0124             !ipv6_addr_equal(remote, &t->parms.raddr) ||
0125             !(t->dev->flags & IFF_UP))
0126             continue;
0127 
0128         if (link == t->parms.link)
0129             return t;
0130         else
0131             cand = t;
0132     }
0133 
0134     memset(&any, 0, sizeof(any));
0135     hash = HASH(&any, local);
0136     for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
0137         if (!ipv6_addr_equal(local, &t->parms.laddr) ||
0138             !ipv6_addr_any(&t->parms.raddr) ||
0139             !(t->dev->flags & IFF_UP))
0140             continue;
0141 
0142         if (link == t->parms.link)
0143             return t;
0144         else if (!cand)
0145             cand = t;
0146     }
0147 
0148     hash = HASH(remote, &any);
0149     for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
0150         if (!ipv6_addr_equal(remote, &t->parms.raddr) ||
0151             !ipv6_addr_any(&t->parms.laddr) ||
0152             !(t->dev->flags & IFF_UP))
0153             continue;
0154 
0155         if (link == t->parms.link)
0156             return t;
0157         else if (!cand)
0158             cand = t;
0159     }
0160 
0161     if (cand)
0162         return cand;
0163 
0164     t = rcu_dereference(ip6n->collect_md_tun);
0165     if (t && t->dev->flags & IFF_UP)
0166         return t;
0167 
0168     t = rcu_dereference(ip6n->tnls_wc[0]);
0169     if (t && (t->dev->flags & IFF_UP))
0170         return t;
0171 
0172     return NULL;
0173 }
0174 
0175 /**
0176  * ip6_tnl_bucket - get head of list matching given tunnel parameters
0177  *   @ip6n: the private data for ip6_vti in the netns
0178  *   @p: parameters containing tunnel end-points
0179  *
0180  * Description:
0181  *   ip6_tnl_bucket() returns the head of the list matching the
0182  *   &struct in6_addr entries laddr and raddr in @p.
0183  *
0184  * Return: head of IPv6 tunnel list
0185  **/
0186 
0187 static struct ip6_tnl __rcu **
0188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
0189 {
0190     const struct in6_addr *remote = &p->raddr;
0191     const struct in6_addr *local = &p->laddr;
0192     unsigned int h = 0;
0193     int prio = 0;
0194 
0195     if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
0196         prio = 1;
0197         h = HASH(remote, local);
0198     }
0199     return &ip6n->tnls[prio][h];
0200 }
0201 
0202 /**
0203  * ip6_tnl_link - add tunnel to hash table
0204  *   @ip6n: the private data for ip6_vti in the netns
0205  *   @t: tunnel to be added
0206  **/
0207 
0208 static void
0209 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
0210 {
0211     struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
0212 
0213     if (t->parms.collect_md)
0214         rcu_assign_pointer(ip6n->collect_md_tun, t);
0215     rcu_assign_pointer(t->next , rtnl_dereference(*tp));
0216     rcu_assign_pointer(*tp, t);
0217 }
0218 
0219 /**
0220  * ip6_tnl_unlink - remove tunnel from hash table
0221  *   @ip6n: the private data for ip6_vti in the netns
0222  *   @t: tunnel to be removed
0223  **/
0224 
0225 static void
0226 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
0227 {
0228     struct ip6_tnl __rcu **tp;
0229     struct ip6_tnl *iter;
0230 
0231     if (t->parms.collect_md)
0232         rcu_assign_pointer(ip6n->collect_md_tun, NULL);
0233 
0234     for (tp = ip6_tnl_bucket(ip6n, &t->parms);
0235          (iter = rtnl_dereference(*tp)) != NULL;
0236          tp = &iter->next) {
0237         if (t == iter) {
0238             rcu_assign_pointer(*tp, t->next);
0239             break;
0240         }
0241     }
0242 }
0243 
0244 static void ip6_dev_free(struct net_device *dev)
0245 {
0246     struct ip6_tnl *t = netdev_priv(dev);
0247 
0248     gro_cells_destroy(&t->gro_cells);
0249     dst_cache_destroy(&t->dst_cache);
0250     free_percpu(dev->tstats);
0251 }
0252 
0253 static int ip6_tnl_create2(struct net_device *dev)
0254 {
0255     struct ip6_tnl *t = netdev_priv(dev);
0256     struct net *net = dev_net(dev);
0257     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
0258     int err;
0259 
0260     dev->rtnl_link_ops = &ip6_link_ops;
0261     err = register_netdevice(dev);
0262     if (err < 0)
0263         goto out;
0264 
0265     strcpy(t->parms.name, dev->name);
0266 
0267     ip6_tnl_link(ip6n, t);
0268     return 0;
0269 
0270 out:
0271     return err;
0272 }
0273 
0274 /**
0275  * ip6_tnl_create - create a new tunnel
0276  *   @net: network namespace
0277  *   @p: tunnel parameters
0278  *
0279  * Description:
0280  *   Create tunnel matching given parameters.
0281  *
0282  * Return:
0283  *   created tunnel or error pointer
0284  **/
0285 
0286 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
0287 {
0288     struct net_device *dev;
0289     struct ip6_tnl *t;
0290     char name[IFNAMSIZ];
0291     int err = -E2BIG;
0292 
0293     if (p->name[0]) {
0294         if (!dev_valid_name(p->name))
0295             goto failed;
0296         strlcpy(name, p->name, IFNAMSIZ);
0297     } else {
0298         sprintf(name, "ip6tnl%%d");
0299     }
0300     err = -ENOMEM;
0301     dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
0302                ip6_tnl_dev_setup);
0303     if (!dev)
0304         goto failed;
0305 
0306     dev_net_set(dev, net);
0307 
0308     t = netdev_priv(dev);
0309     t->parms = *p;
0310     t->net = dev_net(dev);
0311     err = ip6_tnl_create2(dev);
0312     if (err < 0)
0313         goto failed_free;
0314 
0315     return t;
0316 
0317 failed_free:
0318     free_netdev(dev);
0319 failed:
0320     return ERR_PTR(err);
0321 }
0322 
0323 /**
0324  * ip6_tnl_locate - find or create tunnel matching given parameters
0325  *   @net: network namespace
0326  *   @p: tunnel parameters
0327  *   @create: != 0 if allowed to create new tunnel if no match found
0328  *
0329  * Description:
0330  *   ip6_tnl_locate() first tries to locate an existing tunnel
0331  *   based on @parms. If this is unsuccessful, but @create is set a new
0332  *   tunnel device is created and registered for use.
0333  *
0334  * Return:
0335  *   matching tunnel or error pointer
0336  **/
0337 
0338 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
0339         struct __ip6_tnl_parm *p, int create)
0340 {
0341     const struct in6_addr *remote = &p->raddr;
0342     const struct in6_addr *local = &p->laddr;
0343     struct ip6_tnl __rcu **tp;
0344     struct ip6_tnl *t;
0345     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
0346 
0347     for (tp = ip6_tnl_bucket(ip6n, p);
0348          (t = rtnl_dereference(*tp)) != NULL;
0349          tp = &t->next) {
0350         if (ipv6_addr_equal(local, &t->parms.laddr) &&
0351             ipv6_addr_equal(remote, &t->parms.raddr) &&
0352             p->link == t->parms.link) {
0353             if (create)
0354                 return ERR_PTR(-EEXIST);
0355 
0356             return t;
0357         }
0358     }
0359     if (!create)
0360         return ERR_PTR(-ENODEV);
0361     return ip6_tnl_create(net, p);
0362 }
0363 
0364 /**
0365  * ip6_tnl_dev_uninit - tunnel device uninitializer
0366  *   @dev: the device to be destroyed
0367  *
0368  * Description:
0369  *   ip6_tnl_dev_uninit() removes tunnel from its list
0370  **/
0371 
0372 static void
0373 ip6_tnl_dev_uninit(struct net_device *dev)
0374 {
0375     struct ip6_tnl *t = netdev_priv(dev);
0376     struct net *net = t->net;
0377     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
0378 
0379     if (dev == ip6n->fb_tnl_dev)
0380         RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
0381     else
0382         ip6_tnl_unlink(ip6n, t);
0383     dst_cache_reset(&t->dst_cache);
0384     netdev_put(dev, &t->dev_tracker);
0385 }
0386 
0387 /**
0388  * ip6_tnl_parse_tlv_enc_lim - handle encapsulation limit option
0389  *   @skb: received socket buffer
0390  *   @raw: the ICMPv6 error message data
0391  *
0392  * Return:
0393  *   0 if none was found,
0394  *   else index to encapsulation limit
0395  **/
0396 
0397 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
0398 {
0399     const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
0400     unsigned int nhoff = raw - skb->data;
0401     unsigned int off = nhoff + sizeof(*ipv6h);
0402     u8 next, nexthdr = ipv6h->nexthdr;
0403 
0404     while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
0405         struct ipv6_opt_hdr *hdr;
0406         u16 optlen;
0407 
0408         if (!pskb_may_pull(skb, off + sizeof(*hdr)))
0409             break;
0410 
0411         hdr = (struct ipv6_opt_hdr *)(skb->data + off);
0412         if (nexthdr == NEXTHDR_FRAGMENT) {
0413             struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
0414             if (frag_hdr->frag_off)
0415                 break;
0416             optlen = 8;
0417         } else if (nexthdr == NEXTHDR_AUTH) {
0418             optlen = ipv6_authlen(hdr);
0419         } else {
0420             optlen = ipv6_optlen(hdr);
0421         }
0422         /* cache hdr->nexthdr, since pskb_may_pull() might
0423          * invalidate hdr
0424          */
0425         next = hdr->nexthdr;
0426         if (nexthdr == NEXTHDR_DEST) {
0427             u16 i = 2;
0428 
0429             /* Remember : hdr is no longer valid at this point. */
0430             if (!pskb_may_pull(skb, off + optlen))
0431                 break;
0432 
0433             while (1) {
0434                 struct ipv6_tlv_tnl_enc_lim *tel;
0435 
0436                 /* No more room for encapsulation limit */
0437                 if (i + sizeof(*tel) > optlen)
0438                     break;
0439 
0440                 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
0441                 /* return index of option if found and valid */
0442                 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
0443                     tel->length == 1)
0444                     return i + off - nhoff;
0445                 /* else jump to next option */
0446                 if (tel->type)
0447                     i += tel->length + 2;
0448                 else
0449                     i++;
0450             }
0451         }
0452         nexthdr = next;
0453         off += optlen;
0454     }
0455     return 0;
0456 }
0457 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
0458 
0459 /* ip6_tnl_err() should handle errors in the tunnel according to the
0460  * specifications in RFC 2473.
