|
|
1.1 root 1: /* ip_input.c 6.1 83/08/16 */
2:
3: #include "../h/param.h"
4: #include "../h/systm.h"
5: #include "../h/mbuf.h"
6: #include "../h/domain.h"
7: #include "../h/protosw.h"
8: #include "../h/socket.h"
9: #include "../h/errno.h"
10: #include "../h/time.h"
11: #include "../h/kernel.h"
12:
13: #include "../net/if.h"
14: #include "../net/route.h"
15:
16: #include "../netinet/in.h"
17: #include "../netinet/in_pcb.h"
18: #include "../netinet/in_systm.h"
19: #include "../netinet/ip.h"
20: #include "../netinet/ip_var.h"
21: #include "../netinet/ip_icmp.h"
22: #include "../netinet/tcp.h"
23:
24: #ifdef NET_PERF
25: # define IPTRACE1 1
26: # define IPTRACE2 2
27: # define IPTRACE(X,Y) { if (ipprintfs & X) Y}
28: # define IPSTUDY(X) {X}
29: #else NET_PERF
30: # define IPTRACE(X,Y)
31: # define IPSTUDY(X)
32: #endif NET_PERF
33:
34:
35: int ipprintfs = 0;
36: u_char ip_protox[IPPROTO_MAX];
37: int ipqmaxlen = IFQ_MAXLEN;
38: struct ifnet *ifinet; /* first inet interface */
39:
40: /*
41: * Changes:
42: * fix mbuf deallocation when length is too short
43: * in ip_init(), change the search limit to end of domain table.
44: */
45: /*
46: * IP initialization: fill in IP protocol switch table.
47: * All protocols not implemented in kernel go to raw IP protocol handler.
48: */
49: ip_init()
50: {
51: register struct protosw *pr;
52: register int i;
53:
54: pr = pffindproto(PF_INET, IPPROTO_RAW);
55: if (pr == 0)
56: panic("ip_init");
57: for (i = 0; i < IPPROTO_MAX; i++)
58: ip_protox[i] = pr - inetsw;
59: for (pr = inetdomain.dom_protosw;
60: pr < inetdomain.dom_protoswNPROTOSW; pr++)
61: if (pr->pr_family == PF_INET &&
62: pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
63: ip_protox[pr->pr_protocol] = pr - inetsw;
64: ipq.next = ipq.prev = &ipq;
65: ip_id = time.tv_sec & 0xffff;
66: ipintrq.ifq_maxlen = ipqmaxlen;
67: ifinet = if_ifwithaf(AF_INET);
68: }
69:
70: u_char ipcksum = 1;
71: struct ip *ip_reass();
72: struct sockaddr_in ipaddr = { AF_INET };
73:
74: /*
75: * Ip input routine. Checksum and byte swap header. If fragmented
76: * try to reassamble. If complete and fragment queue exists, discard.
77: * Process options. Pass to next level.
78: */
79: ipintr()
80: {
81: register struct ip *ip;
82: register struct mbuf *m;
83: struct mbuf *m0;
84: register int i;
85: register struct ipq *fp;
86: int hlen, s;
87:
88: next:
89: /*
90: * Get next datagram off input queue and get IP header
91: * in first mbuf.
92: */
93: s = splimp();
94: IF_DEQUEUE(&ipintrq, m);
95: splx(s);
96: if (m == 0)
97: return;
98: IPTRACE(IPTRACE1, printf("IPintr: DEQUEUE\n");)
99:
100: if ((m->m_off > MMAXOFF || m->m_len < sizeof (struct ip)) &&
101: (m = m_pullup(m, sizeof (struct ip))) == 0) {
102: ipstat.ips_toosmall++;
103: goto next;
104: }
105: ip = mtod(m, struct ip *);
106: if ((hlen = ip->ip_hl << 2) > m->m_len) {
107: if ((m = m_pullup(m, hlen)) == 0) {
108: ipstat.ips_badhlen++;
109: goto next;
110: }
111: ip = mtod(m, struct ip *);
112: }
113: if (ipcksum)
114: if (ip->ip_sum = in_cksum(m, hlen)) {
115: ipstat.ips_badsum++;
116: goto bad;
117: }
118:
119: /*
120: * Convert fields to host representation.
