blob: 48b18dad6763b45845b75cb2b852ce2b63273033 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 * PACKET - implements raw packet sockets.
7 *
Jesper Juhl02c30a82005-05-05 16:16:16 -07008 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 * Alan Cox, <gw4pts@gw4pts.ampr.org>
11 *
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +090012 * Fixes:
Linus Torvalds1da177e2005-04-16 15:20:36 -070013 * Alan Cox : verify_area() now used correctly
14 * Alan Cox : new skbuff lists, look ma no backlogs!
15 * Alan Cox : tidied skbuff lists.
16 * Alan Cox : Now uses generic datagram routines I
17 * added. Also fixed the peek/read crash
18 * from all old Linux datagram code.
19 * Alan Cox : Uses the improved datagram code.
20 * Alan Cox : Added NULL's for socket options.
21 * Alan Cox : Re-commented the code.
22 * Alan Cox : Use new kernel side addressing
23 * Rob Janssen : Correct MTU usage.
24 * Dave Platt : Counter leaks caused by incorrect
25 * interrupt locking and some slightly
26 * dubious gcc output. Can you read
27 * compiler: it said _VOLATILE_
28 * Richard Kooijman : Timestamp fixes.
29 * Alan Cox : New buffers. Use sk->mac.raw.
30 * Alan Cox : sendmsg/recvmsg support.
31 * Alan Cox : Protocol setting support
32 * Alexey Kuznetsov : Untied from IPv4 stack.
33 * Cyrus Durgin : Fixed kerneld for kmod.
34 * Michal Ostrowski : Module initialization cleanup.
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +090035 * Ulises Alonso : Frame number limit removal and
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 * packet_set_ring memory leak.
Eric W. Biederman0fb375f2005-09-21 00:11:37 -070037 * Eric Biederman : Allow for > 8 byte hardware addresses.
38 * The convention is that longer addresses
39 * will simply extend the hardware address
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +090040 * byte arrays at the end of sockaddr_ll
Eric W. Biederman0fb375f2005-09-21 00:11:37 -070041 * and packet_mreq.
Johann Baudy69e3c752009-05-18 22:11:22 -070042 * Johann Baudy : Added TX RING.
Linus Torvalds1da177e2005-04-16 15:20:36 -070043 *
44 * This program is free software; you can redistribute it and/or
45 * modify it under the terms of the GNU General Public License
46 * as published by the Free Software Foundation; either version
47 * 2 of the License, or (at your option) any later version.
48 *
49 */
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +090050
Linus Torvalds1da177e2005-04-16 15:20:36 -070051#include <linux/types.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070052#include <linux/mm.h>
Randy Dunlap4fc268d2006-01-11 12:17:47 -080053#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070054#include <linux/fcntl.h>
55#include <linux/socket.h>
56#include <linux/in.h>
57#include <linux/inet.h>
58#include <linux/netdevice.h>
59#include <linux/if_packet.h>
60#include <linux/wireless.h>
Herbert Xuffbc6112007-02-04 23:33:10 -080061#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <linux/kmod.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +020063#include <net/net_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070064#include <net/ip.h>
65#include <net/protocol.h>
66#include <linux/skbuff.h>
67#include <net/sock.h>
68#include <linux/errno.h>
69#include <linux/timer.h>
70#include <asm/system.h>
71#include <asm/uaccess.h>
72#include <asm/ioctls.h>
73#include <asm/page.h>
Al Viroa1f8e7f72006-10-19 16:08:53 -040074#include <asm/cacheflush.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#include <asm/io.h>
76#include <linux/proc_fs.h>
77#include <linux/seq_file.h>
78#include <linux/poll.h>
79#include <linux/module.h>
80#include <linux/init.h>
Herbert Xu905db442009-01-30 14:12:06 -080081#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
83#ifdef CONFIG_INET
84#include <net/inet_common.h>
85#endif
86
Linus Torvalds1da177e2005-04-16 15:20:36 -070087/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 Assumptions:
89 - if device has no dev->hard_header routine, it adds and removes ll header
90 inside itself. In this case ll header is invisible outside of device,
91 but higher levels still should reserve dev->hard_header_len.
92 Some devices are enough clever to reallocate skb, when header
93 will not fit to reserved space (tunnel), another ones are silly
94 (PPP).
95 - packet socket receives packets with pulled ll header,
96 so that SOCK_RAW should push it back.
97
98On receive:
99-----------
100
101Incoming, dev->hard_header!=NULL
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700102 mac_header -> ll header
103 data -> data
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104
105Outgoing, dev->hard_header!=NULL
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700106 mac_header -> ll header
107 data -> ll header
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108
109Incoming, dev->hard_header==NULL
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700110 mac_header -> UNKNOWN position. It is very likely, that it points to ll
111 header. PPP makes it, that is wrong, because introduce
YOSHIFUJI Hideakidb0c58f2007-07-19 10:44:35 +0900112 assymetry between rx and tx paths.
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700113 data -> data
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
115Outgoing, dev->hard_header==NULL
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700116 mac_header -> data. ll header is still not built!
117 data -> data
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118
119Resume
120 If dev->hard_header==NULL we are unlikely to restore sensible ll header.
121
122
123On transmit:
124------------
125
126dev->hard_header != NULL
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700127 mac_header -> ll header
128 data -> ll header
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129
130dev->hard_header == NULL (ll header is added by device, we cannot control it)
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700131 mac_header -> data
132 data -> data
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133
134 We should set nh.raw on output to correct posistion,
135 packet classifier depends on it.
136 */
137
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138/* Private packet socket structures. */
139
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000140struct packet_mclist {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 struct packet_mclist *next;
142 int ifindex;
143 int count;
144 unsigned short type;
145 unsigned short alen;
Eric W. Biederman0fb375f2005-09-21 00:11:37 -0700146 unsigned char addr[MAX_ADDR_LEN];
147};
148/* identical to struct packet_mreq except it has
149 * a longer address field.
150 */
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000151struct packet_mreq_max {
Eric W. Biederman0fb375f2005-09-21 00:11:37 -0700152 int mr_ifindex;
153 unsigned short mr_type;
154 unsigned short mr_alen;
155 unsigned char mr_address[MAX_ADDR_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156};
David S. Millera2efcfa2007-05-29 13:12:50 -0700157
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158#ifdef CONFIG_PACKET_MMAP
Johann Baudy69e3c752009-05-18 22:11:22 -0700159static int packet_set_ring(struct sock *sk, struct tpacket_req *req,
160 int closing, int tx_ring);
161
162struct packet_ring_buffer {
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000163 char **pg_vec;
Johann Baudy69e3c752009-05-18 22:11:22 -0700164 unsigned int head;
165 unsigned int frames_per_block;
166 unsigned int frame_size;
167 unsigned int frame_max;
168
169 unsigned int pg_vec_order;
170 unsigned int pg_vec_pages;
171 unsigned int pg_vec_len;
172
173 atomic_t pending;
174};
175
176struct packet_sock;
177static int tpacket_snd(struct packet_sock *po, struct msghdr *msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178#endif
179
180static void packet_flush_mclist(struct sock *sk);
181
182struct packet_sock {
183 /* struct sock has to be the first member of packet_sock */
184 struct sock sk;
185 struct tpacket_stats stats;
186#ifdef CONFIG_PACKET_MMAP
Johann Baudy69e3c752009-05-18 22:11:22 -0700187 struct packet_ring_buffer rx_ring;
188 struct packet_ring_buffer tx_ring;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 int copy_thresh;
190#endif
191 struct packet_type prot_hook;
192 spinlock_t bind_lock;
Herbert Xu905db442009-01-30 14:12:06 -0800193 struct mutex pg_vec_lock;
Herbert Xu8dc41942007-02-04 23:31:32 -0800194 unsigned int running:1, /* prot_hook is attached*/
Peter P. Waskiewicz Jr80feaac2007-04-20 16:05:39 -0700195 auxdata:1,
196 origdev:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 int ifindex; /* bound device */
Al Viro0e11c912006-11-08 00:26:29 -0800198 __be16 num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 struct packet_mclist *mclist;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200#ifdef CONFIG_PACKET_MMAP
201 atomic_t mapped;
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700202 enum tpacket_versions tp_version;
203 unsigned int tp_hdrlen;
Patrick McHardy8913336a2008-07-18 18:05:19 -0700204 unsigned int tp_reserve;
Johann Baudy69e3c752009-05-18 22:11:22 -0700205 unsigned int tp_loss:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206#endif
207};
208
Herbert Xuffbc6112007-02-04 23:33:10 -0800209struct packet_skb_cb {
210 unsigned int origlen;
211 union {
212 struct sockaddr_pkt pkt;
213 struct sockaddr_ll ll;
214 } sa;
215};
216
217#define PACKET_SKB_CB(__skb) ((struct packet_skb_cb *)((__skb)->cb))
Herbert Xu8dc41942007-02-04 23:31:32 -0800218
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219#ifdef CONFIG_PACKET_MMAP
220
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700221static void __packet_set_status(struct packet_sock *po, void *frame, int status)
222{
223 union {
224 struct tpacket_hdr *h1;
225 struct tpacket2_hdr *h2;
226 void *raw;
227 } h;
228
229 h.raw = frame;
230 switch (po->tp_version) {
231 case TPACKET_V1:
232 h.h1->tp_status = status;
Johann Baudy69e3c752009-05-18 22:11:22 -0700233 flush_dcache_page(virt_to_page(&h.h1->tp_status));
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700234 break;
235 case TPACKET_V2:
236 h.h2->tp_status = status;
Johann Baudy69e3c752009-05-18 22:11:22 -0700237 flush_dcache_page(virt_to_page(&h.h2->tp_status));
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700238 break;
Johann Baudy69e3c752009-05-18 22:11:22 -0700239 default:
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000240 pr_err("TPACKET version not supported\n");
Johann Baudy69e3c752009-05-18 22:11:22 -0700241 BUG();
242 }
243
244 smp_wmb();
245}
246
247static int __packet_get_status(struct packet_sock *po, void *frame)
248{
249 union {
250 struct tpacket_hdr *h1;
251 struct tpacket2_hdr *h2;
252 void *raw;
253 } h;
254
255 smp_rmb();
256
257 h.raw = frame;
258 switch (po->tp_version) {
259 case TPACKET_V1:
260 flush_dcache_page(virt_to_page(&h.h1->tp_status));
261 return h.h1->tp_status;
262 case TPACKET_V2:
263 flush_dcache_page(virt_to_page(&h.h2->tp_status));
264 return h.h2->tp_status;
265 default:
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000266 pr_err("TPACKET version not supported\n");
Johann Baudy69e3c752009-05-18 22:11:22 -0700267 BUG();
268 return 0;
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700269 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270}
Johann Baudy69e3c752009-05-18 22:11:22 -0700271
272static void *packet_lookup_frame(struct packet_sock *po,
273 struct packet_ring_buffer *rb,
274 unsigned int position,
275 int status)
276{
277 unsigned int pg_vec_pos, frame_offset;
278 union {
279 struct tpacket_hdr *h1;
280 struct tpacket2_hdr *h2;
281 void *raw;
282 } h;
283
284 pg_vec_pos = position / rb->frames_per_block;
285 frame_offset = position % rb->frames_per_block;
286
287 h.raw = rb->pg_vec[pg_vec_pos] + (frame_offset * rb->frame_size);
288
289 if (status != __packet_get_status(po, h.raw))
290 return NULL;
291
292 return h.raw;
293}
294
295static inline void *packet_current_frame(struct packet_sock *po,
296 struct packet_ring_buffer *rb,
297 int status)
298{
299 return packet_lookup_frame(po, rb, rb->head, status);
300}
301
302static inline void *packet_previous_frame(struct packet_sock *po,
303 struct packet_ring_buffer *rb,
304 int status)
305{
306 unsigned int previous = rb->head ? rb->head - 1 : rb->frame_max;
307 return packet_lookup_frame(po, rb, previous, status);
308}
309
310static inline void packet_increment_head(struct packet_ring_buffer *buff)
311{
312 buff->head = buff->head != buff->frame_max ? buff->head+1 : 0;
313}
314
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315#endif
316
317static inline struct packet_sock *pkt_sk(struct sock *sk)
318{
319 return (struct packet_sock *)sk;
320}
321
322static void packet_sock_destruct(struct sock *sk)
323{
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700324 WARN_ON(atomic_read(&sk->sk_rmem_alloc));
325 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326
327 if (!sock_flag(sk, SOCK_DEAD)) {
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000328 pr_err("Attempt to release alive packet socket: %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 return;
330 }
331
Pavel Emelyanov17ab56a2007-11-10 21:38:48 -0800332 sk_refcnt_debug_dec(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333}
334
335
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800336static const struct proto_ops packet_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800338static const struct proto_ops packet_ops_spkt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000340static int packet_rcv_spkt(struct sk_buff *skb, struct net_device *dev,
341 struct packet_type *pt, struct net_device *orig_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342{
343 struct sock *sk;
344 struct sockaddr_pkt *spkt;
345
346 /*
347 * When we registered the protocol we saved the socket in the data
348 * field for just this event.
