]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * PACKET - implements raw packet sockets. | |
7 | * | |
02c30a84 | 8 | * Authors: Ross Biro |
1da177e4 LT |
9 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
10 | * Alan Cox, <gw4pts@gw4pts.ampr.org> | |
11 | * | |
1ce4f28b | 12 | * Fixes: |
1da177e4 LT |
13 | * 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. | |
1ce4f28b | 35 | * Ulises Alonso : Frame number limit removal and |
1da177e4 | 36 | * packet_set_ring memory leak. |
0fb375fb EB |
37 | * Eric Biederman : Allow for > 8 byte hardware addresses. |
38 | * The convention is that longer addresses | |
39 | * will simply extend the hardware address | |
1ce4f28b | 40 | * byte arrays at the end of sockaddr_ll |
0fb375fb | 41 | * and packet_mreq. |
69e3c75f | 42 | * Johann Baudy : Added TX RING. |
f6fb8f10 | 43 | * Chetan Loke : Implemented TPACKET_V3 block abstraction |
44 | * layer. | |
45 | * Copyright (C) 2011, <lokec@ccs.neu.edu> | |
46 | * | |
1da177e4 LT |
47 | * |
48 | * This program is free software; you can redistribute it and/or | |
49 | * modify it under the terms of the GNU General Public License | |
50 | * as published by the Free Software Foundation; either version | |
51 | * 2 of the License, or (at your option) any later version. | |
52 | * | |
53 | */ | |
1ce4f28b | 54 | |
1da177e4 | 55 | #include <linux/types.h> |
1da177e4 | 56 | #include <linux/mm.h> |
4fc268d2 | 57 | #include <linux/capability.h> |
1da177e4 LT |
58 | #include <linux/fcntl.h> |
59 | #include <linux/socket.h> | |
60 | #include <linux/in.h> | |
61 | #include <linux/inet.h> | |
62 | #include <linux/netdevice.h> | |
63 | #include <linux/if_packet.h> | |
64 | #include <linux/wireless.h> | |
ffbc6111 | 65 | #include <linux/kernel.h> |
1da177e4 | 66 | #include <linux/kmod.h> |
5a0e3ad6 | 67 | #include <linux/slab.h> |
0e3125c7 | 68 | #include <linux/vmalloc.h> |
457c4cbc | 69 | #include <net/net_namespace.h> |
1da177e4 LT |
70 | #include <net/ip.h> |
71 | #include <net/protocol.h> | |
72 | #include <linux/skbuff.h> | |
73 | #include <net/sock.h> | |
74 | #include <linux/errno.h> | |
75 | #include <linux/timer.h> | |
7c0f6ba6 | 76 | #include <linux/uaccess.h> |
1da177e4 LT |
77 | #include <asm/ioctls.h> |
78 | #include <asm/page.h> | |
a1f8e7f7 | 79 | #include <asm/cacheflush.h> |
1da177e4 LT |
80 | #include <asm/io.h> |
81 | #include <linux/proc_fs.h> | |
82 | #include <linux/seq_file.h> | |
83 | #include <linux/poll.h> | |
84 | #include <linux/module.h> | |
85 | #include <linux/init.h> | |
905db440 | 86 | #include <linux/mutex.h> |
05423b24 | 87 | #include <linux/if_vlan.h> |
bfd5f4a3 | 88 | #include <linux/virtio_net.h> |
ed85b565 | 89 | #include <linux/errqueue.h> |
614f60fa | 90 | #include <linux/net_tstamp.h> |
b0138408 | 91 | #include <linux/percpu.h> |
1da177e4 LT |
92 | #ifdef CONFIG_INET |
93 | #include <net/inet_common.h> | |
94 | #endif | |
47dceb8e | 95 | #include <linux/bpf.h> |
719c44d3 | 96 | #include <net/compat.h> |
1da177e4 | 97 | |
2787b04b PE |
98 | #include "internal.h" |
99 | ||
1da177e4 LT |
100 | /* |
101 | Assumptions: | |
102 | - if device has no dev->hard_header routine, it adds and removes ll header | |
103 | inside itself. In this case ll header is invisible outside of device, | |
104 | but higher levels still should reserve dev->hard_header_len. | |
105 | Some devices are enough clever to reallocate skb, when header | |
106 | will not fit to reserved space (tunnel), another ones are silly | |
107 | (PPP). | |
108 | - packet socket receives packets with pulled ll header, | |
109 | so that SOCK_RAW should push it back. | |
110 | ||
111 | On receive: | |
112 | ----------- | |
113 | ||
114 | Incoming, dev->hard_header!=NULL | |
b0e380b1 ACM |
115 | mac_header -> ll header |
116 | data -> data | |
1da177e4 LT |
117 | |
118 | Outgoing, dev->hard_header!=NULL | |
b0e380b1 ACM |
119 | mac_header -> ll header |
120 | data -> ll header | |
1da177e4 LT |
121 | |
122 | Incoming, dev->hard_header==NULL | |
b0e380b1 ACM |
123 | mac_header -> UNKNOWN position. It is very likely, that it points to ll |
124 | header. PPP makes it, that is wrong, because introduce | |
db0c58f9 | 125 | assymetry between rx and tx paths. |
b0e380b1 | 126 | data -> data |
1da177e4 LT |
127 | |
128 | Outgoing, dev->hard_header==NULL | |
b0e380b1 ACM |
129 | mac_header -> data. ll header is still not built! |
130 | data -> data | |
1da177e4 LT |
131 | |
132 | Resume | |
133 | If dev->hard_header==NULL we are unlikely to restore sensible ll header. | |
134 | ||
135 | ||
136 | On transmit: | |
137 | ------------ | |
138 | ||
139 | dev->hard_header != NULL | |
b0e380b1 ACM |
140 | mac_header -> ll header |
141 | data -> ll header | |
1da177e4 LT |
142 | |
143 | dev->hard_header == NULL (ll header is added by device, we cannot control it) | |
b0e380b1 ACM |
144 | mac_header -> data |
145 | data -> data | |
1da177e4 LT |
146 | |
147 | We should set nh.raw on output to correct posistion, | |
148 | packet classifier depends on it. | |
149 | */ | |
150 | ||
1da177e4 LT |
151 | /* Private packet socket structures. */ |
152 | ||
0fb375fb EB |
153 | /* identical to struct packet_mreq except it has |
154 | * a longer address field. | |
155 | */ | |
40d4e3df | 156 | struct packet_mreq_max { |
0fb375fb EB |
157 | int mr_ifindex; |
158 | unsigned short mr_type; | |
159 | unsigned short mr_alen; | |
160 | unsigned char mr_address[MAX_ADDR_LEN]; | |
1da177e4 | 161 | }; |
a2efcfa0 | 162 | |
184f489e DB |
163 | union tpacket_uhdr { |
164 | struct tpacket_hdr *h1; | |
165 | struct tpacket2_hdr *h2; | |
166 | struct tpacket3_hdr *h3; | |
167 | void *raw; | |
168 | }; | |
169 | ||
f6fb8f10 | 170 | static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u, |
69e3c75f JB |
171 | int closing, int tx_ring); |
172 | ||
f6fb8f10 | 173 | #define V3_ALIGNMENT (8) |
174 | ||
bc59ba39 | 175 | #define BLK_HDR_LEN (ALIGN(sizeof(struct tpacket_block_desc), V3_ALIGNMENT)) |
f6fb8f10 | 176 | |
177 | #define BLK_PLUS_PRIV(sz_of_priv) \ | |
178 | (BLK_HDR_LEN + ALIGN((sz_of_priv), V3_ALIGNMENT)) | |
179 | ||
f6fb8f10 | 180 | #define PGV_FROM_VMALLOC 1 |
69e3c75f | 181 | |
f6fb8f10 | 182 | #define BLOCK_STATUS(x) ((x)->hdr.bh1.block_status) |
183 | #define BLOCK_NUM_PKTS(x) ((x)->hdr.bh1.num_pkts) | |
184 | #define BLOCK_O2FP(x) ((x)->hdr.bh1.offset_to_first_pkt) | |
185 | #define BLOCK_LEN(x) ((x)->hdr.bh1.blk_len) | |
186 | #define BLOCK_SNUM(x) ((x)->hdr.bh1.seq_num) | |
187 | #define BLOCK_O2PRIV(x) ((x)->offset_to_priv) | |
188 | #define BLOCK_PRIV(x) ((void *)((char *)(x) + BLOCK_O2PRIV(x))) | |
189 | ||
69e3c75f | 190 | struct packet_sock; |
77f65ebd WB |
191 | static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, |
192 | struct packet_type *pt, struct net_device *orig_dev); | |
1da177e4 | 193 | |
f6fb8f10 | 194 | static void *packet_previous_frame(struct packet_sock *po, |
195 | struct packet_ring_buffer *rb, | |
196 | int status); | |
197 | static void packet_increment_head(struct packet_ring_buffer *buff); | |
878cd3ba | 198 | static int prb_curr_blk_in_use(struct tpacket_block_desc *); |
bc59ba39 | 199 | static void *prb_dispatch_next_block(struct tpacket_kbdq_core *, |
f6fb8f10 | 200 | struct packet_sock *); |
bc59ba39 | 201 | static void prb_retire_current_block(struct tpacket_kbdq_core *, |
f6fb8f10 | 202 | struct packet_sock *, unsigned int status); |
bc59ba39 | 203 | static int prb_queue_frozen(struct tpacket_kbdq_core *); |
204 | static void prb_open_block(struct tpacket_kbdq_core *, | |
205 | struct tpacket_block_desc *); | |
f6fb8f10 | 206 | static void prb_retire_rx_blk_timer_expired(unsigned long); |
bc59ba39 | 207 | static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *); |
208 | static void prb_init_blk_timer(struct packet_sock *, | |
209 | struct tpacket_kbdq_core *, | |
210 | void (*func) (unsigned long)); | |
211 | static void prb_fill_rxhash(struct tpacket_kbdq_core *, struct tpacket3_hdr *); | |
212 | static void prb_clear_rxhash(struct tpacket_kbdq_core *, | |
213 | struct tpacket3_hdr *); | |
214 | static void prb_fill_vlan_info(struct tpacket_kbdq_core *, | |
215 | struct tpacket3_hdr *); | |
1da177e4 | 216 | static void packet_flush_mclist(struct sock *sk); |
ccd4eb49 | 217 | static void packet_pick_tx_queue(struct net_device *dev, struct sk_buff *skb); |
1da177e4 | 218 | |
ffbc6111 | 219 | struct packet_skb_cb { |
ffbc6111 HX |
220 | union { |
221 | struct sockaddr_pkt pkt; | |
2472d761 EB |
222 | union { |
223 | /* Trick: alias skb original length with | |
224 | * ll.sll_family and ll.protocol in order | |
225 | * to save room. | |
226 | */ | |
227 | unsigned int origlen; | |
228 | struct sockaddr_ll ll; | |
229 | }; | |
ffbc6111 HX |
230 | } sa; |
231 | }; | |
232 | ||
d3869efe DW |
233 | #define vio_le() virtio_legacy_is_little_endian() |
234 | ||
ffbc6111 | 235 | #define PACKET_SKB_CB(__skb) ((struct packet_skb_cb *)((__skb)->cb)) |
8dc41944 | 236 | |
bc59ba39 | 237 | #define GET_PBDQC_FROM_RB(x) ((struct tpacket_kbdq_core *)(&(x)->prb_bdqc)) |
f6fb8f10 | 238 | #define GET_PBLOCK_DESC(x, bid) \ |
bc59ba39 | 239 | ((struct tpacket_block_desc *)((x)->pkbdq[(bid)].buffer)) |
f6fb8f10 | 240 | #define GET_CURR_PBLOCK_DESC_FROM_CORE(x) \ |
bc59ba39 | 241 | ((struct tpacket_block_desc *)((x)->pkbdq[(x)->kactive_blk_num].buffer)) |
f6fb8f10 | 242 | #define GET_NEXT_PRB_BLK_NUM(x) \ |
243 | (((x)->kactive_blk_num < ((x)->knum_blocks-1)) ? \ | |
244 | ((x)->kactive_blk_num+1) : 0) | |
245 | ||
dc99f600 DM |
246 | static void __fanout_unlink(struct sock *sk, struct packet_sock *po); |
247 | static void __fanout_link(struct sock *sk, struct packet_sock *po); | |
248 | ||
d346a3fa DB |
249 | static int packet_direct_xmit(struct sk_buff *skb) |
250 | { | |
251 | struct net_device *dev = skb->dev; | |
104ba78c | 252 | struct sk_buff *orig_skb = skb; |
d346a3fa | 253 | struct netdev_queue *txq; |
43279500 | 254 | int ret = NETDEV_TX_BUSY; |
d346a3fa DB |
255 | |
256 | if (unlikely(!netif_running(dev) || | |
43279500 DB |
257 | !netif_carrier_ok(dev))) |
258 | goto drop; | |
d346a3fa | 259 | |
104ba78c WB |
260 | skb = validate_xmit_skb_list(skb, dev); |
261 | if (skb != orig_skb) | |
43279500 | 262 | goto drop; |
d346a3fa | 263 | |
ccd4eb49 | 264 | packet_pick_tx_queue(dev, skb); |
10c51b56 | 265 | txq = skb_get_tx_queue(dev, skb); |
d346a3fa | 266 | |
43279500 DB |
267 | local_bh_disable(); |
268 | ||
269 | HARD_TX_LOCK(dev, txq, smp_processor_id()); | |
10b3ad8c | 270 | if (!netif_xmit_frozen_or_drv_stopped(txq)) |
fa2dbdc2 | 271 | ret = netdev_start_xmit(skb, dev, txq, false); |
43279500 | 272 | HARD_TX_UNLOCK(dev, txq); |
d346a3fa | 273 | |
43279500 DB |
274 | local_bh_enable(); |
275 | ||
276 | if (!dev_xmit_complete(ret)) | |
d346a3fa | 277 | kfree_skb(skb); |
43279500 | 278 | |
d346a3fa | 279 | return ret; |
43279500 | 280 | drop: |
0f97ede4 | 281 | atomic_long_inc(&dev->tx_dropped); |
104ba78c | 282 | kfree_skb_list(skb); |
43279500 | 283 | return NET_XMIT_DROP; |
d346a3fa DB |
284 | } |
285 | ||
66e56cd4 DB |
286 | static struct net_device *packet_cached_dev_get(struct packet_sock *po) |
287 | { | |
288 | struct net_device *dev; | |
289 | ||
290 | rcu_read_lock(); | |
291 | dev = rcu_dereference(po->cached_dev); | |
292 | if (likely(dev)) | |
293 | dev_hold(dev); | |
294 | rcu_read_unlock(); | |
295 | ||
296 | return dev; | |
297 | } | |
298 | ||
299 | static void packet_cached_dev_assign(struct packet_sock *po, | |
300 | struct net_device *dev) | |
301 | { | |
302 | rcu_assign_pointer(po->cached_dev, dev); | |
303 | } | |
304 | ||
305 | static void packet_cached_dev_reset(struct packet_sock *po) | |
306 | { | |
307 | RCU_INIT_POINTER(po->cached_dev, NULL); | |
308 | } | |
309 | ||
d346a3fa DB |
310 | static bool packet_use_direct_xmit(const struct packet_sock *po) |
311 | { | |
312 | return po->xmit == packet_direct_xmit; | |
313 | } | |
314 | ||
0fd5d57b | 315 | static u16 __packet_pick_tx_queue(struct net_device *dev, struct sk_buff *skb) |
d346a3fa | 316 | { |
1cbac010 | 317 | return (u16) raw_smp_processor_id() % dev->real_num_tx_queues; |
d346a3fa DB |
318 | } |
319 | ||
0fd5d57b DB |
320 | static void packet_pick_tx_queue(struct net_device *dev, struct sk_buff *skb) |
321 | { | |
322 | const struct net_device_ops *ops = dev->netdev_ops; | |
323 | u16 queue_index; | |
324 | ||
325 | if (ops->ndo_select_queue) { | |
326 | queue_index = ops->ndo_select_queue(dev, skb, NULL, | |
327 | __packet_pick_tx_queue); | |
328 | queue_index = netdev_cap_txqueue(dev, queue_index); | |
329 | } else { | |
330 | queue_index = __packet_pick_tx_queue(dev, skb); | |
331 | } | |
332 | ||
333 | skb_set_queue_mapping(skb, queue_index); | |
334 | } | |
335 | ||
ce06b03e DM |
336 | /* register_prot_hook must be invoked with the po->bind_lock held, |
337 | * or from a context in which asynchronous accesses to the packet | |
338 | * socket is not possible (packet_create()). | |
339 | */ | |
340 | static void register_prot_hook(struct sock *sk) | |
341 | { | |
342 | struct packet_sock *po = pkt_sk(sk); | |
e40526cb | 343 | |
ce06b03e | 344 | if (!po->running) { |
66e56cd4 | 345 | if (po->fanout) |
dc99f600 | 346 | __fanout_link(sk, po); |
66e56cd4 | 347 | else |
dc99f600 | 348 | dev_add_pack(&po->prot_hook); |
e40526cb | 349 | |
ce06b03e DM |
350 | sock_hold(sk); |
351 | po->running = 1; | |
352 | } | |
353 | } | |
354 | ||
355 | /* {,__}unregister_prot_hook() must be invoked with the po->bind_lock | |
356 | * held. If the sync parameter is true, we will temporarily drop | |
357 | * the po->bind_lock and do a synchronize_net to make sure no | |
358 | * asynchronous packet processing paths still refer to the elements | |
359 | * of po->prot_hook. If the sync parameter is false, it is the | |
360 | * callers responsibility to take care of this. | |
361 | */ | |
362 | static void __unregister_prot_hook(struct sock *sk, bool sync) | |
363 | { | |
364 | struct packet_sock *po = pkt_sk(sk); | |
365 | ||
366 | po->running = 0; | |
66e56cd4 DB |
367 | |
368 | if (po->fanout) | |
dc99f600 | 369 | __fanout_unlink(sk, po); |
66e56cd4 | 370 | else |
dc99f600 | 371 | __dev_remove_pack(&po->prot_hook); |
e40526cb | 372 | |
ce06b03e DM |
373 | __sock_put(sk); |
374 | ||
375 | if (sync) { | |
376 | spin_unlock(&po->bind_lock); | |
377 | synchronize_net(); | |
378 | spin_lock(&po->bind_lock); | |
379 | } | |
380 | } | |
381 | ||
382 | static void unregister_prot_hook(struct sock *sk, bool sync) | |
383 | { | |
384 | struct packet_sock *po = pkt_sk(sk); | |
385 | ||
386 | if (po->running) | |
387 | __unregister_prot_hook(sk, sync); | |
388 | } | |
389 | ||
6e58040b | 390 | static inline struct page * __pure pgv_to_page(void *addr) |
0af55bb5 CG |
391 | { |
392 | if (is_vmalloc_addr(addr)) | |
393 | return vmalloc_to_page(addr); | |
394 | return virt_to_page(addr); | |
395 | } | |
396 | ||
69e3c75f | 397 | static void __packet_set_status(struct packet_sock *po, void *frame, int status) |
1da177e4 | 398 | { |
184f489e | 399 | union tpacket_uhdr h; |
1da177e4 | 400 | |
69e3c75f | 401 | h.raw = frame; |
bbd6ef87 PM |
402 | switch (po->tp_version) { |
403 | case TPACKET_V1: | |
69e3c75f | 404 | h.h1->tp_status = status; |
0af55bb5 | 405 | flush_dcache_page(pgv_to_page(&h.h1->tp_status)); |
bbd6ef87 PM |
406 | break; |
407 | case TPACKET_V2: | |
69e3c75f | 408 | h.h2->tp_status = status; |
0af55bb5 | 409 | flush_dcache_page(pgv_to_page(&h.h2->tp_status)); |
bbd6ef87 | 410 | break; |
f6fb8f10 | 411 | case TPACKET_V3: |
7f953ab2 SV |
412 | h.h3->tp_status = status; |
413 | flush_dcache_page(pgv_to_page(&h.h3->tp_status)); | |
414 | break; | |
69e3c75f | 415 | default: |
f6fb8f10 | 416 | WARN(1, "TPACKET version not supported.\n"); |
69e3c75f | 417 | BUG(); |
bbd6ef87 | 418 | } |
69e3c75f JB |
419 | |
420 | smp_wmb(); | |
bbd6ef87 PM |
421 | } |
422 | ||
69e3c75f | 423 | static int __packet_get_status(struct packet_sock *po, void *frame) |
bbd6ef87 | 424 | { |
184f489e | 425 | union tpacket_uhdr h; |
bbd6ef87 | 426 | |
69e3c75f JB |
427 | smp_rmb(); |
428 | ||
bbd6ef87 PM |
429 | h.raw = frame; |
430 | switch (po->tp_version) { | |
431 | case TPACKET_V1: | |
0af55bb5 | 432 | flush_dcache_page(pgv_to_page(&h.h1->tp_status)); |
69e3c75f | 433 | return h.h1->tp_status; |
bbd6ef87 | 434 | case TPACKET_V2: |
0af55bb5 | 435 | flush_dcache_page(pgv_to_page(&h.h2->tp_status)); |
69e3c75f | 436 | return h.h2->tp_status; |
f6fb8f10 | 437 | case TPACKET_V3: |
7f953ab2 SV |
438 | flush_dcache_page(pgv_to_page(&h.h3->tp_status)); |
439 | return h.h3->tp_status; | |
69e3c75f | 440 | default: |
f6fb8f10 | 441 | WARN(1, "TPACKET version not supported.\n"); |
69e3c75f JB |
442 | BUG(); |
443 | return 0; | |
bbd6ef87 | 444 | } |
1da177e4 | 445 | } |
69e3c75f | 446 | |
b9c32fb2 DB |
447 | static __u32 tpacket_get_timestamp(struct sk_buff *skb, struct timespec *ts, |
448 | unsigned int flags) | |
7a51384c DB |
449 | { |
450 | struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb); | |
451 | ||
68a360e8 WB |
452 | if (shhwtstamps && |
453 | (flags & SOF_TIMESTAMPING_RAW_HARDWARE) && | |
454 | ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts)) | |
455 | return TP_STATUS_TS_RAW_HARDWARE; | |
7a51384c DB |
456 | |
457 | if (ktime_to_timespec_cond(skb->tstamp, ts)) | |
b9c32fb2 | 458 | return TP_STATUS_TS_SOFTWARE; |
7a51384c | 459 | |
b9c32fb2 | 460 | return 0; |
7a51384c DB |
461 | } |
462 | ||
b9c32fb2 DB |
463 | static __u32 __packet_set_timestamp(struct packet_sock *po, void *frame, |
464 | struct sk_buff *skb) | |
2e31396f WB |
465 | { |
466 | union tpacket_uhdr h; | |
467 | struct timespec ts; | |
b9c32fb2 | 468 | __u32 ts_status; |
2e31396f | 469 | |
b9c32fb2 DB |
470 | if (!(ts_status = tpacket_get_timestamp(skb, &ts, po->tp_tstamp))) |
471 | return 0; | |
2e31396f WB |
472 | |
473 | h.raw = frame; | |
474 | switch (po->tp_version) { | |
475 | case TPACKET_V1: | |
476 | h.h1->tp_sec = ts.tv_sec; | |
477 | h.h1->tp_usec = ts.tv_nsec / NSEC_PER_USEC; | |
478 | break; | |
479 | case TPACKET_V2: | |
480 | h.h2->tp_sec = ts.tv_sec; | |
481 | h.h2->tp_nsec = ts.tv_nsec; | |
482 | break; | |
483 | case TPACKET_V3: | |
57ea884b DB |
484 | h.h3->tp_sec = ts.tv_sec; |
485 | h.h3->tp_nsec = ts.tv_nsec; | |
486 | break; | |
2e31396f WB |
487 | default: |
488 | WARN(1, "TPACKET version not supported.\n"); | |
489 | BUG(); | |
490 | } | |
491 | ||
492 | /* one flush is safe, as both fields always lie on the same cacheline */ | |
493 | flush_dcache_page(pgv_to_page(&h.h1->tp_sec)); | |
494 | smp_wmb(); | |
b9c32fb2 DB |
495 | |
496 | return ts_status; | |
2e31396f WB |
497 | } |
498 | ||
69e3c75f JB |
499 | static void *packet_lookup_frame(struct packet_sock *po, |
500 | struct packet_ring_buffer *rb, | |
501 | unsigned int position, | |
502 | int status) | |
503 | { | |
504 | unsigned int pg_vec_pos, frame_offset; | |
184f489e | 505 | union tpacket_uhdr h; |
69e3c75f JB |
506 | |
507 | pg_vec_pos = position / rb->frames_per_block; | |
508 | frame_offset = position % rb->frames_per_block; | |
509 | ||
0e3125c7 NH |
510 | h.raw = rb->pg_vec[pg_vec_pos].buffer + |
511 | (frame_offset * rb->frame_size); | |
69e3c75f JB |
512 | |
513 | if (status != __packet_get_status(po, h.raw)) | |
514 | return NULL; | |
515 | ||
516 | return h.raw; | |
517 | } | |
518 | ||
eea49cc9 | 519 | static void *packet_current_frame(struct packet_sock *po, |
69e3c75f JB |
520 | struct packet_ring_buffer *rb, |
521 | int status) | |
522 | { | |
523 | return packet_lookup_frame(po, rb, rb->head, status); | |
524 | } | |
525 | ||
bc59ba39 | 526 | static void prb_del_retire_blk_timer(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 527 | { |
528 | del_timer_sync(&pkc->retire_blk_timer); | |
529 | } | |
530 | ||
531 | static void prb_shutdown_retire_blk_timer(struct packet_sock *po, | |
f6fb8f10 | 532 | struct sk_buff_head *rb_queue) |
533 | { | |
bc59ba39 | 534 | struct tpacket_kbdq_core *pkc; |
f6fb8f10 | 535 | |
73d0fcf2 | 536 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 537 | |
ec6f809f | 538 | spin_lock_bh(&rb_queue->lock); |
f6fb8f10 | 539 | pkc->delete_blk_timer = 1; |
ec6f809f | 540 | spin_unlock_bh(&rb_queue->lock); |
f6fb8f10 | 541 | |
542 | prb_del_retire_blk_timer(pkc); | |
543 | } | |
544 | ||
545 | static void prb_init_blk_timer(struct packet_sock *po, | |
bc59ba39 | 546 | struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 547 | void (*func) (unsigned long)) |
548 | { | |
549 | init_timer(&pkc->retire_blk_timer); | |
550 | pkc->retire_blk_timer.data = (long)po; | |
551 | pkc->retire_blk_timer.function = func; | |
552 | pkc->retire_blk_timer.expires = jiffies; | |
553 | } | |
554 | ||
e8e85cc5 | 555 | static void prb_setup_retire_blk_timer(struct packet_sock *po) |
f6fb8f10 | 556 | { |
bc59ba39 | 557 | struct tpacket_kbdq_core *pkc; |
f6fb8f10 | 558 | |
e8e85cc5 | 559 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 560 | prb_init_blk_timer(po, pkc, prb_retire_rx_blk_timer_expired); |
561 | } | |