0461  */
0462 static int
0463 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
0464         u8 *type, u8 *code, int *msg, __u32 *info, int offset)
0465 {
0466     const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
0467     struct net *net = dev_net(skb->dev);
0468     u8 rel_type = ICMPV6_DEST_UNREACH;
0469     u8 rel_code = ICMPV6_ADDR_UNREACH;
0470     __u32 rel_info = 0;
0471     struct ip6_tnl *t;
0472     int err = -ENOENT;
0473     int rel_msg = 0;
0474     u8 tproto;
0475     __u16 len;
0476 
0477     /* If the packet doesn't contain the original IPv6 header we are
0478        in trouble since we might need the source address for further
0479        processing of the error. */
0480 
0481     rcu_read_lock();
0482     t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->daddr, &ipv6h->saddr);
0483     if (!t)
0484         goto out;
0485 
0486     tproto = READ_ONCE(t->parms.proto);
0487     if (tproto != ipproto && tproto != 0)
0488         goto out;
0489 
0490     err = 0;
0491 
0492     switch (*type) {
0493     case ICMPV6_DEST_UNREACH:
0494         net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
0495                     t->parms.name);
0496         rel_msg = 1;
0497         break;
0498     case ICMPV6_TIME_EXCEED:
0499         if ((*code) == ICMPV6_EXC_HOPLIMIT) {
0500             net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
0501                         t->parms.name);
0502             rel_msg = 1;
0503         }
0504         break;
0505     case ICMPV6_PARAMPROB: {
0506         struct ipv6_tlv_tnl_enc_lim *tel;
0507         __u32 teli;
0508 
0509         teli = 0;
0510         if ((*code) == ICMPV6_HDR_FIELD)
0511             teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
0512 
0513         if (teli && teli == *info - 2) {
0514             tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
0515             if (tel->encap_limit == 0) {
0516                 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
0517                             t->parms.name);
0518                 rel_msg = 1;
0519             }
0520         } else {
0521             net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
0522                         t->parms.name);
0523         }
0524         break;
0525     }
0526     case ICMPV6_PKT_TOOBIG: {
0527         __u32 mtu;
0528 
0529         ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
0530                 sock_net_uid(net, NULL));
0531         mtu = *info - offset;
0532         if (mtu < IPV6_MIN_MTU)
0533             mtu = IPV6_MIN_MTU;
0534         len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
0535         if (len > mtu) {
0536             rel_type = ICMPV6_PKT_TOOBIG;
0537             rel_code = 0;
0538             rel_info = mtu;
0539             rel_msg = 1;
0540         }
0541         break;
0542     }
0543     case NDISC_REDIRECT:
0544         ip6_redirect(skb, net, skb->dev->ifindex, 0,
0545                  sock_net_uid(net, NULL));
0546         break;
0547     }
0548 
0549     *type = rel_type;
0550     *code = rel_code;
0551     *info = rel_info;
0552     *msg = rel_msg;
0553 
0554 out:
0555     rcu_read_unlock();
0556     return err;
0557 }
0558 
0559 static int
0560 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
0561        u8 type, u8 code, int offset, __be32 info)
0562 {
0563     __u32 rel_info = ntohl(info);
0564     const struct iphdr *eiph;
0565     struct sk_buff *skb2;
0566     int err, rel_msg = 0;
0567     u8 rel_type = type;
0568     u8 rel_code = code;
0569     struct rtable *rt;
0570     struct flowi4 fl4;
0571 
0572     err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
0573               &rel_msg, &rel_info, offset);
0574     if (err < 0)
0575         return err;
0576 
0577     if (rel_msg == 0)
0578         return 0;
0579 
0580     switch (rel_type) {
0581     case ICMPV6_DEST_UNREACH:
0582         if (rel_code != ICMPV6_ADDR_UNREACH)
0583             return 0;
0584         rel_type = ICMP_DEST_UNREACH;
0585         rel_code = ICMP_HOST_UNREACH;
0586         break;
0587     case ICMPV6_PKT_TOOBIG:
0588         if (rel_code != 0)
0589             return 0;
0590         rel_type = ICMP_DEST_UNREACH;
0591         rel_code = ICMP_FRAG_NEEDED;
0592         break;
0593     default:
0594         return 0;
0595     }
0596 
0597     if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
0598         return 0;
0599 
0600     skb2 = skb_clone(skb, GFP_ATOMIC);
0601     if (!skb2)
0602         return 0;
0603 
0604     skb_dst_drop(skb2);
0605 
0606     skb_pull(skb2, offset);
0607     skb_reset_network_header(skb2);
0608     eiph = ip_hdr(skb2);
0609 
0610     /* Try to guess incoming interface */
0611     rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
0612                    0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
0613     if (IS_ERR(rt))
0614         goto out;
0615 
0616     skb2->dev = rt->dst.dev;
0617     ip_rt_put(rt);
0618 
0619     /* route "incoming" packet */
0620     if (rt->rt_flags & RTCF_LOCAL) {
0621         rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
0622                        eiph->daddr, eiph->saddr, 0, 0,
0623                        IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
0624         if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL6) {
0625             if (!IS_ERR(rt))
0626                 ip_rt_put(rt);
0627             goto out;
0628         }
0629         skb_dst_set(skb2, &rt->dst);
0630     } else {
0631         if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
0632                    skb2->dev) ||
0633             skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
0634             goto out;
0635     }
0636 
0637     /* change mtu on this route */
0638     if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
0639         if (rel_info > dst_mtu(skb_dst(skb2)))
0640             goto out;
0641 
0642         skb_dst_update_pmtu_no_confirm(skb2, rel_info);
0643     }
0644 
0645     icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
0646 
0647 out:
0648     kfree_skb(skb2);
0649     return 0;
0650 }
0651 
0652 static int
0653 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
0654        u8 type, u8 code, int offset, __be32 info)
0655 {
0656     __u32 rel_info = ntohl(info);
0657     int err, rel_msg = 0;
0658     u8 rel_type = type;
0659     u8 rel_code = code;
0660 
0661     err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
0662               &rel_msg, &rel_info, offset);
0663     if (err < 0)
0664         return err;
0665 
0666     if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
0667         struct rt6_info *rt;
0668         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
0669 
0670         if (!skb2)
0671             return 0;
0672 
0673         skb_dst_drop(skb2);
0674         skb_pull(skb2, offset);
0675         skb_reset_network_header(skb2);
0676 
0677         /* Try to guess incoming interface */
0678         rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
0679                 NULL, 0, skb2, 0);
0680 
0681         if (rt && rt->dst.dev)
0682             skb2->dev = rt->dst.dev;
0683 
0684         icmpv6_send(skb2, rel_type, rel_code, rel_info);
0685 
0686         ip6_rt_put(rt);
0687 
0688         kfree_skb(skb2);
0689     }
0690 
0691     return 0;
0692 }
0693 
0694 static int
0695 mplsip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
0696         u8 type, u8 code, int offset, __be32 info)
0697 {
0698     __u32 rel_info = ntohl(info);
0699     int err, rel_msg = 0;
0700     u8 rel_type = type;
0701     u8 rel_code = code;
0702 
0703     err = ip6_tnl_err(skb, IPPROTO_MPLS, opt, &rel_type, &rel_code,
0704               &rel_msg, &rel_info, offset);
0705     return err;
0706 }
0707 
0708 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
0709                        const struct ipv6hdr *ipv6h,
0710                        struct sk_buff *skb)
0711 {
0712     __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
0713 
0714     if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
0715         ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
0716 
0717     return IP6_ECN_decapsulate(ipv6h, skb);
0718 }
0719 
0720 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
0721                        const struct ipv6hdr *ipv6h,
0722                        struct sk_buff *skb)
0723 {
0724     if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
0725         ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
0726 
0727     return IP6_ECN_decapsulate(ipv6h, skb);
0728 }
0729 
0730 static inline int mplsip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
0731                            const struct ipv6hdr *ipv6h,
0732                            struct sk_buff *skb)
0733 {
0734     /* ECN is not supported in AF_MPLS */
0735     return 0;
0736 }
0737 
0738 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
0739                  const struct in6_addr *laddr,
0740                  const struct in6_addr *raddr)
0741 {
0742     struct __ip6_tnl_parm *p = &t->parms;