349 */
350
351 sk = pt->af_packet_priv;
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900352
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 /*
354 * Yank back the headers [hope the device set this
355 * right or kerboom...]
356 *
357 * Incoming packets have ll header pulled,
358 * push it back.
359 *
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -0700360 * For outgoing ones skb->data == skb_mac_header(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 * so that this procedure is noop.
362 */
363
364 if (skb->pkt_type == PACKET_LOOPBACK)
365 goto out;
366
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900367 if (dev_net(dev) != sock_net(sk))
Denis V. Lunevd12d01d2007-11-19 22:28:35 -0800368 goto out;
369
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000370 skb = skb_share_check(skb, GFP_ATOMIC);
371 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 goto oom;
373
374 /* drop any routing info */
Eric Dumazetadf30902009-06-02 05:19:30 +0000375 skb_dst_drop(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376
Phil Oester84531c22005-07-12 11:57:52 -0700377 /* drop conntrack reference */
378 nf_reset(skb);
379
Herbert Xuffbc6112007-02-04 23:33:10 -0800380 spkt = &PACKET_SKB_CB(skb)->sa.pkt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -0700382 skb_push(skb, skb->data - skb_mac_header(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383
384 /*
385 * The SOCK_PACKET socket receives _all_ frames.
386 */
387
388 spkt->spkt_family = dev->type;
389 strlcpy(spkt->spkt_device, dev->name, sizeof(spkt->spkt_device));
390 spkt->spkt_protocol = skb->protocol;
391
392 /*
393 * Charge the memory to the socket. This is done specifically
394 * to prevent sockets using all the memory up.
395 */
396
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000397 if (sock_queue_rcv_skb(sk, skb) == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 return 0;
399
400out:
401 kfree_skb(skb);
402oom:
403 return 0;
404}
405
406
407/*
408 * Output a raw packet to a device layer. This bypasses all the other
409 * protocol layers and you must therefore supply it with a complete frame
410 */
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900411
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412static int packet_sendmsg_spkt(struct kiocb *iocb, struct socket *sock,
413 struct msghdr *msg, size_t len)
414{
415 struct sock *sk = sock->sk;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000416 struct sockaddr_pkt *saddr = (struct sockaddr_pkt *)msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 struct sk_buff *skb;
418 struct net_device *dev;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000419 __be16 proto = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 int err;
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900421
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 /*
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900423 * Get and verify the address.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 */
425
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000426 if (saddr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 if (msg->msg_namelen < sizeof(struct sockaddr))
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000428 return -EINVAL;
429 if (msg->msg_namelen == sizeof(struct sockaddr_pkt))
430 proto = saddr->spkt_protocol;
431 } else
432 return -ENOTCONN; /* SOCK_PACKET must be sent giving an address */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433
434 /*
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900435 * Find the device first to size check it
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 */
437
438 saddr->spkt_device[13] = 0;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900439 dev = dev_get_by_name(sock_net(sk), saddr->spkt_device);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 err = -ENODEV;
441 if (dev == NULL)
442 goto out_unlock;
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900443
David S. Millerd5e76b02007-01-25 19:30:36 -0800444 err = -ENETDOWN;
445 if (!(dev->flags & IFF_UP))
446 goto out_unlock;
447
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 /*
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000449 * You may not queue a frame bigger than the mtu. This is the lowest level
450 * raw protocol and you must do your own fragmentation at this level.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 */
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900452
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 err = -EMSGSIZE;
Kris Katterjohn8ae55f02006-01-23 16:28:02 -0800454 if (len > dev->mtu + dev->hard_header_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 goto out_unlock;
456
457 err = -ENOBUFS;
458 skb = sock_wmalloc(sk, len + LL_RESERVED_SPACE(dev), 0, GFP_KERNEL);
459
460 /*
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000461 * If the write buffer is full, then tough. At this level the user
462 * gets to deal with the problem - do your own algorithmic backoffs.
463 * That's far more flexible.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 */
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900465
466 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 goto out_unlock;
468
469 /*
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900470 * Fill it in
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 */
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 /* FIXME: Save some space for broken drivers that write a
474 * hard header at transmission time by themselves. PPP is the
475 * notable one here. This should really be fixed at the driver level.
476 */
477 skb_reserve(skb, LL_RESERVED_SPACE(dev));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700478 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479
480 /* Try to align data part correctly */
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700481 if (dev->header_ops) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 skb->data -= dev->hard_header_len;
483 skb->tail -= dev->hard_header_len;
484 if (len < dev->hard_header_len)
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -0700485 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 }
487
488 /* Returns -EFAULT on error */
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000489 err = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 skb->protocol = proto;
491 skb->dev = dev;
492 skb->priority = sk->sk_priority;
493 if (err)
494 goto out_free;
495
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 /*
497 * Now send it
498 */
499
500 dev_queue_xmit(skb);
501 dev_put(dev);
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000502 return len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
504out_free:
505 kfree_skb(skb);
506out_unlock:
507 if (dev)
508 dev_put(dev);
509 return err;
510}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511
David S. Millerdbcb5852007-01-24 15:21:02 -0800512static inline unsigned int run_filter(struct sk_buff *skb, struct sock *sk,
513 unsigned int res)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514{
515 struct sk_filter *filter;
516
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700517 rcu_read_lock_bh();
518 filter = rcu_dereference(sk->sk_filter);
David S. Millerdbcb5852007-01-24 15:21:02 -0800519 if (filter != NULL)
520 res = sk_run_filter(skb, filter->insns, filter->len);
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700521 rcu_read_unlock_bh();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522
David S. Millerdbcb5852007-01-24 15:21:02 -0800523 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524}
525
526/*
527 This function makes lazy skb cloning in hope that most of packets
528 are discarded by BPF.
529
530 Note tricky part: we DO mangle shared skb! skb->data, skb->len
531 and skb->cb are mangled. It works because (and until) packets
532 falling here are owned by current CPU. Output packets are cloned
533 by dev_queue_xmit_nit(), input packets are processed by net_bh
534 sequencially, so that if we return skb to original state on exit,
535 we will not harm anyone.
536 */
537
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000538static int packet_rcv(struct sk_buff *skb, struct net_device *dev,
539 struct packet_type *pt, struct net_device *orig_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540{
541 struct sock *sk;
542 struct sockaddr_ll *sll;
543 struct packet_sock *po;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000544 u8 *skb_head = skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 int skb_len = skb->len;
David S. Millerdbcb5852007-01-24 15:21:02 -0800546 unsigned int snaplen, res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547
548 if (skb->pkt_type == PACKET_LOOPBACK)
549 goto drop;
550
551 sk = pt->af_packet_priv;
552 po = pkt_sk(sk);
553
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900554 if (dev_net(dev) != sock_net(sk))
Denis V. Lunevd12d01d2007-11-19 22:28:35 -0800555 goto drop;
556
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 skb->dev = dev;
558
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700559 if (dev->header_ops) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 /* The device has an explicit notion of ll header,
561 exported to higher levels.
562
563 Otherwise, the device hides datails of it frame
564 structure, so that corresponding packet head
565 never delivered to user.
566 */
567 if (sk->sk_type != SOCK_DGRAM)
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -0700568 skb_push(skb, skb->data - skb_mac_header(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 else if (skb->pkt_type == PACKET_OUTGOING) {
570 /* Special case: outgoing packets have ll header at head */
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -0300571 skb_pull(skb, skb_network_offset(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 }
573 }
574
575 snaplen = skb->len;
576
David S. Millerdbcb5852007-01-24 15:21:02 -0800577 res = run_filter(skb, sk, snaplen);
578 if (!res)
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700579 goto drop_n_restore;
David S. Millerdbcb5852007-01-24 15:21:02 -0800580 if (snaplen > res)
581 snaplen = res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
583 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
584 (unsigned)sk->sk_rcvbuf)
585 goto drop_n_acct;
586
587 if (skb_shared(skb)) {
588 struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
589 if (nskb == NULL)
590 goto drop_n_acct;
591
592 if (skb_head != skb->data) {
593 skb->data = skb_head;
594 skb->len = skb_len;
595 }
596 kfree_skb(skb);
597 skb = nskb;
598 }
599
Herbert Xuffbc6112007-02-04 23:33:10 -0800600 BUILD_BUG_ON(sizeof(*PACKET_SKB_CB(skb)) + MAX_ADDR_LEN - 8 >
601 sizeof(skb->cb));
602
603 sll = &PACKET_SKB_CB(skb)->sa.ll;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 sll->sll_family = AF_PACKET;
605 sll->sll_hatype = dev->type;
606 sll->sll_protocol = skb->protocol;
607 sll->sll_pkttype = skb->pkt_type;
Peter P Waskiewicz Jr8032b462007-11-10 22:03:25 -0800608 if (unlikely(po->origdev))