121: */
122: ip->ip_len = ntohs((u_short)ip->ip_len);
123: if (ip->ip_len < hlen) {
124: ipstat.ips_badlen++;
125: goto bad;
126: }
127: ip->ip_id = ntohs(ip->ip_id);
128: ip->ip_off = ntohs((u_short)ip->ip_off);
129:
130: /*
131: * Check that the amount of data in the buffers
132: * is as at least much as the IP header would have us expect.
133: * Trim mbufs if longer than we expect.
134: * Drop packet if shorter than we expect.
135: */
136: i = -ip->ip_len;
137: m0 = m;
138: for (;;) {
139: i += m->m_len;
140: if (m->m_next == 0)
141: break;
142: m = m->m_next;
143: }
144: if (i != 0) {
145: if (i < 0) {
146: ipstat.ips_tooshort++;
147: m = m0;
148: goto bad;
149: }
150: if (i <= m->m_len)
151: m->m_len -= i;
152: else
153: m_adj(m0, -i);
154: }
155: m = m0;
156:
157: /*
158: * Process options and, if not destined for us,
159: * ship it on. ip_dooptions returns 1 when an
160: * error was detected (causing an icmp message
161: * to be sent).
162: */
163: if (hlen > sizeof (struct ip) && ip_dooptions(ip))
164: goto next;
165:
166: /*
167: * Fast check on the first internet
168: * interface in the list.
169: */
170: if (ifinet) {
171: struct sockaddr_in *sin;
172:
173: sin = (struct sockaddr_in *)&ifinet->if_addr;
174: if (sin->sin_addr.s_addr == ip->ip_dst.s_addr)
175: goto ours;
176: sin = (struct sockaddr_in *)&ifinet->if_broadaddr;
177: if ((ifinet->if_flags & IFF_BROADCAST) &&
178: sin->sin_addr.s_addr == ip->ip_dst.s_addr)
179: goto ours;
180: }
181: /* BEGIN GROT */
182: #include "nd.h"
183: #if NND > 0
184: /*
185: * Diskless machines don't initially know
186: * their address, so take packets from them
187: * if we're acting as a network disk server.
188: */
189: if (in_netof(ip->ip_dst) == INADDR_ANY &&
190: (in_netof(ip->ip_src) == INADDR_ANY &&
191: in_lnaof(ip->ip_src) != INADDR_ANY))
192: goto ours;
193: #endif
194: /* END GROT */
195: ipaddr.sin_addr = ip->ip_dst;
196: if (if_ifwithaddr((struct sockaddr *)&ipaddr) == 0) {
197: ip_forward(ip);
198: goto next;
199: }
200:
201: ours:
202: /*
203: * Look for queue of fragments
204: * of this datagram.
205: */
206: for (fp = ipq.next; fp != &ipq; fp = fp->next)
207: if (ip->ip_id == fp->ipq_id &&
208: ip->ip_src.s_addr == fp->ipq_src.s_addr &&
209: ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
210: ip->ip_p == fp->ipq_p)
211: goto found;
212: fp = 0;
213: found:
214:
215: /*
216: * Adjust ip_len to not reflect header,
217: * set ip_mff if more fragments are expected,
218: * convert offset of this to bytes.
219: */
220: ip->ip_len -= hlen;
221: ((struct ipasfrag *)ip)->ipf_mff = 0;
222: if (ip->ip_off & IP_MF)
223: ((struct ipasfrag *)ip)->ipf_mff = 1;
224: ip->ip_off <<= 3;
225:
226: /*
227: * If datagram marked as having more fragments
228: * or if this is not the first fragment,
229: * attempt reassembly; if it succeeds, proceed.
230: */
231: if (((struct ipasfrag *)ip)->ipf_mff || ip->ip_off) {
232: ip = ip_reass((struct ipasfrag *)ip, fp);
233: if (ip == 0)
234: goto next;
235: hlen = ip->ip_hl << 2;
236: m = dtom(ip);
237: } else
238: if (fp)
239: ip_freef(fp);
240:
241: /*
242: * Switch out to protocol's input routine.
