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 | ||
b673e4df PE |
83 | static struct raw_hashinfo raw_v4_hashinfo = { |
84 | .lock = __RW_LOCK_UNLOCKED(), | |
85 | }; | |
1da177e4 | 86 | |
65b4c50b | 87 | void raw_hash_sk(struct sock *sk, struct raw_hashinfo *h) |
1da177e4 | 88 | { |
65b4c50b | 89 | struct hlist_head *head; |
1da177e4 | 90 | |
65b4c50b PE |
91 | head = &h->ht[inet_sk(sk)->num & (RAW_HTABLE_SIZE - 1)]; |
92 | ||
93 | write_lock_bh(&h->lock); | |
1da177e4 | 94 | sk_add_node(sk, head); |
65f76517 | 95 | sock_prot_inuse_add(sk->sk_prot, 1); |
65b4c50b PE |
96 | write_unlock_bh(&h->lock); |
97 | } | |
98 | EXPORT_SYMBOL_GPL(raw_hash_sk); | |
99 | ||
ab70768e PE |
100 | void raw_unhash_sk(struct sock *sk, struct raw_hashinfo *h) |
101 | { | |
102 | write_lock_bh(&h->lock); | |
103 | if (sk_del_node_init(sk)) | |
65f76517 | 104 | sock_prot_inuse_add(sk->sk_prot, -1); |
ab70768e PE |
105 | write_unlock_bh(&h->lock); |
106 | } | |
107 | EXPORT_SYMBOL_GPL(raw_unhash_sk); | |
108 | ||
65b4c50b PE |
109 | static void raw_v4_hash(struct sock *sk) |
110 | { | |
111 | raw_hash_sk(sk, &raw_v4_hashinfo); | |
1da177e4 LT |
112 | } |
113 | ||
114 | static void raw_v4_unhash(struct sock *sk) | |
115 | { | |
ab70768e | 116 | raw_unhash_sk(sk, &raw_v4_hashinfo); |
1da177e4 LT |
117 | } |
118 | ||
be185884 PE |
119 | static struct sock *__raw_v4_lookup(struct net *net, struct sock *sk, |
120 | unsigned short num, __be32 raddr, __be32 laddr, int dif) | |
1da177e4 LT |
121 | { |
122 | struct hlist_node *node; | |
123 | ||
124 | sk_for_each_from(sk, node) { | |
125 | struct inet_sock *inet = inet_sk(sk); | |
126 | ||
be185884 | 127 | if (sk->sk_net == net && inet->num == num && |
1da177e4 LT |
128 | !(inet->daddr && inet->daddr != raddr) && |
129 | !(inet->rcv_saddr && inet->rcv_saddr != laddr) && | |
130 | !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) | |
131 | goto found; /* gotcha */ | |
132 | } | |
133 | sk = NULL; | |
134 | found: | |
135 | return sk; | |
136 | } | |
137 | ||
138 | /* | |
139 | * 0 - deliver | |
140 | * 1 - block | |
141 | */ | |
142 | static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb) | |
143 | { | |
144 | int type; | |
145 | ||
146 | if (!pskb_may_pull(skb, sizeof(struct icmphdr))) | |
147 | return 1; | |
148 | ||
88c7664f | 149 | type = icmp_hdr(skb)->type; |
1da177e4 LT |
150 | if (type < 32) { |
151 | __u32 data = raw_sk(sk)->filter.data; | |
152 | ||
153 | return ((1 << type) & data) != 0; | |
154 | } | |
155 | ||
156 | /* Do not block unknown ICMP types */ | |
157 | return 0; | |
158 | } | |
159 | ||
160 | /* IP input processing comes here for RAW socket delivery. | |
161 | * Caller owns SKB, so we must make clones. | |
162 | * | |
163 | * RFC 1122: SHOULD pass TOS value up to the transport layer. | |
164 | * -> It does. And not only TOS, but all IP header. | |
165 | */ | |
7bc54c90 | 166 | static int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash) |
1da177e4 LT |
167 | { |
168 | struct sock *sk; | |
169 | struct hlist_head *head; | |
d13964f4 | 170 | int delivered = 0; |
be185884 | 171 | struct net *net; |
1da177e4 | 172 | |
b673e4df PE |
173 | read_lock(&raw_v4_hashinfo.lock); |
174 | head = &raw_v4_hashinfo.ht[hash]; | |
1da177e4 LT |
175 | if (hlist_empty(head)) |
176 | goto out; | |
be185884 PE |
177 | |
178 | net = skb->dev->nd_net; | |
179 | sk = __raw_v4_lookup(net, __sk_head(head), iph->protocol, | |
1da177e4 LT |
180 | iph->saddr, iph->daddr, |
181 | skb->dev->ifindex); | |
182 | ||
183 | while (sk) { | |
d13964f4 | 184 | delivered = 1; |
1da177e4 LT |
185 | if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) { |
186 | struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); | |
187 | ||
188 | /* Not releasing hash table! */ | |
189 | if (clone) | |
190 | raw_rcv(sk, clone); | |
191 | } | |
be185884 | 192 | sk = __raw_v4_lookup(net, sk_next(sk), iph->protocol, |
1da177e4 LT |
193 | iph->saddr, iph->daddr, |
