release/src-rt-6.x.4708/linux/linux-2.6.36/net/ipv4/netfilter/nf_nat_helper.c

   1 /* ip_nat_helper.c - generic support functions for NAT helpers
   2  *
   3  * (C) 2000-2002 Harald Welte <laforge@netfilter.org>
   4  * (C) 2003-2006 Netfilter Core Team <coreteam@netfilter.org>
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 as
   8  * published by the Free Software Foundation.
   9  */
  10 #include <linux/module.h>
  11 #include <linux/gfp.h>
  12 #include <linux/kmod.h>
  13 #include <linux/types.h>
  14 #include <linux/timer.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/tcp.h>
  17 #include <linux/udp.h>
  18 #include <net/checksum.h>
  19 #include <net/tcp.h>
  20 #include <net/route.h>
  21
  22 #include <linux/netfilter_ipv4.h>
  23 #include <net/netfilter/nf_conntrack.h>
  24 #include <net/netfilter/nf_conntrack_helper.h>
  25 #include <net/netfilter/nf_conntrack_ecache.h>
  26 #include <net/netfilter/nf_conntrack_expect.h>
  27 #include <net/netfilter/nf_nat.h>
  28 #include <net/netfilter/nf_nat_protocol.h>
  29 #include <net/netfilter/nf_nat_core.h>
  30 #include <net/netfilter/nf_nat_helper.h>
  31
  32 #define DUMP_OFFSET(x) \
  33         pr_debug("offset_before=%d, offset_after=%d, correction_pos=%u\n", \
  34                  x->offset_before, x->offset_after, x->correction_pos);
  35
  36 static DEFINE_SPINLOCK(nf_nat_seqofs_lock);
  37
  38 /* Setup TCP sequence correction given this change at this sequence */
  39 static inline void
  40 adjust_tcp_sequence(u32 seq,
  41                     int sizediff,
  42                     struct nf_conn *ct,
  43                     enum ip_conntrack_info ctinfo)
  44 {
  45         enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
  46         struct nf_conn_nat *nat = nfct_nat(ct);
  47         struct nf_nat_seq *this_way = &nat->seq[dir];
  48
  49         pr_debug("adjust_tcp_sequence: seq = %u, sizediff = %d\n",
  50                  seq, sizediff);
  51
  52         pr_debug("adjust_tcp_sequence: Seq_offset before: ");
  53         DUMP_OFFSET(this_way);
  54
  55         spin_lock_bh(&nf_nat_seqofs_lock);
  56
  57         /* SYN adjust. If it's uninitialized, or this is after last
  58          * correction, record it: we don't handle more than one
  59          * adjustment in the window, but do deal with common case of a
  60          * retransmit */
  61         if (this_way->offset_before == this_way->offset_after ||
  62             before(this_way->correction_pos, seq)) {
  63                 this_way->correction_pos = seq;
  64                 this_way->offset_before = this_way->offset_after;
  65                 this_way->offset_after += sizediff;
  66         }
  67         spin_unlock_bh(&nf_nat_seqofs_lock);
  68
  69         pr_debug("adjust_tcp_sequence: Seq_offset after: ");
  70         DUMP_OFFSET(this_way);
  71 }
  72
  73 /* Get the offset value, for conntrack */
  74 s16 nf_nat_get_offset(const struct nf_conn *ct,
  75                       enum ip_conntrack_dir dir,
  76                       u32 seq)
  77 {
  78         struct nf_conn_nat *nat = nfct_nat(ct);
  79         struct nf_nat_seq *this_way;
  80         s16 offset;
  81
  82         if (!nat)
  83                 return 0;
  84
  85         this_way = &nat->seq[dir];
  86         spin_lock_bh(&nf_nat_seqofs_lock);
  87         offset = after(seq, this_way->correction_pos)
  88                  ? this_way->offset_after : this_way->offset_before;
  89         spin_unlock_bh(&nf_nat_seqofs_lock);
  90
  91         return offset;
  92 }
  93 EXPORT_SYMBOL_GPL(nf_nat_get_offset);
  94
  95 /* Frobs data inside this packet, which is linear. */
  96 static void mangle_contents(struct sk_buff *skb,
  97                             unsigned int dataoff,
