]>
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 BLOCK_STATUS(x) ((x)->hdr.bh1.block_status) |
181 | #define BLOCK_NUM_PKTS(x) ((x)->hdr.bh1.num_pkts) | |
182 | #define BLOCK_O2FP(x) ((x)->hdr.bh1.offset_to_first_pkt) | |
183 | #define BLOCK_LEN(x) ((x)->hdr.bh1.blk_len) | |
184 | #define BLOCK_SNUM(x) ((x)->hdr.bh1.seq_num) | |
185 | #define BLOCK_O2PRIV(x) ((x)->offset_to_priv) | |
186 | #define BLOCK_PRIV(x) ((void *)((char *)(x) + BLOCK_O2PRIV(x))) | |
187 | ||
69e3c75f | 188 | struct packet_sock; |
77f65ebd WB |
189 | static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, |
190 | struct packet_type *pt, struct net_device *orig_dev); | |
1da177e4 | 191 | |
f6fb8f10 | 192 | static void *packet_previous_frame(struct packet_sock *po, |
193 | struct packet_ring_buffer *rb, | |
194 | int status); | |
195 | static void packet_increment_head(struct packet_ring_buffer *buff); | |
878cd3ba | 196 | static int prb_curr_blk_in_use(struct tpacket_block_desc *); |
bc59ba39 | 197 | static void *prb_dispatch_next_block(struct tpacket_kbdq_core *, |
f6fb8f10 | 198 | struct packet_sock *); |
bc59ba39 | 199 | static void prb_retire_current_block(struct tpacket_kbdq_core *, |
f6fb8f10 | 200 | struct packet_sock *, unsigned int status); |
bc59ba39 | 201 | static int prb_queue_frozen(struct tpacket_kbdq_core *); |
202 | static void prb_open_block(struct tpacket_kbdq_core *, | |
203 | struct tpacket_block_desc *); | |
17bfd8c8 | 204 | static void prb_retire_rx_blk_timer_expired(struct timer_list *); |
bc59ba39 | 205 | static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *); |
bc59ba39 | 206 | static void prb_fill_rxhash(struct tpacket_kbdq_core *, struct tpacket3_hdr *); |
207 | static void prb_clear_rxhash(struct tpacket_kbdq_core *, | |
208 | struct tpacket3_hdr *); | |
209 | static void prb_fill_vlan_info(struct tpacket_kbdq_core *, | |
210 | struct tpacket3_hdr *); | |
1da177e4 | 211 | static void packet_flush_mclist(struct sock *sk); |
865b03f2 | 212 | static u16 packet_pick_tx_queue(struct sk_buff *skb); |
1da177e4 | 213 | |
ffbc6111 | 214 | struct packet_skb_cb { |
ffbc6111 HX |
215 | union { |
216 | struct sockaddr_pkt pkt; | |
2472d761 EB |
217 | union { |
218 | /* Trick: alias skb original length with | |
219 | * ll.sll_family and ll.protocol in order | |
220 | * to save room. | |
221 | */ | |
222 | unsigned int origlen; | |
223 | struct sockaddr_ll ll; | |
224 | }; | |
ffbc6111 HX |
225 | } sa; |
226 | }; | |
227 | ||
d3869efe DW |
228 | #define vio_le() virtio_legacy_is_little_endian() |
229 | ||
ffbc6111 | 230 | #define PACKET_SKB_CB(__skb) ((struct packet_skb_cb *)((__skb)->cb)) |
8dc41944 | 231 | |
bc59ba39 | 232 | #define GET_PBDQC_FROM_RB(x) ((struct tpacket_kbdq_core *)(&(x)->prb_bdqc)) |
f6fb8f10 | 233 | #define GET_PBLOCK_DESC(x, bid) \ |
bc59ba39 | 234 | ((struct tpacket_block_desc *)((x)->pkbdq[(bid)].buffer)) |
f6fb8f10 | 235 | #define GET_CURR_PBLOCK_DESC_FROM_CORE(x) \ |
bc59ba39 | 236 | ((struct tpacket_block_desc *)((x)->pkbdq[(x)->kactive_blk_num].buffer)) |
f6fb8f10 | 237 | #define GET_NEXT_PRB_BLK_NUM(x) \ |
238 | (((x)->kactive_blk_num < ((x)->knum_blocks-1)) ? \ | |
239 | ((x)->kactive_blk_num+1) : 0) | |
240 | ||
dc99f600 DM |
241 | static void __fanout_unlink(struct sock *sk, struct packet_sock *po); |
242 | static void __fanout_link(struct sock *sk, struct packet_sock *po); | |
243 | ||
d346a3fa DB |
244 | static int packet_direct_xmit(struct sk_buff *skb) |
245 | { | |
865b03f2 | 246 | return dev_direct_xmit(skb, packet_pick_tx_queue(skb)); |
d346a3fa DB |
247 | } |
248 | ||
66e56cd4 DB |
249 | static struct net_device *packet_cached_dev_get(struct packet_sock *po) |
250 | { | |
251 | struct net_device *dev; | |
252 | ||
253 | rcu_read_lock(); | |
254 | dev = rcu_dereference(po->cached_dev); | |
255 | if (likely(dev)) | |
256 | dev_hold(dev); | |
257 | rcu_read_unlock(); | |
258 | ||
259 | return dev; | |
260 | } | |
261 | ||
262 | static void packet_cached_dev_assign(struct packet_sock *po, | |
263 | struct net_device *dev) | |
264 | { | |
265 | rcu_assign_pointer(po->cached_dev, dev); | |
266 | } | |
267 | ||
268 | static void packet_cached_dev_reset(struct packet_sock *po) | |
269 | { | |
270 | RCU_INIT_POINTER(po->cached_dev, NULL); | |
271 | } | |
272 | ||
d346a3fa DB |
273 | static bool packet_use_direct_xmit(const struct packet_sock *po) |
274 | { | |
275 | return po->xmit == packet_direct_xmit; | |
276 | } | |
277 | ||
8ec56fc3 AD |
278 | static u16 __packet_pick_tx_queue(struct net_device *dev, struct sk_buff *skb, |
279 | struct net_device *sb_dev) | |
d346a3fa | 280 | { |
8ec56fc3 | 281 | return dev_pick_tx_cpu_id(dev, skb, sb_dev, NULL); |
d346a3fa DB |
282 | } |
283 | ||
865b03f2 | 284 | static u16 packet_pick_tx_queue(struct sk_buff *skb) |
0fd5d57b | 285 | { |
865b03f2 | 286 | struct net_device *dev = skb->dev; |
0fd5d57b DB |
287 | const struct net_device_ops *ops = dev->netdev_ops; |
288 | u16 queue_index; | |
289 | ||
290 | if (ops->ndo_select_queue) { | |
291 | queue_index = ops->ndo_select_queue(dev, skb, NULL, | |
292 | __packet_pick_tx_queue); | |
293 | queue_index = netdev_cap_txqueue(dev, queue_index); | |
294 | } else { | |
8ec56fc3 | 295 | queue_index = __packet_pick_tx_queue(dev, skb, NULL); |
0fd5d57b DB |
296 | } |
297 | ||
865b03f2 | 298 | return queue_index; |
0fd5d57b DB |
299 | } |
300 | ||
a6361f0c | 301 | /* __register_prot_hook must be invoked through register_prot_hook |
ce06b03e DM |
302 | * or from a context in which asynchronous accesses to the packet |
303 | * socket is not possible (packet_create()). | |
304 | */ | |
a6361f0c | 305 | static void __register_prot_hook(struct sock *sk) |
ce06b03e DM |
306 | { |
307 | struct packet_sock *po = pkt_sk(sk); | |
e40526cb | 308 | |
ce06b03e | 309 | if (!po->running) { |
66e56cd4 | 310 | if (po->fanout) |
dc99f600 | 311 | __fanout_link(sk, po); |
66e56cd4 | 312 | else |
dc99f600 | 313 | dev_add_pack(&po->prot_hook); |
e40526cb | 314 | |
ce06b03e DM |
315 | sock_hold(sk); |
316 | po->running = 1; | |
317 | } | |
318 | } | |
319 | ||
a6361f0c WB |
320 | static void register_prot_hook(struct sock *sk) |
321 | { | |
322 | lockdep_assert_held_once(&pkt_sk(sk)->bind_lock); | |
323 | __register_prot_hook(sk); | |
324 | } | |
325 | ||
326 | /* If the sync parameter is true, we will temporarily drop | |
ce06b03e DM |
327 | * the po->bind_lock and do a synchronize_net to make sure no |
328 | * asynchronous packet processing paths still refer to the elements | |
329 | * of po->prot_hook. If the sync parameter is false, it is the | |
330 | * callers responsibility to take care of this. | |
331 | */ | |
332 | static void __unregister_prot_hook(struct sock *sk, bool sync) | |
333 | { | |
334 | struct packet_sock *po = pkt_sk(sk); | |
335 | ||
a6361f0c WB |
336 | lockdep_assert_held_once(&po->bind_lock); |
337 | ||
ce06b03e | 338 | po->running = 0; |
66e56cd4 DB |
339 | |
340 | if (po->fanout) | |
dc99f600 | 341 | __fanout_unlink(sk, po); |
66e56cd4 | 342 | else |
dc99f600 | 343 | __dev_remove_pack(&po->prot_hook); |
e40526cb | 344 | |
ce06b03e DM |
345 | __sock_put(sk); |
346 | ||
347 | if (sync) { | |
348 | spin_unlock(&po->bind_lock); | |
349 | synchronize_net(); | |
350 | spin_lock(&po->bind_lock); | |
351 | } | |
352 | } | |
353 | ||
354 | static void unregister_prot_hook(struct sock *sk, bool sync) | |
355 | { | |
356 | struct packet_sock *po = pkt_sk(sk); | |
357 | ||
358 | if (po->running) | |
359 | __unregister_prot_hook(sk, sync); | |
360 | } | |
361 | ||
6e58040b | 362 | static inline struct page * __pure pgv_to_page(void *addr) |
0af55bb5 CG |
363 | { |
364 | if (is_vmalloc_addr(addr)) | |
365 | return vmalloc_to_page(addr); | |
366 | return virt_to_page(addr); | |
367 | } | |
368 | ||
69e3c75f | 369 | static void __packet_set_status(struct packet_sock *po, void *frame, int status) |
1da177e4 | 370 | { |
184f489e | 371 | union tpacket_uhdr h; |
1da177e4 | 372 | |
69e3c75f | 373 | h.raw = frame; |
bbd6ef87 PM |
374 | switch (po->tp_version) { |
375 | case TPACKET_V1: | |
69e3c75f | 376 | h.h1->tp_status = status; |
0af55bb5 | 377 | flush_dcache_page(pgv_to_page(&h.h1->tp_status)); |
bbd6ef87 PM |
378 | break; |
379 | case TPACKET_V2: | |
69e3c75f | 380 | h.h2->tp_status = status; |
0af55bb5 | 381 | flush_dcache_page(pgv_to_page(&h.h2->tp_status)); |
bbd6ef87 | 382 | break; |
f6fb8f10 | 383 | case TPACKET_V3: |
7f953ab2 SV |
384 | h.h3->tp_status = status; |
385 | flush_dcache_page(pgv_to_page(&h.h3->tp_status)); | |
386 | break; | |
69e3c75f | 387 | default: |
f6fb8f10 | 388 | WARN(1, "TPACKET version not supported.\n"); |
69e3c75f | 389 | BUG(); |
bbd6ef87 | 390 | } |
69e3c75f JB |
391 | |
392 | smp_wmb(); | |
bbd6ef87 PM |
393 | } |
394 | ||
69e3c75f | 395 | static int __packet_get_status(struct packet_sock *po, void *frame) |
bbd6ef87 | 396 | { |
184f489e | 397 | union tpacket_uhdr h; |
bbd6ef87 | 398 | |
69e3c75f JB |
399 | smp_rmb(); |
400 | ||
bbd6ef87 PM |
401 | h.raw = frame; |
402 | switch (po->tp_version) { | |
403 | case TPACKET_V1: | |
0af55bb5 | 404 | flush_dcache_page(pgv_to_page(&h.h1->tp_status)); |
69e3c75f | 405 | return h.h1->tp_status; |
bbd6ef87 | 406 | case TPACKET_V2: |
0af55bb5 | 407 | flush_dcache_page(pgv_to_page(&h.h2->tp_status)); |
69e3c75f | 408 | return h.h2->tp_status; |
f6fb8f10 | 409 | case TPACKET_V3: |
7f953ab2 SV |
410 | flush_dcache_page(pgv_to_page(&h.h3->tp_status)); |
411 | return h.h3->tp_status; | |
69e3c75f | 412 | default: |
f6fb8f10 | 413 | WARN(1, "TPACKET version not supported.\n"); |
69e3c75f JB |
414 | BUG(); |
415 | return 0; | |
bbd6ef87 | 416 | } |
1da177e4 | 417 | } |
69e3c75f | 418 | |
b9c32fb2 DB |
419 | static __u32 tpacket_get_timestamp(struct sk_buff *skb, struct timespec *ts, |
420 | unsigned int flags) | |
7a51384c DB |
421 | { |
422 | struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb); | |
423 | ||
68a360e8 WB |
424 | if (shhwtstamps && |
425 | (flags & SOF_TIMESTAMPING_RAW_HARDWARE) && | |
426 | ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts)) | |
427 | return TP_STATUS_TS_RAW_HARDWARE; | |
7a51384c DB |
428 | |
429 | if (ktime_to_timespec_cond(skb->tstamp, ts)) | |
b9c32fb2 | 430 | return TP_STATUS_TS_SOFTWARE; |
7a51384c | 431 | |
b9c32fb2 | 432 | return 0; |
7a51384c DB |
433 | } |
434 | ||
b9c32fb2 DB |
435 | static __u32 __packet_set_timestamp(struct packet_sock *po, void *frame, |
436 | struct sk_buff *skb) | |
2e31396f WB |
437 | { |
438 | union tpacket_uhdr h; | |
439 | struct timespec ts; | |
b9c32fb2 | 440 | __u32 ts_status; |
2e31396f | 441 | |
b9c32fb2 DB |
442 | if (!(ts_status = tpacket_get_timestamp(skb, &ts, po->tp_tstamp))) |
443 | return 0; | |
2e31396f WB |
444 | |
445 | h.raw = frame; | |
446 | switch (po->tp_version) { | |
447 | case TPACKET_V1: | |
448 | h.h1->tp_sec = ts.tv_sec; | |
449 | h.h1->tp_usec = ts.tv_nsec / NSEC_PER_USEC; | |
450 | break; | |
451 | case TPACKET_V2: | |
452 | h.h2->tp_sec = ts.tv_sec; | |
453 | h.h2->tp_nsec = ts.tv_nsec; | |
454 | break; | |
455 | case TPACKET_V3: | |
57ea884b DB |
456 | h.h3->tp_sec = ts.tv_sec; |
457 | h.h3->tp_nsec = ts.tv_nsec; | |
458 | break; | |
2e31396f WB |
459 | default: |
460 | WARN(1, "TPACKET version not supported.\n"); | |
461 | BUG(); | |
462 | } | |
463 | ||
464 | /* one flush is safe, as both fields always lie on the same cacheline */ | |
465 | flush_dcache_page(pgv_to_page(&h.h1->tp_sec)); | |
466 | smp_wmb(); | |
b9c32fb2 DB |
467 | |
468 | return ts_status; | |
2e31396f WB |
469 | } |
470 | ||
69e3c75f JB |
471 | static void *packet_lookup_frame(struct packet_sock *po, |
472 | struct packet_ring_buffer *rb, | |
473 | unsigned int position, | |
474 | int status) | |
475 | { | |
476 | unsigned int pg_vec_pos, frame_offset; | |
184f489e | 477 | union tpacket_uhdr h; |
69e3c75f JB |
478 | |
479 | pg_vec_pos = position / rb->frames_per_block; | |
480 | frame_offset = position % rb->frames_per_block; | |
481 | ||
0e3125c7 NH |
482 | h.raw = rb->pg_vec[pg_vec_pos].buffer + |
483 | (frame_offset * rb->frame_size); | |
69e3c75f JB |
484 | |
485 | if (status != __packet_get_status(po, h.raw)) | |
486 | return NULL; | |
487 | ||
488 | return h.raw; | |
489 | } | |
490 | ||
eea49cc9 | 491 | static void *packet_current_frame(struct packet_sock *po, |
69e3c75f JB |
492 | struct packet_ring_buffer *rb, |
493 | int status) | |
494 | { | |
495 | return packet_lookup_frame(po, rb, rb->head, status); | |
496 | } | |
497 | ||
bc59ba39 | 498 | static void prb_del_retire_blk_timer(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 499 | { |
500 | del_timer_sync(&pkc->retire_blk_timer); | |
501 | } | |
502 | ||
503 | static void prb_shutdown_retire_blk_timer(struct packet_sock *po, | |
f6fb8f10 | 504 | struct sk_buff_head *rb_queue) |
505 | { | |
bc59ba39 | 506 | struct tpacket_kbdq_core *pkc; |
f6fb8f10 | 507 | |
73d0fcf2 | 508 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 509 | |
ec6f809f | 510 | spin_lock_bh(&rb_queue->lock); |
f6fb8f10 | 511 | pkc->delete_blk_timer = 1; |
ec6f809f | 512 | spin_unlock_bh(&rb_queue->lock); |
f6fb8f10 | 513 | |
514 | prb_del_retire_blk_timer(pkc); | |
515 | } | |
516 | ||
e8e85cc5 | 517 | static void prb_setup_retire_blk_timer(struct packet_sock *po) |
f6fb8f10 | 518 | { |
bc59ba39 | 519 | struct tpacket_kbdq_core *pkc; |
f6fb8f10 | 520 | |
e8e85cc5 | 521 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
17bfd8c8 KC |
522 | timer_setup(&pkc->retire_blk_timer, prb_retire_rx_blk_timer_expired, |
523 | 0); | |
524 | pkc->retire_blk_timer.expires = jiffies; | |
f6fb8f10 | 525 | } |
526 | ||
527 | static int prb_calc_retire_blk_tmo(struct packet_sock *po, | |
528 | int blk_size_in_bytes) | |
529 | { | |
530 | struct net_device *dev; | |
531 | unsigned int mbits = 0, msec = 0, div = 0, tmo = 0; | |
7cad1bac | 532 | struct ethtool_link_ksettings ecmd; |
4bc71cb9 | 533 | int err; |
f6fb8f10 | 534 | |
4bc71cb9 JP |
535 | rtnl_lock(); |
536 | dev = __dev_get_by_index(sock_net(&po->sk), po->ifindex); | |
537 | if (unlikely(!dev)) { | |
538 | rtnl_unlock(); | |
f6fb8f10 | 539 | return DEFAULT_PRB_RETIRE_TOV; |
4bc71cb9 | 540 | } |
7cad1bac | 541 | err = __ethtool_get_link_ksettings(dev, &ecmd); |
4bc71cb9 JP |
542 | rtnl_unlock(); |
543 | if (!err) { | |
4bc71cb9 JP |
544 | /* |
545 | * If the link speed is so slow you don't really | |
546 | * need to worry about perf anyways | |
547 | */ | |
7cad1bac DD |
548 | if (ecmd.base.speed < SPEED_1000 || |
549 | ecmd.base.speed == SPEED_UNKNOWN) { | |
4bc71cb9 | 550 | return DEFAULT_PRB_RETIRE_TOV; |
e440cf2c | 551 | } else { |
552 | msec = 1; | |
7cad1bac | 553 | div = ecmd.base.speed / 1000; |
f6fb8f10 | 554 | } |
555 | } | |
556 | ||
557 | mbits = (blk_size_in_bytes * 8) / (1024 * 1024); | |
558 | ||
559 | if (div) | |
560 | mbits /= div; | |
561 | ||
562 | tmo = mbits * msec; | |
563 | ||
564 | if (div) | |
565 | return tmo+1; | |
566 | return tmo; | |
567 | } | |
568 | ||
bc59ba39 | 569 | static void prb_init_ft_ops(struct tpacket_kbdq_core *p1, |
f6fb8f10 | 570 | union tpacket_req_u *req_u) |
571 | { | |
572 | p1->feature_req_word = req_u->req3.tp_feature_req_word; | |
573 | } | |
574 | ||
575 | static void init_prb_bdqc(struct packet_sock *po, | |
576 | struct packet_ring_buffer *rb, | |
577 | struct pgv *pg_vec, | |
e8e85cc5 | 578 | union tpacket_req_u *req_u) |
f6fb8f10 | 579 | { |
22781a5b | 580 | struct tpacket_kbdq_core *p1 = GET_PBDQC_FROM_RB(rb); |
bc59ba39 | 581 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 582 | |
583 | memset(p1, 0x0, sizeof(*p1)); | |
584 | ||
585 | p1->knxt_seq_num = 1; | |
586 | p1->pkbdq = pg_vec; | |
bc59ba39 | 587 | pbd = (struct tpacket_block_desc *)pg_vec[0].buffer; |
e3192690 | 588 | p1->pkblk_start = pg_vec[0].buffer; |
f6fb8f10 | 589 | p1->kblk_size = req_u->req3.tp_block_size; |
590 | p1->knum_blocks = req_u->req3.tp_block_nr; | |
591 | p1->hdrlen = po->tp_hdrlen; | |
592 | p1->version = po->tp_version; | |
593 | p1->last_kactive_blk_num = 0; | |
ee80fbf3 | 594 | po->stats.stats3.tp_freeze_q_cnt = 0; |
f6fb8f10 | 595 | if (req_u->req3.tp_retire_blk_tov) |
596 | p1->retire_blk_tov = req_u->req3.tp_retire_blk_tov; | |
597 | else | |
598 | p1->retire_blk_tov = prb_calc_retire_blk_tmo(po, | |
599 | req_u->req3.tp_block_size); | |
600 | p1->tov_in_jiffies = msecs_to_jiffies(p1->retire_blk_tov); | |
601 | p1->blk_sizeof_priv = req_u->req3.tp_sizeof_priv; | |
602 | ||
dc808110 | 603 | p1->max_frame_len = p1->kblk_size - BLK_PLUS_PRIV(p1->blk_sizeof_priv); |
f6fb8f10 | 604 | prb_init_ft_ops(p1, req_u); |
e8e85cc5 | 605 | prb_setup_retire_blk_timer(po); |
f6fb8f10 | 606 | prb_open_block(p1, pbd); |
607 | } | |
608 | ||
609 | /* Do NOT update the last_blk_num first. | |
610 | * Assumes sk_buff_head lock is held. | |
611 | */ | |
bc59ba39 | 612 | static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 613 | { |
614 | mod_timer(&pkc->retire_blk_timer, | |
615 | jiffies + pkc->tov_in_jiffies); | |
616 | pkc->last_kactive_blk_num = pkc->kactive_blk_num; | |
617 | } | |
618 | ||
619 | /* | |
620 | * Timer logic: | |
621 | * 1) We refresh the timer only when we open a block. | |
622 | * By doing this we don't waste cycles refreshing the timer | |
623 | * on packet-by-packet basis. | |
624 | * | |
625 | * With a 1MB block-size, on a 1Gbps line, it will take | |
626 | * i) ~8 ms to fill a block + ii) memcpy etc. | |
627 | * In this cut we are not accounting for the memcpy time. | |
628 | * | |
629 | * So, if the user sets the 'tmo' to 10ms then the timer | |
630 | * will never fire while the block is still getting filled | |
631 | * (which is what we want). However, the user could choose | |
632 | * to close a block early and that's fine. | |
633 | * | |
634 | * But when the timer does fire, we check whether or not to refresh it. | |
635 | * Since the tmo granularity is in msecs, it is not too expensive | |
636 | * to refresh the timer, lets say every '8' msecs. | |
637 | * Either the user can set the 'tmo' or we can derive it based on | |
638 | * a) line-speed and b) block-size. | |
639 | * prb_calc_retire_blk_tmo() calculates the tmo. | |
640 | * | |
641 | */ | |
17bfd8c8 | 642 | static void prb_retire_rx_blk_timer_expired(struct timer_list *t) |
f6fb8f10 | 643 | { |
17bfd8c8 KC |
644 | struct packet_sock *po = |
645 | from_timer(po, t, rx_ring.prb_bdqc.retire_blk_timer); | |
22781a5b | 646 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 647 | unsigned int frozen; |
bc59ba39 | 648 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 649 | |
650 | spin_lock(&po->sk.sk_receive_queue.lock); | |
651 | ||
652 | frozen = prb_queue_frozen(pkc); | |
653 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
654 | ||
655 | if (unlikely(pkc->delete_blk_timer)) | |
656 | goto out; | |
657 | ||
658 | /* We only need to plug the race when the block is partially filled. | |
659 | * tpacket_rcv: | |
660 | * lock(); increment BLOCK_NUM_PKTS; unlock() | |
661 | * copy_bits() is in progress ... | |
662 | * timer fires on other cpu: | |
663 | * we can't retire the current block because copy_bits | |
664 | * is in progress. | |
665 | * | |
666 | */ | |
667 | if (BLOCK_NUM_PKTS(pbd)) { | |
668 | while (atomic_read(&pkc->blk_fill_in_prog)) { | |
669 | /* Waiting for skb_copy_bits to finish... */ | |
670 | cpu_relax(); | |
671 | } | |
672 | } | |
673 | ||
674 | if (pkc->last_kactive_blk_num == pkc->kactive_blk_num) { | |
675 | if (!frozen) { | |
41a50d62 AD |
676 | if (!BLOCK_NUM_PKTS(pbd)) { |
677 | /* An empty block. Just refresh the timer. */ | |
678 | goto refresh_timer; | |
679 | } | |
f6fb8f10 | 680 | prb_retire_current_block(pkc, po, TP_STATUS_BLK_TMO); |
681 | if (!prb_dispatch_next_block(pkc, po)) | |
682 | goto refresh_timer; | |
683 | else | |
684 | goto out; | |
685 | } else { | |
686 | /* Case 1. Queue was frozen because user-space was | |
687 | * lagging behind. | |
688 | */ | |
878cd3ba | 689 | if (prb_curr_blk_in_use(pbd)) { |
f6fb8f10 | 690 | /* |
691 | * Ok, user-space is still behind. | |
692 | * So just refresh the timer. | |
693 | */ | |
694 | goto refresh_timer; | |
695 | } else { | |
696 | /* Case 2. queue was frozen,user-space caught up, | |
697 | * now the link went idle && the timer fired. | |
698 | * We don't have a block to close.So we open this | |
