]>
Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
1da177e4 LT |
2 | /* |
3 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
4 | * operating system. INET is implemented using the BSD Socket | |
5 | * interface as the means of communication with the user level. | |
6 | * | |
7 | * PACKET - implements raw packet sockets. | |
8 | * | |
02c30a84 | 9 | * Authors: Ross Biro |
1da177e4 LT |
10 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
11 | * Alan Cox, <gw4pts@gw4pts.ampr.org> | |
12 | * | |
1ce4f28b | 13 | * Fixes: |
1da177e4 LT |
14 | * Alan Cox : verify_area() now used correctly |
15 | * Alan Cox : new skbuff lists, look ma no backlogs! | |
16 | * Alan Cox : tidied skbuff lists. | |
17 | * Alan Cox : Now uses generic datagram routines I | |
18 | * added. Also fixed the peek/read crash | |
19 | * from all old Linux datagram code. | |
20 | * Alan Cox : Uses the improved datagram code. | |
21 | * Alan Cox : Added NULL's for socket options. | |
22 | * Alan Cox : Re-commented the code. | |
23 | * Alan Cox : Use new kernel side addressing | |
24 | * Rob Janssen : Correct MTU usage. | |
25 | * Dave Platt : Counter leaks caused by incorrect | |
26 | * interrupt locking and some slightly | |
27 | * dubious gcc output. Can you read | |
28 | * compiler: it said _VOLATILE_ | |
29 | * Richard Kooijman : Timestamp fixes. | |
30 | * Alan Cox : New buffers. Use sk->mac.raw. | |
31 | * Alan Cox : sendmsg/recvmsg support. | |
32 | * Alan Cox : Protocol setting support | |
33 | * Alexey Kuznetsov : Untied from IPv4 stack. | |
34 | * Cyrus Durgin : Fixed kerneld for kmod. | |
35 | * Michal Ostrowski : Module initialization cleanup. | |
1ce4f28b | 36 | * Ulises Alonso : Frame number limit removal and |
1da177e4 | 37 | * packet_set_ring memory leak. |
0fb375fb EB |
38 | * Eric Biederman : Allow for > 8 byte hardware addresses. |
39 | * The convention is that longer addresses | |
40 | * will simply extend the hardware address | |
1ce4f28b | 41 | * byte arrays at the end of sockaddr_ll |
0fb375fb | 42 | * and packet_mreq. |
69e3c75f | 43 | * Johann Baudy : Added TX RING. |
f6fb8f10 | 44 | * Chetan Loke : Implemented TPACKET_V3 block abstraction |
45 | * layer. | |
46 | * Copyright (C) 2011, <lokec@ccs.neu.edu> | |
1da177e4 | 47 | */ |
1ce4f28b | 48 | |
1da177e4 | 49 | #include <linux/types.h> |
1da177e4 | 50 | #include <linux/mm.h> |
4fc268d2 | 51 | #include <linux/capability.h> |
1da177e4 LT |
52 | #include <linux/fcntl.h> |
53 | #include <linux/socket.h> | |
54 | #include <linux/in.h> | |
55 | #include <linux/inet.h> | |
56 | #include <linux/netdevice.h> | |
57 | #include <linux/if_packet.h> | |
58 | #include <linux/wireless.h> | |
ffbc6111 | 59 | #include <linux/kernel.h> |
1da177e4 | 60 | #include <linux/kmod.h> |
5a0e3ad6 | 61 | #include <linux/slab.h> |
0e3125c7 | 62 | #include <linux/vmalloc.h> |
457c4cbc | 63 | #include <net/net_namespace.h> |
1da177e4 LT |
64 | #include <net/ip.h> |
65 | #include <net/protocol.h> | |
66 | #include <linux/skbuff.h> | |
67 | #include <net/sock.h> | |
68 | #include <linux/errno.h> | |
69 | #include <linux/timer.h> | |
7c0f6ba6 | 70 | #include <linux/uaccess.h> |
1da177e4 LT |
71 | #include <asm/ioctls.h> |
72 | #include <asm/page.h> | |
a1f8e7f7 | 73 | #include <asm/cacheflush.h> |
1da177e4 LT |
74 | #include <asm/io.h> |
75 | #include <linux/proc_fs.h> | |
76 | #include <linux/seq_file.h> | |
77 | #include <linux/poll.h> | |
78 | #include <linux/module.h> | |
79 | #include <linux/init.h> | |
905db440 | 80 | #include <linux/mutex.h> |
05423b24 | 81 | #include <linux/if_vlan.h> |
bfd5f4a3 | 82 | #include <linux/virtio_net.h> |
ed85b565 | 83 | #include <linux/errqueue.h> |
614f60fa | 84 | #include <linux/net_tstamp.h> |
b0138408 | 85 | #include <linux/percpu.h> |
1da177e4 LT |
86 | #ifdef CONFIG_INET |
87 | #include <net/inet_common.h> | |
88 | #endif | |
47dceb8e | 89 | #include <linux/bpf.h> |
719c44d3 | 90 | #include <net/compat.h> |
1da177e4 | 91 | |
2787b04b PE |
92 | #include "internal.h" |
93 | ||
1da177e4 LT |
94 | /* |
95 | Assumptions: | |
b4c58814 XH |
96 | - If the device has no dev->header_ops, there is no LL header visible |
97 | above the device. In this case, its hard_header_len should be 0. | |
98 | The device may prepend its own header internally. In this case, its | |
99 | needed_headroom should be set to the space needed for it to add its | |
100 | internal header. | |
101 | For example, a WiFi driver pretending to be an Ethernet driver should | |
102 | set its hard_header_len to be the Ethernet header length, and set its | |
103 | needed_headroom to be (the real WiFi header length - the fake Ethernet | |
104 | header length). | |
1da177e4 LT |
105 | - packet socket receives packets with pulled ll header, |
106 | so that SOCK_RAW should push it back. | |
107 | ||
108 | On receive: | |
109 | ----------- | |
110 | ||
b79a80bd | 111 | Incoming, dev->header_ops != NULL |
b0e380b1 ACM |
112 | mac_header -> ll header |
113 | data -> data | |
1da177e4 | 114 | |
b79a80bd | 115 | Outgoing, dev->header_ops != NULL |
b0e380b1 ACM |
116 | mac_header -> ll header |
117 | data -> ll header | |
1da177e4 | 118 | |
b79a80bd XH |
119 | Incoming, dev->header_ops == NULL |
120 | mac_header -> data | |
121 | However drivers often make it point to the ll header. | |
122 | This is incorrect because the ll header should be invisible to us. | |
b0e380b1 | 123 | data -> data |
1da177e4 | 124 | |
b79a80bd XH |
125 | Outgoing, dev->header_ops == NULL |
126 | mac_header -> data. ll header is invisible to us. | |
b0e380b1 | 127 | data -> data |
1da177e4 LT |
128 | |
129 | Resume | |
b79a80bd XH |
130 | If dev->header_ops == NULL we are unable to restore the ll header, |
131 | because it is invisible to us. | |
1da177e4 LT |
132 | |
133 | ||
134 | On transmit: | |
135 | ------------ | |
136 | ||
b79a80bd | 137 | dev->header_ops != NULL |
b0e380b1 ACM |
138 | mac_header -> ll header |
139 | data -> ll header | |
1da177e4 | 140 | |
b79a80bd | 141 | dev->header_ops == NULL (ll header is invisible to us) |
b0e380b1 ACM |
142 | mac_header -> data |
143 | data -> data | |
1da177e4 | 144 | |
09599729 | 145 | We should set network_header on output to the correct position, |
1da177e4 LT |
146 | packet classifier depends on it. |
147 | */ | |
148 | ||
1da177e4 LT |
149 | /* Private packet socket structures. */ |
150 | ||
0fb375fb EB |
151 | /* identical to struct packet_mreq except it has |
152 | * a longer address field. | |
153 | */ | |
40d4e3df | 154 | struct packet_mreq_max { |
0fb375fb EB |
155 | int mr_ifindex; |
156 | unsigned short mr_type; | |
157 | unsigned short mr_alen; | |
158 | unsigned char mr_address[MAX_ADDR_LEN]; | |
1da177e4 | 159 | }; |
a2efcfa0 | 160 | |
184f489e DB |
161 | union tpacket_uhdr { |
162 | struct tpacket_hdr *h1; | |
163 | struct tpacket2_hdr *h2; | |
164 | struct tpacket3_hdr *h3; | |
165 | void *raw; | |
166 | }; | |
167 | ||
f6fb8f10 | 168 | static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u, |
69e3c75f JB |
169 | int closing, int tx_ring); |
170 | ||
f6fb8f10 | 171 | #define V3_ALIGNMENT (8) |
172 | ||
bc59ba39 | 173 | #define BLK_HDR_LEN (ALIGN(sizeof(struct tpacket_block_desc), V3_ALIGNMENT)) |
f6fb8f10 | 174 | |
175 | #define BLK_PLUS_PRIV(sz_of_priv) \ | |
176 | (BLK_HDR_LEN + ALIGN((sz_of_priv), V3_ALIGNMENT)) | |
177 | ||
f6fb8f10 | 178 | #define BLOCK_STATUS(x) ((x)->hdr.bh1.block_status) |
179 | #define BLOCK_NUM_PKTS(x) ((x)->hdr.bh1.num_pkts) | |
180 | #define BLOCK_O2FP(x) ((x)->hdr.bh1.offset_to_first_pkt) | |
181 | #define BLOCK_LEN(x) ((x)->hdr.bh1.blk_len) | |
182 | #define BLOCK_SNUM(x) ((x)->hdr.bh1.seq_num) | |
183 | #define BLOCK_O2PRIV(x) ((x)->offset_to_priv) | |
f6fb8f10 | 184 | |
69e3c75f | 185 | struct packet_sock; |
77f65ebd WB |
186 | static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, |
187 | struct packet_type *pt, struct net_device *orig_dev); | |
1da177e4 | 188 | |
f6fb8f10 | 189 | static void *packet_previous_frame(struct packet_sock *po, |
190 | struct packet_ring_buffer *rb, | |
191 | int status); | |
192 | static void packet_increment_head(struct packet_ring_buffer *buff); | |
878cd3ba | 193 | static int prb_curr_blk_in_use(struct tpacket_block_desc *); |
bc59ba39 | 194 | static void *prb_dispatch_next_block(struct tpacket_kbdq_core *, |
f6fb8f10 | 195 | struct packet_sock *); |
bc59ba39 | 196 | static void prb_retire_current_block(struct tpacket_kbdq_core *, |
f6fb8f10 | 197 | struct packet_sock *, unsigned int status); |
bc59ba39 | 198 | static int prb_queue_frozen(struct tpacket_kbdq_core *); |
199 | static void prb_open_block(struct tpacket_kbdq_core *, | |
200 | struct tpacket_block_desc *); | |
17bfd8c8 | 201 | static void prb_retire_rx_blk_timer_expired(struct timer_list *); |
bc59ba39 | 202 | static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *); |
bc59ba39 | 203 | static void prb_fill_rxhash(struct tpacket_kbdq_core *, struct tpacket3_hdr *); |
204 | static void prb_clear_rxhash(struct tpacket_kbdq_core *, | |
205 | struct tpacket3_hdr *); | |
206 | static void prb_fill_vlan_info(struct tpacket_kbdq_core *, | |
207 | struct tpacket3_hdr *); | |
1da177e4 | 208 | static void packet_flush_mclist(struct sock *sk); |
865b03f2 | 209 | static u16 packet_pick_tx_queue(struct sk_buff *skb); |
1da177e4 | 210 | |
ffbc6111 | 211 | struct packet_skb_cb { |
ffbc6111 HX |
212 | union { |
213 | struct sockaddr_pkt pkt; | |
2472d761 EB |
214 | union { |
215 | /* Trick: alias skb original length with | |
216 | * ll.sll_family and ll.protocol in order | |
217 | * to save room. | |
218 | */ | |
219 | unsigned int origlen; | |
220 | struct sockaddr_ll ll; | |
221 | }; | |
ffbc6111 HX |
222 | } sa; |
223 | }; | |
224 | ||
d3869efe DW |
225 | #define vio_le() virtio_legacy_is_little_endian() |
226 | ||
ffbc6111 | 227 | #define PACKET_SKB_CB(__skb) ((struct packet_skb_cb *)((__skb)->cb)) |
8dc41944 | 228 | |
bc59ba39 | 229 | #define GET_PBDQC_FROM_RB(x) ((struct tpacket_kbdq_core *)(&(x)->prb_bdqc)) |
f6fb8f10 | 230 | #define GET_PBLOCK_DESC(x, bid) \ |
bc59ba39 | 231 | ((struct tpacket_block_desc *)((x)->pkbdq[(bid)].buffer)) |
f6fb8f10 | 232 | #define GET_CURR_PBLOCK_DESC_FROM_CORE(x) \ |
bc59ba39 | 233 | ((struct tpacket_block_desc *)((x)->pkbdq[(x)->kactive_blk_num].buffer)) |
f6fb8f10 | 234 | #define GET_NEXT_PRB_BLK_NUM(x) \ |
235 | (((x)->kactive_blk_num < ((x)->knum_blocks-1)) ? \ | |
236 | ((x)->kactive_blk_num+1) : 0) | |
237 | ||
dc99f600 DM |
238 | static void __fanout_unlink(struct sock *sk, struct packet_sock *po); |
239 | static void __fanout_link(struct sock *sk, struct packet_sock *po); | |
240 | ||
d346a3fa DB |
241 | static int packet_direct_xmit(struct sk_buff *skb) |
242 | { | |
865b03f2 | 243 | return dev_direct_xmit(skb, packet_pick_tx_queue(skb)); |
d346a3fa DB |
244 | } |
245 | ||
66e56cd4 DB |
246 | static struct net_device *packet_cached_dev_get(struct packet_sock *po) |
247 | { | |
248 | struct net_device *dev; | |
249 | ||
250 | rcu_read_lock(); | |
251 | dev = rcu_dereference(po->cached_dev); | |
252 | if (likely(dev)) | |
253 | dev_hold(dev); | |
254 | rcu_read_unlock(); | |
255 | ||
256 | return dev; | |
257 | } | |
258 | ||
