![[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 *
99 *
100 *
101 * To Fix:
102 * IP fragmentation wants rewriting cleanly. The RFC815 algorithm is much more efficient
103 * and could be made very efficient with the addition of some virtual memory hacks to permit
104 * the allocation of a buffer that can then be 'grown' by twiddling page tables.
105 * Output fragmentation wants updating along with the buffer management to use a single
106 * interleaved copy algorithm so that fragmenting has a one copy overhead. Actual packet
107 * output should probably do its own fragmentation at the UDP/RAW layer. TCP shouldn't cause
108 * fragmentation anyway.
109 *
110 * FIXME: copy frag 0 iph to qp->iph
111 *
112 * This program is free software; you can redistribute it and/or
113 * modify it under the terms of the GNU General Public License
114 * as published by the Free Software Foundation; either version
115 * 2 of the License, or (at your option) any later version.
116 */
117
118 #include <asm/segment.h>
119 #include <asm/system.h>
120 #include <linux/types.h>
121 #include <linux/kernel.h>
122 #include <linux/sched.h>
123 #include <linux/mm.h>
124 #include <linux/string.h>
125 #include <linux/errno.h>
126 #include <linux/config.h>
127
128 #include <linux/socket.h>
129 #include <linux/sockios.h>
130 #include <linux/in.h>
131 #include <linux/inet.h>
132 #include <linux/netdevice.h>
133 #include <linux/etherdevice.h>
134 #include <linux/proc_fs.h>
135 #include <linux/stat.h>
136
137 #include <net/snmp.h>
138 #include <net/ip.h>
139 #include <net/protocol.h>
140 #include <net/route.h>
141 #include <net/tcp.h>
142 #include <net/udp.h>
143 #include <linux/skbuff.h>
144 #include <net/sock.h>
145 #include <net/arp.h>
146 #include <net/icmp.h>
147 #include <net/raw.h>
148 #include <net/checksum.h>
149 #include <linux/igmp.h>
150 #include <linux/ip_fw.h>
151 #include <linux/firewall.h>
152 #include <linux/mroute.h>
153 #include <net/netlink.h>
154
155 extern int last_retran;
156 extern void sort_send(struct sock *sk);
157
158 #define min(a,b) ((a)<(b)?(a):(b))
159
160 /*
161 * SNMP management statistics
162 */
163
164 #ifdef CONFIG_IP_FORWARD
165 struct ip_mib ip_statistics={1,64,}; /* Forwarding=Yes, Default TTL=64 */
166 #else
167 struct ip_mib ip_statistics={2,64,}; /* Forwarding=No, Default TTL=64 */
168 #endif
169
170 /*
171 * Handle the issuing of an ioctl() request
172 * for the ip device. This is scheduled to
173 * disappear
174 */
175
176 int ip_ioctl(struct sock *sk, int cmd, unsigned long arg)
/* ![[next]](../icons/right.png)
![[first]](../icons/n_first.png)
![[top]](../icons/top.png)
*/
177 {
178 switch(cmd)
179 {
180 default:
181 return(-EINVAL);
182 }
183 }
184
185
186
187 /*
188 * This function receives all incoming IP datagrams.
189 *
190 * On entry skb->data points to the start of the IP header and
191 * the MAC header has been removed.
192 */
193
194 int ip_rcv(struct sk_buff *skb, struct device *dev, struct packet_type *pt)
/* ![[previous]](../icons/left.png)
![[last]](../icons/n_last.png)
*/
195 {
196 struct iphdr *iph = skb->h.iph;
197 struct sock *raw_sk=NULL;
198 unsigned char hash;
199 unsigned char flag = 0;
200 struct inet_protocol *ipprot;
201 int brd=IS_MYADDR;
202 struct options * opt = NULL;
203 int is_frag=0;
204 #ifdef CONFIG_FIREWALL
205 int err;
206 #endif
207 #ifdef CONFIG_IP_MROUTE
208 int mroute_pkt=0;
209 #endif
210
211 #ifdef CONFIG_NET_IPV6
212 /*
213 * Intercept IPv6 frames. We dump ST-II and invalid types just below..