562 | ||
563 | static int prb_calc_retire_blk_tmo(struct packet_sock *po, | |
564 | int blk_size_in_bytes) | |
565 | { | |
566 | struct net_device *dev; | |
567 | unsigned int mbits = 0, msec = 0, div = 0, tmo = 0; | |
7cad1bac | 568 | struct ethtool_link_ksettings ecmd; |
4bc71cb9 | 569 | int err; |
f6fb8f10 | 570 | |
4bc71cb9 JP |
571 | rtnl_lock(); |
572 | dev = __dev_get_by_index(sock_net(&po->sk), po->ifindex); | |
573 | if (unlikely(!dev)) { | |
574 | rtnl_unlock(); | |
f6fb8f10 | 575 | return DEFAULT_PRB_RETIRE_TOV; |
4bc71cb9 | 576 | } |
7cad1bac | 577 | err = __ethtool_get_link_ksettings(dev, &ecmd); |
4bc71cb9 JP |
578 | rtnl_unlock(); |
579 | if (!err) { | |
4bc71cb9 JP |
580 | /* |
581 | * If the link speed is so slow you don't really | |
582 | * need to worry about perf anyways | |
583 | */ | |
7cad1bac DD |
584 | if (ecmd.base.speed < SPEED_1000 || |
585 | ecmd.base.speed == SPEED_UNKNOWN) { | |
4bc71cb9 | 586 | return DEFAULT_PRB_RETIRE_TOV; |
e440cf2c | 587 | } else { |
588 | msec = 1; | |
7cad1bac | 589 | div = ecmd.base.speed / 1000; |
f6fb8f10 | 590 | } |
591 | } | |
592 | ||
593 | mbits = (blk_size_in_bytes * 8) / (1024 * 1024); | |
594 | ||
595 | if (div) | |
596 | mbits /= div; | |
597 | ||
598 | tmo = mbits * msec; | |
599 | ||
600 | if (div) | |
601 | return tmo+1; | |
602 | return tmo; | |
603 | } | |
604 | ||
bc59ba39 | 605 | static void prb_init_ft_ops(struct tpacket_kbdq_core *p1, |
f6fb8f10 | 606 | union tpacket_req_u *req_u) |
607 | { | |
608 | p1->feature_req_word = req_u->req3.tp_feature_req_word; | |
609 | } | |
610 | ||
611 | static void init_prb_bdqc(struct packet_sock *po, | |
612 | struct packet_ring_buffer *rb, | |
613 | struct pgv *pg_vec, | |
e8e85cc5 | 614 | union tpacket_req_u *req_u) |
f6fb8f10 | 615 | { |
22781a5b | 616 | struct tpacket_kbdq_core *p1 = GET_PBDQC_FROM_RB(rb); |
bc59ba39 | 617 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 618 | |
619 | memset(p1, 0x0, sizeof(*p1)); | |
620 | ||
621 | p1->knxt_seq_num = 1; | |
622 | p1->pkbdq = pg_vec; | |
bc59ba39 | 623 | pbd = (struct tpacket_block_desc *)pg_vec[0].buffer; |
e3192690 | 624 | p1->pkblk_start = pg_vec[0].buffer; |
f6fb8f10 | 625 | p1->kblk_size = req_u->req3.tp_block_size; |
626 | p1->knum_blocks = req_u->req3.tp_block_nr; | |
627 | p1->hdrlen = po->tp_hdrlen; | |
628 | p1->version = po->tp_version; | |
629 | p1->last_kactive_blk_num = 0; | |
ee80fbf3 | 630 | po->stats.stats3.tp_freeze_q_cnt = 0; |
f6fb8f10 | 631 | if (req_u->req3.tp_retire_blk_tov) |
632 | p1->retire_blk_tov = req_u->req3.tp_retire_blk_tov; | |
633 | else | |
634 | p1->retire_blk_tov = prb_calc_retire_blk_tmo(po, | |
635 | req_u->req3.tp_block_size); | |
636 | p1->tov_in_jiffies = msecs_to_jiffies(p1->retire_blk_tov); | |
637 | p1->blk_sizeof_priv = req_u->req3.tp_sizeof_priv; | |
638 | ||
dc808110 | 639 | p1->max_frame_len = p1->kblk_size - BLK_PLUS_PRIV(p1->blk_sizeof_priv); |
f6fb8f10 | 640 | prb_init_ft_ops(p1, req_u); |
e8e85cc5 | 641 | prb_setup_retire_blk_timer(po); |
f6fb8f10 | 642 | prb_open_block(p1, pbd); |
643 | } | |
644 | ||
645 | /* Do NOT update the last_blk_num first. | |
646 | * Assumes sk_buff_head lock is held. | |
647 | */ | |
bc59ba39 | 648 | static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 649 | { |
650 | mod_timer(&pkc->retire_blk_timer, | |
651 | jiffies + pkc->tov_in_jiffies); | |
652 | pkc->last_kactive_blk_num = pkc->kactive_blk_num; | |
653 | } | |
654 | ||
655 | /* | |
656 | * Timer logic: | |
657 | * 1) We refresh the timer only when we open a block. | |
658 | * By doing this we don't waste cycles refreshing the timer | |
659 | * on packet-by-packet basis. | |
660 | * | |
661 | * With a 1MB block-size, on a 1Gbps line, it will take | |
662 | * i) ~8 ms to fill a block + ii) memcpy etc. | |
663 | * In this cut we are not accounting for the memcpy time. | |
664 | * | |
665 | * So, if the user sets the 'tmo' to 10ms then the timer | |
666 | * will never fire while the block is still getting filled | |
667 | * (which is what we want). However, the user could choose | |
668 | * to close a block early and that's fine. | |
669 | * | |
670 | * But when the timer does fire, we check whether or not to refresh it. | |
671 | * Since the tmo granularity is in msecs, it is not too expensive | |
672 | * to refresh the timer, lets say every '8' msecs. | |
673 | * Either the user can set the 'tmo' or we can derive it based on | |
674 | * a) line-speed and b) block-size. | |
675 | * prb_calc_retire_blk_tmo() calculates the tmo. | |
676 | * | |
677 | */ | |
678 | static void prb_retire_rx_blk_timer_expired(unsigned long data) | |
679 | { | |
680 | struct packet_sock *po = (struct packet_sock *)data; | |
22781a5b | 681 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 682 | unsigned int frozen; |
bc59ba39 | 683 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 684 | |
685 | spin_lock(&po->sk.sk_receive_queue.lock); | |
686 | ||
687 | frozen = prb_queue_frozen(pkc); | |
688 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
689 | ||
690 | if (unlikely(pkc->delete_blk_timer)) | |
691 | goto out; | |
692 | ||
693 | /* We only need to plug the race when the block is partially filled. | |
694 | * tpacket_rcv: | |
695 | * lock(); increment BLOCK_NUM_PKTS; unlock() | |
696 | * copy_bits() is in progress ... | |
697 | * timer fires on other cpu: | |
698 | * we can't retire the current block because copy_bits | |
699 | * is in progress. | |
700 | * | |
701 | */ | |
702 | if (BLOCK_NUM_PKTS(pbd)) { | |
703 | while (atomic_read(&pkc->blk_fill_in_prog)) { | |
704 | /* Waiting for skb_copy_bits to finish... */ | |
705 | cpu_relax(); | |
706 | } | |
707 | } | |
708 | ||
709 | if (pkc->last_kactive_blk_num == pkc->kactive_blk_num) { | |
710 | if (!frozen) { | |
41a50d62 AD |
711 | if (!BLOCK_NUM_PKTS(pbd)) { |
712 | /* An empty block. Just refresh the timer. */ | |
713 | goto refresh_timer; | |
714 | } | |
f6fb8f10 | 715 | prb_retire_current_block(pkc, po, TP_STATUS_BLK_TMO); |
716 | if (!prb_dispatch_next_block(pkc, po)) | |
717 | goto refresh_timer; | |
718 | else | |
719 | goto out; | |
720 | } else { | |
721 | /* Case 1. Queue was frozen because user-space was | |
722 | * lagging behind. | |
723 | */ | |
878cd3ba | 724 | if (prb_curr_blk_in_use(pbd)) { |
f6fb8f10 | 725 | /* |
726 | * Ok, user-space is still behind. | |
727 | * So just refresh the timer. | |
728 | */ | |
729 | goto refresh_timer; | |
730 | } else { | |
731 | /* Case 2. queue was frozen,user-space caught up, | |
732 | * now the link went idle && the timer fired. | |
733 | * We don't have a block to close.So we open this | |
734 | * block and restart the timer. | |
735 | * opening a block thaws the queue,restarts timer | |
736 | * Thawing/timer-refresh is a side effect. | |
737 | */ | |
738 | prb_open_block(pkc, pbd); | |
739 | goto out; | |
740 | } | |
741 | } | |
742 | } | |
743 | ||
744 | refresh_timer: | |
745 | _prb_refresh_rx_retire_blk_timer(pkc); | |
746 | ||
747 | out: | |
748 | spin_unlock(&po->sk.sk_receive_queue.lock); | |
749 | } | |
750 | ||
eea49cc9 | 751 | static void prb_flush_block(struct tpacket_kbdq_core *pkc1, |
bc59ba39 | 752 | struct tpacket_block_desc *pbd1, __u32 status) |
f6fb8f10 | 753 | { |
754 | /* Flush everything minus the block header */ | |
755 | ||
756 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
757 | u8 *start, *end; | |
758 | ||
759 | start = (u8 *)pbd1; | |
760 | ||
761 | /* Skip the block header(we know header WILL fit in 4K) */ | |
762 | start += PAGE_SIZE; | |
763 | ||
764 | end = (u8 *)PAGE_ALIGN((unsigned long)pkc1->pkblk_end); | |
765 | for (; start < end; start += PAGE_SIZE) | |
766 | flush_dcache_page(pgv_to_page(start)); | |
767 | ||
768 | smp_wmb(); | |
769 | #endif | |
770 | ||
771 | /* Now update the block status. */ | |
772 | ||
773 | BLOCK_STATUS(pbd1) = status; | |
774 | ||
775 | /* Flush the block header */ | |
776 | ||
777 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
778 | start = (u8 *)pbd1; | |
779 | flush_dcache_page(pgv_to_page(start)); | |
780 | ||
781 | smp_wmb(); | |
782 | #endif | |
783 | } | |
784 | ||
785 | /* | |
786 | * Side effect: | |
787 | * | |
788 | * 1) flush the block | |
789 | * 2) Increment active_blk_num | |
790 | * | |
791 | * Note:We DONT refresh the timer on purpose. | |
792 | * Because almost always the next block will be opened. | |
793 | */ | |
bc59ba39 | 794 | static void prb_close_block(struct tpacket_kbdq_core *pkc1, |
795 | struct tpacket_block_desc *pbd1, | |
f6fb8f10 | 796 | struct packet_sock *po, unsigned int stat) |
797 | { | |
798 | __u32 status = TP_STATUS_USER | stat; | |
799 | ||
800 | struct tpacket3_hdr *last_pkt; | |
bc59ba39 | 801 | struct tpacket_hdr_v1 *h1 = &pbd1->hdr.bh1; |
da413eec | 802 | struct sock *sk = &po->sk; |
f6fb8f10 | 803 | |
ee80fbf3 | 804 | if (po->stats.stats3.tp_drops) |
f6fb8f10 | 805 | status |= TP_STATUS_LOSING; |
806 | ||
807 | last_pkt = (struct tpacket3_hdr *)pkc1->prev; | |
808 | last_pkt->tp_next_offset = 0; | |
809 | ||
810 | /* Get the ts of the last pkt */ | |
811 | if (BLOCK_NUM_PKTS(pbd1)) { | |
812 | h1->ts_last_pkt.ts_sec = last_pkt->tp_sec; | |
813 | h1->ts_last_pkt.ts_nsec = last_pkt->tp_nsec; | |
814 | } else { | |
41a50d62 AD |
815 | /* Ok, we tmo'd - so get the current time. |
816 | * | |
817 | * It shouldn't really happen as we don't close empty | |
818 | * blocks. See prb_retire_rx_blk_timer_expired(). | |
819 | */ | |
f6fb8f10 | 820 | struct timespec ts; |
821 | getnstimeofday(&ts); | |
822 | h1->ts_last_pkt.ts_sec = ts.tv_sec; | |
823 | h1->ts_last_pkt.ts_nsec = ts.tv_nsec; | |
824 | } | |
825 | ||
826 | smp_wmb(); | |
827 | ||
828 | /* Flush the block */ | |
829 | prb_flush_block(pkc1, pbd1, status); | |
830 | ||
da413eec DC |
831 | sk->sk_data_ready(sk); |
832 | ||
f6fb8f10 | 833 | pkc1->kactive_blk_num = GET_NEXT_PRB_BLK_NUM(pkc1); |
834 | } | |
835 | ||
eea49cc9 | 836 | static void prb_thaw_queue(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 837 | { |
838 | pkc->reset_pending_on_curr_blk = 0; | |
839 | } | |
840 | ||
841 | /* | |
842 | * Side effect of opening a block: | |
843 | * | |
844 | * 1) prb_queue is thawed. | |
845 | * 2) retire_blk_timer is refreshed. | |
846 | * | |
847 | */ | |
bc59ba39 | 848 | static void prb_open_block(struct tpacket_kbdq_core *pkc1, |
849 | struct tpacket_block_desc *pbd1) | |
f6fb8f10 | 850 | { |
851 | struct timespec ts; | |
bc59ba39 | 852 | struct tpacket_hdr_v1 *h1 = &pbd1->hdr.bh1; |
f6fb8f10 | 853 | |
854 | smp_rmb(); | |
855 | ||
8da3056c DB |
856 | /* We could have just memset this but we will lose the |
857 | * flexibility of making the priv area sticky | |
858 | */ | |
f6fb8f10 | 859 | |
8da3056c DB |
860 | BLOCK_SNUM(pbd1) = pkc1->knxt_seq_num++; |
861 | BLOCK_NUM_PKTS(pbd1) = 0; | |
862 | BLOCK_LEN(pbd1) = BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
f6fb8f10 | 863 | |
8da3056c DB |
864 | getnstimeofday(&ts); |
865 | ||
866 | h1->ts_first_pkt.ts_sec = ts.tv_sec; | |
867 | h1->ts_first_pkt.ts_nsec = ts.tv_nsec; | |
f6fb8f10 | 868 | |
8da3056c DB |
869 | pkc1->pkblk_start = (char *)pbd1; |
870 | pkc1->nxt_offset = pkc1->pkblk_start + BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
871 | ||
872 | BLOCK_O2FP(pbd1) = (__u32)BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
873 | BLOCK_O2PRIV(pbd1) = BLK_HDR_LEN; | |
874 | ||
875 | pbd1->version = pkc1->version; | |
876 | pkc1->prev = pkc1->nxt_offset; | |
877 | pkc1->pkblk_end = pkc1->pkblk_start + pkc1->kblk_size; | |
878 | ||
879 | prb_thaw_queue(pkc1); | |
880 | _prb_refresh_rx_retire_blk_timer(pkc1); | |
881 | ||
882 | smp_wmb(); | |
f6fb8f10 | 883 | } |
884 | ||
885 | /* | |
886 | * Queue freeze logic: | |
887 | * 1) Assume tp_block_nr = 8 blocks. | |
888 | * 2) At time 't0', user opens Rx ring. | |
889 | * 3) Some time past 't0', kernel starts filling blocks starting from 0 .. 7 | |
890 | * 4) user-space is either sleeping or processing block '0'. | |
891 | * 5) tpacket_rcv is currently filling block '7', since there is no space left, | |
892 | * it will close block-7,loop around and try to fill block '0'. | |
893 | * call-flow: | |
894 | * __packet_lookup_frame_in_block | |
895 | * prb_retire_current_block() | |
896 | * prb_dispatch_next_block() | |
897 | * |->(BLOCK_STATUS == USER) evaluates to true | |
898 | * 5.1) Since block-0 is currently in-use, we just freeze the queue. | |
899 | * 6) Now there are two cases: | |
900 | * 6.1) Link goes idle right after the queue is frozen. | |
901 | * But remember, the last open_block() refreshed the timer. | |
902 | * When this timer expires,it will refresh itself so that we can | |
903 | * re-open block-0 in near future. | |
904 | * 6.2) Link is busy and keeps on receiving packets. This is a simple | |
905 | * case and __packet_lookup_frame_in_block will check if block-0 | |
906 | * is free and can now be re-used. | |
907 | */ | |
eea49cc9 | 908 | static void prb_freeze_queue(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 909 | struct packet_sock *po) |
910 | { | |
911 | pkc->reset_pending_on_curr_blk = 1; | |
ee80fbf3 | 912 | po->stats.stats3.tp_freeze_q_cnt++; |
f6fb8f10 | 913 | } |
914 | ||
915 | #define TOTAL_PKT_LEN_INCL_ALIGN(length) (ALIGN((length), V3_ALIGNMENT)) | |
916 | ||
917 | /* | |
918 | * If the next block is free then we will dispatch it | |
919 | * and return a good offset. | |
920 | * Else, we will freeze the queue. | |
921 | * So, caller must check the return value. | |
922 | */ | |
bc59ba39 | 923 | static void *prb_dispatch_next_block(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 924 | struct packet_sock *po) |
925 | { | |
bc59ba39 | 926 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 927 | |
928 | smp_rmb(); | |
929 | ||
930 | /* 1. Get current block num */ | |
931 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
932 | ||
933 | /* 2. If this block is currently in_use then freeze the queue */ | |
934 | if (TP_STATUS_USER & BLOCK_STATUS(pbd)) { | |
935 | prb_freeze_queue(pkc, po); | |
936 | return NULL; | |
937 | } | |
938 | ||
939 | /* | |
940 | * 3. | |
941 | * open this block and return the offset where the first packet | |
942 | * needs to get stored. | |
943 | */ | |
944 | prb_open_block(pkc, pbd); | |
945 | return (void *)pkc->nxt_offset; | |
946 | } | |
947 | ||
bc59ba39 | 948 | static void prb_retire_current_block(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 949 | struct packet_sock *po, unsigned int status) |
950 | { | |
bc59ba39 | 951 | struct tpacket_block_desc *pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); |
f6fb8f10 | 952 | |
953 | /* retire/close the current block */ | |
954 | if (likely(TP_STATUS_KERNEL == BLOCK_STATUS(pbd))) { | |
955 | /* | |
956 | * Plug the case where copy_bits() is in progress on | |
957 | * cpu-0 and tpacket_rcv() got invoked on cpu-1, didn't | |
958 | * have space to copy the pkt in the current block and | |
959 | * called prb_retire_current_block() | |
960 | * | |
961 | * We don't need to worry about the TMO case because | |
962 | * the timer-handler already handled this case. | |
963 | */ | |
964 | if (!(status & TP_STATUS_BLK_TMO)) { | |
965 | while (atomic_read(&pkc->blk_fill_in_prog)) { | |
966 | /* Waiting for skb_copy_bits to finish... */ | |
967 | cpu_relax(); | |
968 | } | |
969 | } | |
970 | prb_close_block(pkc, pbd, po, status); | |
971 | return; | |
972 | } | |
f6fb8f10 | 973 | } |
974 | ||
878cd3ba | 975 | static int prb_curr_blk_in_use(struct tpacket_block_desc *pbd) |
f6fb8f10 | 976 | { |
977 | return TP_STATUS_USER & BLOCK_STATUS(pbd); | |
978 | } | |
979 | ||
eea49cc9 | 980 | static int prb_queue_frozen(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 981 | { |
982 | return pkc->reset_pending_on_curr_blk; | |
983 | } | |
984 | ||
eea49cc9 | 985 | static void prb_clear_blk_fill_status(struct packet_ring_buffer *rb) |
f6fb8f10 | 986 | { |
bc59ba39 | 987 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(rb); |
f6fb8f10 | 988 | atomic_dec(&pkc->blk_fill_in_prog); |
989 | } | |
990 | ||
eea49cc9 | 991 | static void prb_fill_rxhash(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 992 | struct tpacket3_hdr *ppd) |
993 | { | |
3958afa1 | 994 | ppd->hv1.tp_rxhash = skb_get_hash(pkc->skb); |
f6fb8f10 | 995 | } |
996 | ||
eea49cc9 | 997 | static void prb_clear_rxhash(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 998 | struct tpacket3_hdr *ppd) |
999 | { | |
1000 | ppd->hv1.tp_rxhash = 0; | |
1001 | } | |
1002 | ||
eea49cc9 | 1003 | static void prb_fill_vlan_info(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 1004 | struct tpacket3_hdr *ppd) |
1005 | { | |
df8a39de JP |
1006 | if (skb_vlan_tag_present(pkc->skb)) { |
1007 | ppd->hv1.tp_vlan_tci = skb_vlan_tag_get(pkc->skb); | |
a0cdfcf3 AW |
1008 | ppd->hv1.tp_vlan_tpid = ntohs(pkc->skb->vlan_proto); |
1009 | ppd->tp_status = TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
f6fb8f10 | 1010 | } else { |
9e67030a | 1011 | ppd->hv1.tp_vlan_tci = 0; |
a0cdfcf3 | 1012 | ppd->hv1.tp_vlan_tpid = 0; |
9e67030a | 1013 | ppd->tp_status = TP_STATUS_AVAILABLE; |
f6fb8f10 | 1014 | } |
1015 | } | |
1016 | ||
bc59ba39 | 1017 | static void prb_run_all_ft_ops(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 1018 | struct tpacket3_hdr *ppd) |
1019 | { | |
a0cdfcf3 | 1020 | ppd->hv1.tp_padding = 0; |
f6fb8f10 | 1021 | prb_fill_vlan_info(pkc, ppd); |
1022 | ||
1023 | if (pkc->feature_req_word & TP_FT_REQ_FILL_RXHASH) | |
1024 | prb_fill_rxhash(pkc, ppd); | |
1025 | else | |
1026 | prb_clear_rxhash(pkc, ppd); | |
1027 | } | |
1028 | ||
eea49cc9 | 1029 | static void prb_fill_curr_block(char *curr, |
bc59ba39 | 1030 | struct tpacket_kbdq_core *pkc, |
1031 | struct tpacket_block_desc *pbd, | |
f6fb8f10 | 1032 | unsigned int len) |
1033 | { | |
1034 | struct tpacket3_hdr *ppd; | |
1035 | ||
1036 | ppd = (struct tpacket3_hdr *)curr; | |
1037 | ppd->tp_next_offset = TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1038 | pkc->prev = curr; | |
1039 | pkc->nxt_offset += TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1040 | BLOCK_LEN(pbd) += TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1041 | BLOCK_NUM_PKTS(pbd) += 1; | |
1042 | atomic_inc(&pkc->blk_fill_in_prog); | |
1043 | prb_run_all_ft_ops(pkc, ppd); | |
1044 | } | |
1045 | ||
1046 | /* Assumes caller has the sk->rx_queue.lock */ | |
1047 | static void *__packet_lookup_frame_in_block(struct packet_sock *po, | |
1048 | struct sk_buff *skb, | |
1049 | int status, | |
1050 | unsigned int len | |
1051 | ) | |
1052 | { | |
bc59ba39 | 1053 | struct tpacket_kbdq_core *pkc; |
1054 | struct tpacket_block_desc *pbd; | |
f6fb8f10 | 1055 | char *curr, *end; |
1056 | ||
e3192690 | 1057 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 1058 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); |
1059 | ||
1060 | /* Queue is frozen when user space is lagging behind */ | |
1061 | if (prb_queue_frozen(pkc)) { | |
1062 | /* | |
1063 | * Check if that last block which caused the queue to freeze, | |
1064 | * is still in_use by user-space. | |
1065 | */ | |
878cd3ba | 1066 | if (prb_curr_blk_in_use(pbd)) { |
f6fb8f10 | 1067 | /* Can't record this packet */ |
1068 | return NULL; | |
1069 | } else { | |
1070 | /* | |
1071 | * Ok, the block was released by user-space. | |
1072 | * Now let's open that block. | |
1073 | * opening a block also thaws the queue. | |
1074 | * Thawing is a side effect. | |
1075 | */ | |
1076 | prb_open_block(pkc, pbd); | |
1077 | } | |
1078 | } | |
1079 | ||
1080 | smp_mb(); | |
1081 | curr = pkc->nxt_offset; | |
1082 | pkc->skb = skb; | |
e3192690 | 1083 | end = (char *)pbd + pkc->kblk_size; |
f6fb8f10 | 1084 | |
1085 | /* first try the current block */ | |
1086 | if (curr+TOTAL_PKT_LEN_INCL_ALIGN(len) < end) { | |
1087 | prb_fill_curr_block(curr, pkc, pbd, len); | |
1088 | return (void *)curr; | |
1089 | } | |
1090 | ||
1091 | /* Ok, close the current block */ | |
1092 | prb_retire_current_block(pkc, po, 0); | |
1093 | ||
1094 | /* Now, try to dispatch the next block */ | |
1095 | curr = (char *)prb_dispatch_next_block(pkc, po); | |
1096 | if (curr) { | |
1097 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
1098 | prb_fill_curr_block(curr, pkc, pbd, len); | |
1099 | return (void *)curr; | |
1100 | } | |
1101 | ||
1102 | /* | |
1103 | * No free blocks are available.user_space hasn't caught up yet. | |
1104 | * Queue was just frozen and now this packet will get dropped. | |
1105 | */ | |
1106 | return NULL; | |
1107 | } | |
1108 | ||
eea49cc9 | 1109 | static void *packet_current_rx_frame(struct packet_sock *po, |
f6fb8f10 | 1110 | struct sk_buff *skb, |
1111 | int status, unsigned int len) | |
1112 | { | |
1113 | char *curr = NULL; | |
1114 | switch (po->tp_version) { | |
1115 | case TPACKET_V1: | |
1116 | case TPACKET_V2: | |
1117 | curr = packet_lookup_frame(po, &po->rx_ring, | |
1118 | po->rx_ring.head, status); | |
1119 | return curr; | |
1120 | case TPACKET_V3: | |