259 | static void packet_cached_dev_assign(struct packet_sock *po, | |
260 | struct net_device *dev) | |
261 | { | |
262 | rcu_assign_pointer(po->cached_dev, dev); | |
263 | } | |
264 | ||
265 | static void packet_cached_dev_reset(struct packet_sock *po) | |
266 | { | |
267 | RCU_INIT_POINTER(po->cached_dev, NULL); | |
268 | } | |
269 | ||
d346a3fa DB |
270 | static bool packet_use_direct_xmit(const struct packet_sock *po) |
271 | { | |
272 | return po->xmit == packet_direct_xmit; | |
273 | } | |
274 | ||
865b03f2 | 275 | static u16 packet_pick_tx_queue(struct sk_buff *skb) |
0fd5d57b | 276 | { |
865b03f2 | 277 | struct net_device *dev = skb->dev; |
0fd5d57b | 278 | const struct net_device_ops *ops = dev->netdev_ops; |
b71b5837 | 279 | int cpu = raw_smp_processor_id(); |
0fd5d57b DB |
280 | u16 queue_index; |
281 | ||
b71b5837 PA |
282 | #ifdef CONFIG_XPS |
283 | skb->sender_cpu = cpu + 1; | |
284 | #endif | |
285 | skb_record_rx_queue(skb, cpu % dev->real_num_tx_queues); | |
0fd5d57b | 286 | if (ops->ndo_select_queue) { |
a350ecce | 287 | queue_index = ops->ndo_select_queue(dev, skb, NULL); |
0fd5d57b DB |
288 | queue_index = netdev_cap_txqueue(dev, queue_index); |
289 | } else { | |
b71b5837 | 290 | queue_index = netdev_pick_tx(dev, skb, NULL); |
0fd5d57b DB |
291 | } |
292 | ||
865b03f2 | 293 | return queue_index; |
0fd5d57b DB |
294 | } |
295 | ||
a6361f0c | 296 | /* __register_prot_hook must be invoked through register_prot_hook |
ce06b03e DM |
297 | * or from a context in which asynchronous accesses to the packet |
298 | * socket is not possible (packet_create()). | |
299 | */ | |
a6361f0c | 300 | static void __register_prot_hook(struct sock *sk) |
ce06b03e DM |
301 | { |
302 | struct packet_sock *po = pkt_sk(sk); | |
e40526cb | 303 | |
ce06b03e | 304 | if (!po->running) { |
66e56cd4 | 305 | if (po->fanout) |
dc99f600 | 306 | __fanout_link(sk, po); |
66e56cd4 | 307 | else |
dc99f600 | 308 | dev_add_pack(&po->prot_hook); |
e40526cb | 309 | |
ce06b03e DM |
310 | sock_hold(sk); |
311 | po->running = 1; | |
312 | } | |
313 | } | |
314 | ||
a6361f0c WB |
315 | static void register_prot_hook(struct sock *sk) |
316 | { | |
317 | lockdep_assert_held_once(&pkt_sk(sk)->bind_lock); | |
318 | __register_prot_hook(sk); | |
319 | } | |
320 | ||
321 | /* If the sync parameter is true, we will temporarily drop | |
ce06b03e DM |
322 | * the po->bind_lock and do a synchronize_net to make sure no |
323 | * asynchronous packet processing paths still refer to the elements | |
324 | * of po->prot_hook. If the sync parameter is false, it is the | |
325 | * callers responsibility to take care of this. | |
326 | */ | |
327 | static void __unregister_prot_hook(struct sock *sk, bool sync) | |
328 | { | |
329 | struct packet_sock *po = pkt_sk(sk); | |
330 | ||
a6361f0c WB |
331 | lockdep_assert_held_once(&po->bind_lock); |
332 | ||
ce06b03e | 333 | po->running = 0; |
66e56cd4 DB |
334 | |
335 | if (po->fanout) | |
dc99f600 | 336 | __fanout_unlink(sk, po); |
66e56cd4 | 337 | else |
dc99f600 | 338 | __dev_remove_pack(&po->prot_hook); |
e40526cb | 339 | |
ce06b03e DM |
340 | __sock_put(sk); |
341 | ||
342 | if (sync) { | |
343 | spin_unlock(&po->bind_lock); | |
344 | synchronize_net(); | |
345 | spin_lock(&po->bind_lock); | |
346 | } | |
347 | } | |
348 | ||
349 | static void unregister_prot_hook(struct sock *sk, bool sync) | |
350 | { | |
351 | struct packet_sock *po = pkt_sk(sk); | |
352 | ||
353 | if (po->running) | |
354 | __unregister_prot_hook(sk, sync); | |
355 | } | |
356 | ||
6e58040b | 357 | static inline struct page * __pure pgv_to_page(void *addr) |
0af55bb5 CG |
358 | { |
359 | if (is_vmalloc_addr(addr)) | |
360 | return vmalloc_to_page(addr); | |
361 | return virt_to_page(addr); | |
362 | } | |
363 | ||
69e3c75f | 364 | static void __packet_set_status(struct packet_sock *po, void *frame, int status) |
1da177e4 | 365 | { |
184f489e | 366 | union tpacket_uhdr h; |
1da177e4 | 367 | |
69e3c75f | 368 | h.raw = frame; |
bbd6ef87 PM |
369 | switch (po->tp_version) { |
370 | case TPACKET_V1: | |
69e3c75f | 371 | h.h1->tp_status = status; |
0af55bb5 | 372 | flush_dcache_page(pgv_to_page(&h.h1->tp_status)); |
bbd6ef87 PM |
373 | break; |
374 | case TPACKET_V2: | |
69e3c75f | 375 | h.h2->tp_status = status; |
0af55bb5 | 376 | flush_dcache_page(pgv_to_page(&h.h2->tp_status)); |
bbd6ef87 | 377 | break; |
f6fb8f10 | 378 | case TPACKET_V3: |
7f953ab2 SV |
379 | h.h3->tp_status = status; |
380 | flush_dcache_page(pgv_to_page(&h.h3->tp_status)); | |
381 | break; | |
69e3c75f | 382 | default: |
f6fb8f10 | 383 | WARN(1, "TPACKET version not supported.\n"); |
69e3c75f | 384 | BUG(); |
bbd6ef87 | 385 | } |
69e3c75f JB |
386 | |
387 | smp_wmb(); | |
bbd6ef87 PM |
388 | } |
389 | ||
96f657e6 | 390 | static int __packet_get_status(const struct packet_sock *po, void *frame) |
bbd6ef87 | 391 | { |
184f489e | 392 | union tpacket_uhdr h; |
bbd6ef87 | 393 | |
69e3c75f JB |
394 | smp_rmb(); |
395 | ||
bbd6ef87 PM |
396 | h.raw = frame; |
397 | switch (po->tp_version) { | |
398 | case TPACKET_V1: | |
0af55bb5 | 399 | flush_dcache_page(pgv_to_page(&h.h1->tp_status)); |
69e3c75f | 400 | return h.h1->tp_status; |
bbd6ef87 | 401 | case TPACKET_V2: |
0af55bb5 | 402 | flush_dcache_page(pgv_to_page(&h.h2->tp_status)); |
69e3c75f | 403 | return h.h2->tp_status; |
f6fb8f10 | 404 | case TPACKET_V3: |
7f953ab2 SV |
405 | flush_dcache_page(pgv_to_page(&h.h3->tp_status)); |
406 | return h.h3->tp_status; | |
69e3c75f | 407 | default: |
f6fb8f10 | 408 | WARN(1, "TPACKET version not supported.\n"); |
69e3c75f JB |
409 | BUG(); |
410 | return 0; | |
bbd6ef87 | 411 | } |
1da177e4 | 412 | } |
69e3c75f | 413 | |
d413fcb4 | 414 | static __u32 tpacket_get_timestamp(struct sk_buff *skb, struct timespec64 *ts, |
b9c32fb2 | 415 | unsigned int flags) |
7a51384c DB |
416 | { |
417 | struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb); | |
418 | ||
68a360e8 WB |
419 | if (shhwtstamps && |
420 | (flags & SOF_TIMESTAMPING_RAW_HARDWARE) && | |
d413fcb4 | 421 | ktime_to_timespec64_cond(shhwtstamps->hwtstamp, ts)) |
68a360e8 | 422 | return TP_STATUS_TS_RAW_HARDWARE; |
7a51384c | 423 | |
d413fcb4 | 424 | if (ktime_to_timespec64_cond(skb->tstamp, ts)) |
b9c32fb2 | 425 | return TP_STATUS_TS_SOFTWARE; |
7a51384c | 426 | |
b9c32fb2 | 427 | return 0; |
7a51384c DB |
428 | } |
429 | ||
b9c32fb2 DB |
430 | static __u32 __packet_set_timestamp(struct packet_sock *po, void *frame, |
431 | struct sk_buff *skb) | |
2e31396f WB |
432 | { |
433 | union tpacket_uhdr h; | |
d413fcb4 | 434 | struct timespec64 ts; |
b9c32fb2 | 435 | __u32 ts_status; |
2e31396f | 436 | |
b9c32fb2 DB |
437 | if (!(ts_status = tpacket_get_timestamp(skb, &ts, po->tp_tstamp))) |
438 | return 0; | |
2e31396f WB |
439 | |
440 | h.raw = frame; | |
d413fcb4 AB |
441 | /* |
442 | * versions 1 through 3 overflow the timestamps in y2106, since they | |
443 | * all store the seconds in a 32-bit unsigned integer. | |
444 | * If we create a version 4, that should have a 64-bit timestamp, | |
445 | * either 64-bit seconds + 32-bit nanoseconds, or just 64-bit | |
446 | * nanoseconds. | |
447 | */ | |
2e31396f WB |
448 | switch (po->tp_version) { |
449 | case TPACKET_V1: | |
450 | h.h1->tp_sec = ts.tv_sec; | |
451 | h.h1->tp_usec = ts.tv_nsec / NSEC_PER_USEC; | |
452 | break; | |
453 | case TPACKET_V2: | |
454 | h.h2->tp_sec = ts.tv_sec; | |
455 | h.h2->tp_nsec = ts.tv_nsec; | |
456 | break; | |
457 | case TPACKET_V3: | |
57ea884b DB |
458 | h.h3->tp_sec = ts.tv_sec; |
459 | h.h3->tp_nsec = ts.tv_nsec; | |
460 | break; | |
2e31396f WB |
461 | default: |
462 | WARN(1, "TPACKET version not supported.\n"); | |
463 | BUG(); | |
464 | } | |
465 | ||
466 | /* one flush is safe, as both fields always lie on the same cacheline */ | |
467 | flush_dcache_page(pgv_to_page(&h.h1->tp_sec)); | |
468 | smp_wmb(); | |
b9c32fb2 DB |
469 | |
470 | return ts_status; | |
2e31396f WB |
471 | } |
472 | ||
d4b5bd98 ED |
473 | static void *packet_lookup_frame(const struct packet_sock *po, |
474 | const struct packet_ring_buffer *rb, | |
475 | unsigned int position, | |
476 | int status) | |
69e3c75f JB |
477 | { |
478 | unsigned int pg_vec_pos, frame_offset; | |
184f489e | 479 | union tpacket_uhdr h; |
69e3c75f JB |
480 | |
481 | pg_vec_pos = position / rb->frames_per_block; | |
482 | frame_offset = position % rb->frames_per_block; | |
483 | ||
0e3125c7 NH |
484 | h.raw = rb->pg_vec[pg_vec_pos].buffer + |
485 | (frame_offset * rb->frame_size); | |
69e3c75f JB |
486 | |
487 | if (status != __packet_get_status(po, h.raw)) | |
488 | return NULL; | |
489 | ||
490 | return h.raw; | |
491 | } | |
492 | ||
eea49cc9 | 493 | static void *packet_current_frame(struct packet_sock *po, |
69e3c75f JB |
494 | struct packet_ring_buffer *rb, |
495 | int status) | |
496 | { | |
497 | return packet_lookup_frame(po, rb, rb->head, status); | |
498 | } | |
499 | ||
bc59ba39 | 500 | static void prb_del_retire_blk_timer(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 501 | { |
502 | del_timer_sync(&pkc->retire_blk_timer); | |
503 | } | |
504 | ||
505 | static void prb_shutdown_retire_blk_timer(struct packet_sock *po, | |
f6fb8f10 | 506 | struct sk_buff_head *rb_queue) |
507 | { | |
bc59ba39 | 508 | struct tpacket_kbdq_core *pkc; |
f6fb8f10 | 509 | |
73d0fcf2 | 510 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 511 | |
ec6f809f | 512 | spin_lock_bh(&rb_queue->lock); |
f6fb8f10 | 513 | pkc->delete_blk_timer = 1; |
ec6f809f | 514 | spin_unlock_bh(&rb_queue->lock); |
f6fb8f10 | 515 | |
516 | prb_del_retire_blk_timer(pkc); | |
517 | } | |
518 | ||
e8e85cc5 | 519 | static void prb_setup_retire_blk_timer(struct packet_sock *po) |
f6fb8f10 | 520 | { |
bc59ba39 | 521 | struct tpacket_kbdq_core *pkc; |
f6fb8f10 | 522 | |
e8e85cc5 | 523 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
17bfd8c8 KC |
524 | timer_setup(&pkc->retire_blk_timer, prb_retire_rx_blk_timer_expired, |
525 | 0); | |
526 | pkc->retire_blk_timer.expires = jiffies; | |
f6fb8f10 | 527 | } |
528 | ||
529 | static int prb_calc_retire_blk_tmo(struct packet_sock *po, | |
530 | int blk_size_in_bytes) | |
531 | { | |
532 | struct net_device *dev; | |
0914d2bb | 533 | unsigned int mbits, div; |
7cad1bac | 534 | struct ethtool_link_ksettings ecmd; |
4bc71cb9 | 535 | int err; |
f6fb8f10 | 536 | |
4bc71cb9 JP |
537 | rtnl_lock(); |
538 | dev = __dev_get_by_index(sock_net(&po->sk), po->ifindex); | |
539 | if (unlikely(!dev)) { | |
540 | rtnl_unlock(); | |
f6fb8f10 | 541 | return DEFAULT_PRB_RETIRE_TOV; |
4bc71cb9 | 542 | } |
7cad1bac | 543 | err = __ethtool_get_link_ksettings(dev, &ecmd); |
4bc71cb9 | 544 | rtnl_unlock(); |
0914d2bb | 545 | if (err) |
b43d1f9f | 546 | return DEFAULT_PRB_RETIRE_TOV; |
f6fb8f10 | 547 | |
0914d2bb MW |
548 | /* If the link speed is so slow you don't really |
549 | * need to worry about perf anyways | |
550 | */ | |
551 | if (ecmd.base.speed < SPEED_1000 || | |
552 | ecmd.base.speed == SPEED_UNKNOWN) | |
553 | return DEFAULT_PRB_RETIRE_TOV; | |
554 | ||
555 | div = ecmd.base.speed / 1000; | |
f6fb8f10 | 556 | mbits = (blk_size_in_bytes * 8) / (1024 * 1024); |
557 | ||
558 | if (div) | |
559 | mbits /= div; | |
560 | ||
f6fb8f10 | 561 | if (div) |
0914d2bb MW |