243: */
244: IPTRACE(IPTRACE2, printf("Proto #%d\n",ip->ip_p);)
245: (*inetsw[ip_protox[ip->ip_p]].pr_input)(m);
246: goto next;
247: bad:
248: m_freem(m);
249: goto next;
250: }
251:
252: /*
253: * Take incoming datagram fragment and try to
254: * reassemble it into whole datagram. If a chain for
255: * reassembly of this datagram already exists, then it
256: * is given as fp; otherwise have to make a chain.
257: */
258: struct ip *
259: ip_reass(ip, fp)
260: register struct ipasfrag *ip;
261: register struct ipq *fp;
262: {
263: register struct mbuf *m = dtom(ip);
264: register struct ipasfrag *q;
265: struct mbuf *t;
266: int hlen = ip->ip_hl << 2;
267: int i, next;
268:
269: /*
270: * Presence of header sizes in mbufs
271: * would confuse code below.
272: */
273: m->m_off += hlen;
274: m->m_len -= hlen;
275:
276: /*
277: * If first fragment to arrive, create a reassembly queue.
278: */
279: if (fp == 0) {
280: t = m_get(M_WAIT, MT_FTABLE);
281: MBUFNULL(7, t); /* for debugging */
282: if (t == NULL)
283: goto dropfrag;
284: fp = mtod(t, struct ipq *);
285: insque(fp, &ipq);
286: fp->ipq_ttl = IPFRAGTTL;
287: fp->ipq_p = ip->ip_p;
288: fp->ipq_id = ip->ip_id;
289: fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
290: fp->ipq_src = ((struct ip *)ip)->ip_src;
291: fp->ipq_dst = ((struct ip *)ip)->ip_dst;
292: q = (struct ipasfrag *)fp;
293: goto insert;
294: }
295:
296: /*
297: * Find a segment which begins after this one does.
298: */
299: for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next)
300: if (q->ip_off > ip->ip_off)
301: break;
302:
303: /*
304: * If there is a preceding segment, it may provide some of
305: * our data already. If so, drop the data from the incoming
306: * segment. If it provides all of our data, drop us.
307: */
308: if (q->ipf_prev != (struct ipasfrag *)fp) {
309: i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off;
310: if (i > 0) {
311: if (i >= ip->ip_len)
312: goto dropfrag;
313: m_adj(dtom(ip), i);
314: ip->ip_off += i;
315: ip->ip_len -= i;
316: }
317: }
318:
319: /*
320: * While we overlap succeeding segments trim them or,
321: * if they are completely covered, dequeue them.
322: */
323: while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) {
324: i = (ip->ip_off + ip->ip_len) - q->ip_off;
325: if (i < q->ip_len) {
326: q->ip_len -= i;
327: q->ip_off += i;
328: m_adj(dtom(q), i);
329: break;
330: }
331: q = q->ipf_next;
332: m_freem(dtom(q->ipf_prev));
333: ip_deq(q->ipf_prev);
334: }
335:
336: insert:
337: /*
338: * Stick new segment in its place;
339: * check for complete reassembly.
340: */
341: ip_enq(ip, q->ipf_prev);
342: next = 0;
343: for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) {
344: if (q->ip_off != next)
345: return (0);
346: next += q->ip_len;
347: }
348: if (q->ipf_prev->ipf_mff)
349: return (0);
350:
351: /*
352: * Reassembly is complete; concatenate fragments.
353: */
354: q = fp->ipq_next;
355: m = dtom(q);
356: t = m->m_next;
357: m->m_next = 0;
358: m_cat(m, t);
359: q = q->ipf_next;
360: while (q != (struct ipasfrag *)fp) {
361: t = dtom(q);
362: q = q->ipf_next;
363: m_cat(m, t);
364: }
365:
366: /*
367: * Create header for new ip packet by
368: * modifying header of first packet;
369: * dequeue and discard fragment reassembly header.
370: * Make header visible.
