net/ipv6/rpl_iptunnel.c

0001 // SPDX-License-Identifier: GPL-2.0-only
0002 /*
0003  * Authors:
0004  * (C) 2020 Alexander Aring <alex.aring@gmail.com>
0005  */
0006
0007 #include <linux/rpl_iptunnel.h>
0008
0009 #include <net/dst_cache.h>
0010 #include <net/ip6_route.h>
0011 #include <net/lwtunnel.h>
0012 #include <net/ipv6.h>
0013 #include <net/rpl.h>
0014
0015 struct rpl_iptunnel_encap {
0016     struct ipv6_rpl_sr_hdr srh[0];
0017 };
0018
0019 struct rpl_lwt {
0020     struct dst_cache cache;
0021     struct rpl_iptunnel_encap tuninfo;
0022 };
0023
0024 static inline struct rpl_lwt *rpl_lwt_lwtunnel(struct lwtunnel_state *lwt)
0025 {
0026     return (struct rpl_lwt *)lwt->data;
0027 }
0028
0029 static inline struct rpl_iptunnel_encap *
0030 rpl_encap_lwtunnel(struct lwtunnel_state *lwt)
0031 {
0032     return &rpl_lwt_lwtunnel(lwt)->tuninfo;
0033 }
0034
0035 static const struct nla_policy rpl_iptunnel_policy[RPL_IPTUNNEL_MAX + 1] = {
0036     [RPL_IPTUNNEL_SRH]  = { .type = NLA_BINARY },
0037 };
0038
0039 static bool rpl_validate_srh(struct net *net, struct ipv6_rpl_sr_hdr *srh,
0040                  size_t seglen)
0041 {
0042     int err;
0043
0044     if ((srh->hdrlen << 3) != seglen)
0045         return false;
0046
0047     /* check at least one segment and seglen fit with segments_left */
0048     if (!srh->segments_left ||
0049         (srh->segments_left * sizeof(struct in6_addr)) != seglen)
0050         return false;
0051
0052     if (srh->cmpri || srh->cmpre)
0053         return false;
0054
0055     err = ipv6_chk_rpl_srh_loop(net, srh->rpl_segaddr,
0056                     srh->segments_left);
0057     if (err)
0058         return false;
0059
0060     if (ipv6_addr_type(&srh->rpl_segaddr[srh->segments_left - 1]) &
0061         IPV6_ADDR_MULTICAST)
0062         return false;
0063
0064     return true;
0065 }
0066
0067 static int rpl_build_state(struct net *net, struct nlattr *nla,
0068                unsigned int family, const void *cfg,
0069                struct lwtunnel_state **ts,
0070                struct netlink_ext_ack *extack)
0071 {
0072     struct nlattr *tb[RPL_IPTUNNEL_MAX + 1];
0073     struct lwtunnel_state *newts;
0074     struct ipv6_rpl_sr_hdr *srh;
0075     struct rpl_lwt *rlwt;
0076     int err, srh_len;
0077
0078     if (family != AF_INET6)
0079         return -EINVAL;
0080
0081     err = nla_parse_nested(tb, RPL_IPTUNNEL_MAX, nla,
0082                    rpl_iptunnel_policy, extack);
0083     if (err < 0)
0084         return err;
0085
0086     if (!tb[RPL_IPTUNNEL_SRH])
0087         return -EINVAL;
0088
0089     srh = nla_data(tb[RPL_IPTUNNEL_SRH]);
0090     srh_len = nla_len(tb[RPL_IPTUNNEL_SRH]);
0091
0092     if (srh_len < sizeof(*srh))
0093         return -EINVAL;
0094
0095     /* verify that SRH is consistent */
0096     if (!rpl_validate_srh(net, srh, srh_len - sizeof(*srh)))
0097         return -EINVAL;
0098
0099     newts = lwtunnel_state_alloc(srh_len + sizeof(*rlwt));
0100     if (!newts)
0101         return -ENOMEM;
0102
0103     rlwt = rpl_lwt_lwtunnel(newts);
0104
0105     err = dst_cache_init(&rlwt->cache, GFP_ATOMIC);
0106     if (err) {
0107         kfree(newts);
0108         return err;
0109     }
0110
0111     memcpy(&rlwt->tuninfo.srh, srh, srh_len);
0112
0113     newts->type = LWTUNNEL_ENCAP_RPL;
0114     newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
0115     newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
0116
0117     *ts = newts;
0118
0119     return 0;
0120 }
0121
0122 static void rpl_destroy_state(struct lwtunnel_state *lwt)
0123 {
0124     dst_cache_destroy(&rpl_lwt_lwtunnel(lwt)->cache);
0125 }
0126