1121 | return __packet_lookup_frame_in_block(po, skb, status, len); | |
1122 | default: | |
1123 | WARN(1, "TPACKET version not supported\n"); | |
1124 | BUG(); | |
99aa3473 | 1125 | return NULL; |
f6fb8f10 | 1126 | } |
1127 | } | |
1128 | ||
eea49cc9 | 1129 | static void *prb_lookup_block(struct packet_sock *po, |
f6fb8f10 | 1130 | struct packet_ring_buffer *rb, |
77f65ebd | 1131 | unsigned int idx, |
f6fb8f10 | 1132 | int status) |
1133 | { | |
bc59ba39 | 1134 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(rb); |
77f65ebd | 1135 | struct tpacket_block_desc *pbd = GET_PBLOCK_DESC(pkc, idx); |
f6fb8f10 | 1136 | |
1137 | if (status != BLOCK_STATUS(pbd)) | |
1138 | return NULL; | |
1139 | return pbd; | |
1140 | } | |
1141 | ||
eea49cc9 | 1142 | static int prb_previous_blk_num(struct packet_ring_buffer *rb) |
f6fb8f10 | 1143 | { |
1144 | unsigned int prev; | |
1145 | if (rb->prb_bdqc.kactive_blk_num) | |
1146 | prev = rb->prb_bdqc.kactive_blk_num-1; | |
1147 | else | |
1148 | prev = rb->prb_bdqc.knum_blocks-1; | |
1149 | return prev; | |
1150 | } | |
1151 | ||
1152 | /* Assumes caller has held the rx_queue.lock */ | |
eea49cc9 | 1153 | static void *__prb_previous_block(struct packet_sock *po, |
f6fb8f10 | 1154 | struct packet_ring_buffer *rb, |
1155 | int status) | |
1156 | { | |
1157 | unsigned int previous = prb_previous_blk_num(rb); | |
1158 | return prb_lookup_block(po, rb, previous, status); | |
1159 | } | |
1160 | ||
eea49cc9 | 1161 | static void *packet_previous_rx_frame(struct packet_sock *po, |
f6fb8f10 | 1162 | struct packet_ring_buffer *rb, |
1163 | int status) | |
1164 | { | |
1165 | if (po->tp_version <= TPACKET_V2) | |
1166 | return packet_previous_frame(po, rb, status); | |
1167 | ||
1168 | return __prb_previous_block(po, rb, status); | |
1169 | } | |
1170 | ||
eea49cc9 | 1171 | static void packet_increment_rx_head(struct packet_sock *po, |
f6fb8f10 | 1172 | struct packet_ring_buffer *rb) |
1173 | { | |
1174 | switch (po->tp_version) { | |
1175 | case TPACKET_V1: | |
1176 | case TPACKET_V2: | |
1177 | return packet_increment_head(rb); | |
1178 | case TPACKET_V3: | |
1179 | default: | |
1180 | WARN(1, "TPACKET version not supported.\n"); | |
1181 | BUG(); | |
1182 | return; | |
1183 | } | |
1184 | } | |
1185 | ||
eea49cc9 | 1186 | static void *packet_previous_frame(struct packet_sock *po, |
69e3c75f JB |
1187 | struct packet_ring_buffer *rb, |
1188 | int status) | |
1189 | { | |
1190 | unsigned int previous = rb->head ? rb->head - 1 : rb->frame_max; | |
1191 | return packet_lookup_frame(po, rb, previous, status); | |
1192 | } | |
1193 | ||
eea49cc9 | 1194 | static void packet_increment_head(struct packet_ring_buffer *buff) |
69e3c75f JB |
1195 | { |
1196 | buff->head = buff->head != buff->frame_max ? buff->head+1 : 0; | |
1197 | } | |
1198 | ||
b0138408 DB |
1199 | static void packet_inc_pending(struct packet_ring_buffer *rb) |
1200 | { | |
1201 | this_cpu_inc(*rb->pending_refcnt); | |
1202 | } | |
1203 | ||
1204 | static void packet_dec_pending(struct packet_ring_buffer *rb) | |
1205 | { | |
1206 | this_cpu_dec(*rb->pending_refcnt); | |
1207 | } | |
1208 | ||
1209 | static unsigned int packet_read_pending(const struct packet_ring_buffer *rb) | |
1210 | { | |
1211 | unsigned int refcnt = 0; | |
1212 | int cpu; | |
1213 | ||
1214 | /* We don't use pending refcount in rx_ring. */ | |
1215 | if (rb->pending_refcnt == NULL) | |
1216 | return 0; | |
1217 | ||
1218 | for_each_possible_cpu(cpu) | |
1219 | refcnt += *per_cpu_ptr(rb->pending_refcnt, cpu); | |
1220 | ||
1221 | return refcnt; | |
1222 | } | |
1223 | ||
1224 | static int packet_alloc_pending(struct packet_sock *po) | |
1225 | { | |
1226 | po->rx_ring.pending_refcnt = NULL; | |
1227 | ||
1228 | po->tx_ring.pending_refcnt = alloc_percpu(unsigned int); | |
1229 | if (unlikely(po->tx_ring.pending_refcnt == NULL)) | |
1230 | return -ENOBUFS; | |
1231 | ||
1232 | return 0; | |
1233 | } | |
1234 | ||
1235 | static void packet_free_pending(struct packet_sock *po) | |
1236 | { | |
1237 | free_percpu(po->tx_ring.pending_refcnt); | |
1238 | } | |
1239 | ||
9954729b WB |
1240 | #define ROOM_POW_OFF 2 |
1241 | #define ROOM_NONE 0x0 | |
1242 | #define ROOM_LOW 0x1 | |
1243 | #define ROOM_NORMAL 0x2 | |
1244 | ||
1245 | static bool __tpacket_has_room(struct packet_sock *po, int pow_off) | |
77f65ebd | 1246 | { |
9954729b WB |
1247 | int idx, len; |
1248 | ||
1249 | len = po->rx_ring.frame_max + 1; | |
1250 | idx = po->rx_ring.head; | |
1251 | if (pow_off) | |
1252 | idx += len >> pow_off; | |
1253 | if (idx >= len) | |
1254 | idx -= len; | |
1255 | return packet_lookup_frame(po, &po->rx_ring, idx, TP_STATUS_KERNEL); | |
1256 | } | |
1257 | ||
1258 | static bool __tpacket_v3_has_room(struct packet_sock *po, int pow_off) | |
1259 | { | |
1260 | int idx, len; | |
1261 | ||
1262 | len = po->rx_ring.prb_bdqc.knum_blocks; | |
1263 | idx = po->rx_ring.prb_bdqc.kactive_blk_num; | |
1264 | if (pow_off) | |
1265 | idx += len >> pow_off; | |
1266 | if (idx >= len) | |
1267 | idx -= len; | |
1268 | return prb_lookup_block(po, &po->rx_ring, idx, TP_STATUS_KERNEL); | |
1269 | } | |
77f65ebd | 1270 | |
2ccdbaa6 | 1271 | static int __packet_rcv_has_room(struct packet_sock *po, struct sk_buff *skb) |
9954729b WB |
1272 | { |
1273 | struct sock *sk = &po->sk; | |
1274 | int ret = ROOM_NONE; | |
1275 | ||
1276 | if (po->prot_hook.func != tpacket_rcv) { | |
1277 | int avail = sk->sk_rcvbuf - atomic_read(&sk->sk_rmem_alloc) | |
2ccdbaa6 | 1278 | - (skb ? skb->truesize : 0); |
9954729b WB |
1279 | if (avail > (sk->sk_rcvbuf >> ROOM_POW_OFF)) |
1280 | return ROOM_NORMAL; | |
1281 | else if (avail > 0) | |
1282 | return ROOM_LOW; | |
1283 | else | |
1284 | return ROOM_NONE; | |
1285 | } | |
77f65ebd | 1286 | |
9954729b WB |
1287 | if (po->tp_version == TPACKET_V3) { |
1288 | if (__tpacket_v3_has_room(po, ROOM_POW_OFF)) | |
1289 | ret = ROOM_NORMAL; | |
1290 | else if (__tpacket_v3_has_room(po, 0)) | |
1291 | ret = ROOM_LOW; | |
1292 | } else { | |
1293 | if (__tpacket_has_room(po, ROOM_POW_OFF)) | |
1294 | ret = ROOM_NORMAL; | |
1295 | else if (__tpacket_has_room(po, 0)) | |
1296 | ret = ROOM_LOW; | |
1297 | } | |
2ccdbaa6 WB |
1298 | |
1299 | return ret; | |
1300 | } | |
1301 | ||
1302 | static int packet_rcv_has_room(struct packet_sock *po, struct sk_buff *skb) | |
1303 | { | |
1304 | int ret; | |
1305 | bool has_room; | |
1306 | ||
54d7c01d WB |
1307 | spin_lock_bh(&po->sk.sk_receive_queue.lock); |
1308 | ret = __packet_rcv_has_room(po, skb); | |
2ccdbaa6 WB |
1309 | has_room = ret == ROOM_NORMAL; |
1310 | if (po->pressure == has_room) | |
54d7c01d WB |
1311 | po->pressure = !has_room; |
1312 | spin_unlock_bh(&po->sk.sk_receive_queue.lock); | |
77f65ebd | 1313 | |
9954729b | 1314 | return ret; |
77f65ebd WB |
1315 | } |
1316 | ||
1da177e4 LT |
1317 | static void packet_sock_destruct(struct sock *sk) |
1318 | { | |
ed85b565 RC |
1319 | skb_queue_purge(&sk->sk_error_queue); |
1320 | ||
547b792c | 1321 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); |
14afee4b | 1322 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
1da177e4 LT |
1323 | |
1324 | if (!sock_flag(sk, SOCK_DEAD)) { | |
40d4e3df | 1325 | pr_err("Attempt to release alive packet socket: %p\n", sk); |
1da177e4 LT |
1326 | return; |
1327 | } | |
1328 | ||
17ab56a2 | 1329 | sk_refcnt_debug_dec(sk); |
1da177e4 LT |
1330 | } |
1331 | ||
3b3a5b0a WB |
1332 | static bool fanout_flow_is_huge(struct packet_sock *po, struct sk_buff *skb) |
1333 | { | |
1334 | u32 rxhash; | |
1335 | int i, count = 0; | |
1336 | ||
1337 | rxhash = skb_get_hash(skb); | |
1338 | for (i = 0; i < ROLLOVER_HLEN; i++) | |
1339 | if (po->rollover->history[i] == rxhash) | |
1340 | count++; | |
1341 | ||
1342 | po->rollover->history[prandom_u32() % ROLLOVER_HLEN] = rxhash; | |
1343 | return count > (ROLLOVER_HLEN >> 1); | |
1344 | } | |
1345 | ||
77f65ebd WB |
1346 | static unsigned int fanout_demux_hash(struct packet_fanout *f, |
1347 | struct sk_buff *skb, | |
1348 | unsigned int num) | |
dc99f600 | 1349 | { |
eb70db87 | 1350 | return reciprocal_scale(__skb_get_hash_symmetric(skb), num); |
dc99f600 DM |
1351 | } |
1352 | ||
77f65ebd WB |
1353 | static unsigned int fanout_demux_lb(struct packet_fanout *f, |
1354 | struct sk_buff *skb, | |
1355 | unsigned int num) | |
dc99f600 | 1356 | { |
468479e6 | 1357 | unsigned int val = atomic_inc_return(&f->rr_cur); |
dc99f600 | 1358 | |
468479e6 | 1359 | return val % num; |
77f65ebd WB |
1360 | } |
1361 | ||
1362 | static unsigned int fanout_demux_cpu(struct packet_fanout *f, | |
1363 | struct sk_buff *skb, | |
1364 | unsigned int num) | |
1365 | { | |
1366 | return smp_processor_id() % num; | |
dc99f600 DM |
1367 | } |
1368 | ||
5df0ddfb DB |
1369 | static unsigned int fanout_demux_rnd(struct packet_fanout *f, |
1370 | struct sk_buff *skb, | |
1371 | unsigned int num) | |
1372 | { | |
f337db64 | 1373 | return prandom_u32_max(num); |
5df0ddfb DB |
1374 | } |
1375 | ||
77f65ebd WB |
1376 | static unsigned int fanout_demux_rollover(struct packet_fanout *f, |
1377 | struct sk_buff *skb, | |
ad377cab | 1378 | unsigned int idx, bool try_self, |
77f65ebd | 1379 | unsigned int num) |
95ec3eb4 | 1380 | { |
4633c9e0 | 1381 | struct packet_sock *po, *po_next, *po_skip = NULL; |
a9b63918 | 1382 | unsigned int i, j, room = ROOM_NONE; |
95ec3eb4 | 1383 | |
0648ab70 | 1384 | po = pkt_sk(f->arr[idx]); |
3b3a5b0a WB |
1385 | |
1386 | if (try_self) { | |
1387 | room = packet_rcv_has_room(po, skb); | |
1388 | if (room == ROOM_NORMAL || | |
1389 | (room == ROOM_LOW && !fanout_flow_is_huge(po, skb))) | |
1390 | return idx; | |
4633c9e0 | 1391 | po_skip = po; |
3b3a5b0a | 1392 | } |
ad377cab | 1393 | |
0648ab70 | 1394 | i = j = min_t(int, po->rollover->sock, num - 1); |
77f65ebd | 1395 | do { |
2ccdbaa6 | 1396 | po_next = pkt_sk(f->arr[i]); |
4633c9e0 | 1397 | if (po_next != po_skip && !po_next->pressure && |
2ccdbaa6 | 1398 | packet_rcv_has_room(po_next, skb) == ROOM_NORMAL) { |
77f65ebd | 1399 | if (i != j) |
0648ab70 | 1400 | po->rollover->sock = i; |
a9b63918 WB |
1401 | atomic_long_inc(&po->rollover->num); |
1402 | if (room == ROOM_LOW) | |
1403 | atomic_long_inc(&po->rollover->num_huge); | |
77f65ebd WB |
1404 | return i; |
1405 | } | |
ad377cab | 1406 | |
77f65ebd WB |
1407 | if (++i == num) |
1408 | i = 0; | |
1409 | } while (i != j); | |
1410 | ||
a9b63918 | 1411 | atomic_long_inc(&po->rollover->num_failed); |
77f65ebd WB |
1412 | return idx; |
1413 | } | |
1414 | ||
2d36097d NH |
1415 | static unsigned int fanout_demux_qm(struct packet_fanout *f, |
1416 | struct sk_buff *skb, | |
1417 | unsigned int num) | |
1418 | { | |
1419 | return skb_get_queue_mapping(skb) % num; | |
1420 | } | |
1421 | ||
47dceb8e WB |
1422 | static unsigned int fanout_demux_bpf(struct packet_fanout *f, |
1423 | struct sk_buff *skb, | |
1424 | unsigned int num) | |
1425 | { | |
1426 | struct bpf_prog *prog; | |
1427 | unsigned int ret = 0; | |
1428 | ||
1429 | rcu_read_lock(); | |
1430 | prog = rcu_dereference(f->bpf_prog); | |
1431 | if (prog) | |
ff936a04 | 1432 | ret = bpf_prog_run_clear_cb(prog, skb) % num; |
47dceb8e WB |
1433 | rcu_read_unlock(); |
1434 | ||
1435 | return ret; | |
1436 | } | |
1437 | ||
77f65ebd WB |
1438 | static bool fanout_has_flag(struct packet_fanout *f, u16 flag) |
1439 | { | |
1440 | return f->flags & (flag >> 8); | |
95ec3eb4 DM |
1441 | } |
1442 | ||
95ec3eb4 DM |
1443 | static int packet_rcv_fanout(struct sk_buff *skb, struct net_device *dev, |
1444 | struct packet_type *pt, struct net_device *orig_dev) | |
dc99f600 DM |
1445 | { |
1446 | struct packet_fanout *f = pt->af_packet_priv; | |
f98f4514 | 1447 | unsigned int num = READ_ONCE(f->num_members); |
19bcf9f2 | 1448 | struct net *net = read_pnet(&f->net); |
dc99f600 | 1449 | struct packet_sock *po; |
77f65ebd | 1450 | unsigned int idx; |
dc99f600 | 1451 | |
19bcf9f2 | 1452 | if (!net_eq(dev_net(dev), net) || !num) { |
dc99f600 DM |
1453 | kfree_skb(skb); |
1454 | return 0; | |
1455 | } | |
1456 | ||
3f34b24a | 1457 | if (fanout_has_flag(f, PACKET_FANOUT_FLAG_DEFRAG)) { |
19bcf9f2 | 1458 | skb = ip_check_defrag(net, skb, IP_DEFRAG_AF_PACKET); |
3f34b24a AD |
1459 | if (!skb) |
1460 | return 0; | |
1461 | } | |
95ec3eb4 DM |
1462 | switch (f->type) { |
1463 | case PACKET_FANOUT_HASH: | |
1464 | default: | |
77f65ebd | 1465 | idx = fanout_demux_hash(f, skb, num); |
95ec3eb4 DM |
1466 | break; |
1467 | case PACKET_FANOUT_LB: | |
77f65ebd | 1468 | idx = fanout_demux_lb(f, skb, num); |
95ec3eb4 DM |
1469 | break; |
1470 | case PACKET_FANOUT_CPU: | |
77f65ebd WB |
1471 | idx = fanout_demux_cpu(f, skb, num); |
1472 | break; | |
5df0ddfb DB |
1473 | case PACKET_FANOUT_RND: |
1474 | idx = fanout_demux_rnd(f, skb, num); | |
1475 | break; | |
2d36097d NH |
1476 | case PACKET_FANOUT_QM: |
1477 | idx = fanout_demux_qm(f, skb, num); | |
1478 | break; | |
77f65ebd | 1479 | case PACKET_FANOUT_ROLLOVER: |
ad377cab | 1480 | idx = fanout_demux_rollover(f, skb, 0, false, num); |
95ec3eb4 | 1481 | break; |
47dceb8e | 1482 | case PACKET_FANOUT_CBPF: |
f2e52095 | 1483 | case PACKET_FANOUT_EBPF: |
47dceb8e WB |
1484 | idx = fanout_demux_bpf(f, skb, num); |
1485 | break; | |
dc99f600 DM |
1486 | } |
1487 | ||
ad377cab WB |
1488 | if (fanout_has_flag(f, PACKET_FANOUT_FLAG_ROLLOVER)) |
1489 | idx = fanout_demux_rollover(f, skb, idx, true, num); | |
dc99f600 | 1490 | |
ad377cab | 1491 | po = pkt_sk(f->arr[idx]); |
dc99f600 DM |
1492 | return po->prot_hook.func(skb, dev, &po->prot_hook, orig_dev); |
1493 | } | |
1494 | ||
fff3321d PE |
1495 | DEFINE_MUTEX(fanout_mutex); |
1496 | EXPORT_SYMBOL_GPL(fanout_mutex); | |
dc99f600 | 1497 | static LIST_HEAD(fanout_list); |
4a69a864 | 1498 | static u16 fanout_next_id; |
dc99f600 DM |
1499 | |
1500 | static void __fanout_link(struct sock *sk, struct packet_sock *po) | |
1501 | { | |
1502 | struct packet_fanout *f = po->fanout; | |
1503 | ||
1504 | spin_lock(&f->lock); | |
1505 | f->arr[f->num_members] = sk; | |
1506 | smp_wmb(); | |
1507 | f->num_members++; | |
2bd624b4 AS |
1508 | if (f->num_members == 1) |
1509 | dev_add_pack(&f->prot_hook); | |
dc99f600 DM |
1510 | spin_unlock(&f->lock); |
1511 | } | |
1512 | ||
1513 | static void __fanout_unlink(struct sock *sk, struct packet_sock *po) | |
1514 | { | |
1515 | struct packet_fanout *f = po->fanout; | |
1516 | int i; | |
1517 | ||
1518 | spin_lock(&f->lock); | |
1519 | for (i = 0; i < f->num_members; i++) { | |
1520 | if (f->arr[i] == sk) | |
1521 | break; | |
1522 | } | |
1523 | BUG_ON(i >= f->num_members); | |
1524 | f->arr[i] = f->arr[f->num_members - 1]; | |
1525 | f->num_members--; | |
2bd624b4 AS |
1526 | if (f->num_members == 0) |
1527 | __dev_remove_pack(&f->prot_hook); | |
dc99f600 DM |
1528 | spin_unlock(&f->lock); |
1529 | } | |
1530 | ||
d4dd8aee | 1531 | static bool match_fanout_group(struct packet_type *ptype, struct sock *sk) |
c0de08d0 | 1532 | { |
161642e2 ED |
1533 | if (sk->sk_family != PF_PACKET) |
1534 | return false; | |
c0de08d0 | 1535 | |
161642e2 | 1536 | return ptype->af_packet_priv == pkt_sk(sk)->fanout; |
c0de08d0 EL |
1537 | } |
1538 | ||
47dceb8e WB |
1539 | static void fanout_init_data(struct packet_fanout *f) |
1540 | { | |
1541 | switch (f->type) { | |
1542 | case PACKET_FANOUT_LB: | |
1543 | atomic_set(&f->rr_cur, 0); | |
1544 | break; | |
1545 | case PACKET_FANOUT_CBPF: | |
f2e52095 | 1546 | case PACKET_FANOUT_EBPF: |
47dceb8e WB |
1547 | RCU_INIT_POINTER(f->bpf_prog, NULL); |
1548 | break; | |
1549 | } | |
1550 | } | |
1551 | ||
1552 | static void __fanout_set_data_bpf(struct packet_fanout *f, struct bpf_prog *new) | |
1553 | { | |
1554 | struct bpf_prog *old; | |
1555 | ||
1556 | spin_lock(&f->lock); | |
1557 | old = rcu_dereference_protected(f->bpf_prog, lockdep_is_held(&f->lock)); | |
1558 | rcu_assign_pointer(f->bpf_prog, new); | |
1559 | spin_unlock(&f->lock); | |
1560 | ||
1561 | if (old) { | |
1562 | synchronize_net(); | |
1563 | bpf_prog_destroy(old); | |
1564 | } | |
1565 | } | |
1566 | ||
1567 | static int fanout_set_data_cbpf(struct packet_sock *po, char __user *data, | |
1568 | unsigned int len) | |
1569 | { | |
1570 | struct bpf_prog *new; | |
1571 | struct sock_fprog fprog; | |
1572 | int ret; | |
1573 | ||
1574 | if (sock_flag(&po->sk, SOCK_FILTER_LOCKED)) | |
1575 | return -EPERM; | |
1576 | if (len != sizeof(fprog)) | |
1577 | return -EINVAL; | |
1578 | if (copy_from_user(&fprog, data, len)) | |
1579 | return -EFAULT; | |
1580 | ||
bab18991 | 1581 | ret = bpf_prog_create_from_user(&new, &fprog, NULL, false); |
47dceb8e WB |
1582 | if (ret) |
1583 | return ret; | |
1584 | ||
1585 | __fanout_set_data_bpf(po->fanout, new); | |
1586 | return 0; | |
1587 | } | |
1588 | ||
f2e52095 WB |
1589 | static int fanout_set_data_ebpf(struct packet_sock *po, char __user *data, |
1590 | unsigned int len) | |
1591 | { | |
1592 | struct bpf_prog *new; | |
1593 | u32 fd; | |
1594 | ||
1595 | if (sock_flag(&po->sk, SOCK_FILTER_LOCKED)) | |
1596 | return -EPERM; | |
1597 | if (len != sizeof(fd)) | |
1598 | return -EINVAL; | |
1599 | if (copy_from_user(&fd, data, len)) | |
1600 | return -EFAULT; | |
1601 | ||
113214be | 1602 | new = bpf_prog_get_type(fd, BPF_PROG_TYPE_SOCKET_FILTER); |
f2e52095 WB |
1603 | if (IS_ERR(new)) |
1604 | return PTR_ERR(new); | |
f2e52095 WB |
1605 | |
1606 | __fanout_set_data_bpf(po->fanout, new); | |
1607 | return 0; | |
1608 | } | |
1609 | ||
47dceb8e WB |
1610 | static int fanout_set_data(struct packet_sock *po, char __user *data, |
1611 | unsigned int len) | |
1612 | { | |
1613 | switch (po->fanout->type) { | |
1614 | case PACKET_FANOUT_CBPF: | |
1615 | return fanout_set_data_cbpf(po, data, len); | |
f2e52095 WB |
1616 | case PACKET_FANOUT_EBPF: |
1617 | return fanout_set_data_ebpf(po, data, len); | |
47dceb8e WB |
1618 | default: |
1619 | return -EINVAL; | |
1620 | }; | |
1621 | } | |
1622 | ||
1623 | static void fanout_release_data(struct packet_fanout *f) | |
1624 | { | |
1625 | switch (f->type) { | |
1626 | case PACKET_FANOUT_CBPF: | |
f2e52095 | 1627 | case PACKET_FANOUT_EBPF: |
47dceb8e WB |
1628 | __fanout_set_data_bpf(f, NULL); |
1629 | }; | |
1630 | } | |
1631 | ||
4a69a864 MM |
1632 | static bool __fanout_id_is_free(struct sock *sk, u16 candidate_id) |
1633 | { | |
1634 | struct packet_fanout *f; | |
1635 | ||
1636 | list_for_each_entry(f, &fanout_list, list) { | |
1637 | if (f->id == candidate_id && | |
1638 | read_pnet(&f->net) == sock_net(sk)) { | |
1639 | return false; | |
1640 | } | |
1641 | } | |
1642 | return true; | |
1643 | } | |
1644 | ||
1645 | static bool fanout_find_new_id(struct sock *sk, u16 *new_id) | |
1646 | { | |
1647 | u16 id = fanout_next_id; | |
1648 | ||
1649 | do { | |
1650 | if (__fanout_id_is_free(sk, id)) { | |
1651 | *new_id = id; | |
1652 | fanout_next_id = id + 1; | |
1653 | return true; | |
1654 | } | |
1655 | ||
1656 | id++; | |
1657 | } while (id != fanout_next_id); | |
1658 | ||
1659 | return false; | |
1660 | } | |
1661 | ||
7736d33f | 1662 | static int fanout_add(struct sock *sk, u16 id, u16 type_flags) |
dc99f600 | 1663 | { |
d199fab6 | 1664 | struct packet_rollover *rollover = NULL; |
dc99f600 DM |
1665 | struct packet_sock *po = pkt_sk(sk); |
1666 | struct packet_fanout *f, *match; | |
7736d33f | 1667 | u8 type = type_flags & 0xff; |
77f65ebd | 1668 | u8 flags = type_flags >> 8; |
dc99f600 DM |
1669 | int err; |
1670 | ||
1671 | switch (type) { | |
77f65ebd WB |
1672 | case PACKET_FANOUT_ROLLOVER: |
1673 | if (type_flags & PACKET_FANOUT_FLAG_ROLLOVER) | |
1674 | return -EINVAL; | |
dc99f600 DM |
1675 | case PACKET_FANOUT_HASH: |
1676 | case PACKET_FANOUT_LB: | |
95ec3eb4 | 1677 | case PACKET_FANOUT_CPU: |
5df0ddfb | 1678 | case PACKET_FANOUT_RND: |
2d36097d | 1679 | case PACKET_FANOUT_QM: |
47dceb8e | 1680 | case PACKET_FANOUT_CBPF: |
f2e52095 | 1681 | case PACKET_FANOUT_EBPF: |
dc99f600 DM |
1682 | break; |
1683 | default: | |
1684 | return -EINVAL; | |
1685 | } | |
1686 | ||
d199fab6 ED |
1687 | mutex_lock(&fanout_mutex); |
1688 | ||
d199fab6 | 1689 | err = -EALREADY; |
dc99f600 | 1690 | if (po->fanout) |
d199fab6 | 1691 | goto out; |
dc99f600 | 1692 | |
4633c9e0 WB |
1693 | if (type == PACKET_FANOUT_ROLLOVER || |
1694 | (type_flags & PACKET_FANOUT_FLAG_ROLLOVER)) { | |
d199fab6 ED |
1695 | err = -ENOMEM; |
1696 | rollover = kzalloc(sizeof(*rollover), GFP_KERNEL); | |
1697 | if (!rollover) | |
1698 | goto out; | |
1699 | atomic_long_set(&rollover->num, 0); | |
1700 | atomic_long_set(&rollover->num_huge, 0); | |
1701 | atomic_long_set(&rollover->num_failed, 0); | |
1702 | po->rollover = rollover; | |
0648ab70 WB |
1703 | } |
1704 | ||
4a69a864 MM |
1705 | if (type_flags & PACKET_FANOUT_FLAG_UNIQUEID) { |
1706 | if (id != 0) { | |
1707 | err = -EINVAL; | |
1708 | goto out; | |
1709 | } | |
1710 | if (!fanout_find_new_id(sk, &id)) { | |
1711 | err = -ENOMEM; | |
1712 | goto out; | |
1713 | } | |
1714 | /* ephemeral flag for the first socket in the group: drop it */ | |