194 | skb->dev->ifindex); | |
195 | } | |
196 | out: | |
b673e4df | 197 | read_unlock(&raw_v4_hashinfo.lock); |
d13964f4 | 198 | return delivered; |
1da177e4 LT |
199 | } |
200 | ||
7bc54c90 PE |
201 | int raw_local_deliver(struct sk_buff *skb, int protocol) |
202 | { | |
203 | int hash; | |
204 | struct sock *raw_sk; | |
205 | ||
b673e4df PE |
206 | hash = protocol & (RAW_HTABLE_SIZE - 1); |
207 | raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); | |
7bc54c90 PE |
208 | |
209 | /* If there maybe a raw socket we must check - if not we | |
210 | * don't care less | |
211 | */ | |
212 | if (raw_sk && !raw_v4_input(skb, ip_hdr(skb), hash)) | |
213 | raw_sk = NULL; | |
214 | ||
215 | return raw_sk != NULL; | |
216 | ||
217 | } | |
218 | ||
219 | static void raw_err(struct sock *sk, struct sk_buff *skb, u32 info) | |
1da177e4 LT |
220 | { |
221 | struct inet_sock *inet = inet_sk(sk); | |
88c7664f ACM |
222 | const int type = icmp_hdr(skb)->type; |
223 | const int code = icmp_hdr(skb)->code; | |
1da177e4 LT |
224 | int err = 0; |
225 | int harderr = 0; | |
226 | ||
227 | /* Report error on raw socket, if: | |
228 | 1. User requested ip_recverr. | |
229 | 2. Socket is connected (otherwise the error indication | |
230 | is useless without ip_recverr and error is hard. | |
231 | */ | |
232 | if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED) | |
233 | return; | |
234 | ||
235 | switch (type) { | |
236 | default: | |
237 | case ICMP_TIME_EXCEEDED: | |
238 | err = EHOSTUNREACH; | |
239 | break; | |
240 | case ICMP_SOURCE_QUENCH: | |
241 | return; | |
242 | case ICMP_PARAMETERPROB: | |
243 | err = EPROTO; | |
244 | harderr = 1; | |
245 | break; | |
246 | case ICMP_DEST_UNREACH: | |
247 | err = EHOSTUNREACH; | |
248 | if (code > NR_ICMP_UNREACH) | |
249 | break; | |
250 | err = icmp_err_convert[code].errno; | |
251 | harderr = icmp_err_convert[code].fatal; | |
252 | if (code == ICMP_FRAG_NEEDED) { | |
253 | harderr = inet->pmtudisc != IP_PMTUDISC_DONT; | |
254 | err = EMSGSIZE; | |
255 | } | |
256 | } | |
257 | ||
258 | if (inet->recverr) { | |
259 | struct iphdr *iph = (struct iphdr*)skb->data; | |
260 | u8 *payload = skb->data + (iph->ihl << 2); | |
261 | ||
262 | if (inet->hdrincl) | |
263 | payload = skb->data; | |
264 | ip_icmp_error(sk, skb, err, 0, info, payload); | |
265 | } | |
266 | ||
267 | if (inet->recverr || harderr) { | |
268 | sk->sk_err = err; | |
269 | sk->sk_error_report(sk); | |
270 | } | |
271 | } | |
272 | ||
7bc54c90 PE |
273 | void raw_icmp_error(struct sk_buff *skb, int protocol, u32 info) |
274 | { | |
275 | int hash; | |
276 | struct sock *raw_sk; | |
277 | struct iphdr *iph; | |
be185884 | 278 | struct net *net; |
7bc54c90 | 279 | |
b673e4df | 280 | hash = protocol & (RAW_HTABLE_SIZE - 1); |
7bc54c90 | 281 | |
b673e4df PE |
282 | read_lock(&raw_v4_hashinfo.lock); |
283 | raw_sk = sk_head(&raw_v4_hashinfo.ht[hash]); | |
7bc54c90 PE |
284 | if (raw_sk != NULL) { |
285 | iph = (struct iphdr *)skb->data; | |
be185884 PE |
286 | net = skb->dev->nd_net; |
287 | ||
288 | while ((raw_sk = __raw_v4_lookup(net, raw_sk, protocol, | |
289 | iph->daddr, iph->saddr, | |
7bc54c90 PE |
290 | skb->dev->ifindex)) != NULL) { |
291 | raw_err(raw_sk, skb, info); | |
292 | raw_sk = sk_next(raw_sk); | |
293 | iph = (struct iphdr *)skb->data; | |
294 | } | |
295 | } | |
b673e4df | 296 | read_unlock(&raw_v4_hashinfo.lock); |
7bc54c90 PE |
297 | } |
298 | ||
1da177e4 LT |
299 | static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb) |
300 | { | |
301 | /* Charge it to the socket. */ | |
e905a9ed | 302 | |
1da177e4 | 303 | if (sock_queue_rcv_skb(sk, skb) < 0) { |
33c732c3 | 304 | atomic_inc(&sk->sk_drops); |
1da177e4 LT |
305 | kfree_skb(skb); |
306 | return NET_RX_DROP; | |
307 | } | |
308 | ||
309 | return NET_RX_SUCCESS; | |
310 | } | |
311 | ||
312 | int raw_rcv(struct sock *sk, struct sk_buff *skb) | |
313 | { | |
314 | if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { | |
