![[first]](../icons/first.png){kind=icon}
![[previous]](../icons/n_left.png){kind=icon}
![[next]](../icons/n_right.png){kind=icon}
![[last]](../icons/last.png){kind=icon}
![[top]](../icons/n_top.png){kind=icon}
![[bottom]](../icons/bottom.png){kind=icon}
![[index]](../icons/index.png){kind=icon}
![[help]](../icons/help.png){kind=icon}
*/
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * The Internet Protocol (IP) module.
7 *
8 * Version: @(#)ip.c 1.0.16b 9/1/93
9 *
10 * Authors: Ross Biro, <bir7@leland.Stanford.Edu>
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Donald Becker, <becker@super.org>
13 * Alan Cox, <Alan.Cox@linux.org>
14 * Richard Underwood
15 * Stefan Becker, <stefanb@yello.ping.de>
16 * Jorge Cwik, <jorge@laser.satlink.net>
17 * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
18 *
19 *
20 * Fixes:
21 * Alan Cox : Commented a couple of minor bits of surplus code
22 * Alan Cox : Undefining IP_FORWARD doesn't include the code
23 * (just stops a compiler warning).
24 * Alan Cox : Frames with >=MAX_ROUTE record routes, strict routes or loose routes
25 * are junked rather than corrupting things.
26 * Alan Cox : Frames to bad broadcast subnets are dumped
27 * We used to process them non broadcast and
28 * boy could that cause havoc.
29 * Alan Cox : ip_forward sets the free flag on the
30 * new frame it queues. Still crap because
31 * it copies the frame but at least it
32 * doesn't eat memory too.
33 * Alan Cox : Generic queue code and memory fixes.
34 * Fred Van Kempen : IP fragment support (borrowed from NET2E)
35 * Gerhard Koerting: Forward fragmented frames correctly.
36 * Gerhard Koerting: Fixes to my fix of the above 8-).
37 * Gerhard Koerting: IP interface addressing fix.
38 * Linus Torvalds : More robustness checks
39 * Alan Cox : Even more checks: Still not as robust as it ought to be
40 * Alan Cox : Save IP header pointer for later
41 * Alan Cox : ip option setting
42 * Alan Cox : Use ip_tos/ip_ttl settings
43 * Alan Cox : Fragmentation bogosity removed
44 * (Thanks to Mark.Bush@prg.ox.ac.uk)
45 * Dmitry Gorodchanin : Send of a raw packet crash fix.
46 * Alan Cox : Silly ip bug when an overlength
47 * fragment turns up. Now frees the
48 * queue.
49 * Linus Torvalds/ : Memory leakage on fragmentation
50 * Alan Cox : handling.
51 * Gerhard Koerting: Forwarding uses IP priority hints
52 * Teemu Rantanen : Fragment problems.
53 * Alan Cox : General cleanup, comments and reformat
54 * Alan Cox : SNMP statistics
55 * Alan Cox : BSD address rule semantics. Also see
56 * UDP as there is a nasty checksum issue
57 * if you do things the wrong way.
58 * Alan Cox : Always defrag, moved IP_FORWARD to the config.in file
59 * Alan Cox : IP options adjust sk->priority.
60 * Pedro Roque : Fix mtu/length error in ip_forward.
61 * Alan Cox : Avoid ip_chk_addr when possible.
62 * Richard Underwood : IP multicasting.
63 * Alan Cox : Cleaned up multicast handlers.
64 * Alan Cox : RAW sockets demultiplex in the BSD style.
65 * Gunther Mayer : Fix the SNMP reporting typo
66 * Alan Cox : Always in group 224.0.0.1
67 * Pauline Middelink : Fast ip_checksum update when forwarding
68 * Masquerading support.
69 * Alan Cox : Multicast loopback error for 224.0.0.1
70 * Alan Cox : IP_MULTICAST_LOOP option.
71 * Alan Cox : Use notifiers.