214 */
215
216 if(iph->version == 6)
217 return ipv6_rcv(skb,dev,pt);
218 #endif
219
220 ip_statistics.IpInReceives++;
221
222 /*
223 * Tag the ip header of this packet so we can find it
224 */
225
226 skb->ip_hdr = iph;
227
228 /*
229 * RFC1122: 3.1.2.2 MUST silently discard any IP frame that fails the checksum.
230 * RFC1122: 3.1.2.3 MUST discard a frame with invalid source address [NEEDS FIXING].
231 *
232 * Is the datagram acceptable?
233 *
234 * 1. Length at least the size of an ip header
235 * 2. Version of 4
236 * 3. Checksums correctly. [Speed optimisation for later, skip loopback checksums]
237 * 4. Doesn't have a bogus length
238 * (5. We ought to check for IP multicast addresses and undefined types.. does this matter ?)
239 */
240
241 if (skb->len<sizeof(struct iphdr) || iph->ihl<5 || iph->version != 4 || ip_fast_csum((unsigned char *)iph, iph->ihl) !=0
242 || skb->len < ntohs(iph->tot_len))
243 {
244 ip_statistics.IpInHdrErrors++;
245 kfree_skb(skb, FREE_WRITE);
246 return(0);
247 }
248
249 /*
250 * Our transport medium may have padded the buffer out. Now we know it
251 * is IP we can trim to the true length of the frame.
252 * Note this now means skb->len holds ntohs(iph->tot_len).
253 */
254
255 skb_trim(skb,ntohs(iph->tot_len));
256
257 if (iph->ihl > 5)
258 {
259 skb->ip_summed = 0;
260 if (ip_options_compile(NULL, skb))
261 return(0);
262 opt = (struct options*)skb->proto_priv;
263 #ifdef CONFIG_IP_NOSR
264 if (opt->srr)
265 {
266 kfree_skb(skb, FREE_READ);
267 return -EINVAL;
268 }
269 #endif
270 }
271
272 /*
273 * See if the firewall wants to dispose of the packet.
274 */
275
276 #ifdef CONFIG_FIREWALL
277
278 if ((err=call_in_firewall(PF_INET, skb, iph))<FW_ACCEPT)
279 {
280 if(err==FW_REJECT)
281 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0, dev);
282 kfree_skb(skb, FREE_WRITE);
283 return 0;
284 }
285
286 #endif
287
288 /*
289 * Remember if the frame is fragmented.
290 */
291
292 if(iph->frag_off)
293 {
294 if (iph->frag_off & htons(IP_MF))
295 is_frag|=1;
296 /*
297 * Last fragment ?
298 */
299
300 if (iph->frag_off & htons(IP_OFFSET))
301 is_frag|=2;
302 }
303
304 /*
305 * Do any IP forwarding required. chk_addr() is expensive -- avoid it someday.
306 *
307 * This is inefficient. While finding out if it is for us we could also compute
308 * the routing table entry. This is where the great unified cache theory comes
309 * in as and when someone implements it
310 *
311 * For most hosts over 99% of packets match the first conditional
312 * and don't go via ip_chk_addr. Note: brd is set to IS_MYADDR at
313 * function entry.
314 */
315
316 if ( iph->daddr == skb->dev->pa_addr || (brd = ip_chk_addr(iph->daddr)) != 0)
317 {
318 if (opt && opt->srr)
319 {
320 int srrspace, srrptr;
321 __u32 nexthop;
322 unsigned char * optptr = ((unsigned char *)iph) + opt->srr;
323
324 if (brd != IS_MYADDR || skb->pkt_type != PACKET_HOST)
325 {
326 kfree_skb(skb, FREE_WRITE);
327 return 0;
328 }
329
330 for ( srrptr=optptr[2], srrspace = optptr[1];
331 srrptr <= srrspace;
332 srrptr += 4
333 )
334 {
335 int brd2;
336 if (srrptr + 3 > srrspace)
337 {
338 icmp_send(skb, ICMP_PARAMETERPROB, 0, opt->srr+2,
339 skb->dev);
340 kfree_skb(skb, FREE_WRITE);
341 return 0;
342 }
343 memcpy(&nexthop, &optptr[srrptr-1], 4);
344 if ((brd2 = ip_chk_addr(nexthop)) == 0)
345 break;
346 if (brd2 != IS_MYADDR)
347 {
348
349 /*
350 * ANK: should we implement weak tunneling of multicasts?