562 | return mbits + 1; |
563 | return mbits; | |
f6fb8f10 | 564 | } |
565 | ||
bc59ba39 | 566 | static void prb_init_ft_ops(struct tpacket_kbdq_core *p1, |
f6fb8f10 | 567 | union tpacket_req_u *req_u) |
568 | { | |
569 | p1->feature_req_word = req_u->req3.tp_feature_req_word; | |
570 | } | |
571 | ||
572 | static void init_prb_bdqc(struct packet_sock *po, | |
573 | struct packet_ring_buffer *rb, | |
574 | struct pgv *pg_vec, | |
e8e85cc5 | 575 | union tpacket_req_u *req_u) |
f6fb8f10 | 576 | { |
22781a5b | 577 | struct tpacket_kbdq_core *p1 = GET_PBDQC_FROM_RB(rb); |
bc59ba39 | 578 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 579 | |
580 | memset(p1, 0x0, sizeof(*p1)); | |
581 | ||
582 | p1->knxt_seq_num = 1; | |
583 | p1->pkbdq = pg_vec; | |
bc59ba39 | 584 | pbd = (struct tpacket_block_desc *)pg_vec[0].buffer; |
e3192690 | 585 | p1->pkblk_start = pg_vec[0].buffer; |
f6fb8f10 | 586 | p1->kblk_size = req_u->req3.tp_block_size; |
587 | p1->knum_blocks = req_u->req3.tp_block_nr; | |
588 | p1->hdrlen = po->tp_hdrlen; | |
589 | p1->version = po->tp_version; | |
590 | p1->last_kactive_blk_num = 0; | |
ee80fbf3 | 591 | po->stats.stats3.tp_freeze_q_cnt = 0; |
f6fb8f10 | 592 | if (req_u->req3.tp_retire_blk_tov) |
593 | p1->retire_blk_tov = req_u->req3.tp_retire_blk_tov; | |
594 | else | |
595 | p1->retire_blk_tov = prb_calc_retire_blk_tmo(po, | |
596 | req_u->req3.tp_block_size); | |
597 | p1->tov_in_jiffies = msecs_to_jiffies(p1->retire_blk_tov); | |
598 | p1->blk_sizeof_priv = req_u->req3.tp_sizeof_priv; | |
632ca50f | 599 | rwlock_init(&p1->blk_fill_in_prog_lock); |
f6fb8f10 | 600 | |
dc808110 | 601 | p1->max_frame_len = p1->kblk_size - BLK_PLUS_PRIV(p1->blk_sizeof_priv); |
f6fb8f10 | 602 | prb_init_ft_ops(p1, req_u); |
e8e85cc5 | 603 | prb_setup_retire_blk_timer(po); |
f6fb8f10 | 604 | prb_open_block(p1, pbd); |
605 | } | |
606 | ||
607 | /* Do NOT update the last_blk_num first. | |
608 | * Assumes sk_buff_head lock is held. | |
609 | */ | |
bc59ba39 | 610 | static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 611 | { |
612 | mod_timer(&pkc->retire_blk_timer, | |
613 | jiffies + pkc->tov_in_jiffies); | |
614 | pkc->last_kactive_blk_num = pkc->kactive_blk_num; | |
615 | } | |
616 | ||
617 | /* | |
618 | * Timer logic: | |
619 | * 1) We refresh the timer only when we open a block. | |
620 | * By doing this we don't waste cycles refreshing the timer | |
621 | * on packet-by-packet basis. | |
622 | * | |
623 | * With a 1MB block-size, on a 1Gbps line, it will take | |
624 | * i) ~8 ms to fill a block + ii) memcpy etc. | |
625 | * In this cut we are not accounting for the memcpy time. | |
626 | * | |
627 | * So, if the user sets the 'tmo' to 10ms then the timer | |
628 | * will never fire while the block is still getting filled | |
629 | * (which is what we want). However, the user could choose | |
630 | * to close a block early and that's fine. | |
631 | * | |
632 | * But when the timer does fire, we check whether or not to refresh it. | |
633 | * Since the tmo granularity is in msecs, it is not too expensive | |
634 | * to refresh the timer, lets say every '8' msecs. | |
635 | * Either the user can set the 'tmo' or we can derive it based on | |
636 | * a) line-speed and b) block-size. | |
637 | * prb_calc_retire_blk_tmo() calculates the tmo. | |
638 | * | |
639 | */ | |
17bfd8c8 | 640 | static void prb_retire_rx_blk_timer_expired(struct timer_list *t) |
f6fb8f10 | 641 | { |
17bfd8c8 KC |
642 | struct packet_sock *po = |
643 | from_timer(po, t, rx_ring.prb_bdqc.retire_blk_timer); | |
22781a5b | 644 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 645 | unsigned int frozen; |
bc59ba39 | 646 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 647 | |
648 | spin_lock(&po->sk.sk_receive_queue.lock); | |
649 | ||
650 | frozen = prb_queue_frozen(pkc); | |
651 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
652 | ||
653 | if (unlikely(pkc->delete_blk_timer)) | |
654 | goto out; | |
655 | ||
656 | /* We only need to plug the race when the block is partially filled. | |
657 | * tpacket_rcv: | |
658 | * lock(); increment BLOCK_NUM_PKTS; unlock() | |
659 | * copy_bits() is in progress ... | |
660 | * timer fires on other cpu: | |
661 | * we can't retire the current block because copy_bits | |
662 | * is in progress. | |
663 | * | |
664 | */ | |
665 | if (BLOCK_NUM_PKTS(pbd)) { | |
632ca50f JO |
666 | /* Waiting for skb_copy_bits to finish... */ |
667 | write_lock(&pkc->blk_fill_in_prog_lock); | |
668 | write_unlock(&pkc->blk_fill_in_prog_lock); | |
f6fb8f10 | 669 | } |
670 | ||
671 | if (pkc->last_kactive_blk_num == pkc->kactive_blk_num) { | |
672 | if (!frozen) { | |
41a50d62 AD |
673 | if (!BLOCK_NUM_PKTS(pbd)) { |
674 | /* An empty block. Just refresh the timer. */ | |
675 | goto refresh_timer; | |
676 | } | |
f6fb8f10 | 677 | prb_retire_current_block(pkc, po, TP_STATUS_BLK_TMO); |
678 | if (!prb_dispatch_next_block(pkc, po)) | |
679 | goto refresh_timer; | |
680 | else | |
681 | goto out; | |
682 | } else { | |
683 | /* Case 1. Queue was frozen because user-space was | |
684 | * lagging behind. | |
685 | */ | |
878cd3ba | 686 | if (prb_curr_blk_in_use(pbd)) { |
f6fb8f10 | 687 | /* |
688 | * Ok, user-space is still behind. | |
689 | * So just refresh the timer. | |
690 | */ | |
691 | goto refresh_timer; | |
692 | } else { | |
693 | /* Case 2. queue was frozen,user-space caught up, | |
694 | * now the link went idle && the timer fired. | |
695 | * We don't have a block to close.So we open this | |
696 | * block and restart the timer. | |
697 | * opening a block thaws the queue,restarts timer | |
698 | * Thawing/timer-refresh is a side effect. | |
699 | */ | |
700 | prb_open_block(pkc, pbd); | |
701 | goto out; | |
702 | } | |
703 | } | |
704 | } | |
705 | ||
706 | refresh_timer: | |
707 | _prb_refresh_rx_retire_blk_timer(pkc); | |
708 | ||
709 | out: | |
710 | spin_unlock(&po->sk.sk_receive_queue.lock); | |
711 | } | |
712 | ||
eea49cc9 | 713 | static void prb_flush_block(struct tpacket_kbdq_core *pkc1, |
bc59ba39 | 714 | struct tpacket_block_desc *pbd1, __u32 status) |
f6fb8f10 | 715 | { |
716 | /* Flush everything minus the block header */ | |
717 | ||
718 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
719 | u8 *start, *end; | |
720 | ||
721 | start = (u8 *)pbd1; | |
722 | ||
723 | /* Skip the block header(we know header WILL fit in 4K) */ | |
724 | start += PAGE_SIZE; | |
725 | ||
726 | end = (u8 *)PAGE_ALIGN((unsigned long)pkc1->pkblk_end); | |
727 | for (; start < end; start += PAGE_SIZE) | |
728 | flush_dcache_page(pgv_to_page(start)); | |
729 | ||
730 | smp_wmb(); | |
731 | #endif | |
732 | ||
733 | /* Now update the block status. */ | |
734 | ||
735 | BLOCK_STATUS(pbd1) = status; | |
736 | ||
737 | /* Flush the block header */ | |
738 | ||
739 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
740 | start = (u8 *)pbd1; | |
741 | flush_dcache_page(pgv_to_page(start)); | |
742 | ||
743 | smp_wmb(); | |
744 | #endif | |
745 | } | |
746 | ||
747 | /* | |
748 | * Side effect: | |
749 | * | |
750 | * 1) flush the block | |
751 | * 2) Increment active_blk_num | |
752 | * | |
753 | * Note:We DONT refresh the timer on purpose. | |
754 | * Because almost always the next block will be opened. | |
755 | */ | |
bc59ba39 | 756 | static void prb_close_block(struct tpacket_kbdq_core *pkc1, |
757 | struct tpacket_block_desc *pbd1, | |
f6fb8f10 | 758 | struct packet_sock *po, unsigned int stat) |
759 | { | |
760 | __u32 status = TP_STATUS_USER | stat; | |
761 | ||
762 | struct tpacket3_hdr *last_pkt; | |
bc59ba39 | 763 | struct tpacket_hdr_v1 *h1 = &pbd1->hdr.bh1; |
da413eec | 764 | struct sock *sk = &po->sk; |
f6fb8f10 | 765 | |
8e8e2951 | 766 | if (atomic_read(&po->tp_drops)) |
f6fb8f10 | 767 | status |= TP_STATUS_LOSING; |
768 | ||
769 | last_pkt = (struct tpacket3_hdr *)pkc1->prev; | |
770 | last_pkt->tp_next_offset = 0; | |
771 | ||
772 | /* Get the ts of the last pkt */ | |
773 | if (BLOCK_NUM_PKTS(pbd1)) { | |
774 | h1->ts_last_pkt.ts_sec = last_pkt->tp_sec; | |
775 | h1->ts_last_pkt.ts_nsec = last_pkt->tp_nsec; | |
776 | } else { | |
41a50d62 AD |
777 | /* Ok, we tmo'd - so get the current time. |
778 | * | |
779 | * It shouldn't really happen as we don't close empty | |
780 | * blocks. See prb_retire_rx_blk_timer_expired(). | |
781 | */ | |
d413fcb4 AB |
782 | struct timespec64 ts; |
783 | ktime_get_real_ts64(&ts); | |
f6fb8f10 | 784 | h1->ts_last_pkt.ts_sec = ts.tv_sec; |
785 | h1->ts_last_pkt.ts_nsec = ts.tv_nsec; | |
786 | } | |
787 | ||
788 | smp_wmb(); | |
789 | ||
790 | /* Flush the block */ | |
791 | prb_flush_block(pkc1, pbd1, status); | |
792 | ||
da413eec DC |
793 | sk->sk_data_ready(sk); |
794 | ||
f6fb8f10 | 795 | pkc1->kactive_blk_num = GET_NEXT_PRB_BLK_NUM(pkc1); |
796 | } | |
797 | ||
eea49cc9 | 798 | static void prb_thaw_queue(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 799 | { |
800 | pkc->reset_pending_on_curr_blk = 0; | |
801 | } | |
802 | ||
803 | /* | |
804 | * Side effect of opening a block: | |
805 | * | |
806 | * 1) prb_queue is thawed. | |
807 | * 2) retire_blk_timer is refreshed. | |
808 | * | |
809 | */ | |
bc59ba39 | 810 | static void prb_open_block(struct tpacket_kbdq_core *pkc1, |
811 | struct tpacket_block_desc *pbd1) | |
f6fb8f10 | 812 | { |
d413fcb4 | 813 | struct timespec64 ts; |
bc59ba39 | 814 | struct tpacket_hdr_v1 *h1 = &pbd1->hdr.bh1; |
f6fb8f10 | 815 | |
816 | smp_rmb(); | |
817 | ||
8da3056c DB |
818 | /* We could have just memset this but we will lose the |
819 | * flexibility of making the priv area sticky | |
820 | */ | |
f6fb8f10 | 821 | |
8da3056c DB |
822 | BLOCK_SNUM(pbd1) = pkc1->knxt_seq_num++; |
823 | BLOCK_NUM_PKTS(pbd1) = 0; | |
824 | BLOCK_LEN(pbd1) = BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
f6fb8f10 | 825 | |
d413fcb4 | 826 | ktime_get_real_ts64(&ts); |
8da3056c DB |
827 | |
828 | h1->ts_first_pkt.ts_sec = ts.tv_sec; | |
829 | h1->ts_first_pkt.ts_nsec = ts.tv_nsec; | |
f6fb8f10 | 830 | |
8da3056c DB |
831 | pkc1->pkblk_start = (char *)pbd1; |
832 | pkc1->nxt_offset = pkc1->pkblk_start + BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
833 | ||
834 | BLOCK_O2FP(pbd1) = (__u32)BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
835 | BLOCK_O2PRIV(pbd1) = BLK_HDR_LEN; | |
836 | ||
837 | pbd1->version = pkc1->version; | |
838 | pkc1->prev = pkc1->nxt_offset; | |
839 | pkc1->pkblk_end = pkc1->pkblk_start + pkc1->kblk_size; | |
840 | ||
841 | prb_thaw_queue(pkc1); | |
842 | _prb_refresh_rx_retire_blk_timer(pkc1); | |
843 | ||
844 | smp_wmb(); | |
f6fb8f10 | 845 | } |
846 | ||
847 | /* | |
848 | * Queue freeze logic: | |
849 | * 1) Assume tp_block_nr = 8 blocks. | |
850 | * 2) At time 't0', user opens Rx ring. | |
851 | * 3) Some time past 't0', kernel starts filling blocks starting from 0 .. 7 | |
852 | * 4) user-space is either sleeping or processing block '0'. | |
853 | * 5) tpacket_rcv is currently filling block '7', since there is no space left, | |
854 | * it will close block-7,loop around and try to fill block '0'. | |
855 | * call-flow: | |
856 | * __packet_lookup_frame_in_block | |
857 | * prb_retire_current_block() | |
858 | * prb_dispatch_next_block() | |