72 * Bjorn Ekwall : Removed ip_csum (from slhc.c too)
73 * Bjorn Ekwall : Moved ip_fast_csum to ip.h (inline!)
74 * Stefan Becker : Send out ICMP HOST REDIRECT
75 * Arnt Gulbrandsen : ip_build_xmit
76 * Alan Cox : Per socket routing cache
77 * Alan Cox : Fixed routing cache, added header cache.
78 * Alan Cox : Loopback didnt work right in original ip_build_xmit - fixed it.
79 * Alan Cox : Only send ICMP_REDIRECT if src/dest are the same net.
80 * Alan Cox : Incoming IP option handling.
81 * Alan Cox : Set saddr on raw output frames as per BSD.
82 * Alan Cox : Stopped broadcast source route explosions.
83 * Alan Cox : Can disable source routing
84 * Takeshi Sone : Masquerading didn't work.
85 * Dave Bonn,Alan Cox : Faster IP forwarding whenever possible.
86 * Alan Cox : Memory leaks, tramples, misc debugging.
87 * Alan Cox : Fixed multicast (by popular demand 8))
88 * Alan Cox : Fixed forwarding (by even more popular demand 8))
89 * Alan Cox : Fixed SNMP statistics [I think]
90 * Gerhard Koerting : IP fragmentation forwarding fix
91 * Alan Cox : Device lock against page fault.
92 * Alan Cox : IP_HDRINCL facility.
93 * Werner Almesberger : Zero fragment bug
94 * Alan Cox : RAW IP frame length bug
95 * Alan Cox : Outgoing firewall on build_xmit
96 * A.N.Kuznetsov : IP_OPTIONS support throughout the kernel
97 * Alan Cox : Multicast routing hooks
98 * Jos Vos : Do accounting *before* call_in_firewall
99 *
100 *
101 *
102 * To Fix:
103 * IP fragmentation wants rewriting cleanly. The RFC815 algorithm is much more efficient
104 * and could be made very efficient with the addition of some virtual memory hacks to permit
105 * the allocation of a buffer that can then be 'grown' by twiddling page tables.
106 * Output fragmentation wants updating along with the buffer management to use a single
107 * interleaved copy algorithm so that fragmenting has a one copy overhead. Actual packet
108 * output should probably do its own fragmentation at the UDP/RAW layer. TCP shouldn't cause
109 * fragmentation anyway.
110 *
111 * FIXME: copy frag 0 iph to qp->iph
112 *
113 * This program is free software; you can redistribute it and/or
114 * modify it under the terms of the GNU General Public License
115 * as published by the Free Software Foundation; either version
116 * 2 of the License, or (at your option) any later version.
117 */
118
119 #include <asm/segment.h>
120 #include <asm/system.h>
121 #include <linux/types.h>
122 #include <linux/kernel.h>
123 #include <linux/sched.h>
124 #include <linux/mm.h>
125 #include <linux/string.h>
126 #include <linux/errno.h>
127 #include <linux/config.h>
128
129 #include <linux/socket.h>
130 #include <linux/sockios.h>
131 #include <linux/in.h>
132 #include <linux/inet.h>
133 #include <linux/netdevice.h>
134 #include <linux/etherdevice.h>
135 #include <linux/proc_fs.h>
136 #include <linux/stat.h>
137
138 #include <net/snmp.h>
139 #include <net/ip.h>
140 #include <net/protocol.h>
141 #include <net/route.h>
142 #include <net/tcp.h>
143 #include <net/udp.h>
144 #include <linux/skbuff.h>
145 #include <net/sock.h>
146 #include <net/arp.h>
147 #include <net/icmp.h>
148 #include <net/raw.h>
149 #include <net/checksum.h>
150 #include <linux/igmp.h>