  98                             unsigned int match_offset,
  99                             unsigned int match_len,
 100                             const char *rep_buffer,
 101                             unsigned int rep_len)
 102 {
 103         unsigned char *data;
 104
 105         BUG_ON(skb_is_nonlinear(skb));
 106         data = skb_network_header(skb) + dataoff;
 107
 108         /* move post-replacement */
 109         memmove(data + match_offset + rep_len,
 110                 data + match_offset + match_len,
 111                 skb->tail - (skb->network_header + dataoff +
 112                              match_offset + match_len));
 113
 114         /* insert data from buffer */
 115         memcpy(data + match_offset, rep_buffer, rep_len);
 116
 117         /* update skb info */
 118         if (rep_len > match_len) {
 119                 pr_debug("nf_nat_mangle_packet: Extending packet by "
 120                          "%u from %u bytes\n", rep_len - match_len, skb->len);
 121                 skb_put(skb, rep_len - match_len);
 122         } else {
 123                 pr_debug("nf_nat_mangle_packet: Shrinking packet from "
 124                          "%u from %u bytes\n", match_len - rep_len, skb->len);
 125                 __skb_trim(skb, skb->len + rep_len - match_len);
 126         }
 127
 128         /* fix IP hdr checksum information */
 129         ip_hdr(skb)->tot_len = htons(skb->len);
 130         ip_send_check(ip_hdr(skb));
 131 }
 132
 133 /* Unusual, but possible case. */
 134 static int enlarge_skb(struct sk_buff *skb, unsigned int extra)
 135 {
 136         if (skb->len + extra > 65535)
 137                 return 0;
 138
 139         if (pskb_expand_head(skb, 0, extra - skb_tailroom(skb), GFP_ATOMIC))
 140                 return 0;
 141
 142         return 1;
 143 }
 144
 145 void nf_nat_set_seq_adjust(struct nf_conn *ct, enum ip_conntrack_info ctinfo,
 146                            __be32 seq, s16 off)
 147 {
 148         if (!off)
 149                 return;
 150         set_bit(IPS_SEQ_ADJUST_BIT, &ct->status);
 151         adjust_tcp_sequence(ntohl(seq), off, ct, ctinfo);
 152         nf_conntrack_event_cache(IPCT_NATSEQADJ, ct);
 153 }
 154 EXPORT_SYMBOL_GPL(nf_nat_set_seq_adjust);
 155
 156 int __nf_nat_mangle_tcp_packet(struct sk_buff *skb,
 157                                struct nf_conn *ct,
 158                                enum ip_conntrack_info ctinfo,
 159                                unsigned int match_offset,
 160                                unsigned int match_len,
 161                                const char *rep_buffer,
 162                                unsigned int rep_len, bool adjust)
 163 {
 164         struct rtable *rt = skb_rtable(skb);
 165         struct iphdr *iph;
 166         struct tcphdr *tcph;
 167         int oldlen, datalen;
 168
 169         if (!skb_make_writable(skb, skb->len))
 170                 return 0;
 171
 172         if (rep_len > match_len &&
 173             rep_len - match_len > skb_tailroom(skb) &&
 174             !enlarge_skb(skb, rep_len - match_len))
 175                 return 0;
 176
 177         SKB_LINEAR_ASSERT(skb);
 178
 179         iph = ip_hdr(skb);
 180         tcph = (void *)iph + iph->ihl*4;
 181
 182         oldlen = skb->len - iph->ihl*4;
 183         mangle_contents(skb, iph->ihl*4 + tcph->doff*4,
 184                         match_offset, match_len, rep_buffer, rep_len);
 185
 186         datalen = skb->len - iph->ihl*4;
 187         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 188                 if (!(rt->rt_flags & RTCF_LOCAL) &&
 189                     skb->dev->features & NETIF_F_V4_CSUM) {
 190                         skb->ip_summed = CHECKSUM_PARTIAL;
 191                         skb->csum_start = skb_headroom(skb) +