1715 | flags &= ~(PACKET_FANOUT_FLAG_UNIQUEID >> 8); | |
1716 | } | |
1717 | ||
dc99f600 DM |
1718 | match = NULL; |
1719 | list_for_each_entry(f, &fanout_list, list) { | |
1720 | if (f->id == id && | |
1721 | read_pnet(&f->net) == sock_net(sk)) { | |
1722 | match = f; | |
1723 | break; | |
1724 | } | |
1725 | } | |
afe62c68 | 1726 | err = -EINVAL; |
77f65ebd | 1727 | if (match && match->flags != flags) |
afe62c68 | 1728 | goto out; |
dc99f600 | 1729 | if (!match) { |
afe62c68 | 1730 | err = -ENOMEM; |
dc99f600 | 1731 | match = kzalloc(sizeof(*match), GFP_KERNEL); |
afe62c68 ED |
1732 | if (!match) |
1733 | goto out; | |
1734 | write_pnet(&match->net, sock_net(sk)); | |
1735 | match->id = id; | |
1736 | match->type = type; | |
77f65ebd | 1737 | match->flags = flags; |
afe62c68 ED |
1738 | INIT_LIST_HEAD(&match->list); |
1739 | spin_lock_init(&match->lock); | |
fb5c2c17 | 1740 | refcount_set(&match->sk_ref, 0); |
47dceb8e | 1741 | fanout_init_data(match); |
afe62c68 ED |
1742 | match->prot_hook.type = po->prot_hook.type; |
1743 | match->prot_hook.dev = po->prot_hook.dev; | |
1744 | match->prot_hook.func = packet_rcv_fanout; | |
1745 | match->prot_hook.af_packet_priv = match; | |
c0de08d0 | 1746 | match->prot_hook.id_match = match_fanout_group; |
afe62c68 | 1747 | list_add(&match->list, &fanout_list); |
dc99f600 | 1748 | } |
afe62c68 | 1749 | err = -EINVAL; |
7dfb100a WB |
1750 | |
1751 | spin_lock(&po->bind_lock); | |
1752 | if (po->running && | |
1753 | match->type == type && | |
afe62c68 ED |
1754 | match->prot_hook.type == po->prot_hook.type && |
1755 | match->prot_hook.dev == po->prot_hook.dev) { | |
1756 | err = -ENOSPC; | |
fb5c2c17 | 1757 | if (refcount_read(&match->sk_ref) < PACKET_FANOUT_MAX) { |
afe62c68 ED |
1758 | __dev_remove_pack(&po->prot_hook); |
1759 | po->fanout = match; | |
fb5c2c17 | 1760 | refcount_set(&match->sk_ref, refcount_read(&match->sk_ref) + 1); |
afe62c68 ED |
1761 | __fanout_link(sk, po); |
1762 | err = 0; | |
dc99f600 DM |
1763 | } |
1764 | } | |
7dfb100a WB |
1765 | spin_unlock(&po->bind_lock); |
1766 | ||
1767 | if (err && !refcount_read(&match->sk_ref)) { | |
1768 | list_del(&match->list); | |
1769 | kfree(match); | |
1770 | } | |
1771 | ||
afe62c68 | 1772 | out: |
d199fab6 ED |
1773 | if (err && rollover) { |
1774 | kfree(rollover); | |
0648ab70 WB |
1775 | po->rollover = NULL; |
1776 | } | |
d199fab6 | 1777 | mutex_unlock(&fanout_mutex); |
dc99f600 DM |
1778 | return err; |
1779 | } | |
1780 | ||
2bd624b4 AS |
1781 | /* If pkt_sk(sk)->fanout->sk_ref is zero, this function removes |
1782 | * pkt_sk(sk)->fanout from fanout_list and returns pkt_sk(sk)->fanout. | |
1783 | * It is the responsibility of the caller to call fanout_release_data() and | |
1784 | * free the returned packet_fanout (after synchronize_net()) | |
1785 | */ | |
1786 | static struct packet_fanout *fanout_release(struct sock *sk) | |
dc99f600 DM |
1787 | { |
1788 | struct packet_sock *po = pkt_sk(sk); | |
1789 | struct packet_fanout *f; | |
1790 | ||
fff3321d | 1791 | mutex_lock(&fanout_mutex); |
d199fab6 ED |
1792 | f = po->fanout; |
1793 | if (f) { | |
1794 | po->fanout = NULL; | |
1795 | ||
fb5c2c17 | 1796 | if (refcount_dec_and_test(&f->sk_ref)) |
d199fab6 | 1797 | list_del(&f->list); |
2bd624b4 AS |
1798 | else |
1799 | f = NULL; | |
dc99f600 | 1800 | |
d199fab6 ED |
1801 | if (po->rollover) |
1802 | kfree_rcu(po->rollover, rcu); | |
dc99f600 DM |
1803 | } |
1804 | mutex_unlock(&fanout_mutex); | |
2bd624b4 AS |
1805 | |
1806 | return f; | |
dc99f600 | 1807 | } |
1da177e4 | 1808 | |
3c70c132 DB |
1809 | static bool packet_extra_vlan_len_allowed(const struct net_device *dev, |
1810 | struct sk_buff *skb) | |
1811 | { | |
1812 | /* Earlier code assumed this would be a VLAN pkt, double-check | |
1813 | * this now that we have the actual packet in hand. We can only | |
1814 | * do this check on Ethernet devices. | |
1815 | */ | |
1816 | if (unlikely(dev->type != ARPHRD_ETHER)) | |
1817 | return false; | |
1818 | ||
1819 | skb_reset_mac_header(skb); | |
1820 | return likely(eth_hdr(skb)->h_proto == htons(ETH_P_8021Q)); | |
1821 | } | |
1822 | ||
90ddc4f0 | 1823 | static const struct proto_ops packet_ops; |
1da177e4 | 1824 | |
90ddc4f0 | 1825 | static const struct proto_ops packet_ops_spkt; |
1da177e4 | 1826 | |
40d4e3df ED |
1827 | static int packet_rcv_spkt(struct sk_buff *skb, struct net_device *dev, |
1828 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
1829 | { |
1830 | struct sock *sk; | |
1831 | struct sockaddr_pkt *spkt; | |
1832 | ||
1833 | /* | |
1834 | * When we registered the protocol we saved the socket in the data | |
1835 | * field for just this event. | |
1836 | */ | |
1837 | ||
1838 | sk = pt->af_packet_priv; | |
1ce4f28b | 1839 | |
1da177e4 LT |
1840 | /* |
1841 | * Yank back the headers [hope the device set this | |
1842 | * right or kerboom...] | |
1843 | * | |
1844 | * Incoming packets have ll header pulled, | |
1845 | * push it back. | |
1846 | * | |
98e399f8 | 1847 | * For outgoing ones skb->data == skb_mac_header(skb) |
1da177e4 LT |
1848 | * so that this procedure is noop. |
1849 | */ | |
1850 | ||
1851 | if (skb->pkt_type == PACKET_LOOPBACK) | |
1852 | goto out; | |
1853 | ||
09ad9bc7 | 1854 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
1855 | goto out; |
1856 | ||
40d4e3df ED |
1857 | skb = skb_share_check(skb, GFP_ATOMIC); |
1858 | if (skb == NULL) | |
1da177e4 LT |
1859 | goto oom; |
1860 | ||
1861 | /* drop any routing info */ | |
adf30907 | 1862 | skb_dst_drop(skb); |
1da177e4 | 1863 | |
84531c24 PO |
1864 | /* drop conntrack reference */ |
1865 | nf_reset(skb); | |
1866 | ||
ffbc6111 | 1867 | spkt = &PACKET_SKB_CB(skb)->sa.pkt; |
1da177e4 | 1868 | |
98e399f8 | 1869 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
1870 | |
1871 | /* | |
1872 | * The SOCK_PACKET socket receives _all_ frames. | |
1873 | */ | |
1874 | ||
1875 | spkt->spkt_family = dev->type; | |
1876 | strlcpy(spkt->spkt_device, dev->name, sizeof(spkt->spkt_device)); | |
1877 | spkt->spkt_protocol = skb->protocol; | |
1878 | ||
1879 | /* | |
1880 | * Charge the memory to the socket. This is done specifically | |
1881 | * to prevent sockets using all the memory up. | |
1882 | */ | |
1883 | ||
40d4e3df | 1884 | if (sock_queue_rcv_skb(sk, skb) == 0) |
1da177e4 LT |
1885 | return 0; |
1886 | ||
1887 | out: | |
1888 | kfree_skb(skb); | |
1889 | oom: | |
1890 | return 0; | |
1891 | } | |
1892 | ||
1893 | ||
1894 | /* | |
1895 | * Output a raw packet to a device layer. This bypasses all the other | |
1896 | * protocol layers and you must therefore supply it with a complete frame | |
1897 | */ | |
1ce4f28b | 1898 | |
1b784140 YX |
1899 | static int packet_sendmsg_spkt(struct socket *sock, struct msghdr *msg, |
1900 | size_t len) | |
1da177e4 LT |
1901 | { |
1902 | struct sock *sk = sock->sk; | |
342dfc30 | 1903 | DECLARE_SOCKADDR(struct sockaddr_pkt *, saddr, msg->msg_name); |
1a35ca80 | 1904 | struct sk_buff *skb = NULL; |
1da177e4 | 1905 | struct net_device *dev; |
c14ac945 | 1906 | struct sockcm_cookie sockc; |
40d4e3df | 1907 | __be16 proto = 0; |
1da177e4 | 1908 | int err; |
3bdc0eba | 1909 | int extra_len = 0; |
1ce4f28b | 1910 | |
1da177e4 | 1911 | /* |
1ce4f28b | 1912 | * Get and verify the address. |
1da177e4 LT |
1913 | */ |
1914 | ||
40d4e3df | 1915 | if (saddr) { |
1da177e4 | 1916 | if (msg->msg_namelen < sizeof(struct sockaddr)) |
40d4e3df ED |
1917 | return -EINVAL; |
1918 | if (msg->msg_namelen == sizeof(struct sockaddr_pkt)) | |
1919 | proto = saddr->spkt_protocol; | |
1920 | } else | |
1921 | return -ENOTCONN; /* SOCK_PACKET must be sent giving an address */ | |
1da177e4 LT |
1922 | |
1923 | /* | |
1ce4f28b | 1924 | * Find the device first to size check it |
1da177e4 LT |
1925 | */ |
1926 | ||
de74e92a | 1927 | saddr->spkt_device[sizeof(saddr->spkt_device) - 1] = 0; |
1a35ca80 | 1928 | retry: |
654d1f8a ED |
1929 | rcu_read_lock(); |
1930 | dev = dev_get_by_name_rcu(sock_net(sk), saddr->spkt_device); | |
1da177e4 LT |
1931 | err = -ENODEV; |
1932 | if (dev == NULL) | |
1933 | goto out_unlock; | |
1ce4f28b | 1934 | |
d5e76b0a DM |
1935 | err = -ENETDOWN; |
1936 | if (!(dev->flags & IFF_UP)) | |
1937 | goto out_unlock; | |
1938 | ||
1da177e4 | 1939 | /* |
40d4e3df ED |
1940 | * You may not queue a frame bigger than the mtu. This is the lowest level |
1941 | * raw protocol and you must do your own fragmentation at this level. | |
1da177e4 | 1942 | */ |
1ce4f28b | 1943 | |
3bdc0eba BG |
1944 | if (unlikely(sock_flag(sk, SOCK_NOFCS))) { |
1945 | if (!netif_supports_nofcs(dev)) { | |
1946 | err = -EPROTONOSUPPORT; | |
1947 | goto out_unlock; | |
1948 | } | |
1949 | extra_len = 4; /* We're doing our own CRC */ | |
1950 | } | |
1951 | ||
1da177e4 | 1952 | err = -EMSGSIZE; |
3bdc0eba | 1953 | if (len > dev->mtu + dev->hard_header_len + VLAN_HLEN + extra_len) |
1da177e4 LT |
1954 | goto out_unlock; |
1955 | ||
1a35ca80 ED |
1956 | if (!skb) { |
1957 | size_t reserved = LL_RESERVED_SPACE(dev); | |
4ce40912 | 1958 | int tlen = dev->needed_tailroom; |
1a35ca80 ED |
1959 | unsigned int hhlen = dev->header_ops ? dev->hard_header_len : 0; |
1960 | ||
1961 | rcu_read_unlock(); | |
4ce40912 | 1962 | skb = sock_wmalloc(sk, len + reserved + tlen, 0, GFP_KERNEL); |
1a35ca80 ED |
1963 | if (skb == NULL) |
1964 | return -ENOBUFS; | |
1965 | /* FIXME: Save some space for broken drivers that write a hard | |
1966 | * header at transmission time by themselves. PPP is the notable | |
1967 | * one here. This should really be fixed at the driver level. | |
1968 | */ | |
1969 | skb_reserve(skb, reserved); | |
1970 | skb_reset_network_header(skb); | |
1971 | ||
1972 | /* Try to align data part correctly */ | |
1973 | if (hhlen) { | |
1974 | skb->data -= hhlen; | |
1975 | skb->tail -= hhlen; | |
1976 | if (len < hhlen) | |
1977 | skb_reset_network_header(skb); | |
1978 | } | |
6ce8e9ce | 1979 | err = memcpy_from_msg(skb_put(skb, len), msg, len); |
1a35ca80 ED |
1980 | if (err) |
1981 | goto out_free; | |
1982 | goto retry; | |
1da177e4 LT |
1983 | } |
1984 | ||
9ed988cd WB |
1985 | if (!dev_validate_header(dev, skb->data, len)) { |
1986 | err = -EINVAL; | |
1987 | goto out_unlock; | |
1988 | } | |
3c70c132 DB |
1989 | if (len > (dev->mtu + dev->hard_header_len + extra_len) && |
1990 | !packet_extra_vlan_len_allowed(dev, skb)) { | |
1991 | err = -EMSGSIZE; | |
1992 | goto out_unlock; | |
57f89bfa | 1993 | } |
1a35ca80 | 1994 | |
edbe7746 | 1995 | sockc.tsflags = sk->sk_tsflags; |
c14ac945 SHY |
1996 | if (msg->msg_controllen) { |
1997 | err = sock_cmsg_send(sk, msg, &sockc); | |
f8e7718c | 1998 | if (unlikely(err)) |
c14ac945 | 1999 | goto out_unlock; |
c14ac945 SHY |
2000 | } |
2001 | ||
1da177e4 LT |
2002 | skb->protocol = proto; |
2003 | skb->dev = dev; | |
2004 | skb->priority = sk->sk_priority; | |
2d37a186 | 2005 | skb->mark = sk->sk_mark; |
bf84a010 | 2006 | |
c14ac945 | 2007 | sock_tx_timestamp(sk, sockc.tsflags, &skb_shinfo(skb)->tx_flags); |
1da177e4 | 2008 | |
3bdc0eba BG |
2009 | if (unlikely(extra_len == 4)) |
2010 | skb->no_fcs = 1; | |
2011 | ||
40893fd0 | 2012 | skb_probe_transport_header(skb, 0); |
c1aad275 | 2013 | |
1da177e4 | 2014 | dev_queue_xmit(skb); |
654d1f8a | 2015 | rcu_read_unlock(); |
40d4e3df | 2016 | return len; |
1da177e4 | 2017 | |
1da177e4 | 2018 | out_unlock: |
654d1f8a | 2019 | rcu_read_unlock(); |
1a35ca80 ED |
2020 | out_free: |
2021 | kfree_skb(skb); | |
1da177e4 LT |
2022 | return err; |
2023 | } | |
1da177e4 | 2024 | |
ff936a04 AS |
2025 | static unsigned int run_filter(struct sk_buff *skb, |
2026 | const struct sock *sk, | |
2027 | unsigned int res) | |
1da177e4 LT |
2028 | { |
2029 | struct sk_filter *filter; | |
fda9ef5d | 2030 | |
80f8f102 ED |
2031 | rcu_read_lock(); |
2032 | filter = rcu_dereference(sk->sk_filter); | |
dbcb5855 | 2033 | if (filter != NULL) |
ff936a04 | 2034 | res = bpf_prog_run_clear_cb(filter->prog, skb); |
80f8f102 | 2035 | rcu_read_unlock(); |
1da177e4 | 2036 | |
dbcb5855 | 2037 | return res; |
1da177e4 LT |
2038 | } |
2039 | ||
16cc1400 WB |
2040 | static int packet_rcv_vnet(struct msghdr *msg, const struct sk_buff *skb, |
2041 | size_t *len) | |
2042 | { | |
2043 | struct virtio_net_hdr vnet_hdr; | |
2044 | ||
2045 | if (*len < sizeof(vnet_hdr)) | |
2046 | return -EINVAL; | |
2047 | *len -= sizeof(vnet_hdr); | |
2048 | ||
6391a448 | 2049 | if (virtio_net_hdr_from_skb(skb, &vnet_hdr, vio_le(), true)) |
16cc1400 WB |
2050 | return -EINVAL; |
2051 | ||
2052 | return memcpy_to_msg(msg, (void *)&vnet_hdr, sizeof(vnet_hdr)); | |
2053 | } | |
2054 | ||
1da177e4 | 2055 | /* |
62ab0812 ED |
2056 | * This function makes lazy skb cloning in hope that most of packets |
2057 | * are discarded by BPF. | |
2058 | * | |
2059 | * Note tricky part: we DO mangle shared skb! skb->data, skb->len | |
2060 | * and skb->cb are mangled. It works because (and until) packets | |
2061 | * falling here are owned by current CPU. Output packets are cloned | |
2062 | * by dev_queue_xmit_nit(), input packets are processed by net_bh | |
2063 | * sequencially, so that if we return skb to original state on exit, | |
2064 | * we will not harm anyone. | |
1da177e4 LT |
2065 | */ |
2066 | ||
40d4e3df ED |
2067 | static int packet_rcv(struct sk_buff *skb, struct net_device *dev, |
2068 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
2069 | { |
2070 | struct sock *sk; | |
2071 | struct sockaddr_ll *sll; | |
2072 | struct packet_sock *po; | |
40d4e3df | 2073 | u8 *skb_head = skb->data; |
1da177e4 | 2074 | int skb_len = skb->len; |
dbcb5855 | 2075 | unsigned int snaplen, res; |
da37845f | 2076 | bool is_drop_n_account = false; |
1da177e4 LT |
2077 | |
2078 | if (skb->pkt_type == PACKET_LOOPBACK) | |
2079 | goto drop; | |
2080 | ||
2081 | sk = pt->af_packet_priv; | |
2082 | po = pkt_sk(sk); | |
2083 | ||
09ad9bc7 | 2084 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
2085 | goto drop; |
2086 | ||
1da177e4 LT |
2087 | skb->dev = dev; |
2088 | ||
3b04ddde | 2089 | if (dev->header_ops) { |
1da177e4 | 2090 | /* The device has an explicit notion of ll header, |
62ab0812 ED |
2091 | * exported to higher levels. |
2092 | * | |
2093 | * Otherwise, the device hides details of its frame | |
2094 | * structure, so that corresponding packet head is | |
2095 | * never delivered to user. | |
1da177e4 LT |
2096 | */ |
2097 | if (sk->sk_type != SOCK_DGRAM) | |
98e399f8 | 2098 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
2099 | else if (skb->pkt_type == PACKET_OUTGOING) { |
2100 | /* Special case: outgoing packets have ll header at head */ | |
bbe735e4 | 2101 | skb_pull(skb, skb_network_offset(skb)); |
1da177e4 LT |
2102 | } |
2103 | } | |
2104 | ||
2105 | snaplen = skb->len; | |
2106 | ||
dbcb5855 DM |
2107 | res = run_filter(skb, sk, snaplen); |
2108 | if (!res) | |
fda9ef5d | 2109 | goto drop_n_restore; |
dbcb5855 DM |
2110 | if (snaplen > res) |
2111 | snaplen = res; | |
1da177e4 | 2112 | |
0fd7bac6 | 2113 | if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) |
1da177e4 LT |
2114 | goto drop_n_acct; |
2115 | ||
2116 | if (skb_shared(skb)) { | |
2117 | struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC); | |
2118 | if (nskb == NULL) | |
2119 | goto drop_n_acct; | |
2120 | ||
2121 | if (skb_head != skb->data) { | |
2122 | skb->data = skb_head; | |
2123 | skb->len = skb_len; | |
2124 | } | |
abc4e4fa | 2125 | consume_skb(skb); |
1da177e4 LT |
2126 | skb = nskb; |
2127 | } | |
2128 | ||
b4772ef8 | 2129 | sock_skb_cb_check_size(sizeof(*PACKET_SKB_CB(skb)) + MAX_ADDR_LEN - 8); |
ffbc6111 HX |
2130 | |
2131 | sll = &PACKET_SKB_CB(skb)->sa.ll; | |
1da177e4 | 2132 | sll->sll_hatype = dev->type; |
1da177e4 | 2133 | sll->sll_pkttype = skb->pkt_type; |
8032b464 | 2134 | if (unlikely(po->origdev)) |
80feaacb PWJ |
2135 | sll->sll_ifindex = orig_dev->ifindex; |
2136 | else | |
2137 | sll->sll_ifindex = dev->ifindex; | |
1da177e4 | 2138 | |
b95cce35 | 2139 | sll->sll_halen = dev_parse_header(skb, sll->sll_addr); |
1da177e4 | 2140 | |
2472d761 EB |
2141 | /* sll->sll_family and sll->sll_protocol are set in packet_recvmsg(). |
2142 | * Use their space for storing the original skb length. | |
2143 | */ | |
2144 | PACKET_SKB_CB(skb)->sa.origlen = skb->len; | |
8dc41944 | 2145 | |
1da177e4 LT |
2146 | if (pskb_trim(skb, snaplen)) |
2147 | goto drop_n_acct; | |
2148 | ||
2149 | skb_set_owner_r(skb, sk); | |
2150 | skb->dev = NULL; | |
adf30907 | 2151 | skb_dst_drop(skb); |
1da177e4 | 2152 | |
84531c24 PO |
2153 | /* drop conntrack reference */ |
2154 | nf_reset(skb); | |
2155 | ||
1da177e4 | 2156 | spin_lock(&sk->sk_receive_queue.lock); |
ee80fbf3 | 2157 | po->stats.stats1.tp_packets++; |
3bc3b96f | 2158 | sock_skb_set_dropcount(sk, skb); |
1da177e4 LT |
2159 | __skb_queue_tail(&sk->sk_receive_queue, skb); |
2160 | spin_unlock(&sk->sk_receive_queue.lock); | |
676d2369 | 2161 | sk->sk_data_ready(sk); |
1da177e4 LT |
2162 | return 0; |
2163 | ||
2164 | drop_n_acct: | |
da37845f | 2165 | is_drop_n_account = true; |
7091fbd8 | 2166 | spin_lock(&sk->sk_receive_queue.lock); |
ee80fbf3 | 2167 | po->stats.stats1.tp_drops++; |
7091fbd8 WB |
2168 | atomic_inc(&sk->sk_drops); |
2169 | spin_unlock(&sk->sk_receive_queue.lock); | |
1da177e4 LT |
2170 | |
2171 | drop_n_restore: | |
2172 | if (skb_head != skb->data && skb_shared(skb)) { | |
2173 | skb->data = skb_head; | |
2174 | skb->len = skb_len; | |
2175 | } | |
2176 | drop: | |
da37845f WJ |
2177 | if (!is_drop_n_account) |
2178 | consume_skb(skb); | |
2179 | else | |
2180 | kfree_skb(skb); | |
1da177e4 LT |
2181 | return 0; |
2182 | } | |
2183 | ||
40d4e3df ED |
2184 | static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, |
2185 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
2186 | { |
2187 | struct sock *sk; | |
2188 | struct packet_sock *po; | |
2189 | struct sockaddr_ll *sll; | |
184f489e | 2190 | union tpacket_uhdr h; |
40d4e3df | 2191 | u8 *skb_head = skb->data; |
1da177e4 | 2192 | int skb_len = skb->len; |
dbcb5855 | 2193 | unsigned int snaplen, res; |
f6fb8f10 | 2194 | unsigned long status = TP_STATUS_USER; |
bbd6ef87 | 2195 | unsigned short macoff, netoff, hdrlen; |
1da177e4 | 2196 | struct sk_buff *copy_skb = NULL; |
bbd6ef87 | 2197 | struct timespec ts; |
b9c32fb2 | 2198 | __u32 ts_status; |
da37845f | 2199 | bool is_drop_n_account = false; |
edbd58be | 2200 | bool do_vnet = false; |
1da177e4 | 2201 | |
51846355 AW |
2202 | /* struct tpacket{2,3}_hdr is aligned to a multiple of TPACKET_ALIGNMENT. |
2203 | * We may add members to them until current aligned size without forcing | |
2204 | * userspace to call getsockopt(..., PACKET_HDRLEN, ...). | |
2205 | */ | |
2206 | BUILD_BUG_ON(TPACKET_ALIGN(sizeof(*h.h2)) != 32); | |
2207 | BUILD_BUG_ON(TPACKET_ALIGN(sizeof(*h.h3)) != 48); | |
2208 | ||
1da177e4 LT |
2209 | if (skb->pkt_type == PACKET_LOOPBACK) |
2210 | goto drop; | |
2211 | ||
2212 | sk = pt->af_packet_priv; | |
2213 | po = pkt_sk(sk); | |
2214 | ||
09ad9bc7 | 2215 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
2216 | goto drop; |
2217 | ||
3b04ddde | 2218 | if (dev->header_ops) { |
1da177e4 | 2219 | if (sk->sk_type != SOCK_DGRAM) |
98e399f8 | 2220 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
2221 | else if (skb->pkt_type == PACKET_OUTGOING) { |
2222 | /* Special case: outgoing packets have ll header at head */ | |
bbe735e4 | 2223 | skb_pull(skb, skb_network_offset(skb)); |
1da177e4 LT |
2224 | } |
2225 | } | |
2226 | ||
2227 | snaplen = skb->len; | |
2228 | ||
dbcb5855 DM |
2229 | res = run_filter(skb, sk, snaplen); |
2230 | if (!res) | |
fda9ef5d | 2231 | goto drop_n_restore; |
68c2e5de AD |
2232 | |
2233 | if (skb->ip_summed == CHECKSUM_PARTIAL) | |
2234 | status |= TP_STATUS_CSUMNOTREADY; | |
682f048b AD |
2235 | else if (skb->pkt_type != PACKET_OUTGOING && |
2236 | (skb->ip_summed == CHECKSUM_COMPLETE || | |
2237 | skb_csum_unnecessary(skb))) | |
2238 | status |= TP_STATUS_CSUM_VALID; | |
68c2e5de | 2239 | |
dbcb5855 DM |
2240 | if (snaplen > res) |
2241 | snaplen = res; | |
1da177e4 LT |
2242 | |
2243 | if (sk->sk_type == SOCK_DGRAM) { | |
8913336a PM |
2244 | macoff = netoff = TPACKET_ALIGN(po->tp_hdrlen) + 16 + |
2245 | po->tp_reserve; | |
1da177e4 | 2246 | } else { |
95c96174 | 2247 | unsigned int maclen = skb_network_offset(skb); |
bbd6ef87 | 2248 | netoff = TPACKET_ALIGN(po->tp_hdrlen + |
8913336a | 2249 | (maclen < 16 ? 16 : maclen)) + |
58d19b19 | 2250 | po->tp_reserve; |
edbd58be | 2251 | if (po->has_vnet_hdr) { |
58d19b19 | 2252 | netoff += sizeof(struct virtio_net_hdr); |
edbd58be BP |
2253 | do_vnet = true; |
2254 | } | |
1da177e4 LT |
2255 | macoff = netoff - maclen; |
2256 | } | |
f6fb8f10 | 2257 | if (po->tp_version <= TPACKET_V2) { |
2258 | if (macoff + snaplen > po->rx_ring.frame_size) { | |
2259 | if (po->copy_thresh && | |
0fd7bac6 | 2260 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) { |
f6fb8f10 | 2261 | if (skb_shared(skb)) { |
2262 | copy_skb = skb_clone(skb, GFP_ATOMIC); | |
2263 | } else { | |
