net/sched/act_csum.c

   1 /*
   2  * Checksum updating actions
   3  *
   4  * Copyright (c) 2010 Gregoire Baron <baronchon@n7mm.org>
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License as published by the Free
   8  * Software Foundation; either version 2 of the License, or (at your option)
   9  * any later version.
  10  *
  11  */
  12
  13 #include <linux/types.h>
  14 #include <linux/init.h>
  15 #include <linux/kernel.h>
  16 #include <linux/module.h>
  17 #include <linux/spinlock.h>
  18
  19 #include <linux/netlink.h>
  20 #include <net/netlink.h>
  21 #include <linux/rtnetlink.h>
  22
  23 #include <linux/skbuff.h>
  24
  25 #include <net/ip.h>
  26 #include <net/ipv6.h>
  27 #include <net/icmp.h>
  28 #include <linux/icmpv6.h>
  29 #include <linux/igmp.h>
  30 #include <net/tcp.h>
  31 #include <net/udp.h>
  32 #include <net/ip6_checksum.h>
  33
  34 #include <net/act_api.h>
  35
  36 #include <linux/tc_act/tc_csum.h>
  37 #include <net/tc_act/tc_csum.h>
  38
  39 #define CSUM_TAB_MASK 15
  40
  41 static const struct nla_policy csum_policy[TCA_CSUM_MAX + 1] = {
  42         [TCA_CSUM_PARMS] = { .len = sizeof(struct tc_csum), },
  43 };
  44
  45 static int csum_net_id;
  46 static struct tc_action_ops act_csum_ops;
  47
  48 static int tcf_csum_init(struct net *net, struct nlattr *nla,
  49                          struct nlattr *est, struct tc_action **a, int ovr,
  50                          int bind)
  51 {
  52         struct tc_action_net *tn = net_generic(net, csum_net_id);
  53         struct nlattr *tb[TCA_CSUM_MAX + 1];
  54         struct tc_csum *parm;
  55         struct tcf_csum *p;
  56         int ret = 0, err;
  57
  58         if (nla == NULL)
  59                 return -EINVAL;
  60
  61         err = nla_parse_nested(tb, TCA_CSUM_MAX, nla, csum_policy);
  62         if (err < 0)
  63                 return err;
  64
  65         if (tb[TCA_CSUM_PARMS] == NULL)
  66                 return -EINVAL;
  67         parm = nla_data(tb[TCA_CSUM_PARMS]);
  68
  69         if (!tcf_hash_check(tn, parm->index, a, bind)) {
  70                 ret = tcf_hash_create(tn, parm->index, est, a,
  71                                       &act_csum_ops, bind, false);
  72                 if (ret)
  73                         return ret;
  74                 ret = ACT_P_CREATED;
  75         } else {
  76                 if (bind)/* dont override defaults */
  77                         return 0;
  78                 tcf_hash_release(*a, bind);
  79                 if (!ovr)
  80                         return -EEXIST;
  81         }
  82
  83         p = to_tcf_csum(*a);
  84         spin_lock_bh(&p->tcf_lock);
  85         p->tcf_action = parm->action;
  86         p->update_flags = parm->update_flags;
  87         spin_unlock_bh(&p->tcf_lock);
  88
  89         if (ret == ACT_P_CREATED)
  90                 tcf_hash_insert(tn, *a);
  91
  92         return ret;
  93 }
  94
  95 /**
  96  * tcf_csum_skb_nextlayer - Get next layer pointer
  97  * @skb: sk_buff to use
  98  * @ihl: previous summed headers length
  99  * @ipl: complete packet length
 100  * @jhl: next header length
 101  *
 102  * Check the expected next layer availability in the specified sk_buff.
 103  * Return the next layer pointer if pass, NULL otherwise.
