net/ipv4/netfilter/ip_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) 2003-2004 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  *      14 Jan 2002 Harald Welte <laforge@gnumonks.org>:
  11  *              - add support for SACK adjustment
  12  *      14 Mar 2002 Harald Welte <laforge@gnumonks.org>:
  13  *              - merge SACK support into newnat API
  14  *      16 Aug 2002 Brian J. Murrell <netfilter@interlinx.bc.ca>:
  15  *              - make ip_nat_resize_packet more generic (TCP and UDP)
  16  *              - add ip_nat_mangle_udp_packet
  17  */
  18 #include <linux/config.h>
  19 #include <linux/module.h>
  20 #include <linux/kmod.h>
  21 #include <linux/types.h>
  22 #include <linux/timer.h>
  23 #include <linux/skbuff.h>
  24 #include <linux/netfilter_ipv4.h>
  25 #include <net/checksum.h>
  26 #include <net/icmp.h>
  27 #include <net/ip.h>
  28 #include <net/tcp.h>
  29 #include <net/udp.h>
  30
  31 #define ASSERT_READ_LOCK(x)
  32 #define ASSERT_WRITE_LOCK(x)
  33
  34 #include <linux/netfilter_ipv4/ip_conntrack.h>
  35 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
  36 #include <linux/netfilter_ipv4/ip_nat.h>
  37 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  38 #include <linux/netfilter_ipv4/ip_nat_core.h>
  39 #include <linux/netfilter_ipv4/ip_nat_helper.h>
  40 #include <linux/netfilter_ipv4/listhelp.h>
  41
  42 #if 0
  43 #define DEBUGP printk
  44 #define DUMP_OFFSET(x)  printk("offset_before=%d, offset_after=%d, correction_pos=%u\n", x->offset_before, x->offset_after, x->correction_pos);
  45 #else
  46 #define DEBUGP(format, args...)
  47 #define DUMP_OFFSET(x)
  48 #endif
  49
  50 static DEFINE_SPINLOCK(ip_nat_seqofs_lock);
  51
  52 /* Setup TCP sequence correction given this change at this sequence */
  53 static inline void
  54 adjust_tcp_sequence(u32 seq,
  55                     int sizediff,
  56                     struct ip_conntrack *ct,
  57                     enum ip_conntrack_info ctinfo)
  58 {
  59         int dir;
  60         struct ip_nat_seq *this_way, *other_way;
  61
  62         DEBUGP("ip_nat_resize_packet: old_size = %u, new_size = %u\n",
  63                 (*skb)->len, new_size);
  64
  65         dir = CTINFO2DIR(ctinfo);
  66
  67         this_way = &ct->nat.info.seq[dir];
  68         other_way = &ct->nat.info.seq[!dir];
  69
  70         DEBUGP("ip_nat_resize_packet: Seq_offset before: ");
  71         DUMP_OFFSET(this_way);
  72
  73         spin_lock_bh(&ip_nat_seqofs_lock);
  74
  75         /* SYN adjust. If it's uninitialized, or this is after last
  76          * correction, record it: we don't handle more than one
  77          * adjustment in the window, but do deal with common case of a
  78          * retransmit */
  79         if (this_way->offset_before == this_way->offset_after
  80             || before(this_way->correction_pos, seq)) {
  81                     this_way->correction_pos = seq;
  82                     this_way->offset_before = this_way->offset_after;
  83                     this_way->offset_after += sizediff;
  84         }
  85         spin_unlock_bh(&ip_nat_seqofs_lock);
  86
  87         DEBUGP("ip_nat_resize_packet: Seq_offset after: ");
  88         DUMP_OFFSET(this_way);
  89 }
  90
  91 /* Frobs data inside this packet, which is linear. */
  92 static void mangle_contents(struct sk_buff *skb,
  93                             unsigned int dataoff,
  94                             unsigned int match_offset,
  95                             unsigned int match_len,
  96                             const char *rep_buffer,
  97                             unsigned int rep_len)
  98 {
  99         unsigned char *data;
 100
 101         BUG_ON(skb_is_nonlinear(skb));
 102         data = (unsigned char *)skb->nh.iph + dataoff;
 103
 104         /* move post-replacement */
 105         memmove(data + match_offset + rep_len,
 106                 data + match_offset + match_len,
 107                 skb->tail - (data + match_offset + match_len));
 108