151 #include <linux/ip_fw.h>
152 #ifdef CONFIG_IP_MASQUERADE
153 #include <net/ip_masq.h>
154 #endif
155 #include <linux/firewall.h>
156 #include <linux/mroute.h>
157 #include <net/netlink.h>
158 #ifdef CONFIG_NET_ALIAS
159 #include <linux/net_alias.h>
160 #endif
161
162 extern int last_retran;
163 extern void sort_send(struct sock *sk);
164
165 #define min(a,b) ((a)<(b)?(a):(b))
166
167 /*
168 * SNMP management statistics
169 */
170
171 #ifdef CONFIG_IP_FORWARD
172 struct ip_mib ip_statistics={1,64,}; /* Forwarding=Yes, Default TTL=64 */
173 #else
174 struct ip_mib ip_statistics={2,64,}; /* Forwarding=No, Default TTL=64 */
175 #endif
176
177 /*
178 * Handle the issuing of an ioctl() request
179 * for the ip device. This is scheduled to
180 * disappear
181 */
182
183 int ip_ioctl(struct sock *sk, int cmd, unsigned long arg)
/* ![[next]](../icons/right.png)
![[first]](../icons/n_first.png)
![[top]](../icons/top.png)
*/
184 {
185 switch(cmd)
186 {
187 default:
188 return(-EINVAL);
189 }
190 }
191
192
193
194 /*
195 * This function receives all incoming IP datagrams.
196 *
197 * On entry skb->data points to the start of the IP header and
198 * the MAC header has been removed.
199 */
200
201 int ip_rcv(struct sk_buff *skb, struct device *dev, struct packet_type *pt)
/* ![[previous]](../icons/left.png)
![[last]](../icons/n_last.png)
*/
202 {
203 struct iphdr *iph = skb->h.iph;
204 struct sock *raw_sk=NULL;
205 unsigned char hash;
206 unsigned char flag = 0;
207 struct inet_protocol *ipprot;
208 int brd=IS_MYADDR;
209 struct options * opt = NULL;
210 int is_frag=0;
211 __u32 daddr;
212
213 #ifdef CONFIG_FIREWALL
214 int err;
215 #endif
216 #ifdef CONFIG_IP_MROUTE
217 int mroute_pkt=0;
218 #endif
219
220 #ifdef CONFIG_NET_IPV6
221 /*
222 * Intercept IPv6 frames. We dump ST-II and invalid types just below..
223 */
224
225 if(iph->version == 6)
226 return ipv6_rcv(skb,dev,pt);
227 #endif
228
229 ip_statistics.IpInReceives++;
230
231 /*
232 * Tag the ip header of this packet so we can find it
233 */
234
235 skb->ip_hdr = iph;
236
237 /*
238 * RFC1122: 3.1.2.2 MUST silently discard any IP frame that fails the checksum.
239 * RFC1122: 3.1.2.3 MUST discard a frame with invalid source address [NEEDS FIXING].
240 *
241 * Is the datagram acceptable?
242 *
243 * 1. Length at least the size of an ip header
244 * 2. Version of 4
245 * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums]
246 * 4. Doesn't have a bogus length
247 * (5. We ought to check for IP multicast addresses and undefined types.. does this matter ?)
248 */
249
250 if (skb->len<sizeof(struct iphdr) || iph->ihl<5 || iph->version != 4 || ip_fast_csum((unsigned char *)iph, iph->ihl) !=0
251 || skb->len < ntohs(iph->tot_len))
252 {
253 ip_statistics.IpInHdrErrors++;
254 kfree_skb(skb, FREE_WRITE);
255 return(0);
256 }
257
258 /*
259 * Our transport medium may have padded the buffer out. Now we know it
260 * is IP we can trim to the true length of the frame.
261 * Note this now means skb->len holds ntohs(iph->tot_len).
262 */
263
264 skb_trim(skb,ntohs(iph->tot_len));
265
266 /*
267 * Try to select closest <src,dst> alias device, if any.
268 * net_alias_dev_rcv_sel32 returns main device if it
269 * fails to found other.
