]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
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 | * RAW - implementation of IP "raw" sockets. | |
7 | * | |
8 | * Version: $Id: raw.c,v 1.64 2002/02/01 22:01:04 davem Exp $ | |
9 | * | |
02c30a84 | 10 | * Authors: Ross Biro |
1da177e4 LT |
11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
12 | * | |
13 | * Fixes: | |
14 | * Alan Cox : verify_area() fixed up | |
15 | * Alan Cox : ICMP error handling | |
16 | * Alan Cox : EMSGSIZE if you send too big a packet | |
17 | * Alan Cox : Now uses generic datagrams and shared | |
18 | * skbuff library. No more peek crashes, | |
19 | * no more backlogs | |
20 | * Alan Cox : Checks sk->broadcast. | |
21 | * Alan Cox : Uses skb_free_datagram/skb_copy_datagram | |
22 | * Alan Cox : Raw passes ip options too | |
23 | * Alan Cox : Setsocketopt added | |
24 | * Alan Cox : Fixed error return for broadcasts | |
25 | * Alan Cox : Removed wake_up calls | |
26 | * Alan Cox : Use ttl/tos | |
27 | * Alan Cox : Cleaned up old debugging | |
28 | * Alan Cox : Use new kernel side addresses | |
29 | * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets. | |
30 | * Alan Cox : BSD style RAW socket demultiplexing. | |
31 | * Alan Cox : Beginnings of mrouted support. | |
32 | * Alan Cox : Added IP_HDRINCL option. | |
33 | * Alan Cox : Skip broadcast check if BSDism set. | |
34 | * David S. Miller : New socket lookup architecture. | |
35 | * | |
36 | * This program is free software; you can redistribute it and/or | |
37 | * modify it under the terms of the GNU General Public License | |
38 | * as published by the Free Software Foundation; either version | |
39 | * 2 of the License, or (at your option) any later version. | |
40 | */ | |
bf0d5249 | 41 | |
715b49ef | 42 | #include <linux/types.h> |
1da177e4 LT |
43 | #include <asm/atomic.h> |
44 | #include <asm/byteorder.h> | |
45 | #include <asm/current.h> | |
46 | #include <asm/uaccess.h> | |
47 | #include <asm/ioctls.h> | |
1da177e4 LT |
48 | #include <linux/stddef.h> |
49 | #include <linux/slab.h> | |
50 | #include <linux/errno.h> | |
51 | #include <linux/aio.h> | |
52 | #include <linux/kernel.h> | |
53 | #include <linux/spinlock.h> | |
54 | #include <linux/sockios.h> | |
55 | #include <linux/socket.h> | |
56 | #include <linux/in.h> | |
57 | #include <linux/mroute.h> | |
58 | #include <linux/netdevice.h> | |
59 | #include <linux/in_route.h> | |
60 | #include <linux/route.h> | |
1da177e4 | 61 | #include <linux/skbuff.h> |
457c4cbc | 62 | #include <net/net_namespace.h> |
1da177e4 LT |
63 | #include <net/dst.h> |
64 | #include <net/sock.h> | |
65 | #include <linux/gfp.h> | |
66 | #include <linux/ip.h> | |
67 | #include <linux/net.h> | |
68 | #include <net/ip.h> | |
69 | #include <net/icmp.h> | |
70 | #include <net/udp.h> | |
71 | #include <net/raw.h> | |
72 | #include <net/snmp.h> | |
c752f073 | 73 | #include <net/tcp_states.h> |
1da177e4 LT |
74 | #include <net/inet_common.h> |
75 | #include <net/checksum.h> | |
76 | #include <net/xfrm.h> | |
77 | #include <linux/rtnetlink.h> | |
78 | #include <linux/proc_fs.h> | |
79 | #include <linux/seq_file.h> | |
80 | #include <linux/netfilter.h> | |
81 | #include <linux/netfilter_ipv4.h> | |
82 | ||
7bc54c90 PE |
83 | #define RAWV4_HTABLE_SIZE MAX_INET_PROTOS |
84 | ||
85 | static struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE]; | |
86 | static DEFINE_RWLOCK(raw_v4_lock); | |
1da177e4 LT |
87 | |
88 | static void raw_v4_hash(struct sock *sk) | |
89 | { | |
90 | struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num & | |
91 | (RAWV4_HTABLE_SIZE - 1)]; | |
92 | ||
93 | write_lock_bh(&raw_v4_lock); | |
94 | sk_add_node(sk, head); | |
95 | sock_prot_inc_use(sk->sk_prot); | |
96 | write_unlock_bh(&raw_v4_lock); | |
97 | } | |
98 | ||
99 | static void raw_v4_unhash(struct sock *sk) | |
100 | { | |
e905a9ed | 101 | write_lock_bh(&raw_v4_lock); |
1da177e4 LT |
102 | if (sk_del_node_init(sk)) |
103 | sock_prot_dec_use(sk->sk_prot); | |
104 | write_unlock_bh(&raw_v4_lock); | |