 109         /* insert data from buffer */
 110         memcpy(data + match_offset, rep_buffer, rep_len);
 111
 112         /* update skb info */
 113         if (rep_len > match_len) {
 114                 DEBUGP("ip_nat_mangle_packet: Extending packet by "
 115                         "%u from %u bytes\n", rep_len - match_len,
 116                        skb->len);
 117                 skb_put(skb, rep_len - match_len);
 118         } else {
 119                 DEBUGP("ip_nat_mangle_packet: Shrinking packet from "
 120                         "%u from %u bytes\n", match_len - rep_len,
 121                        skb->len);
 122                 __skb_trim(skb, skb->len + rep_len - match_len);
 123         }
 124
 125         /* fix IP hdr checksum information */
 126         skb->nh.iph->tot_len = htons(skb->len);
 127         ip_send_check(skb->nh.iph);
 128 }
 129
 130 /* Unusual, but possible case. */
 131 static int enlarge_skb(struct sk_buff **pskb, unsigned int extra)
 132 {
 133         struct sk_buff *nskb;
 134
 135         if ((*pskb)->len + extra > 65535)
 136                 return 0;
 137
 138         nskb = skb_copy_expand(*pskb, skb_headroom(*pskb), extra, GFP_ATOMIC);
 139         if (!nskb)
 140                 return 0;
 141
 142         /* Transfer socket to new skb. */
 143         if ((*pskb)->sk)
 144                 skb_set_owner_w(nskb, (*pskb)->sk);
 145         kfree_skb(*pskb);
 146         *pskb = nskb;
 147         return 1;
 148 }
 149
 150 /* Generic function for mangling variable-length address changes inside
 151  * NATed TCP connections (like the PORT XXX,XXX,XXX,XXX,XXX,XXX
 152  * command in FTP).
 153  *
 154  * Takes care about all the nasty sequence number changes, checksumming,
 155  * skb enlargement, ...
 156  *
 157  * */
 158 int
 159 ip_nat_mangle_tcp_packet(struct sk_buff **pskb,
 160                          struct ip_conntrack *ct,
 161                          enum ip_conntrack_info ctinfo,
 162                          unsigned int match_offset,
 163                          unsigned int match_len,
 164                          const char *rep_buffer,
 165                          unsigned int rep_len)
 166 {
 167         struct iphdr *iph;
 168         struct tcphdr *tcph;
 169         int datalen;
 170
 171         if (!skb_make_writable(pskb, (*pskb)->len))
 172                 return 0;
 173
 174         if (rep_len > match_len
 175             && rep_len - match_len > skb_tailroom(*pskb)
 176             && !enlarge_skb(pskb, rep_len - match_len))
 177                 return 0;
 178
 179         SKB_LINEAR_ASSERT(*pskb);
 180
 181         iph = (*pskb)->nh.iph;
 182         tcph = (void *)iph + iph->ihl*4;
 183
 184         mangle_contents(*pskb, iph->ihl*4 + tcph->doff*4,
 185                         match_offset, match_len, rep_buffer, rep_len);
 186
 187         datalen = (*pskb)->len - iph->ihl*4;
 188         tcph->check = 0;
 189         tcph->check = tcp_v4_check(tcph, datalen, iph->saddr, iph->daddr,
 190                                    csum_partial((char *)tcph, datalen, 0));
 191
 192         if (rep_len != match_len) {
 193                 set_bit(IPS_SEQ_ADJUST_BIT, &ct->status);
 194                 adjust_tcp_sequence(ntohl(tcph->seq),
 195                                     (int)rep_len - (int)match_len,
 196                                     ct, ctinfo);
 197                 /* Tell TCP window tracking about seq change */
 198                 ip_conntrack_tcp_update(*pskb, ct, CTINFO2DIR(ctinfo));
 199         }
 200         return 1;
 201 }
 202 EXPORT_SYMBOL(ip_nat_mangle_tcp_packet);
 203
 204 /* Generic function for mangling variable-length address changes inside
 205  * NATed UDP connections (like the CONNECT DATA XXXXX MESG XXXXX INDEX XXXXX
 206  * command in the Amanda protocol)
 207  *
 208  * Takes care about all the nasty sequence number changes, checksumming,
 209  * skb enlargement, ...
 210  *
 211  * XXX - This function could be merged with ip_nat_mangle_tcp_packet which
 212  *       should be fairly easy to do.