33c732c3 | 315 | atomic_inc(&sk->sk_drops); |
1da177e4 LT |
316 | kfree_skb(skb); |
317 | return NET_RX_DROP; | |
318 | } | |
b59c2701 | 319 | nf_reset(skb); |
1da177e4 | 320 | |
d56f90a7 | 321 | skb_push(skb, skb->data - skb_network_header(skb)); |
1da177e4 LT |
322 | |
323 | raw_rcv_skb(sk, skb); | |
324 | return 0; | |
325 | } | |
326 | ||
f7d7fc03 | 327 | static int raw_send_hdrinc(struct sock *sk, void *from, size_t length, |
e905a9ed | 328 | struct rtable *rt, |
1da177e4 LT |
329 | unsigned int flags) |
330 | { | |
331 | struct inet_sock *inet = inet_sk(sk); | |
332 | int hh_len; | |
333 | struct iphdr *iph; | |
334 | struct sk_buff *skb; | |
f844c74f | 335 | unsigned int iphlen; |
1da177e4 LT |
336 | int err; |
337 | ||
338 | if (length > rt->u.dst.dev->mtu) { | |
339 | ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport, | |
340 | rt->u.dst.dev->mtu); | |
341 | return -EMSGSIZE; | |
342 | } | |
343 | if (flags&MSG_PROBE) | |
344 | goto out; | |
345 | ||
346 | hh_len = LL_RESERVED_SPACE(rt->u.dst.dev); | |
347 | ||
348 | skb = sock_alloc_send_skb(sk, length+hh_len+15, | |
349 | flags&MSG_DONTWAIT, &err); | |
350 | if (skb == NULL) | |
e905a9ed | 351 | goto error; |
1da177e4 LT |
352 | skb_reserve(skb, hh_len); |
353 | ||
354 | skb->priority = sk->sk_priority; | |
355 | skb->dst = dst_clone(&rt->u.dst); | |
356 | ||
7e28ecc2 | 357 | skb_reset_network_header(skb); |
eddc9ec5 | 358 | iph = ip_hdr(skb); |
7e28ecc2 | 359 | skb_put(skb, length); |
1da177e4 LT |
360 | |
361 | skb->ip_summed = CHECKSUM_NONE; | |
362 | ||
b0e380b1 | 363 | skb->transport_header = skb->network_header; |
1da177e4 LT |
364 | err = memcpy_fromiovecend((void *)iph, from, 0, length); |
365 | if (err) | |
366 | goto error_fault; | |
367 | ||
368 | /* We don't modify invalid header */ | |
f844c74f HX |
369 | iphlen = iph->ihl * 4; |
370 | if (iphlen >= sizeof(*iph) && iphlen <= length) { | |
1da177e4 LT |
371 | if (!iph->saddr) |
372 | iph->saddr = rt->rt_src; | |
373 | iph->check = 0; | |
374 | iph->tot_len = htons(length); | |
375 | if (!iph->id) | |
376 | ip_select_ident(iph, &rt->u.dst, NULL); | |
377 | ||
378 | iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl); | |
379 | } | |
96793b48 DS |
380 | if (iph->protocol == IPPROTO_ICMP) |
381 | icmp_out_count(((struct icmphdr *) | |
382 | skb_transport_header(skb))->type); | |
1da177e4 | 383 | |
6e23ae2a | 384 | err = NF_HOOK(PF_INET, NF_INET_LOCAL_OUT, skb, NULL, rt->u.dst.dev, |
1da177e4 LT |
385 | dst_output); |
386 | if (err > 0) | |
387 | err = inet->recverr ? net_xmit_errno(err) : 0; | |
388 | if (err) | |
389 | goto error; | |
390 | out: | |
391 | return 0; | |
392 | ||
393 | error_fault: | |
394 | err = -EFAULT; | |
395 | kfree_skb(skb); | |
396 | error: | |
397 | IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS); | |
e905a9ed | 398 | return err; |
1da177e4 LT |
399 | } |
400 | ||
a27b58fe | 401 | static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg) |
1da177e4 LT |
402 | { |
403 | struct iovec *iov; | |
404 | u8 __user *type = NULL; | |
405 | u8 __user *code = NULL; | |
406 | int probed = 0; | |
93765d8a | 407 | unsigned int i; |
1da177e4 LT |
408 | |
409 | if (!msg->msg_iov) | |
a27b58fe | 410 | return 0; |
1da177e4 LT |
411 | |
412 | for (i = 0; i < msg->msg_iovlen; i++) { | |
413 | iov = &msg->msg_iov[i]; | |
414 | if (!iov) | |
415 | continue; | |
416 | ||
417 | switch (fl->proto) { | |
418 | case IPPROTO_ICMP: | |
419 | /* check if one-byte field is readable or not. */ | |
420 | if (iov->iov_base && iov->iov_len < 1) | |
421 | break; | |
422 | ||
423 | if (!type) { | |
424 | type = iov->iov_base; | |
425 | /* check if code field is readable or not. */ | |
426 | if (iov->iov_len > 1) | |
427 | code = type + 1; | |
428 | } else if (!code) | |
429 | code = iov->iov_base; | |
430 | ||
431 | if (type && code) { | |
a27b58fe HC |
432 | if (get_user(fl->fl_icmp_type, type) || |
433 | get_user(fl->fl_icmp_code, code)) | |
434 | return -EFAULT; | |
1da177e4 LT |