 104  */
 105 static void *tcf_csum_skb_nextlayer(struct sk_buff *skb,
 106                                     unsigned int ihl, unsigned int ipl,
 107                                     unsigned int jhl)
 108 {
 109         int ntkoff = skb_network_offset(skb);
 110         int hl = ihl + jhl;
 111
 112         if (!pskb_may_pull(skb, ipl + ntkoff) || (ipl < hl) ||
 113             skb_try_make_writable(skb, hl + ntkoff))
 114                 return NULL;
 115         else
 116                 return (void *)(skb_network_header(skb) + ihl);
 117 }
 118
 119 static int tcf_csum_ipv4_icmp(struct sk_buff *skb, unsigned int ihl,
 120                               unsigned int ipl)
 121 {
 122         struct icmphdr *icmph;
 123
 124         icmph = tcf_csum_skb_nextlayer(skb, ihl, ipl, sizeof(*icmph));
 125         if (icmph == NULL)
 126                 return 0;
 127
 128         icmph->checksum = 0;
 129         skb->csum = csum_partial(icmph, ipl - ihl, 0);
 130         icmph->checksum = csum_fold(skb->csum);
 131
 132         skb->ip_summed = CHECKSUM_NONE;
 133
 134         return 1;
 135 }
 136
 137 static int tcf_csum_ipv4_igmp(struct sk_buff *skb,
 138                               unsigned int ihl, unsigned int ipl)
 139 {
 140         struct igmphdr *igmph;
 141
 142         igmph = tcf_csum_skb_nextlayer(skb, ihl, ipl, sizeof(*igmph));
 143         if (igmph == NULL)
 144                 return 0;
 145
 146         igmph->csum = 0;
 147         skb->csum = csum_partial(igmph, ipl - ihl, 0);
 148         igmph->csum = csum_fold(skb->csum);
 149
 150         skb->ip_summed = CHECKSUM_NONE;
 151
 152         return 1;
 153 }
 154
 155 static int tcf_csum_ipv6_icmp(struct sk_buff *skb, unsigned int ihl,
 156                               unsigned int ipl)
 157 {
 158         struct icmp6hdr *icmp6h;
 159         const struct ipv6hdr *ip6h;
 160
 161         icmp6h = tcf_csum_skb_nextlayer(skb, ihl, ipl, sizeof(*icmp6h));
 162         if (icmp6h == NULL)
 163                 return 0;
 164
 165         ip6h = ipv6_hdr(skb);
 166         icmp6h->icmp6_cksum = 0;
 167         skb->csum = csum_partial(icmp6h, ipl - ihl, 0);
 168         icmp6h->icmp6_cksum = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 169                                               ipl - ihl, IPPROTO_ICMPV6,
 170                                               skb->csum);
 171
 172         skb->ip_summed = CHECKSUM_NONE;
 173
 174         return 1;
 175 }
 176
 177 static int tcf_csum_ipv4_tcp(struct sk_buff *skb, unsigned int ihl,
 178                              unsigned int ipl)
 179 {
 180         struct tcphdr *tcph;
 181         const struct iphdr *iph;
 182
 183         if (skb_is_gso(skb) && skb_shinfo(skb)->gso_type & SKB_GSO_TCPV4)
 184                 return 1;
 185
 186         tcph = tcf_csum_skb_nextlayer(skb, ihl, ipl, sizeof(*tcph));
 187         if (tcph == NULL)
 188                 return 0;
 189
 190         iph = ip_hdr(skb);
 191         tcph->check = 0;
 192         skb->csum = csum_partial(tcph, ipl - ihl, 0);
 193         tcph->check = tcp_v4_check(ipl - ihl,
 194                                    iph->saddr, iph->daddr, skb->csum);
 195
 196         skb->ip_summed = CHECKSUM_NONE;
 197
 198         return 1;
 199 }
 200
 201 static int tcf_csum_ipv6_tcp(struct sk_buff *skb, unsigned int ihl,
 202                              unsigned int ipl)
 203 {
 204         struct tcphdr *tcph;
 205         const struct ipv6hdr *ip6h;
 206
 207         if (skb_is_gso(skb) && skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6)
 208                 return 1;
 209
 210         tcph = tcf_csum_skb_nextlayer(skb, ihl, ipl, sizeof(*tcph));
 211         if (tcph == NULL)
 212                 return 0;
 213
 214         ip6h = ipv6_hdr(skb);
 215         tcph->check = 0;
 216         skb->csum = csum_partial(tcph, ipl - ihl, 0);
 217         tcph->check = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr,
 218                                       ipl - ihl, IPPROTO_TCP,
 219                                       skb->csum);
 220
 221         skb->ip_summed = CHECKSUM_NONE;
 222
 223         return 1;
 224 }
 225
 226 static int tcf_csum_ipv4_udp(struct sk_buff *skb, unsigned int ihl,
 227                              unsigned int ipl, int udplite)
 228 {
 229         struct udphdr *udph;
 230         const struct iphdr *iph;
 231         u16 ul;
 232
 233         if (skb_is_gso(skb) && skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
 234                 return 1;
 235
 236         /*
 237          * Support both UDP and UDPLITE checksum algorithms, Don't use
 238          * udph->len to get the real length without any protocol check,
 239          * UDPLITE uses udph->len for another thing,
 240          * Use iph->tot_len, or just ipl.