0743     int ltype = ipv6_addr_type(laddr);
0744     int rtype = ipv6_addr_type(raddr);
0745     __u32 flags = 0;
0746 
0747     if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
0748         flags = IP6_TNL_F_CAP_PER_PACKET;
0749     } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
0750            rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
0751            !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
0752            (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
0753         if (ltype&IPV6_ADDR_UNICAST)
0754             flags |= IP6_TNL_F_CAP_XMIT;
0755         if (rtype&IPV6_ADDR_UNICAST)
0756             flags |= IP6_TNL_F_CAP_RCV;
0757     }
0758     return flags;
0759 }
0760 EXPORT_SYMBOL(ip6_tnl_get_cap);
0761 
0762 /* called with rcu_read_lock() */
0763 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
0764                   const struct in6_addr *laddr,
0765                   const struct in6_addr *raddr)
0766 {
0767     struct __ip6_tnl_parm *p = &t->parms;
0768     int ret = 0;
0769     struct net *net = t->net;
0770 
0771     if ((p->flags & IP6_TNL_F_CAP_RCV) ||
0772         ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
0773          (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
0774         struct net_device *ldev = NULL;
0775 
0776         if (p->link)
0777             ldev = dev_get_by_index_rcu(net, p->link);
0778 
0779         if ((ipv6_addr_is_multicast(laddr) ||
0780              likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
0781                             0, IFA_F_TENTATIVE))) &&
0782             ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
0783              likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
0784                              0, IFA_F_TENTATIVE))))
0785             ret = 1;
0786     }
0787     return ret;
0788 }
0789 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
0790 
0791 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
0792              const struct tnl_ptk_info *tpi,
0793              struct metadata_dst *tun_dst,
0794              int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
0795                         const struct ipv6hdr *ipv6h,
0796                         struct sk_buff *skb),
0797              bool log_ecn_err)
0798 {
0799     const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
0800     int err;
0801 
0802     if ((!(tpi->flags & TUNNEL_CSUM) &&
0803          (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
0804         ((tpi->flags & TUNNEL_CSUM) &&
0805          !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
0806         tunnel->dev->stats.rx_crc_errors++;
0807         tunnel->dev->stats.rx_errors++;
0808         goto drop;
0809     }
0810 
0811     if (tunnel->parms.i_flags & TUNNEL_SEQ) {
0812         if (!(tpi->flags & TUNNEL_SEQ) ||
0813             (tunnel->i_seqno &&
0814              (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
0815             tunnel->dev->stats.rx_fifo_errors++;
0816             tunnel->dev->stats.rx_errors++;
0817             goto drop;
0818         }
0819         tunnel->i_seqno = ntohl(tpi->seq) + 1;
0820     }
0821 
0822     skb->protocol = tpi->proto;
0823 
0824     /* Warning: All skb pointers will be invalidated! */
0825     if (tunnel->dev->type == ARPHRD_ETHER) {
0826         if (!pskb_may_pull(skb, ETH_HLEN)) {
0827             tunnel->dev->stats.rx_length_errors++;
0828             tunnel->dev->stats.rx_errors++;
0829             goto drop;
0830         }
0831 
0832         ipv6h = ipv6_hdr(skb);
0833         skb->protocol = eth_type_trans(skb, tunnel->dev);
0834         skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
0835     } else {
0836         skb->dev = tunnel->dev;
0837         skb_reset_mac_header(skb);
0838     }
0839 
0840     skb_reset_network_header(skb);
0841     memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
0842 
0843     __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
0844 
0845     err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
0846     if (unlikely(err)) {
0847         if (log_ecn_err)
0848             net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
0849                          &ipv6h->saddr,
0850                          ipv6_get_dsfield(ipv6h));
0851         if (err > 1) {
0852             ++tunnel->dev->stats.rx_frame_errors;
0853             ++tunnel->dev->stats.rx_errors;
0854             goto drop;
0855         }
0856     }
0857 
0858     dev_sw_netstats_rx_add(tunnel->dev, skb->len);
0859 
0860     skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
0861 
0862     if (tun_dst)
0863         skb_dst_set(skb, (struct dst_entry *)tun_dst);
0864 
0865     gro_cells_receive(&tunnel->gro_cells, skb);
0866     return 0;
0867 
0868 drop:
0869     if (tun_dst)
0870         dst_release((struct dst_entry *)tun_dst);
0871     kfree_skb(skb);
0872     return 0;
0873 }
0874 
0875 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
0876         const struct tnl_ptk_info *tpi,
0877         struct metadata_dst *tun_dst,
0878         bool log_ecn_err)
0879 {
0880     int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
0881                     const struct ipv6hdr *ipv6h,
0882                     struct sk_buff *skb);
0883 
0884     dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
0885     if (tpi->proto == htons(ETH_P_IP))
0886         dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
0887 
0888     return __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
0889                  log_ecn_err);
0890 }
0891 EXPORT_SYMBOL(ip6_tnl_rcv);
0892 
0893 static const struct tnl_ptk_info tpi_v6 = {
0894     /* no tunnel info required for ipxip6. */
0895     .proto = htons(ETH_P_IPV6),
0896 };
0897 
0898 static const struct tnl_ptk_info tpi_v4 = {
0899     /* no tunnel info required for ipxip6. */
0900     .proto = htons(ETH_P_IP),
0901 };
0902 
0903 static const struct tnl_ptk_info tpi_mpls = {
0904     /* no tunnel info required for mplsip6. */
0905     .proto = htons(ETH_P_MPLS_UC),
0906 };
0907 
0908 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
0909               const struct tnl_ptk_info *tpi,
0910               int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
0911                           const struct ipv6hdr *ipv6h,
0912                           struct sk_buff *skb))
0913 {
0914     struct ip6_tnl *t;
0915     const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
0916     struct metadata_dst *tun_dst = NULL;
0917     int ret = -1;
0918 
0919     rcu_read_lock();
0920     t = ip6_tnl_lookup(dev_net(skb->dev), skb->dev->ifindex, &ipv6h->saddr, &ipv6h->daddr);
0921 
0922     if (t) {
0923         u8 tproto = READ_ONCE(t->parms.proto);
0924 
0925         if (tproto != ipproto && tproto != 0)
0926             goto drop;
0927         if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
0928             goto drop;
0929         ipv6h = ipv6_hdr(skb);
0930         if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
0931             goto drop;
0932         if (iptunnel_pull_header(skb, 0, tpi->proto, false))
0933             goto drop;
0934         if (t->parms.collect_md) {
0935             tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
0936             if (!tun_dst)
0937                 goto drop;
0938         }
0939         ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
0940                     log_ecn_error);
0941     }
0942 
0943     rcu_read_unlock();
0944 
0945     return ret;
0946 
0947 drop:
0948     rcu_read_unlock();
0949     kfree_skb(skb);
0950     return 0;
0951 }
0952 
0953 static int ip4ip6_rcv(struct sk_buff *skb)
0954 {
0955     return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
0956               ip4ip6_dscp_ecn_decapsulate);
0957 }
0958 
0959 static int ip6ip6_rcv(struct sk_buff *skb)
0960 {
0961     return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
0962               ip6ip6_dscp_ecn_decapsulate);
0963 }
0964 
0965 static int mplsip6_rcv(struct sk_buff *skb)
0966 {
0967     return ipxip6_rcv(skb, IPPROTO_MPLS, &tpi_mpls,
0968               mplsip6_dscp_ecn_decapsulate);
0969 }
0970 
0971 struct ipv6_tel_txoption {
0972     struct ipv6_txoptions ops;
0973     __u8 dst_opt[8];
0974 };
0975 
0976 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
0977 {
0978     memset(opt, 0, sizeof(struct ipv6_tel_txoption));
0979 
0980     opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
0981     opt->dst_opt[3] = 1;
0982     opt->dst_opt[4] = encap_limit;
0983     opt->dst_opt[5] = IPV6_TLV_PADN;
0984     opt->dst_opt[6] = 1;
0985 
0986     opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
0987     opt->ops.opt_nflen = 8;
0988 }
0989 
0990 /**
0991  * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
0992  *   @t: the outgoing tunnel device
0993  *   @hdr: IPv6 header from the incoming packet
0994  *
0995  * Description:
0996  *   Avoid trivial tunneling loop by checking that tunnel exit-point
0997  *   doesn't match source of incoming packet.