 213  */
 214 int
 215 ip_nat_mangle_udp_packet(struct sk_buff **pskb,
 216                          struct ip_conntrack *ct,
 217                          enum ip_conntrack_info ctinfo,
 218                          unsigned int match_offset,
 219                          unsigned int match_len,
 220                          const char *rep_buffer,
 221                          unsigned int rep_len)
 222 {
 223         struct iphdr *iph;
 224         struct udphdr *udph;
 225
 226         /* UDP helpers might accidentally mangle the wrong packet */
 227         iph = (*pskb)->nh.iph;
 228         if ((*pskb)->len < iph->ihl*4 + sizeof(*udph) +
 229                                match_offset + match_len)
 230                 return 0;
 231
 232         if (!skb_make_writable(pskb, (*pskb)->len))
 233                 return 0;
 234
 235         if (rep_len > match_len
 236             && rep_len - match_len > skb_tailroom(*pskb)
 237             && !enlarge_skb(pskb, rep_len - match_len))
 238                 return 0;
 239
 240         iph = (*pskb)->nh.iph;
 241         udph = (void *)iph + iph->ihl*4;
 242         mangle_contents(*pskb, iph->ihl*4 + sizeof(*udph),
 243                         match_offset, match_len, rep_buffer, rep_len);
 244
 245         /* update the length of the UDP packet */
 246         udph->len = htons((*pskb)->len - iph->ihl*4);
 247
 248         /* fix udp checksum if udp checksum was previously calculated */
 249         if (udph->check) {
 250                 int datalen = (*pskb)->len - iph->ihl * 4;
 251                 udph->check = 0;
 252                 udph->check = csum_tcpudp_magic(iph->saddr, iph->daddr,
 253                                                 datalen, IPPROTO_UDP,
 254                                                 csum_partial((char *)udph,
 255                                                              datalen, 0));
 256         }
 257
 258         return 1;
 259 }
 260 EXPORT_SYMBOL(ip_nat_mangle_udp_packet);
 261
 262 /* Adjust one found SACK option including checksum correction */
 263 static void
 264 sack_adjust(struct sk_buff *skb,
 265             struct tcphdr *tcph,
 266             unsigned int sackoff,
 267             unsigned int sackend,
 268             struct ip_nat_seq *natseq)
 269 {
 270         while (sackoff < sackend) {
 271                 struct tcp_sack_block *sack;
 272                 u_int32_t new_start_seq, new_end_seq;
 273
 274                 sack = (void *)skb->data + sackoff;
 275                 if (after(ntohl(sack->start_seq) - natseq->offset_before,
 276                           natseq->correction_pos))
 277                         new_start_seq = ntohl(sack->start_seq)
 278                                         - natseq->offset_after;
 279                 else
 280                         new_start_seq = ntohl(sack->start_seq)
 281                                         - natseq->offset_before;
 282                 new_start_seq = htonl(new_start_seq);
 283
 284                 if (after(ntohl(sack->end_seq) - natseq->offset_before,
 285                           natseq->correction_pos))
 286                         new_end_seq = ntohl(sack->end_seq)
 287                                       - natseq->offset_after;
 288                 else
 289                         new_end_seq = ntohl(sack->end_seq)
 290                                       - natseq->offset_before;
 291                 new_end_seq = htonl(new_end_seq);
 292
 293                 DEBUGP("sack_adjust: start_seq: %d->%d, end_seq: %d->%d\n",
 294                         ntohl(sack->start_seq), new_start_seq,
 295                         ntohl(sack->end_seq), new_end_seq);
 296
 297                 tcph->check =
 298                         ip_nat_cheat_check(~sack->start_seq, new_start_seq,
 299                                            ip_nat_cheat_check(~sack->end_seq,
 300                                                               new_end_seq,
 301                                                               tcph->check));
 302                 sack->start_seq = new_start_seq;
 303                 sack->end_seq = new_end_seq;
 304                 sackoff += sizeof(*sack);
 305         }
 306 }
 307
 308 /* TCP SACK sequence number adjustment */
 309 static inline unsigned int
 310 ip_nat_sack_adjust(struct sk_buff **pskb,
 311                    struct tcphdr *tcph,
 312                    struct ip_conntrack *ct,
 313                    enum ip_conntrack_info ctinfo)
 314 {
 315         unsigned int dir, optoff, optend;
 316
 317         optoff = (*pskb)->nh.iph->ihl*4 + sizeof(struct tcphdr);
 318         optend = (*pskb)->nh.iph->ihl*4 + tcph->doff*4;
 319