435 | probed = 1; |
436 | } | |
437 | break; | |
438 | default: | |
439 | probed = 1; | |
440 | break; | |
441 | } | |
442 | if (probed) | |
443 | break; | |
444 | } | |
a27b58fe | 445 | return 0; |
1da177e4 LT |
446 | } |
447 | ||
448 | static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
449 | size_t len) | |
450 | { | |
451 | struct inet_sock *inet = inet_sk(sk); | |
452 | struct ipcm_cookie ipc; | |
453 | struct rtable *rt = NULL; | |
454 | int free = 0; | |
3ca3c68e | 455 | __be32 daddr; |
c1d18f9f | 456 | __be32 saddr; |
1da177e4 LT |
457 | u8 tos; |
458 | int err; | |
459 | ||
460 | err = -EMSGSIZE; | |
926d4b81 | 461 | if (len > 0xFFFF) |
1da177e4 LT |
462 | goto out; |
463 | ||
464 | /* | |
465 | * Check the flags. | |
466 | */ | |
467 | ||
468 | err = -EOPNOTSUPP; | |
469 | if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ | |
470 | goto out; /* compatibility */ | |
e905a9ed | 471 | |
1da177e4 | 472 | /* |
e905a9ed | 473 | * Get and verify the address. |
1da177e4 LT |
474 | */ |
475 | ||
476 | if (msg->msg_namelen) { | |
477 | struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name; | |
478 | err = -EINVAL; | |
479 | if (msg->msg_namelen < sizeof(*usin)) | |
480 | goto out; | |
481 | if (usin->sin_family != AF_INET) { | |
482 | static int complained; | |
483 | if (!complained++) | |
484 | printk(KERN_INFO "%s forgot to set AF_INET in " | |
485 | "raw sendmsg. Fix it!\n", | |
486 | current->comm); | |
487 | err = -EAFNOSUPPORT; | |
488 | if (usin->sin_family) | |
489 | goto out; | |
490 | } | |
491 | daddr = usin->sin_addr.s_addr; | |
492 | /* ANK: I did not forget to get protocol from port field. | |
493 | * I just do not know, who uses this weirdness. | |
494 | * IP_HDRINCL is much more convenient. | |
495 | */ | |
496 | } else { | |
497 | err = -EDESTADDRREQ; | |
e905a9ed | 498 | if (sk->sk_state != TCP_ESTABLISHED) |
1da177e4 LT |
499 | goto out; |
500 | daddr = inet->daddr; | |
501 | } | |
502 | ||
503 | ipc.addr = inet->saddr; | |
504 | ipc.opt = NULL; | |
505 | ipc.oif = sk->sk_bound_dev_if; | |
506 | ||
507 | if (msg->msg_controllen) { | |
508 | err = ip_cmsg_send(msg, &ipc); | |
509 | if (err) | |
510 | goto out; | |
511 | if (ipc.opt) | |
512 | free = 1; | |
513 | } | |
514 | ||
515 | saddr = ipc.addr; | |
516 | ipc.addr = daddr; | |
517 | ||
518 | if (!ipc.opt) | |
519 | ipc.opt = inet->opt; | |
520 | ||
521 | if (ipc.opt) { | |
522 | err = -EINVAL; | |
523 | /* Linux does not mangle headers on raw sockets, | |
524 | * so that IP options + IP_HDRINCL is non-sense. | |
525 | */ | |
526 | if (inet->hdrincl) | |
527 | goto done; | |
528 | if (ipc.opt->srr) { | |
529 | if (!daddr) | |
530 | goto done; | |
531 | daddr = ipc.opt->faddr; | |
532 | } | |
533 | } | |
534 | tos = RT_CONN_FLAGS(sk); | |
535 | if (msg->msg_flags & MSG_DONTROUTE) | |
536 | tos |= RTO_ONLINK; | |
537 | ||
f97c1e0c | 538 | if (ipv4_is_multicast(daddr)) { |
1da177e4 LT |
539 | if (!ipc.oif) |
540 | ipc.oif = inet->mc_index; | |
541 | if (!saddr) | |
542 | saddr = inet->mc_addr; | |
543 | } | |
544 | ||
545 | { | |
546 | struct flowi fl = { .oif = ipc.oif, | |
547 | .nl_u = { .ip4_u = | |
548 | { .daddr = daddr, | |
549 | .saddr = saddr, | |
550 | .tos = tos } }, | |
551 | .proto = inet->hdrincl ? IPPROTO_RAW : | |
e905a9ed | 552 | sk->sk_protocol, |
1da177e4 | 553 | }; |
a27b58fe HC |
554 | if (!inet->hdrincl) { |
555 | err = raw_probe_proto_opt(&fl, msg); | |
556 | if (err) | |
557 | goto done; | |
558 | } | |
1da177e4 | 559 | |
beb8d13b | 560 | security_sk_classify_flow(sk, &fl); |
f1b050bf | 561 | err = ip_route_output_flow(&init_net, &rt, &fl, sk, 1); |
1da177e4 LT |
562 | } |
563 | if (err) | |
564 | goto done; | |
565 | ||
566 | err = -EACCES; | |
567 | if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) | |
568 | goto done; | |
569 | ||
570 | if (msg->msg_flags & MSG_CONFIRM) | |
571 | goto do_confirm; | |
572 | back_from_confirm: | |
573 | ||
574 | if (inet->hdrincl) | |