699 | * block and restart the timer. | |
700 | * opening a block thaws the queue,restarts timer | |
701 | * Thawing/timer-refresh is a side effect. | |
702 | */ | |
703 | prb_open_block(pkc, pbd); | |
704 | goto out; | |
705 | } | |
706 | } | |
707 | } | |
708 | ||
709 | refresh_timer: | |
710 | _prb_refresh_rx_retire_blk_timer(pkc); | |
711 | ||
712 | out: | |
713 | spin_unlock(&po->sk.sk_receive_queue.lock); | |
714 | } | |
715 | ||
eea49cc9 | 716 | static void prb_flush_block(struct tpacket_kbdq_core *pkc1, |
bc59ba39 | 717 | struct tpacket_block_desc *pbd1, __u32 status) |
f6fb8f10 | 718 | { |
719 | /* Flush everything minus the block header */ | |
720 | ||
721 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
722 | u8 *start, *end; | |
723 | ||
724 | start = (u8 *)pbd1; | |
725 | ||
726 | /* Skip the block header(we know header WILL fit in 4K) */ | |
727 | start += PAGE_SIZE; | |
728 | ||
729 | end = (u8 *)PAGE_ALIGN((unsigned long)pkc1->pkblk_end); | |
730 | for (; start < end; start += PAGE_SIZE) | |
731 | flush_dcache_page(pgv_to_page(start)); | |
732 | ||
733 | smp_wmb(); | |
734 | #endif | |
735 | ||
736 | /* Now update the block status. */ | |
737 | ||
738 | BLOCK_STATUS(pbd1) = status; | |
739 | ||
740 | /* Flush the block header */ | |
741 | ||
742 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
743 | start = (u8 *)pbd1; | |
744 | flush_dcache_page(pgv_to_page(start)); | |
745 | ||
746 | smp_wmb(); | |
747 | #endif | |
748 | } | |
749 | ||
750 | /* | |
751 | * Side effect: | |
752 | * | |
753 | * 1) flush the block | |
754 | * 2) Increment active_blk_num | |
755 | * | |
756 | * Note:We DONT refresh the timer on purpose. | |
757 | * Because almost always the next block will be opened. | |
758 | */ | |
bc59ba39 | 759 | static void prb_close_block(struct tpacket_kbdq_core *pkc1, |
760 | struct tpacket_block_desc *pbd1, | |
f6fb8f10 | 761 | struct packet_sock *po, unsigned int stat) |
762 | { | |
763 | __u32 status = TP_STATUS_USER | stat; | |
764 | ||
765 | struct tpacket3_hdr *last_pkt; | |
bc59ba39 | 766 | struct tpacket_hdr_v1 *h1 = &pbd1->hdr.bh1; |
da413eec | 767 | struct sock *sk = &po->sk; |
f6fb8f10 | 768 | |
ee80fbf3 | 769 | if (po->stats.stats3.tp_drops) |
f6fb8f10 | 770 | status |= TP_STATUS_LOSING; |
771 | ||
772 | last_pkt = (struct tpacket3_hdr *)pkc1->prev; | |
773 | last_pkt->tp_next_offset = 0; | |
774 | ||
775 | /* Get the ts of the last pkt */ | |
776 | if (BLOCK_NUM_PKTS(pbd1)) { | |
777 | h1->ts_last_pkt.ts_sec = last_pkt->tp_sec; | |
778 | h1->ts_last_pkt.ts_nsec = last_pkt->tp_nsec; | |
779 | } else { | |
41a50d62 AD |
780 | /* Ok, we tmo'd - so get the current time. |
781 | * | |
782 | * It shouldn't really happen as we don't close empty | |
783 | * blocks. See prb_retire_rx_blk_timer_expired(). | |
784 | */ | |
f6fb8f10 | 785 | struct timespec ts; |
786 | getnstimeofday(&ts); | |
787 | h1->ts_last_pkt.ts_sec = ts.tv_sec; | |
788 | h1->ts_last_pkt.ts_nsec = ts.tv_nsec; | |
789 | } | |
790 | ||
791 | smp_wmb(); | |
792 | ||
793 | /* Flush the block */ | |
794 | prb_flush_block(pkc1, pbd1, status); | |
795 | ||
da413eec DC |
796 | sk->sk_data_ready(sk); |
797 | ||
f6fb8f10 | 798 | pkc1->kactive_blk_num = GET_NEXT_PRB_BLK_NUM(pkc1); |
799 | } | |
800 | ||
eea49cc9 | 801 | static void prb_thaw_queue(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 802 | { |
803 | pkc->reset_pending_on_curr_blk = 0; | |
804 | } | |
805 | ||
806 | /* | |
807 | * Side effect of opening a block: | |
808 | * | |
809 | * 1) prb_queue is thawed. | |
810 | * 2) retire_blk_timer is refreshed. | |
811 | * | |
812 | */ | |
bc59ba39 | 813 | static void prb_open_block(struct tpacket_kbdq_core *pkc1, |
814 | struct tpacket_block_desc *pbd1) | |
f6fb8f10 | 815 | { |
816 | struct timespec ts; | |
bc59ba39 | 817 | struct tpacket_hdr_v1 *h1 = &pbd1->hdr.bh1; |
f6fb8f10 | 818 | |
819 | smp_rmb(); | |
820 | ||
8da3056c DB |
821 | /* We could have just memset this but we will lose the |
822 | * flexibility of making the priv area sticky | |
823 | */ | |
f6fb8f10 | 824 | |
8da3056c DB |
825 | BLOCK_SNUM(pbd1) = pkc1->knxt_seq_num++; |
826 | BLOCK_NUM_PKTS(pbd1) = 0; | |
827 | BLOCK_LEN(pbd1) = BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
f6fb8f10 | 828 | |
8da3056c DB |
829 | getnstimeofday(&ts); |
830 | ||
831 | h1->ts_first_pkt.ts_sec = ts.tv_sec; | |
832 | h1->ts_first_pkt.ts_nsec = ts.tv_nsec; | |
f6fb8f10 | 833 | |
8da3056c DB |
834 | pkc1->pkblk_start = (char *)pbd1; |
835 | pkc1->nxt_offset = pkc1->pkblk_start + BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
836 | ||
837 | BLOCK_O2FP(pbd1) = (__u32)BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
838 | BLOCK_O2PRIV(pbd1) = BLK_HDR_LEN; | |
839 | ||
840 | pbd1->version = pkc1->version; | |
841 | pkc1->prev = pkc1->nxt_offset; | |
842 | pkc1->pkblk_end = pkc1->pkblk_start + pkc1->kblk_size; | |
843 | ||
844 | prb_thaw_queue(pkc1); | |
845 | _prb_refresh_rx_retire_blk_timer(pkc1); | |
846 | ||
847 | smp_wmb(); | |
f6fb8f10 | 848 | } |
849 | ||
850 | /* | |
851 | * Queue freeze logic: | |
852 | * 1) Assume tp_block_nr = 8 blocks. | |
853 | * 2) At time 't0', user opens Rx ring. | |
854 | * 3) Some time past 't0', kernel starts filling blocks starting from 0 .. 7 | |
855 | * 4) user-space is either sleeping or processing block '0'. | |
856 | * 5) tpacket_rcv is currently filling block '7', since there is no space left, | |
857 | * it will close block-7,loop around and try to fill block '0'. | |
858 | * call-flow: | |
859 | * __packet_lookup_frame_in_block | |
860 | * prb_retire_current_block() | |
861 | * prb_dispatch_next_block() | |
862 | * |->(BLOCK_STATUS == USER) evaluates to true | |
863 | * 5.1) Since block-0 is currently in-use, we just freeze the queue. | |
864 | * 6) Now there are two cases: | |
865 | * 6.1) Link goes idle right after the queue is frozen. | |
866 | * But remember, the last open_block() refreshed the timer. | |
867 | * When this timer expires,it will refresh itself so that we can | |
868 | * re-open block-0 in near future. | |
869 | * 6.2) Link is busy and keeps on receiving packets. This is a simple | |
870 | * case and __packet_lookup_frame_in_block will check if block-0 | |
871 | * is free and can now be re-used. | |
872 | */ | |
eea49cc9 | 873 | static void prb_freeze_queue(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 874 | struct packet_sock *po) |
875 | { | |
876 | pkc->reset_pending_on_curr_blk = 1; | |
ee80fbf3 | 877 | po->stats.stats3.tp_freeze_q_cnt++; |
f6fb8f10 | 878 | } |
879 | ||
880 | #define TOTAL_PKT_LEN_INCL_ALIGN(length) (ALIGN((length), V3_ALIGNMENT)) | |
881 | ||
882 | /* | |
883 | * If the next block is free then we will dispatch it | |
884 | * and return a good offset. | |
885 | * Else, we will freeze the queue. | |
886 | * So, caller must check the return value. | |
887 | */ | |
bc59ba39 | 888 | static void *prb_dispatch_next_block(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 889 | struct packet_sock *po) |
890 | { | |
bc59ba39 | 891 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 892 | |
893 | smp_rmb(); | |
894 | ||
895 | /* 1. Get current block num */ | |
896 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
897 | ||
898 | /* 2. If this block is currently in_use then freeze the queue */ | |
899 | if (TP_STATUS_USER & BLOCK_STATUS(pbd)) { | |
900 | prb_freeze_queue(pkc, po); | |
901 | return NULL; | |
902 | } | |
903 | ||
904 | /* | |
905 | * 3. | |
906 | * open this block and return the offset where the first packet | |
907 | * needs to get stored. | |
908 | */ | |
909 | prb_open_block(pkc, pbd); | |
910 | return (void *)pkc->nxt_offset; | |
911 | } | |
912 | ||
bc59ba39 | 913 | static void prb_retire_current_block(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 914 | struct packet_sock *po, unsigned int status) |
915 | { | |
bc59ba39 | 916 | struct tpacket_block_desc *pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); |
f6fb8f10 | 917 | |
918 | /* retire/close the current block */ | |
919 | if (likely(TP_STATUS_KERNEL == BLOCK_STATUS(pbd))) { | |
920 | /* | |
921 | * Plug the case where copy_bits() is in progress on | |
922 | * cpu-0 and tpacket_rcv() got invoked on cpu-1, didn't | |
923 | * have space to copy the pkt in the current block and | |
924 | * called prb_retire_current_block() | |
925 | * | |
926 | * We don't need to worry about the TMO case because | |
927 | * the timer-handler already handled this case. | |
928 | */ | |
929 | if (!(status & TP_STATUS_BLK_TMO)) { | |
930 | while (atomic_read(&pkc->blk_fill_in_prog)) { | |
931 | /* Waiting for skb_copy_bits to finish... */ | |
932 | cpu_relax(); | |
933 | } | |
934 | } | |
935 | prb_close_block(pkc, pbd, po, status); | |
936 | return; | |
937 | } | |
f6fb8f10 | 938 | } |
939 | ||
878cd3ba | 940 | static int prb_curr_blk_in_use(struct tpacket_block_desc *pbd) |
f6fb8f10 | 941 | { |
942 | return TP_STATUS_USER & BLOCK_STATUS(pbd); | |
943 | } | |
944 | ||
eea49cc9 | 945 | static int prb_queue_frozen(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 946 | { |
947 | return pkc->reset_pending_on_curr_blk; | |
948 | } | |
949 | ||
eea49cc9 | 950 | static void prb_clear_blk_fill_status(struct packet_ring_buffer *rb) |
f6fb8f10 | 951 | { |
bc59ba39 | 952 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(rb); |
f6fb8f10 | 953 | atomic_dec(&pkc->blk_fill_in_prog); |
954 | } | |
955 | ||
eea49cc9 | 956 | static void prb_fill_rxhash(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 957 | struct tpacket3_hdr *ppd) |
958 | { | |
3958afa1 | 959 | ppd->hv1.tp_rxhash = skb_get_hash(pkc->skb); |
f6fb8f10 | 960 | } |
961 | ||
eea49cc9 | 962 | static void prb_clear_rxhash(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 963 | struct tpacket3_hdr *ppd) |
964 | { | |
965 | ppd->hv1.tp_rxhash = 0; | |
966 | } | |
967 | ||
eea49cc9 | 968 | static void prb_fill_vlan_info(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 969 | struct tpacket3_hdr *ppd) |
970 | { | |
df8a39de JP |
971 | if (skb_vlan_tag_present(pkc->skb)) { |
972 | ppd->hv1.tp_vlan_tci = skb_vlan_tag_get(pkc->skb); | |
a0cdfcf3 AW |
973 | ppd->hv1.tp_vlan_tpid = ntohs(pkc->skb->vlan_proto); |
974 | ppd->tp_status = TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
f6fb8f10 | 975 | } else { |
9e67030a | 976 | ppd->hv1.tp_vlan_tci = 0; |
a0cdfcf3 | 977 | ppd->hv1.tp_vlan_tpid = 0; |
9e67030a | 978 | ppd->tp_status = TP_STATUS_AVAILABLE; |
f6fb8f10 | 979 | } |
980 | } | |
981 | ||
bc59ba39 | 982 | static void prb_run_all_ft_ops(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 983 | struct tpacket3_hdr *ppd) |
984 | { | |
a0cdfcf3 | 985 | ppd->hv1.tp_padding = 0; |
f6fb8f10 | 986 | prb_fill_vlan_info(pkc, ppd); |
987 | ||
988 | if (pkc->feature_req_word & TP_FT_REQ_FILL_RXHASH) | |
989 | prb_fill_rxhash(pkc, ppd); | |
990 | else | |
991 | prb_clear_rxhash(pkc, ppd); | |
992 | } | |
993 | ||
eea49cc9 | 994 | static void prb_fill_curr_block(char *curr, |
bc59ba39 | 995 | struct tpacket_kbdq_core *pkc, |
996 | struct tpacket_block_desc *pbd, | |
f6fb8f10 | 997 | unsigned int len) |
998 | { | |
999 | struct tpacket3_hdr *ppd; | |
1000 | ||
1001 | ppd = (struct tpacket3_hdr *)curr; | |
1002 | ppd->tp_next_offset = TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1003 | pkc->prev = curr; | |
1004 | pkc->nxt_offset += TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1005 | BLOCK_LEN(pbd) += TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1006 | BLOCK_NUM_PKTS(pbd) += 1; | |
1007 | atomic_inc(&pkc->blk_fill_in_prog); | |
1008 | prb_run_all_ft_ops(pkc, ppd); | |
1009 | } | |
1010 | ||
1011 | /* Assumes caller has the sk->rx_queue.lock */ | |
1012 | static void *__packet_lookup_frame_in_block(struct packet_sock *po, | |
1013 | struct sk_buff *skb, | |
1014 | int status, | |
1015 | unsigned int len | |
1016 | ) | |
1017 | { | |
bc59ba39 | 1018 | struct tpacket_kbdq_core *pkc; |
1019 | struct tpacket_block_desc *pbd; | |
f6fb8f10 | 1020 | char *curr, *end; |
1021 | ||
e3192690 | 1022 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 1023 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); |
1024 | ||
1025 | /* Queue is frozen when user space is lagging behind */ | |
1026 | if (prb_queue_frozen(pkc)) { | |
1027 | /* | |
1028 | * Check if that last block which caused the queue to freeze, | |
1029 | * is still in_use by user-space. | |
1030 | */ | |
878cd3ba | 1031 | if (prb_curr_blk_in_use(pbd)) { |
f6fb8f10 | 1032 | /* Can't record this packet */ |
1033 | return NULL; | |
1034 | } else { | |
1035 | /* | |
1036 | * Ok, the block was released by user-space. | |
1037 | * Now let's open that block. | |
1038 | * opening a block also thaws the queue. | |
1039 | * Thawing is a side effect. | |
1040 | */ | |
1041 | prb_open_block(pkc, pbd); | |
1042 | } | |
1043 | } | |
1044 | ||
1045 | smp_mb(); | |
1046 | curr = pkc->nxt_offset; | |
1047 | pkc->skb = skb; | |
e3192690 | 1048 | end = (char *)pbd + pkc->kblk_size; |
f6fb8f10 | 1049 | |
1050 | /* first try the current block */ | |
1051 | if (curr+TOTAL_PKT_LEN_INCL_ALIGN(len) < end) { | |
1052 | prb_fill_curr_block(curr, pkc, pbd, len); | |
1053 | return (void *)curr; | |
1054 | } | |
1055 | ||
1056 | /* Ok, close the current block */ | |
1057 | prb_retire_current_block(pkc, po, 0); | |
1058 | ||
1059 | /* Now, try to dispatch the next block */ | |
1060 | curr = (char *)prb_dispatch_next_block(pkc, po); | |
1061 | if (curr) { | |
1062 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
1063 | prb_fill_curr_block(curr, pkc, pbd, len); | |
1064 | return (void *)curr; | |
1065 | } | |
1066 | ||
1067 | /* | |
1068 | * No free blocks are available.user_space hasn't caught up yet. | |
1069 | * Queue was just frozen and now this packet will get dropped. | |
1070 | */ | |
1071 | return NULL; | |
1072 | } | |
1073 | ||
eea49cc9 | 1074 | static void *packet_current_rx_frame(struct packet_sock *po, |
f6fb8f10 | 1075 | struct sk_buff *skb, |
1076 | int status, unsigned int len) | |
1077 | { | |
1078 | char *curr = NULL; | |
1079 | switch (po->tp_version) { | |
1080 | case TPACKET_V1: | |
1081 | case TPACKET_V2: | |
1082 | curr = packet_lookup_frame(po, &po->rx_ring, | |
1083 | po->rx_ring.head, status); | |
1084 | return curr; | |
1085 | case TPACKET_V3: | |
1086 | return __packet_lookup_frame_in_block(po, skb, status, len); | |
1087 | default: | |
1088 | WARN(1, "TPACKET version not supported\n"); | |
1089 | BUG(); | |
99aa3473 | 1090 | return NULL; |
f6fb8f10 | 1091 | } |
1092 | } | |
1093 | ||
eea49cc9 | 1094 | static void *prb_lookup_block(struct packet_sock *po, |
f6fb8f10 | 1095 | struct packet_ring_buffer *rb, |
77f65ebd | 1096 | unsigned int idx, |
f6fb8f10 | 1097 | int status) |
1098 | { | |
bc59ba39 | 1099 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(rb); |
77f65ebd | 1100 | struct tpacket_block_desc *pbd = GET_PBLOCK_DESC(pkc, idx); |
f6fb8f10 | 1101 | |
1102 | if (status != BLOCK_STATUS(pbd)) | |
1103 | return NULL; | |
1104 | return pbd; | |
1105 | } | |
1106 | ||
eea49cc9 | 1107 | static int prb_previous_blk_num(struct packet_ring_buffer *rb) |
f6fb8f10 | 1108 | { |
1109 | unsigned int prev; | |
1110 | if (rb->prb_bdqc.kactive_blk_num) | |
1111 | prev = rb->prb_bdqc.kactive_blk_num-1; | |
1112 | else | |
1113 | prev = rb->prb_bdqc.knum_blocks-1; | |
1114 | return prev; | |
1115 | } | |
1116 | ||
1117 | /* Assumes caller has held the rx_queue.lock */ | |
eea49cc9 | 1118 | static void *__prb_previous_block(struct packet_sock *po, |
f6fb8f10 | 1119 | struct packet_ring_buffer *rb, |
1120 | int status) | |
1121 | { | |
1122 | unsigned int previous = prb_previous_blk_num(rb); | |
1123 | return prb_lookup_block(po, rb, previous, status); | |
1124 | } | |
1125 | ||
eea49cc9 | 1126 | static void *packet_previous_rx_frame(struct packet_sock *po, |
f6fb8f10 | 1127 | struct packet_ring_buffer *rb, |
1128 | int status) | |
1129 | { | |
1130 | if (po->tp_version <= TPACKET_V2) | |
1131 | return packet_previous_frame(po, rb, status); | |
1132 | ||
1133 | return __prb_previous_block(po, rb, status); | |
1134 | } | |
1135 | ||
eea49cc9 | 1136 | static void packet_increment_rx_head(struct packet_sock *po, |
f6fb8f10 | 1137 | struct packet_ring_buffer *rb) |
1138 | { | |
1139 | switch (po->tp_version) { | |
1140 | case TPACKET_V1: | |
1141 | case TPACKET_V2: | |
1142 | return packet_increment_head(rb); | |
1143 | case TPACKET_V3: | |
1144 | default: | |
1145 | WARN(1, "TPACKET version not supported.\n"); | |
1146 | BUG(); | |
1147 | return; | |
1148 | } | |
1149 | } | |
1150 | ||
eea49cc9 | 1151 | static void *packet_previous_frame(struct packet_sock *po, |
69e3c75f JB |
1152 | struct packet_ring_buffer *rb, |
1153 | int status) | |
1154 | { | |
1155 | unsigned int previous = rb->head ? rb->head - 1 : rb->frame_max; | |
1156 | return packet_lookup_frame(po, rb, previous, status); | |
1157 | } | |
1158 | ||
eea49cc9 | 1159 | static void packet_increment_head(struct packet_ring_buffer *buff) |
69e3c75f JB |
1160 | { |
1161 | buff->head = buff->head != buff->frame_max ? buff->head+1 : 0; | |
1162 | } | |
1163 | ||
b0138408 DB |
1164 | static void packet_inc_pending(struct packet_ring_buffer *rb) |
1165 | { | |
1166 | this_cpu_inc(*rb->pending_refcnt); | |
1167 | } | |
1168 | ||
1169 | static void packet_dec_pending(struct packet_ring_buffer *rb) | |
1170 | { | |
1171 | this_cpu_dec(*rb->pending_refcnt); | |
1172 | } | |
1173 | ||
1174 | static unsigned int packet_read_pending(const struct packet_ring_buffer *rb) | |
1175 | { | |
1176 | unsigned int refcnt = 0; | |
1177 | int cpu; | |
1178 | ||
1179 | /* We don't use pending refcount in rx_ring. */ | |
1180 | if (rb->pending_refcnt == NULL) | |
1181 | return 0; | |
1182 | ||
1183 | for_each_possible_cpu(cpu) | |
1184 | refcnt += *per_cpu_ptr(rb->pending_refcnt, cpu); | |
1185 | ||
1186 | return refcnt; | |
1187 | } | |
1188 | ||
1189 | static int packet_alloc_pending(struct packet_sock *po) | |
1190 | { | |
1191 | po->rx_ring.pending_refcnt = NULL; | |
1192 | ||
1193 | po->tx_ring.pending_refcnt = alloc_percpu(unsigned int); | |
1194 | if (unlikely(po->tx_ring.pending_refcnt == NULL)) | |
1195 | return -ENOBUFS; | |
1196 | ||
1197 | return 0; | |
1198 | } | |
1199 | ||
1200 | static void packet_free_pending(struct packet_sock *po) | |
1201 | { | |
1202 | free_percpu(po->tx_ring.pending_refcnt); | |
1203 | } | |
1204 | ||
9954729b WB |
1205 | #define ROOM_POW_OFF 2 |
1206 | #define ROOM_NONE 0x0 | |
1207 | #define ROOM_LOW 0x1 | |
1208 | #define ROOM_NORMAL 0x2 | |
1209 | ||
1210 | static bool __tpacket_has_room(struct packet_sock *po, int pow_off) | |
77f65ebd | 1211 | { |
9954729b WB |
1212 | int idx, len; |
1213 | ||
1214 | len = po->rx_ring.frame_max + 1; | |
1215 | idx = po->rx_ring.head; | |
1216 | if (pow_off) | |
1217 | idx += len >> pow_off; | |
1218 | if (idx >= len) | |
1219 | idx -= len; | |
1220 | return packet_lookup_frame(po, &po->rx_ring, idx, TP_STATUS_KERNEL); | |
1221 | } | |
1222 | ||
1223 | static bool __tpacket_v3_has_room(struct packet_sock *po, int pow_off) | |
1224 | { | |
1225 | int idx, len; | |
1226 | ||
1227 | len = po->rx_ring.prb_bdqc.knum_blocks; | |
1228 | idx = po->rx_ring.prb_bdqc.kactive_blk_num; | |
1229 | if (pow_off) | |
1230 | idx += len >> pow_off; | |
1231 | if (idx >= len) | |
1232 | idx -= len; | |
1233 | return prb_lookup_block(po, &po->rx_ring, idx, TP_STATUS_KERNEL); | |
1234 | } | |
77f65ebd | 1235 | |
2ccdbaa6 | 1236 | static int __packet_rcv_has_room(struct packet_sock *po, struct sk_buff *skb) |
9954729b WB |
1237 | { |
1238 | struct sock *sk = &po->sk; | |
1239 | int ret = ROOM_NONE; | |
1240 | ||
1241 | if (po->prot_hook.func != tpacket_rcv) { | |
1242 | int avail = sk->sk_rcvbuf - atomic_read(&sk->sk_rmem_alloc) | |
2ccdbaa6 | 1243 | - (skb ? skb->truesize : 0); |
9954729b WB |
1244 | if (avail > (sk->sk_rcvbuf >> ROOM_POW_OFF)) |
1245 | return ROOM_NORMAL; | |
1246 | else if (avail > 0) | |
1247 | return ROOM_LOW; | |
1248 | else | |
1249 | return ROOM_NONE; | |
1250 | } | |
77f65ebd | 1251 | |
9954729b WB |
1252 | if (po->tp_version == TPACKET_V3) { |
1253 | if (__tpacket_v3_has_room(po, ROOM_POW_OFF)) | |
1254 | ret = ROOM_NORMAL; | |
1255 | else if (__tpacket_v3_has_room(po, 0)) | |