2264 | copy_skb = skb_get(skb); | |
2265 | skb_head = skb->data; | |
2266 | } | |
2267 | if (copy_skb) | |
2268 | skb_set_owner_r(copy_skb, sk); | |
1da177e4 | 2269 | } |
f6fb8f10 | 2270 | snaplen = po->rx_ring.frame_size - macoff; |
edbd58be | 2271 | if ((int)snaplen < 0) { |
f6fb8f10 | 2272 | snaplen = 0; |
edbd58be BP |
2273 | do_vnet = false; |
2274 | } | |
1da177e4 | 2275 | } |
dc808110 ED |
2276 | } else if (unlikely(macoff + snaplen > |
2277 | GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len)) { | |
2278 | u32 nval; | |
2279 | ||
2280 | nval = GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len - macoff; | |
2281 | pr_err_once("tpacket_rcv: packet too big, clamped from %u to %u. macoff=%u\n", | |
2282 | snaplen, nval, macoff); | |
2283 | snaplen = nval; | |
2284 | if (unlikely((int)snaplen < 0)) { | |
2285 | snaplen = 0; | |
2286 | macoff = GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len; | |
edbd58be | 2287 | do_vnet = false; |
dc808110 | 2288 | } |
1da177e4 | 2289 | } |
1da177e4 | 2290 | spin_lock(&sk->sk_receive_queue.lock); |
f6fb8f10 | 2291 | h.raw = packet_current_rx_frame(po, skb, |
2292 | TP_STATUS_KERNEL, (macoff+snaplen)); | |
bbd6ef87 | 2293 | if (!h.raw) |
58d19b19 | 2294 | goto drop_n_account; |
f6fb8f10 | 2295 | if (po->tp_version <= TPACKET_V2) { |
2296 | packet_increment_rx_head(po, &po->rx_ring); | |
2297 | /* | |
2298 | * LOSING will be reported till you read the stats, | |
2299 | * because it's COR - Clear On Read. | |
2300 | * Anyways, moving it for V1/V2 only as V3 doesn't need this | |
2301 | * at packet level. | |
2302 | */ | |
ee80fbf3 | 2303 | if (po->stats.stats1.tp_drops) |
f6fb8f10 | 2304 | status |= TP_STATUS_LOSING; |
2305 | } | |
ee80fbf3 | 2306 | po->stats.stats1.tp_packets++; |
1da177e4 LT |
2307 | if (copy_skb) { |
2308 | status |= TP_STATUS_COPY; | |
2309 | __skb_queue_tail(&sk->sk_receive_queue, copy_skb); | |
2310 | } | |
1da177e4 LT |
2311 | spin_unlock(&sk->sk_receive_queue.lock); |
2312 | ||
edbd58be | 2313 | if (do_vnet) { |
5a213881 JR |
2314 | if (virtio_net_hdr_from_skb(skb, h.raw + macoff - |
2315 | sizeof(struct virtio_net_hdr), | |
6391a448 | 2316 | vio_le(), true)) { |
58d19b19 WB |
2317 | spin_lock(&sk->sk_receive_queue.lock); |
2318 | goto drop_n_account; | |
2319 | } | |
2320 | } | |
2321 | ||
bbd6ef87 | 2322 | skb_copy_bits(skb, 0, h.raw + macoff, snaplen); |
b9c32fb2 DB |
2323 | |
2324 | if (!(ts_status = tpacket_get_timestamp(skb, &ts, po->tp_tstamp))) | |
7a51384c | 2325 | getnstimeofday(&ts); |
1da177e4 | 2326 | |
b9c32fb2 DB |
2327 | status |= ts_status; |
2328 | ||
bbd6ef87 PM |
2329 | switch (po->tp_version) { |
2330 | case TPACKET_V1: | |
2331 | h.h1->tp_len = skb->len; | |
2332 | h.h1->tp_snaplen = snaplen; | |
2333 | h.h1->tp_mac = macoff; | |
2334 | h.h1->tp_net = netoff; | |
4b457bdf DB |
2335 | h.h1->tp_sec = ts.tv_sec; |
2336 | h.h1->tp_usec = ts.tv_nsec / NSEC_PER_USEC; | |
bbd6ef87 PM |
2337 | hdrlen = sizeof(*h.h1); |
2338 | break; | |
2339 | case TPACKET_V2: | |
2340 | h.h2->tp_len = skb->len; | |
2341 | h.h2->tp_snaplen = snaplen; | |
2342 | h.h2->tp_mac = macoff; | |
2343 | h.h2->tp_net = netoff; | |
bbd6ef87 PM |
2344 | h.h2->tp_sec = ts.tv_sec; |
2345 | h.h2->tp_nsec = ts.tv_nsec; | |
df8a39de JP |
2346 | if (skb_vlan_tag_present(skb)) { |
2347 | h.h2->tp_vlan_tci = skb_vlan_tag_get(skb); | |
a0cdfcf3 AW |
2348 | h.h2->tp_vlan_tpid = ntohs(skb->vlan_proto); |
2349 | status |= TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
a3bcc23e BG |
2350 | } else { |
2351 | h.h2->tp_vlan_tci = 0; | |
a0cdfcf3 | 2352 | h.h2->tp_vlan_tpid = 0; |
a3bcc23e | 2353 | } |
e4d26f4b | 2354 | memset(h.h2->tp_padding, 0, sizeof(h.h2->tp_padding)); |
bbd6ef87 PM |
2355 | hdrlen = sizeof(*h.h2); |
2356 | break; | |
f6fb8f10 | 2357 | case TPACKET_V3: |
2358 | /* tp_nxt_offset,vlan are already populated above. | |
2359 | * So DONT clear those fields here | |
2360 | */ | |
2361 | h.h3->tp_status |= status; | |
2362 | h.h3->tp_len = skb->len; | |
2363 | h.h3->tp_snaplen = snaplen; | |
2364 | h.h3->tp_mac = macoff; | |
2365 | h.h3->tp_net = netoff; | |
f6fb8f10 | 2366 | h.h3->tp_sec = ts.tv_sec; |
2367 | h.h3->tp_nsec = ts.tv_nsec; | |
e4d26f4b | 2368 | memset(h.h3->tp_padding, 0, sizeof(h.h3->tp_padding)); |
f6fb8f10 | 2369 | hdrlen = sizeof(*h.h3); |
2370 | break; | |
bbd6ef87 PM |
2371 | default: |
2372 | BUG(); | |
2373 | } | |
1da177e4 | 2374 | |
bbd6ef87 | 2375 | sll = h.raw + TPACKET_ALIGN(hdrlen); |
b95cce35 | 2376 | sll->sll_halen = dev_parse_header(skb, sll->sll_addr); |
1da177e4 LT |
2377 | sll->sll_family = AF_PACKET; |
2378 | sll->sll_hatype = dev->type; | |
2379 | sll->sll_protocol = skb->protocol; | |
2380 | sll->sll_pkttype = skb->pkt_type; | |
8032b464 | 2381 | if (unlikely(po->origdev)) |
80feaacb PWJ |
2382 | sll->sll_ifindex = orig_dev->ifindex; |
2383 | else | |
2384 | sll->sll_ifindex = dev->ifindex; | |
1da177e4 | 2385 | |
e16aa207 | 2386 | smp_mb(); |
f0d4eb29 | 2387 | |
f6dafa95 | 2388 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 |
f0d4eb29 | 2389 | if (po->tp_version <= TPACKET_V2) { |
0af55bb5 CG |
2390 | u8 *start, *end; |
2391 | ||
f0d4eb29 DB |
2392 | end = (u8 *) PAGE_ALIGN((unsigned long) h.raw + |
2393 | macoff + snaplen); | |
2394 | ||
2395 | for (start = h.raw; start < end; start += PAGE_SIZE) | |
2396 | flush_dcache_page(pgv_to_page(start)); | |
1da177e4 | 2397 | } |
f0d4eb29 | 2398 | smp_wmb(); |
f6dafa95 | 2399 | #endif |
f0d4eb29 | 2400 | |
da413eec | 2401 | if (po->tp_version <= TPACKET_V2) { |
f6fb8f10 | 2402 | __packet_set_status(po, h.raw, status); |
da413eec DC |
2403 | sk->sk_data_ready(sk); |
2404 | } else { | |
f6fb8f10 | 2405 | prb_clear_blk_fill_status(&po->rx_ring); |
da413eec | 2406 | } |
1da177e4 LT |
2407 | |
2408 | drop_n_restore: | |
2409 | if (skb_head != skb->data && skb_shared(skb)) { | |
2410 | skb->data = skb_head; | |
2411 | skb->len = skb_len; | |
2412 | } | |
2413 | drop: | |
da37845f WJ |
2414 | if (!is_drop_n_account) |
2415 | consume_skb(skb); | |
2416 | else | |
2417 | kfree_skb(skb); | |
1da177e4 LT |
2418 | return 0; |
2419 | ||
58d19b19 | 2420 | drop_n_account: |
da37845f | 2421 | is_drop_n_account = true; |
ee80fbf3 | 2422 | po->stats.stats1.tp_drops++; |
1da177e4 LT |
2423 | spin_unlock(&sk->sk_receive_queue.lock); |
2424 | ||
676d2369 | 2425 | sk->sk_data_ready(sk); |
acb5d75b | 2426 | kfree_skb(copy_skb); |
1da177e4 LT |
2427 | goto drop_n_restore; |
2428 | } | |
2429 | ||
69e3c75f JB |
2430 | static void tpacket_destruct_skb(struct sk_buff *skb) |
2431 | { | |
2432 | struct packet_sock *po = pkt_sk(skb->sk); | |
1da177e4 | 2433 | |
69e3c75f | 2434 | if (likely(po->tx_ring.pg_vec)) { |
f0d4eb29 | 2435 | void *ph; |
b9c32fb2 DB |
2436 | __u32 ts; |
2437 | ||
69e3c75f | 2438 | ph = skb_shinfo(skb)->destructor_arg; |
b0138408 | 2439 | packet_dec_pending(&po->tx_ring); |
b9c32fb2 DB |
2440 | |
2441 | ts = __packet_set_timestamp(po, ph, skb); | |
2442 | __packet_set_status(po, ph, TP_STATUS_AVAILABLE | ts); | |
69e3c75f JB |
2443 | } |
2444 | ||
2445 | sock_wfree(skb); | |
2446 | } | |
2447 | ||
c72219b7 DB |
2448 | static void tpacket_set_protocol(const struct net_device *dev, |
2449 | struct sk_buff *skb) | |
2450 | { | |
2451 | if (dev->type == ARPHRD_ETHER) { | |
2452 | skb_reset_mac_header(skb); | |
2453 | skb->protocol = eth_hdr(skb)->h_proto; | |
2454 | } | |
2455 | } | |
2456 | ||
16cc1400 WB |
2457 | static int __packet_snd_vnet_parse(struct virtio_net_hdr *vnet_hdr, size_t len) |
2458 | { | |
16cc1400 WB |
2459 | if ((vnet_hdr->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
2460 | (__virtio16_to_cpu(vio_le(), vnet_hdr->csum_start) + | |
2461 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_offset) + 2 > | |
2462 | __virtio16_to_cpu(vio_le(), vnet_hdr->hdr_len))) | |
2463 | vnet_hdr->hdr_len = __cpu_to_virtio16(vio_le(), | |
2464 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_start) + | |
2465 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_offset) + 2); | |
2466 | ||
2467 | if (__virtio16_to_cpu(vio_le(), vnet_hdr->hdr_len) > len) | |
2468 | return -EINVAL; | |
2469 | ||
16cc1400 WB |
2470 | return 0; |
2471 | } | |
2472 | ||
2473 | static int packet_snd_vnet_parse(struct msghdr *msg, size_t *len, | |
2474 | struct virtio_net_hdr *vnet_hdr) | |
2475 | { | |
16cc1400 WB |
2476 | if (*len < sizeof(*vnet_hdr)) |
2477 | return -EINVAL; | |
2478 | *len -= sizeof(*vnet_hdr); | |
2479 | ||
cbbd26b8 | 2480 | if (!copy_from_iter_full(vnet_hdr, sizeof(*vnet_hdr), &msg->msg_iter)) |
16cc1400 WB |
2481 | return -EFAULT; |
2482 | ||
2483 | return __packet_snd_vnet_parse(vnet_hdr, *len); | |
2484 | } | |
2485 | ||
40d4e3df | 2486 | static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, |
8d39b4a6 | 2487 | void *frame, struct net_device *dev, void *data, int tp_len, |
c14ac945 SHY |
2488 | __be16 proto, unsigned char *addr, int hlen, int copylen, |
2489 | const struct sockcm_cookie *sockc) | |
69e3c75f | 2490 | { |
184f489e | 2491 | union tpacket_uhdr ph; |
8d39b4a6 | 2492 | int to_write, offset, len, nr_frags, len_max; |
69e3c75f JB |
2493 | struct socket *sock = po->sk.sk_socket; |
2494 | struct page *page; | |
69e3c75f JB |
2495 | int err; |
2496 | ||
2497 | ph.raw = frame; | |
2498 | ||
2499 | skb->protocol = proto; | |
2500 | skb->dev = dev; | |
2501 | skb->priority = po->sk.sk_priority; | |
2d37a186 | 2502 | skb->mark = po->sk.sk_mark; |
c14ac945 | 2503 | sock_tx_timestamp(&po->sk, sockc->tsflags, &skb_shinfo(skb)->tx_flags); |
69e3c75f JB |
2504 | skb_shinfo(skb)->destructor_arg = ph.raw; |
2505 | ||
ae641949 | 2506 | skb_reserve(skb, hlen); |
69e3c75f | 2507 | skb_reset_network_header(skb); |
c1aad275 | 2508 | |
69e3c75f JB |
2509 | to_write = tp_len; |
2510 | ||
2511 | if (sock->type == SOCK_DGRAM) { | |
2512 | err = dev_hard_header(skb, dev, ntohs(proto), addr, | |
2513 | NULL, tp_len); | |
2514 | if (unlikely(err < 0)) | |
2515 | return -EINVAL; | |
1d036d25 | 2516 | } else if (copylen) { |
9ed988cd WB |
2517 | int hdrlen = min_t(int, copylen, tp_len); |
2518 | ||
69e3c75f | 2519 | skb_push(skb, dev->hard_header_len); |
1d036d25 | 2520 | skb_put(skb, copylen - dev->hard_header_len); |
9ed988cd | 2521 | err = skb_store_bits(skb, 0, data, hdrlen); |
69e3c75f JB |
2522 | if (unlikely(err)) |
2523 | return err; | |
9ed988cd WB |
2524 | if (!dev_validate_header(dev, skb->data, hdrlen)) |
2525 | return -EINVAL; | |
c72219b7 DB |
2526 | if (!skb->protocol) |
2527 | tpacket_set_protocol(dev, skb); | |
69e3c75f | 2528 | |
9ed988cd WB |
2529 | data += hdrlen; |
2530 | to_write -= hdrlen; | |
69e3c75f JB |
2531 | } |
2532 | ||
69e3c75f JB |
2533 | offset = offset_in_page(data); |
2534 | len_max = PAGE_SIZE - offset; | |
2535 | len = ((to_write > len_max) ? len_max : to_write); | |
2536 | ||
2537 | skb->data_len = to_write; | |
2538 | skb->len += to_write; | |
2539 | skb->truesize += to_write; | |
14afee4b | 2540 | refcount_add(to_write, &po->sk.sk_wmem_alloc); |
69e3c75f JB |
2541 | |
2542 | while (likely(to_write)) { | |
2543 | nr_frags = skb_shinfo(skb)->nr_frags; | |
2544 | ||
2545 | if (unlikely(nr_frags >= MAX_SKB_FRAGS)) { | |
40d4e3df ED |
2546 | pr_err("Packet exceed the number of skb frags(%lu)\n", |
2547 | MAX_SKB_FRAGS); | |
69e3c75f JB |
2548 | return -EFAULT; |
2549 | } | |
2550 | ||
0af55bb5 CG |
2551 | page = pgv_to_page(data); |
2552 | data += len; | |
69e3c75f JB |
2553 | flush_dcache_page(page); |
2554 | get_page(page); | |
0af55bb5 | 2555 | skb_fill_page_desc(skb, nr_frags, page, offset, len); |
69e3c75f JB |
2556 | to_write -= len; |
2557 | offset = 0; | |
2558 | len_max = PAGE_SIZE; | |
2559 | len = ((to_write > len_max) ? len_max : to_write); | |
2560 | } | |
2561 | ||
8fd6c80d | 2562 | skb_probe_transport_header(skb, 0); |
efdfa2f7 | 2563 | |
69e3c75f JB |
2564 | return tp_len; |
2565 | } | |
2566 | ||
8d39b4a6 WB |
2567 | static int tpacket_parse_header(struct packet_sock *po, void *frame, |
2568 | int size_max, void **data) | |
2569 | { | |
2570 | union tpacket_uhdr ph; | |
2571 | int tp_len, off; | |
2572 | ||
2573 | ph.raw = frame; | |
2574 | ||
2575 | switch (po->tp_version) { | |
7f953ab2 SV |
2576 | case TPACKET_V3: |
2577 | if (ph.h3->tp_next_offset != 0) { | |
2578 | pr_warn_once("variable sized slot not supported"); | |
2579 | return -EINVAL; | |
2580 | } | |
2581 | tp_len = ph.h3->tp_len; | |
2582 | break; | |
8d39b4a6 WB |
2583 | case TPACKET_V2: |
2584 | tp_len = ph.h2->tp_len; | |
2585 | break; | |
2586 | default: | |
2587 | tp_len = ph.h1->tp_len; | |
2588 | break; | |
2589 | } | |
2590 | if (unlikely(tp_len > size_max)) { | |
2591 | pr_err("packet size is too long (%d > %d)\n", tp_len, size_max); | |
2592 | return -EMSGSIZE; | |
2593 | } | |
2594 | ||
2595 | if (unlikely(po->tp_tx_has_off)) { | |
2596 | int off_min, off_max; | |
2597 | ||
2598 | off_min = po->tp_hdrlen - sizeof(struct sockaddr_ll); | |
2599 | off_max = po->tx_ring.frame_size - tp_len; | |
2600 | if (po->sk.sk_type == SOCK_DGRAM) { | |
2601 | switch (po->tp_version) { | |
7f953ab2 SV |
2602 | case TPACKET_V3: |
2603 | off = ph.h3->tp_net; | |
2604 | break; | |
8d39b4a6 WB |
2605 | case TPACKET_V2: |
2606 | off = ph.h2->tp_net; | |
2607 | break; | |
2608 | default: | |
2609 | off = ph.h1->tp_net; | |
2610 | break; | |
2611 | } | |
2612 | } else { | |
2613 | switch (po->tp_version) { | |
7f953ab2 SV |
2614 | case TPACKET_V3: |
2615 | off = ph.h3->tp_mac; | |
2616 | break; | |
8d39b4a6 WB |
2617 | case TPACKET_V2: |
2618 | off = ph.h2->tp_mac; | |
2619 | break; | |
2620 | default: | |
2621 | off = ph.h1->tp_mac; | |
2622 | break; | |
2623 | } | |
2624 | } | |
2625 | if (unlikely((off < off_min) || (off_max < off))) | |
2626 | return -EINVAL; | |
2627 | } else { | |
2628 | off = po->tp_hdrlen - sizeof(struct sockaddr_ll); | |
2629 | } | |
2630 | ||
2631 | *data = frame + off; | |
2632 | return tp_len; | |
2633 | } | |
2634 | ||
69e3c75f JB |
2635 | static int tpacket_snd(struct packet_sock *po, struct msghdr *msg) |
2636 | { | |
69e3c75f JB |
2637 | struct sk_buff *skb; |
2638 | struct net_device *dev; | |
1d036d25 | 2639 | struct virtio_net_hdr *vnet_hdr = NULL; |
c14ac945 | 2640 | struct sockcm_cookie sockc; |
69e3c75f | 2641 | __be16 proto; |
09effa67 | 2642 | int err, reserve = 0; |
40d4e3df | 2643 | void *ph; |
342dfc30 | 2644 | DECLARE_SOCKADDR(struct sockaddr_ll *, saddr, msg->msg_name); |
87a2fd28 | 2645 | bool need_wait = !(msg->msg_flags & MSG_DONTWAIT); |
69e3c75f JB |
2646 | int tp_len, size_max; |
2647 | unsigned char *addr; | |
8d39b4a6 | 2648 | void *data; |
69e3c75f | 2649 | int len_sum = 0; |
9e67030a | 2650 | int status = TP_STATUS_AVAILABLE; |
1d036d25 | 2651 | int hlen, tlen, copylen = 0; |
69e3c75f | 2652 | |
69e3c75f JB |
2653 | mutex_lock(&po->pg_vec_lock); |
2654 | ||
66e56cd4 | 2655 | if (likely(saddr == NULL)) { |
e40526cb | 2656 | dev = packet_cached_dev_get(po); |
69e3c75f JB |
2657 | proto = po->num; |
2658 | addr = NULL; | |
2659 | } else { | |
2660 | err = -EINVAL; | |
2661 | if (msg->msg_namelen < sizeof(struct sockaddr_ll)) | |
2662 | goto out; | |
2663 | if (msg->msg_namelen < (saddr->sll_halen | |
2664 | + offsetof(struct sockaddr_ll, | |
2665 | sll_addr))) | |
2666 | goto out; | |
69e3c75f JB |
2667 | proto = saddr->sll_protocol; |
2668 | addr = saddr->sll_addr; | |
827d9780 | 2669 | dev = dev_get_by_index(sock_net(&po->sk), saddr->sll_ifindex); |
69e3c75f JB |
2670 | } |
2671 | ||
69e3c75f JB |
2672 | err = -ENXIO; |
2673 | if (unlikely(dev == NULL)) | |
2674 | goto out; | |
69e3c75f JB |
2675 | err = -ENETDOWN; |
2676 | if (unlikely(!(dev->flags & IFF_UP))) | |
2677 | goto out_put; | |
2678 | ||
d19b183c DCS |
2679 | sockc.tsflags = po->sk.sk_tsflags; |
2680 | if (msg->msg_controllen) { | |
2681 | err = sock_cmsg_send(&po->sk, msg, &sockc); | |
2682 | if (unlikely(err)) | |
2683 | goto out_put; | |
2684 | } | |
2685 | ||
5cfb4c8d DB |
2686 | if (po->sk.sk_socket->type == SOCK_RAW) |
2687 | reserve = dev->hard_header_len; | |
69e3c75f | 2688 | size_max = po->tx_ring.frame_size |
b5dd884e | 2689 | - (po->tp_hdrlen - sizeof(struct sockaddr_ll)); |
69e3c75f | 2690 | |
1d036d25 | 2691 | if ((size_max > dev->mtu + reserve + VLAN_HLEN) && !po->has_vnet_hdr) |
5cfb4c8d | 2692 | size_max = dev->mtu + reserve + VLAN_HLEN; |
09effa67 | 2693 | |
69e3c75f JB |
2694 | do { |
2695 | ph = packet_current_frame(po, &po->tx_ring, | |
87a2fd28 | 2696 | TP_STATUS_SEND_REQUEST); |
69e3c75f | 2697 | if (unlikely(ph == NULL)) { |
87a2fd28 DB |
2698 | if (need_wait && need_resched()) |
2699 | schedule(); | |
69e3c75f JB |
2700 | continue; |
2701 | } | |
2702 | ||
8d39b4a6 WB |
2703 | skb = NULL; |
2704 | tp_len = tpacket_parse_header(po, ph, size_max, &data); | |
2705 | if (tp_len < 0) | |
2706 | goto tpacket_error; | |
2707 | ||
69e3c75f | 2708 | status = TP_STATUS_SEND_REQUEST; |
ae641949 HX |
2709 | hlen = LL_RESERVED_SPACE(dev); |
2710 | tlen = dev->needed_tailroom; | |
1d036d25 WB |
2711 | if (po->has_vnet_hdr) { |
2712 | vnet_hdr = data; | |
2713 | data += sizeof(*vnet_hdr); | |
2714 | tp_len -= sizeof(*vnet_hdr); | |
2715 | if (tp_len < 0 || | |
2716 | __packet_snd_vnet_parse(vnet_hdr, tp_len)) { | |
2717 | tp_len = -EINVAL; | |
2718 | goto tpacket_error; | |
2719 | } | |
2720 | copylen = __virtio16_to_cpu(vio_le(), | |
2721 | vnet_hdr->hdr_len); | |
2722 | } | |
9ed988cd | 2723 | copylen = max_t(int, copylen, dev->hard_header_len); |
69e3c75f | 2724 | skb = sock_alloc_send_skb(&po->sk, |
1d036d25 WB |
2725 | hlen + tlen + sizeof(struct sockaddr_ll) + |
2726 | (copylen - dev->hard_header_len), | |
fbf33a28 | 2727 | !need_wait, &err); |
69e3c75f | 2728 | |
fbf33a28 KM |
2729 | if (unlikely(skb == NULL)) { |
2730 | /* we assume the socket was initially writeable ... */ | |
2731 | if (likely(len_sum > 0)) | |
2732 | err = len_sum; | |
69e3c75f | 2733 | goto out_status; |
fbf33a28 | 2734 | } |
8d39b4a6 | 2735 | tp_len = tpacket_fill_skb(po, skb, ph, dev, data, tp_len, proto, |
c14ac945 | 2736 | addr, hlen, copylen, &sockc); |
dbd46ab4 | 2737 | if (likely(tp_len >= 0) && |
5cfb4c8d | 2738 | tp_len > dev->mtu + reserve && |
1d036d25 | 2739 | !po->has_vnet_hdr && |
3c70c132 DB |
2740 | !packet_extra_vlan_len_allowed(dev, skb)) |
2741 | tp_len = -EMSGSIZE; | |
69e3c75f JB |
2742 | |
2743 | if (unlikely(tp_len < 0)) { | |
8d39b4a6 | 2744 | tpacket_error: |
69e3c75f JB |
2745 | if (po->tp_loss) { |
2746 | __packet_set_status(po, ph, | |
2747 | TP_STATUS_AVAILABLE); | |
2748 | packet_increment_head(&po->tx_ring); | |
2749 | kfree_skb(skb); | |
2750 | continue; | |
2751 | } else { | |
2752 | status = TP_STATUS_WRONG_FORMAT; | |
2753 | err = tp_len; | |
2754 | goto out_status; | |
2755 | } | |
2756 | } | |
2757 | ||
db60eb5f JR |
2758 | if (po->has_vnet_hdr && virtio_net_hdr_to_skb(skb, vnet_hdr, |
2759 | vio_le())) { | |
1d036d25 WB |
2760 | tp_len = -EINVAL; |
2761 | goto tpacket_error; | |
2762 | } | |
2763 | ||
69e3c75f JB |
2764 | skb->destructor = tpacket_destruct_skb; |
2765 | __packet_set_status(po, ph, TP_STATUS_SENDING); | |
b0138408 | 2766 | packet_inc_pending(&po->tx_ring); |
69e3c75f JB |
2767 | |
2768 | status = TP_STATUS_SEND_REQUEST; | |
d346a3fa | 2769 | err = po->xmit(skb); |
eb70df13 JP |
2770 | if (unlikely(err > 0)) { |
2771 | err = net_xmit_errno(err); | |
2772 | if (err && __packet_get_status(po, ph) == | |
2773 | TP_STATUS_AVAILABLE) { | |
2774 | /* skb was destructed already */ | |
2775 | skb = NULL; | |
2776 | goto out_status; | |
2777 | } | |
2778 | /* | |
2779 | * skb was dropped but not destructed yet; | |
2780 | * let's treat it like congestion or err < 0 | |
2781 | */ | |
2782 | err = 0; | |
2783 | } | |
69e3c75f JB |
2784 | packet_increment_head(&po->tx_ring); |
2785 | len_sum += tp_len; | |
b0138408 DB |
2786 | } while (likely((ph != NULL) || |
2787 | /* Note: packet_read_pending() might be slow if we have | |
2788 | * to call it as it's per_cpu variable, but in fast-path | |
2789 | * we already short-circuit the loop with the first | |
2790 | * condition, and luckily don't have to go that path | |
2791 | * anyway. | |
2792 | */ | |
2793 | (need_wait && packet_read_pending(&po->tx_ring)))); | |
69e3c75f JB |
2794 | |
2795 | err = len_sum; | |
2796 | goto out_put; | |
2797 | ||
69e3c75f JB |
2798 | out_status: |
2799 | __packet_set_status(po, ph, status); | |
2800 | kfree_skb(skb); | |
2801 | out_put: | |
e40526cb | 2802 | dev_put(dev); |
69e3c75f JB |
2803 | out: |
2804 | mutex_unlock(&po->pg_vec_lock); | |
2805 | return err; | |
2806 | } | |
69e3c75f | 2807 | |
eea49cc9 OJ |
2808 | static struct sk_buff *packet_alloc_skb(struct sock *sk, size_t prepad, |
2809 | size_t reserve, size_t len, | |
2810 | size_t linear, int noblock, | |
2811 | int *err) | |
bfd5f4a3 SS |
2812 | { |
2813 | struct sk_buff *skb; | |
2814 | ||
2815 | /* Under a page? Don't bother with paged skb. */ | |
2816 | if (prepad + len < PAGE_SIZE || !linear) | |
2817 | linear = len; | |
2818 | ||
2819 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, | |