105 | } | |
106 | ||
7bc54c90 | 107 | static struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num, |
6d741653 | 108 | __be32 raddr, __be32 laddr, |
1da177e4 LT |
109 | int dif) |
110 | { | |
111 | struct hlist_node *node; | |
112 | ||
113 | sk_for_each_from(sk, node) { | |
114 | struct inet_sock *inet = inet_sk(sk); | |
115 | ||
116 | if (inet->num == num && | |
117 | !(inet->daddr && inet->daddr != raddr) && | |
118 | !(inet->rcv_saddr && inet->rcv_saddr != laddr) && | |
119 | !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) | |
120 | goto found; /* gotcha */ | |
121 | } | |
122 | sk = NULL; | |
123 | found: | |
124 | return sk; | |
125 | } | |
126 | ||
127 | /* | |
128 | * 0 - deliver | |
129 | * 1 - block | |
130 | */ | |
131 | static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb) | |
132 | { | |
133 | int type; | |
134 | ||
135 | if (!pskb_may_pull(skb, sizeof(struct icmphdr))) | |
136 | return 1; | |
137 | ||
88c7664f | 138 | type = icmp_hdr(skb)->type; |
1da177e4 LT |
139 | if (type < 32) { |
140 | __u32 data = raw_sk(sk)->filter.data; | |
141 | ||
142 | return ((1 << type) & data) != 0; | |
143 | } | |
144 | ||
145 | /* Do not block unknown ICMP types */ | |
146 | return 0; | |
147 | } | |
148 | ||
149 | /* IP input processing comes here for RAW socket delivery. | |
150 | * Caller owns SKB, so we must make clones. | |
151 | * | |
152 | * RFC 1122: SHOULD pass TOS value up to the transport layer. | |
153 | * -> It does. And not only TOS, but all IP header. | |
154 | */ | |
7bc54c90 | 155 | static int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash) |
1da177e4 LT |
156 | { |
157 | struct sock *sk; | |
158 | struct hlist_head *head; | |
d13964f4 | 159 | int delivered = 0; |
1da177e4 LT |
160 | |
161 | read_lock(&raw_v4_lock); | |
162 | head = &raw_v4_htable[hash]; | |
163 | if (hlist_empty(head)) | |
164 | goto out; | |
165 | sk = __raw_v4_lookup(__sk_head(head), iph->protocol, | |
166 | iph->saddr, iph->daddr, | |
167 | skb->dev->ifindex); | |
168 | ||
169 | while (sk) { | |
d13964f4 | 170 | delivered = 1; |
1da177e4 LT |
171 | if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) { |
172 | struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); | |
173 | ||
174 | /* Not releasing hash table! */ | |
175 | if (clone) | |
176 | raw_rcv(sk, clone); | |
177 | } | |
178 | sk = __raw_v4_lookup(sk_next(sk), iph->protocol, | |
179 | iph->saddr, iph->daddr, | |
180 | skb->dev->ifindex); | |
181 | } | |
182 | out: | |
183 | read_unlock(&raw_v4_lock); | |
d13964f4 | 184 | return delivered; |
1da177e4 LT |
185 | } |
186 | ||
7bc54c90 PE |
187 | int raw_local_deliver(struct sk_buff *skb, int protocol) |
188 | { | |
189 | int hash; | |
190 | struct sock *raw_sk; | |
191 | ||
192 | hash = protocol & (RAWV4_HTABLE_SIZE - 1); | |
193 | raw_sk = sk_head(&raw_v4_htable[hash]); | |
194 | ||
195 | /* If there maybe a raw socket we must check - if not we | |
196 | * don't care less | |
197 | */ | |
198 | if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash)) | |
199 | raw_sk = NULL; | |
200 | ||
201 | return raw_sk != NULL; | |
202 | ||
203 | } | |
204 | ||
205 | static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info) | |
1da177e4 LT |
206 | { |
207 | struct inet_sock *inet = inet_sk(sk); | |
88c7664f ACM |
208 | const int type = icmp_hdr(skb)->type; |
209 | const int code = icmp_hdr(skb)->code; | |
1da177e4 LT |
210 | int err = 0; |
211 | int harderr = 0; | |
212 | ||
213 | /* Report error on raw socket, if: | |
214 | 1. User requested ip_recverr. | |
215 | 2. Socket is connected (otherwise the error indication | |
216 | is useless without ip_recverr and error is hard. | |
217 | */ | |
218 | if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED) | |
219 | return; | |
220 | ||
221 | switch (type) { | |
222 | default: | |
223 | case ICMP_TIME_EXCEEDED: | |
224 | err = EHOSTUNREACH; | |
225 | break; | |
226 | case ICMP_SOURCE_QUENCH: | |
227 | return; | |
228 | case ICMP_PARAMETERPROB: | |