0998  *
0999  * Return:
1000  *   1 if conflict,
1001  *   0 else
1002  **/
1003 
1004 static inline bool
1005 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
1006 {
1007     return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
1008 }
1009 
1010 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
1011              const struct in6_addr *laddr,
1012              const struct in6_addr *raddr)
1013 {
1014     struct __ip6_tnl_parm *p = &t->parms;
1015     int ret = 0;
1016     struct net *net = t->net;
1017 
1018     if (t->parms.collect_md)
1019         return 1;
1020 
1021     if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1022         ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1023          (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1024         struct net_device *ldev = NULL;
1025 
1026         rcu_read_lock();
1027         if (p->link)
1028             ldev = dev_get_by_index_rcu(net, p->link);
1029 
1030         if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
1031                               0, IFA_F_TENTATIVE)))
1032             pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1033                         p->name);
1034         else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1035              !ipv6_addr_is_multicast(raddr) &&
1036              unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1037                               true, 0, IFA_F_TENTATIVE)))
1038             pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1039                         p->name);
1040         else
1041             ret = 1;
1042         rcu_read_unlock();
1043     }
1044     return ret;
1045 }
1046 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1047 
1048 /**
1049  * ip6_tnl_xmit - encapsulate packet and send
1050  *   @skb: the outgoing socket buffer
1051  *   @dev: the outgoing tunnel device
1052  *   @dsfield: dscp code for outer header
1053  *   @fl6: flow of tunneled packet
1054  *   @encap_limit: encapsulation limit
1055  *   @pmtu: Path MTU is stored if packet is too big
1056  *   @proto: next header value
1057  *
1058  * Description:
1059  *   Build new header and do some sanity checks on the packet before sending
1060  *   it.
1061  *
1062  * Return:
1063  *   0 on success
1064  *   -1 fail
1065  *   %-EMSGSIZE message too big. return mtu in this case.
1066  **/
1067 
1068 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1069          struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1070          __u8 proto)
1071 {
1072     struct ip6_tnl *t = netdev_priv(dev);
1073     struct net *net = t->net;
1074     struct net_device_stats *stats = &t->dev->stats;
1075     struct ipv6hdr *ipv6h;
1076     struct ipv6_tel_txoption opt;
1077     struct dst_entry *dst = NULL, *ndst = NULL;
1078     struct net_device *tdev;
1079     int mtu;
1080     unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1081     unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1082     unsigned int max_headroom = psh_hlen;
1083     __be16 payload_protocol;
1084     bool use_cache = false;
1085     u8 hop_limit;
1086     int err = -1;
1087 
1088     payload_protocol = skb_protocol(skb, true);
1089 
1090     if (t->parms.collect_md) {
1091         hop_limit = skb_tunnel_info(skb)->key.ttl;
1092         goto route_lookup;
1093     } else {
1094         hop_limit = t->parms.hop_limit;
1095     }
1096 
1097     /* NBMA tunnel */
1098     if (ipv6_addr_any(&t->parms.raddr)) {
1099         if (payload_protocol == htons(ETH_P_IPV6)) {
1100             struct in6_addr *addr6;
1101             struct neighbour *neigh;
1102             int addr_type;
1103 
1104             if (!skb_dst(skb))
1105                 goto tx_err_link_failure;
1106 
1107             neigh = dst_neigh_lookup(skb_dst(skb),
1108                          &ipv6_hdr(skb)->daddr);
1109             if (!neigh)
1110                 goto tx_err_link_failure;
1111 
1112             addr6 = (struct in6_addr *)&neigh->primary_key;
1113             addr_type = ipv6_addr_type(addr6);
1114 
1115             if (addr_type == IPV6_ADDR_ANY)
1116                 addr6 = &ipv6_hdr(skb)->daddr;
1117 
1118             memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1119             neigh_release(neigh);
1120         } else if (payload_protocol == htons(ETH_P_IP)) {
1121             const struct rtable *rt = skb_rtable(skb);
1122 
1123             if (!rt)
1124                 goto tx_err_link_failure;
1125 
1126             if (rt->rt_gw_family == AF_INET6)
1127                 memcpy(&fl6->daddr, &rt->rt_gw6, sizeof(fl6->daddr));
1128         }
1129     } else if (t->parms.proto != 0 && !(t->parms.flags &
1130                         (IP6_TNL_F_USE_ORIG_TCLASS |
1131                          IP6_TNL_F_USE_ORIG_FWMARK))) {
1132         /* enable the cache only if neither the outer protocol nor the
1133          * routing decision depends on the current inner header value
1134          */
1135         use_cache = true;
1136     }
1137 
1138     if (use_cache)
1139         dst = dst_cache_get(&t->dst_cache);
1140 
1141     if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1142         goto tx_err_link_failure;
1143 
1144     if (!dst) {
1145 route_lookup:
1146         /* add dsfield to flowlabel for route lookup */
1147         fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1148 
1149         dst = ip6_route_output(net, NULL, fl6);
1150 
1151         if (dst->error)
1152             goto tx_err_link_failure;
1153         dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1154         if (IS_ERR(dst)) {
1155             err = PTR_ERR(dst);
1156             dst = NULL;
1157             goto tx_err_link_failure;
1158         }
1159         if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1160             ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1161                        &fl6->daddr, 0, &fl6->saddr))
1162             goto tx_err_link_failure;
1163         ndst = dst;
1164     }
1165 
1166     tdev = dst->dev;
1167 
1168     if (tdev == dev) {
1169         stats->collisions++;
1170         net_warn_ratelimited("%s: Local routing loop detected!\n",
1171                      t->parms.name);
1172         goto tx_err_dst_release;
1173     }
1174     mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1175     if (encap_limit >= 0) {
1176         max_headroom += 8;
1177         mtu -= 8;
1178     }
1179     mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1180                IPV6_MIN_MTU : IPV4_MIN_MTU);
1181 
1182     skb_dst_update_pmtu_no_confirm(skb, mtu);
1183     if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1184         *pmtu = mtu;
1185         err = -EMSGSIZE;
1186         goto tx_err_dst_release;
1187     }
1188 
1189     if (t->err_count > 0) {
1190         if (time_before(jiffies,
1191                 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1192             t->err_count--;
1193 
1194             dst_link_failure(skb);
1195         } else {
1196             t->err_count = 0;
1197         }
1198     }
1199 
1200     skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1201 
1202     /*
1203      * Okay, now see if we can stuff it in the buffer as-is.
1204      */
1205     max_headroom += LL_RESERVED_SPACE(tdev);
1206 
1207     if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1208         (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1209         struct sk_buff *new_skb;
1210 
1211         new_skb = skb_realloc_headroom(skb, max_headroom);
1212         if (!new_skb)
1213             goto tx_err_dst_release;
1214 
1215         if (skb->sk)
1216             skb_set_owner_w(new_skb, skb->sk);
1217         consume_skb(skb);
1218         skb = new_skb;
1219     }
1220 
1221     if (t->parms.collect_md) {
1222         if (t->encap.type != TUNNEL_ENCAP_NONE)
1223             goto tx_err_dst_release;
1224     } else {
1225         if (use_cache && ndst)
1226             dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1227     }
1228     skb_dst_set(skb, dst);
1229 
1230     if (hop_limit == 0) {
1231         if (payload_protocol == htons(ETH_P_IP))
1232             hop_limit = ip_hdr(skb)->ttl;
1233         else if (payload_protocol == htons(ETH_P_IPV6))
1234             hop_limit = ipv6_hdr(skb)->hop_limit;
1235         else
1236             hop_limit = ip6_dst_hoplimit(dst);
1237     }
1238 
1239     /* Calculate max headroom for all the headers and adjust
1240      * needed_headroom if necessary.