859 | * |->(BLOCK_STATUS == USER) evaluates to true | |
860 | * 5.1) Since block-0 is currently in-use, we just freeze the queue. | |
861 | * 6) Now there are two cases: | |
862 | * 6.1) Link goes idle right after the queue is frozen. | |
863 | * But remember, the last open_block() refreshed the timer. | |
864 | * When this timer expires,it will refresh itself so that we can | |
865 | * re-open block-0 in near future. | |
866 | * 6.2) Link is busy and keeps on receiving packets. This is a simple | |
867 | * case and __packet_lookup_frame_in_block will check if block-0 | |
868 | * is free and can now be re-used. | |
869 | */ | |
eea49cc9 | 870 | static void prb_freeze_queue(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 871 | struct packet_sock *po) |
872 | { | |
873 | pkc->reset_pending_on_curr_blk = 1; | |
ee80fbf3 | 874 | po->stats.stats3.tp_freeze_q_cnt++; |
f6fb8f10 | 875 | } |
876 | ||
877 | #define TOTAL_PKT_LEN_INCL_ALIGN(length) (ALIGN((length), V3_ALIGNMENT)) | |
878 | ||
879 | /* | |
880 | * If the next block is free then we will dispatch it | |
881 | * and return a good offset. | |
882 | * Else, we will freeze the queue. | |
883 | * So, caller must check the return value. | |
884 | */ | |
bc59ba39 | 885 | static void *prb_dispatch_next_block(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 886 | struct packet_sock *po) |
887 | { | |
bc59ba39 | 888 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 889 | |
890 | smp_rmb(); | |
891 | ||
892 | /* 1. Get current block num */ | |
893 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
894 | ||
895 | /* 2. If this block is currently in_use then freeze the queue */ | |
896 | if (TP_STATUS_USER & BLOCK_STATUS(pbd)) { | |
897 | prb_freeze_queue(pkc, po); | |
898 | return NULL; | |
899 | } | |
900 | ||
901 | /* | |
902 | * 3. | |
903 | * open this block and return the offset where the first packet | |
904 | * needs to get stored. | |
905 | */ | |
906 | prb_open_block(pkc, pbd); | |
907 | return (void *)pkc->nxt_offset; | |
908 | } | |
909 | ||
bc59ba39 | 910 | static void prb_retire_current_block(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 911 | struct packet_sock *po, unsigned int status) |
912 | { | |
bc59ba39 | 913 | struct tpacket_block_desc *pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); |
f6fb8f10 | 914 | |
915 | /* retire/close the current block */ | |
916 | if (likely(TP_STATUS_KERNEL == BLOCK_STATUS(pbd))) { | |
917 | /* | |
918 | * Plug the case where copy_bits() is in progress on | |
919 | * cpu-0 and tpacket_rcv() got invoked on cpu-1, didn't | |
920 | * have space to copy the pkt in the current block and | |
921 | * called prb_retire_current_block() | |
922 | * | |
923 | * We don't need to worry about the TMO case because | |
924 | * the timer-handler already handled this case. | |
925 | */ | |
926 | if (!(status & TP_STATUS_BLK_TMO)) { | |
632ca50f JO |
927 | /* Waiting for skb_copy_bits to finish... */ |
928 | write_lock(&pkc->blk_fill_in_prog_lock); | |
929 | write_unlock(&pkc->blk_fill_in_prog_lock); | |
f6fb8f10 | 930 | } |
931 | prb_close_block(pkc, pbd, po, status); | |
932 | return; | |
933 | } | |
f6fb8f10 | 934 | } |
935 | ||
878cd3ba | 936 | static int prb_curr_blk_in_use(struct tpacket_block_desc *pbd) |
f6fb8f10 | 937 | { |
938 | return TP_STATUS_USER & BLOCK_STATUS(pbd); | |
939 | } | |
940 | ||
eea49cc9 | 941 | static int prb_queue_frozen(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 942 | { |
943 | return pkc->reset_pending_on_curr_blk; | |
944 | } | |
945 | ||
eea49cc9 | 946 | static void prb_clear_blk_fill_status(struct packet_ring_buffer *rb) |
88fd1cb8 | 947 | __releases(&pkc->blk_fill_in_prog_lock) |
f6fb8f10 | 948 | { |
bc59ba39 | 949 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(rb); |
632ca50f JO |
950 | |
951 | read_unlock(&pkc->blk_fill_in_prog_lock); | |
f6fb8f10 | 952 | } |
953 | ||
eea49cc9 | 954 | static void prb_fill_rxhash(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 955 | struct tpacket3_hdr *ppd) |
956 | { | |
3958afa1 | 957 | ppd->hv1.tp_rxhash = skb_get_hash(pkc->skb); |
f6fb8f10 | 958 | } |
959 | ||
eea49cc9 | 960 | static void prb_clear_rxhash(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 961 | struct tpacket3_hdr *ppd) |
962 | { | |
963 | ppd->hv1.tp_rxhash = 0; | |
964 | } | |
965 | ||
eea49cc9 | 966 | static void prb_fill_vlan_info(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 967 | struct tpacket3_hdr *ppd) |
968 | { | |
df8a39de JP |
969 | if (skb_vlan_tag_present(pkc->skb)) { |
970 | ppd->hv1.tp_vlan_tci = skb_vlan_tag_get(pkc->skb); | |
a0cdfcf3 AW |
971 | ppd->hv1.tp_vlan_tpid = ntohs(pkc->skb->vlan_proto); |
972 | ppd->tp_status = TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
f6fb8f10 | 973 | } else { |
9e67030a | 974 | ppd->hv1.tp_vlan_tci = 0; |
a0cdfcf3 | 975 | ppd->hv1.tp_vlan_tpid = 0; |
9e67030a | 976 | ppd->tp_status = TP_STATUS_AVAILABLE; |
f6fb8f10 | 977 | } |
978 | } | |
979 | ||
bc59ba39 | 980 | static void prb_run_all_ft_ops(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 981 | struct tpacket3_hdr *ppd) |
982 | { | |
a0cdfcf3 | 983 | ppd->hv1.tp_padding = 0; |
f6fb8f10 | 984 | prb_fill_vlan_info(pkc, ppd); |
985 | ||
986 | if (pkc->feature_req_word & TP_FT_REQ_FILL_RXHASH) | |
987 | prb_fill_rxhash(pkc, ppd); | |
988 | else | |
989 | prb_clear_rxhash(pkc, ppd); | |
990 | } | |
991 | ||
eea49cc9 | 992 | static void prb_fill_curr_block(char *curr, |
bc59ba39 | 993 | struct tpacket_kbdq_core *pkc, |
994 | struct tpacket_block_desc *pbd, | |
f6fb8f10 | 995 | unsigned int len) |
88fd1cb8 | 996 | __acquires(&pkc->blk_fill_in_prog_lock) |
f6fb8f10 | 997 | { |
998 | struct tpacket3_hdr *ppd; | |
999 | ||
1000 | ppd = (struct tpacket3_hdr *)curr; | |
1001 | ppd->tp_next_offset = TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1002 | pkc->prev = curr; | |
1003 | pkc->nxt_offset += TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1004 | BLOCK_LEN(pbd) += TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1005 | BLOCK_NUM_PKTS(pbd) += 1; | |
632ca50f | 1006 | read_lock(&pkc->blk_fill_in_prog_lock); |
f6fb8f10 | 1007 | prb_run_all_ft_ops(pkc, ppd); |
1008 | } | |
1009 | ||
1010 | /* Assumes caller has the sk->rx_queue.lock */ | |
1011 | static void *__packet_lookup_frame_in_block(struct packet_sock *po, | |
1012 | struct sk_buff *skb, | |
f6fb8f10 | 1013 | unsigned int len |
1014 | ) | |
1015 | { | |
bc59ba39 | 1016 | struct tpacket_kbdq_core *pkc; |
1017 | struct tpacket_block_desc *pbd; | |
f6fb8f10 | 1018 | char *curr, *end; |
1019 | ||
e3192690 | 1020 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 1021 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); |
1022 | ||
1023 | /* Queue is frozen when user space is lagging behind */ | |
1024 | if (prb_queue_frozen(pkc)) { | |
1025 | /* | |
1026 | * Check if that last block which caused the queue to freeze, | |
1027 | * is still in_use by user-space. | |
1028 | */ | |
878cd3ba | 1029 | if (prb_curr_blk_in_use(pbd)) { |
f6fb8f10 | 1030 | /* Can't record this packet */ |
1031 | return NULL; | |
1032 | } else { | |
1033 | /* | |
1034 | * Ok, the block was released by user-space. | |
1035 | * Now let's open that block. | |
1036 | * opening a block also thaws the queue. | |
1037 | * Thawing is a side effect. | |
1038 | */ | |
1039 | prb_open_block(pkc, pbd); | |
1040 | } | |
1041 | } | |
1042 | ||
1043 | smp_mb(); | |
1044 | curr = pkc->nxt_offset; | |
1045 | pkc->skb = skb; | |
e3192690 | 1046 | end = (char *)pbd + pkc->kblk_size; |
f6fb8f10 | 1047 | |
1048 | /* first try the current block */ | |
1049 | if (curr+TOTAL_PKT_LEN_INCL_ALIGN(len) < end) { | |
1050 | prb_fill_curr_block(curr, pkc, pbd, len); | |
1051 | return (void *)curr; | |
1052 | } | |
1053 | ||
1054 | /* Ok, close the current block */ | |
1055 | prb_retire_current_block(pkc, po, 0); | |
1056 | ||
1057 | /* Now, try to dispatch the next block */ | |
1058 | curr = (char *)prb_dispatch_next_block(pkc, po); | |
1059 | if (curr) { | |
1060 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
1061 | prb_fill_curr_block(curr, pkc, pbd, len); | |
1062 | return (void *)curr; | |
1063 | } | |
1064 | ||
1065 | /* | |
1066 | * No free blocks are available.user_space hasn't caught up yet. | |
1067 | * Queue was just frozen and now this packet will get dropped. | |
1068 | */ | |
1069 | return NULL; | |
1070 | } | |
1071 | ||
eea49cc9 | 1072 | static void *packet_current_rx_frame(struct packet_sock *po, |
f6fb8f10 | 1073 | struct sk_buff *skb, |
1074 | int status, unsigned int len) | |
1075 | { | |
1076 | char *curr = NULL; | |
1077 | switch (po->tp_version) { | |
1078 | case TPACKET_V1: | |
1079 | case TPACKET_V2: | |
1080 | curr = packet_lookup_frame(po, &po->rx_ring, | |
1081 | po->rx_ring.head, status); | |
1082 | return curr; | |
1083 | case TPACKET_V3: | |
46088059 | 1084 | return __packet_lookup_frame_in_block(po, skb, len); |
f6fb8f10 | 1085 | default: |
1086 | WARN(1, "TPACKET version not supported\n"); | |
1087 | BUG(); | |
99aa3473 | 1088 | return NULL; |
f6fb8f10 | 1089 | } |
1090 | } | |
1091 | ||
dcf70cef ED |
1092 | static void *prb_lookup_block(const struct packet_sock *po, |
1093 | const struct packet_ring_buffer *rb, | |
1094 | unsigned int idx, | |
1095 | int status) | |
f6fb8f10 | 1096 | { |
bc59ba39 | 1097 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(rb); |
77f65ebd | 1098 | struct tpacket_block_desc *pbd = GET_PBLOCK_DESC(pkc, idx); |
f6fb8f10 | 1099 | |
1100 | if (status != BLOCK_STATUS(pbd)) | |
1101 | return NULL; | |
1102 | return pbd; | |
1103 | } | |
1104 | ||
eea49cc9 | 1105 | static int prb_previous_blk_num(struct packet_ring_buffer *rb) |
f6fb8f10 | 1106 | { |
1107 | unsigned int prev; | |
1108 | if (rb->prb_bdqc.kactive_blk_num) | |
1109 | prev = rb->prb_bdqc.kactive_blk_num-1; | |
1110 | else | |
1111 | prev = rb->prb_bdqc.knum_blocks-1; | |
1112 | return prev; | |
1113 | } | |
1114 | ||
1115 | /* Assumes caller has held the rx_queue.lock */ | |
eea49cc9 | 1116 | static void *__prb_previous_block(struct packet_sock *po, |
f6fb8f10 | 1117 | struct packet_ring_buffer *rb, |
1118 | int status) | |
1119 | { | |
1120 | unsigned int previous = prb_previous_blk_num(rb); | |
1121 | return prb_lookup_block(po, rb, previous, status); | |
1122 | } | |
1123 | ||
eea49cc9 | 1124 | static void *packet_previous_rx_frame(struct packet_sock *po, |
f6fb8f10 | 1125 | struct packet_ring_buffer *rb, |
1126 | int status) | |
1127 | { | |
1128 | if (po->tp_version <= TPACKET_V2) | |
1129 | return packet_previous_frame(po, rb, status); | |
1130 | ||
1131 | return __prb_previous_block(po, rb, status); | |
1132 | } | |
1133 | ||
eea49cc9 | 1134 | static void packet_increment_rx_head(struct packet_sock *po, |
f6fb8f10 | 1135 | struct packet_ring_buffer *rb) |
1136 | { | |
1137 | switch (po->tp_version) { | |
1138 | case TPACKET_V1: | |
1139 | case TPACKET_V2: | |
1140 | return packet_increment_head(rb); | |
1141 | case TPACKET_V3: | |
1142 | default: | |
1143 | WARN(1, "TPACKET version not supported.\n"); | |
1144 | BUG(); | |
1145 | return; | |
1146 | } | |
1147 | } | |
1148 | ||
eea49cc9 | 1149 | static void *packet_previous_frame(struct packet_sock *po, |