229 | err = EPROTO; | |
230 | harderr = 1; | |
231 | break; | |
232 | case ICMP_DEST_UNREACH: | |
233 | err = EHOSTUNREACH; | |
234 | if (code > NR_ICMP_UNREACH) | |
235 | break; | |
236 | err = icmp_err_convert[code].errno; | |
237 | harderr = icmp_err_convert[code].fatal; | |
238 | if (code == ICMP_FRAG_NEEDED) { | |
239 | harderr = inet->pmtudisc != IP_PMTUDISC_DONT; | |
240 | err = EMSGSIZE; | |
241 | } | |
242 | } | |
243 | ||
244 | if (inet->recverr) { | |
245 | struct iphdr *iph = (struct iphdr*)skb->data; | |
246 | u8 *payload = skb->data + (iph->ihl << 2); | |
247 | ||
248 | if (inet->hdrincl) | |
249 | payload = skb->data; | |
250 | ip_icmp_error(sk, skb, err, 0, info, payload); | |
251 | } | |
252 | ||
253 | if (inet->recverr || harderr) { | |
254 | sk->sk_err = err; | |
255 | sk->sk_error_report(sk); | |
256 | } | |
257 | } | |
258 | ||
7bc54c90 PE |
259 | void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info) |
260 | { | |
261 | int hash; | |
262 | struct sock *raw_sk; | |
263 | struct iphdr *iph; | |
264 | ||
265 | hash = protocol & (RAWV4_HTABLE_SIZE - 1); | |
266 | ||
267 | read_lock(&raw_v4_lock); | |
268 | raw_sk = sk_head(&raw_v4_htable[hash]); | |
269 | if (raw_sk != NULL) { | |
270 | iph = (struct iphdr *)skb->data; | |
271 | while ((raw_sk = __raw_v4_lookup(raw_sk, protocol, iph->daddr, | |
272 | iph->saddr, | |
273 | skb->dev->ifindex)) != NULL) { | |
274 | raw_err(raw_sk, skb, info); | |
275 | raw_sk = sk_next(raw_sk); | |
276 | iph = (struct iphdr *)skb->data; | |
277 | } | |
278 | } | |
279 | read_unlock(&raw_v4_lock); | |
280 | } | |
281 | ||
1da177e4 LT |
282 | static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb) |
283 | { | |
284 | /* Charge it to the socket. */ | |
e905a9ed | 285 | |
1da177e4 | 286 | if (sock_queue_rcv_skb(sk, skb) < 0) { |
33c732c3 | 287 | atomic_inc(&sk->sk_drops); |
1da177e4 LT |
288 | kfree_skb(skb); |
289 | return NET_RX_DROP; | |
290 | } | |
291 | ||
292 | return NET_RX_SUCCESS; | |
293 | } | |
294 | ||
295 | int raw_rcv(struct sock *sk, struct sk_buff *skb) | |
296 | { | |
297 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { | |
33c732c3 | 298 | atomic_inc(&sk->sk_drops); |
1da177e4 LT |
299 | kfree_skb(skb); |
300 | return NET_RX_DROP; | |
301 | } | |
b59c2701 | 302 | nf_reset(skb); |
1da177e4 | 303 | |
d56f90a7 | 304 | skb_push(skb, skb->data - skb_network_header(skb)); |
1da177e4 LT |
305 | |
306 | raw_rcv_skb(sk, skb); | |
307 | return 0; | |
308 | } | |
309 | ||
f7d7fc03 | 310 | static int raw_send_hdrinc(struct sock *sk, void *from, size_t length, |
e905a9ed | 311 | struct rtable *rt, |
1da177e4 LT |
312 | unsigned int flags) |
313 | { | |
314 | struct inet_sock *inet = inet_sk(sk); | |
315 | int hh_len; | |
316 | struct iphdr *iph; | |
317 | struct sk_buff *skb; | |
f844c74f | 318 | unsigned int iphlen; |
1da177e4 LT |
319 | int err; |
320 | ||
321 | if (length > rt->u.dst.dev->mtu) { | |
322 | ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport, | |
323 | rt->u.dst.dev->mtu); | |
324 | return -EMSGSIZE; | |
325 | } | |
326 | if (flags&MSG_PROBE) | |
327 | goto out; | |
328 | ||
329 | hh_len = LL_RESERVED_SPACE(rt->u.dst.dev); | |
330 | ||
331 | skb = sock_alloc_send_skb(sk, length+hh_len+15, | |
332 | flags&MSG_DONTWAIT, &err); | |
333 | if (skb == NULL) | |
e905a9ed | 334 | goto error; |
1da177e4 LT |
335 | skb_reserve(skb, hh_len); |
336 | ||
337 | skb->priority = sk->sk_priority; | |
338 | skb->dst = dst_clone(&rt->u.dst); | |
339 | ||
7e28ecc2 | 340 | skb_reset_network_header(skb); |
eddc9ec5 | 341 | iph = ip_hdr(skb); |
7e28ecc2 | 342 | skb_put(skb, length); |
1da177e4 LT |
343 | |
344 | skb->ip_summed = CHECKSUM_NONE; | |
345 | ||
b0e380b1 | 346 | skb->transport_header = skb->network_header; |
1da177e4 LT |
347 | err = memcpy_fromiovecend((void *)iph, from, 0, length); |
348 | if (err) | |
349 | goto error_fault; | |
350 | ||
351 | /* We don't modify invalid header */ | |
f844c74f HX |
352 | iphlen = iph->ihl * 4; |
353 | if (iphlen >= sizeof(*iph) && iphlen <= length) { | |