Peter P. Waskiewicz Jr80feaac2007-04-20 16:05:39 -0700609 sll->sll_ifindex = orig_dev->ifindex;
610 else
611 sll->sll_ifindex = dev->ifindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
Stephen Hemmingerb95cce32007-09-26 22:13:38 -0700613 sll->sll_halen = dev_parse_header(skb, sll->sll_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614
Herbert Xuffbc6112007-02-04 23:33:10 -0800615 PACKET_SKB_CB(skb)->origlen = skb->len;
Herbert Xu8dc41942007-02-04 23:31:32 -0800616
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 if (pskb_trim(skb, snaplen))
618 goto drop_n_acct;
619
620 skb_set_owner_r(skb, sk);
621 skb->dev = NULL;
Eric Dumazetadf30902009-06-02 05:19:30 +0000622 skb_dst_drop(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623
Phil Oester84531c22005-07-12 11:57:52 -0700624 /* drop conntrack reference */
625 nf_reset(skb);
626
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 spin_lock(&sk->sk_receive_queue.lock);
628 po->stats.tp_packets++;
629 __skb_queue_tail(&sk->sk_receive_queue, skb);
630 spin_unlock(&sk->sk_receive_queue.lock);
631 sk->sk_data_ready(sk, skb->len);
632 return 0;
633
634drop_n_acct:
635 spin_lock(&sk->sk_receive_queue.lock);
636 po->stats.tp_drops++;
637 spin_unlock(&sk->sk_receive_queue.lock);
638
639drop_n_restore:
640 if (skb_head != skb->data && skb_shared(skb)) {
641 skb->data = skb_head;
642 skb->len = skb_len;
643 }
644drop:
Neil Hormanead2ceb2009-03-11 09:49:55 +0000645 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 return 0;
647}
648
649#ifdef CONFIG_PACKET_MMAP
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000650static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev,
651 struct packet_type *pt, struct net_device *orig_dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652{
653 struct sock *sk;
654 struct packet_sock *po;
655 struct sockaddr_ll *sll;
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700656 union {
657 struct tpacket_hdr *h1;
658 struct tpacket2_hdr *h2;
659 void *raw;
660 } h;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000661 u8 *skb_head = skb->data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 int skb_len = skb->len;
David S. Millerdbcb5852007-01-24 15:21:02 -0800663 unsigned int snaplen, res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 unsigned long status = TP_STATUS_LOSING|TP_STATUS_USER;
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700665 unsigned short macoff, netoff, hdrlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 struct sk_buff *copy_skb = NULL;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700667 struct timeval tv;
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700668 struct timespec ts;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
670 if (skb->pkt_type == PACKET_LOOPBACK)
671 goto drop;
672
673 sk = pt->af_packet_priv;
674 po = pkt_sk(sk);
675
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900676 if (dev_net(dev) != sock_net(sk))
Denis V. Lunevd12d01d2007-11-19 22:28:35 -0800677 goto drop;
678
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700679 if (dev->header_ops) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 if (sk->sk_type != SOCK_DGRAM)
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -0700681 skb_push(skb, skb->data - skb_mac_header(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 else if (skb->pkt_type == PACKET_OUTGOING) {
683 /* Special case: outgoing packets have ll header at head */
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -0300684 skb_pull(skb, skb_network_offset(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 }
686 }
687
Herbert Xu8dc41942007-02-04 23:31:32 -0800688 if (skb->ip_summed == CHECKSUM_PARTIAL)
689 status |= TP_STATUS_CSUMNOTREADY;
690
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 snaplen = skb->len;
692
David S. Millerdbcb5852007-01-24 15:21:02 -0800693 res = run_filter(skb, sk, snaplen);
694 if (!res)
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700695 goto drop_n_restore;
David S. Millerdbcb5852007-01-24 15:21:02 -0800696 if (snaplen > res)
697 snaplen = res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
699 if (sk->sk_type == SOCK_DGRAM) {
Patrick McHardy8913336a2008-07-18 18:05:19 -0700700 macoff = netoff = TPACKET_ALIGN(po->tp_hdrlen) + 16 +
701 po->tp_reserve;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 } else {
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -0300703 unsigned maclen = skb_network_offset(skb);
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700704 netoff = TPACKET_ALIGN(po->tp_hdrlen +
Patrick McHardy8913336a2008-07-18 18:05:19 -0700705 (maclen < 16 ? 16 : maclen)) +
706 po->tp_reserve;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 macoff = netoff - maclen;
708 }
709
Johann Baudy69e3c752009-05-18 22:11:22 -0700710 if (macoff + snaplen > po->rx_ring.frame_size) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 if (po->copy_thresh &&
712 atomic_read(&sk->sk_rmem_alloc) + skb->truesize <
713 (unsigned)sk->sk_rcvbuf) {
714 if (skb_shared(skb)) {
715 copy_skb = skb_clone(skb, GFP_ATOMIC);
716 } else {
717 copy_skb = skb_get(skb);
718 skb_head = skb->data;
719 }
720 if (copy_skb)
721 skb_set_owner_r(copy_skb, sk);
722 }
Johann Baudy69e3c752009-05-18 22:11:22 -0700723 snaplen = po->rx_ring.frame_size - macoff;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 if ((int)snaplen < 0)
725 snaplen = 0;
726 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
728 spin_lock(&sk->sk_receive_queue.lock);
Johann Baudy69e3c752009-05-18 22:11:22 -0700729 h.raw = packet_current_frame(po, &po->rx_ring, TP_STATUS_KERNEL);
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700730 if (!h.raw)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 goto ring_is_full;
Johann Baudy69e3c752009-05-18 22:11:22 -0700732 packet_increment_head(&po->rx_ring);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 po->stats.tp_packets++;
734 if (copy_skb) {
735 status |= TP_STATUS_COPY;
736 __skb_queue_tail(&sk->sk_receive_queue, copy_skb);
737 }
738 if (!po->stats.tp_drops)
739 status &= ~TP_STATUS_LOSING;
740 spin_unlock(&sk->sk_receive_queue.lock);
741
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700742 skb_copy_bits(skb, 0, h.raw + macoff, snaplen);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700744 switch (po->tp_version) {
745 case TPACKET_V1:
746 h.h1->tp_len = skb->len;
747 h.h1->tp_snaplen = snaplen;
748 h.h1->tp_mac = macoff;
749 h.h1->tp_net = netoff;
750 if (skb->tstamp.tv64)
751 tv = ktime_to_timeval(skb->tstamp);
752 else
753 do_gettimeofday(&tv);
754 h.h1->tp_sec = tv.tv_sec;
755 h.h1->tp_usec = tv.tv_usec;
756 hdrlen = sizeof(*h.h1);
757 break;
758 case TPACKET_V2:
759 h.h2->tp_len = skb->len;
760 h.h2->tp_snaplen = snaplen;
761 h.h2->tp_mac = macoff;
762 h.h2->tp_net = netoff;
763 if (skb->tstamp.tv64)
764 ts = ktime_to_timespec(skb->tstamp);
765 else
766 getnstimeofday(&ts);
767 h.h2->tp_sec = ts.tv_sec;
768 h.h2->tp_nsec = ts.tv_nsec;
Patrick McHardy393e52e2008-07-14 22:50:39 -0700769 h.h2->tp_vlan_tci = skb->vlan_tci;
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700770 hdrlen = sizeof(*h.h2);
771 break;
772 default:
773 BUG();
774 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700776 sll = h.raw + TPACKET_ALIGN(hdrlen);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -0700777 sll->sll_halen = dev_parse_header(skb, sll->sll_addr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 sll->sll_family = AF_PACKET;
779 sll->sll_hatype = dev->type;
780 sll->sll_protocol = skb->protocol;
781 sll->sll_pkttype = skb->pkt_type;
Peter P Waskiewicz Jr8032b462007-11-10 22:03:25 -0800782 if (unlikely(po->origdev))
Peter P. Waskiewicz Jr80feaac2007-04-20 16:05:39 -0700783 sll->sll_ifindex = orig_dev->ifindex;
784 else
785 sll->sll_ifindex = dev->ifindex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700787 __packet_set_status(po, h.raw, status);
Ralf Baechlee16aa202006-12-07 00:11:33 -0800788 smp_mb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 {
790 struct page *p_start, *p_end;
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700791 u8 *h_end = h.raw + macoff + snaplen - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
Patrick McHardybbd6ef82008-07-14 22:50:15 -0700793 p_start = virt_to_page(h.raw);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 p_end = virt_to_page(h_end);
795 while (p_start <= p_end) {
796 flush_dcache_page(p_start);
797 p_start++;
798 }
799 }
800
801 sk->sk_data_ready(sk, 0);
802
803drop_n_restore:
804 if (skb_head != skb->data && skb_shared(skb)) {
805 skb->data = skb_head;
806 skb->len = skb_len;
807 }
808drop:
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +0900809 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 return 0;
811
812ring_is_full:
813 po->stats.tp_drops++;
814 spin_unlock(&sk->sk_receive_queue.lock);
815
816 sk->sk_data_ready(sk, 0);
Wei Yongjunacb5d752009-02-25 00:36:42 +0000817 kfree_skb(copy_skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 goto drop_n_restore;
819}
820
Johann Baudy69e3c752009-05-18 22:11:22 -0700821static void tpacket_destruct_skb(struct sk_buff *skb)
822{
823 struct packet_sock *po = pkt_sk(skb->sk);
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000824 void *ph;
Johann Baudy69e3c752009-05-18 22:11:22 -0700825
826 BUG_ON(skb == NULL);
827
828 if (likely(po->tx_ring.pg_vec)) {
829 ph = skb_shinfo(skb)->destructor_arg;
830 BUG_ON(__packet_get_status(po, ph) != TP_STATUS_SENDING);
831 BUG_ON(atomic_read(&po->tx_ring.pending) == 0);
832 atomic_dec(&po->tx_ring.pending);
833 __packet_set_status(po, ph, TP_STATUS_AVAILABLE);
834 }
835
836 sock_wfree(skb);
837}
838
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000839static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb,
840 void *frame, struct net_device *dev, int size_max,
841 __be16 proto, unsigned char *addr)
Johann Baudy69e3c752009-05-18 22:11:22 -0700842{
843 union {
844 struct tpacket_hdr *h1;
845 struct tpacket2_hdr *h2;
846 void *raw;
847 } ph;
848 int to_write, offset, len, tp_len, nr_frags, len_max;
849 struct socket *sock = po->sk.sk_socket;
850 struct page *page;
851 void *data;
852 int err;
853
854 ph.raw = frame;
855
856 skb->protocol = proto;
857 skb->dev = dev;
858 skb->priority = po->sk.sk_priority;
859 skb_shinfo(skb)->destructor_arg = ph.raw;
860
861 switch (po->tp_version) {
862 case TPACKET_V2:
863 tp_len = ph.h2->tp_len;
864 break;
865 default:
866 tp_len = ph.h1->tp_len;
867 break;
868 }
869 if (unlikely(tp_len > size_max)) {
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000870 pr_err("packet size is too long (%d > %d)\n", tp_len, size_max);
Johann Baudy69e3c752009-05-18 22:11:22 -0700871 return -EMSGSIZE;
872 }
873
874 skb_reserve(skb, LL_RESERVED_SPACE(dev));
875 skb_reset_network_header(skb);
876
877 data = ph.raw + po->tp_hdrlen - sizeof(struct sockaddr_ll);
878 to_write = tp_len;
879
880 if (sock->type == SOCK_DGRAM) {
881 err = dev_hard_header(skb, dev, ntohs(proto), addr,
882 NULL, tp_len);
883 if (unlikely(err < 0))
884 return -EINVAL;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000885 } else if (dev->hard_header_len) {