69e3c75f JB |
1150 | struct packet_ring_buffer *rb, |
1151 | int status) | |
1152 | { | |
1153 | unsigned int previous = rb->head ? rb->head - 1 : rb->frame_max; | |
1154 | return packet_lookup_frame(po, rb, previous, status); | |
1155 | } | |
1156 | ||
eea49cc9 | 1157 | static void packet_increment_head(struct packet_ring_buffer *buff) |
69e3c75f JB |
1158 | { |
1159 | buff->head = buff->head != buff->frame_max ? buff->head+1 : 0; | |
1160 | } | |
1161 | ||
b0138408 DB |
1162 | static void packet_inc_pending(struct packet_ring_buffer *rb) |
1163 | { | |
1164 | this_cpu_inc(*rb->pending_refcnt); | |
1165 | } | |
1166 | ||
1167 | static void packet_dec_pending(struct packet_ring_buffer *rb) | |
1168 | { | |
1169 | this_cpu_dec(*rb->pending_refcnt); | |
1170 | } | |
1171 | ||
1172 | static unsigned int packet_read_pending(const struct packet_ring_buffer *rb) | |
1173 | { | |
1174 | unsigned int refcnt = 0; | |
1175 | int cpu; | |
1176 | ||
1177 | /* We don't use pending refcount in rx_ring. */ | |
1178 | if (rb->pending_refcnt == NULL) | |
1179 | return 0; | |
1180 | ||
1181 | for_each_possible_cpu(cpu) | |
1182 | refcnt += *per_cpu_ptr(rb->pending_refcnt, cpu); | |
1183 | ||
1184 | return refcnt; | |
1185 | } | |
1186 | ||
1187 | static int packet_alloc_pending(struct packet_sock *po) | |
1188 | { | |
1189 | po->rx_ring.pending_refcnt = NULL; | |
1190 | ||
1191 | po->tx_ring.pending_refcnt = alloc_percpu(unsigned int); | |
1192 | if (unlikely(po->tx_ring.pending_refcnt == NULL)) | |
1193 | return -ENOBUFS; | |
1194 | ||
1195 | return 0; | |
1196 | } | |
1197 | ||
1198 | static void packet_free_pending(struct packet_sock *po) | |
1199 | { | |
1200 | free_percpu(po->tx_ring.pending_refcnt); | |
1201 | } | |
1202 | ||
9954729b WB |
1203 | #define ROOM_POW_OFF 2 |
1204 | #define ROOM_NONE 0x0 | |
1205 | #define ROOM_LOW 0x1 | |
1206 | #define ROOM_NORMAL 0x2 | |
1207 | ||
d4b5bd98 | 1208 | static bool __tpacket_has_room(const struct packet_sock *po, int pow_off) |
77f65ebd | 1209 | { |
9954729b WB |
1210 | int idx, len; |
1211 | ||
d4b5bd98 ED |
1212 | len = READ_ONCE(po->rx_ring.frame_max) + 1; |
1213 | idx = READ_ONCE(po->rx_ring.head); | |
9954729b WB |
1214 | if (pow_off) |
1215 | idx += len >> pow_off; | |
1216 | if (idx >= len) | |
1217 | idx -= len; | |
1218 | return packet_lookup_frame(po, &po->rx_ring, idx, TP_STATUS_KERNEL); | |
1219 | } | |
1220 | ||
dcf70cef | 1221 | static bool __tpacket_v3_has_room(const struct packet_sock *po, int pow_off) |
9954729b WB |
1222 | { |
1223 | int idx, len; | |
1224 | ||
dcf70cef ED |
1225 | len = READ_ONCE(po->rx_ring.prb_bdqc.knum_blocks); |
1226 | idx = READ_ONCE(po->rx_ring.prb_bdqc.kactive_blk_num); | |
9954729b WB |
1227 | if (pow_off) |
1228 | idx += len >> pow_off; | |
1229 | if (idx >= len) | |
1230 | idx -= len; | |
1231 | return prb_lookup_block(po, &po->rx_ring, idx, TP_STATUS_KERNEL); | |
1232 | } | |
77f65ebd | 1233 | |
0338a145 ED |
1234 | static int __packet_rcv_has_room(const struct packet_sock *po, |
1235 | const struct sk_buff *skb) | |
9954729b | 1236 | { |
0338a145 | 1237 | const struct sock *sk = &po->sk; |
9954729b WB |
1238 | int ret = ROOM_NONE; |
1239 | ||
1240 | if (po->prot_hook.func != tpacket_rcv) { | |
0338a145 ED |
1241 | int rcvbuf = READ_ONCE(sk->sk_rcvbuf); |
1242 | int avail = rcvbuf - atomic_read(&sk->sk_rmem_alloc) | |
1243 | - (skb ? skb->truesize : 0); | |
1244 | ||
1245 | if (avail > (rcvbuf >> ROOM_POW_OFF)) | |
9954729b WB |
1246 | return ROOM_NORMAL; |
1247 | else if (avail > 0) | |
1248 | return ROOM_LOW; | |
1249 | else | |
1250 | return ROOM_NONE; | |
1251 | } | |
77f65ebd | 1252 | |
9954729b WB |
1253 | if (po->tp_version == TPACKET_V3) { |
1254 | if (__tpacket_v3_has_room(po, ROOM_POW_OFF)) | |
1255 | ret = ROOM_NORMAL; | |
1256 | else if (__tpacket_v3_has_room(po, 0)) | |
1257 | ret = ROOM_LOW; | |
1258 | } else { | |
1259 | if (__tpacket_has_room(po, ROOM_POW_OFF)) | |
1260 | ret = ROOM_NORMAL; | |
1261 | else if (__tpacket_has_room(po, 0)) | |
1262 | ret = ROOM_LOW; | |
1263 | } | |
2ccdbaa6 WB |
1264 | |
1265 | return ret; | |
1266 | } | |
1267 | ||
1268 | static int packet_rcv_has_room(struct packet_sock *po, struct sk_buff *skb) | |
1269 | { | |
3a2bb84e | 1270 | int pressure, ret; |
2ccdbaa6 | 1271 | |
54d7c01d | 1272 | ret = __packet_rcv_has_room(po, skb); |
3a2bb84e ED |
1273 | pressure = ret != ROOM_NORMAL; |
1274 | ||
1275 | if (READ_ONCE(po->pressure) != pressure) | |
1276 | WRITE_ONCE(po->pressure, pressure); | |
77f65ebd | 1277 | |
9954729b | 1278 | return ret; |
77f65ebd WB |
1279 | } |
1280 | ||
9bb6cd65 ED |
1281 | static void packet_rcv_try_clear_pressure(struct packet_sock *po) |
1282 | { | |
1283 | if (READ_ONCE(po->pressure) && | |
1284 | __packet_rcv_has_room(po, NULL) == ROOM_NORMAL) | |
1285 | WRITE_ONCE(po->pressure, 0); | |
1286 | } | |
1287 | ||
1da177e4 LT |
1288 | static void packet_sock_destruct(struct sock *sk) |
1289 | { | |
ed85b565 RC |
1290 | skb_queue_purge(&sk->sk_error_queue); |
1291 | ||
547b792c | 1292 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); |
14afee4b | 1293 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
1da177e4 LT |
1294 | |
1295 | if (!sock_flag(sk, SOCK_DEAD)) { | |
40d4e3df | 1296 | pr_err("Attempt to release alive packet socket: %p\n", sk); |
1da177e4 LT |
1297 | return; |
1298 | } | |
1299 | ||
17ab56a2 | 1300 | sk_refcnt_debug_dec(sk); |
1da177e4 LT |
1301 | } |
1302 | ||
3b3a5b0a WB |
1303 | static bool fanout_flow_is_huge(struct packet_sock *po, struct sk_buff *skb) |
1304 | { | |
b756ad92 ED |
1305 | u32 *history = po->rollover->history; |
1306 | u32 victim, rxhash; | |
3b3a5b0a WB |
1307 | int i, count = 0; |
1308 | ||
1309 | rxhash = skb_get_hash(skb); | |
1310 | for (i = 0; i < ROLLOVER_HLEN; i++) | |
b756ad92 | 1311 | if (READ_ONCE(history[i]) == rxhash) |
3b3a5b0a WB |
1312 | count++; |
1313 | ||
b756ad92 ED |
1314 | victim = prandom_u32() % ROLLOVER_HLEN; |
1315 | ||
1316 | /* Avoid dirtying the cache line if possible */ | |
1317 | if (READ_ONCE(history[victim]) != rxhash) | |
1318 | WRITE_ONCE(history[victim], rxhash); | |
1319 | ||
3b3a5b0a WB |
1320 | return count > (ROLLOVER_HLEN >> 1); |
1321 | } | |
1322 | ||
77f65ebd WB |
1323 | static unsigned int fanout_demux_hash(struct packet_fanout *f, |
1324 | struct sk_buff *skb, | |
1325 | unsigned int num) | |
dc99f600 | 1326 | { |
eb70db87 | 1327 | return reciprocal_scale(__skb_get_hash_symmetric(skb), num); |
dc99f600 DM |
1328 | } |
1329 | ||
77f65ebd WB |
1330 | static unsigned int fanout_demux_lb(struct packet_fanout *f, |
1331 | struct sk_buff *skb, | |
1332 | unsigned int num) | |
dc99f600 | 1333 | { |
468479e6 | 1334 | unsigned int val = atomic_inc_return(&f->rr_cur); |
dc99f600 | 1335 | |
468479e6 | 1336 | return val % num; |
77f65ebd WB |
1337 | } |
1338 | ||
1339 | static unsigned int fanout_demux_cpu(struct packet_fanout *f, | |
1340 | struct sk_buff *skb, | |
1341 | unsigned int num) | |
1342 | { | |
1343 | return smp_processor_id() % num; | |
dc99f600 DM |
1344 | } |
1345 | ||
5df0ddfb DB |
1346 | static unsigned int fanout_demux_rnd(struct packet_fanout *f, |
1347 | struct sk_buff *skb, | |
1348 | unsigned int num) | |
1349 | { | |
f337db64 | 1350 | return prandom_u32_max(num); |
5df0ddfb DB |
1351 | } |
1352 | ||
77f65ebd WB |
1353 | static unsigned int fanout_demux_rollover(struct packet_fanout *f, |
1354 | struct sk_buff *skb, | |
ad377cab | 1355 | unsigned int idx, bool try_self, |
77f65ebd | 1356 | unsigned int num) |
95ec3eb4 | 1357 | { |
4633c9e0 | 1358 | struct packet_sock *po, *po_next, *po_skip = NULL; |
a9b63918 | 1359 | unsigned int i, j, room = ROOM_NONE; |
95ec3eb4 | 1360 | |
0648ab70 | 1361 | po = pkt_sk(f->arr[idx]); |
3b3a5b0a WB |
1362 | |
1363 | if (try_self) { | |
1364 | room = packet_rcv_has_room(po, skb); | |
1365 | if (room == ROOM_NORMAL || | |
1366 | (room == ROOM_LOW && !fanout_flow_is_huge(po, skb))) | |
1367 | return idx; | |
4633c9e0 | 1368 | po_skip = po; |
3b3a5b0a | 1369 | } |
ad377cab | 1370 | |
0648ab70 | 1371 | i = j = min_t(int, po->rollover->sock, num - 1); |
77f65ebd | 1372 | do { |
2ccdbaa6 | 1373 | po_next = pkt_sk(f->arr[i]); |
3a2bb84e | 1374 | if (po_next != po_skip && !READ_ONCE(po_next->pressure) && |
2ccdbaa6 | 1375 | packet_rcv_has_room(po_next, skb) == ROOM_NORMAL) { |
77f65ebd | 1376 | if (i != j) |
0648ab70 | 1377 | po->rollover->sock = i; |
a9b63918 WB |
1378 | atomic_long_inc(&po->rollover->num); |
1379 | if (room == ROOM_LOW) | |
1380 | atomic_long_inc(&po->rollover->num_huge); | |
77f65ebd WB |
1381 | return i; |
1382 | } | |
ad377cab | 1383 | |
77f65ebd WB |
1384 | if (++i == num) |
1385 | i = 0; | |
1386 | } while (i != j); | |
1387 | ||
a9b63918 | 1388 | atomic_long_inc(&po->rollover->num_failed); |
77f65ebd WB |
1389 | return idx; |
1390 | } | |
1391 | ||
2d36097d NH |
1392 | static unsigned int fanout_demux_qm(struct packet_fanout *f, |
1393 | struct sk_buff *skb, | |
1394 | unsigned int num) | |
1395 | { | |
1396 | return skb_get_queue_mapping(skb) % num; | |
1397 | } | |
1398 | ||
47dceb8e WB |
1399 | static unsigned int fanout_demux_bpf(struct packet_fanout *f, |
1400 | struct sk_buff *skb, | |
1401 | unsigned int num) | |
1402 | { | |
1403 | struct bpf_prog *prog; | |
1404 | unsigned int ret = 0; | |
1405 | ||
1406 | rcu_read_lock(); | |
1407 | prog = rcu_dereference(f->bpf_prog); | |
1408 | if (prog) | |
ff936a04 | 1409 | ret = bpf_prog_run_clear_cb(prog, skb) % num; |
47dceb8e WB |
1410 | rcu_read_unlock(); |
1411 | ||
1412 | return ret; | |
1413 | } | |
1414 | ||
77f65ebd WB |
1415 | static bool fanout_has_flag(struct packet_fanout *f, u16 flag) |
1416 | { | |
1417 | return f->flags & (flag >> 8); | |
95ec3eb4 DM |
1418 | } |
1419 | ||
95ec3eb4 DM |
1420 | static int packet_rcv_fanout(struct sk_buff *skb, struct net_device *dev, |
1421 | struct packet_type *pt, struct net_device *orig_dev) | |
dc99f600 DM |
1422 | { |
1423 | struct packet_fanout *f = pt->af_packet_priv; | |
f98f4514 | 1424 | unsigned int num = READ_ONCE(f->num_members); |
19bcf9f2 | 1425 | struct net *net = read_pnet(&f->net); |
dc99f600 | 1426 | struct packet_sock *po; |
77f65ebd | 1427 | unsigned int idx; |
dc99f600 | 1428 | |
19bcf9f2 | 1429 | if (!net_eq(dev_net(dev), net) || !num) { |
dc99f600 DM |
1430 | kfree_skb(skb); |
1431 | return 0; | |
1432 | } | |
1433 | ||
3f34b24a | 1434 | if (fanout_has_flag(f, PACKET_FANOUT_FLAG_DEFRAG)) { |
19bcf9f2 | 1435 | skb = ip_check_defrag(net, skb, IP_DEFRAG_AF_PACKET); |
3f34b24a AD |
1436 | if (!skb) |
1437 | return 0; | |
1438 | } | |
95ec3eb4 DM |
1439 | switch (f->type) { |
1440 | case PACKET_FANOUT_HASH: | |
1441 | default: | |
77f65ebd | 1442 | idx = fanout_demux_hash(f, skb, num); |
95ec3eb4 DM |
1443 | break; |
1444 | case PACKET_FANOUT_LB: | |
77f65ebd | 1445 | idx = fanout_demux_lb(f, skb, num); |
95ec3eb4 DM |
1446 | break; |
1447 | case PACKET_FANOUT_CPU: | |
77f65ebd WB |
1448 | idx = fanout_demux_cpu(f, skb, num); |
1449 | break; | |
5df0ddfb DB |
1450 | case PACKET_FANOUT_RND: |
1451 | idx = fanout_demux_rnd(f, skb, num); | |
1452 | break; | |
2d36097d NH |
1453 | case PACKET_FANOUT_QM: |
1454 | idx = fanout_demux_qm(f, skb, num); | |