270 */
271
272 #ifdef CONFIG_NET_ALIAS
273 if (iph->daddr != skb->dev->pa_addr && net_alias_has(skb->dev))
274 skb->dev = dev = net_alias_dev_rcv_sel32(skb->dev, AF_INET, iph->saddr, iph->daddr);
275 #endif
276
277 if (iph->ihl > 5)
278 {
279 skb->ip_summed = 0;
280 if (ip_options_compile(NULL, skb))
281 return(0);
282 opt = (struct options*)skb->proto_priv;
283 #ifdef CONFIG_IP_NOSR
284 if (opt->srr)
285 {
286 kfree_skb(skb, FREE_READ);
287 return -EINVAL;
288 }
289 #endif
290 }
291
292 /*
293 * Account for the packet (even if the packet is
294 * not accepted by the firewall!).
295 */
296
297 #ifdef CONFIG_IP_ACCT
298 ip_fw_chk(iph,dev,ip_acct_chain,IP_FW_F_ACCEPT,1);
299 #endif
300
301 /*
302 * See if the firewall wants to dispose of the packet.
303 */
304
305 #ifdef CONFIG_FIREWALL
306
307 if ((err=call_in_firewall(PF_INET, skb->dev, iph))<FW_ACCEPT)
308 {
309 if(err==FW_REJECT)
310 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0, dev);
311 kfree_skb(skb, FREE_WRITE);
312 return 0;
313 }
314
315 #endif
316
317 /*
318 * Remember if the frame is fragmented.
319 */
320
321 if(iph->frag_off)
322 {
323 if (iph->frag_off & htons(IP_MF))
324 is_frag|=IPFWD_FRAGMENT;
325 /*
326 * Last fragment ?
327 */
328
329 if (iph->frag_off & htons(IP_OFFSET))
330 is_frag|=IPFWD_LASTFRAG;
331 }
332
333 /*
334 * Do any IP forwarding required. chk_addr() is expensive -- avoid it someday.
335 *
336 * This is inefficient. While finding out if it is for us we could also compute
337 * the routing table entry. This is where the great unified cache theory comes
338 * in as and when someone implements it
339 *
340 * For most hosts over 99% of packets match the first conditional
341 * and don't go via ip_chk_addr. Note: brd is set to IS_MYADDR at
342 * function entry.
343 */
344 daddr = iph->daddr;
345 if ( iph->daddr == skb->dev->pa_addr || (brd = ip_chk_addr(iph->daddr)) != 0)
346 {
347 if (opt && opt->srr)
348 {
349 int srrspace, srrptr;
350 __u32 nexthop;
351 unsigned char * optptr = ((unsigned char *)iph) + opt->srr;
352
353 if (brd != IS_MYADDR || skb->pkt_type != PACKET_HOST)
354 {
355 kfree_skb(skb, FREE_WRITE);
356 return 0;
357 }
358
359 for ( srrptr=optptr[2], srrspace = optptr[1];
360 srrptr <= srrspace;
361 srrptr += 4
362 )
363 {
364 int brd2;
365 if (srrptr + 3 > srrspace)
366 {
367 icmp_send(skb, ICMP_PARAMETERPROB, 0, opt->srr+2,
368 skb->dev);
369 kfree_skb(skb, FREE_WRITE);
370 return 0;
371 }
372 memcpy(&nexthop, &optptr[srrptr-1], 4);
373 if ((brd2 = ip_chk_addr(nexthop)) == 0)
374 break;
375 if (brd2 != IS_MYADDR)
376 {
377
378 /*
379 * ANK: should we implement weak tunneling of multicasts?
380 * Are they obsolete? DVMRP specs (RFC-1075) is old enough...