28d64271 | 2820 | err, 0); |
bfd5f4a3 SS |
2821 | if (!skb) |
2822 | return NULL; | |
2823 | ||
2824 | skb_reserve(skb, reserve); | |
2825 | skb_put(skb, linear); | |
2826 | skb->data_len = len - linear; | |
2827 | skb->len += len - linear; | |
2828 | ||
2829 | return skb; | |
2830 | } | |
2831 | ||
d346a3fa | 2832 | static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 LT |
2833 | { |
2834 | struct sock *sk = sock->sk; | |
342dfc30 | 2835 | DECLARE_SOCKADDR(struct sockaddr_ll *, saddr, msg->msg_name); |
1da177e4 LT |
2836 | struct sk_buff *skb; |
2837 | struct net_device *dev; | |
0e11c91e | 2838 | __be16 proto; |
1da177e4 | 2839 | unsigned char *addr; |
827d9780 | 2840 | int err, reserve = 0; |
c7d39e32 | 2841 | struct sockcm_cookie sockc; |
bfd5f4a3 SS |
2842 | struct virtio_net_hdr vnet_hdr = { 0 }; |
2843 | int offset = 0; | |
bfd5f4a3 | 2844 | struct packet_sock *po = pkt_sk(sk); |
4dbd212d | 2845 | bool has_vnet_hdr = false; |
57031eb7 | 2846 | int hlen, tlen, linear; |
3bdc0eba | 2847 | int extra_len = 0; |
1da177e4 LT |
2848 | |
2849 | /* | |
1ce4f28b | 2850 | * Get and verify the address. |
1da177e4 | 2851 | */ |
1ce4f28b | 2852 | |
66e56cd4 | 2853 | if (likely(saddr == NULL)) { |
e40526cb | 2854 | dev = packet_cached_dev_get(po); |
1da177e4 LT |
2855 | proto = po->num; |
2856 | addr = NULL; | |
2857 | } else { | |
2858 | err = -EINVAL; | |
2859 | if (msg->msg_namelen < sizeof(struct sockaddr_ll)) | |
2860 | goto out; | |
0fb375fb EB |
2861 | if (msg->msg_namelen < (saddr->sll_halen + offsetof(struct sockaddr_ll, sll_addr))) |
2862 | goto out; | |
1da177e4 LT |
2863 | proto = saddr->sll_protocol; |
2864 | addr = saddr->sll_addr; | |
827d9780 | 2865 | dev = dev_get_by_index(sock_net(sk), saddr->sll_ifindex); |
1da177e4 LT |
2866 | } |
2867 | ||
1da177e4 | 2868 | err = -ENXIO; |
e40526cb | 2869 | if (unlikely(dev == NULL)) |
1da177e4 | 2870 | goto out_unlock; |
d5e76b0a | 2871 | err = -ENETDOWN; |
e40526cb | 2872 | if (unlikely(!(dev->flags & IFF_UP))) |
d5e76b0a DM |
2873 | goto out_unlock; |
2874 | ||
edbe7746 | 2875 | sockc.tsflags = sk->sk_tsflags; |
c7d39e32 EJ |
2876 | sockc.mark = sk->sk_mark; |
2877 | if (msg->msg_controllen) { | |
2878 | err = sock_cmsg_send(sk, msg, &sockc); | |
2879 | if (unlikely(err)) | |
2880 | goto out_unlock; | |
2881 | } | |
2882 | ||
e40526cb DB |
2883 | if (sock->type == SOCK_RAW) |
2884 | reserve = dev->hard_header_len; | |
bfd5f4a3 | 2885 | if (po->has_vnet_hdr) { |
16cc1400 WB |
2886 | err = packet_snd_vnet_parse(msg, &len, &vnet_hdr); |
2887 | if (err) | |
bfd5f4a3 | 2888 | goto out_unlock; |
4dbd212d | 2889 | has_vnet_hdr = true; |
bfd5f4a3 SS |
2890 | } |
2891 | ||
3bdc0eba BG |
2892 | if (unlikely(sock_flag(sk, SOCK_NOFCS))) { |
2893 | if (!netif_supports_nofcs(dev)) { | |
2894 | err = -EPROTONOSUPPORT; | |
2895 | goto out_unlock; | |
2896 | } | |
2897 | extra_len = 4; /* We're doing our own CRC */ | |
2898 | } | |
2899 | ||
1da177e4 | 2900 | err = -EMSGSIZE; |
16cc1400 WB |
2901 | if (!vnet_hdr.gso_type && |
2902 | (len > dev->mtu + reserve + VLAN_HLEN + extra_len)) | |
1da177e4 LT |
2903 | goto out_unlock; |
2904 | ||
bfd5f4a3 | 2905 | err = -ENOBUFS; |
ae641949 HX |
2906 | hlen = LL_RESERVED_SPACE(dev); |
2907 | tlen = dev->needed_tailroom; | |
57031eb7 WB |
2908 | linear = __virtio16_to_cpu(vio_le(), vnet_hdr.hdr_len); |
2909 | linear = max(linear, min_t(int, len, dev->hard_header_len)); | |
2910 | skb = packet_alloc_skb(sk, hlen + tlen, hlen, len, linear, | |
bfd5f4a3 | 2911 | msg->msg_flags & MSG_DONTWAIT, &err); |
40d4e3df | 2912 | if (skb == NULL) |
1da177e4 LT |
2913 | goto out_unlock; |
2914 | ||
bfd5f4a3 | 2915 | skb_set_network_header(skb, reserve); |
1da177e4 | 2916 | |
0c4e8581 | 2917 | err = -EINVAL; |
9c707762 WB |
2918 | if (sock->type == SOCK_DGRAM) { |
2919 | offset = dev_hard_header(skb, dev, ntohs(proto), addr, NULL, len); | |
46d2cfb1 | 2920 | if (unlikely(offset < 0)) |
9c707762 | 2921 | goto out_free; |
9c707762 | 2922 | } |
1da177e4 LT |
2923 | |
2924 | /* Returns -EFAULT on error */ | |
c0371da6 | 2925 | err = skb_copy_datagram_from_iter(skb, offset, &msg->msg_iter, len); |
1da177e4 LT |
2926 | if (err) |
2927 | goto out_free; | |
bf84a010 | 2928 | |
9ed988cd WB |
2929 | if (sock->type == SOCK_RAW && |
2930 | !dev_validate_header(dev, skb->data, len)) { | |
2931 | err = -EINVAL; | |
2932 | goto out_free; | |
2933 | } | |
2934 | ||
c14ac945 | 2935 | sock_tx_timestamp(sk, sockc.tsflags, &skb_shinfo(skb)->tx_flags); |
1da177e4 | 2936 | |
16cc1400 | 2937 | if (!vnet_hdr.gso_type && (len > dev->mtu + reserve + extra_len) && |
3c70c132 DB |
2938 | !packet_extra_vlan_len_allowed(dev, skb)) { |
2939 | err = -EMSGSIZE; | |
2940 | goto out_free; | |
57f89bfa BG |
2941 | } |
2942 | ||
09effa67 DM |
2943 | skb->protocol = proto; |
2944 | skb->dev = dev; | |
1da177e4 | 2945 | skb->priority = sk->sk_priority; |
c7d39e32 | 2946 | skb->mark = sockc.mark; |
0fd5d57b | 2947 | |
4dbd212d | 2948 | if (has_vnet_hdr) { |
db60eb5f | 2949 | err = virtio_net_hdr_to_skb(skb, &vnet_hdr, vio_le()); |
16cc1400 WB |
2950 | if (err) |
2951 | goto out_free; | |
2952 | len += sizeof(vnet_hdr); | |
bfd5f4a3 SS |
2953 | } |
2954 | ||
8fd6c80d DB |
2955 | skb_probe_transport_header(skb, reserve); |
2956 | ||
3bdc0eba BG |
2957 | if (unlikely(extra_len == 4)) |
2958 | skb->no_fcs = 1; | |
2959 | ||
d346a3fa | 2960 | err = po->xmit(skb); |
1da177e4 LT |
2961 | if (err > 0 && (err = net_xmit_errno(err)) != 0) |
2962 | goto out_unlock; | |
2963 | ||
e40526cb | 2964 | dev_put(dev); |
1da177e4 | 2965 | |
40d4e3df | 2966 | return len; |
1da177e4 LT |
2967 | |
2968 | out_free: | |
2969 | kfree_skb(skb); | |
2970 | out_unlock: | |
e40526cb | 2971 | if (dev) |
1da177e4 LT |
2972 | dev_put(dev); |
2973 | out: | |
2974 | return err; | |
2975 | } | |
2976 | ||
1b784140 | 2977 | static int packet_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
69e3c75f | 2978 | { |
69e3c75f JB |
2979 | struct sock *sk = sock->sk; |
2980 | struct packet_sock *po = pkt_sk(sk); | |
d346a3fa | 2981 | |
69e3c75f JB |
2982 | if (po->tx_ring.pg_vec) |
2983 | return tpacket_snd(po, msg); | |
2984 | else | |
69e3c75f JB |
2985 | return packet_snd(sock, msg, len); |
2986 | } | |
2987 | ||
1da177e4 LT |
2988 | /* |
2989 | * Close a PACKET socket. This is fairly simple. We immediately go | |
2990 | * to 'closed' state and remove our protocol entry in the device list. | |
2991 | */ | |
2992 | ||
2993 | static int packet_release(struct socket *sock) | |
2994 | { | |
2995 | struct sock *sk = sock->sk; | |
2996 | struct packet_sock *po; | |
2bd624b4 | 2997 | struct packet_fanout *f; |
d12d01d6 | 2998 | struct net *net; |
f6fb8f10 | 2999 | union tpacket_req_u req_u; |
1da177e4 LT |
3000 | |
3001 | if (!sk) | |
3002 | return 0; | |
3003 | ||
3b1e0a65 | 3004 | net = sock_net(sk); |
1da177e4 LT |
3005 | po = pkt_sk(sk); |
3006 | ||
0fa7fa98 | 3007 | mutex_lock(&net->packet.sklist_lock); |
808f5114 | 3008 | sk_del_node_init_rcu(sk); |
0fa7fa98 PE |
3009 | mutex_unlock(&net->packet.sklist_lock); |
3010 | ||
3011 | preempt_disable(); | |
920de804 | 3012 | sock_prot_inuse_add(net, sk->sk_prot, -1); |
0fa7fa98 | 3013 | preempt_enable(); |
1da177e4 | 3014 | |
808f5114 | 3015 | spin_lock(&po->bind_lock); |
ce06b03e | 3016 | unregister_prot_hook(sk, false); |
66e56cd4 DB |
3017 | packet_cached_dev_reset(po); |
3018 | ||
160ff18a BG |
3019 | if (po->prot_hook.dev) { |
3020 | dev_put(po->prot_hook.dev); | |
3021 | po->prot_hook.dev = NULL; | |
3022 | } | |
808f5114 | 3023 | spin_unlock(&po->bind_lock); |
1da177e4 | 3024 | |
1da177e4 | 3025 | packet_flush_mclist(sk); |
1da177e4 | 3026 | |
9665d5d6 PS |
3027 | if (po->rx_ring.pg_vec) { |
3028 | memset(&req_u, 0, sizeof(req_u)); | |
f6fb8f10 | 3029 | packet_set_ring(sk, &req_u, 1, 0); |
9665d5d6 | 3030 | } |
69e3c75f | 3031 | |
9665d5d6 PS |
3032 | if (po->tx_ring.pg_vec) { |
3033 | memset(&req_u, 0, sizeof(req_u)); | |
f6fb8f10 | 3034 | packet_set_ring(sk, &req_u, 1, 1); |
9665d5d6 | 3035 | } |
1da177e4 | 3036 | |
2bd624b4 | 3037 | f = fanout_release(sk); |
dc99f600 | 3038 | |
808f5114 | 3039 | synchronize_net(); |
2bd624b4 AS |
3040 | |
3041 | if (f) { | |
3042 | fanout_release_data(f); | |
3043 | kfree(f); | |
3044 | } | |
1da177e4 LT |
3045 | /* |
3046 | * Now the socket is dead. No more input will appear. | |
3047 | */ | |
1da177e4 LT |
3048 | sock_orphan(sk); |
3049 | sock->sk = NULL; | |
3050 | ||
3051 | /* Purge queues */ | |
3052 | ||
3053 | skb_queue_purge(&sk->sk_receive_queue); | |
b0138408 | 3054 | packet_free_pending(po); |
17ab56a2 | 3055 | sk_refcnt_debug_release(sk); |
1da177e4 LT |
3056 | |
3057 | sock_put(sk); | |
3058 | return 0; | |
3059 | } | |
3060 | ||
3061 | /* | |
3062 | * Attach a packet hook. | |
3063 | */ | |
3064 | ||
30f7ea1c FR |
3065 | static int packet_do_bind(struct sock *sk, const char *name, int ifindex, |
3066 | __be16 proto) | |
1da177e4 LT |
3067 | { |
3068 | struct packet_sock *po = pkt_sk(sk); | |
158cd4af | 3069 | struct net_device *dev_curr; |
902fefb8 DB |
3070 | __be16 proto_curr; |
3071 | bool need_rehook; | |
30f7ea1c FR |
3072 | struct net_device *dev = NULL; |
3073 | int ret = 0; | |
3074 | bool unlisted = false; | |
dc99f600 | 3075 | |
1da177e4 | 3076 | lock_sock(sk); |
1da177e4 | 3077 | spin_lock(&po->bind_lock); |
30f7ea1c FR |
3078 | rcu_read_lock(); |
3079 | ||
03e1d3d4 WB |
3080 | if (po->fanout) { |
3081 | ret = -EINVAL; | |
3082 | goto out_unlock; | |
3083 | } | |
3084 | ||
30f7ea1c FR |
3085 | if (name) { |
3086 | dev = dev_get_by_name_rcu(sock_net(sk), name); | |
3087 | if (!dev) { | |
3088 | ret = -ENODEV; | |
3089 | goto out_unlock; | |
3090 | } | |
3091 | } else if (ifindex) { | |
3092 | dev = dev_get_by_index_rcu(sock_net(sk), ifindex); | |
3093 | if (!dev) { | |
3094 | ret = -ENODEV; | |
3095 | goto out_unlock; | |
3096 | } | |
3097 | } | |
3098 | ||
3099 | if (dev) | |
3100 | dev_hold(dev); | |
66e56cd4 | 3101 | |
902fefb8 DB |
3102 | proto_curr = po->prot_hook.type; |
3103 | dev_curr = po->prot_hook.dev; | |
3104 | ||
3105 | need_rehook = proto_curr != proto || dev_curr != dev; | |
3106 | ||
3107 | if (need_rehook) { | |
30f7ea1c FR |
3108 | if (po->running) { |
3109 | rcu_read_unlock(); | |
3110 | __unregister_prot_hook(sk, true); | |
3111 | rcu_read_lock(); | |
3112 | dev_curr = po->prot_hook.dev; | |
3113 | if (dev) | |
3114 | unlisted = !dev_get_by_index_rcu(sock_net(sk), | |
3115 | dev->ifindex); | |
3116 | } | |
1da177e4 | 3117 | |
902fefb8 DB |
3118 | po->num = proto; |
3119 | po->prot_hook.type = proto; | |
902fefb8 | 3120 | |
30f7ea1c FR |
3121 | if (unlikely(unlisted)) { |
3122 | dev_put(dev); | |
3123 | po->prot_hook.dev = NULL; | |
3124 | po->ifindex = -1; | |
3125 | packet_cached_dev_reset(po); | |
3126 | } else { | |
3127 | po->prot_hook.dev = dev; | |
3128 | po->ifindex = dev ? dev->ifindex : 0; | |
3129 | packet_cached_dev_assign(po, dev); | |
3130 | } | |
902fefb8 | 3131 | } |
158cd4af LW |
3132 | if (dev_curr) |
3133 | dev_put(dev_curr); | |
66e56cd4 | 3134 | |
902fefb8 | 3135 | if (proto == 0 || !need_rehook) |
1da177e4 LT |
3136 | goto out_unlock; |
3137 | ||
30f7ea1c | 3138 | if (!unlisted && (!dev || (dev->flags & IFF_UP))) { |
ce06b03e | 3139 | register_prot_hook(sk); |
be85d4ad UT |
3140 | } else { |
3141 | sk->sk_err = ENETDOWN; | |
3142 | if (!sock_flag(sk, SOCK_DEAD)) | |
3143 | sk->sk_error_report(sk); | |
1da177e4 LT |
3144 | } |
3145 | ||
3146 | out_unlock: | |
30f7ea1c | 3147 | rcu_read_unlock(); |
1da177e4 LT |
3148 | spin_unlock(&po->bind_lock); |
3149 | release_sock(sk); | |
30f7ea1c | 3150 | return ret; |
1da177e4 LT |
3151 | } |
3152 | ||
3153 | /* | |
3154 | * Bind a packet socket to a device | |
3155 | */ | |
3156 | ||
40d4e3df ED |
3157 | static int packet_bind_spkt(struct socket *sock, struct sockaddr *uaddr, |
3158 | int addr_len) | |
1da177e4 | 3159 | { |
40d4e3df | 3160 | struct sock *sk = sock->sk; |
540e2894 | 3161 | char name[sizeof(uaddr->sa_data) + 1]; |
1ce4f28b | 3162 | |
1da177e4 LT |
3163 | /* |
3164 | * Check legality | |
3165 | */ | |
1ce4f28b | 3166 | |
8ae55f04 | 3167 | if (addr_len != sizeof(struct sockaddr)) |
1da177e4 | 3168 | return -EINVAL; |
540e2894 AP |
3169 | /* uaddr->sa_data comes from the userspace, it's not guaranteed to be |
3170 | * zero-terminated. | |
3171 | */ | |
3172 | memcpy(name, uaddr->sa_data, sizeof(uaddr->sa_data)); | |
3173 | name[sizeof(uaddr->sa_data)] = 0; | |
1da177e4 | 3174 | |
30f7ea1c | 3175 | return packet_do_bind(sk, name, 0, pkt_sk(sk)->num); |
1da177e4 | 3176 | } |
1da177e4 LT |
3177 | |
3178 | static int packet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
3179 | { | |
40d4e3df ED |
3180 | struct sockaddr_ll *sll = (struct sockaddr_ll *)uaddr; |
3181 | struct sock *sk = sock->sk; | |
1da177e4 LT |
3182 | |
3183 | /* | |
3184 | * Check legality | |
3185 | */ | |
1ce4f28b | 3186 | |
1da177e4 LT |
3187 | if (addr_len < sizeof(struct sockaddr_ll)) |
3188 | return -EINVAL; | |
3189 | if (sll->sll_family != AF_PACKET) | |
3190 | return -EINVAL; | |
3191 | ||
30f7ea1c FR |
3192 | return packet_do_bind(sk, NULL, sll->sll_ifindex, |
3193 | sll->sll_protocol ? : pkt_sk(sk)->num); | |
1da177e4 LT |
3194 | } |
3195 | ||
3196 | static struct proto packet_proto = { | |
3197 | .name = "PACKET", | |
3198 | .owner = THIS_MODULE, | |
3199 | .obj_size = sizeof(struct packet_sock), | |
3200 | }; | |
3201 | ||
3202 | /* | |
1ce4f28b | 3203 | * Create a packet of type SOCK_PACKET. |
1da177e4 LT |
3204 | */ |
3205 | ||
3f378b68 EP |
3206 | static int packet_create(struct net *net, struct socket *sock, int protocol, |
3207 | int kern) | |
1da177e4 LT |
3208 | { |
3209 | struct sock *sk; | |
3210 | struct packet_sock *po; | |
0e11c91e | 3211 | __be16 proto = (__force __be16)protocol; /* weird, but documented */ |
1da177e4 LT |
3212 | int err; |
3213 | ||
df008c91 | 3214 | if (!ns_capable(net->user_ns, CAP_NET_RAW)) |
1da177e4 | 3215 | return -EPERM; |
be02097c DM |
3216 | if (sock->type != SOCK_DGRAM && sock->type != SOCK_RAW && |
3217 | sock->type != SOCK_PACKET) | |
1da177e4 LT |
3218 | return -ESOCKTNOSUPPORT; |
3219 | ||
3220 | sock->state = SS_UNCONNECTED; | |
3221 | ||
3222 | err = -ENOBUFS; | |
11aa9c28 | 3223 | sk = sk_alloc(net, PF_PACKET, GFP_KERNEL, &packet_proto, kern); |
1da177e4 LT |
3224 | if (sk == NULL) |
3225 | goto out; | |
3226 | ||
3227 | sock->ops = &packet_ops; | |
1da177e4 LT |
3228 | if (sock->type == SOCK_PACKET) |
3229 | sock->ops = &packet_ops_spkt; | |
be02097c | 3230 | |
1da177e4 LT |
3231 | sock_init_data(sock, sk); |
3232 | ||
3233 | po = pkt_sk(sk); | |
3234 | sk->sk_family = PF_PACKET; | |
0e11c91e | 3235 | po->num = proto; |
d346a3fa | 3236 | po->xmit = dev_queue_xmit; |
66e56cd4 | 3237 | |
b0138408 DB |
3238 | err = packet_alloc_pending(po); |
3239 | if (err) | |
3240 | goto out2; | |
3241 | ||
66e56cd4 | 3242 | packet_cached_dev_reset(po); |
1da177e4 LT |
3243 | |
3244 | sk->sk_destruct = packet_sock_destruct; | |
17ab56a2 | 3245 | sk_refcnt_debug_inc(sk); |
1da177e4 LT |
3246 | |
3247 | /* | |
3248 | * Attach a protocol block | |
3249 | */ | |
3250 | ||
3251 | spin_lock_init(&po->bind_lock); | |
905db440 | 3252 | mutex_init(&po->pg_vec_lock); |
0648ab70 | 3253 | po->rollover = NULL; |
1da177e4 | 3254 | po->prot_hook.func = packet_rcv; |
be02097c | 3255 | |
1da177e4 LT |
3256 | if (sock->type == SOCK_PACKET) |
3257 | po->prot_hook.func = packet_rcv_spkt; | |
be02097c | 3258 | |
1da177e4 LT |
3259 | po->prot_hook.af_packet_priv = sk; |
3260 | ||
0e11c91e AV |
3261 | if (proto) { |
3262 | po->prot_hook.type = proto; | |
ce06b03e | 3263 | register_prot_hook(sk); |
1da177e4 LT |
3264 | } |
3265 | ||
0fa7fa98 | 3266 | mutex_lock(&net->packet.sklist_lock); |
808f5114 | 3267 | sk_add_node_rcu(sk, &net->packet.sklist); |
0fa7fa98 PE |
3268 | mutex_unlock(&net->packet.sklist_lock); |
3269 | ||
3270 | preempt_disable(); | |
3680453c | 3271 | sock_prot_inuse_add(net, &packet_proto, 1); |
0fa7fa98 | 3272 | preempt_enable(); |
808f5114 | 3273 | |
40d4e3df | 3274 | return 0; |
b0138408 DB |
3275 | out2: |
3276 | sk_free(sk); | |
1da177e4 LT |
3277 | out: |
3278 | return err; | |
3279 | } | |
3280 | ||
3281 | /* | |
3282 | * Pull a packet from our receive queue and hand it to the user. | |
3283 | * If necessary we block. | |
3284 | */ | |
3285 | ||
1b784140 YX |
3286 | static int packet_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
3287 | int flags) | |
1da177e4 LT |
3288 | { |
3289 | struct sock *sk = sock->sk; | |
3290 | struct sk_buff *skb; | |
3291 | int copied, err; | |
bfd5f4a3 | 3292 | int vnet_hdr_len = 0; |
2472d761 | 3293 | unsigned int origlen = 0; |
1da177e4 LT |
3294 | |
3295 | err = -EINVAL; | |
ed85b565 | 3296 | if (flags & ~(MSG_PEEK|MSG_DONTWAIT|MSG_TRUNC|MSG_CMSG_COMPAT|MSG_ERRQUEUE)) |
1da177e4 LT |
3297 | goto out; |
3298 | ||
3299 | #if 0 | |
3300 | /* What error should we return now? EUNATTACH? */ | |
3301 | if (pkt_sk(sk)->ifindex < 0) | |
3302 | return -ENODEV; | |
3303 | #endif | |
3304 | ||
ed85b565 | 3305 | if (flags & MSG_ERRQUEUE) { |
cb820f8e RC |
3306 | err = sock_recv_errqueue(sk, msg, len, |
3307 | SOL_PACKET, PACKET_TX_TIMESTAMP); | |
ed85b565 RC |
3308 | goto out; |
3309 | } | |
3310 | ||
1da177e4 LT |
3311 | /* |
3312 | * Call the generic datagram receiver. This handles all sorts | |
3313 | * of horrible races and re-entrancy so we can forget about it | |
3314 | * in the protocol layers. | |
3315 | * | |
3316 | * Now it will return ENETDOWN, if device have just gone down, | |
3317 | * but then it will block. | |
3318 | */ | |
3319 | ||
40d4e3df | 3320 | skb = skb_recv_datagram(sk, flags, flags & MSG_DONTWAIT, &err); |
1da177e4 LT |
3321 | |
3322 | /* | |
1ce4f28b | 3323 | * An error occurred so return it. Because skb_recv_datagram() |
1da177e4 LT |
3324 | * handles the blocking we don't see and worry about blocking |
3325 | * retries. | |
3326 | */ | |
3327 | ||
8ae55f04 | 3328 | if (skb == NULL) |
1da177e4 LT |
3329 | goto out; |
3330 | ||
2ccdbaa6 WB |
3331 | if (pkt_sk(sk)->pressure) |
3332 | packet_rcv_has_room(pkt_sk(sk), NULL); | |
3333 | ||
bfd5f4a3 | 3334 | if (pkt_sk(sk)->has_vnet_hdr) { |
16cc1400 WB |
3335 | err = packet_rcv_vnet(msg, skb, &len); |
3336 | if (err) | |
bfd5f4a3 | 3337 | goto out_free; |
16cc1400 | 3338 | vnet_hdr_len = sizeof(struct virtio_net_hdr); |
bfd5f4a3 SS |
3339 | } |
3340 | ||
f3d33426 HFS |
3341 | /* You lose any data beyond the buffer you gave. If it worries |
3342 | * a user program they can ask the device for its MTU | |
3343 | * anyway. | |
1da177e4 | 3344 | */ |
1da177e4 | 3345 | copied = skb->len; |
40d4e3df ED |
3346 | if (copied > len) { |
3347 | copied = len; | |
3348 | msg->msg_flags |= MSG_TRUNC; | |
1da177e4 LT |
3349 | } |
3350 | ||
51f3d02b | 3351 | err = skb_copy_datagram_msg(skb, 0, msg, copied); |
1da177e4 LT |
3352 | if (err) |
3353 | goto out_free; | |
3354 | ||
2472d761 EB |
3355 | if (sock->type != SOCK_PACKET) { |
3356 | struct sockaddr_ll *sll = &PACKET_SKB_CB(skb)->sa.ll; | |
3357 | ||
3358 | /* Original length was stored in sockaddr_ll fields */ | |
3359 | origlen = PACKET_SKB_CB(skb)->sa.origlen; | |
3360 | sll->sll_family = AF_PACKET; | |
3361 | sll->sll_protocol = skb->protocol; | |
3362 | } | |
3363 | ||
3b885787 | 3364 | sock_recv_ts_and_drops(msg, sk, skb); |
1da177e4 | 3365 | |
f3d33426 HFS |
3366 | if (msg->msg_name) { |
3367 | /* If the address length field is there to be filled | |
3368 | * in, we fill it in now. | |
3369 | */ | |
3370 | if (sock->type == SOCK_PACKET) { | |
342dfc30 | 3371 | __sockaddr_check_size(sizeof(struct sockaddr_pkt)); |
f3d33426 HFS |
3372 | msg->msg_namelen = sizeof(struct sockaddr_pkt); |
3373 | } else { | |
3374 | struct sockaddr_ll *sll = &PACKET_SKB_CB(skb)->sa.ll; | |
2472d761 | 3375 | |
f3d33426 HFS |
3376 | msg->msg_namelen = sll->sll_halen + |
3377 | offsetof(struct sockaddr_ll, sll_addr); | |
3378 | } | |
ffbc6111 HX |
3379 | memcpy(msg->msg_name, &PACKET_SKB_CB(skb)->sa, |
3380 | msg->msg_namelen); | |
f3d33426 | 3381 | } |
1da177e4 | 3382 | |
8dc41944 | 3383 | if (pkt_sk(sk)->auxdata) { |
ffbc6111 HX |
3384 | struct tpacket_auxdata aux; |
3385 | ||
3386 | aux.tp_status = TP_STATUS_USER; | |
3387 | if (skb->ip_summed == CHECKSUM_PARTIAL) | |
3388 | aux.tp_status |= TP_STATUS_CSUMNOTREADY; | |
682f048b AD |
3389 | else if (skb->pkt_type != PACKET_OUTGOING && |
3390 | (skb->ip_summed == CHECKSUM_COMPLETE || | |
3391 | skb_csum_unnecessary(skb))) | |
3392 | aux.tp_status |= TP_STATUS_CSUM_VALID; | |
3393 | ||
2472d761 | 3394 | aux.tp_len = origlen; |
ffbc6111 HX |
3395 | aux.tp_snaplen = skb->len; |
3396 | aux.tp_mac = 0; | |
bbe735e4 | 3397 | aux.tp_net = skb_network_offset(skb); |
df8a39de JP |
3398 | if (skb_vlan_tag_present(skb)) { |
3399 | aux.tp_vlan_tci = skb_vlan_tag_get(skb); | |
a0cdfcf3 AW |
3400 | aux.tp_vlan_tpid = ntohs(skb->vlan_proto); |
3401 | aux.tp_status |= TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
a3bcc23e BG |
3402 | } else { |