 192                                           skb_network_offset(skb) +
 193                                           iph->ihl * 4;
 194                         skb->csum_offset = offsetof(struct tcphdr, check);
 195                         tcph->check = ~tcp_v4_check(datalen,
 196                                                     iph->saddr, iph->daddr, 0);
 197                 } else {
 198                         tcph->check = 0;
 199                         tcph->check = tcp_v4_check(datalen,
 200                                                    iph->saddr, iph->daddr,
 201                                                    csum_partial(tcph,
 202                                                                 datalen, 0));
 203                 }
 204         } else
 205                 inet_proto_csum_replace2(&tcph->check, skb,
 206                                          htons(oldlen), htons(datalen), 1);
 207
 208         if (adjust && rep_len != match_len)
 209                 nf_nat_set_seq_adjust(ct, ctinfo, tcph->seq,
 210                                       (int)rep_len - (int)match_len);
 211
 212         return 1;
 213 }
 214 EXPORT_SYMBOL(__nf_nat_mangle_tcp_packet);
 215
 216 int
 217 nf_nat_mangle_udp_packet(struct sk_buff *skb,
 218                          struct nf_conn *ct,
 219                          enum ip_conntrack_info ctinfo,
 220                          unsigned int match_offset,
 221                          unsigned int match_len,
 222                          const char *rep_buffer,
 223                          unsigned int rep_len)
 224 {
 225         struct rtable *rt = skb_rtable(skb);
 226         struct iphdr *iph;
 227         struct udphdr *udph;
 228         int datalen, oldlen;
 229
 230         /* UDP helpers might accidentally mangle the wrong packet */
 231         iph = ip_hdr(skb);
 232         if (skb->len < iph->ihl*4 + sizeof(*udph) +
 233                                match_offset + match_len)
 234                 return 0;
 235
 236         if (!skb_make_writable(skb, skb->len))
 237                 return 0;
 238
 239         if (rep_len > match_len &&
 240             rep_len - match_len > skb_tailroom(skb) &&
 241             !enlarge_skb(skb, rep_len - match_len))
 242                 return 0;
 243
 244         iph = ip_hdr(skb);
 245         udph = (void *)iph + iph->ihl*4;
 246
 247         oldlen = skb->len - iph->ihl*4;
 248         mangle_contents(skb, iph->ihl*4 + sizeof(*udph),
 249                         match_offset, match_len, rep_buffer, rep_len);
 250
 251         /* update the length of the UDP packet */
 252         datalen = skb->len - iph->ihl*4;
 253         udph->len = htons(datalen);
 254
 255         /* fix udp checksum if udp checksum was previously calculated */
 256         if (!udph->check && skb->ip_summed != CHECKSUM_PARTIAL)
 257                 return 1;
 258
 259         if (skb->ip_summed != CHECKSUM_PARTIAL) {
 260                 if (!(rt->rt_flags & RTCF_LOCAL) &&
 261                     skb->dev->features & NETIF_F_V4_CSUM) {
 262                         skb->ip_summed = CHECKSUM_PARTIAL;
 263                         skb->csum_start = skb_headroom(skb) +
 264                                           skb_network_offset(skb) +
 265                                           iph->ihl * 4;
 266                         skb->csum_offset = offsetof(struct udphdr, check);
 267                         udph->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
 268                                                          datalen, IPPROTO_UDP,
 269                                                          0);
 270                 } else {
 271                         udph->check = 0;