e905a9ed | 575 | err = raw_send_hdrinc(sk, msg->msg_iov, len, |
1da177e4 | 576 | rt, msg->msg_flags); |
e905a9ed | 577 | |
1da177e4 LT |
578 | else { |
579 | if (!ipc.addr) | |
580 | ipc.addr = rt->rt_dst; | |
581 | lock_sock(sk); | |
582 | err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0, | |
583 | &ipc, rt, msg->msg_flags); | |
584 | if (err) | |
585 | ip_flush_pending_frames(sk); | |
586 | else if (!(msg->msg_flags & MSG_MORE)) | |
587 | err = ip_push_pending_frames(sk); | |
588 | release_sock(sk); | |
589 | } | |
590 | done: | |
591 | if (free) | |
592 | kfree(ipc.opt); | |
593 | ip_rt_put(rt); | |
594 | ||
5418c692 JJ |
595 | out: |
596 | if (err < 0) | |
597 | return err; | |
598 | return len; | |
1da177e4 LT |
599 | |
600 | do_confirm: | |
601 | dst_confirm(&rt->u.dst); | |
602 | if (!(msg->msg_flags & MSG_PROBE) || len) | |
603 | goto back_from_confirm; | |
604 | err = 0; | |
605 | goto done; | |
606 | } | |
607 | ||
608 | static void raw_close(struct sock *sk, long timeout) | |
609 | { | |
e905a9ed | 610 | /* |
1da177e4 LT |
611 | * Raw sockets may have direct kernel refereneces. Kill them. |
612 | */ | |
613 | ip_ra_control(sk, 0, NULL); | |
614 | ||
615 | sk_common_release(sk); | |
616 | } | |
617 | ||
618 | /* This gets rid of all the nasties in af_inet. -DaveM */ | |
619 | static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) | |
620 | { | |
621 | struct inet_sock *inet = inet_sk(sk); | |
622 | struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; | |
623 | int ret = -EINVAL; | |
624 | int chk_addr_ret; | |
625 | ||
626 | if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) | |
627 | goto out; | |
e5ba31f1 | 628 | chk_addr_ret = inet_addr_type(sk->sk_net, addr->sin_addr.s_addr); |
1da177e4 LT |
629 | ret = -EADDRNOTAVAIL; |
630 | if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL && | |
631 | chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST) | |
632 | goto out; | |
633 | inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr; | |
634 | if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) | |
635 | inet->saddr = 0; /* Use device */ | |
636 | sk_dst_reset(sk); | |
637 | ret = 0; | |
638 | out: return ret; | |
639 | } | |
640 | ||
641 | /* | |
642 | * This should be easy, if there is something there | |
643 | * we return it, otherwise we block. | |
644 | */ | |
645 | ||
646 | static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, | |
647 | size_t len, int noblock, int flags, int *addr_len) | |
648 | { | |
649 | struct inet_sock *inet = inet_sk(sk); | |
650 | size_t copied = 0; | |
651 | int err = -EOPNOTSUPP; | |
652 | struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; | |
653 | struct sk_buff *skb; | |
654 | ||
655 | if (flags & MSG_OOB) | |
656 | goto out; | |
657 | ||
658 | if (addr_len) | |
659 | *addr_len = sizeof(*sin); | |
660 | ||
661 | if (flags & MSG_ERRQUEUE) { | |
662 | err = ip_recv_error(sk, msg, len); | |
663 | goto out; | |
664 | } | |
665 | ||
666 | skb = skb_recv_datagram(sk, flags, noblock, &err); | |
667 | if (!skb) | |
668 | goto out; | |
669 | ||
670 | copied = skb->len; | |
671 | if (len < copied) { | |
672 | msg->msg_flags |= MSG_TRUNC; | |
673 | copied = len; | |
674 | } | |
675 | ||
676 | err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); | |
677 | if (err) | |
678 | goto done; | |
679 | ||
680 | sock_recv_timestamp(msg, sk, skb); | |
681 | ||
682 | /* Copy the address. */ | |
683 | if (sin) { | |
684 | sin->sin_family = AF_INET; | |
eddc9ec5 | 685 | sin->sin_addr.s_addr = ip_hdr(skb)->saddr; |
f59fc7f3 | 686 | sin->sin_port = 0; |
1da177e4 LT |
687 | memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); |
688 | } | |
689 | if (inet->cmsg_flags) | |
690 | ip_cmsg_recv(msg, skb); | |
691 | if (flags & MSG_TRUNC) | |
692 | copied = skb->len; | |
693 | done: | |
694 | skb_free_datagram(sk, skb); | |
5418c692 JJ |
695 | out: |
696 | if (err) | |
697 | return err; | |
698 | return copied; | |