371: */
372: ip = fp->ipq_next;
373: ip->ip_len = next;
374: ((struct ip *)ip)->ip_src = fp->ipq_src;
375: ((struct ip *)ip)->ip_dst = fp->ipq_dst;
376: remque(fp);
377: (void) m_free(dtom(fp));
378: m = dtom(ip);
379: m->m_len += sizeof (struct ipasfrag);
380: m->m_off -= sizeof (struct ipasfrag);
381: return ((struct ip *)ip);
382:
383: dropfrag:
384: m_freem(m);
385: return (0);
386: }
387:
388: /*
389: * Free a fragment reassembly header and all
390: * associated datagrams.
391: */
392: ip_freef(fp)
393: struct ipq *fp;
394: {
395: register struct ipasfrag *q, *p;
396:
397: for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) {
398: p = q->ipf_next;
399: ip_deq(q);
400: m_freem(dtom(q));
401: }
402: remque(fp);
403: (void) m_free(dtom(fp));
404: }
405:
406: /*
407: * Put an ip fragment on a reassembly chain.
408: * Like insque, but pointers in middle of structure.
409: */
410: ip_enq(p, prev)
411: register struct ipasfrag *p, *prev;
412: {
413:
414: p->ipf_prev = prev;
415: p->ipf_next = prev->ipf_next;
416: prev->ipf_next->ipf_prev = p;
417: prev->ipf_next = p;
418: }
419:
420: /*
421: * To ip_enq as remque is to insque.
422: */
423: ip_deq(p)
424: register struct ipasfrag *p;
425: {
426:
427: p->ipf_prev->ipf_next = p->ipf_next;
428: p->ipf_next->ipf_prev = p->ipf_prev;
429: }
430:
431: /*
432: * IP timer processing;
433: * if a timer expires on a reassembly
434: * queue, discard it.
435: */
436: ip_slowtimo()
437: {
438: register struct ipq *fp;
439: int s = splnet();
440:
441: fp = ipq.next;
442: if (fp == 0) {
443: splx(s);
444: return;
445: }
446: while (fp != &ipq) {
447: --fp->ipq_ttl;
448: fp = fp->next;
449: if (fp->prev->ipq_ttl == 0)
450: ip_freef(fp->prev);
451: }
452: splx(s);
453: }
454:
455: /*
456: * Drain off all datagram fragments.
457: */
458: ip_drain()
459: {
460:
461: while (ipq.next != &ipq)
462: ip_freef(ipq.next);
463: }
464:
465: /*
466: * Do option processing on a datagram,
467: * possibly discarding it if bad options
468: * are encountered.
469: */
470: ip_dooptions(ip)
471: struct ip *ip;
472: {
473: register u_char *cp;
474: int opt, optlen, cnt, code, type;
475: struct in_addr *sin;
476: register struct ip_timestamp *ipt;
477: register struct ifnet *ifp;
478: struct in_addr t;
479:
480: cp = (u_char *)(ip + 1);
481: cnt = (ip->ip_hl << 2) - sizeof (struct ip);
482: for (; cnt > 0; cnt -= optlen, cp += optlen) {
483: opt = cp[0];
484: if (opt == IPOPT_EOL)
485: break;
486: if (opt == IPOPT_NOP)
487: optlen = 1;
488: else
489: optlen = cp[1];
490: switch (opt) {
491:
492: default:
493: break;
494:
495: /*
496: * Source routing with record.
497: * Find interface with current destination address.
498: * If none on this machine then drop if strictly routed,
499: * or do nothing if loosely routed.
500: * Record interface address and bring up next address
501: * component. If strictly routed make sure next
502: * address on directly accessible net.