 241          */
 242
 243         udph = tcf_csum_skb_nextlayer(skb, ihl, ipl, sizeof(*udph));
 244         if (udph == NULL)
 245                 return 0;
 246
 247         iph = ip_hdr(skb);
 248         ul = ntohs(udph->len);
 249
 250         if (udplite || udph->check) {
 251
 252                 udph->check = 0;
 253
 254                 if (udplite) {
 255                         if (ul == 0)
 256                                 skb->csum = csum_partial(udph, ipl - ihl, 0);
 257                         else if ((ul >= sizeof(*udph)) && (ul <= ipl - ihl))
 258                                 skb->csum = csum_partial(udph, ul, 0);
 259                         else
 260                                 goto ignore_obscure_skb;
 261                 } else {
 262                         if (ul != ipl - ihl)
 263                                 goto ignore_obscure_skb;
 264
 265                         skb->csum = csum_partial(udph, ul, 0);
 266                 }
 267
 268                 udph->check = csum_tcpudp_magic(iph->saddr, iph->daddr,
 269                                                 ul, iph->protocol,
 270                                                 skb->csum);
 271
 272                 if (!udph->check)
 273                         udph->check = CSUM_MANGLED_0;
 274         }
 275
 276         skb->ip_summed = CHECKSUM_NONE;
 277
 278 ignore_obscure_skb:
 279         return 1;
 280 }
 281
 282 static int tcf_csum_ipv6_udp(struct sk_buff *skb, unsigned int ihl,
 283                              unsigned int ipl, int udplite)
 284 {
 285         struct udphdr *udph;
 286         const struct ipv6hdr *ip6h;
 287         u16 ul;
 288
 289         if (skb_is_gso(skb) && skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
 290                 return 1;
 291
 292         /*
 293          * Support both UDP and UDPLITE checksum algorithms, Don't use
 294          * udph->len to get the real length without any protocol check,
 295          * UDPLITE uses udph->len for another thing,
 296          * Use ip6h->payload_len + sizeof(*ip6h) ... , or just ipl.