1241      */
1242     max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1243             + dst->header_len + t->hlen;
1244     if (max_headroom > dev->needed_headroom)
1245         dev->needed_headroom = max_headroom;
1246 
1247     err = ip6_tnl_encap(skb, t, &proto, fl6);
1248     if (err)
1249         return err;
1250 
1251     if (encap_limit >= 0) {
1252         init_tel_txopt(&opt, encap_limit);
1253         ipv6_push_frag_opts(skb, &opt.ops, &proto);
1254     }
1255 
1256     skb_push(skb, sizeof(struct ipv6hdr));
1257     skb_reset_network_header(skb);
1258     ipv6h = ipv6_hdr(skb);
1259     ip6_flow_hdr(ipv6h, dsfield,
1260              ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1261     ipv6h->hop_limit = hop_limit;
1262     ipv6h->nexthdr = proto;
1263     ipv6h->saddr = fl6->saddr;
1264     ipv6h->daddr = fl6->daddr;
1265     ip6tunnel_xmit(NULL, skb, dev);
1266     return 0;
1267 tx_err_link_failure:
1268     stats->tx_carrier_errors++;
1269     dst_link_failure(skb);
1270 tx_err_dst_release:
1271     dst_release(dst);
1272     return err;
1273 }
1274 EXPORT_SYMBOL(ip6_tnl_xmit);
1275 
1276 static inline int
1277 ipxip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev,
1278         u8 protocol)
1279 {
1280     struct ip6_tnl *t = netdev_priv(dev);
1281     struct ipv6hdr *ipv6h;
1282     const struct iphdr  *iph;
1283     int encap_limit = -1;
1284     __u16 offset;
1285     struct flowi6 fl6;
1286     __u8 dsfield, orig_dsfield;
1287     __u32 mtu;
1288     u8 tproto;
1289     int err;
1290 
1291     tproto = READ_ONCE(t->parms.proto);
1292     if (tproto != protocol && tproto != 0)
1293         return -1;
1294 
1295     if (t->parms.collect_md) {
1296         struct ip_tunnel_info *tun_info;
1297         const struct ip_tunnel_key *key;
1298 
1299         tun_info = skb_tunnel_info(skb);
1300         if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1301                  ip_tunnel_info_af(tun_info) != AF_INET6))
1302             return -1;
1303         key = &tun_info->key;
1304         memset(&fl6, 0, sizeof(fl6));
1305         fl6.flowi6_proto = protocol;
1306         fl6.saddr = key->u.ipv6.src;
1307         fl6.daddr = key->u.ipv6.dst;
1308         fl6.flowlabel = key->label;
1309         dsfield =  key->tos;
1310         switch (protocol) {
1311         case IPPROTO_IPIP:
1312             iph = ip_hdr(skb);
1313             orig_dsfield = ipv4_get_dsfield(iph);
1314             break;
1315         case IPPROTO_IPV6:
1316             ipv6h = ipv6_hdr(skb);
1317             orig_dsfield = ipv6_get_dsfield(ipv6h);
1318             break;
1319         default:
1320             orig_dsfield = dsfield;
1321             break;
1322         }
1323     } else {
1324         if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1325             encap_limit = t->parms.encap_limit;
1326         if (protocol == IPPROTO_IPV6) {
1327             offset = ip6_tnl_parse_tlv_enc_lim(skb,
1328                         skb_network_header(skb));
1329             /* ip6_tnl_parse_tlv_enc_lim() might have
1330              * reallocated skb->head
1331              */
1332             if (offset > 0) {
1333                 struct ipv6_tlv_tnl_enc_lim *tel;
1334 
1335                 tel = (void *)&skb_network_header(skb)[offset];
1336                 if (tel->encap_limit == 0) {
1337                     icmpv6_ndo_send(skb, ICMPV6_PARAMPROB,
1338                             ICMPV6_HDR_FIELD, offset + 2);
1339                     return -1;
1340                 }
1341                 encap_limit = tel->encap_limit - 1;
1342             }
1343         }
1344 
1345         memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1346         fl6.flowi6_proto = protocol;
1347 
1348         if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1349             fl6.flowi6_mark = skb->mark;
1350         else
1351             fl6.flowi6_mark = t->parms.fwmark;
1352         switch (protocol) {
1353         case IPPROTO_IPIP:
1354             iph = ip_hdr(skb);
1355             orig_dsfield = ipv4_get_dsfield(iph);
1356             if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1357                 dsfield = orig_dsfield;
1358             else
1359                 dsfield = ip6_tclass(t->parms.flowinfo);
1360             break;
1361         case IPPROTO_IPV6:
1362             ipv6h = ipv6_hdr(skb);
1363             orig_dsfield = ipv6_get_dsfield(ipv6h);
1364             if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1365                 dsfield = orig_dsfield;
1366             else
1367                 dsfield = ip6_tclass(t->parms.flowinfo);
1368             if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1369                 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1370             break;
1371         default:
1372             orig_dsfield = dsfield = ip6_tclass(t->parms.flowinfo);
1373             break;
1374         }
1375     }
1376 
1377     fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1378     dsfield = INET_ECN_encapsulate(dsfield, orig_dsfield);
1379 
1380     if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1381         return -1;
1382 
1383     skb_set_inner_ipproto(skb, protocol);
1384 
1385     err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1386                protocol);
1387     if (err != 0) {
1388         /* XXX: send ICMP error even if DF is not set. */
1389         if (err == -EMSGSIZE)
1390             switch (protocol) {
1391             case IPPROTO_IPIP:
1392                 icmp_ndo_send(skb, ICMP_DEST_UNREACH,
1393                           ICMP_FRAG_NEEDED, htonl(mtu));
1394                 break;
1395             case IPPROTO_IPV6:
1396                 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1397                 break;
1398             default:
1399                 break;
1400             }
1401         return -1;
1402     }
1403 
1404     return 0;
1405 }
1406 
1407 static netdev_tx_t
1408 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1409 {
1410     struct ip6_tnl *t = netdev_priv(dev);
1411     struct net_device_stats *stats = &t->dev->stats;
1412     u8 ipproto;
1413     int ret;
1414 
1415     if (!pskb_inet_may_pull(skb))
1416         goto tx_err;
1417 
1418     switch (skb->protocol) {
1419     case htons(ETH_P_IP):
1420         ipproto = IPPROTO_IPIP;
1421         break;
1422     case htons(ETH_P_IPV6):
1423         if (ip6_tnl_addr_conflict(t, ipv6_hdr(skb)))
1424             goto tx_err;
1425         ipproto = IPPROTO_IPV6;
1426         break;
1427     case htons(ETH_P_MPLS_UC):
1428         ipproto = IPPROTO_MPLS;
1429         break;
1430     default:
1431         goto tx_err;
1432     }
1433 
1434     ret = ipxip6_tnl_xmit(skb, dev, ipproto);
1435     if (ret < 0)
1436         goto tx_err;
1437 
1438     return NETDEV_TX_OK;
1439 
1440 tx_err:
1441     stats->tx_errors++;
1442     stats->tx_dropped++;
1443     kfree_skb(skb);
1444     return NETDEV_TX_OK;
1445 }
1446 
1447 static void ip6_tnl_link_config(struct ip6_tnl *t)
1448 {
1449     struct net_device *dev = t->dev;
1450     struct net_device *tdev = NULL;
1451     struct __ip6_tnl_parm *p = &t->parms;
1452     struct flowi6 *fl6 = &t->fl.u.ip6;
1453     unsigned int mtu;
1454     int t_hlen;
1455 
1456     __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
1457     memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1458 
1459     /* Set up flowi template */
1460     fl6->saddr = p->laddr;
1461     fl6->daddr = p->raddr;
1462     fl6->flowi6_oif = p->link;
1463     fl6->flowlabel = 0;
1464 
1465     if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1466         fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1467     if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1468         fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1469 
1470     p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1471     p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1472 
1473     if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1474         dev->flags |= IFF_POINTOPOINT;
1475     else
1476         dev->flags &= ~IFF_POINTOPOINT;
1477 
1478     t->tun_hlen = 0;
1479     t->hlen = t->encap_hlen + t->tun_hlen;
1480     t_hlen = t->hlen + sizeof(struct ipv6hdr);
1481 
1482     if (p->flags & IP6_TNL_F_CAP_XMIT) {
1483         int strict = (ipv6_addr_type(&p->raddr) &
1484                   (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1485 
1486         struct rt6_info *rt = rt6_lookup(t->net,
1487                          &p->raddr, &p->laddr,
1488                          p->link, NULL, strict);
1489         if (rt) {
1490             tdev = rt->dst.dev;
1491             ip6_rt_put(rt);
1492         }
1493 
1494         if (!tdev && p->link)
1495             tdev = __dev_get_by_index(t->net, p->link);
1496 
1497         if (tdev) {
1498             dev->hard_header_len = tdev->hard_header_len + t_hlen;
1499             mtu = min_t(unsigned int, tdev->mtu, IP6_MAX_MTU);
1500 
1501             dev->mtu = mtu - t_hlen;
1502             if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1503                 dev->mtu -= 8;
1504 
1505             if (dev->mtu < IPV6_MIN_MTU)
1506                 dev->mtu = IPV6_MIN_MTU;
1507         }
1508     }
1509 }
1510 
1511 /**
1512  * ip6_tnl_change - update the tunnel parameters
1513  *   @t: tunnel to be changed
1514  *   @p: tunnel configuration parameters
1515  *
1516  * Description:
1517  *   ip6_tnl_change() updates the tunnel parameters
1518  **/
1519 
1520 static void
1521 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1522 {
1523     t->parms.laddr = p->laddr;
1524     t->parms.raddr = p->raddr;
1525     t->parms.flags = p->flags;
1526     t->parms.hop_limit = p->hop_limit;
1527     t->parms.encap_limit = p->encap_limit;
1528     t->parms.flowinfo = p->flowinfo;
1529     t->parms.link = p->link;
1530     t->parms.proto = p->proto;
1531     t->parms.fwmark = p->fwmark;
1532     dst_cache_reset(&t->dst_cache);
1533     ip6_tnl_link_config(t);
1534 }
1535 
1536 static void ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1537 {
1538     struct net *net = t->net;
1539     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1540 
1541     ip6_tnl_unlink(ip6n, t);
1542     synchronize_net();
1543     ip6_tnl_change(t, p);
1544     ip6_tnl_link(ip6n, t);
1545     netdev_state_change(t->dev);
1546 }
1547 
1548 static void ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1549 {
1550     /* for default tnl0 device allow to change only the proto */
1551     t->parms.proto = p->proto;
1552     netdev_state_change(t->dev);
1553 }
1554 
1555 static void
1556 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1557 {
1558     p->laddr = u->laddr;
1559     p->raddr = u->raddr;
1560     p->flags = u->flags;
1561     p->hop_limit = u->hop_limit;
1562     p->encap_limit = u->encap_limit;
1563     p->flowinfo = u->flowinfo;
1564     p->link = u->link;
1565     p->proto = u->proto;
1566     memcpy(p->name, u->name, sizeof(u->name));
1567 }
1568 
1569 static void
1570 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1571 {
1572     u->laddr = p->laddr;
1573     u->raddr = p->raddr;
1574     u->flags = p->flags;
1575     u->hop_limit = p->hop_limit;
1576     u->encap_limit = p->encap_limit;
1577     u->flowinfo = p->flowinfo;
1578     u->link = p->link;
1579     u->proto = p->proto;
1580     memcpy(u->name, p->name, sizeof(u->name));
1581 }
1582 
1583 /**
1584  * ip6_tnl_siocdevprivate - configure ipv6 tunnels from userspace
1585  *   @dev: virtual device associated with tunnel
1586  *   @ifr: unused
1587  *   @data: parameters passed from userspace
1588  *   @cmd: command to be performed
1589  *
1590  * Description:
1591  *   ip6_tnl_ioctl() is used for managing IPv6 tunnels
1592  *   from userspace.