Johann Baudy69e3c752009-05-18 22:11:22 -0700886 /* net device doesn't like empty head */
887 if (unlikely(tp_len <= dev->hard_header_len)) {
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000888 pr_err("packet size is too short (%d < %d)\n",
889 tp_len, dev->hard_header_len);
Johann Baudy69e3c752009-05-18 22:11:22 -0700890 return -EINVAL;
891 }
892
893 skb_push(skb, dev->hard_header_len);
894 err = skb_store_bits(skb, 0, data,
895 dev->hard_header_len);
896 if (unlikely(err))
897 return err;
898
899 data += dev->hard_header_len;
900 to_write -= dev->hard_header_len;
901 }
902
903 err = -EFAULT;
904 page = virt_to_page(data);
905 offset = offset_in_page(data);
906 len_max = PAGE_SIZE - offset;
907 len = ((to_write > len_max) ? len_max : to_write);
908
909 skb->data_len = to_write;
910 skb->len += to_write;
911 skb->truesize += to_write;
912 atomic_add(to_write, &po->sk.sk_wmem_alloc);
913
914 while (likely(to_write)) {
915 nr_frags = skb_shinfo(skb)->nr_frags;
916
917 if (unlikely(nr_frags >= MAX_SKB_FRAGS)) {
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000918 pr_err("Packet exceed the number of skb frags(%lu)\n",
919 MAX_SKB_FRAGS);
Johann Baudy69e3c752009-05-18 22:11:22 -0700920 return -EFAULT;
921 }
922
923 flush_dcache_page(page);
924 get_page(page);
925 skb_fill_page_desc(skb,
926 nr_frags,
927 page++, offset, len);
928 to_write -= len;
929 offset = 0;
930 len_max = PAGE_SIZE;
931 len = ((to_write > len_max) ? len_max : to_write);
932 }
933
934 return tp_len;
935}
936
937static int tpacket_snd(struct packet_sock *po, struct msghdr *msg)
938{
939 struct socket *sock;
940 struct sk_buff *skb;
941 struct net_device *dev;
942 __be16 proto;
943 int ifindex, err, reserve = 0;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +0000944 void *ph;
945 struct sockaddr_ll *saddr = (struct sockaddr_ll *)msg->msg_name;
Johann Baudy69e3c752009-05-18 22:11:22 -0700946 int tp_len, size_max;
947 unsigned char *addr;
948 int len_sum = 0;
949 int status = 0;
950
951 sock = po->sk.sk_socket;
952
953 mutex_lock(&po->pg_vec_lock);
954
955 err = -EBUSY;
956 if (saddr == NULL) {
957 ifindex = po->ifindex;
958 proto = po->num;
959 addr = NULL;
960 } else {
961 err = -EINVAL;
962 if (msg->msg_namelen < sizeof(struct sockaddr_ll))
963 goto out;
964 if (msg->msg_namelen < (saddr->sll_halen
965 + offsetof(struct sockaddr_ll,
966 sll_addr)))
967 goto out;
968 ifindex = saddr->sll_ifindex;
969 proto = saddr->sll_protocol;
970 addr = saddr->sll_addr;
971 }
972
973 dev = dev_get_by_index(sock_net(&po->sk), ifindex);
974 err = -ENXIO;
975 if (unlikely(dev == NULL))
976 goto out;
977
978 reserve = dev->hard_header_len;
979
980 err = -ENETDOWN;
981 if (unlikely(!(dev->flags & IFF_UP)))
982 goto out_put;
983
984 size_max = po->tx_ring.frame_size
Gabor Gombasb5dd8842009-10-29 03:19:11 -0700985 - (po->tp_hdrlen - sizeof(struct sockaddr_ll));
Johann Baudy69e3c752009-05-18 22:11:22 -0700986
987 if (size_max > dev->mtu + reserve)
988 size_max = dev->mtu + reserve;
989
990 do {
991 ph = packet_current_frame(po, &po->tx_ring,
992 TP_STATUS_SEND_REQUEST);
993
994 if (unlikely(ph == NULL)) {
995 schedule();
996 continue;
997 }
998
999 status = TP_STATUS_SEND_REQUEST;
1000 skb = sock_alloc_send_skb(&po->sk,
1001 LL_ALLOCATED_SPACE(dev)
1002 + sizeof(struct sockaddr_ll),
1003 0, &err);
1004
1005 if (unlikely(skb == NULL))
1006 goto out_status;
1007
1008 tp_len = tpacket_fill_skb(po, skb, ph, dev, size_max, proto,
1009 addr);
1010
1011 if (unlikely(tp_len < 0)) {
1012 if (po->tp_loss) {
1013 __packet_set_status(po, ph,
1014 TP_STATUS_AVAILABLE);
1015 packet_increment_head(&po->tx_ring);
1016 kfree_skb(skb);
1017 continue;
1018 } else {
1019 status = TP_STATUS_WRONG_FORMAT;
1020 err = tp_len;
1021 goto out_status;
1022 }
1023 }
1024
1025 skb->destructor = tpacket_destruct_skb;
1026 __packet_set_status(po, ph, TP_STATUS_SENDING);
1027 atomic_inc(&po->tx_ring.pending);
1028
1029 status = TP_STATUS_SEND_REQUEST;
1030 err = dev_queue_xmit(skb);
1031 if (unlikely(err > 0 && (err = net_xmit_errno(err)) != 0))
1032 goto out_xmit;
1033 packet_increment_head(&po->tx_ring);
1034 len_sum += tp_len;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001035 } while (likely((ph != NULL) || ((!(msg->msg_flags & MSG_DONTWAIT))
Johann Baudy69e3c752009-05-18 22:11:22 -07001036 && (atomic_read(&po->tx_ring.pending))))
1037 );
1038
1039 err = len_sum;
1040 goto out_put;
1041
1042out_xmit:
1043 skb->destructor = sock_wfree;
1044 atomic_dec(&po->tx_ring.pending);
1045out_status:
1046 __packet_set_status(po, ph, status);
1047 kfree_skb(skb);
1048out_put:
1049 dev_put(dev);
1050out:
1051 mutex_unlock(&po->pg_vec_lock);
1052 return err;
1053}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054#endif
1055
Johann Baudy69e3c752009-05-18 22:11:22 -07001056static int packet_snd(struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 struct msghdr *msg, size_t len)
1058{
1059 struct sock *sk = sock->sk;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001060 struct sockaddr_ll *saddr = (struct sockaddr_ll *)msg->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 struct sk_buff *skb;
1062 struct net_device *dev;
Al Viro0e11c912006-11-08 00:26:29 -08001063 __be16 proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 unsigned char *addr;
1065 int ifindex, err, reserve = 0;
1066
1067 /*
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09001068 * Get and verify the address.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 */
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09001070
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 if (saddr == NULL) {
1072 struct packet_sock *po = pkt_sk(sk);
1073
1074 ifindex = po->ifindex;
1075 proto = po->num;
1076 addr = NULL;
1077 } else {
1078 err = -EINVAL;
1079 if (msg->msg_namelen < sizeof(struct sockaddr_ll))
1080 goto out;
Eric W. Biederman0fb375f2005-09-21 00:11:37 -07001081 if (msg->msg_namelen < (saddr->sll_halen + offsetof(struct sockaddr_ll, sll_addr)))
1082 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 ifindex = saddr->sll_ifindex;
1084 proto = saddr->sll_protocol;
1085 addr = saddr->sll_addr;
1086 }
1087
1088
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001089 dev = dev_get_by_index(sock_net(sk), ifindex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 err = -ENXIO;
1091 if (dev == NULL)
1092 goto out_unlock;
1093 if (sock->type == SOCK_RAW)
1094 reserve = dev->hard_header_len;
1095
David S. Millerd5e76b02007-01-25 19:30:36 -08001096 err = -ENETDOWN;
1097 if (!(dev->flags & IFF_UP))
1098 goto out_unlock;
1099
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 err = -EMSGSIZE;
1101 if (len > dev->mtu+reserve)
1102 goto out_unlock;
1103
Johannes Bergf5184d22008-05-12 20:48:31 -07001104 skb = sock_alloc_send_skb(sk, len + LL_ALLOCATED_SPACE(dev),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 msg->msg_flags & MSG_DONTWAIT, &err);
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001106 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 goto out_unlock;
1108
1109 skb_reserve(skb, LL_RESERVED_SPACE(dev));
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001110 skb_reset_network_header(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07001112 err = -EINVAL;
1113 if (sock->type == SOCK_DGRAM &&
1114 dev_hard_header(skb, dev, ntohs(proto), addr, NULL, len) < 0)
1115 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116
1117 /* Returns -EFAULT on error */
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001118 err = memcpy_fromiovec(skb_put(skb, len), msg->msg_iov, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 if (err)
1120 goto out_free;
1121
1122 skb->protocol = proto;
1123 skb->dev = dev;
1124 skb->priority = sk->sk_priority;
1125
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 /*
1127 * Now send it
1128 */
1129
1130 err = dev_queue_xmit(skb);
1131 if (err > 0 && (err = net_xmit_errno(err)) != 0)
1132 goto out_unlock;
1133
1134 dev_put(dev);
1135
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001136 return len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137
1138out_free:
1139 kfree_skb(skb);
1140out_unlock:
1141 if (dev)
1142 dev_put(dev);
1143out:
1144 return err;
1145}
1146
Johann Baudy69e3c752009-05-18 22:11:22 -07001147static int packet_sendmsg(struct kiocb *iocb, struct socket *sock,
1148 struct msghdr *msg, size_t len)
1149{
1150#ifdef CONFIG_PACKET_MMAP
1151 struct sock *sk = sock->sk;
1152 struct packet_sock *po = pkt_sk(sk);
1153 if (po->tx_ring.pg_vec)
1154 return tpacket_snd(po, msg);
1155 else
1156#endif
1157 return packet_snd(sock, msg, len);
1158}
1159
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160/*
1161 * Close a PACKET socket. This is fairly simple. We immediately go
1162 * to 'closed' state and remove our protocol entry in the device list.
1163 */
1164
1165static int packet_release(struct socket *sock)
1166{
1167 struct sock *sk = sock->sk;
1168 struct packet_sock *po;
Denis V. Lunevd12d01d2007-11-19 22:28:35 -08001169 struct net *net;
Johann Baudy69e3c752009-05-18 22:11:22 -07001170#ifdef CONFIG_PACKET_MMAP
1171 struct tpacket_req req;
1172#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173
1174 if (!sk)
1175 return 0;
1176
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001177 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 po = pkt_sk(sk);
1179
Denis V. Lunev2aaef4e2007-12-11 04:19:54 -08001180 write_lock_bh(&net->packet.sklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 sk_del_node_init(sk);
Eric Dumazet920de802008-11-24 00:09:29 -08001182 sock_prot_inuse_add(net, sk->sk_prot, -1);
Denis V. Lunev2aaef4e2007-12-11 04:19:54 -08001183 write_unlock_bh(&net->packet.sklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184
1185 /*
1186 * Unhook packet receive handler.
1187 */
1188
1189 if (po->running) {
1190 /*
1191 * Remove the protocol hook
1192 */
1193 dev_remove_pack(&po->prot_hook);
1194 po->running = 0;
1195 po->num = 0;
1196 __sock_put(sk);
1197 }
1198
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 packet_flush_mclist(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200
1201#ifdef CONFIG_PACKET_MMAP
Johann Baudy69e3c752009-05-18 22:11:22 -07001202 memset(&req, 0, sizeof(req));
1203
1204 if (po->rx_ring.pg_vec)
1205 packet_set_ring(sk, &req, 1, 0);
1206
1207 if (po->tx_ring.pg_vec)
1208 packet_set_ring(sk, &req, 1, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209#endif
1210
1211 /*
1212 * Now the socket is dead. No more input will appear.
1213 */
1214
1215 sock_orphan(sk);
1216 sock->sk = NULL;
1217
1218 /* Purge queues */
1219
1220 skb_queue_purge(&sk->sk_receive_queue);
Pavel Emelyanov17ab56a2007-11-10 21:38:48 -08001221 sk_refcnt_debug_release(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222
1223 sock_put(sk);
1224 return 0;
1225}
1226
1227/*
1228 * Attach a packet hook.
1229 */
1230
Al Viro0e11c912006-11-08 00:26:29 -08001231static int packet_do_bind(struct sock *sk, struct net_device *dev, __be16 protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232{
1233 struct packet_sock *po = pkt_sk(sk);
1234 /*
1235 * Detach an existing hook if present.