351 * Are they obsolete? DVMRP specs (RFC-1075) is old enough...
352 * [They are obsolete]
353 */
354 kfree_skb(skb, FREE_WRITE);
355 return -EINVAL;
356 }
357 }
358 if (srrptr <= srrspace)
359 {
360 opt->srr_is_hit = 1;
361 opt->is_changed = 1;
362 #ifdef CONFIG_IP_FORWARD
363 if (ip_forward(skb, dev, is_frag, nexthop))
364 kfree_skb(skb, FREE_WRITE);
365 #else
366 ip_statistics.IpInAddrErrors++;
367 kfree_skb(skb, FREE_WRITE);
368 #endif
369 return 0;
370 }
371 }
372
373 #ifdef CONFIG_IP_MULTICAST
374 if(!(dev->flags&IFF_ALLMULTI) && brd==IS_MULTICAST && iph->daddr!=IGMP_ALL_HOSTS && !(dev->flags&IFF_LOOPBACK))
375 {
376 /*
377 * Check it is for one of our groups
378 */
379 struct ip_mc_list *ip_mc=dev->ip_mc_list;
380 do
381 {
382 if(ip_mc==NULL)
383 {
384 kfree_skb(skb, FREE_WRITE);
385 return 0;
386 }
387 if(ip_mc->multiaddr==iph->daddr)
388 break;
389 ip_mc=ip_mc->next;
390 }
391 while(1);
392 }
393 #endif
394
395 #ifdef CONFIG_IP_MASQUERADE
396 /*
397 * Do we need to de-masquerade this fragment?
398 */
399 if (ip_fw_demasquerade(skb))
400 {
401 struct iphdr *iph=skb->h.iph;
402 if (ip_forward(skb, dev, is_frag|4, iph->daddr))
403 kfree_skb(skb, FREE_WRITE);
404 return(0);
405 }
406 #endif
407
408 /*
409 * Account for the packet
410 */
411
412 #ifdef CONFIG_IP_ACCT
413 ip_fw_chk(iph,dev,ip_acct_chain,IP_FW_F_ACCEPT,1);
414 #endif
415
416 /*
417 * Reassemble IP fragments.
418 */
419
420 if(is_frag)
421 {
422 /* Defragment. Obtain the complete packet if there is one */
423 skb=ip_defrag(iph,skb,dev);
424 if(skb==NULL)
425 return 0;
426 skb->dev = dev;
427 iph=skb->h.iph;
428 }
429
430 /*
431 * Point into the IP datagram, just past the header.
432 */
433
434 skb->ip_hdr = iph;
435 skb->h.raw += iph->ihl*4;
436
437 #ifdef CONFIG_IP_MROUTE
438 /*
439 * Check the state on multicast routing (multicast and not 224.0.0.z)
440 */
441
442 if(brd==IS_MULTICAST && (iph->daddr&htonl(0xFFFFFF00))!=htonl(0xE0000000))
443 mroute_pkt=1;
444
445 #endif
446 /*
447 * Deliver to raw sockets. This is fun as to avoid copies we want to make no surplus copies.
448 *
449 * RFC 1122: SHOULD pass TOS value up to the transport layer.
450 */
451
452 hash = iph->protocol & (SOCK_ARRAY_SIZE-1);
453
454 /*
455 * If there maybe a raw socket we must check - if not we don't care less
456 */
457
458 if((raw_sk=raw_prot.sock_array[hash])!=NULL)
459 {
460 struct sock *sknext=NULL;
461 struct sk_buff *skb1;
462 raw_sk=get_sock_raw(raw_sk, iph->protocol, iph->saddr, iph->daddr);
463 if(raw_sk) /* Any raw sockets */
464 {
465 do
466 {
467 /* Find the next */
468 sknext=get_sock_raw(raw_sk->next, iph->protocol, iph->saddr, iph->daddr);
469 if(sknext)
470 skb1=skb_clone(skb, GFP_ATOMIC);
471 else
472 break; /* One pending raw socket left */
473 if(skb1)
474 raw_rcv(raw_sk, skb1, dev, iph->saddr,iph->daddr);
475 raw_sk=sknext;
476 }
477 while(raw_sk!=NULL);
478
479 /*
480 * Here either raw_sk is the last raw socket, or NULL if none
481 */
482
483 /*
484 * We deliver to the last raw socket AFTER the protocol checks as it avoids a surplus copy
485 */
486 }
487 }
488
489 /*
490 * skb->h.raw now points at the protocol beyond the IP header.