1da177e4 LT |
354 | if (!iph->saddr) |
355 | iph->saddr = rt->rt_src; | |
356 | iph->check = 0; | |
357 | iph->tot_len = htons(length); | |
358 | if (!iph->id) | |
359 | ip_select_ident(iph, &rt->u.dst, NULL); | |
360 | ||
361 | iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl); | |
362 | } | |
96793b48 DS |
363 | if (iph->protocol == IPPROTO_ICMP) |
364 | icmp_out_count(((struct icmphdr *) | |
365 | skb_transport_header(skb))->type); | |
1da177e4 | 366 | |
6e23ae2a | 367 | err = NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev, |
1da177e4 LT |
368 | dst_output); |
369 | if (err > 0) | |
370 | err = inet->recverr ? net_xmit_errno(err) : 0; | |
371 | if (err) | |
372 | goto error; | |
373 | out: | |
374 | return 0; | |
375 | ||
376 | error_fault: | |
377 | err = -EFAULT; | |
378 | kfree_skb(skb); | |
379 | error: | |
380 | IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS); | |
e905a9ed | 381 | return err; |
1da177e4 LT |
382 | } |
383 | ||
a27b58fe | 384 | static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg) |
1da177e4 LT |
385 | { |
386 | struct iovec *iov; | |
387 | u8 __user *type = NULL; | |
388 | u8 __user *code = NULL; | |
389 | int probed = 0; | |
93765d8a | 390 | unsigned int i; |
1da177e4 LT |
391 | |
392 | if (!msg->msg_iov) | |
a27b58fe | 393 | return 0; |
1da177e4 LT |
394 | |
395 | for (i = 0; i < msg->msg_iovlen; i++) { | |
396 | iov = &msg->msg_iov[i]; | |
397 | if (!iov) | |
398 | continue; | |
399 | ||
400 | switch (fl->proto) { | |
401 | case IPPROTO_ICMP: | |
402 | /* check if one-byte field is readable or not. */ | |
403 | if (iov->iov_base && iov->iov_len < 1) | |
404 | break; | |
405 | ||
406 | if (!type) { | |
407 | type = iov->iov_base; | |
408 | /* check if code field is readable or not. */ | |
409 | if (iov->iov_len > 1) | |
410 | code = type + 1; | |
411 | } else if (!code) | |
412 | code = iov->iov_base; | |
413 | ||
414 | if (type && code) { | |
a27b58fe HC |
415 | if (get_user(fl->fl_icmp_type, type) || |
416 | get_user(fl->fl_icmp_code, code)) | |
417 | return -EFAULT; | |
1da177e4 LT |
418 | probed = 1; |
419 | } | |
420 | break; | |
421 | default: | |
422 | probed = 1; | |
423 | break; | |
424 | } | |
425 | if (probed) | |
426 | break; | |
427 | } | |
a27b58fe | 428 | return 0; |
1da177e4 LT |
429 | } |
430 | ||
431 | static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
432 | size_t len) | |
433 | { | |
434 | struct inet_sock *inet = inet_sk(sk); | |
435 | struct ipcm_cookie ipc; | |
436 | struct rtable *rt = NULL; | |
437 | int free = 0; | |
3ca3c68e | 438 | __be32 daddr; |
c1d18f9f | 439 | __be32 saddr; |
1da177e4 LT |
440 | u8 tos; |
441 | int err; | |
442 | ||
443 | err = -EMSGSIZE; | |
926d4b81 | 444 | if (len > 0xFFFF) |
1da177e4 LT |
445 | goto out; |
446 | ||
447 | /* | |
448 | * Check the flags. | |
449 | */ | |
450 | ||
451 | err = -EOPNOTSUPP; | |
452 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ | |
453 | goto out; /* compatibility */ | |
e905a9ed | 454 | |
1da177e4 | 455 | /* |
e905a9ed | 456 | * Get and verify the address. |
1da177e4 LT |
457 | */ |
458 | ||
459 | if (msg->msg_namelen) { | |
460 | struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name; | |
461 | err = -EINVAL; | |
462 | if (msg->msg_namelen < sizeof(*usin)) | |
463 | goto out; | |
464 | if (usin->sin_family != AF_INET) { | |
465 | static int complained; | |
466 | if (!complained++) | |
467 | printk(KERN_INFO "%s forgot to set AF_INET in " | |
468 | "raw sendmsg. Fix it!\n", | |
469 | current->comm); | |
470 | err = -EAFNOSUPPORT; | |
471 | if (usin->sin_family) | |
472 | goto out; | |
473 | } | |
474 | daddr = usin->sin_addr.s_addr; | |
475 | /* ANK: I did not forget to get protocol from port field. | |
476 | * I just do not know, who uses this weirdness. | |
477 | * IP_HDRINCL is much more convenient. | |
478 | */ | |
479 | } else { | |
480 | err = -EDESTADDRREQ; | |
e905a9ed | 481 | if (sk->sk_state != TCP_ESTABLISHED) |
1da177e4 LT |
482 | goto out; |