1236 */
1237
1238 lock_sock(sk);
1239
1240 spin_lock(&po->bind_lock);
1241 if (po->running) {
1242 __sock_put(sk);
1243 po->running = 0;
1244 po->num = 0;
1245 spin_unlock(&po->bind_lock);
1246 dev_remove_pack(&po->prot_hook);
1247 spin_lock(&po->bind_lock);
1248 }
1249
1250 po->num = protocol;
1251 po->prot_hook.type = protocol;
1252 po->prot_hook.dev = dev;
1253
1254 po->ifindex = dev ? dev->ifindex : 0;
1255
1256 if (protocol == 0)
1257 goto out_unlock;
1258
Urs Thuermannbe85d4a2007-11-12 21:05:20 -08001259 if (!dev || (dev->flags & IFF_UP)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 dev_add_pack(&po->prot_hook);
1261 sock_hold(sk);
1262 po->running = 1;
Urs Thuermannbe85d4a2007-11-12 21:05:20 -08001263 } else {
1264 sk->sk_err = ENETDOWN;
1265 if (!sock_flag(sk, SOCK_DEAD))
1266 sk->sk_error_report(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 }
1268
1269out_unlock:
1270 spin_unlock(&po->bind_lock);
1271 release_sock(sk);
1272 return 0;
1273}
1274
1275/*
1276 * Bind a packet socket to a device
1277 */
1278
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001279static int packet_bind_spkt(struct socket *sock, struct sockaddr *uaddr,
1280 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281{
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001282 struct sock *sk = sock->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 char name[15];
1284 struct net_device *dev;
1285 int err = -ENODEV;
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09001286
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 /*
1288 * Check legality
1289 */
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09001290
Kris Katterjohn8ae55f02006-01-23 16:28:02 -08001291 if (addr_len != sizeof(struct sockaddr))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 return -EINVAL;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001293 strlcpy(name, uaddr->sa_data, sizeof(name));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001295 dev = dev_get_by_name(sock_net(sk), name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 if (dev) {
1297 err = packet_do_bind(sk, dev, pkt_sk(sk)->num);
1298 dev_put(dev);
1299 }
1300 return err;
1301}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302
1303static int packet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
1304{
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001305 struct sockaddr_ll *sll = (struct sockaddr_ll *)uaddr;
1306 struct sock *sk = sock->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 struct net_device *dev = NULL;
1308 int err;
1309
1310
1311 /*
1312 * Check legality
1313 */
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09001314
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 if (addr_len < sizeof(struct sockaddr_ll))
1316 return -EINVAL;
1317 if (sll->sll_family != AF_PACKET)
1318 return -EINVAL;
1319
1320 if (sll->sll_ifindex) {
1321 err = -ENODEV;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001322 dev = dev_get_by_index(sock_net(sk), sll->sll_ifindex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 if (dev == NULL)
1324 goto out;
1325 }
1326 err = packet_do_bind(sk, dev, sll->sll_protocol ? : pkt_sk(sk)->num);
1327 if (dev)
1328 dev_put(dev);
1329
1330out:
1331 return err;
1332}
1333
1334static struct proto packet_proto = {
1335 .name = "PACKET",
1336 .owner = THIS_MODULE,
1337 .obj_size = sizeof(struct packet_sock),
1338};
1339
1340/*
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09001341 * Create a packet of type SOCK_PACKET.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 */
1343
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001344static int packet_create(struct net *net, struct socket *sock, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345{
1346 struct sock *sk;
1347 struct packet_sock *po;
Al Viro0e11c912006-11-08 00:26:29 -08001348 __be16 proto = (__force __be16)protocol; /* weird, but documented */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 int err;
1350
1351 if (!capable(CAP_NET_RAW))
1352 return -EPERM;
David S. Millerbe020972007-05-29 13:16:31 -07001353 if (sock->type != SOCK_DGRAM && sock->type != SOCK_RAW &&
1354 sock->type != SOCK_PACKET)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355 return -ESOCKTNOSUPPORT;
1356
1357 sock->state = SS_UNCONNECTED;
1358
1359 err = -ENOBUFS;
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001360 sk = sk_alloc(net, PF_PACKET, GFP_KERNEL, &packet_proto);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 if (sk == NULL)
1362 goto out;
1363
1364 sock->ops = &packet_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 if (sock->type == SOCK_PACKET)
1366 sock->ops = &packet_ops_spkt;
David S. Millerbe020972007-05-29 13:16:31 -07001367
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 sock_init_data(sock, sk);
1369
1370 po = pkt_sk(sk);
1371 sk->sk_family = PF_PACKET;
Al Viro0e11c912006-11-08 00:26:29 -08001372 po->num = proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373
1374 sk->sk_destruct = packet_sock_destruct;
Pavel Emelyanov17ab56a2007-11-10 21:38:48 -08001375 sk_refcnt_debug_inc(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376
1377 /*
1378 * Attach a protocol block
1379 */
1380
1381 spin_lock_init(&po->bind_lock);
Herbert Xu905db442009-01-30 14:12:06 -08001382 mutex_init(&po->pg_vec_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 po->prot_hook.func = packet_rcv;
David S. Millerbe020972007-05-29 13:16:31 -07001384
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 if (sock->type == SOCK_PACKET)
1386 po->prot_hook.func = packet_rcv_spkt;
David S. Millerbe020972007-05-29 13:16:31 -07001387
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 po->prot_hook.af_packet_priv = sk;
1389
Al Viro0e11c912006-11-08 00:26:29 -08001390 if (proto) {
1391 po->prot_hook.type = proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 dev_add_pack(&po->prot_hook);
1393 sock_hold(sk);
1394 po->running = 1;
1395 }
1396
Denis V. Lunev2aaef4e2007-12-11 04:19:54 -08001397 write_lock_bh(&net->packet.sklist_lock);
1398 sk_add_node(sk, &net->packet.sklist);
Eric Dumazet36804532008-11-19 14:25:35 -08001399 sock_prot_inuse_add(net, &packet_proto, 1);
Eric Dumazet920de802008-11-24 00:09:29 -08001400 write_unlock_bh(&net->packet.sklist_lock);
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001401 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402out:
1403 return err;
1404}
1405
1406/*
1407 * Pull a packet from our receive queue and hand it to the user.
1408 * If necessary we block.
1409 */
1410
1411static int packet_recvmsg(struct kiocb *iocb, struct socket *sock,
1412 struct msghdr *msg, size_t len, int flags)
1413{
1414 struct sock *sk = sock->sk;
1415 struct sk_buff *skb;
1416 int copied, err;
Eric W. Biederman0fb375f2005-09-21 00:11:37 -07001417 struct sockaddr_ll *sll;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
1419 err = -EINVAL;
1420 if (flags & ~(MSG_PEEK|MSG_DONTWAIT|MSG_TRUNC|MSG_CMSG_COMPAT))
1421 goto out;
1422
1423#if 0
1424 /* What error should we return now? EUNATTACH? */
1425 if (pkt_sk(sk)->ifindex < 0)
1426 return -ENODEV;
1427#endif
1428
1429 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 * Call the generic datagram receiver. This handles all sorts
1431 * of horrible races and re-entrancy so we can forget about it
1432 * in the protocol layers.
1433 *
1434 * Now it will return ENETDOWN, if device have just gone down,
1435 * but then it will block.
1436 */
1437
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001438 skb = skb_recv_datagram(sk, flags, flags & MSG_DONTWAIT, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439
1440 /*
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09001441 * An error occurred so return it. Because skb_recv_datagram()
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 * handles the blocking we don't see and worry about blocking
1443 * retries.
1444 */
1445
Kris Katterjohn8ae55f02006-01-23 16:28:02 -08001446 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 goto out;
1448
1449 /*
Eric W. Biederman0fb375f2005-09-21 00:11:37 -07001450 * If the address length field is there to be filled in, we fill
1451 * it in now.
1452 */
1453
Herbert Xuffbc6112007-02-04 23:33:10 -08001454 sll = &PACKET_SKB_CB(skb)->sa.ll;
Eric W. Biederman0fb375f2005-09-21 00:11:37 -07001455 if (sock->type == SOCK_PACKET)
1456 msg->msg_namelen = sizeof(struct sockaddr_pkt);
1457 else
1458 msg->msg_namelen = sll->sll_halen + offsetof(struct sockaddr_ll, sll_addr);
1459
1460 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 * You lose any data beyond the buffer you gave. If it worries a
1462 * user program they can ask the device for its MTU anyway.
1463 */
1464
1465 copied = skb->len;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001466 if (copied > len) {
1467 copied = len;
1468 msg->msg_flags |= MSG_TRUNC;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 }
1470
1471 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1472 if (err)
1473 goto out_free;
1474
1475 sock_recv_timestamp(msg, sk, skb);
1476
1477 if (msg->msg_name)
Herbert Xuffbc6112007-02-04 23:33:10 -08001478 memcpy(msg->msg_name, &PACKET_SKB_CB(skb)->sa,
1479 msg->msg_namelen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480
Herbert Xu8dc41942007-02-04 23:31:32 -08001481 if (pkt_sk(sk)->auxdata) {
Herbert Xuffbc6112007-02-04 23:33:10 -08001482 struct tpacket_auxdata aux;
1483
1484 aux.tp_status = TP_STATUS_USER;
1485 if (skb->ip_summed == CHECKSUM_PARTIAL)
1486 aux.tp_status |= TP_STATUS_CSUMNOTREADY;
1487 aux.tp_len = PACKET_SKB_CB(skb)->origlen;
1488 aux.tp_snaplen = skb->len;
1489 aux.tp_mac = 0;
Arnaldo Carvalho de Melobbe735e2007-03-10 22:16:10 -03001490 aux.tp_net = skb_network_offset(skb);
Patrick McHardy393e52e2008-07-14 22:50:39 -07001491 aux.tp_vlan_tci = skb->vlan_tci;
Herbert Xuffbc6112007-02-04 23:33:10 -08001492
1493 put_cmsg(msg, SOL_PACKET, PACKET_AUXDATA, sizeof(aux), &aux);
Herbert Xu8dc41942007-02-04 23:31:32 -08001494 }
1495
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 /*
1497 * Free or return the buffer as appropriate. Again this
1498 * hides all the races and re-entrancy issues from us.
1499 */
1500 err = (flags&MSG_TRUNC) ? skb->len : copied;
1501
1502out_free:
1503 skb_free_datagram(sk, skb);
1504out:
1505 return err;
1506}
1507
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508static int packet_getname_spkt(struct socket *sock, struct sockaddr *uaddr,
1509 int *uaddr_len, int peer)
1510{
1511 struct net_device *dev;
1512 struct sock *sk = sock->sk;
1513
1514 if (peer)
1515 return -EOPNOTSUPP;
1516
1517 uaddr->sa_family = AF_PACKET;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001518 dev = dev_get_by_index(sock_net(sk), pkt_sk(sk)->ifindex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 if (dev) {
1520 strlcpy(uaddr->sa_data, dev->name, 15);
1521 dev_put(dev);
1522 } else
1523 memset(uaddr->sa_data, 0, 14);
1524 *uaddr_len = sizeof(*uaddr);
1525
1526 return 0;
1527}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528
1529static int packet_getname(struct socket *sock, struct sockaddr *uaddr,
1530 int *uaddr_len, int peer)
1531{
1532 struct net_device *dev;
1533 struct sock *sk = sock->sk;
1534 struct packet_sock *po = pkt_sk(sk);
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001535 struct sockaddr_ll *sll = (struct sockaddr_ll *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536
1537 if (peer)
1538 return -EOPNOTSUPP;
1539
1540 sll->sll_family = AF_PACKET;
1541 sll->sll_ifindex = po->ifindex;
1542 sll->sll_protocol = po->num;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001543 dev = dev_get_by_index(sock_net(sk), po->ifindex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 if (dev) {
1545 sll->sll_hatype = dev->type;
1546 sll->sll_halen = dev->addr_len;
1547 memcpy(sll->sll_addr, dev->dev_addr, dev->addr_len);
1548 dev_put(dev);
1549 } else {
1550 sll->sll_hatype = 0; /* Bad: we have no ARPHRD_UNSPEC */
1551 sll->sll_halen = 0;
1552 }
Eric W. Biederman0fb375f2005-09-21 00:11:37 -07001553 *uaddr_len = offsetof(struct sockaddr_ll, sll_addr) + sll->sll_halen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554
1555 return 0;
1556}
1557
Wang Chen2aeb0b82008-07-14 20:49:46 -07001558static int packet_dev_mc(struct net_device *dev, struct packet_mclist *i,
1559 int what)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560{