1256 | ret = ROOM_LOW; | |
1257 | } else { | |
1258 | if (__tpacket_has_room(po, ROOM_POW_OFF)) | |
1259 | ret = ROOM_NORMAL; | |
1260 | else if (__tpacket_has_room(po, 0)) | |
1261 | ret = ROOM_LOW; | |
1262 | } | |
2ccdbaa6 WB |
1263 | |
1264 | return ret; | |
1265 | } | |
1266 | ||
1267 | static int packet_rcv_has_room(struct packet_sock *po, struct sk_buff *skb) | |
1268 | { | |
1269 | int ret; | |
1270 | bool has_room; | |
1271 | ||
54d7c01d WB |
1272 | spin_lock_bh(&po->sk.sk_receive_queue.lock); |
1273 | ret = __packet_rcv_has_room(po, skb); | |
2ccdbaa6 WB |
1274 | has_room = ret == ROOM_NORMAL; |
1275 | if (po->pressure == has_room) | |
54d7c01d WB |
1276 | po->pressure = !has_room; |
1277 | spin_unlock_bh(&po->sk.sk_receive_queue.lock); | |
77f65ebd | 1278 | |
9954729b | 1279 | return ret; |
77f65ebd WB |
1280 | } |
1281 | ||
1da177e4 LT |
1282 | static void packet_sock_destruct(struct sock *sk) |
1283 | { | |
ed85b565 RC |
1284 | skb_queue_purge(&sk->sk_error_queue); |
1285 | ||
547b792c | 1286 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); |
14afee4b | 1287 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
1da177e4 LT |
1288 | |
1289 | if (!sock_flag(sk, SOCK_DEAD)) { | |
40d4e3df | 1290 | pr_err("Attempt to release alive packet socket: %p\n", sk); |
1da177e4 LT |
1291 | return; |
1292 | } | |
1293 | ||
17ab56a2 | 1294 | sk_refcnt_debug_dec(sk); |
1da177e4 LT |
1295 | } |
1296 | ||
3b3a5b0a WB |
1297 | static bool fanout_flow_is_huge(struct packet_sock *po, struct sk_buff *skb) |
1298 | { | |
1299 | u32 rxhash; | |
1300 | int i, count = 0; | |
1301 | ||
1302 | rxhash = skb_get_hash(skb); | |
1303 | for (i = 0; i < ROLLOVER_HLEN; i++) | |
1304 | if (po->rollover->history[i] == rxhash) | |
1305 | count++; | |
1306 | ||
1307 | po->rollover->history[prandom_u32() % ROLLOVER_HLEN] = rxhash; | |
1308 | return count > (ROLLOVER_HLEN >> 1); | |
1309 | } | |
1310 | ||
77f65ebd WB |
1311 | static unsigned int fanout_demux_hash(struct packet_fanout *f, |
1312 | struct sk_buff *skb, | |
1313 | unsigned int num) | |
dc99f600 | 1314 | { |
eb70db87 | 1315 | return reciprocal_scale(__skb_get_hash_symmetric(skb), num); |
dc99f600 DM |
1316 | } |
1317 | ||
77f65ebd WB |
1318 | static unsigned int fanout_demux_lb(struct packet_fanout *f, |
1319 | struct sk_buff *skb, | |
1320 | unsigned int num) | |
dc99f600 | 1321 | { |
468479e6 | 1322 | unsigned int val = atomic_inc_return(&f->rr_cur); |
dc99f600 | 1323 | |
468479e6 | 1324 | return val % num; |
77f65ebd WB |
1325 | } |
1326 | ||
1327 | static unsigned int fanout_demux_cpu(struct packet_fanout *f, | |
1328 | struct sk_buff *skb, | |
1329 | unsigned int num) | |
1330 | { | |
1331 | return smp_processor_id() % num; | |
dc99f600 DM |
1332 | } |
1333 | ||
5df0ddfb DB |
1334 | static unsigned int fanout_demux_rnd(struct packet_fanout *f, |
1335 | struct sk_buff *skb, | |
1336 | unsigned int num) | |
1337 | { | |
f337db64 | 1338 | return prandom_u32_max(num); |
5df0ddfb DB |
1339 | } |
1340 | ||
77f65ebd WB |
1341 | static unsigned int fanout_demux_rollover(struct packet_fanout *f, |
1342 | struct sk_buff *skb, | |
ad377cab | 1343 | unsigned int idx, bool try_self, |
77f65ebd | 1344 | unsigned int num) |
95ec3eb4 | 1345 | { |
4633c9e0 | 1346 | struct packet_sock *po, *po_next, *po_skip = NULL; |
a9b63918 | 1347 | unsigned int i, j, room = ROOM_NONE; |
95ec3eb4 | 1348 | |
0648ab70 | 1349 | po = pkt_sk(f->arr[idx]); |
3b3a5b0a WB |
1350 | |
1351 | if (try_self) { | |
1352 | room = packet_rcv_has_room(po, skb); | |
1353 | if (room == ROOM_NORMAL || | |
1354 | (room == ROOM_LOW && !fanout_flow_is_huge(po, skb))) | |
1355 | return idx; | |
4633c9e0 | 1356 | po_skip = po; |
3b3a5b0a | 1357 | } |
ad377cab | 1358 | |
0648ab70 | 1359 | i = j = min_t(int, po->rollover->sock, num - 1); |
77f65ebd | 1360 | do { |
2ccdbaa6 | 1361 | po_next = pkt_sk(f->arr[i]); |
4633c9e0 | 1362 | if (po_next != po_skip && !po_next->pressure && |
2ccdbaa6 | 1363 | packet_rcv_has_room(po_next, skb) == ROOM_NORMAL) { |
77f65ebd | 1364 | if (i != j) |
0648ab70 | 1365 | po->rollover->sock = i; |
a9b63918 WB |
1366 | atomic_long_inc(&po->rollover->num); |
1367 | if (room == ROOM_LOW) | |
1368 | atomic_long_inc(&po->rollover->num_huge); | |
77f65ebd WB |
1369 | return i; |
1370 | } | |
ad377cab | 1371 | |
77f65ebd WB |
1372 | if (++i == num) |
1373 | i = 0; | |
1374 | } while (i != j); | |
1375 | ||
a9b63918 | 1376 | atomic_long_inc(&po->rollover->num_failed); |
77f65ebd WB |
1377 | return idx; |
1378 | } | |
1379 | ||
2d36097d NH |
1380 | static unsigned int fanout_demux_qm(struct packet_fanout *f, |
1381 | struct sk_buff *skb, | |
1382 | unsigned int num) | |
1383 | { | |
1384 | return skb_get_queue_mapping(skb) % num; | |
1385 | } | |
1386 | ||
47dceb8e WB |
1387 | static unsigned int fanout_demux_bpf(struct packet_fanout *f, |
1388 | struct sk_buff *skb, | |
1389 | unsigned int num) | |
1390 | { | |
1391 | struct bpf_prog *prog; | |
1392 | unsigned int ret = 0; | |
1393 | ||
1394 | rcu_read_lock(); | |
1395 | prog = rcu_dereference(f->bpf_prog); | |
1396 | if (prog) | |
ff936a04 | 1397 | ret = bpf_prog_run_clear_cb(prog, skb) % num; |
47dceb8e WB |
1398 | rcu_read_unlock(); |
1399 | ||
1400 | return ret; | |
1401 | } | |
1402 | ||
77f65ebd WB |
1403 | static bool fanout_has_flag(struct packet_fanout *f, u16 flag) |
1404 | { | |
1405 | return f->flags & (flag >> 8); | |
95ec3eb4 DM |
1406 | } |
1407 | ||
95ec3eb4 DM |
1408 | static int packet_rcv_fanout(struct sk_buff *skb, struct net_device *dev, |
1409 | struct packet_type *pt, struct net_device *orig_dev) | |
dc99f600 DM |
1410 | { |
1411 | struct packet_fanout *f = pt->af_packet_priv; | |
f98f4514 | 1412 | unsigned int num = READ_ONCE(f->num_members); |
19bcf9f2 | 1413 | struct net *net = read_pnet(&f->net); |
dc99f600 | 1414 | struct packet_sock *po; |
77f65ebd | 1415 | unsigned int idx; |
dc99f600 | 1416 | |
19bcf9f2 | 1417 | if (!net_eq(dev_net(dev), net) || !num) { |
dc99f600 DM |
1418 | kfree_skb(skb); |
1419 | return 0; | |
1420 | } | |
1421 | ||
3f34b24a | 1422 | if (fanout_has_flag(f, PACKET_FANOUT_FLAG_DEFRAG)) { |
19bcf9f2 | 1423 | skb = ip_check_defrag(net, skb, IP_DEFRAG_AF_PACKET); |
3f34b24a AD |
1424 | if (!skb) |
1425 | return 0; | |
1426 | } | |
95ec3eb4 DM |
1427 | switch (f->type) { |
1428 | case PACKET_FANOUT_HASH: | |
1429 | default: | |
77f65ebd | 1430 | idx = fanout_demux_hash(f, skb, num); |
95ec3eb4 DM |
1431 | break; |
1432 | case PACKET_FANOUT_LB: | |
77f65ebd | 1433 | idx = fanout_demux_lb(f, skb, num); |
95ec3eb4 DM |
1434 | break; |
1435 | case PACKET_FANOUT_CPU: | |
77f65ebd WB |
1436 | idx = fanout_demux_cpu(f, skb, num); |
1437 | break; | |
5df0ddfb DB |
1438 | case PACKET_FANOUT_RND: |
1439 | idx = fanout_demux_rnd(f, skb, num); | |
1440 | break; | |
2d36097d NH |
1441 | case PACKET_FANOUT_QM: |
1442 | idx = fanout_demux_qm(f, skb, num); | |
1443 | break; | |
77f65ebd | 1444 | case PACKET_FANOUT_ROLLOVER: |
ad377cab | 1445 | idx = fanout_demux_rollover(f, skb, 0, false, num); |
95ec3eb4 | 1446 | break; |
47dceb8e | 1447 | case PACKET_FANOUT_CBPF: |
f2e52095 | 1448 | case PACKET_FANOUT_EBPF: |
47dceb8e WB |
1449 | idx = fanout_demux_bpf(f, skb, num); |
1450 | break; | |
dc99f600 DM |
1451 | } |
1452 | ||
ad377cab WB |
1453 | if (fanout_has_flag(f, PACKET_FANOUT_FLAG_ROLLOVER)) |
1454 | idx = fanout_demux_rollover(f, skb, idx, true, num); | |
dc99f600 | 1455 | |
ad377cab | 1456 | po = pkt_sk(f->arr[idx]); |
dc99f600 DM |
1457 | return po->prot_hook.func(skb, dev, &po->prot_hook, orig_dev); |
1458 | } | |
1459 | ||
fff3321d PE |
1460 | DEFINE_MUTEX(fanout_mutex); |
1461 | EXPORT_SYMBOL_GPL(fanout_mutex); | |
dc99f600 | 1462 | static LIST_HEAD(fanout_list); |
4a69a864 | 1463 | static u16 fanout_next_id; |
dc99f600 DM |
1464 | |
1465 | static void __fanout_link(struct sock *sk, struct packet_sock *po) | |
1466 | { | |
1467 | struct packet_fanout *f = po->fanout; | |
1468 | ||
1469 | spin_lock(&f->lock); | |
1470 | f->arr[f->num_members] = sk; | |
1471 | smp_wmb(); | |
1472 | f->num_members++; | |
2bd624b4 AS |
1473 | if (f->num_members == 1) |
1474 | dev_add_pack(&f->prot_hook); | |
dc99f600 DM |
1475 | spin_unlock(&f->lock); |
1476 | } | |
1477 | ||
1478 | static void __fanout_unlink(struct sock *sk, struct packet_sock *po) | |
1479 | { | |
1480 | struct packet_fanout *f = po->fanout; | |
1481 | int i; | |
1482 | ||
1483 | spin_lock(&f->lock); | |
1484 | for (i = 0; i < f->num_members; i++) { | |
1485 | if (f->arr[i] == sk) | |
1486 | break; | |
1487 | } | |
1488 | BUG_ON(i >= f->num_members); | |
1489 | f->arr[i] = f->arr[f->num_members - 1]; | |
1490 | f->num_members--; | |
2bd624b4 AS |
1491 | if (f->num_members == 0) |
1492 | __dev_remove_pack(&f->prot_hook); | |
dc99f600 DM |
1493 | spin_unlock(&f->lock); |
1494 | } | |
1495 | ||
d4dd8aee | 1496 | static bool match_fanout_group(struct packet_type *ptype, struct sock *sk) |
c0de08d0 | 1497 | { |
161642e2 ED |
1498 | if (sk->sk_family != PF_PACKET) |
1499 | return false; | |
c0de08d0 | 1500 | |
161642e2 | 1501 | return ptype->af_packet_priv == pkt_sk(sk)->fanout; |
c0de08d0 EL |
1502 | } |
1503 | ||
47dceb8e WB |
1504 | static void fanout_init_data(struct packet_fanout *f) |
1505 | { | |
1506 | switch (f->type) { | |
1507 | case PACKET_FANOUT_LB: | |
1508 | atomic_set(&f->rr_cur, 0); | |
1509 | break; | |
1510 | case PACKET_FANOUT_CBPF: | |
f2e52095 | 1511 | case PACKET_FANOUT_EBPF: |
47dceb8e WB |
1512 | RCU_INIT_POINTER(f->bpf_prog, NULL); |
1513 | break; | |
1514 | } | |
1515 | } | |
1516 | ||
1517 | static void __fanout_set_data_bpf(struct packet_fanout *f, struct bpf_prog *new) | |
1518 | { | |
1519 | struct bpf_prog *old; | |
1520 | ||
1521 | spin_lock(&f->lock); | |
1522 | old = rcu_dereference_protected(f->bpf_prog, lockdep_is_held(&f->lock)); | |
1523 | rcu_assign_pointer(f->bpf_prog, new); | |
1524 | spin_unlock(&f->lock); | |
1525 | ||
1526 | if (old) { | |
1527 | synchronize_net(); | |
1528 | bpf_prog_destroy(old); | |
1529 | } | |
1530 | } | |
1531 | ||
1532 | static int fanout_set_data_cbpf(struct packet_sock *po, char __user *data, | |
1533 | unsigned int len) | |
1534 | { | |
1535 | struct bpf_prog *new; | |
1536 | struct sock_fprog fprog; | |
1537 | int ret; | |
1538 | ||
1539 | if (sock_flag(&po->sk, SOCK_FILTER_LOCKED)) | |
1540 | return -EPERM; | |
1541 | if (len != sizeof(fprog)) | |
1542 | return -EINVAL; | |
1543 | if (copy_from_user(&fprog, data, len)) | |
1544 | return -EFAULT; | |
1545 | ||
bab18991 | 1546 | ret = bpf_prog_create_from_user(&new, &fprog, NULL, false); |
47dceb8e WB |
1547 | if (ret) |
1548 | return ret; | |
1549 | ||
1550 | __fanout_set_data_bpf(po->fanout, new); | |
1551 | return 0; | |
1552 | } | |
1553 | ||
f2e52095 WB |
1554 | static int fanout_set_data_ebpf(struct packet_sock *po, char __user *data, |
1555 | unsigned int len) | |
1556 | { | |
1557 | struct bpf_prog *new; | |
1558 | u32 fd; | |
1559 | ||
1560 | if (sock_flag(&po->sk, SOCK_FILTER_LOCKED)) | |
1561 | return -EPERM; | |
1562 | if (len != sizeof(fd)) | |
1563 | return -EINVAL; | |
1564 | if (copy_from_user(&fd, data, len)) | |
1565 | return -EFAULT; | |
1566 | ||
113214be | 1567 | new = bpf_prog_get_type(fd, BPF_PROG_TYPE_SOCKET_FILTER); |
f2e52095 WB |
1568 | if (IS_ERR(new)) |
1569 | return PTR_ERR(new); | |
f2e52095 WB |
1570 | |
1571 | __fanout_set_data_bpf(po->fanout, new); | |
1572 | return 0; | |
1573 | } | |
1574 | ||
47dceb8e WB |
1575 | static int fanout_set_data(struct packet_sock *po, char __user *data, |
1576 | unsigned int len) | |
1577 | { | |
1578 | switch (po->fanout->type) { | |
1579 | case PACKET_FANOUT_CBPF: | |
1580 | return fanout_set_data_cbpf(po, data, len); | |
f2e52095 WB |
1581 | case PACKET_FANOUT_EBPF: |
1582 | return fanout_set_data_ebpf(po, data, len); | |
47dceb8e WB |
1583 | default: |
1584 | return -EINVAL; | |
07d53ae4 | 1585 | } |
47dceb8e WB |
1586 | } |
1587 | ||
1588 | static void fanout_release_data(struct packet_fanout *f) | |
1589 | { | |
1590 | switch (f->type) { | |
1591 | case PACKET_FANOUT_CBPF: | |
f2e52095 | 1592 | case PACKET_FANOUT_EBPF: |
47dceb8e | 1593 | __fanout_set_data_bpf(f, NULL); |
07d53ae4 | 1594 | } |
47dceb8e WB |
1595 | } |
1596 | ||
4a69a864 MM |
1597 | static bool __fanout_id_is_free(struct sock *sk, u16 candidate_id) |
1598 | { | |
1599 | struct packet_fanout *f; | |
1600 | ||
1601 | list_for_each_entry(f, &fanout_list, list) { | |
1602 | if (f->id == candidate_id && | |
1603 | read_pnet(&f->net) == sock_net(sk)) { | |
1604 | return false; | |
1605 | } | |
1606 | } | |
1607 | return true; | |
1608 | } | |
1609 | ||
1610 | static bool fanout_find_new_id(struct sock *sk, u16 *new_id) | |
1611 | { | |
1612 | u16 id = fanout_next_id; | |
1613 | ||
1614 | do { | |
1615 | if (__fanout_id_is_free(sk, id)) { | |
1616 | *new_id = id; | |
1617 | fanout_next_id = id + 1; | |
1618 | return true; | |
1619 | } | |
1620 | ||
1621 | id++; | |
1622 | } while (id != fanout_next_id); | |
1623 | ||
1624 | return false; | |
1625 | } | |
1626 | ||
7736d33f | 1627 | static int fanout_add(struct sock *sk, u16 id, u16 type_flags) |
dc99f600 | 1628 | { |
d199fab6 | 1629 | struct packet_rollover *rollover = NULL; |
dc99f600 DM |
1630 | struct packet_sock *po = pkt_sk(sk); |
1631 | struct packet_fanout *f, *match; | |
7736d33f | 1632 | u8 type = type_flags & 0xff; |
77f65ebd | 1633 | u8 flags = type_flags >> 8; |
dc99f600 DM |
1634 | int err; |
1635 | ||
1636 | switch (type) { | |
77f65ebd WB |
1637 | case PACKET_FANOUT_ROLLOVER: |
1638 | if (type_flags & PACKET_FANOUT_FLAG_ROLLOVER) | |
1639 | return -EINVAL; | |
dc99f600 DM |
1640 | case PACKET_FANOUT_HASH: |
1641 | case PACKET_FANOUT_LB: | |
95ec3eb4 | 1642 | case PACKET_FANOUT_CPU: |
5df0ddfb | 1643 | case PACKET_FANOUT_RND: |
2d36097d | 1644 | case PACKET_FANOUT_QM: |
47dceb8e | 1645 | case PACKET_FANOUT_CBPF: |
f2e52095 | 1646 | case PACKET_FANOUT_EBPF: |
dc99f600 DM |
1647 | break; |
1648 | default: | |
1649 | return -EINVAL; | |
1650 | } | |
1651 | ||
d199fab6 ED |
1652 | mutex_lock(&fanout_mutex); |
1653 | ||
d199fab6 | 1654 | err = -EALREADY; |
dc99f600 | 1655 | if (po->fanout) |
d199fab6 | 1656 | goto out; |
dc99f600 | 1657 | |
4633c9e0 WB |
1658 | if (type == PACKET_FANOUT_ROLLOVER || |
1659 | (type_flags & PACKET_FANOUT_FLAG_ROLLOVER)) { | |
d199fab6 ED |
1660 | err = -ENOMEM; |
1661 | rollover = kzalloc(sizeof(*rollover), GFP_KERNEL); | |
1662 | if (!rollover) | |
1663 | goto out; | |
1664 | atomic_long_set(&rollover->num, 0); | |
1665 | atomic_long_set(&rollover->num_huge, 0); | |
1666 | atomic_long_set(&rollover->num_failed, 0); | |
0648ab70 WB |
1667 | } |
1668 | ||
4a69a864 MM |
1669 | if (type_flags & PACKET_FANOUT_FLAG_UNIQUEID) { |
1670 | if (id != 0) { | |
1671 | err = -EINVAL; | |
1672 | goto out; | |
1673 | } | |
1674 | if (!fanout_find_new_id(sk, &id)) { | |
1675 | err = -ENOMEM; | |
1676 | goto out; | |
1677 | } | |
1678 | /* ephemeral flag for the first socket in the group: drop it */ | |
1679 | flags &= ~(PACKET_FANOUT_FLAG_UNIQUEID >> 8); | |
1680 | } | |
1681 | ||
dc99f600 DM |
1682 | match = NULL; |
1683 | list_for_each_entry(f, &fanout_list, list) { | |
1684 | if (f->id == id && | |
1685 | read_pnet(&f->net) == sock_net(sk)) { | |
1686 | match = f; | |
1687 | break; | |
1688 | } | |
1689 | } | |
afe62c68 | 1690 | err = -EINVAL; |
77f65ebd | 1691 | if (match && match->flags != flags) |
afe62c68 | 1692 | goto out; |
dc99f600 | 1693 | if (!match) { |
afe62c68 | 1694 | err = -ENOMEM; |
dc99f600 | 1695 | match = kzalloc(sizeof(*match), GFP_KERNEL); |
afe62c68 ED |
1696 | if (!match) |
1697 | goto out; | |
1698 | write_pnet(&match->net, sock_net(sk)); | |
1699 | match->id = id; | |
1700 | match->type = type; | |
77f65ebd | 1701 | match->flags = flags; |
afe62c68 ED |
1702 | INIT_LIST_HEAD(&match->list); |
1703 | spin_lock_init(&match->lock); | |
fb5c2c17 | 1704 | refcount_set(&match->sk_ref, 0); |
47dceb8e | 1705 | fanout_init_data(match); |
afe62c68 ED |
1706 | match->prot_hook.type = po->prot_hook.type; |
1707 | match->prot_hook.dev = po->prot_hook.dev; | |
1708 | match->prot_hook.func = packet_rcv_fanout; | |
1709 | match->prot_hook.af_packet_priv = match; | |
c0de08d0 | 1710 | match->prot_hook.id_match = match_fanout_group; |
afe62c68 | 1711 | list_add(&match->list, &fanout_list); |
dc99f600 | 1712 | } |
afe62c68 | 1713 | err = -EINVAL; |
008ba2a1 WB |
1714 | |
1715 | spin_lock(&po->bind_lock); | |
1716 | if (po->running && | |
1717 | match->type == type && | |
afe62c68 ED |
1718 | match->prot_hook.type == po->prot_hook.type && |
1719 | match->prot_hook.dev == po->prot_hook.dev) { | |
1720 | err = -ENOSPC; | |
fb5c2c17 | 1721 | if (refcount_read(&match->sk_ref) < PACKET_FANOUT_MAX) { |
afe62c68 ED |
1722 | __dev_remove_pack(&po->prot_hook); |
1723 | po->fanout = match; | |
57f015f5 MM |
1724 | po->rollover = rollover; |
1725 | rollover = NULL; | |
fb5c2c17 | 1726 | refcount_set(&match->sk_ref, refcount_read(&match->sk_ref) + 1); |
afe62c68 ED |
1727 | __fanout_link(sk, po); |
1728 | err = 0; | |
dc99f600 DM |
1729 | } |
1730 | } | |
008ba2a1 WB |
1731 | spin_unlock(&po->bind_lock); |
1732 | ||
1733 | if (err && !refcount_read(&match->sk_ref)) { | |
1734 | list_del(&match->list); | |
1735 | kfree(match); | |
1736 | } | |
1737 | ||
afe62c68 | 1738 | out: |
57f015f5 | 1739 | kfree(rollover); |
d199fab6 | 1740 | mutex_unlock(&fanout_mutex); |
dc99f600 DM |
1741 | return err; |
1742 | } | |
1743 | ||
2bd624b4 AS |
1744 | /* If pkt_sk(sk)->fanout->sk_ref is zero, this function removes |
1745 | * pkt_sk(sk)->fanout from fanout_list and returns pkt_sk(sk)->fanout. | |
1746 | * It is the responsibility of the caller to call fanout_release_data() and | |
1747 | * free the returned packet_fanout (after synchronize_net()) | |
1748 | */ | |
1749 | static struct packet_fanout *fanout_release(struct sock *sk) | |
dc99f600 DM |
1750 | { |
1751 | struct packet_sock *po = pkt_sk(sk); | |
1752 | struct packet_fanout *f; | |
1753 | ||
fff3321d | 1754 | mutex_lock(&fanout_mutex); |
d199fab6 ED |
1755 | f = po->fanout; |
1756 | if (f) { | |
1757 | po->fanout = NULL; | |
1758 | ||
fb5c2c17 | 1759 | if (refcount_dec_and_test(&f->sk_ref)) |
d199fab6 | 1760 | list_del(&f->list); |
2bd624b4 AS |
1761 | else |
1762 | f = NULL; | |
dc99f600 DM |
1763 | } |
1764 | mutex_unlock(&fanout_mutex); | |
2bd624b4 AS |
1765 | |
1766 | return f; | |
dc99f600 | 1767 | } |
1da177e4 | 1768 | |
3c70c132 DB |
1769 | static bool packet_extra_vlan_len_allowed(const struct net_device *dev, |
1770 | struct sk_buff *skb) | |
1771 | { | |
1772 | /* Earlier code assumed this would be a VLAN pkt, double-check | |
1773 | * this now that we have the actual packet in hand. We can only | |
1774 | * do this check on Ethernet devices. | |
1775 | */ | |
1776 | if (unlikely(dev->type != ARPHRD_ETHER)) | |
1777 | return false; | |
1778 | ||
1779 | skb_reset_mac_header(skb); | |
1780 | return likely(eth_hdr(skb)->h_proto == htons(ETH_P_8021Q)); | |
1781 | } | |
1782 | ||
90ddc4f0 | 1783 | static const struct proto_ops packet_ops; |
1da177e4 | 1784 | |
90ddc4f0 | 1785 | static const struct proto_ops packet_ops_spkt; |
1da177e4 | 1786 | |
40d4e3df ED |
1787 | static int packet_rcv_spkt(struct sk_buff *skb, struct net_device *dev, |
1788 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
1789 | { |
1790 | struct sock *sk; | |
1791 | struct sockaddr_pkt *spkt; | |
1792 | ||
1793 | /* | |
1794 | * When we registered the protocol we saved the socket in the data | |
1795 | * field for just this event. | |
1796 | */ | |
1797 | ||
1798 | sk = pt->af_packet_priv; | |
1ce4f28b | 1799 | |
1da177e4 LT |
1800 | /* |
1801 | * Yank back the headers [hope the device set this | |
1802 | * right or kerboom...] | |
1803 | * | |
1804 | * Incoming packets have ll header pulled, | |
1805 | * push it back. | |
1806 | * | |
98e399f8 | 1807 | * For outgoing ones skb->data == skb_mac_header(skb) |
1da177e4 LT |
1808 | * so that this procedure is noop. |
1809 | */ | |
1810 | ||
1811 | if (skb->pkt_type == PACKET_LOOPBACK) | |
1812 | goto out; | |
1813 | ||
09ad9bc7 | 1814 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
1815 | goto out; |
1816 | ||
40d4e3df ED |
1817 | skb = skb_share_check(skb, GFP_ATOMIC); |