483 | daddr = inet->daddr; | |
484 | } | |
485 | ||
486 | ipc.addr = inet->saddr; | |
487 | ipc.opt = NULL; | |
488 | ipc.oif = sk->sk_bound_dev_if; | |
489 | ||
490 | if (msg->msg_controllen) { | |
491 | err = ip_cmsg_send(msg, &ipc); | |
492 | if (err) | |
493 | goto out; | |
494 | if (ipc.opt) | |
495 | free = 1; | |
496 | } | |
497 | ||
498 | saddr = ipc.addr; | |
499 | ipc.addr = daddr; | |
500 | ||
501 | if (!ipc.opt) | |
502 | ipc.opt = inet->opt; | |
503 | ||
504 | if (ipc.opt) { | |
505 | err = -EINVAL; | |
506 | /* Linux does not mangle headers on raw sockets, | |
507 | * so that IP options + IP_HDRINCL is non-sense. | |
508 | */ | |
509 | if (inet->hdrincl) | |
510 | goto done; | |
511 | if (ipc.opt->srr) { | |
512 | if (!daddr) | |
513 | goto done; | |
514 | daddr = ipc.opt->faddr; | |
515 | } | |
516 | } | |
517 | tos = RT_CONN_FLAGS(sk); | |
518 | if (msg->msg_flags & MSG_DONTROUTE) | |
519 | tos |= RTO_ONLINK; | |
520 | ||
521 | if (MULTICAST(daddr)) { | |
522 | if (!ipc.oif) | |
523 | ipc.oif = inet->mc_index; | |
524 | if (!saddr) | |
525 | saddr = inet->mc_addr; | |
526 | } | |
527 | ||
528 | { | |
529 | struct flowi fl = { .oif = ipc.oif, | |
530 | .nl_u = { .ip4_u = | |
531 | { .daddr = daddr, | |
532 | .saddr = saddr, | |
533 | .tos = tos } }, | |
534 | .proto = inet->hdrincl ? IPPROTO_RAW : | |
e905a9ed | 535 | sk->sk_protocol, |
1da177e4 | 536 | }; |
a27b58fe HC |
537 | if (!inet->hdrincl) { |
538 | err = raw_probe_proto_opt(&fl, msg); | |
539 | if (err) | |
540 | goto done; | |
541 | } | |
1da177e4 | 542 | |
beb8d13b | 543 | security_sk_classify_flow(sk, &fl); |
8eb9086f | 544 | err = ip_route_output_flow(&rt, &fl, sk, 1); |
1da177e4 LT |
545 | } |
546 | if (err) | |
547 | goto done; | |
548 | ||
549 | err = -EACCES; | |
550 | if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) | |
551 | goto done; | |
552 | ||
553 | if (msg->msg_flags & MSG_CONFIRM) | |
554 | goto do_confirm; | |
555 | back_from_confirm: | |
556 | ||
557 | if (inet->hdrincl) | |
e905a9ed | 558 | err = raw_send_hdrinc(sk, msg->msg_iov, len, |
1da177e4 | 559 | rt, msg->msg_flags); |
e905a9ed | 560 | |
1da177e4 LT |
561 | else { |
562 | if (!ipc.addr) | |
563 | ipc.addr = rt->rt_dst; | |
564 | lock_sock(sk); | |
565 | err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0, | |
566 | &ipc, rt, msg->msg_flags); | |
567 | if (err) | |
568 | ip_flush_pending_frames(sk); | |
569 | else if (!(msg->msg_flags & MSG_MORE)) | |
570 | err = ip_push_pending_frames(sk); | |
571 | release_sock(sk); | |
572 | } | |
573 | done: | |
574 | if (free) | |
575 | kfree(ipc.opt); | |
576 | ip_rt_put(rt); | |
577 | ||
5418c692 JJ |
578 | out: |
579 | if (err < 0) | |
580 | return err; | |
581 | return len; | |
1da177e4 LT |
582 | |
583 | do_confirm: | |
584 | dst_confirm(&rt->u.dst); | |
585 | if (!(msg->msg_flags & MSG_PROBE) || len) | |
586 | goto back_from_confirm; | |
587 | err = 0; | |
588 | goto done; | |
589 | } | |
590 | ||
591 | static void raw_close(struct sock *sk, long timeout) | |
592 | { | |
e905a9ed | 593 | /* |
1da177e4 LT |
594 | * Raw sockets may have direct kernel refereneces. Kill them. |
595 | */ | |
596 | ip_ra_control(sk, 0, NULL); | |
597 | ||
598 | sk_common_release(sk); | |
599 | } | |
600 | ||
601 | /* This gets rid of all the nasties in af_inet. -DaveM */ | |
602 | static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) | |
603 | { | |
604 | struct inet_sock *inet = inet_sk(sk); | |
605 | struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; | |
606 | int ret = -EINVAL; | |
607 | int chk_addr_ret; | |
608 | ||
609 | if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) | |
610 | goto out; | |
611 | chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr); | |
612 | ret = -EADDRNOTAVAIL; | |
613 | if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL && | |
614 | chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST) | |
615 | goto out; | |
616 | inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr; | |