 297          */
 298
 299         udph = tcf_csum_skb_nextlayer(skb, ihl, ipl, sizeof(*udph));
 300         if (udph == NULL)
 301                 return 0;
 302
 303         ip6h = ipv6_hdr(skb);
 304         ul = ntohs(udph->len);
 305
 306         udph->check = 0;
 307
 308         if (udplite) {
 309                 if (ul == 0)
 310                         skb->csum = csum_partial(udph, ipl - ihl, 0);
 311
 312                 else if ((ul >= sizeof(*udph)) && (ul <= ipl - ihl))
 313                         skb->csum = csum_partial(udph, ul, 0);
 314
 315                 else
 316                         goto ignore_obscure_skb;
 317         } else {
 318                 if (ul != ipl - ihl)
 319                         goto ignore_obscure_skb;
 320
 321                 skb->csum = csum_partial(udph, ul, 0);
 322         }
 323
 324         udph->check = csum_ipv6_magic(&ip6h->saddr, &ip6h->daddr, ul,
 325                                       udplite ? IPPROTO_UDPLITE : IPPROTO_UDP,
 326                                       skb->csum);
 327
 328         if (!udph->check)
 329                 udph->check = CSUM_MANGLED_0;
 330
 331         skb->ip_summed = CHECKSUM_NONE;
 332
 333 ignore_obscure_skb:
 334         return 1;
 335 }
 336
 337 static int tcf_csum_ipv4(struct sk_buff *skb, u32 update_flags)
 338 {
 339         const struct iphdr *iph;
 340         int ntkoff;
 341
 342         ntkoff = skb_network_offset(skb);
 343
 344         if (!pskb_may_pull(skb, sizeof(*iph) + ntkoff))
 345                 goto fail;
 346
 347         iph = ip_hdr(skb);
 348
 349         switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
 350         case IPPROTO_ICMP:
 351                 if (update_flags & TCA_CSUM_UPDATE_FLAG_ICMP)
 352                         if (!tcf_csum_ipv4_icmp(skb, iph->ihl * 4,
 353                                                 ntohs(iph->tot_len)))
 354                                 goto fail;
 355                 break;
 356         case IPPROTO_IGMP:
 357                 if (update_flags & TCA_CSUM_UPDATE_FLAG_IGMP)
 358                         if (!tcf_csum_ipv4_igmp(skb, iph->ihl * 4,
 359                                                 ntohs(iph->tot_len)))
 360                                 goto fail;
 361                 break;
 362         case IPPROTO_TCP:
 363                 if (update_flags & TCA_CSUM_UPDATE_FLAG_TCP)
 364                         if (!tcf_csum_ipv4_tcp(skb, iph->ihl * 4,
 365                                                ntohs(iph->tot_len)))
 366                                 goto fail;
 367                 break;
 368         case IPPROTO_UDP:
 369                 if (update_flags & TCA_CSUM_UPDATE_FLAG_UDP)
 370                         if (!tcf_csum_ipv4_udp(skb, iph->ihl * 4,
 371                                                ntohs(iph->tot_len), 0))
 372                                 goto fail;
 373                 break;
 374         case IPPROTO_UDPLITE:
 375                 if (update_flags & TCA_CSUM_UPDATE_FLAG_UDPLITE)
 376                         if (!tcf_csum_ipv4_udp(skb, iph->ihl * 4,
 377                                                ntohs(iph->tot_len), 1))
 378                                 goto fail;
 379                 break;
 380         }
 381
 382         if (update_flags & TCA_CSUM_UPDATE_FLAG_IPV4HDR) {
 383                 if (skb_try_make_writable(skb, sizeof(*iph) + ntkoff))
 384                         goto fail;
 385
 386                 ip_send_check(ip_hdr(skb));
 387         }
 388
 389         return 1;
 390
 391 fail:
 392         return 0;
 393 }
 394
 395 static int tcf_csum_ipv6_hopopts(struct ipv6_opt_hdr *ip6xh, unsigned int ixhl,
 396                                  unsigned int *pl)
 397 {
 398         int off, len, optlen;
 399         unsigned char *xh = (void *)ip6xh;
 400
 401         off = sizeof(*ip6xh);
 402         len = ixhl - off;
 403
 404         while (len > 1) {
 405                 switch (xh[off]) {
 406                 case IPV6_TLV_PAD1:
 407                         optlen = 1;
 408                         break;
 409                 case IPV6_TLV_JUMBO:
 410                         optlen = xh[off + 1] + 2;