 272                         udph->check = csum_tcpudp_magic(iph->saddr, iph->daddr,
 273                                                         datalen, IPPROTO_UDP,
 274                                                         csum_partial(udph,
 275                                                                      datalen, 0));
 276                         if (!udph->check)
 277                                 udph->check = CSUM_MANGLED_0;
 278                 }
 279         } else
 280                 inet_proto_csum_replace2(&udph->check, skb,
 281                                          htons(oldlen), htons(datalen), 1);
 282
 283         return 1;
 284 }
 285 EXPORT_SYMBOL(nf_nat_mangle_udp_packet);
 286
 287 /* Adjust one found SACK option including checksum correction */
 288 static void
 289 sack_adjust(struct sk_buff *skb,
 290             struct tcphdr *tcph,
 291             unsigned int sackoff,
 292             unsigned int sackend,
 293             struct nf_nat_seq *natseq)
 294 {
 295         while (sackoff < sackend) {
 296                 struct tcp_sack_block_wire *sack;
 297                 __be32 new_start_seq, new_end_seq;
 298
 299                 sack = (void *)skb->data + sackoff;
 300                 if (after(ntohl(sack->start_seq) - natseq->offset_before,
 301                           natseq->correction_pos))
 302                         new_start_seq = htonl(ntohl(sack->start_seq)
 303                                         - natseq->offset_after);
 304                 else
 305                         new_start_seq = htonl(ntohl(sack->start_seq)
 306                                         - natseq->offset_before);
 307
 308                 if (after(ntohl(sack->end_seq) - natseq->offset_before,
 309                           natseq->correction_pos))
 310                         new_end_seq = htonl(ntohl(sack->end_seq)
 311                                       - natseq->offset_after);
 312                 else
 313                         new_end_seq = htonl(ntohl(sack->end_seq)
 314                                       - natseq->offset_before);
 315
 316                 pr_debug("sack_adjust: start_seq: %d->%d, end_seq: %d->%d\n",
 317                          ntohl(sack->start_seq), new_start_seq,
 318                          ntohl(sack->end_seq), new_end_seq);
 319
 320                 inet_proto_csum_replace4(&tcph->check, skb,
 321                                          sack->start_seq, new_start_seq, 0);
 322                 inet_proto_csum_replace4(&tcph->check, skb,
 323                                          sack->end_seq, new_end_seq, 0);
 324                 sack->start_seq = new_start_seq;
 325                 sack->end_seq = new_end_seq;
 326                 sackoff += sizeof(*sack);
 327         }
 328 }
 329
 330 /* TCP SACK sequence number adjustment */
 331 static inline unsigned int
 332 nf_nat_sack_adjust(struct sk_buff *skb,
 333                    struct tcphdr *tcph,
 334                    struct nf_conn *ct,
 335                    enum ip_conntrack_info ctinfo)
 336 {
 337         unsigned int dir, optoff, optend;
 338         struct nf_conn_nat *nat = nfct_nat(ct);
 339
 340         optoff = ip_hdrlen(skb) + sizeof(struct tcphdr);
 341         optend = ip_hdrlen(skb) + tcph->doff * 4;
 342
 343         if (!skb_make_writable(skb, optend))
 344                 return 0;
 345
 346         dir = CTINFO2DIR(ctinfo);
 347
 348         while (optoff < optend) {
 349                 /* Usually: option, length. */
 350                 unsigned char *op = skb->data + optoff;
 351
 352                 switch (op[0]) {
 353                 case TCPOPT_EOL:
 354                         return 1;
 355                 case TCPOPT_NOP:
 356                         optoff++;