1593  *
1594  *   The possible commands are the following:
1595  *     %SIOCGETTUNNEL: get tunnel parameters for device
1596  *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1597  *     %SIOCCHGTUNNEL: change tunnel parameters to those given
1598  *     %SIOCDELTUNNEL: delete tunnel
1599  *
1600  *   The fallback device "ip6tnl0", created during module
1601  *   initialization, can be used for creating other tunnel devices.
1602  *
1603  * Return:
1604  *   0 on success,
1605  *   %-EFAULT if unable to copy data to or from userspace,
1606  *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
1607  *   %-EINVAL if passed tunnel parameters are invalid,
1608  *   %-EEXIST if changing a tunnel's parameters would cause a conflict
1609  *   %-ENODEV if attempting to change or delete a nonexisting device
1610  **/
1611 
1612 static int
1613 ip6_tnl_siocdevprivate(struct net_device *dev, struct ifreq *ifr,
1614                void __user *data, int cmd)
1615 {
1616     int err = 0;
1617     struct ip6_tnl_parm p;
1618     struct __ip6_tnl_parm p1;
1619     struct ip6_tnl *t = netdev_priv(dev);
1620     struct net *net = t->net;
1621     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1622 
1623     memset(&p1, 0, sizeof(p1));
1624 
1625     switch (cmd) {
1626     case SIOCGETTUNNEL:
1627         if (dev == ip6n->fb_tnl_dev) {
1628             if (copy_from_user(&p, data, sizeof(p))) {
1629                 err = -EFAULT;
1630                 break;
1631             }
1632             ip6_tnl_parm_from_user(&p1, &p);
1633             t = ip6_tnl_locate(net, &p1, 0);
1634             if (IS_ERR(t))
1635                 t = netdev_priv(dev);
1636         } else {
1637             memset(&p, 0, sizeof(p));
1638         }
1639         ip6_tnl_parm_to_user(&p, &t->parms);
1640         if (copy_to_user(data, &p, sizeof(p)))
1641             err = -EFAULT;
1642         break;
1643     case SIOCADDTUNNEL:
1644     case SIOCCHGTUNNEL:
1645         err = -EPERM;
1646         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1647             break;
1648         err = -EFAULT;
1649         if (copy_from_user(&p, data, sizeof(p)))
1650             break;
1651         err = -EINVAL;
1652         if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1653             p.proto != 0)
1654             break;
1655         ip6_tnl_parm_from_user(&p1, &p);
1656         t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1657         if (cmd == SIOCCHGTUNNEL) {
1658             if (!IS_ERR(t)) {
1659                 if (t->dev != dev) {
1660                     err = -EEXIST;
1661                     break;
1662                 }
1663             } else
1664                 t = netdev_priv(dev);
1665             if (dev == ip6n->fb_tnl_dev)
1666                 ip6_tnl0_update(t, &p1);
1667             else
1668                 ip6_tnl_update(t, &p1);
1669         }
1670         if (!IS_ERR(t)) {
1671             err = 0;
1672             ip6_tnl_parm_to_user(&p, &t->parms);
1673             if (copy_to_user(data, &p, sizeof(p)))
1674                 err = -EFAULT;
1675 
1676         } else {
1677             err = PTR_ERR(t);
1678         }
1679         break;
1680     case SIOCDELTUNNEL:
1681         err = -EPERM;
1682         if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1683             break;
1684 
1685         if (dev == ip6n->fb_tnl_dev) {
1686             err = -EFAULT;
1687             if (copy_from_user(&p, data, sizeof(p)))
1688                 break;
1689             err = -ENOENT;
1690             ip6_tnl_parm_from_user(&p1, &p);
1691             t = ip6_tnl_locate(net, &p1, 0);
1692             if (IS_ERR(t))
1693                 break;
1694             err = -EPERM;
1695             if (t->dev == ip6n->fb_tnl_dev)
1696                 break;
1697             dev = t->dev;
1698         }
1699         err = 0;
1700         unregister_netdevice(dev);
1701         break;
1702     default:
1703         err = -EINVAL;
1704     }
1705     return err;
1706 }
1707 
1708 /**
1709  * ip6_tnl_change_mtu - change mtu manually for tunnel device
1710  *   @dev: virtual device associated with tunnel
1711  *   @new_mtu: the new mtu
1712  *
1713  * Return:
1714  *   0 on success,
1715  *   %-EINVAL if mtu too small
1716  **/
1717 
1718 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1719 {
1720     struct ip6_tnl *tnl = netdev_priv(dev);
1721 
1722     if (tnl->parms.proto == IPPROTO_IPV6) {
1723         if (new_mtu < IPV6_MIN_MTU)
1724             return -EINVAL;
1725     } else {
1726         if (new_mtu < ETH_MIN_MTU)
1727             return -EINVAL;
1728     }
1729     if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1730         if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1731             return -EINVAL;
1732     } else {
1733         if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1734             return -EINVAL;
1735     }
1736     dev->mtu = new_mtu;
1737     return 0;
1738 }
1739 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1740 
1741 int ip6_tnl_get_iflink(const struct net_device *dev)
1742 {
1743     struct ip6_tnl *t = netdev_priv(dev);
1744 
1745     return t->parms.link;
1746 }
1747 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1748 
1749 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1750               unsigned int num)
1751 {
1752     if (num >= MAX_IPTUN_ENCAP_OPS)
1753         return -ERANGE;
1754 
1755     return !cmpxchg((const struct ip6_tnl_encap_ops **)
1756             &ip6tun_encaps[num],
1757             NULL, ops) ? 0 : -1;
1758 }
1759 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1760 
1761 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1762               unsigned int num)
1763 {
1764     int ret;
1765 
1766     if (num >= MAX_IPTUN_ENCAP_OPS)
1767         return -ERANGE;
1768 
1769     ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1770                &ip6tun_encaps[num],
1771                ops, NULL) == ops) ? 0 : -1;
1772 
1773     synchronize_net();
1774 
1775     return ret;
1776 }
1777 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1778 
1779 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1780             struct ip_tunnel_encap *ipencap)
1781 {
1782     int hlen;
1783 
1784     memset(&t->encap, 0, sizeof(t->encap));
1785 
1786     hlen = ip6_encap_hlen(ipencap);
1787     if (hlen < 0)
1788         return hlen;
1789 
1790     t->encap.type = ipencap->type;
1791     t->encap.sport = ipencap->sport;
1792     t->encap.dport = ipencap->dport;
1793     t->encap.flags = ipencap->flags;
1794 
1795     t->encap_hlen = hlen;
1796     t->hlen = t->encap_hlen + t->tun_hlen;
1797 
1798     return 0;
1799 }
1800 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1801 
1802 static const struct net_device_ops ip6_tnl_netdev_ops = {
1803     .ndo_init   = ip6_tnl_dev_init,
1804     .ndo_uninit = ip6_tnl_dev_uninit,
1805     .ndo_start_xmit = ip6_tnl_start_xmit,
1806     .ndo_siocdevprivate = ip6_tnl_siocdevprivate,
1807     .ndo_change_mtu = ip6_tnl_change_mtu,
1808     .ndo_get_stats64 = dev_get_tstats64,
1809     .ndo_get_iflink = ip6_tnl_get_iflink,
1810 };
1811 
1812 #define IPXIPX_FEATURES (NETIF_F_SG |       \
1813              NETIF_F_FRAGLIST | \
1814              NETIF_F_HIGHDMA |  \
1815              NETIF_F_GSO_SOFTWARE | \
1816              NETIF_F_HW_CSUM)
1817 
1818 /**
1819  * ip6_tnl_dev_setup - setup virtual tunnel device
1820  *   @dev: virtual device associated with tunnel
1821  *
1822  * Description:
1823  *   Initialize function pointers and device parameters
1824  **/
1825 
1826 static void ip6_tnl_dev_setup(struct net_device *dev)
1827 {
1828     dev->netdev_ops = &ip6_tnl_netdev_ops;
1829     dev->header_ops = &ip_tunnel_header_ops;
1830     dev->needs_free_netdev = true;
1831     dev->priv_destructor = ip6_dev_free;
1832 
1833     dev->type = ARPHRD_TUNNEL6;
1834     dev->flags |= IFF_NOARP;
1835     dev->addr_len = sizeof(struct in6_addr);
1836     dev->features |= NETIF_F_LLTX;
1837     netif_keep_dst(dev);
1838 
1839     dev->features       |= IPXIPX_FEATURES;
1840     dev->hw_features    |= IPXIPX_FEATURES;
1841 
1842     /* This perm addr will be used as interface identifier by IPv6 */
1843     dev->addr_assign_type = NET_ADDR_RANDOM;
1844     eth_random_addr(dev->perm_addr);
1845 }
1846 
1847 
1848 /**
1849  * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1850  *   @dev: virtual device associated with tunnel
1851  **/
1852 
1853 static inline int
1854 ip6_tnl_dev_init_gen(struct net_device *dev)
1855 {
1856     struct ip6_tnl *t = netdev_priv(dev);
1857     int ret;
1858     int t_hlen;
1859 
1860     t->dev = dev;
1861     t->net = dev_net(dev);
1862     dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1863     if (!dev->tstats)