381 * [They are obsolete]
382 */
383 kfree_skb(skb, FREE_WRITE);
384 return -EINVAL;
385 }
386 memcpy(&daddr, &optptr[srrptr-1], 4);
387 }
388 if (srrptr <= srrspace)
389 {
390 opt->srr_is_hit = 1;
391 opt->is_changed = 1;
392 #ifdef CONFIG_IP_FORWARD
393 if (ip_forward(skb, dev, is_frag, nexthop))
394 kfree_skb(skb, FREE_WRITE);
395 #else
396 ip_statistics.IpInAddrErrors++;
397 kfree_skb(skb, FREE_WRITE);
398 #endif
399 return 0;
400 }
401 }
402
403 #ifdef CONFIG_IP_MULTICAST
404 if(!(dev->flags&IFF_ALLMULTI) && brd==IS_MULTICAST && iph->daddr!=IGMP_ALL_HOSTS && !(dev->flags&IFF_LOOPBACK))
405 {
406 /*
407 * Check it is for one of our groups
408 */
409 struct ip_mc_list *ip_mc=dev->ip_mc_list;
410 do
411 {
412 if(ip_mc==NULL)
413 {
414 kfree_skb(skb, FREE_WRITE);
415 return 0;
416 }
417 if(ip_mc->multiaddr==iph->daddr)
418 break;
419 ip_mc=ip_mc->next;
420 }
421 while(1);
422 }
423 #endif
424
425 #ifdef CONFIG_IP_MASQUERADE
426 /*
427 * Do we need to de-masquerade this fragment?
428 */
429 if (ip_fw_demasquerade(&skb,dev))
430 {
431 struct iphdr *iph=skb->h.iph;
432 if (ip_forward(skb, dev, is_frag|IPFWD_MASQUERADED, iph->daddr))
433 kfree_skb(skb, FREE_WRITE);
434 return(0);
435 }
436 #endif
437
438 /*
439 * Reassemble IP fragments.
440 */
441
442 if(is_frag)
443 {
444 /* Defragment. Obtain the complete packet if there is one */
445 skb=ip_defrag(iph,skb,dev);
446 if(skb==NULL)
447 return 0;
448 skb->dev = dev;
449 iph=skb->h.iph;
450 }
451
452 /*
453 * Point into the IP datagram, just past the header.
454 */
455
456 skb->ip_hdr = iph;
457 skb->h.raw += iph->ihl*4;
458
459 #ifdef CONFIG_IP_MROUTE
460 /*
461 * Check the state on multicast routing (multicast and not 224.0.0.z)
462 */
463
464 if(brd==IS_MULTICAST && (iph->daddr&htonl(0xFFFFFF00))!=htonl(0xE0000000))
465 mroute_pkt=1;
466
467 #endif
468 /*
469 * Deliver to raw sockets. This is fun as to avoid copies we want to make no surplus copies.
470 *
471 * RFC 1122: SHOULD pass TOS value up to the transport layer.
472 */
473
474 hash = iph->protocol & (SOCK_ARRAY_SIZE-1);
475
476 /*
477 * If there maybe a raw socket we must check - if not we don't care less
478 */
479
480 if((raw_sk=raw_prot.sock_array[hash])!=NULL)
481 {
482 struct sock *sknext=NULL;
483 struct sk_buff *skb1;
484 raw_sk=get_sock_raw(raw_sk, iph->protocol, iph->saddr, iph->daddr);
485 if(raw_sk) /* Any raw sockets */
486 {
487 do
488 {
489 /* Find the next */
490 sknext=get_sock_raw(raw_sk->next, iph->protocol, iph->saddr, iph->daddr);
491 if(sknext)
492 skb1=skb_clone(skb, GFP_ATOMIC);
493 else
494 break; /* One pending raw socket left */
495 if(skb1)
496 raw_rcv(raw_sk, skb1, dev, iph->saddr,daddr);
497 raw_sk=sknext;
498 }
499 while(raw_sk!=NULL);
500
501 /*
502 * Here either raw_sk is the last raw socket, or NULL if none
503 */
504
505 /*
506 * We deliver to the last raw socket AFTER the protocol checks as it avoids a surplus copy
507 */
508 }
509 }
510
511 /*
512 * skb->h.raw now points at the protocol beyond the IP header.