 411                         if (optlen != 6 || len < 6 || (off & 3) != 2)
 412                                 /* wrong jumbo option length/alignment */
 413                                 return 0;
 414                         *pl = ntohl(*(__be32 *)(xh + off + 2));
 415                         goto done;
 416                 default:
 417                         optlen = xh[off + 1] + 2;
 418                         if (optlen > len)
 419                                 /* ignore obscure options */
 420                                 goto done;
 421                         break;
 422                 }
 423                 off += optlen;
 424                 len -= optlen;
 425         }
 426
 427 done:
 428         return 1;
 429 }
 430
 431 static int tcf_csum_ipv6(struct sk_buff *skb, u32 update_flags)
 432 {
 433         struct ipv6hdr *ip6h;
 434         struct ipv6_opt_hdr *ip6xh;
 435         unsigned int hl, ixhl;
 436         unsigned int pl;
 437         int ntkoff;
 438         u8 nexthdr;
 439
 440         ntkoff = skb_network_offset(skb);
 441
 442         hl = sizeof(*ip6h);
 443
 444         if (!pskb_may_pull(skb, hl + ntkoff))
 445                 goto fail;
 446
 447         ip6h = ipv6_hdr(skb);
 448
 449         pl = ntohs(ip6h->payload_len);
 450         nexthdr = ip6h->nexthdr;
 451
 452         do {
 453                 switch (nexthdr) {
 454                 case NEXTHDR_FRAGMENT:
 455                         goto ignore_skb;
 456                 case NEXTHDR_ROUTING:
 457                 case NEXTHDR_HOP:
 458                 case NEXTHDR_DEST:
 459                         if (!pskb_may_pull(skb, hl + sizeof(*ip6xh) + ntkoff))
 460                                 goto fail;
 461                         ip6xh = (void *)(skb_network_header(skb) + hl);
 462                         ixhl = ipv6_optlen(ip6xh);
 463                         if (!pskb_may_pull(skb, hl + ixhl + ntkoff))
 464                                 goto fail;
 465                         ip6xh = (void *)(skb_network_header(skb) + hl);
 466                         if ((nexthdr == NEXTHDR_HOP) &&
 467                             !(tcf_csum_ipv6_hopopts(ip6xh, ixhl, &pl)))
 468                                 goto fail;
 469                         nexthdr = ip6xh->nexthdr;
 470                         hl += ixhl;
 471                         break;
 472                 case IPPROTO_ICMPV6:
 473                         if (update_flags & TCA_CSUM_UPDATE_FLAG_ICMP)
 474                                 if (!tcf_csum_ipv6_icmp(skb,
 475                                                         hl, pl + sizeof(*ip6h)))
 476                                         goto fail;
 477                         goto done;
 478                 case IPPROTO_TCP:
 479                         if (update_flags & TCA_CSUM_UPDATE_FLAG_TCP)
 480                                 if (!tcf_csum_ipv6_tcp(skb,
 481                                                        hl, pl + sizeof(*ip6h)))
 482                                         goto fail;
 483                         goto done;
 484                 case IPPROTO_UDP:
 485                         if (update_flags & TCA_CSUM_UPDATE_FLAG_UDP)
 486                                 if (!tcf_csum_ipv6_udp(skb, hl,
 487                                                        pl + sizeof(*ip6h), 0))
 488                                         goto fail;
 489                         goto done;
 490                 case IPPROTO_UDPLITE:
 491                         if (update_flags & TCA_CSUM_UPDATE_FLAG_UDPLITE)
 492                                 if (!tcf_csum_ipv6_udp(skb, hl,
 493                                                        pl + sizeof(*ip6h), 1))
 494                                         goto fail;
 495                         goto done;
 496                 default:
 497                         goto ignore_skb;
 498                 }
 499         } while (pskb_may_pull(skb, hl + 1 + ntkoff));
 500
 501 done:
 502 ignore_skb:
 503         return 1;
 504
 505 fail:
 506         return 0;
 507 }
 508
 509 static int tcf_csum(struct sk_buff *skb, const struct tc_action *a,
 510                     struct tcf_result *res)