1818 | if (skb == NULL) | |
1da177e4 LT |
1819 | goto oom; |
1820 | ||
1821 | /* drop any routing info */ | |
adf30907 | 1822 | skb_dst_drop(skb); |
1da177e4 | 1823 | |
84531c24 PO |
1824 | /* drop conntrack reference */ |
1825 | nf_reset(skb); | |
1826 | ||
ffbc6111 | 1827 | spkt = &PACKET_SKB_CB(skb)->sa.pkt; |
1da177e4 | 1828 | |
98e399f8 | 1829 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
1830 | |
1831 | /* | |
1832 | * The SOCK_PACKET socket receives _all_ frames. | |
1833 | */ | |
1834 | ||
1835 | spkt->spkt_family = dev->type; | |
1836 | strlcpy(spkt->spkt_device, dev->name, sizeof(spkt->spkt_device)); | |
1837 | spkt->spkt_protocol = skb->protocol; | |
1838 | ||
1839 | /* | |
1840 | * Charge the memory to the socket. This is done specifically | |
1841 | * to prevent sockets using all the memory up. | |
1842 | */ | |
1843 | ||
40d4e3df | 1844 | if (sock_queue_rcv_skb(sk, skb) == 0) |
1da177e4 LT |
1845 | return 0; |
1846 | ||
1847 | out: | |
1848 | kfree_skb(skb); | |
1849 | oom: | |
1850 | return 0; | |
1851 | } | |
1852 | ||
1853 | ||
1854 | /* | |
1855 | * Output a raw packet to a device layer. This bypasses all the other | |
1856 | * protocol layers and you must therefore supply it with a complete frame | |
1857 | */ | |
1ce4f28b | 1858 | |
1b784140 YX |
1859 | static int packet_sendmsg_spkt(struct socket *sock, struct msghdr *msg, |
1860 | size_t len) | |
1da177e4 LT |
1861 | { |
1862 | struct sock *sk = sock->sk; | |
342dfc30 | 1863 | DECLARE_SOCKADDR(struct sockaddr_pkt *, saddr, msg->msg_name); |
1a35ca80 | 1864 | struct sk_buff *skb = NULL; |
1da177e4 | 1865 | struct net_device *dev; |
c14ac945 | 1866 | struct sockcm_cookie sockc; |
40d4e3df | 1867 | __be16 proto = 0; |
1da177e4 | 1868 | int err; |
3bdc0eba | 1869 | int extra_len = 0; |
1ce4f28b | 1870 | |
1da177e4 | 1871 | /* |
1ce4f28b | 1872 | * Get and verify the address. |
1da177e4 LT |
1873 | */ |
1874 | ||
40d4e3df | 1875 | if (saddr) { |
1da177e4 | 1876 | if (msg->msg_namelen < sizeof(struct sockaddr)) |
40d4e3df ED |
1877 | return -EINVAL; |
1878 | if (msg->msg_namelen == sizeof(struct sockaddr_pkt)) | |
1879 | proto = saddr->spkt_protocol; | |
1880 | } else | |
1881 | return -ENOTCONN; /* SOCK_PACKET must be sent giving an address */ | |
1da177e4 LT |
1882 | |
1883 | /* | |
1ce4f28b | 1884 | * Find the device first to size check it |
1da177e4 LT |
1885 | */ |
1886 | ||
de74e92a | 1887 | saddr->spkt_device[sizeof(saddr->spkt_device) - 1] = 0; |
1a35ca80 | 1888 | retry: |
654d1f8a ED |
1889 | rcu_read_lock(); |
1890 | dev = dev_get_by_name_rcu(sock_net(sk), saddr->spkt_device); | |
1da177e4 LT |
1891 | err = -ENODEV; |
1892 | if (dev == NULL) | |
1893 | goto out_unlock; | |
1ce4f28b | 1894 | |
d5e76b0a DM |
1895 | err = -ENETDOWN; |
1896 | if (!(dev->flags & IFF_UP)) | |
1897 | goto out_unlock; | |
1898 | ||
1da177e4 | 1899 | /* |
40d4e3df ED |
1900 | * You may not queue a frame bigger than the mtu. This is the lowest level |
1901 | * raw protocol and you must do your own fragmentation at this level. | |
1da177e4 | 1902 | */ |
1ce4f28b | 1903 | |
3bdc0eba BG |
1904 | if (unlikely(sock_flag(sk, SOCK_NOFCS))) { |
1905 | if (!netif_supports_nofcs(dev)) { | |
1906 | err = -EPROTONOSUPPORT; | |
1907 | goto out_unlock; | |
1908 | } | |
1909 | extra_len = 4; /* We're doing our own CRC */ | |
1910 | } | |
1911 | ||
1da177e4 | 1912 | err = -EMSGSIZE; |
3bdc0eba | 1913 | if (len > dev->mtu + dev->hard_header_len + VLAN_HLEN + extra_len) |
1da177e4 LT |
1914 | goto out_unlock; |
1915 | ||
1a35ca80 ED |
1916 | if (!skb) { |
1917 | size_t reserved = LL_RESERVED_SPACE(dev); | |
4ce40912 | 1918 | int tlen = dev->needed_tailroom; |
1a35ca80 ED |
1919 | unsigned int hhlen = dev->header_ops ? dev->hard_header_len : 0; |
1920 | ||
1921 | rcu_read_unlock(); | |
4ce40912 | 1922 | skb = sock_wmalloc(sk, len + reserved + tlen, 0, GFP_KERNEL); |
1a35ca80 ED |
1923 | if (skb == NULL) |
1924 | return -ENOBUFS; | |
1925 | /* FIXME: Save some space for broken drivers that write a hard | |
1926 | * header at transmission time by themselves. PPP is the notable | |
1927 | * one here. This should really be fixed at the driver level. | |
1928 | */ | |
1929 | skb_reserve(skb, reserved); | |
1930 | skb_reset_network_header(skb); | |
1931 | ||
1932 | /* Try to align data part correctly */ | |
1933 | if (hhlen) { | |
1934 | skb->data -= hhlen; | |
1935 | skb->tail -= hhlen; | |
1936 | if (len < hhlen) | |
1937 | skb_reset_network_header(skb); | |
1938 | } | |
6ce8e9ce | 1939 | err = memcpy_from_msg(skb_put(skb, len), msg, len); |
1a35ca80 ED |
1940 | if (err) |
1941 | goto out_free; | |
1942 | goto retry; | |
1da177e4 LT |
1943 | } |
1944 | ||
9ed988cd WB |
1945 | if (!dev_validate_header(dev, skb->data, len)) { |
1946 | err = -EINVAL; | |
1947 | goto out_unlock; | |
1948 | } | |
3c70c132 DB |
1949 | if (len > (dev->mtu + dev->hard_header_len + extra_len) && |
1950 | !packet_extra_vlan_len_allowed(dev, skb)) { | |
1951 | err = -EMSGSIZE; | |
1952 | goto out_unlock; | |
57f89bfa | 1953 | } |
1a35ca80 | 1954 | |
657a0667 | 1955 | sockcm_init(&sockc, sk); |
c14ac945 SHY |
1956 | if (msg->msg_controllen) { |
1957 | err = sock_cmsg_send(sk, msg, &sockc); | |
f8e7718c | 1958 | if (unlikely(err)) |
c14ac945 | 1959 | goto out_unlock; |
c14ac945 SHY |
1960 | } |
1961 | ||
1da177e4 LT |
1962 | skb->protocol = proto; |
1963 | skb->dev = dev; | |
1964 | skb->priority = sk->sk_priority; | |
2d37a186 | 1965 | skb->mark = sk->sk_mark; |
3d0ba8c0 | 1966 | skb->tstamp = sockc.transmit_time; |
bf84a010 | 1967 | |
8f932f76 | 1968 | skb_setup_tx_timestamp(skb, sockc.tsflags); |
1da177e4 | 1969 | |
3bdc0eba BG |
1970 | if (unlikely(extra_len == 4)) |
1971 | skb->no_fcs = 1; | |
1972 | ||
40893fd0 | 1973 | skb_probe_transport_header(skb, 0); |
c1aad275 | 1974 | |
1da177e4 | 1975 | dev_queue_xmit(skb); |
654d1f8a | 1976 | rcu_read_unlock(); |
40d4e3df | 1977 | return len; |
1da177e4 | 1978 | |
1da177e4 | 1979 | out_unlock: |
654d1f8a | 1980 | rcu_read_unlock(); |
1a35ca80 ED |
1981 | out_free: |
1982 | kfree_skb(skb); | |
1da177e4 LT |
1983 | return err; |
1984 | } | |
1da177e4 | 1985 | |
ff936a04 AS |
1986 | static unsigned int run_filter(struct sk_buff *skb, |
1987 | const struct sock *sk, | |
1988 | unsigned int res) | |
1da177e4 LT |
1989 | { |
1990 | struct sk_filter *filter; | |
fda9ef5d | 1991 | |
80f8f102 ED |
1992 | rcu_read_lock(); |
1993 | filter = rcu_dereference(sk->sk_filter); | |
dbcb5855 | 1994 | if (filter != NULL) |
ff936a04 | 1995 | res = bpf_prog_run_clear_cb(filter->prog, skb); |
80f8f102 | 1996 | rcu_read_unlock(); |
1da177e4 | 1997 | |
dbcb5855 | 1998 | return res; |
1da177e4 LT |
1999 | } |
2000 | ||
16cc1400 WB |
2001 | static int packet_rcv_vnet(struct msghdr *msg, const struct sk_buff *skb, |
2002 | size_t *len) | |
2003 | { | |
2004 | struct virtio_net_hdr vnet_hdr; | |
2005 | ||
2006 | if (*len < sizeof(vnet_hdr)) | |
2007 | return -EINVAL; | |
2008 | *len -= sizeof(vnet_hdr); | |
2009 | ||
fd3a8862 | 2010 | if (virtio_net_hdr_from_skb(skb, &vnet_hdr, vio_le(), true, 0)) |
16cc1400 WB |
2011 | return -EINVAL; |
2012 | ||
2013 | return memcpy_to_msg(msg, (void *)&vnet_hdr, sizeof(vnet_hdr)); | |
2014 | } | |
2015 | ||
1da177e4 | 2016 | /* |
62ab0812 ED |
2017 | * This function makes lazy skb cloning in hope that most of packets |
2018 | * are discarded by BPF. | |
2019 | * | |
2020 | * Note tricky part: we DO mangle shared skb! skb->data, skb->len | |
2021 | * and skb->cb are mangled. It works because (and until) packets | |
2022 | * falling here are owned by current CPU. Output packets are cloned | |
2023 | * by dev_queue_xmit_nit(), input packets are processed by net_bh | |
2024 | * sequencially, so that if we return skb to original state on exit, | |
2025 | * we will not harm anyone. | |
1da177e4 LT |
2026 | */ |
2027 | ||
40d4e3df ED |
2028 | static int packet_rcv(struct sk_buff *skb, struct net_device *dev, |
2029 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
2030 | { |
2031 | struct sock *sk; | |
2032 | struct sockaddr_ll *sll; | |
2033 | struct packet_sock *po; | |
40d4e3df | 2034 | u8 *skb_head = skb->data; |
1da177e4 | 2035 | int skb_len = skb->len; |
dbcb5855 | 2036 | unsigned int snaplen, res; |
da37845f | 2037 | bool is_drop_n_account = false; |
1da177e4 LT |
2038 | |
2039 | if (skb->pkt_type == PACKET_LOOPBACK) | |
2040 | goto drop; | |
2041 | ||
2042 | sk = pt->af_packet_priv; | |
2043 | po = pkt_sk(sk); | |
2044 | ||
09ad9bc7 | 2045 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
2046 | goto drop; |
2047 | ||
1da177e4 LT |
2048 | skb->dev = dev; |
2049 | ||
3b04ddde | 2050 | if (dev->header_ops) { |
1da177e4 | 2051 | /* The device has an explicit notion of ll header, |
62ab0812 ED |
2052 | * exported to higher levels. |
2053 | * | |
2054 | * Otherwise, the device hides details of its frame | |
2055 | * structure, so that corresponding packet head is | |
2056 | * never delivered to user. | |
1da177e4 LT |
2057 | */ |
2058 | if (sk->sk_type != SOCK_DGRAM) | |
98e399f8 | 2059 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
2060 | else if (skb->pkt_type == PACKET_OUTGOING) { |
2061 | /* Special case: outgoing packets have ll header at head */ | |
bbe735e4 | 2062 | skb_pull(skb, skb_network_offset(skb)); |
1da177e4 LT |
2063 | } |
2064 | } | |
2065 | ||
2066 | snaplen = skb->len; | |
2067 | ||
dbcb5855 DM |
2068 | res = run_filter(skb, sk, snaplen); |
2069 | if (!res) | |
fda9ef5d | 2070 | goto drop_n_restore; |
dbcb5855 DM |
2071 | if (snaplen > res) |
2072 | snaplen = res; | |
1da177e4 | 2073 | |
0fd7bac6 | 2074 | if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) |
1da177e4 LT |
2075 | goto drop_n_acct; |
2076 | ||
2077 | if (skb_shared(skb)) { | |
2078 | struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC); | |
2079 | if (nskb == NULL) | |
2080 | goto drop_n_acct; | |
2081 | ||
2082 | if (skb_head != skb->data) { | |
2083 | skb->data = skb_head; | |
2084 | skb->len = skb_len; | |
2085 | } | |
abc4e4fa | 2086 | consume_skb(skb); |
1da177e4 LT |
2087 | skb = nskb; |
2088 | } | |
2089 | ||
b4772ef8 | 2090 | sock_skb_cb_check_size(sizeof(*PACKET_SKB_CB(skb)) + MAX_ADDR_LEN - 8); |
ffbc6111 HX |
2091 | |
2092 | sll = &PACKET_SKB_CB(skb)->sa.ll; | |
1da177e4 | 2093 | sll->sll_hatype = dev->type; |
1da177e4 | 2094 | sll->sll_pkttype = skb->pkt_type; |
8032b464 | 2095 | if (unlikely(po->origdev)) |
80feaacb PWJ |
2096 | sll->sll_ifindex = orig_dev->ifindex; |
2097 | else | |
2098 | sll->sll_ifindex = dev->ifindex; | |
1da177e4 | 2099 | |
b95cce35 | 2100 | sll->sll_halen = dev_parse_header(skb, sll->sll_addr); |
1da177e4 | 2101 | |
2472d761 EB |
2102 | /* sll->sll_family and sll->sll_protocol are set in packet_recvmsg(). |
2103 | * Use their space for storing the original skb length. | |
2104 | */ | |
2105 | PACKET_SKB_CB(skb)->sa.origlen = skb->len; | |
8dc41944 | 2106 | |
1da177e4 LT |
2107 | if (pskb_trim(skb, snaplen)) |
2108 | goto drop_n_acct; | |
2109 | ||
2110 | skb_set_owner_r(skb, sk); | |
2111 | skb->dev = NULL; | |
adf30907 | 2112 | skb_dst_drop(skb); |
1da177e4 | 2113 | |
84531c24 PO |
2114 | /* drop conntrack reference */ |
2115 | nf_reset(skb); | |
2116 | ||
1da177e4 | 2117 | spin_lock(&sk->sk_receive_queue.lock); |
ee80fbf3 | 2118 | po->stats.stats1.tp_packets++; |
3bc3b96f | 2119 | sock_skb_set_dropcount(sk, skb); |
1da177e4 LT |
2120 | __skb_queue_tail(&sk->sk_receive_queue, skb); |
2121 | spin_unlock(&sk->sk_receive_queue.lock); | |
676d2369 | 2122 | sk->sk_data_ready(sk); |
1da177e4 LT |
2123 | return 0; |
2124 | ||
2125 | drop_n_acct: | |
da37845f | 2126 | is_drop_n_account = true; |
7091fbd8 | 2127 | spin_lock(&sk->sk_receive_queue.lock); |
ee80fbf3 | 2128 | po->stats.stats1.tp_drops++; |
7091fbd8 WB |
2129 | atomic_inc(&sk->sk_drops); |
2130 | spin_unlock(&sk->sk_receive_queue.lock); | |
1da177e4 LT |
2131 | |
2132 | drop_n_restore: | |
2133 | if (skb_head != skb->data && skb_shared(skb)) { | |
2134 | skb->data = skb_head; | |
2135 | skb->len = skb_len; | |
2136 | } | |
2137 | drop: | |
da37845f WJ |
2138 | if (!is_drop_n_account) |
2139 | consume_skb(skb); | |
2140 | else | |
2141 | kfree_skb(skb); | |
1da177e4 LT |
2142 | return 0; |
2143 | } | |
2144 | ||
40d4e3df ED |
2145 | static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, |
2146 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
2147 | { |
2148 | struct sock *sk; | |
2149 | struct packet_sock *po; | |
2150 | struct sockaddr_ll *sll; | |
184f489e | 2151 | union tpacket_uhdr h; |
40d4e3df | 2152 | u8 *skb_head = skb->data; |
1da177e4 | 2153 | int skb_len = skb->len; |
dbcb5855 | 2154 | unsigned int snaplen, res; |
f6fb8f10 | 2155 | unsigned long status = TP_STATUS_USER; |
bbd6ef87 | 2156 | unsigned short macoff, netoff, hdrlen; |
1da177e4 | 2157 | struct sk_buff *copy_skb = NULL; |
bbd6ef87 | 2158 | struct timespec ts; |
b9c32fb2 | 2159 | __u32 ts_status; |
da37845f | 2160 | bool is_drop_n_account = false; |
edbd58be | 2161 | bool do_vnet = false; |
1da177e4 | 2162 | |
51846355 AW |
2163 | /* struct tpacket{2,3}_hdr is aligned to a multiple of TPACKET_ALIGNMENT. |
2164 | * We may add members to them until current aligned size without forcing | |
2165 | * userspace to call getsockopt(..., PACKET_HDRLEN, ...). | |
2166 | */ | |
2167 | BUILD_BUG_ON(TPACKET_ALIGN(sizeof(*h.h2)) != 32); | |
2168 | BUILD_BUG_ON(TPACKET_ALIGN(sizeof(*h.h3)) != 48); | |
2169 | ||
1da177e4 LT |
2170 | if (skb->pkt_type == PACKET_LOOPBACK) |
2171 | goto drop; | |
2172 | ||
2173 | sk = pt->af_packet_priv; | |
2174 | po = pkt_sk(sk); | |
2175 | ||
09ad9bc7 | 2176 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
2177 | goto drop; |
2178 | ||
3b04ddde | 2179 | if (dev->header_ops) { |
1da177e4 | 2180 | if (sk->sk_type != SOCK_DGRAM) |
98e399f8 | 2181 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
2182 | else if (skb->pkt_type == PACKET_OUTGOING) { |
2183 | /* Special case: outgoing packets have ll header at head */ | |
bbe735e4 | 2184 | skb_pull(skb, skb_network_offset(skb)); |
1da177e4 LT |
2185 | } |
2186 | } | |
2187 | ||
2188 | snaplen = skb->len; | |
2189 | ||
dbcb5855 DM |
2190 | res = run_filter(skb, sk, snaplen); |
2191 | if (!res) | |
fda9ef5d | 2192 | goto drop_n_restore; |
68c2e5de AD |
2193 | |
2194 | if (skb->ip_summed == CHECKSUM_PARTIAL) | |
2195 | status |= TP_STATUS_CSUMNOTREADY; | |
682f048b AD |
2196 | else if (skb->pkt_type != PACKET_OUTGOING && |
2197 | (skb->ip_summed == CHECKSUM_COMPLETE || | |
2198 | skb_csum_unnecessary(skb))) | |
2199 | status |= TP_STATUS_CSUM_VALID; | |
68c2e5de | 2200 | |
dbcb5855 DM |
2201 | if (snaplen > res) |
2202 | snaplen = res; | |
1da177e4 LT |
2203 | |
2204 | if (sk->sk_type == SOCK_DGRAM) { | |
8913336a PM |
2205 | macoff = netoff = TPACKET_ALIGN(po->tp_hdrlen) + 16 + |
2206 | po->tp_reserve; | |
1da177e4 | 2207 | } else { |
95c96174 | 2208 | unsigned int maclen = skb_network_offset(skb); |
bbd6ef87 | 2209 | netoff = TPACKET_ALIGN(po->tp_hdrlen + |
8913336a | 2210 | (maclen < 16 ? 16 : maclen)) + |
58d19b19 | 2211 | po->tp_reserve; |
edbd58be | 2212 | if (po->has_vnet_hdr) { |
58d19b19 | 2213 | netoff += sizeof(struct virtio_net_hdr); |
edbd58be BP |
2214 | do_vnet = true; |
2215 | } | |
1da177e4 LT |
2216 | macoff = netoff - maclen; |
2217 | } | |
f6fb8f10 | 2218 | if (po->tp_version <= TPACKET_V2) { |
2219 | if (macoff + snaplen > po->rx_ring.frame_size) { | |
2220 | if (po->copy_thresh && | |
0fd7bac6 | 2221 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) { |
f6fb8f10 | 2222 | if (skb_shared(skb)) { |
2223 | copy_skb = skb_clone(skb, GFP_ATOMIC); | |
2224 | } else { | |
2225 | copy_skb = skb_get(skb); | |
2226 | skb_head = skb->data; | |
2227 | } | |
2228 | if (copy_skb) | |
2229 | skb_set_owner_r(copy_skb, sk); | |
1da177e4 | 2230 | } |
f6fb8f10 | 2231 | snaplen = po->rx_ring.frame_size - macoff; |
edbd58be | 2232 | if ((int)snaplen < 0) { |
f6fb8f10 | 2233 | snaplen = 0; |
edbd58be BP |
2234 | do_vnet = false; |
2235 | } | |
1da177e4 | 2236 | } |
dc808110 ED |
2237 | } else if (unlikely(macoff + snaplen > |
2238 | GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len)) { | |
2239 | u32 nval; | |
2240 | ||
2241 | nval = GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len - macoff; | |
2242 | pr_err_once("tpacket_rcv: packet too big, clamped from %u to %u. macoff=%u\n", | |
2243 | snaplen, nval, macoff); | |
2244 | snaplen = nval; | |
2245 | if (unlikely((int)snaplen < 0)) { | |
2246 | snaplen = 0; | |
2247 | macoff = GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len; | |
edbd58be | 2248 | do_vnet = false; |
dc808110 | 2249 | } |
1da177e4 | 2250 | } |
1da177e4 | 2251 | spin_lock(&sk->sk_receive_queue.lock); |
f6fb8f10 | 2252 | h.raw = packet_current_rx_frame(po, skb, |
2253 | TP_STATUS_KERNEL, (macoff+snaplen)); | |
bbd6ef87 | 2254 | if (!h.raw) |
58d19b19 | 2255 | goto drop_n_account; |
f6fb8f10 | 2256 | if (po->tp_version <= TPACKET_V2) { |
2257 | packet_increment_rx_head(po, &po->rx_ring); | |
2258 | /* | |
2259 | * LOSING will be reported till you read the stats, | |
2260 | * because it's COR - Clear On Read. | |
2261 | * Anyways, moving it for V1/V2 only as V3 doesn't need this | |
2262 | * at packet level. | |
2263 | */ | |
ee80fbf3 | 2264 | if (po->stats.stats1.tp_drops) |
f6fb8f10 | 2265 | status |= TP_STATUS_LOSING; |
2266 | } | |
945d015e ED |
2267 | |
2268 | if (do_vnet && | |
2269 | virtio_net_hdr_from_skb(skb, h.raw + macoff - | |
2270 | sizeof(struct virtio_net_hdr), | |
2271 | vio_le(), true, 0)) | |
2272 | goto drop_n_account; | |
2273 | ||
ee80fbf3 | 2274 | po->stats.stats1.tp_packets++; |
1da177e4 LT |
2275 | if (copy_skb) { |
2276 | status |= TP_STATUS_COPY; | |
2277 | __skb_queue_tail(&sk->sk_receive_queue, copy_skb); | |
2278 | } | |
1da177e4 LT |
2279 | spin_unlock(&sk->sk_receive_queue.lock); |
2280 | ||
bbd6ef87 | 2281 | skb_copy_bits(skb, 0, h.raw + macoff, snaplen); |
b9c32fb2 DB |
2282 | |
2283 | if (!(ts_status = tpacket_get_timestamp(skb, &ts, po->tp_tstamp))) | |
7a51384c | 2284 | getnstimeofday(&ts); |
1da177e4 | 2285 | |
b9c32fb2 DB |
2286 | status |= ts_status; |
2287 | ||
bbd6ef87 PM |
2288 | switch (po->tp_version) { |
2289 | case TPACKET_V1: | |
2290 | h.h1->tp_len = skb->len; | |
2291 | h.h1->tp_snaplen = snaplen; | |
2292 | h.h1->tp_mac = macoff; | |
2293 | h.h1->tp_net = netoff; | |
4b457bdf DB |
2294 | h.h1->tp_sec = ts.tv_sec; |
2295 | h.h1->tp_usec = ts.tv_nsec / NSEC_PER_USEC; | |
bbd6ef87 PM |
2296 | hdrlen = sizeof(*h.h1); |
2297 | break; | |
2298 | case TPACKET_V2: | |
2299 | h.h2->tp_len = skb->len; | |
2300 | h.h2->tp_snaplen = snaplen; | |
2301 | h.h2->tp_mac = macoff; | |
2302 | h.h2->tp_net = netoff; | |
bbd6ef87 PM |
2303 | h.h2->tp_sec = ts.tv_sec; |
2304 | h.h2->tp_nsec = ts.tv_nsec; | |
df8a39de JP |
2305 | if (skb_vlan_tag_present(skb)) { |
2306 | h.h2->tp_vlan_tci = skb_vlan_tag_get(skb); | |
a0cdfcf3 AW |
2307 | h.h2->tp_vlan_tpid = ntohs(skb->vlan_proto); |
2308 | status |= TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
a3bcc23e BG |
2309 | } else { |
2310 | h.h2->tp_vlan_tci = 0; | |
a0cdfcf3 | 2311 | h.h2->tp_vlan_tpid = 0; |
a3bcc23e | 2312 | } |
e4d26f4b | 2313 | memset(h.h2->tp_padding, 0, sizeof(h.h2->tp_padding)); |
bbd6ef87 PM |
2314 | hdrlen = sizeof(*h.h2); |
2315 | break; | |
f6fb8f10 | 2316 | case TPACKET_V3: |
2317 | /* tp_nxt_offset,vlan are already populated above. | |
2318 | * So DONT clear those fields here | |
2319 | */ | |
2320 | h.h3->tp_status |= status; | |
2321 | h.h3->tp_len = skb->len; | |
2322 | h.h3->tp_snaplen = snaplen; | |
2323 | h.h3->tp_mac = macoff; | |
2324 | h.h3->tp_net = netoff; | |
f6fb8f10 | 2325 | h.h3->tp_sec = ts.tv_sec; |
2326 | h.h3->tp_nsec = ts.tv_nsec; | |
e4d26f4b | 2327 | memset(h.h3->tp_padding, 0, sizeof(h.h3->tp_padding)); |
f6fb8f10 | 2328 | hdrlen = sizeof(*h.h3); |
2329 | break; | |
bbd6ef87 PM |
2330 | default: |
2331 | BUG(); | |
2332 | } | |
1da177e4 | 2333 | |
bbd6ef87 | 2334 | sll = h.raw + TPACKET_ALIGN(hdrlen); |
b95cce35 | 2335 | sll->sll_halen = dev_parse_header(skb, sll->sll_addr); |
1da177e4 LT |
2336 | sll->sll_family = AF_PACKET; |
2337 | sll->sll_hatype = dev->type; | |
2338 | sll->sll_protocol = skb->protocol; | |
2339 | sll->sll_pkttype = skb->pkt_type; | |