1da177e4 LT |
699 | } |
700 | ||
701 | static int raw_init(struct sock *sk) | |
702 | { | |
703 | struct raw_sock *rp = raw_sk(sk); | |
704 | ||
705 | if (inet_sk(sk)->num == IPPROTO_ICMP) | |
706 | memset(&rp->filter, 0, sizeof(rp->filter)); | |
707 | return 0; | |
708 | } | |
709 | ||
710 | static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen) | |
711 | { | |
712 | if (optlen > sizeof(struct icmp_filter)) | |
713 | optlen = sizeof(struct icmp_filter); | |
714 | if (copy_from_user(&raw_sk(sk)->filter, optval, optlen)) | |
715 | return -EFAULT; | |
716 | return 0; | |
717 | } | |
718 | ||
719 | static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) | |
720 | { | |
721 | int len, ret = -EFAULT; | |
722 | ||
723 | if (get_user(len, optlen)) | |
724 | goto out; | |
725 | ret = -EINVAL; | |
726 | if (len < 0) | |
727 | goto out; | |
728 | if (len > sizeof(struct icmp_filter)) | |
729 | len = sizeof(struct icmp_filter); | |
730 | ret = -EFAULT; | |
731 | if (put_user(len, optlen) || | |
732 | copy_to_user(optval, &raw_sk(sk)->filter, len)) | |
733 | goto out; | |
734 | ret = 0; | |
735 | out: return ret; | |
736 | } | |
737 | ||
3fdadf7d | 738 | static int do_raw_setsockopt(struct sock *sk, int level, int optname, |
1da177e4 LT |
739 | char __user *optval, int optlen) |
740 | { | |
1da177e4 LT |
741 | if (optname == ICMP_FILTER) { |
742 | if (inet_sk(sk)->num != IPPROTO_ICMP) | |
743 | return -EOPNOTSUPP; | |
744 | else | |
745 | return raw_seticmpfilter(sk, optval, optlen); | |
746 | } | |
747 | return -ENOPROTOOPT; | |
748 | } | |
749 | ||
3fdadf7d DM |
750 | static int raw_setsockopt(struct sock *sk, int level, int optname, |
751 | char __user *optval, int optlen) | |
1da177e4 LT |
752 | { |
753 | if (level != SOL_RAW) | |
3fdadf7d DM |
754 | return ip_setsockopt(sk, level, optname, optval, optlen); |
755 | return do_raw_setsockopt(sk, level, optname, optval, optlen); | |
756 | } | |
1da177e4 | 757 | |
3fdadf7d DM |
758 | #ifdef CONFIG_COMPAT |
759 | static int compat_raw_setsockopt(struct sock *sk, int level, int optname, | |
543d9cfe | 760 | char __user *optval, int optlen) |
3fdadf7d DM |
761 | { |
762 | if (level != SOL_RAW) | |
543d9cfe | 763 | return compat_ip_setsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
764 | return do_raw_setsockopt(sk, level, optname, optval, optlen); |
765 | } | |
766 | #endif | |
767 | ||
768 | static int do_raw_getsockopt(struct sock *sk, int level, int optname, | |
769 | char __user *optval, int __user *optlen) | |
770 | { | |
1da177e4 LT |
771 | if (optname == ICMP_FILTER) { |
772 | if (inet_sk(sk)->num != IPPROTO_ICMP) | |
773 | return -EOPNOTSUPP; | |
774 | else | |
775 | return raw_geticmpfilter(sk, optval, optlen); | |
776 | } | |
777 | return -ENOPROTOOPT; | |
778 | } | |
779 | ||
3fdadf7d DM |
780 | static int raw_getsockopt(struct sock *sk, int level, int optname, |
781 | char __user *optval, int __user *optlen) | |
782 | { | |
783 | if (level != SOL_RAW) | |
784 | return ip_getsockopt(sk, level, optname, optval, optlen); | |
785 | return do_raw_getsockopt(sk, level, optname, optval, optlen); | |
786 | } | |
787 | ||
788 | #ifdef CONFIG_COMPAT | |
789 | static int compat_raw_getsockopt(struct sock *sk, int level, int optname, | |
543d9cfe | 790 | char __user *optval, int __user *optlen) |
3fdadf7d DM |
791 | { |
792 | if (level != SOL_RAW) | |
543d9cfe | 793 | return compat_ip_getsockopt(sk, level, optname, optval, optlen); |
3fdadf7d DM |
794 | return do_raw_getsockopt(sk, level, optname, optval, optlen); |
795 | } | |
796 | #endif | |
797 | ||
1da177e4 LT |
798 | static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg) |
799 | { | |
800 | switch (cmd) { | |
801 | case SIOCOUTQ: { | |
802 | int amount = atomic_read(&sk->sk_wmem_alloc); | |
803 | return put_user(amount, (int __user *)arg); | |
804 | } | |
805 | case SIOCINQ: { | |
806 | struct sk_buff *skb; | |
807 | int amount = 0; | |
808 | ||
e0f9f858 | 809 | spin_lock_bh(&sk->sk_receive_queue.lock); |