513 */
514
515 hash = iph->protocol & (MAX_INET_PROTOS -1);
516 for (ipprot = (struct inet_protocol *)inet_protos[hash];ipprot != NULL;ipprot=(struct inet_protocol *)ipprot->next)
517 {
518 struct sk_buff *skb2;
519
520 if (ipprot->protocol != iph->protocol)
521 continue;
522 /*
523 * See if we need to make a copy of it. This will
524 * only be set if more than one protocol wants it.
525 * and then not for the last one. If there is a pending
526 * raw delivery wait for that
527 */
528
529 #ifdef CONFIG_IP_MROUTE
530 if (ipprot->copy || raw_sk || mroute_pkt)
531 #else
532 if (ipprot->copy || raw_sk)
533 #endif
534 {
535 skb2 = skb_clone(skb, GFP_ATOMIC);
536 if(skb2==NULL)
537 continue;
538 }
539 else
540 {
541 skb2 = skb;
542 }
543 flag = 1;
544
545 /*
546 * Pass on the datagram to each protocol that wants it,
547 * based on the datagram protocol. We should really
548 * check the protocol handler's return values here...
549 */
550
551 ipprot->handler(skb2, dev, opt, daddr,
552 (ntohs(iph->tot_len) - (iph->ihl * 4)),
553 iph->saddr, 0, ipprot);
554 }
555
556 /*
557 * All protocols checked.
558 * If this packet was a broadcast, we may *not* reply to it, since that
559 * causes (proven, grin) ARP storms and a leakage of memory (i.e. all
560 * ICMP reply messages get queued up for transmission...)
561 */
562
563 #ifdef CONFIG_IP_MROUTE
564 /*
565 * Forward the last copy to the multicast router. If
566 * there is a pending raw deliery however make a copy
567 * and forward that.
568 */
569
570 if(mroute_pkt)
571 {
572 flag=1;
573 if(raw_sk==NULL)
574 ipmr_forward(skb, is_frag);
575 else
576 {
577 struct sk_buff *skb2=skb_clone(skb, GFP_ATOMIC);
578 if(skb2)
579 {
580 skb2->free=1;
581 ipmr_forward(skb2, is_frag);
582 }
583 }
584 }
585 #endif
586
587 if(raw_sk!=NULL) /* Shift to last raw user */
588 raw_rcv(raw_sk, skb, dev, iph->saddr, daddr);
589 else if (!flag) /* Free and report errors */
590 {
591 if (brd != IS_BROADCAST && brd!=IS_MULTICAST)
592 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PROT_UNREACH, 0, dev);
593 kfree_skb(skb, FREE_WRITE);
594 }
595
596 return(0);
597 }
598
599 /*
600 * Do any unicast IP forwarding required.
601 */
602
603 /*
604 * Don't forward multicast or broadcast frames.
605 */
606
607 if(skb->pkt_type!=PACKET_HOST || brd==IS_BROADCAST)
608 {
609 kfree_skb(skb,FREE_WRITE);
610 return 0;
611 }
612
613 /*
614 * The packet is for another target. Forward the frame
615 */
616
617 #ifdef CONFIG_IP_FORWARD
618 if (opt && opt->is_strictroute)
619 {
620 icmp_send(skb, ICMP_PARAMETERPROB, 0, 16, skb->dev);
621 kfree_skb(skb, FREE_WRITE);
622 return -1;
623 }
624 if (ip_forward(skb, dev, is_frag, iph->daddr))
625 kfree_skb(skb, FREE_WRITE);
626 #else
627 /* printk("Machine %lx tried to use us as a forwarder to %lx but we have forwarding disabled!\n",
628 iph->saddr,iph->daddr);*/
629 ip_statistics.IpInAddrErrors++;
630 kfree_skb(skb, FREE_WRITE);
631 #endif
632 return(0);
633 }
634
635
![[bottom]](../icons/n_bottom.png){kind=icon}
*/