617 | if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) | |
618 | inet->saddr = 0; /* Use device */ | |
619 | sk_dst_reset(sk); | |
620 | ret = 0; | |
621 | out: return ret; | |
622 | } | |
623 | ||
624 | /* | |
625 | * This should be easy, if there is something there | |
626 | * we return it, otherwise we block. | |
627 | */ | |
628 | ||
629 | static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
630 | size_t len, int noblock, int flags, int *addr_len) | |
631 | { | |
632 | struct inet_sock *inet = inet_sk(sk); | |
633 | size_t copied = 0; | |
634 | int err = -EOPNOTSUPP; | |
635 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
636 | struct sk_buff *skb; | |
637 | ||
638 | if (flags & MSG_OOB) | |
639 | goto out; | |
640 | ||
641 | if (addr_len) | |
642 | *addr_len = sizeof(*sin); | |
643 | ||
644 | if (flags & MSG_ERRQUEUE) { | |
645 | err = ip_recv_error(sk, msg, len); | |
646 | goto out; | |
647 | } | |
648 | ||
649 | skb = skb_recv_datagram(sk, flags, noblock, &err); | |
650 | if (!skb) | |
651 | goto out; | |
652 | ||
653 | copied = skb->len; | |
654 | if (len < copied) { | |
655 | msg->msg_flags |= MSG_TRUNC; | |
656 | copied = len; | |
657 | } | |
658 | ||
659 | err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); | |
660 | if (err) | |
661 | goto done; | |
662 | ||
663 | sock_recv_timestamp(msg, sk, skb); | |
664 | ||
665 | /* Copy the address. */ | |
666 | if (sin) { | |
667 | sin->sin_family = AF_INET; | |
eddc9ec5 | 668 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; |
f59fc7f3 | 669 | sin->sin_port = 0; |
1da177e4 LT |
670 | memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); |
671 | } | |
672 | if (inet->cmsg_flags) | |
673 | ip_cmsg_recv(msg, skb); | |
674 | if (flags & MSG_TRUNC) | |
675 | copied = skb->len; | |
676 | done: | |
677 | skb_free_datagram(sk, skb); | |
5418c692 JJ |
678 | out: |
679 | if (err) | |
680 | return err; | |
681 | return copied; | |
1da177e4 LT |
682 | } |
683 | ||
684 | static int raw_init(struct sock *sk) | |
685 | { | |
686 | struct raw_sock *rp = raw_sk(sk); | |
687 | ||
688 | if (inet_sk(sk)->num == IPPROTO_ICMP) | |
689 | memset(&rp->filter, 0, sizeof(rp->filter)); | |
690 | return 0; | |
691 | } | |
692 | ||
693 | static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen) | |
694 | { | |
695 | if (optlen > sizeof(struct icmp_filter)) | |
696 | optlen = sizeof(struct icmp_filter); | |
697 | if (copy_from_user(&raw_sk(sk)->filter, optval, optlen)) | |
698 | return -EFAULT; | |
699 | return 0; | |
700 | } | |
701 | ||
702 | static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) | |
703 | { | |
704 | int len, ret = -EFAULT; | |
705 | ||
706 | if (get_user(len, optlen)) | |
707 | goto out; | |
708 | ret = -EINVAL; | |
709 | if (len < 0) | |
710 | goto out; | |
711 | if (len > sizeof(struct icmp_filter)) | |
712 | len = sizeof(struct icmp_filter); | |
713 | ret = -EFAULT; | |
714 | if (put_user(len, optlen) || | |
715 | copy_to_user(optval, &raw_sk(sk)->filter, len)) | |
716 | goto out; | |
717 | ret = 0; | |
718 | out: return ret; | |
719 | } | |
720 | ||
3fdadf7d | 721 | static int do_raw_setsockopt(struct sock *sk, int level, int optname, |
1da177e4 LT |
722 | char __user *optval, int optlen) |
723 | { | |
1da177e4 LT |
724 | if (optname == ICMP_FILTER) { |
725 | if (inet_sk(sk)->num != IPPROTO_ICMP) | |
726 | return -EOPNOTSUPP; | |
727 | else | |
728 | return raw_seticmpfilter(sk, optval, optlen); | |
729 | } | |
730 | return -ENOPROTOOPT; | |
731 | } | |
732 | ||
3fdadf7d DM |
733 | static int raw_setsockopt(struct sock *sk, int level, int optname, |
734 | char __user *optval, int optlen) | |
1da177e4 LT |
735 | { |
736 | if (level != SOL_RAW) | |
3fdadf7d DM |
737 | return ip_setsockopt(sk, level, optname, optval, optlen); |
738 | return do_raw_setsockopt(sk, level, optname, optval, optlen); | |
739 | } | |
1da177e4 | 740 | |
3fdadf7d DM |
741 | #ifdef CONFIG_COMPAT |
742 | static int compat_raw_setsockopt(struct sock *sk, int level, int optname, | |