0127 static int rpl_do_srh_inline(struct sk_buff *skb, const struct rpl_lwt *rlwt,
0128                  const struct ipv6_rpl_sr_hdr *srh)
0129 {
0130     struct ipv6_rpl_sr_hdr *isrh, *csrh;
0131     const struct ipv6hdr *oldhdr;
0132     struct ipv6hdr *hdr;
0133     unsigned char *buf;
0134     size_t hdrlen;
0135     int err;
0136
0137     oldhdr = ipv6_hdr(skb);
0138
0139     buf = kcalloc(struct_size(srh, segments.addr, srh->segments_left), 2, GFP_ATOMIC);
0140     if (!buf)
0141         return -ENOMEM;
0142
0143     isrh = (struct ipv6_rpl_sr_hdr *)buf;
0144     csrh = (struct ipv6_rpl_sr_hdr *)(buf + ((srh->hdrlen + 1) << 3));
0145
0146     memcpy(isrh, srh, sizeof(*isrh));
0147     memcpy(isrh->rpl_segaddr, &srh->rpl_segaddr[1],
0148            (srh->segments_left - 1) * 16);
0149     isrh->rpl_segaddr[srh->segments_left - 1] = oldhdr->daddr;
0150
0151     ipv6_rpl_srh_compress(csrh, isrh, &srh->rpl_segaddr[0],
0152                   isrh->segments_left - 1);
0153
0154     hdrlen = ((csrh->hdrlen + 1) << 3);
0155
0156     err = skb_cow_head(skb, hdrlen + skb->mac_len);
0157     if (unlikely(err)) {
0158         kfree(buf);
0159         return err;
0160     }
0161
0162     skb_pull(skb, sizeof(struct ipv6hdr));
0163     skb_postpull_rcsum(skb, skb_network_header(skb),
0164                sizeof(struct ipv6hdr));
0165
0166     skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
0167     skb_reset_network_header(skb);
0168     skb_mac_header_rebuild(skb);
0169
0170     hdr = ipv6_hdr(skb);
0171     memmove(hdr, oldhdr, sizeof(*hdr));
0172     isrh = (void *)hdr + sizeof(*hdr);
0173     memcpy(isrh, csrh, hdrlen);
0174
0175     isrh->nexthdr = hdr->nexthdr;
0176     hdr->nexthdr = NEXTHDR_ROUTING;
0177     hdr->daddr = srh->rpl_segaddr[0];
0178
0179     ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
0180     skb_set_transport_header(skb, sizeof(struct ipv6hdr));
0181
0182     skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
0183
0184     kfree(buf);
0185
0186     return 0;
0187 }
0188
0189 static int rpl_do_srh(struct sk_buff *skb, const struct rpl_lwt *rlwt)
0190 {
0191     struct dst_entry *dst = skb_dst(skb);
0192     struct rpl_iptunnel_encap *tinfo;
0193
0194     if (skb->protocol != htons(ETH_P_IPV6))
0195         return -EINVAL;
0196
0197     tinfo = rpl_encap_lwtunnel(dst->lwtstate);
0198
0199     return rpl_do_srh_inline(skb, rlwt, tinfo->srh);
0200 }
0201
0202 static int rpl_output(struct net *net, struct sock *sk, struct sk_buff *skb)
0203 {
0204     struct dst_entry *orig_dst = skb_dst(skb);
0205     struct dst_entry *dst = NULL;
0206     struct rpl_lwt *rlwt;
0207     int err;
0208
0209     rlwt = rpl_lwt_lwtunnel(orig_dst->lwtstate);
0210
0211     err = rpl_do_srh(skb, rlwt);
0212     if (unlikely(err))
0213         goto drop;
0214
0215     preempt_disable();
0216     dst = dst_cache_get(&rlwt->cache);
0217     preempt_enable();
0218
0219     if (unlikely(!dst)) {
0220         struct ipv6hdr *hdr = ipv6_hdr(skb);
0221         struct flowi6 fl6;
0222
0223         memset(&fl6, 0, sizeof(fl6));
0224         fl6.daddr = hdr->daddr;
0225         fl6.saddr = hdr->saddr;
0226         fl6.flowlabel = ip6_flowinfo(hdr);
0227         fl6.flowi6_mark = skb->mark;
0228         fl6.flowi6_proto = hdr->nexthdr;
0229
0230         dst = ip6_route_output(net, NULL, &fl6);
0231         if (dst->error) {
0232             err = dst->error;
0233             dst_release(dst);
0234             goto drop;
0235         }
0236
0237         preempt_disable();
0238         dst_cache_set_ip6(&rlwt->cache, dst, &fl6.saddr);
0239         preempt_enable();
0240     }
0241