503: */
504: case IPOPT_LSRR:
505: case IPOPT_SSRR:
506: if (cp[2] < 4 || cp[2] > optlen - (sizeof (long) - 1))
507: break;
508: sin = (struct in_addr *)(cp + cp[2]);
509: ipaddr.sin_addr = *sin;
510: ifp = if_ifwithaddr((struct sockaddr *)&ipaddr);
511: type = ICMP_UNREACH, code = ICMP_UNREACH_SRCFAIL;
512: if (ifp == 0) {
513: if (opt == IPOPT_SSRR) {
514: IPSTUDY(ipstudy.ips_ssnoifaddr++;)
515: goto bad;
516: }
517: break;
518: }
519: t = ip->ip_dst; ip->ip_dst = *sin; *sin = t;
520: cp[2] += 4;
521: if (cp[2] > optlen - (sizeof (long) - 1))
522: break;
523: ip->ip_dst = sin[1];
524: if (opt == IPOPT_SSRR &&
525: if_ifonnetof(in_netof(ip->ip_dst)) == 0) {
526: IPSTUDY(ipstudy.ips_ssnoifnet++;)
527: goto bad;
528: }
529: break;
530:
531: case IPOPT_TS:
532: code = cp - (u_char *)ip;
533: type = ICMP_PARAMPROB;
534: ipt = (struct ip_timestamp *)cp;
535: if (ipt->ipt_len < 5) {
536: IPSTUDY(ipstudy.ips_oplen++;)
537: goto bad;
538: }
539: if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) {
540: if (++ipt->ipt_oflw == 0) {
541: IPSTUDY(ipstudy.ips_oplen++;)
542: goto bad;
543: }
544: break;
545: }
546: sin = (struct in_addr *)(cp+cp[2]);
547: switch (ipt->ipt_flg) {
548:
549: case IPOPT_TS_TSONLY:
550: break;
551:
552: case IPOPT_TS_TSANDADDR:
553: if (ipt->ipt_ptr + 8 > ipt->ipt_len) {
554: IPSTUDY(ipstudy.ips_oplen++;)
555: goto bad;
556: }
557: if (ifinet == 0) {
558: IPSTUDY(ipstudy.ips_noif++;)
559: goto bad; /* ??? */
560: }
561: *sin++ = ((struct sockaddr_in *)&ifinet->if_addr)->sin_addr;
562: break;
563:
564: case IPOPT_TS_PRESPEC:
565: ipaddr.sin_addr = *sin;
566: if (if_ifwithaddr((struct sockaddr *)&ipaddr) == 0)
567: continue;
568: if (ipt->ipt_ptr + 8 > ipt->ipt_len) {
569: IPSTUDY(ipstudy.ips_oplen++;)
570: goto bad;
571: }
572: ipt->ipt_ptr += 4;
573: break;
574:
575: default:
576: IPSTUDY(ipstudy.ips_badopt++;)
577: goto bad;
578: }
579: *(n_time *)sin = iptime();
580: ipt->ipt_ptr += 4;
581: }
582: }
583: return (0);
584: bad:
585: icmp_error(ip, type, code);
586: return (1);
587: }
588:
589: /*
590: * Strip out IP options, at higher
591: * level protocol in the kernel.
592: * Second argument is buffer to which options
593: * will be moved, and return value is their length.