543d9cfe | 743 | char __user *optval, int optlen) |
3fdadf7d DM |
744 | { |
745 | if (level != SOL_RAW) | |
543d9cfe | 746 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
747 | return do_raw_setsockopt(sk, level, optname, optval, optlen); |
748 | } | |
749 | #endif | |
750 | ||
751 | static int do_raw_getsockopt(struct sock *sk, int level, int optname, | |
752 | char __user *optval, int __user *optlen) | |
753 | { | |
1da177e4 LT |
754 | if (optname == ICMP_FILTER) { |
755 | if (inet_sk(sk)->num != IPPROTO_ICMP) | |
756 | return -EOPNOTSUPP; | |
757 | else | |
758 | return raw_geticmpfilter(sk, optval, optlen); | |
759 | } | |
760 | return -ENOPROTOOPT; | |
761 | } | |
762 | ||
3fdadf7d DM |
763 | static int raw_getsockopt(struct sock *sk, int level, int optname, |
764 | char __user *optval, int __user *optlen) | |
765 | { | |
766 | if (level != SOL_RAW) | |
767 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
768 | return do_raw_getsockopt(sk, level, optname, optval, optlen); | |
769 | } | |
770 | ||
771 | #ifdef CONFIG_COMPAT | |
772 | static int compat_raw_getsockopt(struct sock *sk, int level, int optname, | |
543d9cfe | 773 | char __user *optval, int __user *optlen) |
3fdadf7d DM |
774 | { |
775 | if (level != SOL_RAW) | |
543d9cfe | 776 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
777 | return do_raw_getsockopt(sk, level, optname, optval, optlen); |
778 | } | |
779 | #endif | |
780 | ||
1da177e4 LT |
781 | static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg) |
782 | { | |
783 | switch (cmd) { | |
784 | case SIOCOUTQ: { | |
785 | int amount = atomic_read(&sk->sk_wmem_alloc); | |
786 | return put_user(amount, (int __user *)arg); | |
787 | } | |
788 | case SIOCINQ: { | |
789 | struct sk_buff *skb; | |
790 | int amount = 0; | |
791 | ||
e0f9f858 | 792 | spin_lock_bh(&sk->sk_receive_queue.lock); |
1da177e4 LT |
793 | skb = skb_peek(&sk->sk_receive_queue); |
794 | if (skb != NULL) | |
795 | amount = skb->len; | |
e0f9f858 | 796 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
1da177e4 LT |
797 | return put_user(amount, (int __user *)arg); |
798 | } | |
799 | ||
800 | default: | |
801 | #ifdef CONFIG_IP_MROUTE | |
802 | return ipmr_ioctl(sk, cmd, (void __user *)arg); | |
803 | #else | |
804 | return -ENOIOCTLCMD; | |
805 | #endif | |
806 | } | |
807 | } | |
808 | ||
47a31a6f ED |
809 | DEFINE_PROTO_INUSE(raw) |
810 | ||
1da177e4 | 811 | struct proto raw_prot = { |
543d9cfe ACM |
812 | .name = "RAW", |
813 | .owner = THIS_MODULE, | |
814 | .close = raw_close, | |
815 | .connect = ip4_datagram_connect, | |
816 | .disconnect = udp_disconnect, | |
817 | .ioctl = raw_ioctl, | |
818 | .init = raw_init, | |
819 | .setsockopt = raw_setsockopt, | |
820 | .getsockopt = raw_getsockopt, | |
821 | .sendmsg = raw_sendmsg, | |
822 | .recvmsg = raw_recvmsg, | |
823 | .bind = raw_bind, | |
824 | .backlog_rcv = raw_rcv_skb, | |
825 | .hash = raw_v4_hash, | |
826 | .unhash = raw_v4_unhash, | |
827 | .obj_size = sizeof(struct raw_sock), | |
3fdadf7d | 828 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
829 | .compat_setsockopt = compat_raw_setsockopt, |
830 | .compat_getsockopt = compat_raw_getsockopt, | |
3fdadf7d | 831 | #endif |
47a31a6f | 832 | REF_PROTO_INUSE(raw) |
1da177e4 LT |
833 | }; |
834 | ||
835 | #ifdef CONFIG_PROC_FS | |
836 | struct raw_iter_state { | |
837 | int bucket; | |
838 | }; | |
839 | ||
840 | #define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private) | |
841 | ||
842 | static struct sock *raw_get_first(struct seq_file *seq) | |
843 | { | |
844 | struct sock *sk; | |
845 | struct raw_iter_state* state = raw_seq_private(seq); | |
846 | ||
847 | for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) { | |
848 | struct hlist_node *node; | |
849 | ||
850 | sk_for_each(sk, node, &raw_v4_htable[state->bucket]) | |
851 | if (sk->sk_family == PF_INET) | |
852 | goto found; | |
853 | } | |
854 | sk = NULL; | |
855 | found: | |
856 | return sk; | |
857 | } | |
858 | ||
859 | static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) | |