1da177e4 LT |
810 | skb = skb_peek(&sk->sk_receive_queue); |
811 | if (skb != NULL) | |
812 | amount = skb->len; | |
e0f9f858 | 813 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
1da177e4 LT |
814 | return put_user(amount, (int __user *)arg); |
815 | } | |
816 | ||
817 | default: | |
818 | #ifdef CONFIG_IP_MROUTE | |
819 | return ipmr_ioctl(sk, cmd, (void __user *)arg); | |
820 | #else | |
821 | return -ENOIOCTLCMD; | |
822 | #endif | |
823 | } | |
824 | } | |
825 | ||
47a31a6f ED |
826 | DEFINE_PROTO_INUSE(raw) |
827 | ||
1da177e4 | 828 | struct proto raw_prot = { |
543d9cfe ACM |
829 | .name = "RAW", |
830 | .owner = THIS_MODULE, | |
831 | .close = raw_close, | |
832 | .connect = ip4_datagram_connect, | |
833 | .disconnect = udp_disconnect, | |
834 | .ioctl = raw_ioctl, | |
835 | .init = raw_init, | |
836 | .setsockopt = raw_setsockopt, | |
837 | .getsockopt = raw_getsockopt, | |
838 | .sendmsg = raw_sendmsg, | |
839 | .recvmsg = raw_recvmsg, | |
840 | .bind = raw_bind, | |
841 | .backlog_rcv = raw_rcv_skb, | |
842 | .hash = raw_v4_hash, | |
843 | .unhash = raw_v4_unhash, | |
844 | .obj_size = sizeof(struct raw_sock), | |
3fdadf7d | 845 | #ifdef CONFIG_COMPAT |
543d9cfe ACM |
846 | .compat_setsockopt = compat_raw_setsockopt, |
847 | .compat_getsockopt = compat_raw_getsockopt, | |
3fdadf7d | 848 | #endif |
47a31a6f | 849 | REF_PROTO_INUSE(raw) |
1da177e4 LT |
850 | }; |
851 | ||
852 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
853 | static struct sock *raw_get_first(struct seq_file *seq) |
854 | { | |
855 | struct sock *sk; | |
856 | struct raw_iter_state* state = raw_seq_private(seq); | |
857 | ||
b673e4df PE |
858 | for (state->bucket = 0; state->bucket < RAW_HTABLE_SIZE; |
859 | ++state->bucket) { | |
1da177e4 LT |
860 | struct hlist_node *node; |
861 | ||
42a73808 | 862 | sk_for_each(sk, node, &state->h->ht[state->bucket]) |
f51d599f PE |
863 | if (sk->sk_net == state->p.net && |
864 | sk->sk_family == state->family) | |
1da177e4 LT |
865 | goto found; |
866 | } | |
867 | sk = NULL; | |
868 | found: | |
869 | return sk; | |
870 | } | |
871 | ||
872 | static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) | |
873 | { | |
874 | struct raw_iter_state* state = raw_seq_private(seq); | |
875 | ||
876 | do { | |
877 | sk = sk_next(sk); | |
878 | try_again: | |
879 | ; | |
f51d599f PE |
880 | } while (sk && sk->sk_net != state->p.net && |
881 | sk->sk_family != state->family); | |
1da177e4 | 882 | |
b673e4df | 883 | if (!sk && ++state->bucket < RAW_HTABLE_SIZE) { |
42a73808 | 884 | sk = sk_head(&state->h->ht[state->bucket]); |
1da177e4 LT |
885 | goto try_again; |
886 | } | |
887 | return sk; | |
888 | } | |
889 | ||
890 | static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) | |
891 | { | |
892 | struct sock *sk = raw_get_first(seq); | |
893 | ||
894 | if (sk) | |
895 | while (pos && (sk = raw_get_next(seq, sk)) != NULL) | |
896 | --pos; | |
897 | return pos ? NULL : sk; | |
898 | } | |
899 | ||
42a73808 | 900 | void *raw_seq_start(struct seq_file *seq, loff_t *pos) |
1da177e4 | 901 | { |
42a73808 PE |
902 | struct raw_iter_state *state = raw_seq_private(seq); |
903 | ||
904 | read_lock(&state->h->lock); | |
1da177e4 LT |
905 | return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
906 | } | |
42a73808 | 907 | EXPORT_SYMBOL_GPL(raw_seq_start); |
1da177e4 | 908 | |
42a73808 | 909 | void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) |
1da177e4 LT |
910 | { |
911 | struct sock *sk; | |
912 | ||
913 | if (v == SEQ_START_TOKEN) | |
914 | sk = raw_get_first(seq); | |
915 | else | |
916 | sk = raw_get_next(seq, v); | |
917 | ++*pos; | |
918 | return sk; | |
919 | } | |
42a73808 | 920 | EXPORT_SYMBOL_GPL(raw_seq_next); |
1da177e4 | 921 | |
42a73808 | 922 | void raw_seq_stop(struct seq_file *seq, void *v) |
1da177e4 | 923 | { |
42a73808 PE |
924 | struct raw_iter_state *state = raw_seq_private(seq); |
925 | ||
926 | read_unlock(&state->h->lock); | |