594: */
595: ip_stripoptions(ip, mopt)
596: struct ip *ip;
597: struct mbuf *mopt;
598: {
599: register int i;
600: register struct mbuf *m;
601: int olen;
602:
603: olen = (ip->ip_hl<<2) - sizeof (struct ip);
604: m = dtom(ip);
605: ip++;
606: if (mopt) {
607: mopt->m_len = olen;
608: mopt->m_off = MMINOFF;
609: bcopy((caddr_t)ip, mtod(m, caddr_t), (unsigned)olen);
610: }
611: i = m->m_len - (sizeof (struct ip) + olen);
612: bcopy((caddr_t)ip+olen, (caddr_t)ip, (unsigned)i);
613: m->m_len -= olen;
614: }
615:
616: u_char inetctlerrmap[PRC_NCMDS] = {
617: ECONNABORTED, ECONNABORTED, 0, 0,
618: 0, 0, EHOSTDOWN, EHOSTUNREACH,
619: ENETUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
620: EMSGSIZE, 0, 0, 0,
621: 0, 0, 0, 0
622: };
623:
624: ip_ctlinput(cmd, arg)
625: int cmd;
626: caddr_t arg;
627: {
628: struct in_addr *in;
629: int tcp_abort(), udp_abort();
630: extern struct inpcb tcb, udb;
631:
632: if (cmd < 0 || cmd > PRC_NCMDS)
633: return;
634: if (inetctlerrmap[cmd] == 0)
635: return; /* XXX */
636: if (cmd == PRC_IFDOWN)
637: in = &((struct sockaddr_in *)arg)->sin_addr;
638: else if (cmd == PRC_HOSTDEAD || cmd == PRC_HOSTUNREACH)
639: in = (struct in_addr *)arg;
640: else
641: in = &((struct icmp *)arg)->icmp_ip.ip_dst;
642: /* THIS IS VERY QUESTIONABLE, SHOULD HIT ALL PROTOCOLS */
643: in_pcbnotify(&tcb, in, (int)inetctlerrmap[cmd], tcp_abort);
644: in_pcbnotify(&udb, in, (int)inetctlerrmap[cmd], udp_abort);
645: }
646:
647: int ipforwarding = 1;
648: /*
649: * Forward a packet. If some error occurs return the sender
650: * and icmp packet. Note we can't always generate a meaningful
651: * icmp message because icmp doesn't have a large enough repetoire
652: * of codes and types.
653: */
654: ip_forward(ip)
655: register struct ip *ip;
656: {
657: register int error, type, code;
658: struct mbuf *mopt, *mcopy;
659:
660: if (ipprintfs)
661: printf("forward: src %x dst %x ttl %x\n", ip->ip_src,
662: ip->ip_dst, ip->ip_ttl);
663: if (ipforwarding == 0) {
664: /* can't tell difference between net and host */
665: IPSTUDY(ipstudy.ips_cantdiffnethost++;)
666: type = ICMP_UNREACH, code = ICMP_UNREACH_NET;
667: goto sendicmp;
668: }
669: if (ip->ip_ttl < IPTTLDEC) {
670: type = ICMP_TIMXCEED, code = ICMP_TIMXCEED_INTRANS;
671: IPSTUDY(ipstudy.ips_timxceed++;)
672: goto sendicmp;
673: }
674: ip->ip_ttl -= IPTTLDEC;
675: mopt = m_get(M_DONTWAIT, MT_DATA);
676: MBUFNULL(8, mopt); /* for debugging */
677: if (mopt == NULL) {
678: m_freem(dtom(ip));
679: return;
680: }
681:
682: /*
683: * Save at most 64 bytes of the packet in case
684: * we need to generate an ICMP message to the src.
685: */
686: mcopy = m_copy(dtom(ip), 0, imin(ip->ip_len, 64));
687: MBUFNULL(35, mcopy); /* for debugging */
688: ip_stripoptions(ip, mopt);
689:
690: error = ip_output(dtom(ip), mopt, (struct route *)0, IP_FORWARDING);
691: if (error == 0) {
692: if (mcopy)
693: m_freem(mcopy);
694: return;
695: }
696: if (mcopy == NULL)
697: return;
698: ip = mtod(mcopy, struct ip *);
699: type = ICMP_UNREACH, code = 0; /* need ``undefined'' */
700: switch (error) {
701:
702: case ENETUNREACH:
703: case ENETDOWN:
704: IPSTUDY(ipstudy.ips_unreach++;)
705: code = ICMP_UNREACH_NET;
706: break;
707:
708: case EMSGSIZE:
709: IPSTUDY(ipstudy.ips_msgsize++;)
710: code = ICMP_UNREACH_NEEDFRAG;
711: break;
712:
713: case EPERM:
714: code = ICMP_UNREACH_PORT;
715: IPSTUDY(ipstudy.ips_badport++;)
716: break;
717:
718: case ENOBUFS:
719: type = ICMP_SOURCEQUENCH;
720: break;
721:
722: case EHOSTDOWN:
723: case EHOSTUNREACH:
724:
725: IPSTUDY(ipstudy.ips_unreach++;)
726: code = ICMP_UNREACH_HOST;
727: break;
728: }
729: sendicmp:
730: icmp_error(ip, type, code);
731: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.