860 | { | |
861 | struct raw_iter_state* state = raw_seq_private(seq); | |
862 | ||
863 | do { | |
864 | sk = sk_next(sk); | |
865 | try_again: | |
866 | ; | |
867 | } while (sk && sk->sk_family != PF_INET); | |
868 | ||
869 | if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) { | |
870 | sk = sk_head(&raw_v4_htable[state->bucket]); | |
871 | goto try_again; | |
872 | } | |
873 | return sk; | |
874 | } | |
875 | ||
876 | static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) | |
877 | { | |
878 | struct sock *sk = raw_get_first(seq); | |
879 | ||
880 | if (sk) | |
881 | while (pos && (sk = raw_get_next(seq, sk)) != NULL) | |
882 | --pos; | |
883 | return pos ? NULL : sk; | |
884 | } | |
885 | ||
886 | static void *raw_seq_start(struct seq_file *seq, loff_t *pos) | |
887 | { | |
888 | read_lock(&raw_v4_lock); | |
889 | return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; | |
890 | } | |
891 | ||
892 | static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
893 | { | |
894 | struct sock *sk; | |
895 | ||
896 | if (v == SEQ_START_TOKEN) | |
897 | sk = raw_get_first(seq); | |
898 | else | |
899 | sk = raw_get_next(seq, v); | |
900 | ++*pos; | |
901 | return sk; | |
902 | } | |
903 | ||
904 | static void raw_seq_stop(struct seq_file *seq, void *v) | |
905 | { | |
906 | read_unlock(&raw_v4_lock); | |
907 | } | |
908 | ||
909 | static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i) | |
910 | { | |
911 | struct inet_sock *inet = inet_sk(sp); | |
714e85be AV |
912 | __be32 dest = inet->daddr, |
913 | src = inet->rcv_saddr; | |
1da177e4 LT |
914 | __u16 destp = 0, |
915 | srcp = inet->num; | |
916 | ||
917 | sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X" | |
33c732c3 | 918 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d", |
e905a9ed | 919 | i, src, srcp, dest, destp, sp->sk_state, |
1da177e4 LT |
920 | atomic_read(&sp->sk_wmem_alloc), |
921 | atomic_read(&sp->sk_rmem_alloc), | |
922 | 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), | |
33c732c3 | 923 | atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops)); |
1da177e4 LT |
924 | return tmpbuf; |
925 | } | |
926 | ||
33c732c3 WC |
927 | #define TMPSZ 128 |
928 | ||
1da177e4 LT |
929 | static int raw_seq_show(struct seq_file *seq, void *v) |
930 | { | |
33c732c3 | 931 | char tmpbuf[TMPSZ+1]; |
1da177e4 LT |
932 | |
933 | if (v == SEQ_START_TOKEN) | |
33c732c3 | 934 | seq_printf(seq, "%-*s\n", TMPSZ-1, |
1da177e4 LT |
935 | " sl local_address rem_address st tx_queue " |
936 | "rx_queue tr tm->when retrnsmt uid timeout " | |
33c732c3 | 937 | "inode drops"); |
1da177e4 LT |
938 | else { |
939 | struct raw_iter_state *state = raw_seq_private(seq); | |
940 | ||
33c732c3 | 941 | seq_printf(seq, "%-*s\n", TMPSZ-1, |
1da177e4 LT |
942 | get_raw_sock(v, tmpbuf, state->bucket)); |
943 | } | |
944 | return 0; | |
945 | } | |
946 | ||
f690808e | 947 | static const struct seq_operations raw_seq_ops = { |
1da177e4 LT |
948 | .start = raw_seq_start, |
949 | .next = raw_seq_next, | |
950 | .stop = raw_seq_stop, | |
951 | .show = raw_seq_show, | |
952 | }; | |
953 | ||
954 | static int raw_seq_open(struct inode *inode, struct file *file) | |
955 | { | |
cf7732e4 PE |
956 | return seq_open_private(file, &raw_seq_ops, |
957 | sizeof(struct raw_iter_state)); | |
1da177e4 LT |
958 | } |
959 | ||
9a32144e | 960 | static const struct file_operations raw_seq_fops = { |
1da177e4 LT |
961 | .owner = THIS_MODULE, |
962 | .open = raw_seq_open, | |
963 | .read = seq_read, | |
964 | .llseek = seq_lseek, | |
965 | .release = seq_release_private, | |
966 | }; | |
967 | ||
968 | int __init raw_proc_init(void) | |
969 | { | |
457c4cbc | 970 | if (!proc_net_fops_create(&init_net, "raw", S_IRUGO, &raw_seq_fops)) |
1da177e4 LT |
971 | return -ENOMEM; |
972 | return 0; | |
973 | } | |
974 | ||
975 | void __init raw_proc_exit(void) | |
976 | { | |
457c4cbc | 977 | proc_net_remove(&init_net, "raw"); |
1da177e4 LT |
978 | } |
979 | #endif /* CONFIG_PROC_FS */ |