8032b464 | 2340 | if (unlikely(po->origdev)) |
80feaacb PWJ |
2341 | sll->sll_ifindex = orig_dev->ifindex; |
2342 | else | |
2343 | sll->sll_ifindex = dev->ifindex; | |
1da177e4 | 2344 | |
e16aa207 | 2345 | smp_mb(); |
f0d4eb29 | 2346 | |
f6dafa95 | 2347 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 |
f0d4eb29 | 2348 | if (po->tp_version <= TPACKET_V2) { |
0af55bb5 CG |
2349 | u8 *start, *end; |
2350 | ||
f0d4eb29 DB |
2351 | end = (u8 *) PAGE_ALIGN((unsigned long) h.raw + |
2352 | macoff + snaplen); | |
2353 | ||
2354 | for (start = h.raw; start < end; start += PAGE_SIZE) | |
2355 | flush_dcache_page(pgv_to_page(start)); | |
1da177e4 | 2356 | } |
f0d4eb29 | 2357 | smp_wmb(); |
f6dafa95 | 2358 | #endif |
f0d4eb29 | 2359 | |
da413eec | 2360 | if (po->tp_version <= TPACKET_V2) { |
f6fb8f10 | 2361 | __packet_set_status(po, h.raw, status); |
da413eec DC |
2362 | sk->sk_data_ready(sk); |
2363 | } else { | |
f6fb8f10 | 2364 | prb_clear_blk_fill_status(&po->rx_ring); |
da413eec | 2365 | } |
1da177e4 LT |
2366 | |
2367 | drop_n_restore: | |
2368 | if (skb_head != skb->data && skb_shared(skb)) { | |
2369 | skb->data = skb_head; | |
2370 | skb->len = skb_len; | |
2371 | } | |
2372 | drop: | |
da37845f WJ |
2373 | if (!is_drop_n_account) |
2374 | consume_skb(skb); | |
2375 | else | |
2376 | kfree_skb(skb); | |
1da177e4 LT |
2377 | return 0; |
2378 | ||
58d19b19 | 2379 | drop_n_account: |
da37845f | 2380 | is_drop_n_account = true; |
ee80fbf3 | 2381 | po->stats.stats1.tp_drops++; |
1da177e4 LT |
2382 | spin_unlock(&sk->sk_receive_queue.lock); |
2383 | ||
676d2369 | 2384 | sk->sk_data_ready(sk); |
acb5d75b | 2385 | kfree_skb(copy_skb); |
1da177e4 LT |
2386 | goto drop_n_restore; |
2387 | } | |
2388 | ||
69e3c75f JB |
2389 | static void tpacket_destruct_skb(struct sk_buff *skb) |
2390 | { | |
2391 | struct packet_sock *po = pkt_sk(skb->sk); | |
1da177e4 | 2392 | |
69e3c75f | 2393 | if (likely(po->tx_ring.pg_vec)) { |
f0d4eb29 | 2394 | void *ph; |
b9c32fb2 DB |
2395 | __u32 ts; |
2396 | ||
5cd8d46e | 2397 | ph = skb_zcopy_get_nouarg(skb); |
b0138408 | 2398 | packet_dec_pending(&po->tx_ring); |
b9c32fb2 DB |
2399 | |
2400 | ts = __packet_set_timestamp(po, ph, skb); | |
2401 | __packet_set_status(po, ph, TP_STATUS_AVAILABLE | ts); | |
69e3c75f JB |
2402 | } |
2403 | ||
2404 | sock_wfree(skb); | |
2405 | } | |
2406 | ||
c72219b7 DB |
2407 | static void tpacket_set_protocol(const struct net_device *dev, |
2408 | struct sk_buff *skb) | |
2409 | { | |
2410 | if (dev->type == ARPHRD_ETHER) { | |
2411 | skb_reset_mac_header(skb); | |
2412 | skb->protocol = eth_hdr(skb)->h_proto; | |
2413 | } | |
2414 | } | |
2415 | ||
16cc1400 WB |
2416 | static int __packet_snd_vnet_parse(struct virtio_net_hdr *vnet_hdr, size_t len) |
2417 | { | |
16cc1400 WB |
2418 | if ((vnet_hdr->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
2419 | (__virtio16_to_cpu(vio_le(), vnet_hdr->csum_start) + | |
2420 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_offset) + 2 > | |
2421 | __virtio16_to_cpu(vio_le(), vnet_hdr->hdr_len))) | |
2422 | vnet_hdr->hdr_len = __cpu_to_virtio16(vio_le(), | |
2423 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_start) + | |
2424 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_offset) + 2); | |
2425 | ||
2426 | if (__virtio16_to_cpu(vio_le(), vnet_hdr->hdr_len) > len) | |
2427 | return -EINVAL; | |
2428 | ||
16cc1400 WB |
2429 | return 0; |
2430 | } | |
2431 | ||
2432 | static int packet_snd_vnet_parse(struct msghdr *msg, size_t *len, | |
2433 | struct virtio_net_hdr *vnet_hdr) | |
2434 | { | |
16cc1400 WB |
2435 | if (*len < sizeof(*vnet_hdr)) |
2436 | return -EINVAL; | |
2437 | *len -= sizeof(*vnet_hdr); | |
2438 | ||
cbbd26b8 | 2439 | if (!copy_from_iter_full(vnet_hdr, sizeof(*vnet_hdr), &msg->msg_iter)) |
16cc1400 WB |
2440 | return -EFAULT; |
2441 | ||
2442 | return __packet_snd_vnet_parse(vnet_hdr, *len); | |
2443 | } | |
2444 | ||
40d4e3df | 2445 | static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, |
8d39b4a6 | 2446 | void *frame, struct net_device *dev, void *data, int tp_len, |
c14ac945 SHY |
2447 | __be16 proto, unsigned char *addr, int hlen, int copylen, |
2448 | const struct sockcm_cookie *sockc) | |
69e3c75f | 2449 | { |
184f489e | 2450 | union tpacket_uhdr ph; |
8d39b4a6 | 2451 | int to_write, offset, len, nr_frags, len_max; |
69e3c75f JB |
2452 | struct socket *sock = po->sk.sk_socket; |
2453 | struct page *page; | |
69e3c75f JB |
2454 | int err; |
2455 | ||
2456 | ph.raw = frame; | |
2457 | ||
2458 | skb->protocol = proto; | |
2459 | skb->dev = dev; | |
2460 | skb->priority = po->sk.sk_priority; | |
2d37a186 | 2461 | skb->mark = po->sk.sk_mark; |
3d0ba8c0 | 2462 | skb->tstamp = sockc->transmit_time; |
8f932f76 | 2463 | skb_setup_tx_timestamp(skb, sockc->tsflags); |
5cd8d46e | 2464 | skb_zcopy_set_nouarg(skb, ph.raw); |
69e3c75f | 2465 | |
ae641949 | 2466 | skb_reserve(skb, hlen); |
69e3c75f | 2467 | skb_reset_network_header(skb); |
c1aad275 | 2468 | |
69e3c75f JB |
2469 | to_write = tp_len; |
2470 | ||
2471 | if (sock->type == SOCK_DGRAM) { | |
2472 | err = dev_hard_header(skb, dev, ntohs(proto), addr, | |
2473 | NULL, tp_len); | |
2474 | if (unlikely(err < 0)) | |
2475 | return -EINVAL; | |
1d036d25 | 2476 | } else if (copylen) { |
9ed988cd WB |
2477 | int hdrlen = min_t(int, copylen, tp_len); |
2478 | ||
69e3c75f | 2479 | skb_push(skb, dev->hard_header_len); |
1d036d25 | 2480 | skb_put(skb, copylen - dev->hard_header_len); |
9ed988cd | 2481 | err = skb_store_bits(skb, 0, data, hdrlen); |
69e3c75f JB |
2482 | if (unlikely(err)) |
2483 | return err; | |
9ed988cd WB |
2484 | if (!dev_validate_header(dev, skb->data, hdrlen)) |
2485 | return -EINVAL; | |
c72219b7 DB |
2486 | if (!skb->protocol) |
2487 | tpacket_set_protocol(dev, skb); | |
69e3c75f | 2488 | |
9ed988cd WB |
2489 | data += hdrlen; |
2490 | to_write -= hdrlen; | |
69e3c75f JB |
2491 | } |
2492 | ||
69e3c75f JB |
2493 | offset = offset_in_page(data); |
2494 | len_max = PAGE_SIZE - offset; | |
2495 | len = ((to_write > len_max) ? len_max : to_write); | |
2496 | ||
2497 | skb->data_len = to_write; | |
2498 | skb->len += to_write; | |
2499 | skb->truesize += to_write; | |
14afee4b | 2500 | refcount_add(to_write, &po->sk.sk_wmem_alloc); |
69e3c75f JB |
2501 | |
2502 | while (likely(to_write)) { | |
2503 | nr_frags = skb_shinfo(skb)->nr_frags; | |
2504 | ||
2505 | if (unlikely(nr_frags >= MAX_SKB_FRAGS)) { | |
40d4e3df ED |
2506 | pr_err("Packet exceed the number of skb frags(%lu)\n", |
2507 | MAX_SKB_FRAGS); | |
69e3c75f JB |
2508 | return -EFAULT; |
2509 | } | |
2510 | ||
0af55bb5 CG |
2511 | page = pgv_to_page(data); |
2512 | data += len; | |
69e3c75f JB |
2513 | flush_dcache_page(page); |
2514 | get_page(page); | |
0af55bb5 | 2515 | skb_fill_page_desc(skb, nr_frags, page, offset, len); |
69e3c75f JB |
2516 | to_write -= len; |
2517 | offset = 0; | |
2518 | len_max = PAGE_SIZE; | |
2519 | len = ((to_write > len_max) ? len_max : to_write); | |
2520 | } | |
2521 | ||
8fd6c80d | 2522 | skb_probe_transport_header(skb, 0); |
efdfa2f7 | 2523 | |
69e3c75f JB |
2524 | return tp_len; |
2525 | } | |
2526 | ||
8d39b4a6 WB |
2527 | static int tpacket_parse_header(struct packet_sock *po, void *frame, |
2528 | int size_max, void **data) | |
2529 | { | |
2530 | union tpacket_uhdr ph; | |
2531 | int tp_len, off; | |
2532 | ||
2533 | ph.raw = frame; | |
2534 | ||
2535 | switch (po->tp_version) { | |
7f953ab2 SV |
2536 | case TPACKET_V3: |
2537 | if (ph.h3->tp_next_offset != 0) { | |
2538 | pr_warn_once("variable sized slot not supported"); | |
2539 | return -EINVAL; | |
2540 | } | |
2541 | tp_len = ph.h3->tp_len; | |
2542 | break; | |
8d39b4a6 WB |
2543 | case TPACKET_V2: |
2544 | tp_len = ph.h2->tp_len; | |
2545 | break; | |
2546 | default: | |
2547 | tp_len = ph.h1->tp_len; | |
2548 | break; | |
2549 | } | |
2550 | if (unlikely(tp_len > size_max)) { | |
2551 | pr_err("packet size is too long (%d > %d)\n", tp_len, size_max); | |
2552 | return -EMSGSIZE; | |
2553 | } | |
2554 | ||
2555 | if (unlikely(po->tp_tx_has_off)) { | |
2556 | int off_min, off_max; | |
2557 | ||
2558 | off_min = po->tp_hdrlen - sizeof(struct sockaddr_ll); | |
2559 | off_max = po->tx_ring.frame_size - tp_len; | |
2560 | if (po->sk.sk_type == SOCK_DGRAM) { | |
2561 | switch (po->tp_version) { | |
7f953ab2 SV |
2562 | case TPACKET_V3: |
2563 | off = ph.h3->tp_net; | |
2564 | break; | |
8d39b4a6 WB |
2565 | case TPACKET_V2: |
2566 | off = ph.h2->tp_net; | |
2567 | break; | |
2568 | default: | |
2569 | off = ph.h1->tp_net; | |
2570 | break; | |
2571 | } | |
2572 | } else { | |
2573 | switch (po->tp_version) { | |
7f953ab2 SV |
2574 | case TPACKET_V3: |
2575 | off = ph.h3->tp_mac; | |
2576 | break; | |
8d39b4a6 WB |
2577 | case TPACKET_V2: |
2578 | off = ph.h2->tp_mac; | |
2579 | break; | |
2580 | default: | |
2581 | off = ph.h1->tp_mac; | |
2582 | break; | |
2583 | } | |
2584 | } | |
2585 | if (unlikely((off < off_min) || (off_max < off))) | |
2586 | return -EINVAL; | |
2587 | } else { | |
2588 | off = po->tp_hdrlen - sizeof(struct sockaddr_ll); | |
2589 | } | |
2590 | ||
2591 | *data = frame + off; | |
2592 | return tp_len; | |
2593 | } | |
2594 | ||
69e3c75f JB |
2595 | static int tpacket_snd(struct packet_sock *po, struct msghdr *msg) |
2596 | { | |
69e3c75f JB |
2597 | struct sk_buff *skb; |
2598 | struct net_device *dev; | |
1d036d25 | 2599 | struct virtio_net_hdr *vnet_hdr = NULL; |
c14ac945 | 2600 | struct sockcm_cookie sockc; |
69e3c75f | 2601 | __be16 proto; |
09effa67 | 2602 | int err, reserve = 0; |
40d4e3df | 2603 | void *ph; |
342dfc30 | 2604 | DECLARE_SOCKADDR(struct sockaddr_ll *, saddr, msg->msg_name); |
87a2fd28 | 2605 | bool need_wait = !(msg->msg_flags & MSG_DONTWAIT); |
69e3c75f JB |
2606 | int tp_len, size_max; |
2607 | unsigned char *addr; | |
8d39b4a6 | 2608 | void *data; |
69e3c75f | 2609 | int len_sum = 0; |
9e67030a | 2610 | int status = TP_STATUS_AVAILABLE; |
1d036d25 | 2611 | int hlen, tlen, copylen = 0; |
69e3c75f | 2612 | |
69e3c75f JB |
2613 | mutex_lock(&po->pg_vec_lock); |
2614 | ||
66e56cd4 | 2615 | if (likely(saddr == NULL)) { |
e40526cb | 2616 | dev = packet_cached_dev_get(po); |
69e3c75f JB |
2617 | proto = po->num; |
2618 | addr = NULL; | |
2619 | } else { | |
2620 | err = -EINVAL; | |
2621 | if (msg->msg_namelen < sizeof(struct sockaddr_ll)) | |
2622 | goto out; | |
2623 | if (msg->msg_namelen < (saddr->sll_halen | |
2624 | + offsetof(struct sockaddr_ll, | |
2625 | sll_addr))) | |
2626 | goto out; | |
69e3c75f | 2627 | proto = saddr->sll_protocol; |
6b8d95f1 | 2628 | addr = saddr->sll_halen ? saddr->sll_addr : NULL; |
827d9780 | 2629 | dev = dev_get_by_index(sock_net(&po->sk), saddr->sll_ifindex); |
99137b78 WB |
2630 | if (addr && dev && saddr->sll_halen < dev->addr_len) |
2631 | goto out; | |
69e3c75f JB |
2632 | } |
2633 | ||
69e3c75f JB |
2634 | err = -ENXIO; |
2635 | if (unlikely(dev == NULL)) | |
2636 | goto out; | |
69e3c75f JB |
2637 | err = -ENETDOWN; |
2638 | if (unlikely(!(dev->flags & IFF_UP))) | |
2639 | goto out_put; | |
2640 | ||
657a0667 | 2641 | sockcm_init(&sockc, &po->sk); |
d19b183c DCS |
2642 | if (msg->msg_controllen) { |
2643 | err = sock_cmsg_send(&po->sk, msg, &sockc); | |
2644 | if (unlikely(err)) | |
2645 | goto out_put; | |
2646 | } | |
2647 | ||
5cfb4c8d DB |
2648 | if (po->sk.sk_socket->type == SOCK_RAW) |
2649 | reserve = dev->hard_header_len; | |
69e3c75f | 2650 | size_max = po->tx_ring.frame_size |
b5dd884e | 2651 | - (po->tp_hdrlen - sizeof(struct sockaddr_ll)); |
69e3c75f | 2652 | |
1d036d25 | 2653 | if ((size_max > dev->mtu + reserve + VLAN_HLEN) && !po->has_vnet_hdr) |
5cfb4c8d | 2654 | size_max = dev->mtu + reserve + VLAN_HLEN; |
09effa67 | 2655 | |
69e3c75f JB |
2656 | do { |
2657 | ph = packet_current_frame(po, &po->tx_ring, | |
87a2fd28 | 2658 | TP_STATUS_SEND_REQUEST); |
69e3c75f | 2659 | if (unlikely(ph == NULL)) { |
87a2fd28 DB |
2660 | if (need_wait && need_resched()) |
2661 | schedule(); | |
69e3c75f JB |
2662 | continue; |
2663 | } | |
2664 | ||
8d39b4a6 WB |
2665 | skb = NULL; |
2666 | tp_len = tpacket_parse_header(po, ph, size_max, &data); | |
2667 | if (tp_len < 0) | |
2668 | goto tpacket_error; | |
2669 | ||
69e3c75f | 2670 | status = TP_STATUS_SEND_REQUEST; |
ae641949 HX |
2671 | hlen = LL_RESERVED_SPACE(dev); |
2672 | tlen = dev->needed_tailroom; | |
1d036d25 WB |
2673 | if (po->has_vnet_hdr) { |
2674 | vnet_hdr = data; | |
2675 | data += sizeof(*vnet_hdr); | |
2676 | tp_len -= sizeof(*vnet_hdr); | |
2677 | if (tp_len < 0 || | |
2678 | __packet_snd_vnet_parse(vnet_hdr, tp_len)) { | |
2679 | tp_len = -EINVAL; | |
2680 | goto tpacket_error; | |
2681 | } | |
2682 | copylen = __virtio16_to_cpu(vio_le(), | |
2683 | vnet_hdr->hdr_len); | |
2684 | } | |
9ed988cd | 2685 | copylen = max_t(int, copylen, dev->hard_header_len); |
69e3c75f | 2686 | skb = sock_alloc_send_skb(&po->sk, |
1d036d25 WB |
2687 | hlen + tlen + sizeof(struct sockaddr_ll) + |
2688 | (copylen - dev->hard_header_len), | |
fbf33a28 | 2689 | !need_wait, &err); |
69e3c75f | 2690 | |
fbf33a28 KM |
2691 | if (unlikely(skb == NULL)) { |
2692 | /* we assume the socket was initially writeable ... */ | |
2693 | if (likely(len_sum > 0)) | |
2694 | err = len_sum; | |
69e3c75f | 2695 | goto out_status; |
fbf33a28 | 2696 | } |
8d39b4a6 | 2697 | tp_len = tpacket_fill_skb(po, skb, ph, dev, data, tp_len, proto, |
c14ac945 | 2698 | addr, hlen, copylen, &sockc); |
dbd46ab4 | 2699 | if (likely(tp_len >= 0) && |
5cfb4c8d | 2700 | tp_len > dev->mtu + reserve && |
1d036d25 | 2701 | !po->has_vnet_hdr && |
3c70c132 DB |
2702 | !packet_extra_vlan_len_allowed(dev, skb)) |
2703 | tp_len = -EMSGSIZE; | |
69e3c75f JB |
2704 | |
2705 | if (unlikely(tp_len < 0)) { | |
8d39b4a6 | 2706 | tpacket_error: |
69e3c75f JB |
2707 | if (po->tp_loss) { |
2708 | __packet_set_status(po, ph, | |
2709 | TP_STATUS_AVAILABLE); | |
2710 | packet_increment_head(&po->tx_ring); | |
2711 | kfree_skb(skb); | |
2712 | continue; | |
2713 | } else { | |
2714 | status = TP_STATUS_WRONG_FORMAT; | |
2715 | err = tp_len; | |
2716 | goto out_status; | |
2717 | } | |
2718 | } | |
2719 | ||
9d2f67e4 JT |
2720 | if (po->has_vnet_hdr) { |
2721 | if (virtio_net_hdr_to_skb(skb, vnet_hdr, vio_le())) { | |
2722 | tp_len = -EINVAL; | |
2723 | goto tpacket_error; | |
2724 | } | |
2725 | virtio_net_hdr_set_proto(skb, vnet_hdr); | |
1d036d25 WB |
2726 | } |
2727 | ||
69e3c75f JB |
2728 | skb->destructor = tpacket_destruct_skb; |
2729 | __packet_set_status(po, ph, TP_STATUS_SENDING); | |
b0138408 | 2730 | packet_inc_pending(&po->tx_ring); |
69e3c75f JB |
2731 | |
2732 | status = TP_STATUS_SEND_REQUEST; | |
d346a3fa | 2733 | err = po->xmit(skb); |
eb70df13 JP |
2734 | if (unlikely(err > 0)) { |
2735 | err = net_xmit_errno(err); | |
2736 | if (err && __packet_get_status(po, ph) == | |
2737 | TP_STATUS_AVAILABLE) { | |
2738 | /* skb was destructed already */ | |
2739 | skb = NULL; | |
2740 | goto out_status; | |
2741 | } | |
2742 | /* | |
2743 | * skb was dropped but not destructed yet; | |
2744 | * let's treat it like congestion or err < 0 | |
2745 | */ | |
2746 | err = 0; | |
2747 | } | |
69e3c75f JB |
2748 | packet_increment_head(&po->tx_ring); |
2749 | len_sum += tp_len; | |
b0138408 DB |
2750 | } while (likely((ph != NULL) || |
2751 | /* Note: packet_read_pending() might be slow if we have | |
2752 | * to call it as it's per_cpu variable, but in fast-path | |
2753 | * we already short-circuit the loop with the first | |
2754 | * condition, and luckily don't have to go that path | |
2755 | * anyway. | |
2756 | */ | |
2757 | (need_wait && packet_read_pending(&po->tx_ring)))); | |
69e3c75f JB |
2758 | |
2759 | err = len_sum; | |
2760 | goto out_put; | |
2761 | ||
69e3c75f JB |
2762 | out_status: |
2763 | __packet_set_status(po, ph, status); | |
2764 | kfree_skb(skb); | |
2765 | out_put: | |
e40526cb | 2766 | dev_put(dev); |
69e3c75f JB |
2767 | out: |
2768 | mutex_unlock(&po->pg_vec_lock); | |
2769 | return err; | |
2770 | } | |
69e3c75f | 2771 | |
eea49cc9 OJ |
2772 | static struct sk_buff *packet_alloc_skb(struct sock *sk, size_t prepad, |
2773 | size_t reserve, size_t len, | |
2774 | size_t linear, int noblock, | |
2775 | int *err) | |
bfd5f4a3 SS |
2776 | { |
2777 | struct sk_buff *skb; | |
2778 | ||
2779 | /* Under a page? Don't bother with paged skb. */ | |
2780 | if (prepad + len < PAGE_SIZE || !linear) | |
2781 | linear = len; | |
2782 | ||
2783 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, | |
28d64271 | 2784 | err, 0); |
bfd5f4a3 SS |
2785 | if (!skb) |
2786 | return NULL; | |
2787 | ||
2788 | skb_reserve(skb, reserve); | |
2789 | skb_put(skb, linear); | |
2790 | skb->data_len = len - linear; | |
2791 | skb->len += len - linear; | |
2792 | ||
2793 | return skb; | |
2794 | } | |
2795 | ||
d346a3fa | 2796 | static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 LT |
2797 | { |
2798 | struct sock *sk = sock->sk; | |
342dfc30 | 2799 | DECLARE_SOCKADDR(struct sockaddr_ll *, saddr, msg->msg_name); |
1da177e4 LT |
2800 | struct sk_buff *skb; |
2801 | struct net_device *dev; | |
0e11c91e | 2802 | __be16 proto; |
1da177e4 | 2803 | unsigned char *addr; |
827d9780 | 2804 | int err, reserve = 0; |
c7d39e32 | 2805 | struct sockcm_cookie sockc; |
bfd5f4a3 SS |
2806 | struct virtio_net_hdr vnet_hdr = { 0 }; |
2807 | int offset = 0; | |
bfd5f4a3 | 2808 | struct packet_sock *po = pkt_sk(sk); |
da7c9561 | 2809 | bool has_vnet_hdr = false; |
57031eb7 | 2810 | int hlen, tlen, linear; |
3bdc0eba | 2811 | int extra_len = 0; |
1da177e4 LT |
2812 | |
2813 | /* | |
1ce4f28b | 2814 | * Get and verify the address. |
1da177e4 | 2815 | */ |
1ce4f28b | 2816 | |
66e56cd4 | 2817 | if (likely(saddr == NULL)) { |
e40526cb | 2818 | dev = packet_cached_dev_get(po); |
1da177e4 LT |
2819 | proto = po->num; |
2820 | addr = NULL; | |
2821 | } else { | |
2822 | err = -EINVAL; | |
2823 | if (msg->msg_namelen < sizeof(struct sockaddr_ll)) | |
2824 | goto out; | |
0fb375fb EB |
2825 | if (msg->msg_namelen < (saddr->sll_halen + offsetof(struct sockaddr_ll, sll_addr))) |
2826 | goto out; | |
1da177e4 | 2827 | proto = saddr->sll_protocol; |
6b8d95f1 | 2828 | addr = saddr->sll_halen ? saddr->sll_addr : NULL; |
827d9780 | 2829 | dev = dev_get_by_index(sock_net(sk), saddr->sll_ifindex); |
99137b78 WB |
2830 | if (addr && dev && saddr->sll_halen < dev->addr_len) |
2831 | goto out; | |
1da177e4 LT |
2832 | } |
2833 | ||
1da177e4 | 2834 | err = -ENXIO; |
e40526cb | 2835 | if (unlikely(dev == NULL)) |
1da177e4 | 2836 | goto out_unlock; |
d5e76b0a | 2837 | err = -ENETDOWN; |
e40526cb | 2838 | if (unlikely(!(dev->flags & IFF_UP))) |
d5e76b0a DM |
2839 | goto out_unlock; |
2840 | ||
657a0667 | 2841 | sockcm_init(&sockc, sk); |
c7d39e32 EJ |
2842 | sockc.mark = sk->sk_mark; |
2843 | if (msg->msg_controllen) { | |
2844 | err = sock_cmsg_send(sk, msg, &sockc); | |
2845 | if (unlikely(err)) | |
2846 | goto out_unlock; | |
2847 | } | |
2848 | ||
e40526cb DB |
2849 | if (sock->type == SOCK_RAW) |
2850 | reserve = dev->hard_header_len; | |
bfd5f4a3 | 2851 | if (po->has_vnet_hdr) { |
16cc1400 WB |
2852 | err = packet_snd_vnet_parse(msg, &len, &vnet_hdr); |
2853 | if (err) | |
bfd5f4a3 | 2854 | goto out_unlock; |
da7c9561 | 2855 | has_vnet_hdr = true; |
bfd5f4a3 SS |
2856 | } |
2857 | ||
3bdc0eba BG |
2858 | if (unlikely(sock_flag(sk, SOCK_NOFCS))) { |
2859 | if (!netif_supports_nofcs(dev)) { | |
2860 | err = -EPROTONOSUPPORT; | |
2861 | goto out_unlock; | |
2862 | } | |
2863 | extra_len = 4; /* We're doing our own CRC */ | |
2864 | } | |
2865 | ||
1da177e4 | 2866 | err = -EMSGSIZE; |
16cc1400 WB |
2867 | if (!vnet_hdr.gso_type && |
2868 | (len > dev->mtu + reserve + VLAN_HLEN + extra_len)) | |
1da177e4 LT |
2869 | goto out_unlock; |
2870 | ||
bfd5f4a3 | 2871 | err = -ENOBUFS; |
ae641949 HX |
2872 | hlen = LL_RESERVED_SPACE(dev); |
2873 | tlen = dev->needed_tailroom; | |
57031eb7 WB |
2874 | linear = __virtio16_to_cpu(vio_le(), vnet_hdr.hdr_len); |
2875 | linear = max(linear, min_t(int, len, dev->hard_header_len)); | |
2876 | skb = packet_alloc_skb(sk, hlen + tlen, hlen, len, linear, | |
bfd5f4a3 | 2877 | msg->msg_flags & MSG_DONTWAIT, &err); |
40d4e3df | 2878 | if (skb == NULL) |
1da177e4 LT |
2879 | goto out_unlock; |
2880 | ||
b84bbaf7 | 2881 | skb_reset_network_header(skb); |