1561 switch (i->type) {
1562 case PACKET_MR_MULTICAST:
1563 if (what > 0)
Eric W. Biedermand95ed922009-05-19 18:27:17 +00001564 return dev_mc_add(dev, i->addr, i->alen, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 else
Eric W. Biedermand95ed922009-05-19 18:27:17 +00001566 return dev_mc_delete(dev, i->addr, i->alen, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 break;
1568 case PACKET_MR_PROMISC:
Wang Chen2aeb0b82008-07-14 20:49:46 -07001569 return dev_set_promiscuity(dev, what);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 break;
1571 case PACKET_MR_ALLMULTI:
Wang Chen2aeb0b82008-07-14 20:49:46 -07001572 return dev_set_allmulti(dev, what);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 break;
Eric W. Biedermand95ed922009-05-19 18:27:17 +00001574 case PACKET_MR_UNICAST:
1575 if (what > 0)
Jiri Pirkoccffad252009-05-22 23:22:17 +00001576 return dev_unicast_add(dev, i->addr);
Eric W. Biedermand95ed922009-05-19 18:27:17 +00001577 else
Jiri Pirkoccffad252009-05-22 23:22:17 +00001578 return dev_unicast_delete(dev, i->addr);
Eric W. Biedermand95ed922009-05-19 18:27:17 +00001579 break;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001580 default:
1581 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 }
Wang Chen2aeb0b82008-07-14 20:49:46 -07001583 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584}
1585
1586static void packet_dev_mclist(struct net_device *dev, struct packet_mclist *i, int what)
1587{
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001588 for ( ; i; i = i->next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 if (i->ifindex == dev->ifindex)
1590 packet_dev_mc(dev, i, what);
1591 }
1592}
1593
Eric W. Biederman0fb375f2005-09-21 00:11:37 -07001594static int packet_mc_add(struct sock *sk, struct packet_mreq_max *mreq)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595{
1596 struct packet_sock *po = pkt_sk(sk);
1597 struct packet_mclist *ml, *i;
1598 struct net_device *dev;
1599 int err;
1600
1601 rtnl_lock();
1602
1603 err = -ENODEV;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001604 dev = __dev_get_by_index(sock_net(sk), mreq->mr_ifindex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 if (!dev)
1606 goto done;
1607
1608 err = -EINVAL;
1609 if (mreq->mr_alen > dev->addr_len)
1610 goto done;
1611
1612 err = -ENOBUFS;
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001613 i = kmalloc(sizeof(*i), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 if (i == NULL)
1615 goto done;
1616
1617 err = 0;
1618 for (ml = po->mclist; ml; ml = ml->next) {
1619 if (ml->ifindex == mreq->mr_ifindex &&
1620 ml->type == mreq->mr_type &&
1621 ml->alen == mreq->mr_alen &&
1622 memcmp(ml->addr, mreq->mr_address, ml->alen) == 0) {
1623 ml->count++;
1624 /* Free the new element ... */
1625 kfree(i);
1626 goto done;
1627 }
1628 }
1629
1630 i->type = mreq->mr_type;
1631 i->ifindex = mreq->mr_ifindex;
1632 i->alen = mreq->mr_alen;
1633 memcpy(i->addr, mreq->mr_address, i->alen);
1634 i->count = 1;
1635 i->next = po->mclist;
1636 po->mclist = i;
Wang Chen2aeb0b82008-07-14 20:49:46 -07001637 err = packet_dev_mc(dev, i, 1);
1638 if (err) {
1639 po->mclist = i->next;
1640 kfree(i);
1641 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642
1643done:
1644 rtnl_unlock();
1645 return err;
1646}
1647
Eric W. Biederman0fb375f2005-09-21 00:11:37 -07001648static int packet_mc_drop(struct sock *sk, struct packet_mreq_max *mreq)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649{
1650 struct packet_mclist *ml, **mlp;
1651
1652 rtnl_lock();
1653
1654 for (mlp = &pkt_sk(sk)->mclist; (ml = *mlp) != NULL; mlp = &ml->next) {
1655 if (ml->ifindex == mreq->mr_ifindex &&
1656 ml->type == mreq->mr_type &&
1657 ml->alen == mreq->mr_alen &&
1658 memcmp(ml->addr, mreq->mr_address, ml->alen) == 0) {
1659 if (--ml->count == 0) {
1660 struct net_device *dev;
1661 *mlp = ml->next;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001662 dev = dev_get_by_index(sock_net(sk), ml->ifindex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 if (dev) {
1664 packet_dev_mc(dev, ml, -1);
1665 dev_put(dev);
1666 }
1667 kfree(ml);
1668 }
1669 rtnl_unlock();
1670 return 0;
1671 }
1672 }
1673 rtnl_unlock();
1674 return -EADDRNOTAVAIL;
1675}
1676
1677static void packet_flush_mclist(struct sock *sk)
1678{
1679 struct packet_sock *po = pkt_sk(sk);
1680 struct packet_mclist *ml;
1681
1682 if (!po->mclist)
1683 return;
1684
1685 rtnl_lock();
1686 while ((ml = po->mclist) != NULL) {
1687 struct net_device *dev;
1688
1689 po->mclist = ml->next;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001690 dev = dev_get_by_index(sock_net(sk), ml->ifindex);
1691 if (dev != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 packet_dev_mc(dev, ml, -1);
1693 dev_put(dev);
1694 }
1695 kfree(ml);
1696 }
1697 rtnl_unlock();
1698}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699
1700static int
David S. Millerb7058842009-09-30 16:12:20 -07001701packet_setsockopt(struct socket *sock, int level, int optname, char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702{
1703 struct sock *sk = sock->sk;
Herbert Xu8dc41942007-02-04 23:31:32 -08001704 struct packet_sock *po = pkt_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705 int ret;
1706
1707 if (level != SOL_PACKET)
1708 return -ENOPROTOOPT;
1709
Johann Baudy69e3c752009-05-18 22:11:22 -07001710 switch (optname) {
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09001711 case PACKET_ADD_MEMBERSHIP:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 case PACKET_DROP_MEMBERSHIP:
1713 {
Eric W. Biederman0fb375f2005-09-21 00:11:37 -07001714 struct packet_mreq_max mreq;
1715 int len = optlen;
1716 memset(&mreq, 0, sizeof(mreq));
1717 if (len < sizeof(struct packet_mreq))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 return -EINVAL;
Eric W. Biederman0fb375f2005-09-21 00:11:37 -07001719 if (len > sizeof(mreq))
1720 len = sizeof(mreq);
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001721 if (copy_from_user(&mreq, optval, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 return -EFAULT;
Eric W. Biederman0fb375f2005-09-21 00:11:37 -07001723 if (len < (mreq.mr_alen + offsetof(struct packet_mreq, mr_address)))
1724 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 if (optname == PACKET_ADD_MEMBERSHIP)
1726 ret = packet_mc_add(sk, &mreq);
1727 else
1728 ret = packet_mc_drop(sk, &mreq);
1729 return ret;
1730 }
David S. Millera2efcfa2007-05-29 13:12:50 -07001731
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732#ifdef CONFIG_PACKET_MMAP
1733 case PACKET_RX_RING:
Johann Baudy69e3c752009-05-18 22:11:22 -07001734 case PACKET_TX_RING:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 {
1736 struct tpacket_req req;
1737
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001738 if (optlen < sizeof(req))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 return -EINVAL;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001740 if (copy_from_user(&req, optval, sizeof(req)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 return -EFAULT;
Johann Baudy69e3c752009-05-18 22:11:22 -07001742 return packet_set_ring(sk, &req, 0, optname == PACKET_TX_RING);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 }
1744 case PACKET_COPY_THRESH:
1745 {
1746 int val;
1747
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001748 if (optlen != sizeof(val))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749 return -EINVAL;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001750 if (copy_from_user(&val, optval, sizeof(val)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 return -EFAULT;
1752
1753 pkt_sk(sk)->copy_thresh = val;
1754 return 0;
1755 }
Patrick McHardybbd6ef82008-07-14 22:50:15 -07001756 case PACKET_VERSION:
1757 {
1758 int val;
1759
1760 if (optlen != sizeof(val))
1761 return -EINVAL;
Johann Baudy69e3c752009-05-18 22:11:22 -07001762 if (po->rx_ring.pg_vec || po->tx_ring.pg_vec)
Patrick McHardybbd6ef82008-07-14 22:50:15 -07001763 return -EBUSY;
1764 if (copy_from_user(&val, optval, sizeof(val)))
1765 return -EFAULT;
1766 switch (val) {
1767 case TPACKET_V1:
1768 case TPACKET_V2:
1769 po->tp_version = val;
1770 return 0;
1771 default:
1772 return -EINVAL;
1773 }
1774 }
Patrick McHardy8913336a2008-07-18 18:05:19 -07001775 case PACKET_RESERVE:
1776 {
1777 unsigned int val;
1778
1779 if (optlen != sizeof(val))
1780 return -EINVAL;
Johann Baudy69e3c752009-05-18 22:11:22 -07001781 if (po->rx_ring.pg_vec || po->tx_ring.pg_vec)
Patrick McHardy8913336a2008-07-18 18:05:19 -07001782 return -EBUSY;
1783 if (copy_from_user(&val, optval, sizeof(val)))
1784 return -EFAULT;
1785 po->tp_reserve = val;
1786 return 0;
1787 }
Johann Baudy69e3c752009-05-18 22:11:22 -07001788 case PACKET_LOSS:
1789 {
1790 unsigned int val;
1791
1792 if (optlen != sizeof(val))
1793 return -EINVAL;
1794 if (po->rx_ring.pg_vec || po->tx_ring.pg_vec)
1795 return -EBUSY;
1796 if (copy_from_user(&val, optval, sizeof(val)))
1797 return -EFAULT;
1798 po->tp_loss = !!val;
1799 return 0;
1800 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801#endif
Herbert Xu8dc41942007-02-04 23:31:32 -08001802 case PACKET_AUXDATA:
1803 {
1804 int val;
1805
1806 if (optlen < sizeof(val))
1807 return -EINVAL;
1808 if (copy_from_user(&val, optval, sizeof(val)))
1809 return -EFAULT;
1810
1811 po->auxdata = !!val;
1812 return 0;
1813 }
Peter P. Waskiewicz Jr80feaac2007-04-20 16:05:39 -07001814 case PACKET_ORIGDEV:
1815 {
1816 int val;
1817
1818 if (optlen < sizeof(val))
1819 return -EINVAL;
1820 if (copy_from_user(&val, optval, sizeof(val)))
1821 return -EFAULT;
1822
1823 po->origdev = !!val;
1824 return 0;
1825 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 default:
1827 return -ENOPROTOOPT;
1828 }
1829}
1830
1831static int packet_getsockopt(struct socket *sock, int level, int optname,
1832 char __user *optval, int __user *optlen)
1833{
1834 int len;
Herbert Xu8dc41942007-02-04 23:31:32 -08001835 int val;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 struct sock *sk = sock->sk;
1837 struct packet_sock *po = pkt_sk(sk);
Herbert Xu8dc41942007-02-04 23:31:32 -08001838 void *data;
1839 struct tpacket_stats st;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840
1841 if (level != SOL_PACKET)
1842 return -ENOPROTOOPT;
1843
Kris Katterjohn8ae55f02006-01-23 16:28:02 -08001844 if (get_user(len, optlen))
1845 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846
1847 if (len < 0)
1848 return -EINVAL;
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09001849
Johann Baudy69e3c752009-05-18 22:11:22 -07001850 switch (optname) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 case PACKET_STATISTICS:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 if (len > sizeof(struct tpacket_stats))
1853 len = sizeof(struct tpacket_stats);
1854 spin_lock_bh(&sk->sk_receive_queue.lock);
1855 st = po->stats;
1856 memset(&po->stats, 0, sizeof(st));
1857 spin_unlock_bh(&sk->sk_receive_queue.lock);
1858 st.tp_packets += st.tp_drops;
1859
Herbert Xu8dc41942007-02-04 23:31:32 -08001860 data = &st;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 break;
Herbert Xu8dc41942007-02-04 23:31:32 -08001862 case PACKET_AUXDATA:
1863 if (len > sizeof(int))
1864 len = sizeof(int);
1865 val = po->auxdata;
1866
1867 data = &val;
1868 break;
Peter P. Waskiewicz Jr80feaac2007-04-20 16:05:39 -07001869 case PACKET_ORIGDEV:
1870 if (len > sizeof(int))
1871 len = sizeof(int);
1872 val = po->origdev;
1873
1874 data = &val;
1875 break;
Patrick McHardybbd6ef82008-07-14 22:50:15 -07001876#ifdef CONFIG_PACKET_MMAP
1877 case PACKET_VERSION:
1878 if (len > sizeof(int))
1879 len = sizeof(int);
1880 val = po->tp_version;
1881 data = &val;
1882 break;
1883 case PACKET_HDRLEN:
1884 if (len > sizeof(int))
1885 len = sizeof(int);
1886 if (copy_from_user(&val, optval, len))
1887 return -EFAULT;
1888 switch (val) {
1889 case TPACKET_V1:
1890 val = sizeof(struct tpacket_hdr);
1891 break;
1892 case TPACKET_V2:
1893 val = sizeof(struct tpacket2_hdr);
1894 break;
1895 default:
1896 return -EINVAL;
1897 }
1898 data = &val;
1899 break;
Patrick McHardy8913336a2008-07-18 18:05:19 -07001900 case PACKET_RESERVE:
1901 if (len > sizeof(unsigned int))
1902 len = sizeof(unsigned int);
1903 val = po->tp_reserve;
1904 data = &val;
1905 break;
Johann Baudy69e3c752009-05-18 22:11:22 -07001906 case PACKET_LOSS:
1907 if (len > sizeof(unsigned int))