3403 | aux.tp_vlan_tci = 0; | |
a0cdfcf3 | 3404 | aux.tp_vlan_tpid = 0; |
a3bcc23e | 3405 | } |
ffbc6111 | 3406 | put_cmsg(msg, SOL_PACKET, PACKET_AUXDATA, sizeof(aux), &aux); |
8dc41944 HX |
3407 | } |
3408 | ||
1da177e4 LT |
3409 | /* |
3410 | * Free or return the buffer as appropriate. Again this | |
3411 | * hides all the races and re-entrancy issues from us. | |
3412 | */ | |
bfd5f4a3 | 3413 | err = vnet_hdr_len + ((flags&MSG_TRUNC) ? skb->len : copied); |
1da177e4 LT |
3414 | |
3415 | out_free: | |
3416 | skb_free_datagram(sk, skb); | |
3417 | out: | |
3418 | return err; | |
3419 | } | |
3420 | ||
1da177e4 LT |
3421 | static int packet_getname_spkt(struct socket *sock, struct sockaddr *uaddr, |
3422 | int *uaddr_len, int peer) | |
3423 | { | |
3424 | struct net_device *dev; | |
3425 | struct sock *sk = sock->sk; | |
3426 | ||
3427 | if (peer) | |
3428 | return -EOPNOTSUPP; | |
3429 | ||
3430 | uaddr->sa_family = AF_PACKET; | |
2dc85bf3 | 3431 | memset(uaddr->sa_data, 0, sizeof(uaddr->sa_data)); |
654d1f8a ED |
3432 | rcu_read_lock(); |
3433 | dev = dev_get_by_index_rcu(sock_net(sk), pkt_sk(sk)->ifindex); | |
3434 | if (dev) | |
2dc85bf3 | 3435 | strlcpy(uaddr->sa_data, dev->name, sizeof(uaddr->sa_data)); |
654d1f8a | 3436 | rcu_read_unlock(); |
1da177e4 LT |
3437 | *uaddr_len = sizeof(*uaddr); |
3438 | ||
3439 | return 0; | |
3440 | } | |
1da177e4 LT |
3441 | |
3442 | static int packet_getname(struct socket *sock, struct sockaddr *uaddr, | |
3443 | int *uaddr_len, int peer) | |
3444 | { | |
3445 | struct net_device *dev; | |
3446 | struct sock *sk = sock->sk; | |
3447 | struct packet_sock *po = pkt_sk(sk); | |
13cfa97b | 3448 | DECLARE_SOCKADDR(struct sockaddr_ll *, sll, uaddr); |
1da177e4 LT |
3449 | |
3450 | if (peer) | |
3451 | return -EOPNOTSUPP; | |
3452 | ||
3453 | sll->sll_family = AF_PACKET; | |
3454 | sll->sll_ifindex = po->ifindex; | |
3455 | sll->sll_protocol = po->num; | |
67286640 | 3456 | sll->sll_pkttype = 0; |
654d1f8a ED |
3457 | rcu_read_lock(); |
3458 | dev = dev_get_by_index_rcu(sock_net(sk), po->ifindex); | |
1da177e4 LT |
3459 | if (dev) { |
3460 | sll->sll_hatype = dev->type; | |
3461 | sll->sll_halen = dev->addr_len; | |
3462 | memcpy(sll->sll_addr, dev->dev_addr, dev->addr_len); | |
1da177e4 LT |
3463 | } else { |
3464 | sll->sll_hatype = 0; /* Bad: we have no ARPHRD_UNSPEC */ | |
3465 | sll->sll_halen = 0; | |
3466 | } | |
654d1f8a | 3467 | rcu_read_unlock(); |
0fb375fb | 3468 | *uaddr_len = offsetof(struct sockaddr_ll, sll_addr) + sll->sll_halen; |
1da177e4 LT |
3469 | |
3470 | return 0; | |
3471 | } | |
3472 | ||
2aeb0b88 WC |
3473 | static int packet_dev_mc(struct net_device *dev, struct packet_mclist *i, |
3474 | int what) | |
1da177e4 LT |
3475 | { |
3476 | switch (i->type) { | |
3477 | case PACKET_MR_MULTICAST: | |
1162563f JP |
3478 | if (i->alen != dev->addr_len) |
3479 | return -EINVAL; | |
1da177e4 | 3480 | if (what > 0) |
22bedad3 | 3481 | return dev_mc_add(dev, i->addr); |
1da177e4 | 3482 | else |
22bedad3 | 3483 | return dev_mc_del(dev, i->addr); |
1da177e4 LT |
3484 | break; |
3485 | case PACKET_MR_PROMISC: | |
2aeb0b88 | 3486 | return dev_set_promiscuity(dev, what); |
1da177e4 | 3487 | case PACKET_MR_ALLMULTI: |
2aeb0b88 | 3488 | return dev_set_allmulti(dev, what); |
d95ed927 | 3489 | case PACKET_MR_UNICAST: |
1162563f JP |
3490 | if (i->alen != dev->addr_len) |
3491 | return -EINVAL; | |
d95ed927 | 3492 | if (what > 0) |
a748ee24 | 3493 | return dev_uc_add(dev, i->addr); |
d95ed927 | 3494 | else |
a748ee24 | 3495 | return dev_uc_del(dev, i->addr); |
d95ed927 | 3496 | break; |
40d4e3df ED |
3497 | default: |
3498 | break; | |
1da177e4 | 3499 | } |
2aeb0b88 | 3500 | return 0; |
1da177e4 LT |
3501 | } |
3502 | ||
82f17091 FR |
3503 | static void packet_dev_mclist_delete(struct net_device *dev, |
3504 | struct packet_mclist **mlp) | |
1da177e4 | 3505 | { |
82f17091 FR |
3506 | struct packet_mclist *ml; |
3507 | ||
3508 | while ((ml = *mlp) != NULL) { | |
3509 | if (ml->ifindex == dev->ifindex) { | |
3510 | packet_dev_mc(dev, ml, -1); | |
3511 | *mlp = ml->next; | |
3512 | kfree(ml); | |
3513 | } else | |
3514 | mlp = &ml->next; | |
1da177e4 LT |
3515 | } |
3516 | } | |
3517 | ||
0fb375fb | 3518 | static int packet_mc_add(struct sock *sk, struct packet_mreq_max *mreq) |
1da177e4 LT |
3519 | { |
3520 | struct packet_sock *po = pkt_sk(sk); | |
3521 | struct packet_mclist *ml, *i; | |
3522 | struct net_device *dev; | |
3523 | int err; | |
3524 | ||
3525 | rtnl_lock(); | |
3526 | ||
3527 | err = -ENODEV; | |
3b1e0a65 | 3528 | dev = __dev_get_by_index(sock_net(sk), mreq->mr_ifindex); |
1da177e4 LT |
3529 | if (!dev) |
3530 | goto done; | |
3531 | ||
3532 | err = -EINVAL; | |
1162563f | 3533 | if (mreq->mr_alen > dev->addr_len) |
1da177e4 LT |
3534 | goto done; |
3535 | ||
3536 | err = -ENOBUFS; | |
8b3a7005 | 3537 | i = kmalloc(sizeof(*i), GFP_KERNEL); |
1da177e4 LT |
3538 | if (i == NULL) |
3539 | goto done; | |
3540 | ||
3541 | err = 0; | |
3542 | for (ml = po->mclist; ml; ml = ml->next) { | |
3543 | if (ml->ifindex == mreq->mr_ifindex && | |
3544 | ml->type == mreq->mr_type && | |
3545 | ml->alen == mreq->mr_alen && | |
3546 | memcmp(ml->addr, mreq->mr_address, ml->alen) == 0) { | |
3547 | ml->count++; | |
3548 | /* Free the new element ... */ | |
3549 | kfree(i); | |
3550 | goto done; | |
3551 | } | |
3552 | } | |
3553 | ||
3554 | i->type = mreq->mr_type; | |
3555 | i->ifindex = mreq->mr_ifindex; | |
3556 | i->alen = mreq->mr_alen; | |
3557 | memcpy(i->addr, mreq->mr_address, i->alen); | |
309cf37f | 3558 | memset(i->addr + i->alen, 0, sizeof(i->addr) - i->alen); |
1da177e4 LT |
3559 | i->count = 1; |
3560 | i->next = po->mclist; | |
3561 | po->mclist = i; | |
2aeb0b88 WC |
3562 | err = packet_dev_mc(dev, i, 1); |
3563 | if (err) { | |
3564 | po->mclist = i->next; | |
3565 | kfree(i); | |
3566 | } | |
1da177e4 LT |
3567 | |
3568 | done: | |
3569 | rtnl_unlock(); | |
3570 | return err; | |
3571 | } | |
3572 | ||
0fb375fb | 3573 | static int packet_mc_drop(struct sock *sk, struct packet_mreq_max *mreq) |
1da177e4 LT |
3574 | { |
3575 | struct packet_mclist *ml, **mlp; | |
3576 | ||
3577 | rtnl_lock(); | |
3578 | ||
3579 | for (mlp = &pkt_sk(sk)->mclist; (ml = *mlp) != NULL; mlp = &ml->next) { | |
3580 | if (ml->ifindex == mreq->mr_ifindex && | |
3581 | ml->type == mreq->mr_type && | |
3582 | ml->alen == mreq->mr_alen && | |
3583 | memcmp(ml->addr, mreq->mr_address, ml->alen) == 0) { | |
3584 | if (--ml->count == 0) { | |
3585 | struct net_device *dev; | |
3586 | *mlp = ml->next; | |
ad959e76 ED |
3587 | dev = __dev_get_by_index(sock_net(sk), ml->ifindex); |
3588 | if (dev) | |
1da177e4 | 3589 | packet_dev_mc(dev, ml, -1); |
1da177e4 LT |
3590 | kfree(ml); |
3591 | } | |
82f17091 | 3592 | break; |
1da177e4 LT |
3593 | } |
3594 | } | |
3595 | rtnl_unlock(); | |
82f17091 | 3596 | return 0; |
1da177e4 LT |
3597 | } |
3598 | ||
3599 | static void packet_flush_mclist(struct sock *sk) | |
3600 | { | |
3601 | struct packet_sock *po = pkt_sk(sk); | |
3602 | struct packet_mclist *ml; | |
3603 | ||
3604 | if (!po->mclist) | |
3605 | return; | |
3606 | ||
3607 | rtnl_lock(); | |
3608 | while ((ml = po->mclist) != NULL) { | |
3609 | struct net_device *dev; | |
3610 | ||
3611 | po->mclist = ml->next; | |
ad959e76 ED |
3612 | dev = __dev_get_by_index(sock_net(sk), ml->ifindex); |
3613 | if (dev != NULL) | |
1da177e4 | 3614 | packet_dev_mc(dev, ml, -1); |
1da177e4 LT |
3615 | kfree(ml); |
3616 | } | |
3617 | rtnl_unlock(); | |
3618 | } | |
1da177e4 LT |
3619 | |
3620 | static int | |
b7058842 | 3621 | packet_setsockopt(struct socket *sock, int level, int optname, char __user *optval, unsigned int optlen) |
1da177e4 LT |
3622 | { |
3623 | struct sock *sk = sock->sk; | |
8dc41944 | 3624 | struct packet_sock *po = pkt_sk(sk); |
1da177e4 LT |
3625 | int ret; |
3626 | ||
3627 | if (level != SOL_PACKET) | |
3628 | return -ENOPROTOOPT; | |
3629 | ||
69e3c75f | 3630 | switch (optname) { |
1ce4f28b | 3631 | case PACKET_ADD_MEMBERSHIP: |
1da177e4 LT |
3632 | case PACKET_DROP_MEMBERSHIP: |
3633 | { | |
0fb375fb EB |
3634 | struct packet_mreq_max mreq; |
3635 | int len = optlen; | |
3636 | memset(&mreq, 0, sizeof(mreq)); | |
3637 | if (len < sizeof(struct packet_mreq)) | |
1da177e4 | 3638 | return -EINVAL; |
0fb375fb EB |
3639 | if (len > sizeof(mreq)) |
3640 | len = sizeof(mreq); | |
40d4e3df | 3641 | if (copy_from_user(&mreq, optval, len)) |
1da177e4 | 3642 | return -EFAULT; |
0fb375fb EB |
3643 | if (len < (mreq.mr_alen + offsetof(struct packet_mreq, mr_address))) |
3644 | return -EINVAL; | |
1da177e4 LT |
3645 | if (optname == PACKET_ADD_MEMBERSHIP) |
3646 | ret = packet_mc_add(sk, &mreq); | |
3647 | else | |
3648 | ret = packet_mc_drop(sk, &mreq); | |
3649 | return ret; | |
3650 | } | |
a2efcfa0 | 3651 | |
1da177e4 | 3652 | case PACKET_RX_RING: |
69e3c75f | 3653 | case PACKET_TX_RING: |
1da177e4 | 3654 | { |
f6fb8f10 | 3655 | union tpacket_req_u req_u; |
3656 | int len; | |
1da177e4 | 3657 | |
f6fb8f10 | 3658 | switch (po->tp_version) { |
3659 | case TPACKET_V1: | |
3660 | case TPACKET_V2: | |
3661 | len = sizeof(req_u.req); | |
3662 | break; | |
3663 | case TPACKET_V3: | |
3664 | default: | |
3665 | len = sizeof(req_u.req3); | |
3666 | break; | |
3667 | } | |
3668 | if (optlen < len) | |
1da177e4 | 3669 | return -EINVAL; |
f6fb8f10 | 3670 | if (copy_from_user(&req_u.req, optval, len)) |
1da177e4 | 3671 | return -EFAULT; |
f6fb8f10 | 3672 | return packet_set_ring(sk, &req_u, 0, |
3673 | optname == PACKET_TX_RING); | |
1da177e4 LT |
3674 | } |
3675 | case PACKET_COPY_THRESH: | |
3676 | { | |
3677 | int val; | |
3678 | ||
40d4e3df | 3679 | if (optlen != sizeof(val)) |
1da177e4 | 3680 | return -EINVAL; |
40d4e3df | 3681 | if (copy_from_user(&val, optval, sizeof(val))) |
1da177e4 LT |
3682 | return -EFAULT; |
3683 | ||
3684 | pkt_sk(sk)->copy_thresh = val; | |
3685 | return 0; | |
3686 | } | |
bbd6ef87 PM |
3687 | case PACKET_VERSION: |
3688 | { | |
3689 | int val; | |
3690 | ||
3691 | if (optlen != sizeof(val)) | |
3692 | return -EINVAL; | |
bbd6ef87 PM |
3693 | if (copy_from_user(&val, optval, sizeof(val))) |
3694 | return -EFAULT; | |
3695 | switch (val) { | |
3696 | case TPACKET_V1: | |
3697 | case TPACKET_V2: | |
f6fb8f10 | 3698 | case TPACKET_V3: |
84ac7260 | 3699 | break; |
bbd6ef87 PM |
3700 | default: |
3701 | return -EINVAL; | |
3702 | } | |
84ac7260 PP |
3703 | lock_sock(sk); |
3704 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3705 | ret = -EBUSY; | |
3706 | } else { | |
3707 | po->tp_version = val; | |
3708 | ret = 0; | |
3709 | } | |
3710 | release_sock(sk); | |
3711 | return ret; | |
bbd6ef87 | 3712 | } |
8913336a PM |
3713 | case PACKET_RESERVE: |
3714 | { | |
3715 | unsigned int val; | |
3716 | ||
3717 | if (optlen != sizeof(val)) | |
3718 | return -EINVAL; | |
8913336a PM |
3719 | if (copy_from_user(&val, optval, sizeof(val))) |
3720 | return -EFAULT; | |
bcc5364b AK |
3721 | if (val > INT_MAX) |
3722 | return -EINVAL; | |
c27927e3 WB |
3723 | lock_sock(sk); |
3724 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3725 | ret = -EBUSY; | |
3726 | } else { | |
3727 | po->tp_reserve = val; | |
3728 | ret = 0; | |
3729 | } | |
3730 | release_sock(sk); | |
3731 | return ret; | |
8913336a | 3732 | } |
69e3c75f JB |
3733 | case PACKET_LOSS: |
3734 | { | |
3735 | unsigned int val; | |
3736 | ||
3737 | if (optlen != sizeof(val)) | |
3738 | return -EINVAL; | |
3739 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) | |
3740 | return -EBUSY; | |
3741 | if (copy_from_user(&val, optval, sizeof(val))) | |
3742 | return -EFAULT; | |
3743 | po->tp_loss = !!val; | |
3744 | return 0; | |
3745 | } | |
8dc41944 HX |
3746 | case PACKET_AUXDATA: |
3747 | { | |
3748 | int val; | |
3749 | ||
3750 | if (optlen < sizeof(val)) | |
3751 | return -EINVAL; | |
3752 | if (copy_from_user(&val, optval, sizeof(val))) | |
3753 | return -EFAULT; | |
3754 | ||
3755 | po->auxdata = !!val; | |
3756 | return 0; | |
3757 | } | |
80feaacb PWJ |
3758 | case PACKET_ORIGDEV: |
3759 | { | |
3760 | int val; | |
3761 | ||
3762 | if (optlen < sizeof(val)) | |
3763 | return -EINVAL; | |
3764 | if (copy_from_user(&val, optval, sizeof(val))) | |
3765 | return -EFAULT; | |
3766 | ||
3767 | po->origdev = !!val; | |
3768 | return 0; | |
3769 | } | |
bfd5f4a3 SS |
3770 | case PACKET_VNET_HDR: |
3771 | { | |
3772 | int val; | |
3773 | ||
3774 | if (sock->type != SOCK_RAW) | |
3775 | return -EINVAL; | |
3776 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) | |
3777 | return -EBUSY; | |
3778 | if (optlen < sizeof(val)) | |
3779 | return -EINVAL; | |
3780 | if (copy_from_user(&val, optval, sizeof(val))) | |
3781 | return -EFAULT; | |
3782 | ||
3783 | po->has_vnet_hdr = !!val; | |
3784 | return 0; | |
3785 | } | |
614f60fa SM |
3786 | case PACKET_TIMESTAMP: |
3787 | { | |
3788 | int val; | |
3789 | ||
3790 | if (optlen != sizeof(val)) | |
3791 | return -EINVAL; | |
3792 | if (copy_from_user(&val, optval, sizeof(val))) | |
3793 | return -EFAULT; | |
3794 | ||
3795 | po->tp_tstamp = val; | |
3796 | return 0; | |
3797 | } | |
dc99f600 DM |
3798 | case PACKET_FANOUT: |
3799 | { | |
3800 | int val; | |
3801 | ||
3802 | if (optlen != sizeof(val)) | |
3803 | return -EINVAL; | |
3804 | if (copy_from_user(&val, optval, sizeof(val))) | |
3805 | return -EFAULT; | |
3806 | ||
3807 | return fanout_add(sk, val & 0xffff, val >> 16); | |
3808 | } | |
47dceb8e WB |
3809 | case PACKET_FANOUT_DATA: |
3810 | { | |
3811 | if (!po->fanout) | |
3812 | return -EINVAL; | |
3813 | ||
3814 | return fanout_set_data(po, optval, optlen); | |
3815 | } | |
5920cd3a PC |
3816 | case PACKET_TX_HAS_OFF: |
3817 | { | |
3818 | unsigned int val; | |
3819 | ||
3820 | if (optlen != sizeof(val)) | |
3821 | return -EINVAL; | |
3822 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) | |
3823 | return -EBUSY; | |
3824 | if (copy_from_user(&val, optval, sizeof(val))) | |
3825 | return -EFAULT; | |
3826 | po->tp_tx_has_off = !!val; | |
3827 | return 0; | |
3828 | } | |
d346a3fa DB |
3829 | case PACKET_QDISC_BYPASS: |
3830 | { | |
3831 | int val; | |
3832 | ||
3833 | if (optlen != sizeof(val)) | |
3834 | return -EINVAL; | |
3835 | if (copy_from_user(&val, optval, sizeof(val))) | |
3836 | return -EFAULT; | |
3837 | ||
3838 | po->xmit = val ? packet_direct_xmit : dev_queue_xmit; | |
3839 | return 0; | |
3840 | } | |
1da177e4 LT |
3841 | default: |
3842 | return -ENOPROTOOPT; | |
3843 | } | |
3844 | } | |
3845 | ||
3846 | static int packet_getsockopt(struct socket *sock, int level, int optname, | |
3847 | char __user *optval, int __user *optlen) | |
3848 | { | |
3849 | int len; | |
c06fff6e | 3850 | int val, lv = sizeof(val); |
1da177e4 LT |
3851 | struct sock *sk = sock->sk; |
3852 | struct packet_sock *po = pkt_sk(sk); | |
c06fff6e | 3853 | void *data = &val; |
ee80fbf3 | 3854 | union tpacket_stats_u st; |
a9b63918 | 3855 | struct tpacket_rollover_stats rstats; |
1da177e4 LT |
3856 | |
3857 | if (level != SOL_PACKET) | |
3858 | return -ENOPROTOOPT; | |
3859 | ||
8ae55f04 KK |
3860 | if (get_user(len, optlen)) |
3861 | return -EFAULT; | |
1da177e4 LT |
3862 | |
3863 | if (len < 0) | |
3864 | return -EINVAL; | |
1ce4f28b | 3865 | |
69e3c75f | 3866 | switch (optname) { |
1da177e4 | 3867 | case PACKET_STATISTICS: |
1da177e4 | 3868 | spin_lock_bh(&sk->sk_receive_queue.lock); |
ee80fbf3 DB |
3869 | memcpy(&st, &po->stats, sizeof(st)); |
3870 | memset(&po->stats, 0, sizeof(po->stats)); | |
3871 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
3872 | ||
f6fb8f10 | 3873 | if (po->tp_version == TPACKET_V3) { |
c06fff6e | 3874 | lv = sizeof(struct tpacket_stats_v3); |
8bcdeaff | 3875 | st.stats3.tp_packets += st.stats3.tp_drops; |
ee80fbf3 | 3876 | data = &st.stats3; |
f6fb8f10 | 3877 | } else { |
c06fff6e | 3878 | lv = sizeof(struct tpacket_stats); |
8bcdeaff | 3879 | st.stats1.tp_packets += st.stats1.tp_drops; |
ee80fbf3 | 3880 | data = &st.stats1; |
f6fb8f10 | 3881 | } |
ee80fbf3 | 3882 | |
8dc41944 HX |
3883 | break; |
3884 | case PACKET_AUXDATA: | |
8dc41944 | 3885 | val = po->auxdata; |
80feaacb PWJ |
3886 | break; |
3887 | case PACKET_ORIGDEV: | |
80feaacb | 3888 | val = po->origdev; |
bfd5f4a3 SS |
3889 | break; |
3890 | case PACKET_VNET_HDR: | |
bfd5f4a3 | 3891 | val = po->has_vnet_hdr; |
1da177e4 | 3892 | break; |
bbd6ef87 | 3893 | case PACKET_VERSION: |
bbd6ef87 | 3894 | val = po->tp_version; |
bbd6ef87 PM |
3895 | break; |
3896 | case PACKET_HDRLEN: | |
3897 | if (len > sizeof(int)) | |
3898 | len = sizeof(int); | |
fd2c83b3 AP |
3899 | if (len < sizeof(int)) |
3900 | return -EINVAL; | |
bbd6ef87 PM |
3901 | if (copy_from_user(&val, optval, len)) |
3902 | return -EFAULT; | |
3903 | switch (val) { | |
3904 | case TPACKET_V1: | |
3905 | val = sizeof(struct tpacket_hdr); | |
3906 | break; | |
3907 | case TPACKET_V2: | |
3908 | val = sizeof(struct tpacket2_hdr); | |
3909 | break; | |
f6fb8f10 | 3910 | case TPACKET_V3: |
3911 | val = sizeof(struct tpacket3_hdr); | |
3912 | break; | |
bbd6ef87 PM |
3913 | default: |
3914 | return -EINVAL; | |
3915 | } | |
bbd6ef87 | 3916 | break; |
8913336a | 3917 | case PACKET_RESERVE: |
8913336a | 3918 | val = po->tp_reserve; |
8913336a | 3919 | break; |
69e3c75f | 3920 | case PACKET_LOSS: |
69e3c75f | 3921 | val = po->tp_loss; |
69e3c75f | 3922 | break; |
614f60fa | 3923 | case PACKET_TIMESTAMP: |
614f60fa | 3924 | val = po->tp_tstamp; |
614f60fa | 3925 | break; |
dc99f600 | 3926 | case PACKET_FANOUT: |
dc99f600 DM |
3927 | val = (po->fanout ? |
3928 | ((u32)po->fanout->id | | |
77f65ebd WB |
3929 | ((u32)po->fanout->type << 16) | |
3930 | ((u32)po->fanout->flags << 24)) : | |
dc99f600 | 3931 | 0); |
dc99f600 | 3932 | break; |
a9b63918 WB |
3933 | case PACKET_ROLLOVER_STATS: |
3934 | if (!po->rollover) | |
3935 | return -EINVAL; | |
3936 | rstats.tp_all = atomic_long_read(&po->rollover->num); | |
3937 | rstats.tp_huge = atomic_long_read(&po->rollover->num_huge); | |
3938 | rstats.tp_failed = atomic_long_read(&po->rollover->num_failed); | |
3939 | data = &rstats; | |
3940 | lv = sizeof(rstats); | |
3941 | break; | |
5920cd3a PC |
3942 | case PACKET_TX_HAS_OFF: |
3943 | val = po->tp_tx_has_off; | |
3944 | break; | |
d346a3fa DB |
3945 | case PACKET_QDISC_BYPASS: |
3946 | val = packet_use_direct_xmit(po); | |
3947 | break; | |
1da177e4 LT |
3948 | default: |
3949 | return -ENOPROTOOPT; | |
3950 | } | |
3951 | ||
c06fff6e ED |
3952 | if (len > lv) |
3953 | len = lv; | |
8ae55f04 KK |
3954 | if (put_user(len, optlen)) |
3955 | return -EFAULT; | |
8dc41944 HX |
3956 | if (copy_to_user(optval, data, len)) |
3957 | return -EFAULT; | |
8ae55f04 | 3958 | return 0; |
1da177e4 LT |
3959 | } |
3960 | ||
3961 | ||
719c44d3 WB |
3962 | #ifdef CONFIG_COMPAT |
3963 | static int compat_packet_setsockopt(struct socket *sock, int level, int optname, | |
3964 | char __user *optval, unsigned int optlen) | |
3965 | { | |
3966 | struct packet_sock *po = pkt_sk(sock->sk); | |
3967 | ||
3968 | if (level != SOL_PACKET) | |
3969 | return -ENOPROTOOPT; | |
3970 | ||
3971 | if (optname == PACKET_FANOUT_DATA && | |
3972 | po->fanout && po->fanout->type == PACKET_FANOUT_CBPF) { | |
3973 | optval = (char __user *)get_compat_bpf_fprog(optval); | |
3974 | if (!optval) | |
3975 | return -EFAULT; | |
3976 | optlen = sizeof(struct sock_fprog); | |
3977 | } | |
3978 | ||
3979 | return packet_setsockopt(sock, level, optname, optval, optlen); | |
3980 | } | |
3981 | #endif | |
3982 | ||
351638e7 JP |
3983 | static int packet_notifier(struct notifier_block *this, |
3984 | unsigned long msg, void *ptr) | |
1da177e4 LT |
3985 | { |
3986 | struct sock *sk; | |
351638e7 | 3987 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); |
c346dca1 | 3988 | struct net *net = dev_net(dev); |
1da177e4 | 3989 | |
808f5114 | 3990 | rcu_read_lock(); |
b67bfe0d | 3991 | sk_for_each_rcu(sk, &net->packet.sklist) { |
1da177e4 LT |
3992 | struct packet_sock *po = pkt_sk(sk); |
3993 | ||
3994 | switch (msg) { | |
3995 | case NETDEV_UNREGISTER: | |
1da177e4 | 3996 | if (po->mclist) |
82f17091 | 3997 | packet_dev_mclist_delete(dev, &po->mclist); |
a2efcfa0 DM |
3998 | /* fallthrough */ |
3999 | ||
1da177e4 LT |
4000 | case NETDEV_DOWN: |
4001 | if (dev->ifindex == po->ifindex) { | |
4002 | spin_lock(&po->bind_lock); | |
4003 | if (po->running) { | |
ce06b03e | 4004 | __unregister_prot_hook(sk, false); |
1da177e4 LT |
4005 | sk->sk_err = ENETDOWN; |
4006 | if (!sock_flag(sk, SOCK_DEAD)) | |
4007 | sk->sk_error_report(sk); | |
4008 | } | |
4009 | if (msg == NETDEV_UNREGISTER) { | |
66e56cd4 | 4010 | packet_cached_dev_reset(po); |
1da177e4 | 4011 | po->ifindex = -1; |
160ff18a BG |
4012 | if (po->prot_hook.dev) |
4013 | dev_put(po->prot_hook.dev); | |
1da177e4 LT |
4014 | po->prot_hook.dev = NULL; |
4015 | } | |
4016 | spin_unlock(&po->bind_lock); | |
4017 | } | |
4018 | break; | |
4019 | case NETDEV_UP: | |
808f5114 | 4020 | if (dev->ifindex == po->ifindex) { |
4021 | spin_lock(&po->bind_lock); | |