1da177e4 | 2882 | |
0c4e8581 | 2883 | err = -EINVAL; |
9c707762 WB |
2884 | if (sock->type == SOCK_DGRAM) { |
2885 | offset = dev_hard_header(skb, dev, ntohs(proto), addr, NULL, len); | |
46d2cfb1 | 2886 | if (unlikely(offset < 0)) |
9c707762 | 2887 | goto out_free; |
b84bbaf7 | 2888 | } else if (reserve) { |
9aad13b0 | 2889 | skb_reserve(skb, -reserve); |
993675a3 WB |
2890 | if (len < reserve) |
2891 | skb_reset_network_header(skb); | |
9c707762 | 2892 | } |
1da177e4 LT |
2893 | |
2894 | /* Returns -EFAULT on error */ | |
c0371da6 | 2895 | err = skb_copy_datagram_from_iter(skb, offset, &msg->msg_iter, len); |
1da177e4 LT |
2896 | if (err) |
2897 | goto out_free; | |
bf84a010 | 2898 | |
9ed988cd WB |
2899 | if (sock->type == SOCK_RAW && |
2900 | !dev_validate_header(dev, skb->data, len)) { | |
2901 | err = -EINVAL; | |
2902 | goto out_free; | |
2903 | } | |
2904 | ||
8f932f76 | 2905 | skb_setup_tx_timestamp(skb, sockc.tsflags); |
1da177e4 | 2906 | |
16cc1400 | 2907 | if (!vnet_hdr.gso_type && (len > dev->mtu + reserve + extra_len) && |
3c70c132 DB |
2908 | !packet_extra_vlan_len_allowed(dev, skb)) { |
2909 | err = -EMSGSIZE; | |
2910 | goto out_free; | |
57f89bfa BG |
2911 | } |
2912 | ||
09effa67 DM |
2913 | skb->protocol = proto; |
2914 | skb->dev = dev; | |
1da177e4 | 2915 | skb->priority = sk->sk_priority; |
c7d39e32 | 2916 | skb->mark = sockc.mark; |
3d0ba8c0 | 2917 | skb->tstamp = sockc.transmit_time; |
0fd5d57b | 2918 | |
da7c9561 | 2919 | if (has_vnet_hdr) { |
db60eb5f | 2920 | err = virtio_net_hdr_to_skb(skb, &vnet_hdr, vio_le()); |
16cc1400 WB |
2921 | if (err) |
2922 | goto out_free; | |
2923 | len += sizeof(vnet_hdr); | |
9d2f67e4 | 2924 | virtio_net_hdr_set_proto(skb, &vnet_hdr); |
bfd5f4a3 SS |
2925 | } |
2926 | ||
8fd6c80d DB |
2927 | skb_probe_transport_header(skb, reserve); |
2928 | ||
3bdc0eba BG |
2929 | if (unlikely(extra_len == 4)) |
2930 | skb->no_fcs = 1; | |
2931 | ||
d346a3fa | 2932 | err = po->xmit(skb); |
1da177e4 LT |
2933 | if (err > 0 && (err = net_xmit_errno(err)) != 0) |
2934 | goto out_unlock; | |
2935 | ||
e40526cb | 2936 | dev_put(dev); |
1da177e4 | 2937 | |
40d4e3df | 2938 | return len; |
1da177e4 LT |
2939 | |
2940 | out_free: | |
2941 | kfree_skb(skb); | |
2942 | out_unlock: | |
e40526cb | 2943 | if (dev) |
1da177e4 LT |
2944 | dev_put(dev); |
2945 | out: | |
2946 | return err; | |
2947 | } | |
2948 | ||
1b784140 | 2949 | static int packet_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
69e3c75f | 2950 | { |
69e3c75f JB |
2951 | struct sock *sk = sock->sk; |
2952 | struct packet_sock *po = pkt_sk(sk); | |
d346a3fa | 2953 | |
69e3c75f JB |
2954 | if (po->tx_ring.pg_vec) |
2955 | return tpacket_snd(po, msg); | |
2956 | else | |
69e3c75f JB |
2957 | return packet_snd(sock, msg, len); |
2958 | } | |
2959 | ||
1da177e4 LT |
2960 | /* |
2961 | * Close a PACKET socket. This is fairly simple. We immediately go | |
2962 | * to 'closed' state and remove our protocol entry in the device list. | |
2963 | */ | |
2964 | ||
2965 | static int packet_release(struct socket *sock) | |
2966 | { | |
2967 | struct sock *sk = sock->sk; | |
2968 | struct packet_sock *po; | |
2bd624b4 | 2969 | struct packet_fanout *f; |
d12d01d6 | 2970 | struct net *net; |
f6fb8f10 | 2971 | union tpacket_req_u req_u; |
1da177e4 LT |
2972 | |
2973 | if (!sk) | |
2974 | return 0; | |
2975 | ||
3b1e0a65 | 2976 | net = sock_net(sk); |
1da177e4 LT |
2977 | po = pkt_sk(sk); |
2978 | ||
0fa7fa98 | 2979 | mutex_lock(&net->packet.sklist_lock); |
808f5114 | 2980 | sk_del_node_init_rcu(sk); |
0fa7fa98 PE |
2981 | mutex_unlock(&net->packet.sklist_lock); |
2982 | ||
2983 | preempt_disable(); | |
920de804 | 2984 | sock_prot_inuse_add(net, sk->sk_prot, -1); |
0fa7fa98 | 2985 | preempt_enable(); |
1da177e4 | 2986 | |
808f5114 | 2987 | spin_lock(&po->bind_lock); |
ce06b03e | 2988 | unregister_prot_hook(sk, false); |
66e56cd4 DB |
2989 | packet_cached_dev_reset(po); |
2990 | ||
160ff18a BG |
2991 | if (po->prot_hook.dev) { |
2992 | dev_put(po->prot_hook.dev); | |
2993 | po->prot_hook.dev = NULL; | |
2994 | } | |
808f5114 | 2995 | spin_unlock(&po->bind_lock); |
1da177e4 | 2996 | |
1da177e4 | 2997 | packet_flush_mclist(sk); |
1da177e4 | 2998 | |
5171b37d | 2999 | lock_sock(sk); |
9665d5d6 PS |
3000 | if (po->rx_ring.pg_vec) { |
3001 | memset(&req_u, 0, sizeof(req_u)); | |
f6fb8f10 | 3002 | packet_set_ring(sk, &req_u, 1, 0); |
9665d5d6 | 3003 | } |
69e3c75f | 3004 | |
9665d5d6 PS |
3005 | if (po->tx_ring.pg_vec) { |
3006 | memset(&req_u, 0, sizeof(req_u)); | |
f6fb8f10 | 3007 | packet_set_ring(sk, &req_u, 1, 1); |
9665d5d6 | 3008 | } |
5171b37d | 3009 | release_sock(sk); |
1da177e4 | 3010 | |
2bd624b4 | 3011 | f = fanout_release(sk); |
dc99f600 | 3012 | |
808f5114 | 3013 | synchronize_net(); |
2bd624b4 AS |
3014 | |
3015 | if (f) { | |
57f015f5 | 3016 | kfree(po->rollover); |
2bd624b4 AS |
3017 | fanout_release_data(f); |
3018 | kfree(f); | |
3019 | } | |
1da177e4 LT |
3020 | /* |
3021 | * Now the socket is dead. No more input will appear. | |
3022 | */ | |
1da177e4 LT |
3023 | sock_orphan(sk); |
3024 | sock->sk = NULL; | |
3025 | ||
3026 | /* Purge queues */ | |
3027 | ||
3028 | skb_queue_purge(&sk->sk_receive_queue); | |
b0138408 | 3029 | packet_free_pending(po); |
17ab56a2 | 3030 | sk_refcnt_debug_release(sk); |
1da177e4 LT |
3031 | |
3032 | sock_put(sk); | |
3033 | return 0; | |
3034 | } | |
3035 | ||
3036 | /* | |
3037 | * Attach a packet hook. | |
3038 | */ | |
3039 | ||
30f7ea1c FR |
3040 | static int packet_do_bind(struct sock *sk, const char *name, int ifindex, |
3041 | __be16 proto) | |
1da177e4 LT |
3042 | { |
3043 | struct packet_sock *po = pkt_sk(sk); | |
158cd4af | 3044 | struct net_device *dev_curr; |
902fefb8 DB |
3045 | __be16 proto_curr; |
3046 | bool need_rehook; | |
30f7ea1c FR |
3047 | struct net_device *dev = NULL; |
3048 | int ret = 0; | |
3049 | bool unlisted = false; | |
dc99f600 | 3050 | |
1da177e4 | 3051 | lock_sock(sk); |
1da177e4 | 3052 | spin_lock(&po->bind_lock); |
30f7ea1c FR |
3053 | rcu_read_lock(); |
3054 | ||
4971613c WB |
3055 | if (po->fanout) { |
3056 | ret = -EINVAL; | |
3057 | goto out_unlock; | |
3058 | } | |
3059 | ||
30f7ea1c FR |
3060 | if (name) { |
3061 | dev = dev_get_by_name_rcu(sock_net(sk), name); | |
3062 | if (!dev) { | |
3063 | ret = -ENODEV; | |
3064 | goto out_unlock; | |
3065 | } | |
3066 | } else if (ifindex) { | |
3067 | dev = dev_get_by_index_rcu(sock_net(sk), ifindex); | |
3068 | if (!dev) { | |
3069 | ret = -ENODEV; | |
3070 | goto out_unlock; | |
3071 | } | |
3072 | } | |
3073 | ||
3074 | if (dev) | |
3075 | dev_hold(dev); | |
66e56cd4 | 3076 | |
902fefb8 DB |
3077 | proto_curr = po->prot_hook.type; |
3078 | dev_curr = po->prot_hook.dev; | |
3079 | ||
3080 | need_rehook = proto_curr != proto || dev_curr != dev; | |
3081 | ||
3082 | if (need_rehook) { | |
30f7ea1c FR |
3083 | if (po->running) { |
3084 | rcu_read_unlock(); | |
15fe076e ED |
3085 | /* prevents packet_notifier() from calling |
3086 | * register_prot_hook() | |
3087 | */ | |
3088 | po->num = 0; | |
30f7ea1c FR |
3089 | __unregister_prot_hook(sk, true); |
3090 | rcu_read_lock(); | |
3091 | dev_curr = po->prot_hook.dev; | |
3092 | if (dev) | |
3093 | unlisted = !dev_get_by_index_rcu(sock_net(sk), | |
3094 | dev->ifindex); | |
3095 | } | |
1da177e4 | 3096 | |
15fe076e | 3097 | BUG_ON(po->running); |
902fefb8 DB |
3098 | po->num = proto; |
3099 | po->prot_hook.type = proto; | |
902fefb8 | 3100 | |
30f7ea1c FR |
3101 | if (unlikely(unlisted)) { |
3102 | dev_put(dev); | |
3103 | po->prot_hook.dev = NULL; | |
3104 | po->ifindex = -1; | |
3105 | packet_cached_dev_reset(po); | |
3106 | } else { | |
3107 | po->prot_hook.dev = dev; | |
3108 | po->ifindex = dev ? dev->ifindex : 0; | |
3109 | packet_cached_dev_assign(po, dev); | |
3110 | } | |
902fefb8 | 3111 | } |
158cd4af LW |
3112 | if (dev_curr) |
3113 | dev_put(dev_curr); | |
66e56cd4 | 3114 | |
902fefb8 | 3115 | if (proto == 0 || !need_rehook) |
1da177e4 LT |
3116 | goto out_unlock; |
3117 | ||
30f7ea1c | 3118 | if (!unlisted && (!dev || (dev->flags & IFF_UP))) { |
ce06b03e | 3119 | register_prot_hook(sk); |
be85d4ad UT |
3120 | } else { |
3121 | sk->sk_err = ENETDOWN; | |
3122 | if (!sock_flag(sk, SOCK_DEAD)) | |
3123 | sk->sk_error_report(sk); | |
1da177e4 LT |
3124 | } |
3125 | ||
3126 | out_unlock: | |
30f7ea1c | 3127 | rcu_read_unlock(); |
1da177e4 LT |
3128 | spin_unlock(&po->bind_lock); |
3129 | release_sock(sk); | |
30f7ea1c | 3130 | return ret; |
1da177e4 LT |
3131 | } |
3132 | ||
3133 | /* | |
3134 | * Bind a packet socket to a device | |
3135 | */ | |
3136 | ||
40d4e3df ED |
3137 | static int packet_bind_spkt(struct socket *sock, struct sockaddr *uaddr, |
3138 | int addr_len) | |
1da177e4 | 3139 | { |
40d4e3df | 3140 | struct sock *sk = sock->sk; |
540e2894 | 3141 | char name[sizeof(uaddr->sa_data) + 1]; |
1ce4f28b | 3142 | |
1da177e4 LT |
3143 | /* |
3144 | * Check legality | |
3145 | */ | |
1ce4f28b | 3146 | |
8ae55f04 | 3147 | if (addr_len != sizeof(struct sockaddr)) |
1da177e4 | 3148 | return -EINVAL; |
540e2894 AP |
3149 | /* uaddr->sa_data comes from the userspace, it's not guaranteed to be |
3150 | * zero-terminated. | |
3151 | */ | |
3152 | memcpy(name, uaddr->sa_data, sizeof(uaddr->sa_data)); | |
3153 | name[sizeof(uaddr->sa_data)] = 0; | |
1da177e4 | 3154 | |
30f7ea1c | 3155 | return packet_do_bind(sk, name, 0, pkt_sk(sk)->num); |
1da177e4 | 3156 | } |
1da177e4 LT |
3157 | |
3158 | static int packet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
3159 | { | |
40d4e3df ED |
3160 | struct sockaddr_ll *sll = (struct sockaddr_ll *)uaddr; |
3161 | struct sock *sk = sock->sk; | |
1da177e4 LT |
3162 | |
3163 | /* | |
3164 | * Check legality | |
3165 | */ | |
1ce4f28b | 3166 | |
1da177e4 LT |
3167 | if (addr_len < sizeof(struct sockaddr_ll)) |
3168 | return -EINVAL; | |
3169 | if (sll->sll_family != AF_PACKET) | |
3170 | return -EINVAL; | |
3171 | ||
30f7ea1c FR |
3172 | return packet_do_bind(sk, NULL, sll->sll_ifindex, |
3173 | sll->sll_protocol ? : pkt_sk(sk)->num); | |
1da177e4 LT |
3174 | } |
3175 | ||
3176 | static struct proto packet_proto = { | |
3177 | .name = "PACKET", | |
3178 | .owner = THIS_MODULE, | |
3179 | .obj_size = sizeof(struct packet_sock), | |
3180 | }; | |
3181 | ||
3182 | /* | |
1ce4f28b | 3183 | * Create a packet of type SOCK_PACKET. |
1da177e4 LT |
3184 | */ |
3185 | ||
3f378b68 EP |
3186 | static int packet_create(struct net *net, struct socket *sock, int protocol, |
3187 | int kern) | |
1da177e4 LT |
3188 | { |
3189 | struct sock *sk; | |
3190 | struct packet_sock *po; | |
0e11c91e | 3191 | __be16 proto = (__force __be16)protocol; /* weird, but documented */ |
1da177e4 LT |
3192 | int err; |
3193 | ||
df008c91 | 3194 | if (!ns_capable(net->user_ns, CAP_NET_RAW)) |
1da177e4 | 3195 | return -EPERM; |
be02097c DM |
3196 | if (sock->type != SOCK_DGRAM && sock->type != SOCK_RAW && |
3197 | sock->type != SOCK_PACKET) | |
1da177e4 LT |
3198 | return -ESOCKTNOSUPPORT; |
3199 | ||
3200 | sock->state = SS_UNCONNECTED; | |
3201 | ||
3202 | err = -ENOBUFS; | |
11aa9c28 | 3203 | sk = sk_alloc(net, PF_PACKET, GFP_KERNEL, &packet_proto, kern); |
1da177e4 LT |
3204 | if (sk == NULL) |
3205 | goto out; | |
3206 | ||
3207 | sock->ops = &packet_ops; | |
1da177e4 LT |
3208 | if (sock->type == SOCK_PACKET) |
3209 | sock->ops = &packet_ops_spkt; | |
be02097c | 3210 | |
1da177e4 LT |
3211 | sock_init_data(sock, sk); |
3212 | ||
3213 | po = pkt_sk(sk); | |
3214 | sk->sk_family = PF_PACKET; | |
0e11c91e | 3215 | po->num = proto; |
d346a3fa | 3216 | po->xmit = dev_queue_xmit; |
66e56cd4 | 3217 | |
b0138408 DB |
3218 | err = packet_alloc_pending(po); |
3219 | if (err) | |
3220 | goto out2; | |
3221 | ||
66e56cd4 | 3222 | packet_cached_dev_reset(po); |
1da177e4 LT |
3223 | |
3224 | sk->sk_destruct = packet_sock_destruct; | |
17ab56a2 | 3225 | sk_refcnt_debug_inc(sk); |
1da177e4 LT |
3226 | |
3227 | /* | |
3228 | * Attach a protocol block | |
3229 | */ | |
3230 | ||
3231 | spin_lock_init(&po->bind_lock); | |
905db440 | 3232 | mutex_init(&po->pg_vec_lock); |
0648ab70 | 3233 | po->rollover = NULL; |
1da177e4 | 3234 | po->prot_hook.func = packet_rcv; |
be02097c | 3235 | |
1da177e4 LT |
3236 | if (sock->type == SOCK_PACKET) |
3237 | po->prot_hook.func = packet_rcv_spkt; | |
be02097c | 3238 | |
1da177e4 LT |
3239 | po->prot_hook.af_packet_priv = sk; |
3240 | ||
0e11c91e AV |
3241 | if (proto) { |
3242 | po->prot_hook.type = proto; | |
a6361f0c | 3243 | __register_prot_hook(sk); |
1da177e4 LT |
3244 | } |
3245 | ||
0fa7fa98 | 3246 | mutex_lock(&net->packet.sklist_lock); |
808f5114 | 3247 | sk_add_node_rcu(sk, &net->packet.sklist); |
0fa7fa98 PE |
3248 | mutex_unlock(&net->packet.sklist_lock); |
3249 | ||
3250 | preempt_disable(); | |
3680453c | 3251 | sock_prot_inuse_add(net, &packet_proto, 1); |
0fa7fa98 | 3252 | preempt_enable(); |
808f5114 | 3253 | |
40d4e3df | 3254 | return 0; |
b0138408 DB |
3255 | out2: |
3256 | sk_free(sk); | |
1da177e4 LT |
3257 | out: |
3258 | return err; | |
3259 | } | |
3260 | ||
3261 | /* | |
3262 | * Pull a packet from our receive queue and hand it to the user. | |
3263 | * If necessary we block. | |
3264 | */ | |
3265 | ||
1b784140 YX |
3266 | static int packet_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
3267 | int flags) | |
1da177e4 LT |
3268 | { |
3269 | struct sock *sk = sock->sk; | |
3270 | struct sk_buff *skb; | |
3271 | int copied, err; | |
bfd5f4a3 | 3272 | int vnet_hdr_len = 0; |
2472d761 | 3273 | unsigned int origlen = 0; |
1da177e4 LT |
3274 | |
3275 | err = -EINVAL; | |
ed85b565 | 3276 | if (flags & ~(MSG_PEEK|MSG_DONTWAIT|MSG_TRUNC|MSG_CMSG_COMPAT|MSG_ERRQUEUE)) |
1da177e4 LT |
3277 | goto out; |
3278 | ||
3279 | #if 0 | |
3280 | /* What error should we return now? EUNATTACH? */ | |
3281 | if (pkt_sk(sk)->ifindex < 0) | |
3282 | return -ENODEV; | |
3283 | #endif | |
3284 | ||
ed85b565 | 3285 | if (flags & MSG_ERRQUEUE) { |
cb820f8e RC |
3286 | err = sock_recv_errqueue(sk, msg, len, |
3287 | SOL_PACKET, PACKET_TX_TIMESTAMP); | |
ed85b565 RC |
3288 | goto out; |
3289 | } | |
3290 | ||
1da177e4 LT |
3291 | /* |
3292 | * Call the generic datagram receiver. This handles all sorts | |
3293 | * of horrible races and re-entrancy so we can forget about it | |
3294 | * in the protocol layers. | |
3295 | * | |
3296 | * Now it will return ENETDOWN, if device have just gone down, | |
3297 | * but then it will block. | |
3298 | */ | |
3299 | ||
40d4e3df | 3300 | skb = skb_recv_datagram(sk, flags, flags & MSG_DONTWAIT, &err); |
1da177e4 LT |
3301 | |
3302 | /* | |
1ce4f28b | 3303 | * An error occurred so return it. Because skb_recv_datagram() |
1da177e4 LT |
3304 | * handles the blocking we don't see and worry about blocking |
3305 | * retries. | |
3306 | */ | |
3307 | ||
8ae55f04 | 3308 | if (skb == NULL) |
1da177e4 LT |
3309 | goto out; |
3310 | ||
2ccdbaa6 WB |
3311 | if (pkt_sk(sk)->pressure) |
3312 | packet_rcv_has_room(pkt_sk(sk), NULL); | |
3313 | ||
bfd5f4a3 | 3314 | if (pkt_sk(sk)->has_vnet_hdr) { |
16cc1400 WB |
3315 | err = packet_rcv_vnet(msg, skb, &len); |
3316 | if (err) | |
bfd5f4a3 | 3317 | goto out_free; |
16cc1400 | 3318 | vnet_hdr_len = sizeof(struct virtio_net_hdr); |
bfd5f4a3 SS |
3319 | } |
3320 | ||
f3d33426 HFS |
3321 | /* You lose any data beyond the buffer you gave. If it worries |
3322 | * a user program they can ask the device for its MTU | |
3323 | * anyway. | |
1da177e4 | 3324 | */ |
1da177e4 | 3325 | copied = skb->len; |
40d4e3df ED |
3326 | if (copied > len) { |
3327 | copied = len; | |
3328 | msg->msg_flags |= MSG_TRUNC; | |
1da177e4 LT |
3329 | } |
3330 | ||
51f3d02b | 3331 | err = skb_copy_datagram_msg(skb, 0, msg, copied); |
1da177e4 LT |
3332 | if (err) |
3333 | goto out_free; | |
3334 | ||
2472d761 EB |
3335 | if (sock->type != SOCK_PACKET) { |
3336 | struct sockaddr_ll *sll = &PACKET_SKB_CB(skb)->sa.ll; | |
3337 | ||
3338 | /* Original length was stored in sockaddr_ll fields */ | |
3339 | origlen = PACKET_SKB_CB(skb)->sa.origlen; | |
3340 | sll->sll_family = AF_PACKET; | |
3341 | sll->sll_protocol = skb->protocol; | |
3342 | } | |
3343 | ||
3b885787 | 3344 | sock_recv_ts_and_drops(msg, sk, skb); |
1da177e4 | 3345 | |
f3d33426 HFS |
3346 | if (msg->msg_name) { |
3347 | /* If the address length field is there to be filled | |
3348 | * in, we fill it in now. | |
3349 | */ | |
3350 | if (sock->type == SOCK_PACKET) { | |
342dfc30 | 3351 | __sockaddr_check_size(sizeof(struct sockaddr_pkt)); |
f3d33426 HFS |
3352 | msg->msg_namelen = sizeof(struct sockaddr_pkt); |
3353 | } else { | |
3354 | struct sockaddr_ll *sll = &PACKET_SKB_CB(skb)->sa.ll; | |
2472d761 | 3355 | |
f3d33426 HFS |
3356 | msg->msg_namelen = sll->sll_halen + |
3357 | offsetof(struct sockaddr_ll, sll_addr); | |
3358 | } | |
ffbc6111 HX |
3359 | memcpy(msg->msg_name, &PACKET_SKB_CB(skb)->sa, |
3360 | msg->msg_namelen); | |
f3d33426 | 3361 | } |
1da177e4 | 3362 | |
8dc41944 | 3363 | if (pkt_sk(sk)->auxdata) { |
ffbc6111 HX |
3364 | struct tpacket_auxdata aux; |
3365 | ||
3366 | aux.tp_status = TP_STATUS_USER; | |
3367 | if (skb->ip_summed == CHECKSUM_PARTIAL) | |
3368 | aux.tp_status |= TP_STATUS_CSUMNOTREADY; | |
682f048b AD |
3369 | else if (skb->pkt_type != PACKET_OUTGOING && |
3370 | (skb->ip_summed == CHECKSUM_COMPLETE || | |
3371 | skb_csum_unnecessary(skb))) | |
3372 | aux.tp_status |= TP_STATUS_CSUM_VALID; | |
3373 | ||
2472d761 | 3374 | aux.tp_len = origlen; |
ffbc6111 HX |
3375 | aux.tp_snaplen = skb->len; |
3376 | aux.tp_mac = 0; | |
bbe735e4 | 3377 | aux.tp_net = skb_network_offset(skb); |
df8a39de JP |
3378 | if (skb_vlan_tag_present(skb)) { |
3379 | aux.tp_vlan_tci = skb_vlan_tag_get(skb); | |
a0cdfcf3 AW |
3380 | aux.tp_vlan_tpid = ntohs(skb->vlan_proto); |
3381 | aux.tp_status |= TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
a3bcc23e BG |
3382 | } else { |
3383 | aux.tp_vlan_tci = 0; | |
a0cdfcf3 | 3384 | aux.tp_vlan_tpid = 0; |
a3bcc23e | 3385 | } |
ffbc6111 | 3386 | put_cmsg(msg, SOL_PACKET, PACKET_AUXDATA, sizeof(aux), &aux); |
8dc41944 HX |
3387 | } |
3388 | ||
1da177e4 LT |
3389 | /* |
3390 | * Free or return the buffer as appropriate. Again this | |
3391 | * hides all the races and re-entrancy issues from us. | |
3392 | */ | |
bfd5f4a3 | 3393 | err = vnet_hdr_len + ((flags&MSG_TRUNC) ? skb->len : copied); |
1da177e4 LT |
3394 | |
3395 | out_free: | |
3396 | skb_free_datagram(sk, skb); | |
3397 | out: | |
3398 | return err; | |
3399 | } | |
3400 | ||
1da177e4 | 3401 | static int packet_getname_spkt(struct socket *sock, struct sockaddr *uaddr, |
9b2c45d4 | 3402 | int peer) |
1da177e4 LT |
3403 | { |
3404 | struct net_device *dev; | |
3405 | struct sock *sk = sock->sk; | |
3406 | ||
3407 | if (peer) | |
3408 | return -EOPNOTSUPP; | |
3409 | ||
3410 | uaddr->sa_family = AF_PACKET; | |
2dc85bf3 | 3411 | memset(uaddr->sa_data, 0, sizeof(uaddr->sa_data)); |
654d1f8a ED |
3412 | rcu_read_lock(); |
3413 | dev = dev_get_by_index_rcu(sock_net(sk), pkt_sk(sk)->ifindex); | |
3414 | if (dev) | |
2dc85bf3 | 3415 | strlcpy(uaddr->sa_data, dev->name, sizeof(uaddr->sa_data)); |
654d1f8a | 3416 | rcu_read_unlock(); |
1da177e4 | 3417 | |
9b2c45d4 | 3418 | return sizeof(*uaddr); |
1da177e4 | 3419 | } |
1da177e4 LT |
3420 | |
3421 | static int packet_getname(struct socket *sock, struct sockaddr *uaddr, | |
9b2c45d4 | 3422 | int peer) |
1da177e4 LT |
3423 | { |
3424 | struct net_device *dev; | |
3425 | struct sock *sk = sock->sk; | |
3426 | struct packet_sock *po = pkt_sk(sk); | |
13cfa97b | 3427 | DECLARE_SOCKADDR(struct sockaddr_ll *, sll, uaddr); |
1da177e4 LT |
3428 | |
3429 | if (peer) | |
3430 | return -EOPNOTSUPP; | |
3431 | ||
3432 | sll->sll_family = AF_PACKET; | |
3433 | sll->sll_ifindex = po->ifindex; | |
3434 | sll->sll_protocol = po->num; | |
67286640 | 3435 | sll->sll_pkttype = 0; |
654d1f8a ED |
3436 | rcu_read_lock(); |
3437 | dev = dev_get_by_index_rcu(sock_net(sk), po->ifindex); | |
1da177e4 LT |
3438 | if (dev) { |
3439 | sll->sll_hatype = dev->type; | |
3440 | sll->sll_halen = dev->addr_len; | |
3441 | memcpy(sll->sll_addr, dev->dev_addr, dev->addr_len); | |
1da177e4 LT |
3442 | } else { |
3443 | sll->sll_hatype = 0; /* Bad: we have no ARPHRD_UNSPEC */ | |
3444 | sll->sll_halen = 0; | |