1864         return -ENOMEM;
1865 
1866     ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1867     if (ret)
1868         goto free_stats;
1869 
1870     ret = gro_cells_init(&t->gro_cells, dev);
1871     if (ret)
1872         goto destroy_dst;
1873 
1874     t->tun_hlen = 0;
1875     t->hlen = t->encap_hlen + t->tun_hlen;
1876     t_hlen = t->hlen + sizeof(struct ipv6hdr);
1877 
1878     dev->type = ARPHRD_TUNNEL6;
1879     dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1880     dev->mtu = ETH_DATA_LEN - t_hlen;
1881     if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1882         dev->mtu -= 8;
1883     dev->min_mtu = ETH_MIN_MTU;
1884     dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1885 
1886     netdev_hold(dev, &t->dev_tracker, GFP_KERNEL);
1887     return 0;
1888 
1889 destroy_dst:
1890     dst_cache_destroy(&t->dst_cache);
1891 free_stats:
1892     free_percpu(dev->tstats);
1893     dev->tstats = NULL;
1894 
1895     return ret;
1896 }
1897 
1898 /**
1899  * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1900  *   @dev: virtual device associated with tunnel
1901  **/
1902 
1903 static int ip6_tnl_dev_init(struct net_device *dev)
1904 {
1905     struct ip6_tnl *t = netdev_priv(dev);
1906     int err = ip6_tnl_dev_init_gen(dev);
1907 
1908     if (err)
1909         return err;
1910     ip6_tnl_link_config(t);
1911     if (t->parms.collect_md)
1912         netif_keep_dst(dev);
1913     return 0;
1914 }
1915 
1916 /**
1917  * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1918  *   @dev: fallback device
1919  *
1920  * Return: 0
1921  **/
1922 
1923 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1924 {
1925     struct ip6_tnl *t = netdev_priv(dev);
1926     struct net *net = dev_net(dev);
1927     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1928 
1929     t->parms.proto = IPPROTO_IPV6;
1930 
1931     rcu_assign_pointer(ip6n->tnls_wc[0], t);
1932     return 0;
1933 }
1934 
1935 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1936                 struct netlink_ext_ack *extack)
1937 {
1938     u8 proto;
1939 
1940     if (!data || !data[IFLA_IPTUN_PROTO])
1941         return 0;
1942 
1943     proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1944     if (proto != IPPROTO_IPV6 &&
1945         proto != IPPROTO_IPIP &&
1946         proto != 0)
1947         return -EINVAL;
1948 
1949     return 0;
1950 }
1951 
1952 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1953                   struct __ip6_tnl_parm *parms)
1954 {
1955     memset(parms, 0, sizeof(*parms));
1956 
1957     if (!data)
1958         return;
1959 
1960     if (data[IFLA_IPTUN_LINK])
1961         parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1962 
1963     if (data[IFLA_IPTUN_LOCAL])
1964         parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1965 
1966     if (data[IFLA_IPTUN_REMOTE])
1967         parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1968 
1969     if (data[IFLA_IPTUN_TTL])
1970         parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1971 
1972     if (data[IFLA_IPTUN_ENCAP_LIMIT])
1973         parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1974 
1975     if (data[IFLA_IPTUN_FLOWINFO])
1976         parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1977 
1978     if (data[IFLA_IPTUN_FLAGS])
1979         parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1980 
1981     if (data[IFLA_IPTUN_PROTO])
1982         parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1983 
1984     if (data[IFLA_IPTUN_COLLECT_METADATA])
1985         parms->collect_md = true;
1986 
1987     if (data[IFLA_IPTUN_FWMARK])
1988         parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1989 }
1990 
1991 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1992                     struct ip_tunnel_encap *ipencap)
1993 {
1994     bool ret = false;
1995 
1996     memset(ipencap, 0, sizeof(*ipencap));
1997 
1998     if (!data)
1999         return ret;
2000 
2001     if (data[IFLA_IPTUN_ENCAP_TYPE]) {
2002         ret = true;
2003         ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
2004     }
2005 
2006     if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
2007         ret = true;
2008         ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
2009     }
2010 
2011     if (data[IFLA_IPTUN_ENCAP_SPORT]) {
2012         ret = true;
2013         ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
2014     }
2015 
2016     if (data[IFLA_IPTUN_ENCAP_DPORT]) {
2017         ret = true;
2018         ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2019     }
2020 
2021     return ret;
2022 }
2023 
2024 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2025                struct nlattr *tb[], struct nlattr *data[],
2026                struct netlink_ext_ack *extack)
2027 {
2028     struct net *net = dev_net(dev);
2029     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2030     struct ip_tunnel_encap ipencap;
2031     struct ip6_tnl *nt, *t;
2032     int err;
2033 
2034     nt = netdev_priv(dev);
2035 
2036     if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2037         err = ip6_tnl_encap_setup(nt, &ipencap);
2038         if (err < 0)
2039             return err;
2040     }
2041 
2042     ip6_tnl_netlink_parms(data, &nt->parms);
2043 
2044     if (nt->parms.collect_md) {
2045         if (rtnl_dereference(ip6n->collect_md_tun))
2046             return -EEXIST;
2047     } else {
2048         t = ip6_tnl_locate(net, &nt->parms, 0);
2049         if (!IS_ERR(t))
2050             return -EEXIST;
2051     }
2052 
2053     err = ip6_tnl_create2(dev);
2054     if (!err && tb[IFLA_MTU])
2055         ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2056 
2057     return err;
2058 }
2059 
2060 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2061                   struct nlattr *data[],
2062                   struct netlink_ext_ack *extack)
2063 {
2064     struct ip6_tnl *t = netdev_priv(dev);
2065     struct __ip6_tnl_parm p;
2066     struct net *net = t->net;
2067     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2068     struct ip_tunnel_encap ipencap;
2069 
2070     if (dev == ip6n->fb_tnl_dev)
2071         return -EINVAL;
2072 
2073     if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2074         int err = ip6_tnl_encap_setup(t, &ipencap);
2075 
2076         if (err < 0)
2077             return err;
2078     }
2079     ip6_tnl_netlink_parms(data, &p);
2080     if (p.collect_md)
2081         return -EINVAL;
2082 
2083     t = ip6_tnl_locate(net, &p, 0);
2084     if (!IS_ERR(t)) {
2085         if (t->dev != dev)
2086             return -EEXIST;
2087     } else
2088         t = netdev_priv(dev);
2089 
2090     ip6_tnl_update(t, &p);
2091     return 0;
2092 }
2093 
2094 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2095 {
2096     struct net *net = dev_net(dev);
2097     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2098 
2099     if (dev != ip6n->fb_tnl_dev)
2100         unregister_netdevice_queue(dev, head);
2101 }
2102 
2103 static size_t ip6_tnl_get_size(const struct net_device *dev)
2104 {
2105     return
2106         /* IFLA_IPTUN_LINK */
2107         nla_total_size(4) +
2108         /* IFLA_IPTUN_LOCAL */
2109         nla_total_size(sizeof(struct in6_addr)) +
2110         /* IFLA_IPTUN_REMOTE */
2111         nla_total_size(sizeof(struct in6_addr)) +
2112         /* IFLA_IPTUN_TTL */
2113         nla_total_size(1) +
2114         /* IFLA_IPTUN_ENCAP_LIMIT */
2115         nla_total_size(1) +
2116         /* IFLA_IPTUN_FLOWINFO */
2117         nla_total_size(4) +
2118         /* IFLA_IPTUN_FLAGS */
2119         nla_total_size(4) +
2120         /* IFLA_IPTUN_PROTO */
2121         nla_total_size(1) +
2122         /* IFLA_IPTUN_ENCAP_TYPE */
2123         nla_total_size(2) +
2124         /* IFLA_IPTUN_ENCAP_FLAGS */
2125         nla_total_size(2) +
2126         /* IFLA_IPTUN_ENCAP_SPORT */
2127         nla_total_size(2) +
2128         /* IFLA_IPTUN_ENCAP_DPORT */
2129         nla_total_size(2) +
2130         /* IFLA_IPTUN_COLLECT_METADATA */
2131         nla_total_size(0) +
2132         /* IFLA_IPTUN_FWMARK */
2133         nla_total_size(4) +
2134         0;
2135 }
2136 
2137 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2138 {
2139     struct ip6_tnl *tunnel = netdev_priv(dev);
2140     struct __ip6_tnl_parm *parm = &tunnel->parms;
2141 
2142     if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2143         nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2144         nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2145         nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2146         nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2147         nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2148         nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2149         nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2150         nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2151         goto nla_put_failure;