1908 len = sizeof(unsigned int);
1909 val = po->tp_loss;
1910 data = &val;
1911 break;
Patrick McHardybbd6ef82008-07-14 22:50:15 -07001912#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 default:
1914 return -ENOPROTOOPT;
1915 }
1916
Kris Katterjohn8ae55f02006-01-23 16:28:02 -08001917 if (put_user(len, optlen))
1918 return -EFAULT;
Herbert Xu8dc41942007-02-04 23:31:32 -08001919 if (copy_to_user(optval, data, len))
1920 return -EFAULT;
Kris Katterjohn8ae55f02006-01-23 16:28:02 -08001921 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922}
1923
1924
1925static int packet_notifier(struct notifier_block *this, unsigned long msg, void *data)
1926{
1927 struct sock *sk;
1928 struct hlist_node *node;
Jason Lunzad930652007-02-20 23:19:54 -08001929 struct net_device *dev = data;
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001930 struct net *net = dev_net(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931
Denis V. Lunev2aaef4e2007-12-11 04:19:54 -08001932 read_lock(&net->packet.sklist_lock);
1933 sk_for_each(sk, node, &net->packet.sklist) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 struct packet_sock *po = pkt_sk(sk);
1935
1936 switch (msg) {
1937 case NETDEV_UNREGISTER:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938 if (po->mclist)
1939 packet_dev_mclist(dev, po->mclist, -1);
David S. Millera2efcfa2007-05-29 13:12:50 -07001940 /* fallthrough */
1941
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942 case NETDEV_DOWN:
1943 if (dev->ifindex == po->ifindex) {
1944 spin_lock(&po->bind_lock);
1945 if (po->running) {
1946 __dev_remove_pack(&po->prot_hook);
1947 __sock_put(sk);
1948 po->running = 0;
1949 sk->sk_err = ENETDOWN;
1950 if (!sock_flag(sk, SOCK_DEAD))
1951 sk->sk_error_report(sk);
1952 }
1953 if (msg == NETDEV_UNREGISTER) {
1954 po->ifindex = -1;
1955 po->prot_hook.dev = NULL;
1956 }
1957 spin_unlock(&po->bind_lock);
1958 }
1959 break;
1960 case NETDEV_UP:
1961 spin_lock(&po->bind_lock);
1962 if (dev->ifindex == po->ifindex && po->num &&
1963 !po->running) {
1964 dev_add_pack(&po->prot_hook);
1965 sock_hold(sk);
1966 po->running = 1;
1967 }
1968 spin_unlock(&po->bind_lock);
1969 break;
1970 }
1971 }
Denis V. Lunev2aaef4e2007-12-11 04:19:54 -08001972 read_unlock(&net->packet.sklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973 return NOTIFY_DONE;
1974}
1975
1976
1977static int packet_ioctl(struct socket *sock, unsigned int cmd,
1978 unsigned long arg)
1979{
1980 struct sock *sk = sock->sk;
1981
Johann Baudy69e3c752009-05-18 22:11:22 -07001982 switch (cmd) {
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001983 case SIOCOUTQ:
1984 {
1985 int amount = sk_wmem_alloc_get(sk);
Eric Dumazet31e6d362009-06-17 19:05:41 -07001986
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001987 return put_user(amount, (int __user *)arg);
1988 }
1989 case SIOCINQ:
1990 {
1991 struct sk_buff *skb;
1992 int amount = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00001994 spin_lock_bh(&sk->sk_receive_queue.lock);
1995 skb = skb_peek(&sk->sk_receive_queue);
1996 if (skb)
1997 amount = skb->len;
1998 spin_unlock_bh(&sk->sk_receive_queue.lock);
1999 return put_user(amount, (int __user *)arg);
2000 }
2001 case SIOCGSTAMP:
2002 return sock_get_timestamp(sk, (struct timeval __user *)arg);
2003 case SIOCGSTAMPNS:
2004 return sock_get_timestampns(sk, (struct timespec __user *)arg);
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09002005
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006#ifdef CONFIG_INET
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002007 case SIOCADDRT:
2008 case SIOCDELRT:
2009 case SIOCDARP:
2010 case SIOCGARP:
2011 case SIOCSARP:
2012 case SIOCGIFADDR:
2013 case SIOCSIFADDR:
2014 case SIOCGIFBRDADDR:
2015 case SIOCSIFBRDADDR:
2016 case SIOCGIFNETMASK:
2017 case SIOCSIFNETMASK:
2018 case SIOCGIFDSTADDR:
2019 case SIOCSIFDSTADDR:
2020 case SIOCSIFFLAGS:
2021 if (!net_eq(sock_net(sk), &init_net))
2022 return -ENOIOCTLCMD;
2023 return inet_dgram_ops.ioctl(sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024#endif
2025
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002026 default:
2027 return -ENOIOCTLCMD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 }
2029 return 0;
2030}
2031
2032#ifndef CONFIG_PACKET_MMAP
2033#define packet_mmap sock_no_mmap
2034#define packet_poll datagram_poll
2035#else
2036
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002037static unsigned int packet_poll(struct file *file, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038 poll_table *wait)
2039{
2040 struct sock *sk = sock->sk;
2041 struct packet_sock *po = pkt_sk(sk);
2042 unsigned int mask = datagram_poll(file, sock, wait);
2043
2044 spin_lock_bh(&sk->sk_receive_queue.lock);
Johann Baudy69e3c752009-05-18 22:11:22 -07002045 if (po->rx_ring.pg_vec) {
2046 if (!packet_previous_frame(po, &po->rx_ring, TP_STATUS_KERNEL))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047 mask |= POLLIN | POLLRDNORM;
2048 }
2049 spin_unlock_bh(&sk->sk_receive_queue.lock);
Johann Baudy69e3c752009-05-18 22:11:22 -07002050 spin_lock_bh(&sk->sk_write_queue.lock);
2051 if (po->tx_ring.pg_vec) {
2052 if (packet_current_frame(po, &po->tx_ring, TP_STATUS_AVAILABLE))
2053 mask |= POLLOUT | POLLWRNORM;
2054 }
2055 spin_unlock_bh(&sk->sk_write_queue.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056 return mask;
2057}
2058
2059
2060/* Dirty? Well, I still did not learn better way to account
2061 * for user mmaps.
2062 */
2063
2064static void packet_mm_open(struct vm_area_struct *vma)
2065{
2066 struct file *file = vma->vm_file;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002067 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09002069
Linus Torvalds1da177e2005-04-16 15:20:36 -07002070 if (sk)
2071 atomic_inc(&pkt_sk(sk)->mapped);
2072}
2073
2074static void packet_mm_close(struct vm_area_struct *vma)
2075{
2076 struct file *file = vma->vm_file;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002077 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09002079
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080 if (sk)
2081 atomic_dec(&pkt_sk(sk)->mapped);
2082}
2083
2084static struct vm_operations_struct packet_mmap_ops = {
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002085 .open = packet_mm_open,
2086 .close = packet_mm_close,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087};
2088
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002089static void free_pg_vec(char **pg_vec, unsigned int order, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090{
2091 int i;
2092
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002093 for (i = 0; i < len; i++) {
2094 if (likely(pg_vec[i]))
2095 free_pages((unsigned long) pg_vec[i], order);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 }
2097 kfree(pg_vec);
2098}
2099
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002100static inline char *alloc_one_pg_vec_page(unsigned long order)
2101{
Eric Dumazet719bfea2009-04-15 03:39:52 -07002102 gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | __GFP_ZERO | __GFP_NOWARN;
2103
2104 return (char *) __get_free_pages(gfp_flags, order);
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002105}
2106
2107static char **alloc_pg_vec(struct tpacket_req *req, int order)
2108{
2109 unsigned int block_nr = req->tp_block_nr;
2110 char **pg_vec;
2111 int i;
2112
2113 pg_vec = kzalloc(block_nr * sizeof(char *), GFP_KERNEL);
2114 if (unlikely(!pg_vec))
2115 goto out;
2116
2117 for (i = 0; i < block_nr; i++) {
2118 pg_vec[i] = alloc_one_pg_vec_page(order);
2119 if (unlikely(!pg_vec[i]))
2120 goto out_free_pgvec;
2121 }
2122
2123out:
2124 return pg_vec;
2125
2126out_free_pgvec:
2127 free_pg_vec(pg_vec, order, block_nr);
2128 pg_vec = NULL;
2129 goto out;
2130}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131
Johann Baudy69e3c752009-05-18 22:11:22 -07002132static int packet_set_ring(struct sock *sk, struct tpacket_req *req,
2133 int closing, int tx_ring)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134{
2135 char **pg_vec = NULL;
2136 struct packet_sock *po = pkt_sk(sk);
Al Viro0e11c912006-11-08 00:26:29 -08002137 int was_running, order = 0;
Johann Baudy69e3c752009-05-18 22:11:22 -07002138 struct packet_ring_buffer *rb;
2139 struct sk_buff_head *rb_queue;
Al Viro0e11c912006-11-08 00:26:29 -08002140 __be16 num;
Johann Baudy69e3c752009-05-18 22:11:22 -07002141 int err;
2142
2143 rb = tx_ring ? &po->tx_ring : &po->rx_ring;
2144 rb_queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue;
2145
2146 err = -EBUSY;
2147 if (!closing) {
2148 if (atomic_read(&po->mapped))
2149 goto out;
2150 if (atomic_read(&rb->pending))
2151 goto out;
2152 }
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09002153
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 if (req->tp_block_nr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155 /* Sanity tests and some calculations */
Johann Baudy69e3c752009-05-18 22:11:22 -07002156 err = -EBUSY;
2157 if (unlikely(rb->pg_vec))
2158 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159
Patrick McHardybbd6ef82008-07-14 22:50:15 -07002160 switch (po->tp_version) {
2161 case TPACKET_V1:
2162 po->tp_hdrlen = TPACKET_HDRLEN;
2163 break;
2164 case TPACKET_V2:
2165 po->tp_hdrlen = TPACKET2_HDRLEN;
2166 break;
2167 }
2168
Johann Baudy69e3c752009-05-18 22:11:22 -07002169 err = -EINVAL;
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002170 if (unlikely((int)req->tp_block_size <= 0))
Johann Baudy69e3c752009-05-18 22:11:22 -07002171 goto out;
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002172 if (unlikely(req->tp_block_size & (PAGE_SIZE - 1)))
Johann Baudy69e3c752009-05-18 22:11:22 -07002173 goto out;
Patrick McHardy8913336a2008-07-18 18:05:19 -07002174 if (unlikely(req->tp_frame_size < po->tp_hdrlen +
Johann Baudy69e3c752009-05-18 22:11:22 -07002175 po->tp_reserve))
2176 goto out;
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002177 if (unlikely(req->tp_frame_size & (TPACKET_ALIGNMENT - 1)))
Johann Baudy69e3c752009-05-18 22:11:22 -07002178 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179
Johann Baudy69e3c752009-05-18 22:11:22 -07002180 rb->frames_per_block = req->tp_block_size/req->tp_frame_size;
2181 if (unlikely(rb->frames_per_block <= 0))
2182 goto out;
2183 if (unlikely((rb->frames_per_block * req->tp_block_nr) !=
2184 req->tp_frame_nr))
2185 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186
2187 err = -ENOMEM;
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002188 order = get_order(req->tp_block_size);
2189 pg_vec = alloc_pg_vec(req, order);
2190 if (unlikely(!pg_vec))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191 goto out;
Johann Baudy69e3c752009-05-18 22:11:22 -07002192 }
2193 /* Done */
2194 else {
2195 err = -EINVAL;
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002196 if (unlikely(req->tp_frame_nr))
Johann Baudy69e3c752009-05-18 22:11:22 -07002197 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198 }
2199
2200 lock_sock(sk);
2201
2202 /* Detach socket from network */
2203 spin_lock(&po->bind_lock);
2204 was_running = po->running;
2205 num = po->num;
2206 if (was_running) {
2207 __dev_remove_pack(&po->prot_hook);
2208 po->num = 0;
2209 po->running = 0;
2210 __sock_put(sk);
2211 }
2212 spin_unlock(&po->bind_lock);
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09002213
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214 synchronize_net();
2215
2216 err = -EBUSY;
Herbert Xu905db442009-01-30 14:12:06 -08002217 mutex_lock(&po->pg_vec_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218 if (closing || atomic_read(&po->mapped) == 0) {
2219 err = 0;
2220#define XC(a, b) ({ __typeof__ ((a)) __t; __t = (a); (a) = (b); __t; })
Johann Baudy69e3c752009-05-18 22:11:22 -07002221 spin_lock_bh(&rb_queue->lock);
2222 pg_vec = XC(rb->pg_vec, pg_vec);
2223 rb->frame_max = (req->tp_frame_nr - 1);
2224 rb->head = 0;
2225 rb->frame_size = req->tp_frame_size;
2226 spin_unlock_bh(&rb_queue->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227
Johann Baudy69e3c752009-05-18 22:11:22 -07002228 order = XC(rb->pg_vec_order, order);
2229 req->tp_block_nr = XC(rb->pg_vec_len, req->tp_block_nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230
Johann Baudy69e3c752009-05-18 22:11:22 -07002231 rb->pg_vec_pages = req->tp_block_size/PAGE_SIZE;
2232 po->prot_hook.func = (po->rx_ring.pg_vec) ?