1455 | break; | |
77f65ebd | 1456 | case PACKET_FANOUT_ROLLOVER: |
ad377cab | 1457 | idx = fanout_demux_rollover(f, skb, 0, false, num); |
95ec3eb4 | 1458 | break; |
47dceb8e | 1459 | case PACKET_FANOUT_CBPF: |
f2e52095 | 1460 | case PACKET_FANOUT_EBPF: |
47dceb8e WB |
1461 | idx = fanout_demux_bpf(f, skb, num); |
1462 | break; | |
dc99f600 DM |
1463 | } |
1464 | ||
ad377cab WB |
1465 | if (fanout_has_flag(f, PACKET_FANOUT_FLAG_ROLLOVER)) |
1466 | idx = fanout_demux_rollover(f, skb, idx, true, num); | |
dc99f600 | 1467 | |
ad377cab | 1468 | po = pkt_sk(f->arr[idx]); |
dc99f600 DM |
1469 | return po->prot_hook.func(skb, dev, &po->prot_hook, orig_dev); |
1470 | } | |
1471 | ||
fff3321d PE |
1472 | DEFINE_MUTEX(fanout_mutex); |
1473 | EXPORT_SYMBOL_GPL(fanout_mutex); | |
dc99f600 | 1474 | static LIST_HEAD(fanout_list); |
4a69a864 | 1475 | static u16 fanout_next_id; |
dc99f600 DM |
1476 | |
1477 | static void __fanout_link(struct sock *sk, struct packet_sock *po) | |
1478 | { | |
1479 | struct packet_fanout *f = po->fanout; | |
1480 | ||
1481 | spin_lock(&f->lock); | |
1482 | f->arr[f->num_members] = sk; | |
1483 | smp_wmb(); | |
1484 | f->num_members++; | |
2bd624b4 AS |
1485 | if (f->num_members == 1) |
1486 | dev_add_pack(&f->prot_hook); | |
dc99f600 DM |
1487 | spin_unlock(&f->lock); |
1488 | } | |
1489 | ||
1490 | static void __fanout_unlink(struct sock *sk, struct packet_sock *po) | |
1491 | { | |
1492 | struct packet_fanout *f = po->fanout; | |
1493 | int i; | |
1494 | ||
1495 | spin_lock(&f->lock); | |
1496 | for (i = 0; i < f->num_members; i++) { | |
1497 | if (f->arr[i] == sk) | |
1498 | break; | |
1499 | } | |
1500 | BUG_ON(i >= f->num_members); | |
1501 | f->arr[i] = f->arr[f->num_members - 1]; | |
1502 | f->num_members--; | |
2bd624b4 AS |
1503 | if (f->num_members == 0) |
1504 | __dev_remove_pack(&f->prot_hook); | |
dc99f600 DM |
1505 | spin_unlock(&f->lock); |
1506 | } | |
1507 | ||
d4dd8aee | 1508 | static bool match_fanout_group(struct packet_type *ptype, struct sock *sk) |
c0de08d0 | 1509 | { |
161642e2 ED |
1510 | if (sk->sk_family != PF_PACKET) |
1511 | return false; | |
c0de08d0 | 1512 | |
161642e2 | 1513 | return ptype->af_packet_priv == pkt_sk(sk)->fanout; |
c0de08d0 EL |
1514 | } |
1515 | ||
47dceb8e WB |
1516 | static void fanout_init_data(struct packet_fanout *f) |
1517 | { | |
1518 | switch (f->type) { | |
1519 | case PACKET_FANOUT_LB: | |
1520 | atomic_set(&f->rr_cur, 0); | |
1521 | break; | |
1522 | case PACKET_FANOUT_CBPF: | |
f2e52095 | 1523 | case PACKET_FANOUT_EBPF: |
47dceb8e WB |
1524 | RCU_INIT_POINTER(f->bpf_prog, NULL); |
1525 | break; | |
1526 | } | |
1527 | } | |
1528 | ||
1529 | static void __fanout_set_data_bpf(struct packet_fanout *f, struct bpf_prog *new) | |
1530 | { | |
1531 | struct bpf_prog *old; | |
1532 | ||
1533 | spin_lock(&f->lock); | |
1534 | old = rcu_dereference_protected(f->bpf_prog, lockdep_is_held(&f->lock)); | |
1535 | rcu_assign_pointer(f->bpf_prog, new); | |
1536 | spin_unlock(&f->lock); | |
1537 | ||
1538 | if (old) { | |
1539 | synchronize_net(); | |
1540 | bpf_prog_destroy(old); | |
1541 | } | |
1542 | } | |
1543 | ||
b1ea9ff6 | 1544 | static int fanout_set_data_cbpf(struct packet_sock *po, sockptr_t data, |
47dceb8e WB |
1545 | unsigned int len) |
1546 | { | |
1547 | struct bpf_prog *new; | |
1548 | struct sock_fprog fprog; | |
1549 | int ret; | |
1550 | ||
1551 | if (sock_flag(&po->sk, SOCK_FILTER_LOCKED)) | |
1552 | return -EPERM; | |
4d295e54 CH |
1553 | |
1554 | ret = copy_bpf_fprog_from_user(&fprog, data, len); | |
1555 | if (ret) | |
1556 | return ret; | |
47dceb8e | 1557 | |
bab18991 | 1558 | ret = bpf_prog_create_from_user(&new, &fprog, NULL, false); |
47dceb8e WB |
1559 | if (ret) |
1560 | return ret; | |
1561 | ||
1562 | __fanout_set_data_bpf(po->fanout, new); | |
1563 | return 0; | |
1564 | } | |
1565 | ||
a7b75c5a | 1566 | static int fanout_set_data_ebpf(struct packet_sock *po, sockptr_t data, |
f2e52095 WB |
1567 | unsigned int len) |
1568 | { | |
1569 | struct bpf_prog *new; | |
1570 | u32 fd; | |
1571 | ||
1572 | if (sock_flag(&po->sk, SOCK_FILTER_LOCKED)) | |
1573 | return -EPERM; | |
1574 | if (len != sizeof(fd)) | |
1575 | return -EINVAL; | |
a7b75c5a | 1576 | if (copy_from_sockptr(&fd, data, len)) |
f2e52095 WB |
1577 | return -EFAULT; |
1578 | ||
113214be | 1579 | new = bpf_prog_get_type(fd, BPF_PROG_TYPE_SOCKET_FILTER); |
f2e52095 WB |
1580 | if (IS_ERR(new)) |
1581 | return PTR_ERR(new); | |
f2e52095 WB |
1582 | |
1583 | __fanout_set_data_bpf(po->fanout, new); | |
1584 | return 0; | |
1585 | } | |
1586 | ||
a7b75c5a | 1587 | static int fanout_set_data(struct packet_sock *po, sockptr_t data, |
47dceb8e WB |
1588 | unsigned int len) |
1589 | { | |
1590 | switch (po->fanout->type) { | |
1591 | case PACKET_FANOUT_CBPF: | |
a7b75c5a | 1592 | return fanout_set_data_cbpf(po, data, len); |
f2e52095 WB |
1593 | case PACKET_FANOUT_EBPF: |
1594 | return fanout_set_data_ebpf(po, data, len); | |
47dceb8e WB |
1595 | default: |
1596 | return -EINVAL; | |
07d53ae4 | 1597 | } |
47dceb8e WB |
1598 | } |
1599 | ||
1600 | static void fanout_release_data(struct packet_fanout *f) | |
1601 | { | |
1602 | switch (f->type) { | |
1603 | case PACKET_FANOUT_CBPF: | |
f2e52095 | 1604 | case PACKET_FANOUT_EBPF: |
47dceb8e | 1605 | __fanout_set_data_bpf(f, NULL); |
07d53ae4 | 1606 | } |
47dceb8e WB |
1607 | } |
1608 | ||
4a69a864 MM |
1609 | static bool __fanout_id_is_free(struct sock *sk, u16 candidate_id) |
1610 | { | |
1611 | struct packet_fanout *f; | |
1612 | ||
1613 | list_for_each_entry(f, &fanout_list, list) { | |
1614 | if (f->id == candidate_id && | |
1615 | read_pnet(&f->net) == sock_net(sk)) { | |
1616 | return false; | |
1617 | } | |
1618 | } | |
1619 | return true; | |
1620 | } | |
1621 | ||
1622 | static bool fanout_find_new_id(struct sock *sk, u16 *new_id) | |
1623 | { | |
1624 | u16 id = fanout_next_id; | |
1625 | ||
1626 | do { | |
1627 | if (__fanout_id_is_free(sk, id)) { | |
1628 | *new_id = id; | |
1629 | fanout_next_id = id + 1; | |
1630 | return true; | |
1631 | } | |
1632 | ||
1633 | id++; | |
1634 | } while (id != fanout_next_id); | |
1635 | ||
1636 | return false; | |
1637 | } | |
1638 | ||
7736d33f | 1639 | static int fanout_add(struct sock *sk, u16 id, u16 type_flags) |
dc99f600 | 1640 | { |
d199fab6 | 1641 | struct packet_rollover *rollover = NULL; |
dc99f600 DM |
1642 | struct packet_sock *po = pkt_sk(sk); |
1643 | struct packet_fanout *f, *match; | |
7736d33f | 1644 | u8 type = type_flags & 0xff; |
77f65ebd | 1645 | u8 flags = type_flags >> 8; |
dc99f600 DM |
1646 | int err; |
1647 | ||
1648 | switch (type) { | |
77f65ebd WB |
1649 | case PACKET_FANOUT_ROLLOVER: |
1650 | if (type_flags & PACKET_FANOUT_FLAG_ROLLOVER) | |
1651 | return -EINVAL; | |
dc99f600 DM |
1652 | case PACKET_FANOUT_HASH: |
1653 | case PACKET_FANOUT_LB: | |
95ec3eb4 | 1654 | case PACKET_FANOUT_CPU: |
5df0ddfb | 1655 | case PACKET_FANOUT_RND: |
2d36097d | 1656 | case PACKET_FANOUT_QM: |
47dceb8e | 1657 | case PACKET_FANOUT_CBPF: |
f2e52095 | 1658 | case PACKET_FANOUT_EBPF: |
dc99f600 DM |
1659 | break; |
1660 | default: | |
1661 | return -EINVAL; | |
1662 | } | |
1663 | ||
d199fab6 ED |
1664 | mutex_lock(&fanout_mutex); |
1665 | ||
d199fab6 | 1666 | err = -EALREADY; |
dc99f600 | 1667 | if (po->fanout) |
d199fab6 | 1668 | goto out; |
dc99f600 | 1669 | |
4633c9e0 WB |
1670 | if (type == PACKET_FANOUT_ROLLOVER || |
1671 | (type_flags & PACKET_FANOUT_FLAG_ROLLOVER)) { | |
d199fab6 ED |
1672 | err = -ENOMEM; |
1673 | rollover = kzalloc(sizeof(*rollover), GFP_KERNEL); | |
1674 | if (!rollover) | |
1675 | goto out; | |
1676 | atomic_long_set(&rollover->num, 0); | |
1677 | atomic_long_set(&rollover->num_huge, 0); | |
1678 | atomic_long_set(&rollover->num_failed, 0); | |
0648ab70 WB |
1679 | } |
1680 | ||
4a69a864 MM |
1681 | if (type_flags & PACKET_FANOUT_FLAG_UNIQUEID) { |
1682 | if (id != 0) { | |
1683 | err = -EINVAL; | |
1684 | goto out; | |
1685 | } | |
1686 | if (!fanout_find_new_id(sk, &id)) { | |
1687 | err = -ENOMEM; | |
1688 | goto out; | |
1689 | } | |
1690 | /* ephemeral flag for the first socket in the group: drop it */ | |
1691 | flags &= ~(PACKET_FANOUT_FLAG_UNIQUEID >> 8); | |
1692 | } | |
1693 | ||
dc99f600 DM |
1694 | match = NULL; |
1695 | list_for_each_entry(f, &fanout_list, list) { | |
1696 | if (f->id == id && | |
1697 | read_pnet(&f->net) == sock_net(sk)) { | |
1698 | match = f; | |
1699 | break; | |
1700 | } | |
1701 | } | |
afe62c68 | 1702 | err = -EINVAL; |
77f65ebd | 1703 | if (match && match->flags != flags) |
afe62c68 | 1704 | goto out; |
dc99f600 | 1705 | if (!match) { |
afe62c68 | 1706 | err = -ENOMEM; |
dc99f600 | 1707 | match = kzalloc(sizeof(*match), GFP_KERNEL); |
afe62c68 ED |
1708 | if (!match) |
1709 | goto out; | |
1710 | write_pnet(&match->net, sock_net(sk)); | |
1711 | match->id = id; | |
1712 | match->type = type; | |
77f65ebd | 1713 | match->flags = flags; |
afe62c68 ED |
1714 | INIT_LIST_HEAD(&match->list); |
1715 | spin_lock_init(&match->lock); | |
fb5c2c17 | 1716 | refcount_set(&match->sk_ref, 0); |
47dceb8e | 1717 | fanout_init_data(match); |
afe62c68 ED |
1718 | match->prot_hook.type = po->prot_hook.type; |
1719 | match->prot_hook.dev = po->prot_hook.dev; | |
1720 | match->prot_hook.func = packet_rcv_fanout; | |
1721 | match->prot_hook.af_packet_priv = match; | |
c0de08d0 | 1722 | match->prot_hook.id_match = match_fanout_group; |
afe62c68 | 1723 | list_add(&match->list, &fanout_list); |
dc99f600 | 1724 | } |
afe62c68 | 1725 | err = -EINVAL; |
008ba2a1 WB |
1726 | |
1727 | spin_lock(&po->bind_lock); | |
1728 | if (po->running && | |
1729 | match->type == type && | |
afe62c68 ED |
1730 | match->prot_hook.type == po->prot_hook.type && |
1731 | match->prot_hook.dev == po->prot_hook.dev) { | |
1732 | err = -ENOSPC; | |
fb5c2c17 | 1733 | if (refcount_read(&match->sk_ref) < PACKET_FANOUT_MAX) { |
afe62c68 ED |
1734 | __dev_remove_pack(&po->prot_hook); |
1735 | po->fanout = match; | |
57f015f5 MM |
1736 | po->rollover = rollover; |
1737 | rollover = NULL; | |
fb5c2c17 | 1738 | refcount_set(&match->sk_ref, refcount_read(&match->sk_ref) + 1); |
afe62c68 ED |
1739 | __fanout_link(sk, po); |
1740 | err = 0; | |
dc99f600 DM |
1741 | } |
1742 | } | |
008ba2a1 WB |
1743 | spin_unlock(&po->bind_lock); |
1744 | ||
1745 | if (err && !refcount_read(&match->sk_ref)) { | |
1746 | list_del(&match->list); | |
1747 | kfree(match); | |
1748 | } | |
1749 | ||
afe62c68 | 1750 | out: |
57f015f5 | 1751 | kfree(rollover); |
d199fab6 | 1752 | mutex_unlock(&fanout_mutex); |
dc99f600 DM |
1753 | return err; |
1754 | } | |
1755 | ||
2bd624b4 AS |
1756 | /* If pkt_sk(sk)->fanout->sk_ref is zero, this function removes |
1757 | * pkt_sk(sk)->fanout from fanout_list and returns pkt_sk(sk)->fanout. | |
1758 | * It is the responsibility of the caller to call fanout_release_data() and | |
1759 | * free the returned packet_fanout (after synchronize_net()) | |
1760 | */ | |
1761 | static struct packet_fanout *fanout_release(struct sock *sk) | |
dc99f600 DM |
1762 | { |
1763 | struct packet_sock *po = pkt_sk(sk); | |
1764 | struct packet_fanout *f; | |
1765 | ||
fff3321d | 1766 | mutex_lock(&fanout_mutex); |