 357                         continue;
 358                 default:
 359                         /* no partial options */
 360                         if (optoff + 1 == optend ||
 361                             optoff + op[1] > optend ||
 362                             op[1] < 2)
 363                                 return 0;
 364                         if (op[0] == TCPOPT_SACK &&
 365                             op[1] >= 2+TCPOLEN_SACK_PERBLOCK &&
 366                             ((op[1] - 2) % TCPOLEN_SACK_PERBLOCK) == 0)
 367                                 sack_adjust(skb, tcph, optoff+2,
 368                                             optoff+op[1], &nat->seq[!dir]);
 369                         optoff += op[1];
 370                 }
 371         }
 372         return 1;
 373 }
 374
 375 /* TCP sequence number adjustment.  Returns 1 on success, 0 on failure */
 376 int
 377 nf_nat_seq_adjust(struct sk_buff *skb,
 378                   struct nf_conn *ct,
 379                   enum ip_conntrack_info ctinfo)
 380 {
 381         struct tcphdr *tcph;
 382         int dir;
 383         __be32 newseq, newack;
 384         s16 seqoff, ackoff;
 385         struct nf_conn_nat *nat = nfct_nat(ct);
 386         struct nf_nat_seq *this_way, *other_way;
 387
 388         dir = CTINFO2DIR(ctinfo);
 389
 390         this_way = &nat->seq[dir];
 391         other_way = &nat->seq[!dir];
 392
 393         if (!skb_make_writable(skb, ip_hdrlen(skb) + sizeof(*tcph)))
 394                 return 0;
 395
 396         tcph = (void *)skb->data + ip_hdrlen(skb);
 397         if (after(ntohl(tcph->seq), this_way->correction_pos))
 398                 seqoff = this_way->offset_after;
 399         else
 400                 seqoff = this_way->offset_before;
 401
 402         if (after(ntohl(tcph->ack_seq) - other_way->offset_before,
 403                   other_way->correction_pos))
 404                 ackoff = other_way->offset_after;
 405         else
 406                 ackoff = other_way->offset_before;
 407
 408         newseq = htonl(ntohl(tcph->seq) + seqoff);
 409         newack = htonl(ntohl(tcph->ack_seq) - ackoff);
 410
 411         inet_proto_csum_replace4(&tcph->check, skb, tcph->seq, newseq, 0);
 412         inet_proto_csum_replace4(&tcph->check, skb, tcph->ack_seq, newack, 0);
 413
 414         pr_debug("Adjusting sequence number from %u->%u, ack from %u->%u\n",
 415                  ntohl(tcph->seq), ntohl(newseq), ntohl(tcph->ack_seq),
 416                  ntohl(newack));
 417
 418         tcph->seq = newseq;
 419         tcph->ack_seq = newack;
 420
 421         return nf_nat_sack_adjust(skb, tcph, ct, ctinfo);
 422 }
 423
 424 /* Setup NAT on this expected conntrack so it follows master. */
 425 /* If we fail to get a free NAT slot, we'll get dropped on confirm */
 426 void nf_nat_follow_master(struct nf_conn *ct,
 427                           struct nf_conntrack_expect *exp)
 428 {
 429         struct nf_nat_range range;
 430
 431         /* This must be a fresh one. */
 432         BUG_ON(ct->status & IPS_NAT_DONE_MASK);
 433
 434         /* Change src to where master sends to */
 435         range.flags = IP_NAT_RANGE_MAP_IPS;
 436         range.min_ip = range.max_ip
 437                 = ct->master->tuplehash[!exp->dir].tuple.dst.u3.ip;
 438         nf_nat_setup_info(ct, &range, IP_NAT_MANIP_SRC);
 439
 440         /* For DST manip, map port here to where it's expected. */
 441         range.flags = (IP_NAT_RANGE_MAP_IPS | IP_NAT_RANGE_PROTO_SPECIFIED);
 442         range.min = range.max = exp->saved_proto;
 443         range.min_ip = range.max_ip
 444                 = ct->master->tuplehash[!exp->dir].tuple.src.u3.ip;
 445         nf_nat_setup_info(ct, &range, IP_NAT_MANIP_DST);
 446 }
 447 EXPORT_SYMBOL(nf_nat_follow_master);