3445 | } | |
654d1f8a | 3446 | rcu_read_unlock(); |
1da177e4 | 3447 | |
9b2c45d4 | 3448 | return offsetof(struct sockaddr_ll, sll_addr) + sll->sll_halen; |
1da177e4 LT |
3449 | } |
3450 | ||
2aeb0b88 WC |
3451 | static int packet_dev_mc(struct net_device *dev, struct packet_mclist *i, |
3452 | int what) | |
1da177e4 LT |
3453 | { |
3454 | switch (i->type) { | |
3455 | case PACKET_MR_MULTICAST: | |
1162563f JP |
3456 | if (i->alen != dev->addr_len) |
3457 | return -EINVAL; | |
1da177e4 | 3458 | if (what > 0) |
22bedad3 | 3459 | return dev_mc_add(dev, i->addr); |
1da177e4 | 3460 | else |
22bedad3 | 3461 | return dev_mc_del(dev, i->addr); |
1da177e4 LT |
3462 | break; |
3463 | case PACKET_MR_PROMISC: | |
2aeb0b88 | 3464 | return dev_set_promiscuity(dev, what); |
1da177e4 | 3465 | case PACKET_MR_ALLMULTI: |
2aeb0b88 | 3466 | return dev_set_allmulti(dev, what); |
d95ed927 | 3467 | case PACKET_MR_UNICAST: |
1162563f JP |
3468 | if (i->alen != dev->addr_len) |
3469 | return -EINVAL; | |
d95ed927 | 3470 | if (what > 0) |
a748ee24 | 3471 | return dev_uc_add(dev, i->addr); |
d95ed927 | 3472 | else |
a748ee24 | 3473 | return dev_uc_del(dev, i->addr); |
d95ed927 | 3474 | break; |
40d4e3df ED |
3475 | default: |
3476 | break; | |
1da177e4 | 3477 | } |
2aeb0b88 | 3478 | return 0; |
1da177e4 LT |
3479 | } |
3480 | ||
82f17091 FR |
3481 | static void packet_dev_mclist_delete(struct net_device *dev, |
3482 | struct packet_mclist **mlp) | |
1da177e4 | 3483 | { |
82f17091 FR |
3484 | struct packet_mclist *ml; |
3485 | ||
3486 | while ((ml = *mlp) != NULL) { | |
3487 | if (ml->ifindex == dev->ifindex) { | |
3488 | packet_dev_mc(dev, ml, -1); | |
3489 | *mlp = ml->next; | |
3490 | kfree(ml); | |
3491 | } else | |
3492 | mlp = &ml->next; | |
1da177e4 LT |
3493 | } |
3494 | } | |
3495 | ||
0fb375fb | 3496 | static int packet_mc_add(struct sock *sk, struct packet_mreq_max *mreq) |
1da177e4 LT |
3497 | { |
3498 | struct packet_sock *po = pkt_sk(sk); | |
3499 | struct packet_mclist *ml, *i; | |
3500 | struct net_device *dev; | |
3501 | int err; | |
3502 | ||
3503 | rtnl_lock(); | |
3504 | ||
3505 | err = -ENODEV; | |
3b1e0a65 | 3506 | dev = __dev_get_by_index(sock_net(sk), mreq->mr_ifindex); |
1da177e4 LT |
3507 | if (!dev) |
3508 | goto done; | |
3509 | ||
3510 | err = -EINVAL; | |
1162563f | 3511 | if (mreq->mr_alen > dev->addr_len) |
1da177e4 LT |
3512 | goto done; |
3513 | ||
3514 | err = -ENOBUFS; | |
8b3a7005 | 3515 | i = kmalloc(sizeof(*i), GFP_KERNEL); |
1da177e4 LT |
3516 | if (i == NULL) |
3517 | goto done; | |
3518 | ||
3519 | err = 0; | |
3520 | for (ml = po->mclist; ml; ml = ml->next) { | |
3521 | if (ml->ifindex == mreq->mr_ifindex && | |
3522 | ml->type == mreq->mr_type && | |
3523 | ml->alen == mreq->mr_alen && | |
3524 | memcmp(ml->addr, mreq->mr_address, ml->alen) == 0) { | |
3525 | ml->count++; | |
3526 | /* Free the new element ... */ | |
3527 | kfree(i); | |
3528 | goto done; | |
3529 | } | |
3530 | } | |
3531 | ||
3532 | i->type = mreq->mr_type; | |
3533 | i->ifindex = mreq->mr_ifindex; | |
3534 | i->alen = mreq->mr_alen; | |
3535 | memcpy(i->addr, mreq->mr_address, i->alen); | |
309cf37f | 3536 | memset(i->addr + i->alen, 0, sizeof(i->addr) - i->alen); |
1da177e4 LT |
3537 | i->count = 1; |
3538 | i->next = po->mclist; | |
3539 | po->mclist = i; | |
2aeb0b88 WC |
3540 | err = packet_dev_mc(dev, i, 1); |
3541 | if (err) { | |
3542 | po->mclist = i->next; | |
3543 | kfree(i); | |
3544 | } | |
1da177e4 LT |
3545 | |
3546 | done: | |
3547 | rtnl_unlock(); | |
3548 | return err; | |
3549 | } | |
3550 | ||
0fb375fb | 3551 | static int packet_mc_drop(struct sock *sk, struct packet_mreq_max *mreq) |
1da177e4 LT |
3552 | { |
3553 | struct packet_mclist *ml, **mlp; | |
3554 | ||
3555 | rtnl_lock(); | |
3556 | ||
3557 | for (mlp = &pkt_sk(sk)->mclist; (ml = *mlp) != NULL; mlp = &ml->next) { | |
3558 | if (ml->ifindex == mreq->mr_ifindex && | |
3559 | ml->type == mreq->mr_type && | |
3560 | ml->alen == mreq->mr_alen && | |
3561 | memcmp(ml->addr, mreq->mr_address, ml->alen) == 0) { | |
3562 | if (--ml->count == 0) { | |
3563 | struct net_device *dev; | |
3564 | *mlp = ml->next; | |
ad959e76 ED |
3565 | dev = __dev_get_by_index(sock_net(sk), ml->ifindex); |
3566 | if (dev) | |
1da177e4 | 3567 | packet_dev_mc(dev, ml, -1); |
1da177e4 LT |
3568 | kfree(ml); |
3569 | } | |
82f17091 | 3570 | break; |
1da177e4 LT |
3571 | } |
3572 | } | |
3573 | rtnl_unlock(); | |
82f17091 | 3574 | return 0; |
1da177e4 LT |
3575 | } |
3576 | ||
3577 | static void packet_flush_mclist(struct sock *sk) | |
3578 | { | |
3579 | struct packet_sock *po = pkt_sk(sk); | |
3580 | struct packet_mclist *ml; | |
3581 | ||
3582 | if (!po->mclist) | |
3583 | return; | |
3584 | ||
3585 | rtnl_lock(); | |
3586 | while ((ml = po->mclist) != NULL) { | |
3587 | struct net_device *dev; | |
3588 | ||
3589 | po->mclist = ml->next; | |
ad959e76 ED |
3590 | dev = __dev_get_by_index(sock_net(sk), ml->ifindex); |
3591 | if (dev != NULL) | |
1da177e4 | 3592 | packet_dev_mc(dev, ml, -1); |
1da177e4 LT |
3593 | kfree(ml); |
3594 | } | |
3595 | rtnl_unlock(); | |
3596 | } | |
1da177e4 LT |
3597 | |
3598 | static int | |
b7058842 | 3599 | packet_setsockopt(struct socket *sock, int level, int optname, char __user *optval, unsigned int optlen) |
1da177e4 LT |
3600 | { |
3601 | struct sock *sk = sock->sk; | |
8dc41944 | 3602 | struct packet_sock *po = pkt_sk(sk); |
1da177e4 LT |
3603 | int ret; |
3604 | ||
3605 | if (level != SOL_PACKET) | |
3606 | return -ENOPROTOOPT; | |
3607 | ||
69e3c75f | 3608 | switch (optname) { |
1ce4f28b | 3609 | case PACKET_ADD_MEMBERSHIP: |
1da177e4 LT |
3610 | case PACKET_DROP_MEMBERSHIP: |
3611 | { | |
0fb375fb EB |
3612 | struct packet_mreq_max mreq; |
3613 | int len = optlen; | |
3614 | memset(&mreq, 0, sizeof(mreq)); | |
3615 | if (len < sizeof(struct packet_mreq)) | |
1da177e4 | 3616 | return -EINVAL; |
0fb375fb EB |
3617 | if (len > sizeof(mreq)) |
3618 | len = sizeof(mreq); | |
40d4e3df | 3619 | if (copy_from_user(&mreq, optval, len)) |
1da177e4 | 3620 | return -EFAULT; |
0fb375fb EB |
3621 | if (len < (mreq.mr_alen + offsetof(struct packet_mreq, mr_address))) |
3622 | return -EINVAL; | |
1da177e4 LT |
3623 | if (optname == PACKET_ADD_MEMBERSHIP) |
3624 | ret = packet_mc_add(sk, &mreq); | |
3625 | else | |
3626 | ret = packet_mc_drop(sk, &mreq); | |
3627 | return ret; | |
3628 | } | |
a2efcfa0 | 3629 | |
1da177e4 | 3630 | case PACKET_RX_RING: |
69e3c75f | 3631 | case PACKET_TX_RING: |
1da177e4 | 3632 | { |
f6fb8f10 | 3633 | union tpacket_req_u req_u; |
3634 | int len; | |
1da177e4 | 3635 | |
5171b37d | 3636 | lock_sock(sk); |
f6fb8f10 | 3637 | switch (po->tp_version) { |
3638 | case TPACKET_V1: | |
3639 | case TPACKET_V2: | |
3640 | len = sizeof(req_u.req); | |
3641 | break; | |
3642 | case TPACKET_V3: | |
3643 | default: | |
3644 | len = sizeof(req_u.req3); | |
3645 | break; | |
3646 | } | |
5171b37d ED |
3647 | if (optlen < len) { |
3648 | ret = -EINVAL; | |
3649 | } else { | |
3650 | if (copy_from_user(&req_u.req, optval, len)) | |
3651 | ret = -EFAULT; | |
3652 | else | |
3653 | ret = packet_set_ring(sk, &req_u, 0, | |
3654 | optname == PACKET_TX_RING); | |
3655 | } | |
3656 | release_sock(sk); | |
3657 | return ret; | |
1da177e4 LT |
3658 | } |
3659 | case PACKET_COPY_THRESH: | |
3660 | { | |
3661 | int val; | |
3662 | ||
40d4e3df | 3663 | if (optlen != sizeof(val)) |
1da177e4 | 3664 | return -EINVAL; |
40d4e3df | 3665 | if (copy_from_user(&val, optval, sizeof(val))) |
1da177e4 LT |
3666 | return -EFAULT; |
3667 | ||
3668 | pkt_sk(sk)->copy_thresh = val; | |
3669 | return 0; | |
3670 | } | |
bbd6ef87 PM |
3671 | case PACKET_VERSION: |
3672 | { | |
3673 | int val; | |
3674 | ||
3675 | if (optlen != sizeof(val)) | |
3676 | return -EINVAL; | |
bbd6ef87 PM |
3677 | if (copy_from_user(&val, optval, sizeof(val))) |
3678 | return -EFAULT; | |
3679 | switch (val) { | |
3680 | case TPACKET_V1: | |
3681 | case TPACKET_V2: | |
f6fb8f10 | 3682 | case TPACKET_V3: |
84ac7260 | 3683 | break; |
bbd6ef87 PM |
3684 | default: |
3685 | return -EINVAL; | |
3686 | } | |
84ac7260 PP |
3687 | lock_sock(sk); |
3688 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3689 | ret = -EBUSY; | |
3690 | } else { | |
3691 | po->tp_version = val; | |
3692 | ret = 0; | |
3693 | } | |
3694 | release_sock(sk); | |
3695 | return ret; | |
bbd6ef87 | 3696 | } |
8913336a PM |
3697 | case PACKET_RESERVE: |
3698 | { | |
3699 | unsigned int val; | |
3700 | ||
3701 | if (optlen != sizeof(val)) | |
3702 | return -EINVAL; | |
8913336a PM |
3703 | if (copy_from_user(&val, optval, sizeof(val))) |
3704 | return -EFAULT; | |
bcc5364b AK |
3705 | if (val > INT_MAX) |
3706 | return -EINVAL; | |
c27927e3 WB |
3707 | lock_sock(sk); |
3708 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3709 | ret = -EBUSY; | |
3710 | } else { | |
3711 | po->tp_reserve = val; | |
3712 | ret = 0; | |
3713 | } | |
3714 | release_sock(sk); | |
3715 | return ret; | |
8913336a | 3716 | } |
69e3c75f JB |
3717 | case PACKET_LOSS: |
3718 | { | |
3719 | unsigned int val; | |
3720 | ||
3721 | if (optlen != sizeof(val)) | |
3722 | return -EINVAL; | |
69e3c75f JB |
3723 | if (copy_from_user(&val, optval, sizeof(val))) |
3724 | return -EFAULT; | |
a6361f0c WB |
3725 | |
3726 | lock_sock(sk); | |
3727 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3728 | ret = -EBUSY; | |
3729 | } else { | |
3730 | po->tp_loss = !!val; | |
3731 | ret = 0; | |
3732 | } | |
3733 | release_sock(sk); | |
3734 | return ret; | |
69e3c75f | 3735 | } |
8dc41944 HX |
3736 | case PACKET_AUXDATA: |
3737 | { | |
3738 | int val; | |
3739 | ||
3740 | if (optlen < sizeof(val)) | |
3741 | return -EINVAL; | |
3742 | if (copy_from_user(&val, optval, sizeof(val))) | |
3743 | return -EFAULT; | |
3744 | ||
a6361f0c | 3745 | lock_sock(sk); |
8dc41944 | 3746 | po->auxdata = !!val; |
a6361f0c | 3747 | release_sock(sk); |
8dc41944 HX |
3748 | return 0; |
3749 | } | |
80feaacb PWJ |
3750 | case PACKET_ORIGDEV: |
3751 | { | |
3752 | int val; | |
3753 | ||
3754 | if (optlen < sizeof(val)) | |
3755 | return -EINVAL; | |
3756 | if (copy_from_user(&val, optval, sizeof(val))) | |
3757 | return -EFAULT; | |
3758 | ||
a6361f0c | 3759 | lock_sock(sk); |
80feaacb | 3760 | po->origdev = !!val; |
a6361f0c | 3761 | release_sock(sk); |
80feaacb PWJ |
3762 | return 0; |
3763 | } | |
bfd5f4a3 SS |
3764 | case PACKET_VNET_HDR: |
3765 | { | |
3766 | int val; | |
3767 | ||
3768 | if (sock->type != SOCK_RAW) | |
3769 | return -EINVAL; | |
bfd5f4a3 SS |
3770 | if (optlen < sizeof(val)) |
3771 | return -EINVAL; | |
3772 | if (copy_from_user(&val, optval, sizeof(val))) | |
3773 | return -EFAULT; | |
3774 | ||
a6361f0c WB |
3775 | lock_sock(sk); |
3776 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3777 | ret = -EBUSY; | |
3778 | } else { | |
3779 | po->has_vnet_hdr = !!val; | |
3780 | ret = 0; | |
3781 | } | |
3782 | release_sock(sk); | |
3783 | return ret; | |
bfd5f4a3 | 3784 | } |
614f60fa SM |
3785 | case PACKET_TIMESTAMP: |
3786 | { | |
3787 | int val; | |
3788 | ||
3789 | if (optlen != sizeof(val)) | |
3790 | return -EINVAL; | |
3791 | if (copy_from_user(&val, optval, sizeof(val))) | |
3792 | return -EFAULT; | |
3793 | ||
3794 | po->tp_tstamp = val; | |
3795 | return 0; | |
3796 | } | |
dc99f600 DM |
3797 | case PACKET_FANOUT: |
3798 | { | |
3799 | int val; | |
3800 | ||
3801 | if (optlen != sizeof(val)) | |
3802 | return -EINVAL; | |
3803 | if (copy_from_user(&val, optval, sizeof(val))) | |
3804 | return -EFAULT; | |
3805 | ||
3806 | return fanout_add(sk, val & 0xffff, val >> 16); | |
3807 | } | |
47dceb8e WB |
3808 | case PACKET_FANOUT_DATA: |
3809 | { | |
3810 | if (!po->fanout) | |
3811 | return -EINVAL; | |
3812 | ||
3813 | return fanout_set_data(po, optval, optlen); | |
3814 | } | |
fa788d98 VW |
3815 | case PACKET_IGNORE_OUTGOING: |
3816 | { | |
3817 | int val; | |
3818 | ||
3819 | if (optlen != sizeof(val)) | |
3820 | return -EINVAL; | |
3821 | if (copy_from_user(&val, optval, sizeof(val))) | |
3822 | return -EFAULT; | |
3823 | if (val < 0 || val > 1) | |
3824 | return -EINVAL; | |
3825 | ||
3826 | po->prot_hook.ignore_outgoing = !!val; | |
3827 | return 0; | |
3828 | } | |
5920cd3a PC |
3829 | case PACKET_TX_HAS_OFF: |
3830 | { | |
3831 | unsigned int val; | |
3832 | ||
3833 | if (optlen != sizeof(val)) | |
3834 | return -EINVAL; | |
5920cd3a PC |
3835 | if (copy_from_user(&val, optval, sizeof(val))) |
3836 | return -EFAULT; | |
a6361f0c WB |
3837 | |
3838 | lock_sock(sk); | |
3839 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3840 | ret = -EBUSY; | |
3841 | } else { | |
3842 | po->tp_tx_has_off = !!val; | |
3843 | ret = 0; | |
3844 | } | |
3845 | release_sock(sk); | |
5920cd3a PC |
3846 | return 0; |
3847 | } | |
d346a3fa DB |
3848 | case PACKET_QDISC_BYPASS: |
3849 | { | |
3850 | int val; | |
3851 | ||
3852 | if (optlen != sizeof(val)) | |
3853 | return -EINVAL; | |
3854 | if (copy_from_user(&val, optval, sizeof(val))) | |
3855 | return -EFAULT; | |
3856 | ||
3857 | po->xmit = val ? packet_direct_xmit : dev_queue_xmit; | |
3858 | return 0; | |
3859 | } | |
1da177e4 LT |
3860 | default: |
3861 | return -ENOPROTOOPT; | |
3862 | } | |
3863 | } | |
3864 | ||
3865 | static int packet_getsockopt(struct socket *sock, int level, int optname, | |
3866 | char __user *optval, int __user *optlen) | |
3867 | { | |
3868 | int len; | |
c06fff6e | 3869 | int val, lv = sizeof(val); |
1da177e4 LT |
3870 | struct sock *sk = sock->sk; |
3871 | struct packet_sock *po = pkt_sk(sk); | |
c06fff6e | 3872 | void *data = &val; |
ee80fbf3 | 3873 | union tpacket_stats_u st; |
a9b63918 | 3874 | struct tpacket_rollover_stats rstats; |
1da177e4 LT |
3875 | |
3876 | if (level != SOL_PACKET) | |
3877 | return -ENOPROTOOPT; | |
3878 | ||
8ae55f04 KK |
3879 | if (get_user(len, optlen)) |
3880 | return -EFAULT; | |
1da177e4 LT |
3881 | |
3882 | if (len < 0) | |
3883 | return -EINVAL; | |
1ce4f28b | 3884 | |
69e3c75f | 3885 | switch (optname) { |
1da177e4 | 3886 | case PACKET_STATISTICS: |
1da177e4 | 3887 | spin_lock_bh(&sk->sk_receive_queue.lock); |
ee80fbf3 DB |
3888 | memcpy(&st, &po->stats, sizeof(st)); |
3889 | memset(&po->stats, 0, sizeof(po->stats)); | |
3890 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
3891 | ||
f6fb8f10 | 3892 | if (po->tp_version == TPACKET_V3) { |
c06fff6e | 3893 | lv = sizeof(struct tpacket_stats_v3); |
8bcdeaff | 3894 | st.stats3.tp_packets += st.stats3.tp_drops; |
ee80fbf3 | 3895 | data = &st.stats3; |
f6fb8f10 | 3896 | } else { |
c06fff6e | 3897 | lv = sizeof(struct tpacket_stats); |
8bcdeaff | 3898 | st.stats1.tp_packets += st.stats1.tp_drops; |
ee80fbf3 | 3899 | data = &st.stats1; |
f6fb8f10 | 3900 | } |
ee80fbf3 | 3901 | |
8dc41944 HX |
3902 | break; |
3903 | case PACKET_AUXDATA: | |
8dc41944 | 3904 | val = po->auxdata; |
80feaacb PWJ |
3905 | break; |
3906 | case PACKET_ORIGDEV: | |
80feaacb | 3907 | val = po->origdev; |
bfd5f4a3 SS |
3908 | break; |
3909 | case PACKET_VNET_HDR: | |
bfd5f4a3 | 3910 | val = po->has_vnet_hdr; |
1da177e4 | 3911 | break; |
bbd6ef87 | 3912 | case PACKET_VERSION: |
bbd6ef87 | 3913 | val = po->tp_version; |
bbd6ef87 PM |
3914 | break; |
3915 | case PACKET_HDRLEN: | |
3916 | if (len > sizeof(int)) | |
3917 | len = sizeof(int); | |
fd2c83b3 AP |
3918 | if (len < sizeof(int)) |
3919 | return -EINVAL; | |
bbd6ef87 PM |
3920 | if (copy_from_user(&val, optval, len)) |
3921 | return -EFAULT; | |
3922 | switch (val) { | |
3923 | case TPACKET_V1: | |
3924 | val = sizeof(struct tpacket_hdr); | |
3925 | break; | |
3926 | case TPACKET_V2: | |
3927 | val = sizeof(struct tpacket2_hdr); | |
3928 | break; | |
f6fb8f10 | 3929 | case TPACKET_V3: |
3930 | val = sizeof(struct tpacket3_hdr); | |
3931 | break; | |
bbd6ef87 PM |
3932 | default: |
3933 | return -EINVAL; | |
3934 | } | |
bbd6ef87 | 3935 | break; |
8913336a | 3936 | case PACKET_RESERVE: |
8913336a | 3937 | val = po->tp_reserve; |
8913336a | 3938 | break; |
69e3c75f | 3939 | case PACKET_LOSS: |
69e3c75f | 3940 | val = po->tp_loss; |
69e3c75f | 3941 | break; |
614f60fa | 3942 | case PACKET_TIMESTAMP: |
614f60fa | 3943 | val = po->tp_tstamp; |
614f60fa | 3944 | break; |
dc99f600 | 3945 | case PACKET_FANOUT: |
dc99f600 DM |
3946 | val = (po->fanout ? |
3947 | ((u32)po->fanout->id | | |
77f65ebd WB |
3948 | ((u32)po->fanout->type << 16) | |
3949 | ((u32)po->fanout->flags << 24)) : | |
dc99f600 | 3950 | 0); |
dc99f600 | 3951 | break; |
fa788d98 VW |
3952 | case PACKET_IGNORE_OUTGOING: |
3953 | val = po->prot_hook.ignore_outgoing; | |
3954 | break; | |
a9b63918 | 3955 | case PACKET_ROLLOVER_STATS: |
57f015f5 | 3956 | if (!po->rollover) |
a9b63918 | 3957 | return -EINVAL; |
57f015f5 MM |
3958 | rstats.tp_all = atomic_long_read(&po->rollover->num); |
3959 | rstats.tp_huge = atomic_long_read(&po->rollover->num_huge); | |
3960 | rstats.tp_failed = atomic_long_read(&po->rollover->num_failed); | |
3961 | data = &rstats; | |
3962 | lv = sizeof(rstats); | |
a9b63918 | 3963 | break; |
5920cd3a PC |
3964 | case PACKET_TX_HAS_OFF: |
3965 | val = po->tp_tx_has_off; | |
3966 | break; | |
d346a3fa DB |
3967 | case PACKET_QDISC_BYPASS: |
3968 | val = packet_use_direct_xmit(po); | |
3969 | break; | |
1da177e4 LT |
3970 | default: |
3971 | return -ENOPROTOOPT; | |
3972 | } | |
3973 | ||
c06fff6e ED |
3974 | if (len > lv) |
3975 | len = lv; | |
8ae55f04 KK |
3976 | if (put_user(len, optlen)) |
3977 | return -EFAULT; | |
8dc41944 HX |
3978 | if (copy_to_user(optval, data, len)) |
3979 | return -EFAULT; | |
8ae55f04 | 3980 | return 0; |
1da177e4 LT |
3981 | } |
3982 | ||
3983 | ||
719c44d3 WB |
3984 | #ifdef CONFIG_COMPAT |
3985 | static int compat_packet_setsockopt(struct socket *sock, int level, int optname, | |
3986 | char __user *optval, unsigned int optlen) | |
3987 | { | |
3988 | struct packet_sock *po = pkt_sk(sock->sk); | |
3989 | ||
3990 | if (level != SOL_PACKET) | |
3991 | return -ENOPROTOOPT; | |
3992 | ||
3993 | if (optname == PACKET_FANOUT_DATA && | |
3994 | po->fanout && po->fanout->type == PACKET_FANOUT_CBPF) { | |
3995 | optval = (char __user *)get_compat_bpf_fprog(optval); | |
3996 | if (!optval) | |
3997 | return -EFAULT; | |
3998 | optlen = sizeof(struct sock_fprog); | |
3999 | } | |
4000 | ||
4001 | return packet_setsockopt(sock, level, optname, optval, optlen); | |
4002 | } | |
4003 | #endif | |
4004 | ||
351638e7 JP |
4005 | static int packet_notifier(struct notifier_block *this, |
4006 | unsigned long msg, void *ptr) | |
1da177e4 LT |
4007 | { |
4008 | struct sock *sk; | |
351638e7 | 4009 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); |
c346dca1 | 4010 | struct net *net = dev_net(dev); |
1da177e4 | 4011 | |
808f5114 | 4012 | rcu_read_lock(); |
b67bfe0d | 4013 | sk_for_each_rcu(sk, &net->packet.sklist) { |
1da177e4 LT |
4014 | struct packet_sock *po = pkt_sk(sk); |
4015 | ||
4016 | switch (msg) { | |
4017 | case NETDEV_UNREGISTER: | |
1da177e4 | 4018 | if (po->mclist) |
82f17091 | 4019 | packet_dev_mclist_delete(dev, &po->mclist); |
a2efcfa0 DM |
4020 | /* fallthrough */ |
4021 | ||
1da177e4 LT |
4022 | case NETDEV_DOWN: |
4023 | if (dev->ifindex == po->ifindex) { | |
4024 | spin_lock(&po->bind_lock); | |
4025 | if (po->running) { | |
ce06b03e | 4026 | __unregister_prot_hook(sk, false); |
1da177e4 LT |
4027 | sk->sk_err = ENETDOWN; |
4028 | if (!sock_flag(sk, SOCK_DEAD)) | |
4029 | sk->sk_error_report(sk); | |
4030 | } | |
4031 | if (msg == NETDEV_UNREGISTER) { | |
66e56cd4 | 4032 | packet_cached_dev_reset(po); |
1da177e4 | 4033 | po->ifindex = -1; |
160ff18a BG |
4034 | if (po->prot_hook.dev) |
4035 | dev_put(po->prot_hook.dev); | |
1da177e4 LT |
4036 | po->prot_hook.dev = NULL; |
4037 | } | |
4038 | spin_unlock(&po->bind_lock); | |
4039 | } | |
4040 | break; | |
4041 | case NETDEV_UP: | |
808f5114 | 4042 | if (dev->ifindex == po->ifindex) { |
4043 | spin_lock(&po->bind_lock); | |
ce06b03e DM |
4044 | if (po->num) |
4045 | register_prot_hook(sk); | |
808f5114 | 4046 | spin_unlock(&po->bind_lock); |
1da177e4 | 4047 | } |
1da177e4 LT |
4048 | break; |
4049 | } | |
4050 | } | |
808f5114 | 4051 | rcu_read_unlock(); |
1da177e4 LT |
4052 | return NOTIFY_DONE; |
4053 | } | |
4054 | ||
4055 | ||
4056 | static int packet_ioctl(struct socket *sock, unsigned int cmd, | |
4057 | unsigned long arg) | |