d199fab6 ED |
1767 | f = po->fanout; |
1768 | if (f) { | |
1769 | po->fanout = NULL; | |
1770 | ||
fb5c2c17 | 1771 | if (refcount_dec_and_test(&f->sk_ref)) |
d199fab6 | 1772 | list_del(&f->list); |
2bd624b4 AS |
1773 | else |
1774 | f = NULL; | |
dc99f600 DM |
1775 | } |
1776 | mutex_unlock(&fanout_mutex); | |
2bd624b4 AS |
1777 | |
1778 | return f; | |
dc99f600 | 1779 | } |
1da177e4 | 1780 | |
3c70c132 DB |
1781 | static bool packet_extra_vlan_len_allowed(const struct net_device *dev, |
1782 | struct sk_buff *skb) | |
1783 | { | |
1784 | /* Earlier code assumed this would be a VLAN pkt, double-check | |
1785 | * this now that we have the actual packet in hand. We can only | |
1786 | * do this check on Ethernet devices. | |
1787 | */ | |
1788 | if (unlikely(dev->type != ARPHRD_ETHER)) | |
1789 | return false; | |
1790 | ||
1791 | skb_reset_mac_header(skb); | |
1792 | return likely(eth_hdr(skb)->h_proto == htons(ETH_P_8021Q)); | |
1793 | } | |
1794 | ||
90ddc4f0 | 1795 | static const struct proto_ops packet_ops; |
1da177e4 | 1796 | |
90ddc4f0 | 1797 | static const struct proto_ops packet_ops_spkt; |
1da177e4 | 1798 | |
40d4e3df ED |
1799 | static int packet_rcv_spkt(struct sk_buff *skb, struct net_device *dev, |
1800 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
1801 | { |
1802 | struct sock *sk; | |
1803 | struct sockaddr_pkt *spkt; | |
1804 | ||
1805 | /* | |
1806 | * When we registered the protocol we saved the socket in the data | |
1807 | * field for just this event. | |
1808 | */ | |
1809 | ||
1810 | sk = pt->af_packet_priv; | |
1ce4f28b | 1811 | |
1da177e4 LT |
1812 | /* |
1813 | * Yank back the headers [hope the device set this | |
1814 | * right or kerboom...] | |
1815 | * | |
1816 | * Incoming packets have ll header pulled, | |
1817 | * push it back. | |
1818 | * | |
98e399f8 | 1819 | * For outgoing ones skb->data == skb_mac_header(skb) |
1da177e4 LT |
1820 | * so that this procedure is noop. |
1821 | */ | |
1822 | ||
1823 | if (skb->pkt_type == PACKET_LOOPBACK) | |
1824 | goto out; | |
1825 | ||
09ad9bc7 | 1826 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
1827 | goto out; |
1828 | ||
40d4e3df ED |
1829 | skb = skb_share_check(skb, GFP_ATOMIC); |
1830 | if (skb == NULL) | |
1da177e4 LT |
1831 | goto oom; |
1832 | ||
1833 | /* drop any routing info */ | |
adf30907 | 1834 | skb_dst_drop(skb); |
1da177e4 | 1835 | |
84531c24 | 1836 | /* drop conntrack reference */ |
895b5c9f | 1837 | nf_reset_ct(skb); |
84531c24 | 1838 | |
ffbc6111 | 1839 | spkt = &PACKET_SKB_CB(skb)->sa.pkt; |
1da177e4 | 1840 | |
98e399f8 | 1841 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
1842 | |
1843 | /* | |
1844 | * The SOCK_PACKET socket receives _all_ frames. | |
1845 | */ | |
1846 | ||
1847 | spkt->spkt_family = dev->type; | |
1848 | strlcpy(spkt->spkt_device, dev->name, sizeof(spkt->spkt_device)); | |
1849 | spkt->spkt_protocol = skb->protocol; | |
1850 | ||
1851 | /* | |
1852 | * Charge the memory to the socket. This is done specifically | |
1853 | * to prevent sockets using all the memory up. | |
1854 | */ | |
1855 | ||
40d4e3df | 1856 | if (sock_queue_rcv_skb(sk, skb) == 0) |
1da177e4 LT |
1857 | return 0; |
1858 | ||
1859 | out: | |
1860 | kfree_skb(skb); | |
1861 | oom: | |
1862 | return 0; | |
1863 | } | |
1864 | ||
75c65772 MM |
1865 | static void packet_parse_headers(struct sk_buff *skb, struct socket *sock) |
1866 | { | |
18bed891 YK |
1867 | if ((!skb->protocol || skb->protocol == htons(ETH_P_ALL)) && |
1868 | sock->type == SOCK_RAW) { | |
75c65772 MM |
1869 | skb_reset_mac_header(skb); |
1870 | skb->protocol = dev_parse_header_protocol(skb); | |
1871 | } | |
1872 | ||
1873 | skb_probe_transport_header(skb); | |
1874 | } | |
1da177e4 LT |
1875 | |
1876 | /* | |
1877 | * Output a raw packet to a device layer. This bypasses all the other | |
1878 | * protocol layers and you must therefore supply it with a complete frame | |
1879 | */ | |
1ce4f28b | 1880 | |
1b784140 YX |
1881 | static int packet_sendmsg_spkt(struct socket *sock, struct msghdr *msg, |
1882 | size_t len) | |
1da177e4 LT |
1883 | { |
1884 | struct sock *sk = sock->sk; | |
342dfc30 | 1885 | DECLARE_SOCKADDR(struct sockaddr_pkt *, saddr, msg->msg_name); |
1a35ca80 | 1886 | struct sk_buff *skb = NULL; |
1da177e4 | 1887 | struct net_device *dev; |
c14ac945 | 1888 | struct sockcm_cookie sockc; |
40d4e3df | 1889 | __be16 proto = 0; |
1da177e4 | 1890 | int err; |
3bdc0eba | 1891 | int extra_len = 0; |
1ce4f28b | 1892 | |
1da177e4 | 1893 | /* |
1ce4f28b | 1894 | * Get and verify the address. |
1da177e4 LT |
1895 | */ |
1896 | ||
40d4e3df | 1897 | if (saddr) { |
1da177e4 | 1898 | if (msg->msg_namelen < sizeof(struct sockaddr)) |
40d4e3df ED |
1899 | return -EINVAL; |
1900 | if (msg->msg_namelen == sizeof(struct sockaddr_pkt)) | |
1901 | proto = saddr->spkt_protocol; | |
1902 | } else | |
1903 | return -ENOTCONN; /* SOCK_PACKET must be sent giving an address */ | |
1da177e4 LT |
1904 | |
1905 | /* | |
1ce4f28b | 1906 | * Find the device first to size check it |
1da177e4 LT |
1907 | */ |
1908 | ||
de74e92a | 1909 | saddr->spkt_device[sizeof(saddr->spkt_device) - 1] = 0; |
1a35ca80 | 1910 | retry: |
654d1f8a ED |
1911 | rcu_read_lock(); |
1912 | dev = dev_get_by_name_rcu(sock_net(sk), saddr->spkt_device); | |
1da177e4 LT |
1913 | err = -ENODEV; |
1914 | if (dev == NULL) | |
1915 | goto out_unlock; | |
1ce4f28b | 1916 | |
d5e76b0a DM |
1917 | err = -ENETDOWN; |
1918 | if (!(dev->flags & IFF_UP)) | |
1919 | goto out_unlock; | |
1920 | ||
1da177e4 | 1921 | /* |
40d4e3df ED |
1922 | * You may not queue a frame bigger than the mtu. This is the lowest level |
1923 | * raw protocol and you must do your own fragmentation at this level. | |
1da177e4 | 1924 | */ |
1ce4f28b | 1925 | |
3bdc0eba BG |
1926 | if (unlikely(sock_flag(sk, SOCK_NOFCS))) { |
1927 | if (!netif_supports_nofcs(dev)) { | |
1928 | err = -EPROTONOSUPPORT; | |
1929 | goto out_unlock; | |
1930 | } | |
1931 | extra_len = 4; /* We're doing our own CRC */ | |
1932 | } | |
1933 | ||
1da177e4 | 1934 | err = -EMSGSIZE; |
3bdc0eba | 1935 | if (len > dev->mtu + dev->hard_header_len + VLAN_HLEN + extra_len) |
1da177e4 LT |
1936 | goto out_unlock; |
1937 | ||
1a35ca80 ED |
1938 | if (!skb) { |
1939 | size_t reserved = LL_RESERVED_SPACE(dev); | |
4ce40912 | 1940 | int tlen = dev->needed_tailroom; |
1a35ca80 ED |
1941 | unsigned int hhlen = dev->header_ops ? dev->hard_header_len : 0; |
1942 | ||
1943 | rcu_read_unlock(); | |
4ce40912 | 1944 | skb = sock_wmalloc(sk, len + reserved + tlen, 0, GFP_KERNEL); |
1a35ca80 ED |
1945 | if (skb == NULL) |
1946 | return -ENOBUFS; | |
1947 | /* FIXME: Save some space for broken drivers that write a hard | |
1948 | * header at transmission time by themselves. PPP is the notable | |
1949 | * one here. This should really be fixed at the driver level. | |
1950 | */ | |
1951 | skb_reserve(skb, reserved); | |
1952 | skb_reset_network_header(skb); | |
1953 | ||
1954 | /* Try to align data part correctly */ | |
1955 | if (hhlen) { | |
1956 | skb->data -= hhlen; | |
1957 | skb->tail -= hhlen; | |
1958 | if (len < hhlen) | |
1959 | skb_reset_network_header(skb); | |
1960 | } | |
6ce8e9ce | 1961 | err = memcpy_from_msg(skb_put(skb, len), msg, len); |
1a35ca80 ED |
1962 | if (err) |
1963 | goto out_free; | |
1964 | goto retry; | |
1da177e4 LT |
1965 | } |
1966 | ||
9ed988cd WB |
1967 | if (!dev_validate_header(dev, skb->data, len)) { |
1968 | err = -EINVAL; | |
1969 | goto out_unlock; | |
1970 | } | |
3c70c132 DB |
1971 | if (len > (dev->mtu + dev->hard_header_len + extra_len) && |
1972 | !packet_extra_vlan_len_allowed(dev, skb)) { | |
1973 | err = -EMSGSIZE; | |
1974 | goto out_unlock; | |
57f89bfa | 1975 | } |
1a35ca80 | 1976 | |
657a0667 | 1977 | sockcm_init(&sockc, sk); |
c14ac945 SHY |
1978 | if (msg->msg_controllen) { |
1979 | err = sock_cmsg_send(sk, msg, &sockc); | |
f8e7718c | 1980 | if (unlikely(err)) |
c14ac945 | 1981 | goto out_unlock; |
c14ac945 SHY |
1982 | } |
1983 | ||
1da177e4 LT |
1984 | skb->protocol = proto; |
1985 | skb->dev = dev; | |
1986 | skb->priority = sk->sk_priority; | |
2d37a186 | 1987 | skb->mark = sk->sk_mark; |
3d0ba8c0 | 1988 | skb->tstamp = sockc.transmit_time; |
bf84a010 | 1989 | |
8f932f76 | 1990 | skb_setup_tx_timestamp(skb, sockc.tsflags); |
1da177e4 | 1991 | |
3bdc0eba BG |
1992 | if (unlikely(extra_len == 4)) |
1993 | skb->no_fcs = 1; | |
1994 | ||
75c65772 | 1995 | packet_parse_headers(skb, sock); |
c1aad275 | 1996 | |
1da177e4 | 1997 | dev_queue_xmit(skb); |
654d1f8a | 1998 | rcu_read_unlock(); |
40d4e3df | 1999 | return len; |
1da177e4 | 2000 | |
1da177e4 | 2001 | out_unlock: |
654d1f8a | 2002 | rcu_read_unlock(); |
1a35ca80 ED |
2003 | out_free: |
2004 | kfree_skb(skb); | |
1da177e4 LT |
2005 | return err; |
2006 | } | |
1da177e4 | 2007 | |
ff936a04 AS |
2008 | static unsigned int run_filter(struct sk_buff *skb, |
2009 | const struct sock *sk, | |
2010 | unsigned int res) | |
1da177e4 LT |
2011 | { |
2012 | struct sk_filter *filter; | |
fda9ef5d | 2013 | |
80f8f102 ED |
2014 | rcu_read_lock(); |
2015 | filter = rcu_dereference(sk->sk_filter); | |
dbcb5855 | 2016 | if (filter != NULL) |
ff936a04 | 2017 | res = bpf_prog_run_clear_cb(filter->prog, skb); |
80f8f102 | 2018 | rcu_read_unlock(); |
1da177e4 | 2019 | |
dbcb5855 | 2020 | return res; |
1da177e4 LT |
2021 | } |
2022 | ||
16cc1400 WB |
2023 | static int packet_rcv_vnet(struct msghdr *msg, const struct sk_buff *skb, |
2024 | size_t *len) | |
2025 | { | |
2026 | struct virtio_net_hdr vnet_hdr; | |
2027 | ||
2028 | if (*len < sizeof(vnet_hdr)) | |
2029 | return -EINVAL; | |
2030 | *len -= sizeof(vnet_hdr); | |
2031 | ||
fd3a8862 | 2032 | if (virtio_net_hdr_from_skb(skb, &vnet_hdr, vio_le(), true, 0)) |
16cc1400 WB |
2033 | return -EINVAL; |
2034 | ||
2035 | return memcpy_to_msg(msg, (void *)&vnet_hdr, sizeof(vnet_hdr)); | |
2036 | } | |
2037 | ||
1da177e4 | 2038 | /* |
62ab0812 ED |
2039 | * This function makes lazy skb cloning in hope that most of packets |
2040 | * are discarded by BPF. | |
2041 | * | |
2042 | * Note tricky part: we DO mangle shared skb! skb->data, skb->len | |
2043 | * and skb->cb are mangled. It works because (and until) packets | |
2044 | * falling here are owned by current CPU. Output packets are cloned | |
2045 | * by dev_queue_xmit_nit(), input packets are processed by net_bh | |
2046 | * sequencially, so that if we return skb to original state on exit, | |
2047 | * we will not harm anyone. | |
1da177e4 LT |
2048 | */ |
2049 | ||
40d4e3df ED |
2050 | static int packet_rcv(struct sk_buff *skb, struct net_device *dev, |
2051 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
2052 | { |
2053 | struct sock *sk; | |
2054 | struct sockaddr_ll *sll; | |
2055 | struct packet_sock *po; | |
40d4e3df | 2056 | u8 *skb_head = skb->data; |
1da177e4 | 2057 | int skb_len = skb->len; |
dbcb5855 | 2058 | unsigned int snaplen, res; |
da37845f | 2059 | bool is_drop_n_account = false; |
1da177e4 LT |
2060 | |
2061 | if (skb->pkt_type == PACKET_LOOPBACK) | |
2062 | goto drop; | |
2063 | ||
2064 | sk = pt->af_packet_priv; | |
2065 | po = pkt_sk(sk); | |
2066 | ||
09ad9bc7 | 2067 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