2152 
2153     if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2154         nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2155         nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2156         nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2157         goto nla_put_failure;
2158 
2159     if (parm->collect_md)
2160         if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2161             goto nla_put_failure;
2162 
2163     return 0;
2164 
2165 nla_put_failure:
2166     return -EMSGSIZE;
2167 }
2168 
2169 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2170 {
2171     struct ip6_tnl *tunnel = netdev_priv(dev);
2172 
2173     return tunnel->net;
2174 }
2175 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2176 
2177 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2178     [IFLA_IPTUN_LINK]       = { .type = NLA_U32 },
2179     [IFLA_IPTUN_LOCAL]      = { .len = sizeof(struct in6_addr) },
2180     [IFLA_IPTUN_REMOTE]     = { .len = sizeof(struct in6_addr) },
2181     [IFLA_IPTUN_TTL]        = { .type = NLA_U8 },
2182     [IFLA_IPTUN_ENCAP_LIMIT]    = { .type = NLA_U8 },
2183     [IFLA_IPTUN_FLOWINFO]       = { .type = NLA_U32 },
2184     [IFLA_IPTUN_FLAGS]      = { .type = NLA_U32 },
2185     [IFLA_IPTUN_PROTO]      = { .type = NLA_U8 },
2186     [IFLA_IPTUN_ENCAP_TYPE]     = { .type = NLA_U16 },
2187     [IFLA_IPTUN_ENCAP_FLAGS]    = { .type = NLA_U16 },
2188     [IFLA_IPTUN_ENCAP_SPORT]    = { .type = NLA_U16 },
2189     [IFLA_IPTUN_ENCAP_DPORT]    = { .type = NLA_U16 },
2190     [IFLA_IPTUN_COLLECT_METADATA]   = { .type = NLA_FLAG },
2191     [IFLA_IPTUN_FWMARK]     = { .type = NLA_U32 },
2192 };
2193 
2194 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2195     .kind       = "ip6tnl",
2196     .maxtype    = IFLA_IPTUN_MAX,
2197     .policy     = ip6_tnl_policy,
2198     .priv_size  = sizeof(struct ip6_tnl),
2199     .setup      = ip6_tnl_dev_setup,
2200     .validate   = ip6_tnl_validate,
2201     .newlink    = ip6_tnl_newlink,
2202     .changelink = ip6_tnl_changelink,
2203     .dellink    = ip6_tnl_dellink,
2204     .get_size   = ip6_tnl_get_size,
2205     .fill_info  = ip6_tnl_fill_info,
2206     .get_link_net   = ip6_tnl_get_link_net,
2207 };
2208 
2209 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2210     .handler    = ip4ip6_rcv,
2211     .err_handler    = ip4ip6_err,
2212     .priority   =   1,
2213 };
2214 
2215 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2216     .handler    = ip6ip6_rcv,
2217     .err_handler    = ip6ip6_err,
2218     .priority   =   1,
2219 };
2220 
2221 static struct xfrm6_tunnel mplsip6_handler __read_mostly = {
2222     .handler    = mplsip6_rcv,
2223     .err_handler    = mplsip6_err,
2224     .priority   =   1,
2225 };
2226 
2227 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2228 {
2229     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2230     struct net_device *dev, *aux;
2231     int h;
2232     struct ip6_tnl *t;
2233 
2234     for_each_netdev_safe(net, dev, aux)
2235         if (dev->rtnl_link_ops == &ip6_link_ops)
2236             unregister_netdevice_queue(dev, list);
2237 
2238     for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2239         t = rtnl_dereference(ip6n->tnls_r_l[h]);
2240         while (t) {
2241             /* If dev is in the same netns, it has already
2242              * been added to the list by the previous loop.
2243              */
2244             if (!net_eq(dev_net(t->dev), net))
2245                 unregister_netdevice_queue(t->dev, list);
2246             t = rtnl_dereference(t->next);
2247         }
2248     }
2249 
2250     t = rtnl_dereference(ip6n->tnls_wc[0]);
2251     while (t) {
2252         /* If dev is in the same netns, it has already
2253          * been added to the list by the previous loop.
2254          */
2255         if (!net_eq(dev_net(t->dev), net))
2256             unregister_netdevice_queue(t->dev, list);
2257         t = rtnl_dereference(t->next);
2258     }
2259 }
2260 
2261 static int __net_init ip6_tnl_init_net(struct net *net)
2262 {
2263     struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2264     struct ip6_tnl *t = NULL;
2265     int err;
2266 
2267     ip6n->tnls[0] = ip6n->tnls_wc;
2268     ip6n->tnls[1] = ip6n->tnls_r_l;
2269 
2270     if (!net_has_fallback_tunnels(net))
2271         return 0;
2272     err = -ENOMEM;
2273     ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2274                     NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2275 
2276     if (!ip6n->fb_tnl_dev)
2277         goto err_alloc_dev;
2278     dev_net_set(ip6n->fb_tnl_dev, net);
2279     ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2280     /* FB netdevice is special: we have one, and only one per netns.
2281      * Allowing to move it to another netns is clearly unsafe.
2282      */
2283     ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2284 
2285     err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2286     if (err < 0)
2287         goto err_register;
2288 
2289     err = register_netdev(ip6n->fb_tnl_dev);
2290     if (err < 0)
2291         goto err_register;
2292 
2293     t = netdev_priv(ip6n->fb_tnl_dev);
2294 
2295     strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2296     return 0;
2297 
2298 err_register:
2299     free_netdev(ip6n->fb_tnl_dev);
2300 err_alloc_dev:
2301     return err;
2302 }
2303 
2304 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2305 {
2306     struct net *net;
2307     LIST_HEAD(list);
2308 
2309     rtnl_lock();
2310     list_for_each_entry(net, net_list, exit_list)
2311         ip6_tnl_destroy_tunnels(net, &list);
2312     unregister_netdevice_many(&list);
2313     rtnl_unlock();
2314 }
2315 
2316 static struct pernet_operations ip6_tnl_net_ops = {
2317     .init = ip6_tnl_init_net,
2318     .exit_batch = ip6_tnl_exit_batch_net,
2319     .id   = &ip6_tnl_net_id,
2320     .size = sizeof(struct ip6_tnl_net),
2321 };
2322 
2323 /**
2324  * ip6_tunnel_init - register protocol and reserve needed resources
2325  *
2326  * Return: 0 on success
2327  **/
2328 
2329 static int __init ip6_tunnel_init(void)
2330 {
2331     int  err;
2332 
2333     if (!ipv6_mod_enabled())
2334         return -EOPNOTSUPP;
2335 
2336     err = register_pernet_device(&ip6_tnl_net_ops);
2337     if (err < 0)
2338         goto out_pernet;
2339 
2340     err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2341     if (err < 0) {
2342         pr_err("%s: can't register ip4ip6\n", __func__);
2343         goto out_ip4ip6;
2344     }
2345 
2346     err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2347     if (err < 0) {
2348         pr_err("%s: can't register ip6ip6\n", __func__);
2349         goto out_ip6ip6;
2350     }
2351 
2352     if (ip6_tnl_mpls_supported()) {
2353         err = xfrm6_tunnel_register(&mplsip6_handler, AF_MPLS);
2354         if (err < 0) {
2355             pr_err("%s: can't register mplsip6\n", __func__);
2356             goto out_mplsip6;
2357         }
2358     }
2359 
2360     err = rtnl_link_register(&ip6_link_ops);
2361     if (err < 0)
2362         goto rtnl_link_failed;
2363 
2364     return 0;
2365 
2366 rtnl_link_failed:
2367     if (ip6_tnl_mpls_supported())
2368         xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS);
2369 out_mplsip6:
2370     xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2371 out_ip6ip6:
2372     xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2373 out_ip4ip6:
2374     unregister_pernet_device(&ip6_tnl_net_ops);
2375 out_pernet:
2376     return err;
2377 }
2378 
2379 /**
2380  * ip6_tunnel_cleanup - free resources and unregister protocol
2381  **/
2382 
2383 static void __exit ip6_tunnel_cleanup(void)
2384 {
2385     rtnl_link_unregister(&ip6_link_ops);
2386     if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2387         pr_info("%s: can't deregister ip4ip6\n", __func__);
2388 
2389     if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2390         pr_info("%s: can't deregister ip6ip6\n", __func__);
2391 
2392     if (ip6_tnl_mpls_supported() &&
2393         xfrm6_tunnel_deregister(&mplsip6_handler, AF_MPLS))
2394         pr_info("%s: can't deregister mplsip6\n", __func__);
2395     unregister_pernet_device(&ip6_tnl_net_ops);
2396 }
2397 
2398 module_init(ip6_tunnel_init);
2399 module_exit(ip6_tunnel_cleanup);
[ Source navigation ] [ Diff markup ] [ Identifier search ] [ general search ]

This page was automatically generated by the 2.1.0 LXR engine.

The LXR team
Valid CSS 2.1! Valid HTML 4.01!