 511 {
 512         struct tcf_csum *p = to_tcf_csum(a);
 513         int action;
 514         u32 update_flags;
 515
 516         spin_lock(&p->tcf_lock);
 517         tcf_lastuse_update(&p->tcf_tm);
 518         bstats_update(&p->tcf_bstats, skb);
 519         action = p->tcf_action;
 520         update_flags = p->update_flags;
 521         spin_unlock(&p->tcf_lock);
 522
 523         if (unlikely(action == TC_ACT_SHOT))
 524                 goto drop;
 525
 526         switch (tc_skb_protocol(skb)) {
 527         case cpu_to_be16(ETH_P_IP):
 528                 if (!tcf_csum_ipv4(skb, update_flags))
 529                         goto drop;
 530                 break;
 531         case cpu_to_be16(ETH_P_IPV6):
 532                 if (!tcf_csum_ipv6(skb, update_flags))
 533                         goto drop;
 534                 break;
 535         }
 536
 537         return action;
 538
 539 drop:
 540         spin_lock(&p->tcf_lock);
 541         p->tcf_qstats.drops++;
 542         spin_unlock(&p->tcf_lock);
 543         return TC_ACT_SHOT;
 544 }
 545
 546 static int tcf_csum_dump(struct sk_buff *skb, struct tc_action *a, int bind,
 547                          int ref)
 548 {
 549         unsigned char *b = skb_tail_pointer(skb);
 550         struct tcf_csum *p = to_tcf_csum(a);
 551         struct tc_csum opt = {
 552                 .update_flags = p->update_flags,
 553                 .index   = p->tcf_index,
 554                 .action  = p->tcf_action,
 555                 .refcnt  = p->tcf_refcnt - ref,
 556                 .bindcnt = p->tcf_bindcnt - bind,
 557         };
 558         struct tcf_t t;
 559
 560         if (nla_put(skb, TCA_CSUM_PARMS, sizeof(opt), &opt))
 561                 goto nla_put_failure;
 562
 563         tcf_tm_dump(&t, &p->tcf_tm);
 564         if (nla_put_64bit(skb, TCA_CSUM_TM, sizeof(t), &t, TCA_CSUM_PAD))
 565                 goto nla_put_failure;
 566
 567         return skb->len;
 568
 569 nla_put_failure:
 570         nlmsg_trim(skb, b);
 571         return -1;
 572 }
 573
 574 static int tcf_csum_walker(struct net *net, struct sk_buff *skb,
 575                            struct netlink_callback *cb, int type,
 576                            const struct tc_action_ops *ops)
 577 {
 578         struct tc_action_net *tn = net_generic(net, csum_net_id);
 579
 580         return tcf_generic_walker(tn, skb, cb, type, ops);
 581 }
 582
 583 static int tcf_csum_search(struct net *net, struct tc_action **a, u32 index)
 584 {
 585         struct tc_action_net *tn = net_generic(net, csum_net_id);
 586
 587         return tcf_hash_search(tn, a, index);
 588 }
 589
 590 static struct tc_action_ops act_csum_ops = {
 591         .kind           = "csum",
 592         .type           = TCA_ACT_CSUM,
 593         .owner          = THIS_MODULE,
 594         .act            = tcf_csum,
 595         .dump           = tcf_csum_dump,
 596         .init           = tcf_csum_init,
 597         .walk           = tcf_csum_walker,
 598         .lookup         = tcf_csum_search,
 599         .size           = sizeof(struct tcf_csum),
 600 };
 601
 602 static __net_init int csum_init_net(struct net *net)
 603 {
 604         struct tc_action_net *tn = net_generic(net, csum_net_id);
 605
 606         return tc_action_net_init(tn, &act_csum_ops, CSUM_TAB_MASK);
 607 }
 608
 609 static void __net_exit csum_exit_net(struct net *net)
 610 {
 611         struct tc_action_net *tn = net_generic(net, csum_net_id);
 612
 613         tc_action_net_exit(tn);
 614 }
 615
 616 static struct pernet_operations csum_net_ops = {
 617         .init = csum_init_net,
 618         .exit = csum_exit_net,
 619         .id   = &csum_net_id,
 620         .size = sizeof(struct tc_action_net),
 621 };
 622
 623 MODULE_DESCRIPTION("Checksum updating actions");
 624 MODULE_LICENSE("GPL");
 625
 626 static int __init csum_init_module(void)
 627 {
 628         return tcf_register_action(&act_csum_ops, &csum_net_ops);
 629 }
 630
 631 static void __exit csum_cleanup_module(void)
 632 {
 633         tcf_unregister_action(&act_csum_ops, &csum_net_ops);
 634 }
 635
 636 module_init(csum_init_module);
 637 module_exit(csum_cleanup_module);