2233 tpacket_rcv : packet_rcv;
2234 skb_queue_purge(rb_queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235#undef XC
2236 if (atomic_read(&po->mapped))
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002237 pr_err("packet_mmap: vma is busy: %d\n",
2238 atomic_read(&po->mapped));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002239 }
Herbert Xu905db442009-01-30 14:12:06 -08002240 mutex_unlock(&po->pg_vec_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241
2242 spin_lock(&po->bind_lock);
2243 if (was_running && !po->running) {
2244 sock_hold(sk);
2245 po->running = 1;
2246 po->num = num;
2247 dev_add_pack(&po->prot_hook);
2248 }
2249 spin_unlock(&po->bind_lock);
2250
2251 release_sock(sk);
2252
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253 if (pg_vec)
2254 free_pg_vec(pg_vec, order, req->tp_block_nr);
2255out:
2256 return err;
2257}
2258
Johann Baudy69e3c752009-05-18 22:11:22 -07002259static int packet_mmap(struct file *file, struct socket *sock,
2260 struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002261{
2262 struct sock *sk = sock->sk;
2263 struct packet_sock *po = pkt_sk(sk);
Johann Baudy69e3c752009-05-18 22:11:22 -07002264 unsigned long size, expected_size;
2265 struct packet_ring_buffer *rb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266 unsigned long start;
2267 int err = -EINVAL;
2268 int i;
2269
2270 if (vma->vm_pgoff)
2271 return -EINVAL;
2272
Herbert Xu905db442009-01-30 14:12:06 -08002273 mutex_lock(&po->pg_vec_lock);
Johann Baudy69e3c752009-05-18 22:11:22 -07002274
2275 expected_size = 0;
2276 for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) {
2277 if (rb->pg_vec) {
2278 expected_size += rb->pg_vec_len
2279 * rb->pg_vec_pages
2280 * PAGE_SIZE;
2281 }
2282 }
2283
2284 if (expected_size == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002285 goto out;
Johann Baudy69e3c752009-05-18 22:11:22 -07002286
2287 size = vma->vm_end - vma->vm_start;
2288 if (size != expected_size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289 goto out;
2290
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291 start = vma->vm_start;
Johann Baudy69e3c752009-05-18 22:11:22 -07002292 for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) {
2293 if (rb->pg_vec == NULL)
2294 continue;
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002295
Johann Baudy69e3c752009-05-18 22:11:22 -07002296 for (i = 0; i < rb->pg_vec_len; i++) {
2297 struct page *page = virt_to_page(rb->pg_vec[i]);
2298 int pg_num;
2299
2300 for (pg_num = 0; pg_num < rb->pg_vec_pages;
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002301 pg_num++, page++) {
Johann Baudy69e3c752009-05-18 22:11:22 -07002302 err = vm_insert_page(vma, start, page);
2303 if (unlikely(err))
2304 goto out;
2305 start += PAGE_SIZE;
2306 }
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002307 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002308 }
Johann Baudy69e3c752009-05-18 22:11:22 -07002309
David S. Miller4ebf0ae2005-12-06 16:38:35 -08002310 atomic_inc(&po->mapped);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311 vma->vm_ops = &packet_mmap_ops;
2312 err = 0;
2313
2314out:
Herbert Xu905db442009-01-30 14:12:06 -08002315 mutex_unlock(&po->pg_vec_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002316 return err;
2317}
2318#endif
2319
2320
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08002321static const struct proto_ops packet_ops_spkt = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002322 .family = PF_PACKET,
2323 .owner = THIS_MODULE,
2324 .release = packet_release,
2325 .bind = packet_bind_spkt,
2326 .connect = sock_no_connect,
2327 .socketpair = sock_no_socketpair,
2328 .accept = sock_no_accept,
2329 .getname = packet_getname_spkt,
2330 .poll = datagram_poll,
2331 .ioctl = packet_ioctl,
2332 .listen = sock_no_listen,
2333 .shutdown = sock_no_shutdown,
2334 .setsockopt = sock_no_setsockopt,
2335 .getsockopt = sock_no_getsockopt,
2336 .sendmsg = packet_sendmsg_spkt,
2337 .recvmsg = packet_recvmsg,
2338 .mmap = sock_no_mmap,
2339 .sendpage = sock_no_sendpage,
2340};
Linus Torvalds1da177e2005-04-16 15:20:36 -07002341
Eric Dumazet90ddc4f2005-12-22 12:49:22 -08002342static const struct proto_ops packet_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 .family = PF_PACKET,
2344 .owner = THIS_MODULE,
2345 .release = packet_release,
2346 .bind = packet_bind,
2347 .connect = sock_no_connect,
2348 .socketpair = sock_no_socketpair,
2349 .accept = sock_no_accept,
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09002350 .getname = packet_getname,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351 .poll = packet_poll,
2352 .ioctl = packet_ioctl,
2353 .listen = sock_no_listen,
2354 .shutdown = sock_no_shutdown,
2355 .setsockopt = packet_setsockopt,
2356 .getsockopt = packet_getsockopt,
2357 .sendmsg = packet_sendmsg,
2358 .recvmsg = packet_recvmsg,
2359 .mmap = packet_mmap,
2360 .sendpage = sock_no_sendpage,
2361};
2362
2363static struct net_proto_family packet_family_ops = {
2364 .family = PF_PACKET,
2365 .create = packet_create,
2366 .owner = THIS_MODULE,
2367};
2368
2369static struct notifier_block packet_netdev_notifier = {
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002370 .notifier_call = packet_notifier,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002371};
2372
2373#ifdef CONFIG_PROC_FS
Denis V. Lunevd12d01d2007-11-19 22:28:35 -08002374static inline struct sock *packet_seq_idx(struct net *net, loff_t off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375{
2376 struct sock *s;
2377 struct hlist_node *node;
2378
Denis V. Lunev2aaef4e2007-12-11 04:19:54 -08002379 sk_for_each(s, node, &net->packet.sklist) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380 if (!off--)
2381 return s;
2382 }
2383 return NULL;
2384}
2385
2386static void *packet_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet40ccbf52008-01-07 22:39:57 -08002387 __acquires(seq_file_net(seq)->packet.sklist_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388{
Denis V. Luneve372c412007-11-19 22:31:54 -08002389 struct net *net = seq_file_net(seq);
Denis V. Lunev2aaef4e2007-12-11 04:19:54 -08002390 read_lock(&net->packet.sklist_lock);
Denis V. Lunevd12d01d2007-11-19 22:28:35 -08002391 return *pos ? packet_seq_idx(net, *pos - 1) : SEQ_START_TOKEN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392}
2393
2394static void *packet_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2395{
Herbert Xu1bf40952007-12-16 14:04:02 -08002396 struct net *net = seq_file_net(seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397 ++*pos;
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09002398 return (v == SEQ_START_TOKEN)
Denis V. Lunev2aaef4e2007-12-11 04:19:54 -08002399 ? sk_head(&net->packet.sklist)
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002400 : sk_next((struct sock *)v) ;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002401}
2402
2403static void packet_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet40ccbf52008-01-07 22:39:57 -08002404 __releases(seq_file_net(seq)->packet.sklist_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002405{
Herbert Xu1bf40952007-12-16 14:04:02 -08002406 struct net *net = seq_file_net(seq);
Denis V. Lunev2aaef4e2007-12-11 04:19:54 -08002407 read_unlock(&net->packet.sklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408}
2409
YOSHIFUJI Hideaki1ce4f282007-02-09 23:25:10 +09002410static int packet_seq_show(struct seq_file *seq, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002411{
2412 if (v == SEQ_START_TOKEN)
2413 seq_puts(seq, "sk RefCnt Type Proto Iface R Rmem User Inode\n");
2414 else {
2415 struct sock *s = v;
2416 const struct packet_sock *po = pkt_sk(s);
2417
2418 seq_printf(seq,
2419 "%p %-6d %-4d %04x %-5d %1d %-6u %-6u %-6lu\n",
2420 s,
2421 atomic_read(&s->sk_refcnt),
2422 s->sk_type,
2423 ntohs(po->num),
2424 po->ifindex,
2425 po->running,
2426 atomic_read(&s->sk_rmem_alloc),
2427 sock_i_uid(s),
Eric Dumazet40d4e3d2009-07-21 21:57:59 +00002428 sock_i_ino(s));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002429 }
2430
2431 return 0;
2432}
2433
Philippe De Muyter56b3d972007-07-10 23:07:31 -07002434static const struct seq_operations packet_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435 .start = packet_seq_start,
2436 .next = packet_seq_next,
2437 .stop = packet_seq_stop,
2438 .show = packet_seq_show,
2439};
2440
2441static int packet_seq_open(struct inode *inode, struct file *file)
2442{
Denis V. Luneve372c412007-11-19 22:31:54 -08002443 return seq_open_net(inode, file, &packet_seq_ops,
2444 sizeof(struct seq_net_private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445}
2446
Arjan van de Venda7071d2007-02-12 00:55:36 -08002447static const struct file_operations packet_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448 .owner = THIS_MODULE,
2449 .open = packet_seq_open,
2450 .read = seq_read,
2451 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08002452 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453};
2454
2455#endif
2456
Denis V. Lunevd12d01d2007-11-19 22:28:35 -08002457static int packet_net_init(struct net *net)
2458{
Denis V. Lunev2aaef4e2007-12-11 04:19:54 -08002459 rwlock_init(&net->packet.sklist_lock);
2460 INIT_HLIST_HEAD(&net->packet.sklist);
Denis V. Lunevd12d01d2007-11-19 22:28:35 -08002461
2462 if (!proc_net_fops_create(net, "packet", 0, &packet_seq_fops))
2463 return -ENOMEM;
2464
2465 return 0;
2466}
2467
2468static void packet_net_exit(struct net *net)
2469{
2470 proc_net_remove(net, "packet");
2471}
2472
2473static struct pernet_operations packet_net_ops = {
2474 .init = packet_net_init,
2475 .exit = packet_net_exit,
2476};
2477
2478
Linus Torvalds1da177e2005-04-16 15:20:36 -07002479static void __exit packet_exit(void)
2480{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002481 unregister_netdevice_notifier(&packet_netdev_notifier);
Denis V. Lunevd12d01d2007-11-19 22:28:35 -08002482 unregister_pernet_subsys(&packet_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002483 sock_unregister(PF_PACKET);
2484 proto_unregister(&packet_proto);
2485}
2486
2487static int __init packet_init(void)
2488{
2489 int rc = proto_register(&packet_proto, 0);
2490
2491 if (rc != 0)
2492 goto out;
2493
2494 sock_register(&packet_family_ops);
Denis V. Lunevd12d01d2007-11-19 22:28:35 -08002495 register_pernet_subsys(&packet_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496 register_netdevice_notifier(&packet_netdev_notifier);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002497out:
2498 return rc;
2499}
2500
2501module_init(packet_init);
2502module_exit(packet_exit);
2503MODULE_LICENSE("GPL");
2504MODULE_ALIAS_NETPROTO(PF_PACKET);