ce06b03e DM |
4022 | if (po->num) |
4023 | register_prot_hook(sk); | |
808f5114 | 4024 | spin_unlock(&po->bind_lock); |
1da177e4 | 4025 | } |
1da177e4 LT |
4026 | break; |
4027 | } | |
4028 | } | |
808f5114 | 4029 | rcu_read_unlock(); |
1da177e4 LT |
4030 | return NOTIFY_DONE; |
4031 | } | |
4032 | ||
4033 | ||
4034 | static int packet_ioctl(struct socket *sock, unsigned int cmd, | |
4035 | unsigned long arg) | |
4036 | { | |
4037 | struct sock *sk = sock->sk; | |
4038 | ||
69e3c75f | 4039 | switch (cmd) { |
40d4e3df ED |
4040 | case SIOCOUTQ: |
4041 | { | |
4042 | int amount = sk_wmem_alloc_get(sk); | |
31e6d363 | 4043 | |
40d4e3df ED |
4044 | return put_user(amount, (int __user *)arg); |
4045 | } | |
4046 | case SIOCINQ: | |
4047 | { | |
4048 | struct sk_buff *skb; | |
4049 | int amount = 0; | |
4050 | ||
4051 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
4052 | skb = skb_peek(&sk->sk_receive_queue); | |
4053 | if (skb) | |
4054 | amount = skb->len; | |
4055 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
4056 | return put_user(amount, (int __user *)arg); | |
4057 | } | |
4058 | case SIOCGSTAMP: | |
4059 | return sock_get_timestamp(sk, (struct timeval __user *)arg); | |
4060 | case SIOCGSTAMPNS: | |
4061 | return sock_get_timestampns(sk, (struct timespec __user *)arg); | |
1ce4f28b | 4062 | |
1da177e4 | 4063 | #ifdef CONFIG_INET |
40d4e3df ED |
4064 | case SIOCADDRT: |
4065 | case SIOCDELRT: | |
4066 | case SIOCDARP: | |
4067 | case SIOCGARP: | |
4068 | case SIOCSARP: | |
4069 | case SIOCGIFADDR: | |
4070 | case SIOCSIFADDR: | |
4071 | case SIOCGIFBRDADDR: | |
4072 | case SIOCSIFBRDADDR: | |
4073 | case SIOCGIFNETMASK: | |
4074 | case SIOCSIFNETMASK: | |
4075 | case SIOCGIFDSTADDR: | |
4076 | case SIOCSIFDSTADDR: | |
4077 | case SIOCSIFFLAGS: | |
40d4e3df | 4078 | return inet_dgram_ops.ioctl(sock, cmd, arg); |
1da177e4 LT |
4079 | #endif |
4080 | ||
40d4e3df ED |
4081 | default: |
4082 | return -ENOIOCTLCMD; | |
1da177e4 LT |
4083 | } |
4084 | return 0; | |
4085 | } | |
4086 | ||
40d4e3df | 4087 | static unsigned int packet_poll(struct file *file, struct socket *sock, |
1da177e4 LT |
4088 | poll_table *wait) |
4089 | { | |
4090 | struct sock *sk = sock->sk; | |
4091 | struct packet_sock *po = pkt_sk(sk); | |
4092 | unsigned int mask = datagram_poll(file, sock, wait); | |
4093 | ||
4094 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
69e3c75f | 4095 | if (po->rx_ring.pg_vec) { |
f6fb8f10 | 4096 | if (!packet_previous_rx_frame(po, &po->rx_ring, |
4097 | TP_STATUS_KERNEL)) | |
1da177e4 LT |
4098 | mask |= POLLIN | POLLRDNORM; |
4099 | } | |
2ccdbaa6 | 4100 | if (po->pressure && __packet_rcv_has_room(po, NULL) == ROOM_NORMAL) |
54d7c01d | 4101 | po->pressure = 0; |
1da177e4 | 4102 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
69e3c75f JB |
4103 | spin_lock_bh(&sk->sk_write_queue.lock); |
4104 | if (po->tx_ring.pg_vec) { | |
4105 | if (packet_current_frame(po, &po->tx_ring, TP_STATUS_AVAILABLE)) | |
4106 | mask |= POLLOUT | POLLWRNORM; | |
4107 | } | |
4108 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
1da177e4 LT |
4109 | return mask; |
4110 | } | |
4111 | ||
4112 | ||
4113 | /* Dirty? Well, I still did not learn better way to account | |
4114 | * for user mmaps. | |
4115 | */ | |
4116 | ||
4117 | static void packet_mm_open(struct vm_area_struct *vma) | |
4118 | { | |
4119 | struct file *file = vma->vm_file; | |
40d4e3df | 4120 | struct socket *sock = file->private_data; |
1da177e4 | 4121 | struct sock *sk = sock->sk; |
1ce4f28b | 4122 | |
1da177e4 LT |
4123 | if (sk) |
4124 | atomic_inc(&pkt_sk(sk)->mapped); | |
4125 | } | |
4126 | ||
4127 | static void packet_mm_close(struct vm_area_struct *vma) | |
4128 | { | |
4129 | struct file *file = vma->vm_file; | |
40d4e3df | 4130 | struct socket *sock = file->private_data; |
1da177e4 | 4131 | struct sock *sk = sock->sk; |
1ce4f28b | 4132 | |
1da177e4 LT |
4133 | if (sk) |
4134 | atomic_dec(&pkt_sk(sk)->mapped); | |
4135 | } | |
4136 | ||
f0f37e2f | 4137 | static const struct vm_operations_struct packet_mmap_ops = { |
40d4e3df ED |
4138 | .open = packet_mm_open, |
4139 | .close = packet_mm_close, | |
1da177e4 LT |
4140 | }; |
4141 | ||
0e3125c7 NH |
4142 | static void free_pg_vec(struct pgv *pg_vec, unsigned int order, |
4143 | unsigned int len) | |
1da177e4 LT |
4144 | { |
4145 | int i; | |
4146 | ||
4ebf0ae2 | 4147 | for (i = 0; i < len; i++) { |
0e3125c7 | 4148 | if (likely(pg_vec[i].buffer)) { |
c56b4d90 | 4149 | if (is_vmalloc_addr(pg_vec[i].buffer)) |
0e3125c7 NH |
4150 | vfree(pg_vec[i].buffer); |
4151 | else | |
4152 | free_pages((unsigned long)pg_vec[i].buffer, | |
4153 | order); | |
4154 | pg_vec[i].buffer = NULL; | |
4155 | } | |
1da177e4 LT |
4156 | } |
4157 | kfree(pg_vec); | |
4158 | } | |
4159 | ||
eea49cc9 | 4160 | static char *alloc_one_pg_vec_page(unsigned long order) |
4ebf0ae2 | 4161 | { |
f0d4eb29 | 4162 | char *buffer; |
0e3125c7 NH |
4163 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | |
4164 | __GFP_ZERO | __GFP_NOWARN | __GFP_NORETRY; | |
4165 | ||
4166 | buffer = (char *) __get_free_pages(gfp_flags, order); | |
0e3125c7 NH |
4167 | if (buffer) |
4168 | return buffer; | |
4169 | ||
f0d4eb29 | 4170 | /* __get_free_pages failed, fall back to vmalloc */ |
bbce5a59 | 4171 | buffer = vzalloc((1 << order) * PAGE_SIZE); |
0e3125c7 NH |
4172 | if (buffer) |
4173 | return buffer; | |
4174 | ||
f0d4eb29 | 4175 | /* vmalloc failed, lets dig into swap here */ |
0e3125c7 | 4176 | gfp_flags &= ~__GFP_NORETRY; |
f0d4eb29 | 4177 | buffer = (char *) __get_free_pages(gfp_flags, order); |
0e3125c7 NH |
4178 | if (buffer) |
4179 | return buffer; | |
4180 | ||
f0d4eb29 | 4181 | /* complete and utter failure */ |
0e3125c7 | 4182 | return NULL; |
4ebf0ae2 DM |
4183 | } |
4184 | ||
0e3125c7 | 4185 | static struct pgv *alloc_pg_vec(struct tpacket_req *req, int order) |
4ebf0ae2 DM |
4186 | { |
4187 | unsigned int block_nr = req->tp_block_nr; | |
0e3125c7 | 4188 | struct pgv *pg_vec; |
4ebf0ae2 DM |
4189 | int i; |
4190 | ||
0e3125c7 | 4191 | pg_vec = kcalloc(block_nr, sizeof(struct pgv), GFP_KERNEL); |
4ebf0ae2 DM |
4192 | if (unlikely(!pg_vec)) |
4193 | goto out; | |
4194 | ||
4195 | for (i = 0; i < block_nr; i++) { | |
c56b4d90 | 4196 | pg_vec[i].buffer = alloc_one_pg_vec_page(order); |
0e3125c7 | 4197 | if (unlikely(!pg_vec[i].buffer)) |
4ebf0ae2 DM |
4198 | goto out_free_pgvec; |
4199 | } | |
4200 | ||
4201 | out: | |
4202 | return pg_vec; | |
4203 | ||
4204 | out_free_pgvec: | |
4205 | free_pg_vec(pg_vec, order, block_nr); | |
4206 | pg_vec = NULL; | |
4207 | goto out; | |
4208 | } | |
1da177e4 | 4209 | |
f6fb8f10 | 4210 | static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u, |
69e3c75f | 4211 | int closing, int tx_ring) |
1da177e4 | 4212 | { |
0e3125c7 | 4213 | struct pgv *pg_vec = NULL; |
1da177e4 | 4214 | struct packet_sock *po = pkt_sk(sk); |
0e11c91e | 4215 | int was_running, order = 0; |
69e3c75f JB |
4216 | struct packet_ring_buffer *rb; |
4217 | struct sk_buff_head *rb_queue; | |
0e11c91e | 4218 | __be16 num; |
f6fb8f10 | 4219 | int err = -EINVAL; |
4220 | /* Added to avoid minimal code churn */ | |
4221 | struct tpacket_req *req = &req_u->req; | |
4222 | ||
84ac7260 | 4223 | lock_sock(sk); |
1ce4f28b | 4224 | |
69e3c75f JB |
4225 | rb = tx_ring ? &po->tx_ring : &po->rx_ring; |
4226 | rb_queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
1da177e4 | 4227 | |
69e3c75f JB |
4228 | err = -EBUSY; |
4229 | if (!closing) { | |
4230 | if (atomic_read(&po->mapped)) | |
4231 | goto out; | |
b0138408 | 4232 | if (packet_read_pending(rb)) |
69e3c75f JB |
4233 | goto out; |
4234 | } | |
1da177e4 | 4235 | |
69e3c75f JB |
4236 | if (req->tp_block_nr) { |
4237 | /* Sanity tests and some calculations */ | |
4238 | err = -EBUSY; | |
4239 | if (unlikely(rb->pg_vec)) | |
4240 | goto out; | |
1da177e4 | 4241 | |
bbd6ef87 PM |
4242 | switch (po->tp_version) { |
4243 | case TPACKET_V1: | |
4244 | po->tp_hdrlen = TPACKET_HDRLEN; | |
4245 | break; | |
4246 | case TPACKET_V2: | |
4247 | po->tp_hdrlen = TPACKET2_HDRLEN; | |
4248 | break; | |
f6fb8f10 | 4249 | case TPACKET_V3: |
4250 | po->tp_hdrlen = TPACKET3_HDRLEN; | |
4251 | break; | |
bbd6ef87 PM |
4252 | } |
4253 | ||
69e3c75f | 4254 | err = -EINVAL; |
4ebf0ae2 | 4255 | if (unlikely((int)req->tp_block_size <= 0)) |
69e3c75f | 4256 | goto out; |
90836b67 | 4257 | if (unlikely(!PAGE_ALIGNED(req->tp_block_size))) |
69e3c75f | 4258 | goto out; |
dc808110 | 4259 | if (po->tp_version >= TPACKET_V3 && |
2b6867c2 AK |
4260 | req->tp_block_size <= |
4261 | BLK_PLUS_PRIV((u64)req_u->req3.tp_sizeof_priv)) | |
dc808110 | 4262 | goto out; |
8913336a | 4263 | if (unlikely(req->tp_frame_size < po->tp_hdrlen + |
69e3c75f JB |
4264 | po->tp_reserve)) |
4265 | goto out; | |
4ebf0ae2 | 4266 | if (unlikely(req->tp_frame_size & (TPACKET_ALIGNMENT - 1))) |
69e3c75f | 4267 | goto out; |
1da177e4 | 4268 | |
4194b491 TK |
4269 | rb->frames_per_block = req->tp_block_size / req->tp_frame_size; |
4270 | if (unlikely(rb->frames_per_block == 0)) | |
69e3c75f | 4271 | goto out; |
8f8d28e4 AK |
4272 | if (unlikely(req->tp_block_size > UINT_MAX / req->tp_block_nr)) |
4273 | goto out; | |
69e3c75f JB |
4274 | if (unlikely((rb->frames_per_block * req->tp_block_nr) != |
4275 | req->tp_frame_nr)) | |
4276 | goto out; | |
1da177e4 LT |
4277 | |
4278 | err = -ENOMEM; | |
4ebf0ae2 DM |
4279 | order = get_order(req->tp_block_size); |
4280 | pg_vec = alloc_pg_vec(req, order); | |
4281 | if (unlikely(!pg_vec)) | |
1da177e4 | 4282 | goto out; |
f6fb8f10 | 4283 | switch (po->tp_version) { |
4284 | case TPACKET_V3: | |
7f953ab2 SV |
4285 | /* Block transmit is not supported yet */ |
4286 | if (!tx_ring) { | |
e8e85cc5 | 4287 | init_prb_bdqc(po, rb, pg_vec, req_u); |
7f953ab2 SV |
4288 | } else { |
4289 | struct tpacket_req3 *req3 = &req_u->req3; | |
4290 | ||
4291 | if (req3->tp_retire_blk_tov || | |
4292 | req3->tp_sizeof_priv || | |
4293 | req3->tp_feature_req_word) { | |
4294 | err = -EINVAL; | |
4295 | goto out; | |
4296 | } | |
4297 | } | |
d7cf0c34 | 4298 | break; |
f6fb8f10 | 4299 | default: |
4300 | break; | |
4301 | } | |
69e3c75f JB |
4302 | } |
4303 | /* Done */ | |
4304 | else { | |
4305 | err = -EINVAL; | |
4ebf0ae2 | 4306 | if (unlikely(req->tp_frame_nr)) |
69e3c75f | 4307 | goto out; |
1da177e4 LT |
4308 | } |
4309 | ||
1da177e4 LT |
4310 | |
4311 | /* Detach socket from network */ | |
4312 | spin_lock(&po->bind_lock); | |
4313 | was_running = po->running; | |
4314 | num = po->num; | |
4315 | if (was_running) { | |
1da177e4 | 4316 | po->num = 0; |
ce06b03e | 4317 | __unregister_prot_hook(sk, false); |
1da177e4 LT |
4318 | } |
4319 | spin_unlock(&po->bind_lock); | |
1ce4f28b | 4320 | |
1da177e4 LT |
4321 | synchronize_net(); |
4322 | ||
4323 | err = -EBUSY; | |
905db440 | 4324 | mutex_lock(&po->pg_vec_lock); |
1da177e4 LT |
4325 | if (closing || atomic_read(&po->mapped) == 0) { |
4326 | err = 0; | |
69e3c75f | 4327 | spin_lock_bh(&rb_queue->lock); |
c053fd96 | 4328 | swap(rb->pg_vec, pg_vec); |
69e3c75f JB |
4329 | rb->frame_max = (req->tp_frame_nr - 1); |
4330 | rb->head = 0; | |
4331 | rb->frame_size = req->tp_frame_size; | |
4332 | spin_unlock_bh(&rb_queue->lock); | |
4333 | ||
c053fd96 CG |
4334 | swap(rb->pg_vec_order, order); |
4335 | swap(rb->pg_vec_len, req->tp_block_nr); | |
69e3c75f JB |
4336 | |
4337 | rb->pg_vec_pages = req->tp_block_size/PAGE_SIZE; | |
4338 | po->prot_hook.func = (po->rx_ring.pg_vec) ? | |
4339 | tpacket_rcv : packet_rcv; | |
4340 | skb_queue_purge(rb_queue); | |
1da177e4 | 4341 | if (atomic_read(&po->mapped)) |
40d4e3df ED |
4342 | pr_err("packet_mmap: vma is busy: %d\n", |
4343 | atomic_read(&po->mapped)); | |
1da177e4 | 4344 | } |
905db440 | 4345 | mutex_unlock(&po->pg_vec_lock); |
1da177e4 LT |
4346 | |
4347 | spin_lock(&po->bind_lock); | |
ce06b03e | 4348 | if (was_running) { |
1da177e4 | 4349 | po->num = num; |
ce06b03e | 4350 | register_prot_hook(sk); |
1da177e4 LT |
4351 | } |
4352 | spin_unlock(&po->bind_lock); | |
c800aaf8 | 4353 | if (pg_vec && (po->tp_version > TPACKET_V2)) { |
f6fb8f10 | 4354 | /* Because we don't support block-based V3 on tx-ring */ |
4355 | if (!tx_ring) | |
73d0fcf2 | 4356 | prb_shutdown_retire_blk_timer(po, rb_queue); |
f6fb8f10 | 4357 | } |
1da177e4 | 4358 | |
1da177e4 LT |
4359 | if (pg_vec) |
4360 | free_pg_vec(pg_vec, order, req->tp_block_nr); | |
4361 | out: | |
84ac7260 | 4362 | release_sock(sk); |
1da177e4 LT |
4363 | return err; |
4364 | } | |
4365 | ||
69e3c75f JB |
4366 | static int packet_mmap(struct file *file, struct socket *sock, |
4367 | struct vm_area_struct *vma) | |
1da177e4 LT |
4368 | { |
4369 | struct sock *sk = sock->sk; | |
4370 | struct packet_sock *po = pkt_sk(sk); | |
69e3c75f JB |
4371 | unsigned long size, expected_size; |
4372 | struct packet_ring_buffer *rb; | |
1da177e4 LT |
4373 | unsigned long start; |
4374 | int err = -EINVAL; | |
4375 | int i; | |
4376 | ||
4377 | if (vma->vm_pgoff) | |
4378 | return -EINVAL; | |
4379 | ||
905db440 | 4380 | mutex_lock(&po->pg_vec_lock); |
69e3c75f JB |
4381 | |
4382 | expected_size = 0; | |
4383 | for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) { | |
4384 | if (rb->pg_vec) { | |
4385 | expected_size += rb->pg_vec_len | |
4386 | * rb->pg_vec_pages | |
4387 | * PAGE_SIZE; | |
4388 | } | |
4389 | } | |
4390 | ||
4391 | if (expected_size == 0) | |
1da177e4 | 4392 | goto out; |
69e3c75f JB |
4393 | |
4394 | size = vma->vm_end - vma->vm_start; | |
4395 | if (size != expected_size) | |
1da177e4 LT |
4396 | goto out; |
4397 | ||
1da177e4 | 4398 | start = vma->vm_start; |
69e3c75f JB |
4399 | for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) { |
4400 | if (rb->pg_vec == NULL) | |
4401 | continue; | |
4402 | ||
4403 | for (i = 0; i < rb->pg_vec_len; i++) { | |
0e3125c7 NH |
4404 | struct page *page; |
4405 | void *kaddr = rb->pg_vec[i].buffer; | |
69e3c75f JB |
4406 | int pg_num; |
4407 | ||
c56b4d90 CG |
4408 | for (pg_num = 0; pg_num < rb->pg_vec_pages; pg_num++) { |
4409 | page = pgv_to_page(kaddr); | |
69e3c75f JB |
4410 | err = vm_insert_page(vma, start, page); |
4411 | if (unlikely(err)) | |
4412 | goto out; | |
4413 | start += PAGE_SIZE; | |
0e3125c7 | 4414 | kaddr += PAGE_SIZE; |
69e3c75f | 4415 | } |
4ebf0ae2 | 4416 | } |
1da177e4 | 4417 | } |
69e3c75f | 4418 | |
4ebf0ae2 | 4419 | atomic_inc(&po->mapped); |
1da177e4 LT |
4420 | vma->vm_ops = &packet_mmap_ops; |
4421 | err = 0; | |
4422 | ||
4423 | out: | |
905db440 | 4424 | mutex_unlock(&po->pg_vec_lock); |
1da177e4 LT |
4425 | return err; |
4426 | } | |
1da177e4 | 4427 | |
90ddc4f0 | 4428 | static const struct proto_ops packet_ops_spkt = { |
1da177e4 LT |
4429 | .family = PF_PACKET, |
4430 | .owner = THIS_MODULE, | |
4431 | .release = packet_release, | |
4432 | .bind = packet_bind_spkt, | |
4433 | .connect = sock_no_connect, | |
4434 | .socketpair = sock_no_socketpair, | |
4435 | .accept = sock_no_accept, | |
4436 | .getname = packet_getname_spkt, | |
4437 | .poll = datagram_poll, | |
4438 | .ioctl = packet_ioctl, | |
4439 | .listen = sock_no_listen, | |
4440 | .shutdown = sock_no_shutdown, | |
4441 | .setsockopt = sock_no_setsockopt, | |
4442 | .getsockopt = sock_no_getsockopt, | |
4443 | .sendmsg = packet_sendmsg_spkt, | |
4444 | .recvmsg = packet_recvmsg, | |
4445 | .mmap = sock_no_mmap, | |
4446 | .sendpage = sock_no_sendpage, | |
4447 | }; | |
1da177e4 | 4448 | |
90ddc4f0 | 4449 | static const struct proto_ops packet_ops = { |
1da177e4 LT |
4450 | .family = PF_PACKET, |
4451 | .owner = THIS_MODULE, | |
4452 | .release = packet_release, | |
4453 | .bind = packet_bind, | |
4454 | .connect = sock_no_connect, | |
4455 | .socketpair = sock_no_socketpair, | |
4456 | .accept = sock_no_accept, | |
1ce4f28b | 4457 | .getname = packet_getname, |
1da177e4 LT |
4458 | .poll = packet_poll, |
4459 | .ioctl = packet_ioctl, | |
4460 | .listen = sock_no_listen, | |
4461 | .shutdown = sock_no_shutdown, | |
4462 | .setsockopt = packet_setsockopt, | |
4463 | .getsockopt = packet_getsockopt, | |
719c44d3 WB |
4464 | #ifdef CONFIG_COMPAT |
4465 | .compat_setsockopt = compat_packet_setsockopt, | |
4466 | #endif | |
1da177e4 LT |
4467 | .sendmsg = packet_sendmsg, |
4468 | .recvmsg = packet_recvmsg, | |
4469 | .mmap = packet_mmap, | |
4470 | .sendpage = sock_no_sendpage, | |
4471 | }; | |
4472 | ||
ec1b4cf7 | 4473 | static const struct net_proto_family packet_family_ops = { |
1da177e4 LT |
4474 | .family = PF_PACKET, |
4475 | .create = packet_create, | |
4476 | .owner = THIS_MODULE, | |
4477 | }; | |
4478 | ||
4479 | static struct notifier_block packet_netdev_notifier = { | |
40d4e3df | 4480 | .notifier_call = packet_notifier, |
1da177e4 LT |
4481 | }; |
4482 | ||
4483 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
4484 | |
4485 | static void *packet_seq_start(struct seq_file *seq, loff_t *pos) | |
808f5114 | 4486 | __acquires(RCU) |
1da177e4 | 4487 | { |
e372c414 | 4488 | struct net *net = seq_file_net(seq); |
808f5114 | 4489 | |
4490 | rcu_read_lock(); | |
4491 | return seq_hlist_start_head_rcu(&net->packet.sklist, *pos); | |
1da177e4 LT |
4492 | } |
4493 | ||
4494 | static void *packet_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
4495 | { | |
1bf40954 | 4496 | struct net *net = seq_file_net(seq); |
808f5114 | 4497 | return seq_hlist_next_rcu(v, &net->packet.sklist, pos); |
1da177e4 LT |
4498 | } |
4499 | ||
4500 | static void packet_seq_stop(struct seq_file *seq, void *v) | |
808f5114 | 4501 | __releases(RCU) |
1da177e4 | 4502 | { |
808f5114 | 4503 | rcu_read_unlock(); |
1da177e4 LT |
4504 | } |
4505 | ||
1ce4f28b | 4506 | static int packet_seq_show(struct seq_file *seq, void *v) |
1da177e4 LT |
4507 | { |
4508 | if (v == SEQ_START_TOKEN) | |
4509 | seq_puts(seq, "sk RefCnt Type Proto Iface R Rmem User Inode\n"); | |
4510 | else { | |
b7ceabd9 | 4511 | struct sock *s = sk_entry(v); |
1da177e4 LT |
4512 | const struct packet_sock *po = pkt_sk(s); |
4513 | ||
4514 | seq_printf(seq, | |
71338aa7 | 4515 | "%pK %-6d %-4d %04x %-5d %1d %-6u %-6u %-6lu\n", |
1da177e4 | 4516 | s, |
41c6d650 | 4517 | refcount_read(&s->sk_refcnt), |
1da177e4 LT |
4518 | s->sk_type, |
4519 | ntohs(po->num), | |
4520 | po->ifindex, | |
4521 | po->running, | |
4522 | atomic_read(&s->sk_rmem_alloc), | |
a7cb5a49 | 4523 | from_kuid_munged(seq_user_ns(seq), sock_i_uid(s)), |
40d4e3df | 4524 | sock_i_ino(s)); |
1da177e4 LT |
4525 | } |
4526 | ||
4527 | return 0; | |
4528 | } | |
4529 | ||
56b3d975 | 4530 | static const struct seq_operations packet_seq_ops = { |
1da177e4 LT |
4531 | .start = packet_seq_start, |
4532 | .next = packet_seq_next, | |
4533 | .stop = packet_seq_stop, | |
4534 | .show = packet_seq_show, | |
4535 | }; | |
4536 | ||
4537 | static int packet_seq_open(struct inode *inode, struct file *file) | |
4538 | { | |
e372c414 DL |
4539 | return seq_open_net(inode, file, &packet_seq_ops, |
4540 | sizeof(struct seq_net_private)); | |
1da177e4 LT |
4541 | } |
4542 | ||
da7071d7 | 4543 | static const struct file_operations packet_seq_fops = { |
1da177e4 LT |
4544 | .owner = THIS_MODULE, |
4545 | .open = packet_seq_open, | |
4546 | .read = seq_read, | |
4547 | .llseek = seq_lseek, | |
e372c414 | 4548 | .release = seq_release_net, |
1da177e4 LT |
4549 | }; |
4550 | ||
4551 | #endif | |
4552 | ||
2c8c1e72 | 4553 | static int __net_init packet_net_init(struct net *net) |
d12d01d6 | 4554 | { |
0fa7fa98 | 4555 | mutex_init(&net->packet.sklist_lock); |
2aaef4e4 | 4556 | INIT_HLIST_HEAD(&net->packet.sklist); |
d12d01d6 | 4557 | |
d4beaa66 | 4558 | if (!proc_create("packet", 0, net->proc_net, &packet_seq_fops)) |
d12d01d6 DL |
4559 | return -ENOMEM; |
4560 | ||
4561 | return 0; | |
4562 | } | |
4563 | ||
2c8c1e72 | 4564 | static void __net_exit packet_net_exit(struct net *net) |
d12d01d6 | 4565 | { |
ece31ffd | 4566 | remove_proc_entry("packet", net->proc_net); |
d12d01d6 DL |
4567 | } |
4568 | ||
4569 | static struct pernet_operations packet_net_ops = { | |
4570 | .init = packet_net_init, | |
4571 | .exit = packet_net_exit, | |
4572 | }; | |
4573 | ||
4574 | ||
1da177e4 LT |
4575 | static void __exit packet_exit(void) |
4576 | { | |
1da177e4 | 4577 | unregister_netdevice_notifier(&packet_netdev_notifier); |
d12d01d6 | 4578 | unregister_pernet_subsys(&packet_net_ops); |
1da177e4 LT |
4579 | sock_unregister(PF_PACKET); |
4580 | proto_unregister(&packet_proto); | |
4581 | } | |
4582 | ||
4583 | static int __init packet_init(void) | |
4584 | { | |
4585 | int rc = proto_register(&packet_proto, 0); | |
4586 | ||
4587 | if (rc != 0) | |
4588 | goto out; | |
4589 | ||
4590 | sock_register(&packet_family_ops); | |
d12d01d6 | 4591 | register_pernet_subsys(&packet_net_ops); |
1da177e4 | 4592 | register_netdevice_notifier(&packet_netdev_notifier); |
1da177e4 LT |
4593 | out: |
4594 | return rc; | |
4595 | } | |
4596 | ||
4597 | module_init(packet_init); | |
4598 | module_exit(packet_exit); | |
4599 | MODULE_LICENSE("GPL"); | |
4600 | MODULE_ALIAS_NETPROTO(PF_PACKET); |