 320         if (!skb_make_writable(pskb, optend))
 321                 return 0;
 322
 323         dir = CTINFO2DIR(ctinfo);
 324
 325         while (optoff < optend) {
 326                 /* Usually: option, length. */
 327                 unsigned char *op = (*pskb)->data + optoff;
 328
 329                 switch (op[0]) {
 330                 case TCPOPT_EOL:
 331                         return 1;
 332                 case TCPOPT_NOP:
 333                         optoff++;
 334                         continue;
 335                 default:
 336                         /* no partial options */
 337                         if (optoff + 1 == optend
 338                             || optoff + op[1] > optend
 339                             || op[1] < 2)
 340                                 return 0;
 341                         if (op[0] == TCPOPT_SACK
 342                             && op[1] >= 2+TCPOLEN_SACK_PERBLOCK
 343                             && ((op[1] - 2) % TCPOLEN_SACK_PERBLOCK) == 0)
 344                                 sack_adjust(*pskb, tcph, optoff+2,
 345                                             optoff+op[1],
 346                                             &ct->nat.info.seq[!dir]);
 347                         optoff += op[1];
 348                 }
 349         }
 350         return 1;
 351 }
 352
 353 /* TCP sequence number adjustment.  Returns 1 on success, 0 on failure */
 354 int
 355 ip_nat_seq_adjust(struct sk_buff **pskb,
 356                   struct ip_conntrack *ct,
 357                   enum ip_conntrack_info ctinfo)
 358 {
 359         struct tcphdr *tcph;
 360         int dir, newseq, newack;
 361         struct ip_nat_seq *this_way, *other_way;
 362
 363         dir = CTINFO2DIR(ctinfo);
 364
 365         this_way = &ct->nat.info.seq[dir];
 366         other_way = &ct->nat.info.seq[!dir];
 367
 368         if (!skb_make_writable(pskb, (*pskb)->nh.iph->ihl*4+sizeof(*tcph)))
 369                 return 0;
 370
 371         tcph = (void *)(*pskb)->data + (*pskb)->nh.iph->ihl*4;
 372         if (after(ntohl(tcph->seq), this_way->correction_pos))
 373                 newseq = ntohl(tcph->seq) + this_way->offset_after;
 374         else
 375                 newseq = ntohl(tcph->seq) + this_way->offset_before;
 376         newseq = htonl(newseq);
 377
 378         if (after(ntohl(tcph->ack_seq) - other_way->offset_before,
 379                   other_way->correction_pos))
 380                 newack = ntohl(tcph->ack_seq) - other_way->offset_after;
 381         else
 382                 newack = ntohl(tcph->ack_seq) - other_way->offset_before;
 383         newack = htonl(newack);
 384
 385         tcph->check = ip_nat_cheat_check(~tcph->seq, newseq,
 386                                          ip_nat_cheat_check(~tcph->ack_seq,
 387                                                             newack,
 388                                                             tcph->check));
 389
 390         DEBUGP("Adjusting sequence number from %u->%u, ack from %u->%u\n",
 391                 ntohl(tcph->seq), ntohl(newseq), ntohl(tcph->ack_seq),
 392                 ntohl(newack));
 393
 394         tcph->seq = newseq;
 395         tcph->ack_seq = newack;
 396
 397         if (!ip_nat_sack_adjust(pskb, tcph, ct, ctinfo))
 398                 return 0;
 399
 400         ip_conntrack_tcp_update(*pskb, ct, dir);
 401
 402         return 1;
 403 }
 404 EXPORT_SYMBOL(ip_nat_seq_adjust);
 405
 406 /* Setup NAT on this expected conntrack so it follows master. */
 407 /* If we fail to get a free NAT slot, we'll get dropped on confirm */
 408 void ip_nat_follow_master(struct ip_conntrack *ct,
 409                           struct ip_conntrack_expect *exp)
 410 {
 411         struct ip_nat_range range;
 412
 413         /* This must be a fresh one. */
 414         BUG_ON(ct->status & IPS_NAT_DONE_MASK);
 415
 416         /* Change src to where master sends to */
 417         range.flags = IP_NAT_RANGE_MAP_IPS;
 418         range.min_ip = range.max_ip
 419                 = ct->master->tuplehash[!exp->dir].tuple.dst.ip;
 420         /* hook doesn't matter, but it has to do source manip */
 421         ip_nat_setup_info(ct, &range, NF_IP_POST_ROUTING);
 422
 423         /* For DST manip, map port here to where it's expected. */
 424         range.flags = (IP_NAT_RANGE_MAP_IPS | IP_NAT_RANGE_PROTO_SPECIFIED);
 425         range.min = range.max = exp->saved_proto;
 426         range.min_ip = range.max_ip
 427                 = ct->master->tuplehash[!exp->dir].tuple.src.ip;
 428         /* hook doesn't matter, but it has to do destination manip */
 429         ip_nat_setup_info(ct, &range, NF_IP_PRE_ROUTING);
 430 }
 431 EXPORT_SYMBOL(ip_nat_follow_master);