491 */
492
493 hash = iph->protocol & (MAX_INET_PROTOS -1);
494 for (ipprot = (struct inet_protocol *)inet_protos[hash];ipprot != NULL;ipprot=(struct inet_protocol *)ipprot->next)
495 {
496 struct sk_buff *skb2;
497
498 if (ipprot->protocol != iph->protocol)
499 continue;
500 /*
501 * See if we need to make a copy of it. This will
502 * only be set if more than one protocol wants it.
503 * and then not for the last one. If there is a pending
504 * raw delivery wait for that
505 */
506
507 #ifdef CONFIG_IP_MROUTE
508 if (ipprot->copy || raw_sk || mroute_pkt)
509 #else
510 if (ipprot->copy || raw_sk)
511 #endif
512 {
513 skb2 = skb_clone(skb, GFP_ATOMIC);
514 if(skb2==NULL)
515 continue;
516 }
517 else
518 {
519 skb2 = skb;
520 }
521 flag = 1;
522
523 /*
524 * Pass on the datagram to each protocol that wants it,
525 * based on the datagram protocol. We should really
526 * check the protocol handler's return values here...
527 */
528
529 ipprot->handler(skb2, dev, opt, iph->daddr,
530 (ntohs(iph->tot_len) - (iph->ihl * 4)),
531 iph->saddr, 0, ipprot);
532 }
533
534 /*
535 * All protocols checked.
536 * If this packet was a broadcast, we may *not* reply to it, since that
537 * causes (proven, grin) ARP storms and a leakage of memory (i.e. all
538 * ICMP reply messages get queued up for transmission...)
539 */
540
541 #ifdef CONFIG_IP_MROUTE
542 /*
543 * Forward the last copy to the multicast router. If
544 * there is a pending raw deliery however make a copy
545 * and forward that.
546 */
547
548 if(mroute_pkt)
549 {
550 flag=1;
551 if(raw_sk==NULL)
552 ipmr_forward(skb, is_frag);
553 else
554 {
555 struct sk_buff *skb2=skb_clone(skb, GFP_ATOMIC);
556 if(skb2)
557 {
558 skb2->free=1;
559 ipmr_forward(skb2, is_frag);
560 }
561 }
562 }
563 #endif
564
565 if(raw_sk!=NULL) /* Shift to last raw user */
566 raw_rcv(raw_sk, skb, dev, iph->saddr, iph->daddr);
567 else if (!flag) /* Free and report errors */
568 {
569 if (brd != IS_BROADCAST && brd!=IS_MULTICAST)
570 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PROT_UNREACH, 0, dev);
571 kfree_skb(skb, FREE_WRITE);
572 }
573
574 return(0);
575 }
576
577 /*
578 * Do any unicast IP forwarding required.
579 */
580
581 /*
582 * Don't forward multicast or broadcast frames.
583 */
584
585 if(skb->pkt_type!=PACKET_HOST || brd==IS_BROADCAST)
586 {
587 kfree_skb(skb,FREE_WRITE);
588 return 0;
589 }
590
591 /*
592 * The packet is for another target. Forward the frame
593 */
594
595 #ifdef CONFIG_IP_FORWARD
596 if (opt && opt->is_strictroute)
597 {
598 icmp_send(skb, ICMP_PARAMETERPROB, 0, 16, skb->dev);
599 kfree_skb(skb, FREE_WRITE);
600 return -1;
601 }
602 if (ip_forward(skb, dev, is_frag, iph->daddr))
603 kfree_skb(skb, FREE_WRITE);
604 #else
605 /* printk("Machine %lx tried to use us as a forwarder to %lx but we have forwarding disabled!\n",
606 iph->saddr,iph->daddr);*/
607 ip_statistics.IpInAddrErrors++;
608 kfree_skb(skb, FREE_WRITE);
609 #endif
610 return(0);
611 }
612
613
![[bottom]](../icons/n_bottom.png){kind=icon}
*/