1da177e4 | 927 | } |
42a73808 | 928 | EXPORT_SYMBOL_GPL(raw_seq_stop); |
1da177e4 LT |
929 | |
930 | static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i) | |
931 | { | |
932 | struct inet_sock *inet = inet_sk(sp); | |
714e85be AV |
933 | __be32 dest = inet->daddr, |
934 | src = inet->rcv_saddr; | |
1da177e4 LT |
935 | __u16 destp = 0, |
936 | srcp = inet->num; | |
937 | ||
938 | sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X" | |
33c732c3 | 939 | " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p %d", |
e905a9ed | 940 | i, src, srcp, dest, destp, sp->sk_state, |
1da177e4 LT |
941 | atomic_read(&sp->sk_wmem_alloc), |
942 | atomic_read(&sp->sk_rmem_alloc), | |
943 | 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), | |
33c732c3 | 944 | atomic_read(&sp->sk_refcnt), sp, atomic_read(&sp->sk_drops)); |
1da177e4 LT |
945 | return tmpbuf; |
946 | } | |
947 | ||
33c732c3 WC |
948 | #define TMPSZ 128 |
949 | ||
1da177e4 LT |
950 | static int raw_seq_show(struct seq_file *seq, void *v) |
951 | { | |
33c732c3 | 952 | char tmpbuf[TMPSZ+1]; |
1da177e4 LT |
953 | |
954 | if (v == SEQ_START_TOKEN) | |
33c732c3 | 955 | seq_printf(seq, "%-*s\n", TMPSZ-1, |
1da177e4 LT |
956 | " sl local_address rem_address st tx_queue " |
957 | "rx_queue tr tm->when retrnsmt uid timeout " | |
33c732c3 | 958 | "inode drops"); |
1da177e4 LT |
959 | else { |
960 | struct raw_iter_state *state = raw_seq_private(seq); | |
961 | ||
33c732c3 | 962 | seq_printf(seq, "%-*s\n", TMPSZ-1, |
1da177e4 LT |
963 | get_raw_sock(v, tmpbuf, state->bucket)); |
964 | } | |
965 | return 0; | |
966 | } | |
967 | ||
f690808e | 968 | static const struct seq_operations raw_seq_ops = { |
1da177e4 LT |
969 | .start = raw_seq_start, |
970 | .next = raw_seq_next, | |
971 | .stop = raw_seq_stop, | |
972 | .show = raw_seq_show, | |
973 | }; | |
974 | ||
f51d599f | 975 | int raw_seq_open(struct inode *ino, struct file *file, struct raw_hashinfo *h, |
42a73808 | 976 | unsigned short family) |
1da177e4 | 977 | { |
f51d599f | 978 | int err; |
42a73808 PE |
979 | struct raw_iter_state *i; |
980 | ||
f51d599f | 981 | err = seq_open_net(ino, file, &raw_seq_ops, |
cf7732e4 | 982 | sizeof(struct raw_iter_state)); |
f51d599f PE |
983 | if (err < 0) |
984 | return err; | |
42a73808 | 985 | |
f51d599f | 986 | i = raw_seq_private((struct seq_file *)file->private_data); |
42a73808 PE |
987 | i->h = h; |
988 | i->family = family; | |
989 | return 0; | |
990 | } | |
991 | EXPORT_SYMBOL_GPL(raw_seq_open); | |
992 | ||
993 | static int raw_v4_seq_open(struct inode *inode, struct file *file) | |
994 | { | |
f51d599f | 995 | return raw_seq_open(inode, file, &raw_v4_hashinfo, PF_INET); |
1da177e4 LT |
996 | } |
997 | ||
9a32144e | 998 | static const struct file_operations raw_seq_fops = { |
1da177e4 | 999 | .owner = THIS_MODULE, |
42a73808 | 1000 | .open = raw_v4_seq_open, |
1da177e4 LT |
1001 | .read = seq_read, |
1002 | .llseek = seq_lseek, | |
f51d599f | 1003 | .release = seq_release_net, |
1da177e4 LT |
1004 | }; |
1005 | ||
a308da16 | 1006 | static __net_init int raw_init_net(struct net *net) |
1da177e4 | 1007 | { |
a308da16 | 1008 | if (!proc_net_fops_create(net, "raw", S_IRUGO, &raw_seq_fops)) |
1da177e4 | 1009 | return -ENOMEM; |
a308da16 | 1010 | |
1da177e4 LT |
1011 | return 0; |
1012 | } | |
1013 | ||
a308da16 PE |
1014 | static __net_exit void raw_exit_net(struct net *net) |
1015 | { | |
1016 | proc_net_remove(net, "raw"); | |
1017 | } | |
1018 | ||
1019 | static __net_initdata struct pernet_operations raw_net_ops = { | |
1020 | .init = raw_init_net, | |
1021 | .exit = raw_exit_net, | |
1022 | }; | |
1023 | ||
1024 | int __init raw_proc_init(void) | |
1025 | { | |
1026 | return register_pernet_subsys(&raw_net_ops); | |
1027 | } | |
1028 | ||
1da177e4 LT |
1029 | void __init raw_proc_exit(void) |
1030 | { | |
a308da16 | 1031 | unregister_pernet_subsys(&raw_net_ops); |
1da177e4 LT |
1032 | } |
1033 | #endif /* CONFIG_PROC_FS */ |