0242     skb_dst_drop(skb);
0243     skb_dst_set(skb, dst);
0244
0245     err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
0246     if (unlikely(err))
0247         goto drop;
0248
0249     return dst_output(net, sk, skb);
0250
0251 drop:
0252     kfree_skb(skb);
0253     return err;
0254 }
0255
0256 static int rpl_input(struct sk_buff *skb)
0257 {
0258     struct dst_entry *orig_dst = skb_dst(skb);
0259     struct dst_entry *dst = NULL;
0260     struct rpl_lwt *rlwt;
0261     int err;
0262
0263     rlwt = rpl_lwt_lwtunnel(orig_dst->lwtstate);
0264
0265     err = rpl_do_srh(skb, rlwt);
0266     if (unlikely(err)) {
0267         kfree_skb(skb);
0268         return err;
0269     }
0270
0271     preempt_disable();
0272     dst = dst_cache_get(&rlwt->cache);
0273     preempt_enable();
0274
0275     skb_dst_drop(skb);
0276
0277     if (!dst) {
0278         ip6_route_input(skb);
0279         dst = skb_dst(skb);
0280         if (!dst->error) {
0281             preempt_disable();
0282             dst_cache_set_ip6(&rlwt->cache, dst,
0283                       &ipv6_hdr(skb)->saddr);
0284             preempt_enable();
0285         }
0286     } else {
0287         skb_dst_set(skb, dst);
0288     }
0289
0290     err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
0291     if (unlikely(err))
0292         return err;
0293
0294     return dst_input(skb);
0295 }
0296
0297 static int nla_put_rpl_srh(struct sk_buff *skb, int attrtype,
0298                struct rpl_iptunnel_encap *tuninfo)
0299 {
0300     struct rpl_iptunnel_encap *data;
0301     struct nlattr *nla;
0302     int len;
0303
0304     len = RPL_IPTUNNEL_SRH_SIZE(tuninfo->srh);
0305
0306     nla = nla_reserve(skb, attrtype, len);
0307     if (!nla)
0308         return -EMSGSIZE;
0309
0310     data = nla_data(nla);
0311     memcpy(data, tuninfo->srh, len);
0312
0313     return 0;
0314 }
0315
0316 static int rpl_fill_encap_info(struct sk_buff *skb,
0317                    struct lwtunnel_state *lwtstate)
0318 {
0319     struct rpl_iptunnel_encap *tuninfo = rpl_encap_lwtunnel(lwtstate);
0320
0321     if (nla_put_rpl_srh(skb, RPL_IPTUNNEL_SRH, tuninfo))
0322         return -EMSGSIZE;
0323
0324     return 0;
0325 }
0326
0327 static int rpl_encap_nlsize(struct lwtunnel_state *lwtstate)
0328 {
0329     struct rpl_iptunnel_encap *tuninfo = rpl_encap_lwtunnel(lwtstate);
0330
0331     return nla_total_size(RPL_IPTUNNEL_SRH_SIZE(tuninfo->srh));
0332 }
0333
0334 static int rpl_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
0335 {
0336     struct rpl_iptunnel_encap *a_hdr = rpl_encap_lwtunnel(a);
0337     struct rpl_iptunnel_encap *b_hdr = rpl_encap_lwtunnel(b);
0338     int len = RPL_IPTUNNEL_SRH_SIZE(a_hdr->srh);
0339
0340     if (len != RPL_IPTUNNEL_SRH_SIZE(b_hdr->srh))
0341         return 1;
0342
0343     return memcmp(a_hdr, b_hdr, len);
0344 }
0345
0346 static const struct lwtunnel_encap_ops rpl_ops = {
0347     .build_state    = rpl_build_state,
0348     .destroy_state  = rpl_destroy_state,
0349     .output     = rpl_output,
0350     .input      = rpl_input,
0351     .fill_encap = rpl_fill_encap_info,
0352     .get_encap_size = rpl_encap_nlsize,
0353     .cmp_encap  = rpl_encap_cmp,
0354     .owner      = THIS_MODULE,
0355 };
0356
0357 int __init rpl_init(void)
0358 {
0359     int err;
0360
0361     err = lwtunnel_encap_add_ops(&rpl_ops, LWTUNNEL_ENCAP_RPL);
0362     if (err)
0363         goto out;
0364
0365     pr_info("RPL Segment Routing with IPv6\n");
0366
0367     return 0;
0368
0369 out:
0370     return err;
0371 }
0372
0373 void rpl_exit(void)
0374 {
0375     lwtunnel_encap_del_ops(&rpl_ops, LWTUNNEL_ENCAP_RPL);
0376 }