4058 | { | |
4059 | struct sock *sk = sock->sk; | |
4060 | ||
69e3c75f | 4061 | switch (cmd) { |
40d4e3df ED |
4062 | case SIOCOUTQ: |
4063 | { | |
4064 | int amount = sk_wmem_alloc_get(sk); | |
31e6d363 | 4065 | |
40d4e3df ED |
4066 | return put_user(amount, (int __user *)arg); |
4067 | } | |
4068 | case SIOCINQ: | |
4069 | { | |
4070 | struct sk_buff *skb; | |
4071 | int amount = 0; | |
4072 | ||
4073 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
4074 | skb = skb_peek(&sk->sk_receive_queue); | |
4075 | if (skb) | |
4076 | amount = skb->len; | |
4077 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
4078 | return put_user(amount, (int __user *)arg); | |
4079 | } | |
4080 | case SIOCGSTAMP: | |
4081 | return sock_get_timestamp(sk, (struct timeval __user *)arg); | |
4082 | case SIOCGSTAMPNS: | |
4083 | return sock_get_timestampns(sk, (struct timespec __user *)arg); | |
1ce4f28b | 4084 | |
1da177e4 | 4085 | #ifdef CONFIG_INET |
40d4e3df ED |
4086 | case SIOCADDRT: |
4087 | case SIOCDELRT: | |
4088 | case SIOCDARP: | |
4089 | case SIOCGARP: | |
4090 | case SIOCSARP: | |
4091 | case SIOCGIFADDR: | |
4092 | case SIOCSIFADDR: | |
4093 | case SIOCGIFBRDADDR: | |
4094 | case SIOCSIFBRDADDR: | |
4095 | case SIOCGIFNETMASK: | |
4096 | case SIOCSIFNETMASK: | |
4097 | case SIOCGIFDSTADDR: | |
4098 | case SIOCSIFDSTADDR: | |
4099 | case SIOCSIFFLAGS: | |
40d4e3df | 4100 | return inet_dgram_ops.ioctl(sock, cmd, arg); |
1da177e4 LT |
4101 | #endif |
4102 | ||
40d4e3df ED |
4103 | default: |
4104 | return -ENOIOCTLCMD; | |
1da177e4 LT |
4105 | } |
4106 | return 0; | |
4107 | } | |
4108 | ||
a11e1d43 LT |
4109 | static __poll_t packet_poll(struct file *file, struct socket *sock, |
4110 | poll_table *wait) | |
1da177e4 LT |
4111 | { |
4112 | struct sock *sk = sock->sk; | |
4113 | struct packet_sock *po = pkt_sk(sk); | |
a11e1d43 | 4114 | __poll_t mask = datagram_poll(file, sock, wait); |
1da177e4 LT |
4115 | |
4116 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
69e3c75f | 4117 | if (po->rx_ring.pg_vec) { |
f6fb8f10 | 4118 | if (!packet_previous_rx_frame(po, &po->rx_ring, |
4119 | TP_STATUS_KERNEL)) | |
a9a08845 | 4120 | mask |= EPOLLIN | EPOLLRDNORM; |
1da177e4 | 4121 | } |
2ccdbaa6 | 4122 | if (po->pressure && __packet_rcv_has_room(po, NULL) == ROOM_NORMAL) |
54d7c01d | 4123 | po->pressure = 0; |
1da177e4 | 4124 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
69e3c75f JB |
4125 | spin_lock_bh(&sk->sk_write_queue.lock); |
4126 | if (po->tx_ring.pg_vec) { | |
4127 | if (packet_current_frame(po, &po->tx_ring, TP_STATUS_AVAILABLE)) | |
a9a08845 | 4128 | mask |= EPOLLOUT | EPOLLWRNORM; |
69e3c75f JB |
4129 | } |
4130 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
1da177e4 LT |
4131 | return mask; |
4132 | } | |
4133 | ||
4134 | ||
4135 | /* Dirty? Well, I still did not learn better way to account | |
4136 | * for user mmaps. | |
4137 | */ | |
4138 | ||
4139 | static void packet_mm_open(struct vm_area_struct *vma) | |
4140 | { | |
4141 | struct file *file = vma->vm_file; | |
40d4e3df | 4142 | struct socket *sock = file->private_data; |
1da177e4 | 4143 | struct sock *sk = sock->sk; |
1ce4f28b | 4144 | |
1da177e4 LT |
4145 | if (sk) |
4146 | atomic_inc(&pkt_sk(sk)->mapped); | |
4147 | } | |
4148 | ||
4149 | static void packet_mm_close(struct vm_area_struct *vma) | |
4150 | { | |
4151 | struct file *file = vma->vm_file; | |
40d4e3df | 4152 | struct socket *sock = file->private_data; |
1da177e4 | 4153 | struct sock *sk = sock->sk; |
1ce4f28b | 4154 | |
1da177e4 LT |
4155 | if (sk) |
4156 | atomic_dec(&pkt_sk(sk)->mapped); | |
4157 | } | |
4158 | ||
f0f37e2f | 4159 | static const struct vm_operations_struct packet_mmap_ops = { |
40d4e3df ED |
4160 | .open = packet_mm_open, |
4161 | .close = packet_mm_close, | |
1da177e4 LT |
4162 | }; |
4163 | ||
3a7ad063 ED |
4164 | static void free_pg_vec(struct pgv *pg_vec, unsigned int order, |
4165 | unsigned int len) | |
1da177e4 LT |
4166 | { |
4167 | int i; | |
4168 | ||
4ebf0ae2 | 4169 | for (i = 0; i < len; i++) { |
0e3125c7 | 4170 | if (likely(pg_vec[i].buffer)) { |
3a7ad063 ED |
4171 | if (is_vmalloc_addr(pg_vec[i].buffer)) |
4172 | vfree(pg_vec[i].buffer); | |
4173 | else | |
4174 | free_pages((unsigned long)pg_vec[i].buffer, | |
4175 | order); | |
0e3125c7 NH |
4176 | pg_vec[i].buffer = NULL; |
4177 | } | |
1da177e4 LT |
4178 | } |
4179 | kfree(pg_vec); | |
4180 | } | |
4181 | ||
3a7ad063 | 4182 | static char *alloc_one_pg_vec_page(unsigned long order) |
4ebf0ae2 | 4183 | { |
f0d4eb29 | 4184 | char *buffer; |
3a7ad063 ED |
4185 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | |
4186 | __GFP_ZERO | __GFP_NOWARN | __GFP_NORETRY; | |
0e3125c7 | 4187 | |
3a7ad063 | 4188 | buffer = (char *) __get_free_pages(gfp_flags, order); |
0e3125c7 NH |
4189 | if (buffer) |
4190 | return buffer; | |
4191 | ||
3a7ad063 ED |
4192 | /* __get_free_pages failed, fall back to vmalloc */ |
4193 | buffer = vzalloc(array_size((1 << order), PAGE_SIZE)); | |
4194 | if (buffer) | |
4195 | return buffer; | |
0e3125c7 | 4196 | |
3a7ad063 ED |
4197 | /* vmalloc failed, lets dig into swap here */ |
4198 | gfp_flags &= ~__GFP_NORETRY; | |
4199 | buffer = (char *) __get_free_pages(gfp_flags, order); | |
4200 | if (buffer) | |
4201 | return buffer; | |
4202 | ||
4203 | /* complete and utter failure */ | |
4204 | return NULL; | |
4ebf0ae2 DM |
4205 | } |
4206 | ||
3a7ad063 | 4207 | static struct pgv *alloc_pg_vec(struct tpacket_req *req, int order) |
4ebf0ae2 DM |
4208 | { |
4209 | unsigned int block_nr = req->tp_block_nr; | |
0e3125c7 | 4210 | struct pgv *pg_vec; |
4ebf0ae2 DM |
4211 | int i; |
4212 | ||
0e3125c7 | 4213 | pg_vec = kcalloc(block_nr, sizeof(struct pgv), GFP_KERNEL); |
4ebf0ae2 DM |
4214 | if (unlikely(!pg_vec)) |
4215 | goto out; | |
4216 | ||
4217 | for (i = 0; i < block_nr; i++) { | |
3a7ad063 | 4218 | pg_vec[i].buffer = alloc_one_pg_vec_page(order); |
0e3125c7 | 4219 | if (unlikely(!pg_vec[i].buffer)) |
4ebf0ae2 DM |
4220 | goto out_free_pgvec; |
4221 | } | |
4222 | ||
4223 | out: | |
4224 | return pg_vec; | |
4225 | ||
4226 | out_free_pgvec: | |
3a7ad063 | 4227 | free_pg_vec(pg_vec, order, block_nr); |
4ebf0ae2 DM |
4228 | pg_vec = NULL; |
4229 | goto out; | |
4230 | } | |
1da177e4 | 4231 | |
f6fb8f10 | 4232 | static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u, |
69e3c75f | 4233 | int closing, int tx_ring) |
1da177e4 | 4234 | { |
0e3125c7 | 4235 | struct pgv *pg_vec = NULL; |
1da177e4 | 4236 | struct packet_sock *po = pkt_sk(sk); |
3a7ad063 | 4237 | int was_running, order = 0; |
69e3c75f JB |
4238 | struct packet_ring_buffer *rb; |
4239 | struct sk_buff_head *rb_queue; | |
0e11c91e | 4240 | __be16 num; |
f6fb8f10 | 4241 | int err = -EINVAL; |
4242 | /* Added to avoid minimal code churn */ | |
4243 | struct tpacket_req *req = &req_u->req; | |
4244 | ||
69e3c75f JB |
4245 | rb = tx_ring ? &po->tx_ring : &po->rx_ring; |
4246 | rb_queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
1da177e4 | 4247 | |
69e3c75f JB |
4248 | err = -EBUSY; |
4249 | if (!closing) { | |
4250 | if (atomic_read(&po->mapped)) | |
4251 | goto out; | |
b0138408 | 4252 | if (packet_read_pending(rb)) |
69e3c75f JB |
4253 | goto out; |
4254 | } | |
1da177e4 | 4255 | |
69e3c75f | 4256 | if (req->tp_block_nr) { |
4576cd46 WB |
4257 | unsigned int min_frame_size; |
4258 | ||
69e3c75f JB |
4259 | /* Sanity tests and some calculations */ |
4260 | err = -EBUSY; | |
4261 | if (unlikely(rb->pg_vec)) | |
4262 | goto out; | |
1da177e4 | 4263 | |
bbd6ef87 PM |
4264 | switch (po->tp_version) { |
4265 | case TPACKET_V1: | |
4266 | po->tp_hdrlen = TPACKET_HDRLEN; | |
4267 | break; | |
4268 | case TPACKET_V2: | |
4269 | po->tp_hdrlen = TPACKET2_HDRLEN; | |
4270 | break; | |
f6fb8f10 | 4271 | case TPACKET_V3: |
4272 | po->tp_hdrlen = TPACKET3_HDRLEN; | |
4273 | break; | |
bbd6ef87 PM |
4274 | } |
4275 | ||
69e3c75f | 4276 | err = -EINVAL; |
4ebf0ae2 | 4277 | if (unlikely((int)req->tp_block_size <= 0)) |
69e3c75f | 4278 | goto out; |
90836b67 | 4279 | if (unlikely(!PAGE_ALIGNED(req->tp_block_size))) |
69e3c75f | 4280 | goto out; |
4576cd46 | 4281 | min_frame_size = po->tp_hdrlen + po->tp_reserve; |
dc808110 | 4282 | if (po->tp_version >= TPACKET_V3 && |
4576cd46 WB |
4283 | req->tp_block_size < |
4284 | BLK_PLUS_PRIV((u64)req_u->req3.tp_sizeof_priv) + min_frame_size) | |
dc808110 | 4285 | goto out; |
4576cd46 | 4286 | if (unlikely(req->tp_frame_size < min_frame_size)) |
69e3c75f | 4287 | goto out; |
4ebf0ae2 | 4288 | if (unlikely(req->tp_frame_size & (TPACKET_ALIGNMENT - 1))) |
69e3c75f | 4289 | goto out; |
1da177e4 | 4290 | |
4194b491 TK |
4291 | rb->frames_per_block = req->tp_block_size / req->tp_frame_size; |
4292 | if (unlikely(rb->frames_per_block == 0)) | |
69e3c75f | 4293 | goto out; |
8f8d28e4 AK |
4294 | if (unlikely(req->tp_block_size > UINT_MAX / req->tp_block_nr)) |
4295 | goto out; | |
69e3c75f JB |
4296 | if (unlikely((rb->frames_per_block * req->tp_block_nr) != |
4297 | req->tp_frame_nr)) | |
4298 | goto out; | |
1da177e4 LT |
4299 | |
4300 | err = -ENOMEM; | |
3a7ad063 ED |
4301 | order = get_order(req->tp_block_size); |
4302 | pg_vec = alloc_pg_vec(req, order); | |
4ebf0ae2 | 4303 | if (unlikely(!pg_vec)) |
1da177e4 | 4304 | goto out; |
f6fb8f10 | 4305 | switch (po->tp_version) { |
4306 | case TPACKET_V3: | |
7f953ab2 SV |
4307 | /* Block transmit is not supported yet */ |
4308 | if (!tx_ring) { | |
e8e85cc5 | 4309 | init_prb_bdqc(po, rb, pg_vec, req_u); |
7f953ab2 SV |
4310 | } else { |
4311 | struct tpacket_req3 *req3 = &req_u->req3; | |
4312 | ||
4313 | if (req3->tp_retire_blk_tov || | |
4314 | req3->tp_sizeof_priv || | |
4315 | req3->tp_feature_req_word) { | |
4316 | err = -EINVAL; | |
4317 | goto out; | |
4318 | } | |
4319 | } | |
d7cf0c34 | 4320 | break; |
f6fb8f10 | 4321 | default: |
4322 | break; | |
4323 | } | |
69e3c75f JB |
4324 | } |
4325 | /* Done */ | |
4326 | else { | |
4327 | err = -EINVAL; | |
4ebf0ae2 | 4328 | if (unlikely(req->tp_frame_nr)) |
69e3c75f | 4329 | goto out; |
1da177e4 LT |
4330 | } |
4331 | ||
1da177e4 LT |
4332 | |
4333 | /* Detach socket from network */ | |
4334 | spin_lock(&po->bind_lock); | |
4335 | was_running = po->running; | |
4336 | num = po->num; | |
4337 | if (was_running) { | |
1da177e4 | 4338 | po->num = 0; |
ce06b03e | 4339 | __unregister_prot_hook(sk, false); |
1da177e4 LT |
4340 | } |
4341 | spin_unlock(&po->bind_lock); | |
1ce4f28b | 4342 | |
1da177e4 LT |
4343 | synchronize_net(); |
4344 | ||
4345 | err = -EBUSY; | |
905db440 | 4346 | mutex_lock(&po->pg_vec_lock); |
1da177e4 LT |
4347 | if (closing || atomic_read(&po->mapped) == 0) { |
4348 | err = 0; | |
69e3c75f | 4349 | spin_lock_bh(&rb_queue->lock); |
c053fd96 | 4350 | swap(rb->pg_vec, pg_vec); |
69e3c75f JB |
4351 | rb->frame_max = (req->tp_frame_nr - 1); |
4352 | rb->head = 0; | |
4353 | rb->frame_size = req->tp_frame_size; | |
4354 | spin_unlock_bh(&rb_queue->lock); | |
4355 | ||
3a7ad063 | 4356 | swap(rb->pg_vec_order, order); |
c053fd96 | 4357 | swap(rb->pg_vec_len, req->tp_block_nr); |
69e3c75f JB |
4358 | |
4359 | rb->pg_vec_pages = req->tp_block_size/PAGE_SIZE; | |
4360 | po->prot_hook.func = (po->rx_ring.pg_vec) ? | |
4361 | tpacket_rcv : packet_rcv; | |
4362 | skb_queue_purge(rb_queue); | |
1da177e4 | 4363 | if (atomic_read(&po->mapped)) |
40d4e3df ED |
4364 | pr_err("packet_mmap: vma is busy: %d\n", |
4365 | atomic_read(&po->mapped)); | |
1da177e4 | 4366 | } |
905db440 | 4367 | mutex_unlock(&po->pg_vec_lock); |
1da177e4 LT |
4368 | |
4369 | spin_lock(&po->bind_lock); | |
ce06b03e | 4370 | if (was_running) { |
1da177e4 | 4371 | po->num = num; |
ce06b03e | 4372 | register_prot_hook(sk); |
1da177e4 LT |
4373 | } |
4374 | spin_unlock(&po->bind_lock); | |
c800aaf8 | 4375 | if (pg_vec && (po->tp_version > TPACKET_V2)) { |
f6fb8f10 | 4376 | /* Because we don't support block-based V3 on tx-ring */ |
4377 | if (!tx_ring) | |
73d0fcf2 | 4378 | prb_shutdown_retire_blk_timer(po, rb_queue); |
f6fb8f10 | 4379 | } |
1da177e4 | 4380 | |
1da177e4 | 4381 | if (pg_vec) |
3a7ad063 | 4382 | free_pg_vec(pg_vec, order, req->tp_block_nr); |
1da177e4 LT |
4383 | out: |
4384 | return err; | |
4385 | } | |
4386 | ||
69e3c75f JB |
4387 | static int packet_mmap(struct file *file, struct socket *sock, |
4388 | struct vm_area_struct *vma) | |
1da177e4 LT |
4389 | { |
4390 | struct sock *sk = sock->sk; | |
4391 | struct packet_sock *po = pkt_sk(sk); | |
69e3c75f JB |
4392 | unsigned long size, expected_size; |
4393 | struct packet_ring_buffer *rb; | |
1da177e4 LT |
4394 | unsigned long start; |
4395 | int err = -EINVAL; | |
4396 | int i; | |
4397 | ||
4398 | if (vma->vm_pgoff) | |
4399 | return -EINVAL; | |
4400 | ||
905db440 | 4401 | mutex_lock(&po->pg_vec_lock); |
69e3c75f JB |
4402 | |
4403 | expected_size = 0; | |
4404 | for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) { | |
4405 | if (rb->pg_vec) { | |
4406 | expected_size += rb->pg_vec_len | |
4407 | * rb->pg_vec_pages | |
4408 | * PAGE_SIZE; | |
4409 | } | |
4410 | } | |
4411 | ||
4412 | if (expected_size == 0) | |
1da177e4 | 4413 | goto out; |
69e3c75f JB |
4414 | |
4415 | size = vma->vm_end - vma->vm_start; | |
4416 | if (size != expected_size) | |
1da177e4 LT |
4417 | goto out; |
4418 | ||
1da177e4 | 4419 | start = vma->vm_start; |
69e3c75f JB |
4420 | for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) { |
4421 | if (rb->pg_vec == NULL) | |
4422 | continue; | |
4423 | ||
4424 | for (i = 0; i < rb->pg_vec_len; i++) { | |
0e3125c7 NH |
4425 | struct page *page; |
4426 | void *kaddr = rb->pg_vec[i].buffer; | |
69e3c75f JB |
4427 | int pg_num; |
4428 | ||
c56b4d90 CG |
4429 | for (pg_num = 0; pg_num < rb->pg_vec_pages; pg_num++) { |
4430 | page = pgv_to_page(kaddr); | |
69e3c75f JB |
4431 | err = vm_insert_page(vma, start, page); |
4432 | if (unlikely(err)) | |
4433 | goto out; | |
4434 | start += PAGE_SIZE; | |
0e3125c7 | 4435 | kaddr += PAGE_SIZE; |
69e3c75f | 4436 | } |
4ebf0ae2 | 4437 | } |
1da177e4 | 4438 | } |
69e3c75f | 4439 | |
4ebf0ae2 | 4440 | atomic_inc(&po->mapped); |
1da177e4 LT |
4441 | vma->vm_ops = &packet_mmap_ops; |
4442 | err = 0; | |
4443 | ||
4444 | out: | |
905db440 | 4445 | mutex_unlock(&po->pg_vec_lock); |
1da177e4 LT |
4446 | return err; |
4447 | } | |
1da177e4 | 4448 | |
90ddc4f0 | 4449 | static const struct proto_ops packet_ops_spkt = { |
1da177e4 LT |
4450 | .family = PF_PACKET, |
4451 | .owner = THIS_MODULE, | |
4452 | .release = packet_release, | |
4453 | .bind = packet_bind_spkt, | |
4454 | .connect = sock_no_connect, | |
4455 | .socketpair = sock_no_socketpair, | |
4456 | .accept = sock_no_accept, | |
4457 | .getname = packet_getname_spkt, | |
a11e1d43 | 4458 | .poll = datagram_poll, |
1da177e4 LT |
4459 | .ioctl = packet_ioctl, |
4460 | .listen = sock_no_listen, | |
4461 | .shutdown = sock_no_shutdown, | |
4462 | .setsockopt = sock_no_setsockopt, | |
4463 | .getsockopt = sock_no_getsockopt, | |
4464 | .sendmsg = packet_sendmsg_spkt, | |
4465 | .recvmsg = packet_recvmsg, | |
4466 | .mmap = sock_no_mmap, | |
4467 | .sendpage = sock_no_sendpage, | |
4468 | }; | |
1da177e4 | 4469 | |
90ddc4f0 | 4470 | static const struct proto_ops packet_ops = { |
1da177e4 LT |
4471 | .family = PF_PACKET, |
4472 | .owner = THIS_MODULE, | |
4473 | .release = packet_release, | |
4474 | .bind = packet_bind, | |
4475 | .connect = sock_no_connect, | |
4476 | .socketpair = sock_no_socketpair, | |
4477 | .accept = sock_no_accept, | |
1ce4f28b | 4478 | .getname = packet_getname, |
a11e1d43 | 4479 | .poll = packet_poll, |
1da177e4 LT |
4480 | .ioctl = packet_ioctl, |
4481 | .listen = sock_no_listen, | |
4482 | .shutdown = sock_no_shutdown, | |
4483 | .setsockopt = packet_setsockopt, | |
4484 | .getsockopt = packet_getsockopt, | |
719c44d3 WB |
4485 | #ifdef CONFIG_COMPAT |
4486 | .compat_setsockopt = compat_packet_setsockopt, | |
4487 | #endif | |
1da177e4 LT |
4488 | .sendmsg = packet_sendmsg, |
4489 | .recvmsg = packet_recvmsg, | |
4490 | .mmap = packet_mmap, | |
4491 | .sendpage = sock_no_sendpage, | |
4492 | }; | |
4493 | ||
ec1b4cf7 | 4494 | static const struct net_proto_family packet_family_ops = { |
1da177e4 LT |
4495 | .family = PF_PACKET, |
4496 | .create = packet_create, | |
4497 | .owner = THIS_MODULE, | |
4498 | }; | |
4499 | ||
4500 | static struct notifier_block packet_netdev_notifier = { | |
40d4e3df | 4501 | .notifier_call = packet_notifier, |
1da177e4 LT |
4502 | }; |
4503 | ||
4504 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
4505 | |
4506 | static void *packet_seq_start(struct seq_file *seq, loff_t *pos) | |
808f5114 | 4507 | __acquires(RCU) |
1da177e4 | 4508 | { |
e372c414 | 4509 | struct net *net = seq_file_net(seq); |
808f5114 | 4510 | |
4511 | rcu_read_lock(); | |
4512 | return seq_hlist_start_head_rcu(&net->packet.sklist, *pos); | |
1da177e4 LT |
4513 | } |
4514 | ||
4515 | static void *packet_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
4516 | { | |
1bf40954 | 4517 | struct net *net = seq_file_net(seq); |
808f5114 | 4518 | return seq_hlist_next_rcu(v, &net->packet.sklist, pos); |
1da177e4 LT |
4519 | } |
4520 | ||
4521 | static void packet_seq_stop(struct seq_file *seq, void *v) | |
808f5114 | 4522 | __releases(RCU) |
1da177e4 | 4523 | { |
808f5114 | 4524 | rcu_read_unlock(); |
1da177e4 LT |
4525 | } |
4526 | ||
1ce4f28b | 4527 | static int packet_seq_show(struct seq_file *seq, void *v) |
1da177e4 LT |
4528 | { |
4529 | if (v == SEQ_START_TOKEN) | |
4530 | seq_puts(seq, "sk RefCnt Type Proto Iface R Rmem User Inode\n"); | |
4531 | else { | |
b7ceabd9 | 4532 | struct sock *s = sk_entry(v); |
1da177e4 LT |
4533 | const struct packet_sock *po = pkt_sk(s); |
4534 | ||
4535 | seq_printf(seq, | |
71338aa7 | 4536 | "%pK %-6d %-4d %04x %-5d %1d %-6u %-6u %-6lu\n", |
1da177e4 | 4537 | s, |
41c6d650 | 4538 | refcount_read(&s->sk_refcnt), |
1da177e4 LT |
4539 | s->sk_type, |
4540 | ntohs(po->num), | |
4541 | po->ifindex, | |
4542 | po->running, | |
4543 | atomic_read(&s->sk_rmem_alloc), | |
a7cb5a49 | 4544 | from_kuid_munged(seq_user_ns(seq), sock_i_uid(s)), |
40d4e3df | 4545 | sock_i_ino(s)); |
1da177e4 LT |
4546 | } |
4547 | ||
4548 | return 0; | |
4549 | } | |
4550 | ||
56b3d975 | 4551 | static const struct seq_operations packet_seq_ops = { |
1da177e4 LT |
4552 | .start = packet_seq_start, |
4553 | .next = packet_seq_next, | |
4554 | .stop = packet_seq_stop, | |
4555 | .show = packet_seq_show, | |
4556 | }; | |
1da177e4 LT |
4557 | #endif |
4558 | ||
2c8c1e72 | 4559 | static int __net_init packet_net_init(struct net *net) |
d12d01d6 | 4560 | { |
0fa7fa98 | 4561 | mutex_init(&net->packet.sklist_lock); |
2aaef4e4 | 4562 | INIT_HLIST_HEAD(&net->packet.sklist); |
d12d01d6 | 4563 | |
c3506372 CH |
4564 | if (!proc_create_net("packet", 0, net->proc_net, &packet_seq_ops, |
4565 | sizeof(struct seq_net_private))) | |
d12d01d6 DL |
4566 | return -ENOMEM; |
4567 | ||
4568 | return 0; | |
4569 | } | |
4570 | ||
2c8c1e72 | 4571 | static void __net_exit packet_net_exit(struct net *net) |
d12d01d6 | 4572 | { |
ece31ffd | 4573 | remove_proc_entry("packet", net->proc_net); |
669f8f1a | 4574 | WARN_ON_ONCE(!hlist_empty(&net->packet.sklist)); |
d12d01d6 DL |
4575 | } |
4576 | ||
4577 | static struct pernet_operations packet_net_ops = { | |
4578 | .init = packet_net_init, | |
4579 | .exit = packet_net_exit, | |
4580 | }; | |
4581 | ||
4582 | ||
1da177e4 LT |
4583 | static void __exit packet_exit(void) |
4584 | { | |
1da177e4 | 4585 | unregister_netdevice_notifier(&packet_netdev_notifier); |
d12d01d6 | 4586 | unregister_pernet_subsys(&packet_net_ops); |
1da177e4 LT |
4587 | sock_unregister(PF_PACKET); |
4588 | proto_unregister(&packet_proto); | |
4589 | } | |
4590 | ||
4591 | static int __init packet_init(void) | |
4592 | { | |
4593 | int rc = proto_register(&packet_proto, 0); | |
4594 | ||
4595 | if (rc != 0) | |
4596 | goto out; | |
4597 | ||
4598 | sock_register(&packet_family_ops); | |
d12d01d6 | 4599 | register_pernet_subsys(&packet_net_ops); |
1da177e4 | 4600 | register_netdevice_notifier(&packet_netdev_notifier); |
1da177e4 LT |
4601 | out: |
4602 | return rc; | |
4603 | } | |
4604 | ||
4605 | module_init(packet_init); | |
4606 | module_exit(packet_exit); | |
4607 | MODULE_LICENSE("GPL"); | |
4608 | MODULE_ALIAS_NETPROTO(PF_PACKET); |