2068 | goto drop; |
2069 | ||
1da177e4 LT |
2070 | skb->dev = dev; |
2071 | ||
3b04ddde | 2072 | if (dev->header_ops) { |
1da177e4 | 2073 | /* The device has an explicit notion of ll header, |
62ab0812 ED |
2074 | * exported to higher levels. |
2075 | * | |
2076 | * Otherwise, the device hides details of its frame | |
2077 | * structure, so that corresponding packet head is | |
2078 | * never delivered to user. | |
1da177e4 LT |
2079 | */ |
2080 | if (sk->sk_type != SOCK_DGRAM) | |
98e399f8 | 2081 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
2082 | else if (skb->pkt_type == PACKET_OUTGOING) { |
2083 | /* Special case: outgoing packets have ll header at head */ | |
bbe735e4 | 2084 | skb_pull(skb, skb_network_offset(skb)); |
1da177e4 LT |
2085 | } |
2086 | } | |
2087 | ||
2088 | snaplen = skb->len; | |
2089 | ||
dbcb5855 DM |
2090 | res = run_filter(skb, sk, snaplen); |
2091 | if (!res) | |
fda9ef5d | 2092 | goto drop_n_restore; |
dbcb5855 DM |
2093 | if (snaplen > res) |
2094 | snaplen = res; | |
1da177e4 | 2095 | |
0fd7bac6 | 2096 | if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) |
1da177e4 LT |
2097 | goto drop_n_acct; |
2098 | ||
2099 | if (skb_shared(skb)) { | |
2100 | struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC); | |
2101 | if (nskb == NULL) | |
2102 | goto drop_n_acct; | |
2103 | ||
2104 | if (skb_head != skb->data) { | |
2105 | skb->data = skb_head; | |
2106 | skb->len = skb_len; | |
2107 | } | |
abc4e4fa | 2108 | consume_skb(skb); |
1da177e4 LT |
2109 | skb = nskb; |
2110 | } | |
2111 | ||
b4772ef8 | 2112 | sock_skb_cb_check_size(sizeof(*PACKET_SKB_CB(skb)) + MAX_ADDR_LEN - 8); |
ffbc6111 HX |
2113 | |
2114 | sll = &PACKET_SKB_CB(skb)->sa.ll; | |
1da177e4 | 2115 | sll->sll_hatype = dev->type; |
1da177e4 | 2116 | sll->sll_pkttype = skb->pkt_type; |
8032b464 | 2117 | if (unlikely(po->origdev)) |
80feaacb PWJ |
2118 | sll->sll_ifindex = orig_dev->ifindex; |
2119 | else | |
2120 | sll->sll_ifindex = dev->ifindex; | |
1da177e4 | 2121 | |
b95cce35 | 2122 | sll->sll_halen = dev_parse_header(skb, sll->sll_addr); |
1da177e4 | 2123 | |
2472d761 EB |
2124 | /* sll->sll_family and sll->sll_protocol are set in packet_recvmsg(). |
2125 | * Use their space for storing the original skb length. | |
2126 | */ | |
2127 | PACKET_SKB_CB(skb)->sa.origlen = skb->len; | |
8dc41944 | 2128 | |
1da177e4 LT |
2129 | if (pskb_trim(skb, snaplen)) |
2130 | goto drop_n_acct; | |
2131 | ||
2132 | skb_set_owner_r(skb, sk); | |
2133 | skb->dev = NULL; | |
adf30907 | 2134 | skb_dst_drop(skb); |
1da177e4 | 2135 | |
84531c24 | 2136 | /* drop conntrack reference */ |
895b5c9f | 2137 | nf_reset_ct(skb); |
84531c24 | 2138 | |
1da177e4 | 2139 | spin_lock(&sk->sk_receive_queue.lock); |
ee80fbf3 | 2140 | po->stats.stats1.tp_packets++; |
3bc3b96f | 2141 | sock_skb_set_dropcount(sk, skb); |
1da177e4 LT |
2142 | __skb_queue_tail(&sk->sk_receive_queue, skb); |
2143 | spin_unlock(&sk->sk_receive_queue.lock); | |
676d2369 | 2144 | sk->sk_data_ready(sk); |
1da177e4 LT |
2145 | return 0; |
2146 | ||
2147 | drop_n_acct: | |
da37845f | 2148 | is_drop_n_account = true; |
8e8e2951 | 2149 | atomic_inc(&po->tp_drops); |
7091fbd8 | 2150 | atomic_inc(&sk->sk_drops); |
1da177e4 LT |
2151 | |
2152 | drop_n_restore: | |
2153 | if (skb_head != skb->data && skb_shared(skb)) { | |
2154 | skb->data = skb_head; | |
2155 | skb->len = skb_len; | |
2156 | } | |
2157 | drop: | |
da37845f WJ |
2158 | if (!is_drop_n_account) |
2159 | consume_skb(skb); | |
2160 | else | |
2161 | kfree_skb(skb); | |
1da177e4 LT |
2162 | return 0; |
2163 | } | |
2164 | ||
40d4e3df ED |
2165 | static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, |
2166 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
2167 | { |
2168 | struct sock *sk; | |
2169 | struct packet_sock *po; | |
2170 | struct sockaddr_ll *sll; | |
184f489e | 2171 | union tpacket_uhdr h; |
40d4e3df | 2172 | u8 *skb_head = skb->data; |
1da177e4 | 2173 | int skb_len = skb->len; |
dbcb5855 | 2174 | unsigned int snaplen, res; |
f6fb8f10 | 2175 | unsigned long status = TP_STATUS_USER; |
acf69c94 OC |
2176 | unsigned short macoff, hdrlen; |
2177 | unsigned int netoff; | |
1da177e4 | 2178 | struct sk_buff *copy_skb = NULL; |
d413fcb4 | 2179 | struct timespec64 ts; |
b9c32fb2 | 2180 | __u32 ts_status; |
da37845f | 2181 | bool is_drop_n_account = false; |
61fad681 | 2182 | unsigned int slot_id = 0; |
edbd58be | 2183 | bool do_vnet = false; |
1da177e4 | 2184 | |
51846355 AW |
2185 | /* struct tpacket{2,3}_hdr is aligned to a multiple of TPACKET_ALIGNMENT. |
2186 | * We may add members to them until current aligned size without forcing | |
2187 | * userspace to call getsockopt(..., PACKET_HDRLEN, ...). | |
2188 | */ | |
2189 | BUILD_BUG_ON(TPACKET_ALIGN(sizeof(*h.h2)) != 32); | |
2190 | BUILD_BUG_ON(TPACKET_ALIGN(sizeof(*h.h3)) != 48); | |
2191 | ||
1da177e4 LT |
2192 | if (skb->pkt_type == PACKET_LOOPBACK) |
2193 | goto drop; | |
2194 | ||
2195 | sk = pt->af_packet_priv; | |
2196 | po = pkt_sk(sk); | |
2197 | ||
09ad9bc7 | 2198 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
2199 | goto drop; |
2200 | ||
3b04ddde | 2201 | if (dev->header_ops) { |
1da177e4 | 2202 | if (sk->sk_type != SOCK_DGRAM) |
98e399f8 | 2203 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
2204 | else if (skb->pkt_type == PACKET_OUTGOING) { |
2205 | /* Special case: outgoing packets have ll header at head */ | |
bbe735e4 | 2206 | skb_pull(skb, skb_network_offset(skb)); |
1da177e4 LT |
2207 | } |
2208 | } | |
2209 | ||
2210 | snaplen = skb->len; | |
2211 | ||
dbcb5855 DM |
2212 | res = run_filter(skb, sk, snaplen); |
2213 | if (!res) | |
fda9ef5d | 2214 | goto drop_n_restore; |
68c2e5de | 2215 | |
2c51c627 ED |
2216 | /* If we are flooded, just give up */ |
2217 | if (__packet_rcv_has_room(po, skb) == ROOM_NONE) { | |
2218 | atomic_inc(&po->tp_drops); | |
2219 | goto drop_n_restore; | |
2220 | } | |
2221 | ||
68c2e5de AD |
2222 | if (skb->ip_summed == CHECKSUM_PARTIAL) |
2223 | status |= TP_STATUS_CSUMNOTREADY; | |
682f048b AD |
2224 | else if (skb->pkt_type != PACKET_OUTGOING && |
2225 | (skb->ip_summed == CHECKSUM_COMPLETE || | |
2226 | skb_csum_unnecessary(skb))) | |
2227 | status |= TP_STATUS_CSUM_VALID; | |
68c2e5de | 2228 | |
dbcb5855 DM |
2229 | if (snaplen > res) |
2230 | snaplen = res; | |
1da177e4 LT |
2231 | |
2232 | if (sk->sk_type == SOCK_DGRAM) { | |
8913336a PM |
2233 | macoff = netoff = TPACKET_ALIGN(po->tp_hdrlen) + 16 + |
2234 | po->tp_reserve; | |
1da177e4 | 2235 | } else { |
95c96174 | 2236 | unsigned int maclen = skb_network_offset(skb); |
bbd6ef87 | 2237 | netoff = TPACKET_ALIGN(po->tp_hdrlen + |
8913336a | 2238 | (maclen < 16 ? 16 : maclen)) + |
58d19b19 | 2239 | po->tp_reserve; |
edbd58be | 2240 | if (po->has_vnet_hdr) { |
58d19b19 | 2241 | netoff += sizeof(struct virtio_net_hdr); |
edbd58be BP |
2242 | do_vnet = true; |
2243 | } | |
1da177e4 LT |
2244 | macoff = netoff - maclen; |
2245 | } | |
acf69c94 OC |
2246 | if (netoff > USHRT_MAX) { |
2247 | atomic_inc(&po->tp_drops); | |
2248 | goto drop_n_restore; | |
2249 | } | |
f6fb8f10 | 2250 | if (po->tp_version <= TPACKET_V2) { |
2251 | if (macoff + snaplen > po->rx_ring.frame_size) { | |
2252 | if (po->copy_thresh && | |
0fd7bac6 | 2253 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) { |
f6fb8f10 | 2254 | if (skb_shared(skb)) { |
2255 | copy_skb = skb_clone(skb, GFP_ATOMIC); | |
2256 | } else { | |
2257 | copy_skb = skb_get(skb); | |
2258 | skb_head = skb->data; | |
2259 | } | |
2260 | if (copy_skb) | |
2261 | skb_set_owner_r(copy_skb, sk); | |
1da177e4 | 2262 | } |
f6fb8f10 | 2263 | snaplen = po->rx_ring.frame_size - macoff; |
edbd58be | 2264 | if ((int)snaplen < 0) { |
f6fb8f10 | 2265 | snaplen = 0; |
edbd58be BP |
2266 | do_vnet = false; |
2267 | } | |
1da177e4 | 2268 | } |
dc808110 ED |
2269 | } else if (unlikely(macoff + snaplen > |
2270 | GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len)) { | |
2271 | u32 nval; | |
2272 | ||
2273 | nval = GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len - macoff; | |
2274 | pr_err_once("tpacket_rcv: packet too big, clamped from %u to %u. macoff=%u\n", | |
2275 | snaplen, nval, macoff); | |
2276 | snaplen = nval; | |
2277 | if (unlikely((int)snaplen < 0)) { | |
2278 | snaplen = 0; | |
2279 | macoff = GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len; | |
edbd58be | 2280 | do_vnet = false; |
dc808110 | 2281 | } |
1da177e4 | 2282 | } |
1da177e4 | 2283 | spin_lock(&sk->sk_receive_queue.lock); |
f6fb8f10 | 2284 | h.raw = packet_current_rx_frame(po, skb, |
2285 | TP_STATUS_KERNEL, (macoff+snaplen)); | |
bbd6ef87 | 2286 | if (!h.raw) |
58d19b19 | 2287 | goto drop_n_account; |
46e4c421 | 2288 | |
61fad681 WB |
2289 | if (po->tp_version <= TPACKET_V2) { |
2290 | slot_id = po->rx_ring.head; | |
2291 | if (test_bit(slot_id, po->rx_ring.rx_owner_map)) | |
2292 | goto drop_n_account; | |
2293 | __set_bit(slot_id, po->rx_ring.rx_owner_map); | |
2294 | } | |
2295 | ||
46e4c421 WB |
2296 | if (do_vnet && |
2297 | virtio_net_hdr_from_skb(skb, h.raw + macoff - | |
2298 | sizeof(struct virtio_net_hdr), | |
88fd1cb8 JO |
2299 | vio_le(), true, 0)) { |
2300 | if (po->tp_version == TPACKET_V3) | |
2301 | prb_clear_blk_fill_status(&po->rx_ring); | |
46e4c421 | 2302 | goto drop_n_account; |
88fd1cb8 | 2303 | } |
46e4c421 | 2304 | |
f6fb8f10 | 2305 | if (po->tp_version <= TPACKET_V2) { |
2306 | packet_increment_rx_head(po, &po->rx_ring); | |
2307 | /* | |
2308 | * LOSING will be reported till you read the stats, | |
2309 | * because it's COR - Clear On Read. | |
2310 | * Anyways, moving it for V1/V2 only as V3 doesn't need this | |
2311 | * at packet level. | |
2312 | */ | |
8e8e2951 | 2313 | if (atomic_read(&po->tp_drops)) |
f6fb8f10 | 2314 | status |= TP_STATUS_LOSING; |
2315 | } | |
945d015e | 2316 | |
ee80fbf3 | 2317 | po->stats.stats1.tp_packets++; |
1da177e4 LT |
2318 | if (copy_skb) { |
2319 | status |= TP_STATUS_COPY; | |
2320 | __skb_queue_tail(&sk->sk_receive_queue, copy_skb); | |
2321 | } | |
1da177e4 LT |
2322 | spin_unlock(&sk->sk_receive_queue.lock); |
2323 | ||
bbd6ef87 | 2324 | skb_copy_bits(skb, 0, h.raw + macoff, snaplen); |
b9c32fb2 DB |
2325 | |
2326 | if (!(ts_status = tpacket_get_timestamp(skb, &ts, po->tp_tstamp))) | |
d413fcb4 | 2327 | ktime_get_real_ts64(&ts); |
1da177e4 | 2328 | |
b9c32fb2 DB |
2329 | status |= ts_status; |
2330 | ||
bbd6ef87 PM |
2331 | switch (po->tp_version) { |
2332 | case TPACKET_V1: | |
2333 | h.h1->tp_len = skb->len; | |
2334 | h.h1->tp_snaplen = snaplen; | |
2335 | h.h1->tp_mac = macoff; | |
2336 | h.h1->tp_net = netoff; | |
4b457bdf DB |
2337 | h.h1->tp_sec = ts.tv_sec; |
2338 | h.h1->tp_usec = ts.tv_nsec / NSEC_PER_USEC; | |
bbd6ef87 PM |
2339 | hdrlen = sizeof(*h.h1); |
2340 | break; | |
2341 | case TPACKET_V2: | |
2342 | h.h2->tp_len = skb->len; | |
2343 | h.h2->tp_snaplen = snaplen; | |
2344 | h.h2->tp_mac = macoff; | |
2345 | h.h2->tp_net = netoff; | |
bbd6ef87 PM |
2346 | h.h2->tp_sec = ts.tv_sec; |
2347 | h.h2->tp_nsec = ts.tv_nsec; | |
df8a39de JP |
2348 | if (skb_vlan_tag_present(skb)) { |
2349 | h.h2->tp_vlan_tci = skb_vlan_tag_get(skb); | |
a0cdfcf3 AW |
2350 | h.h2->tp_vlan_tpid = ntohs(skb->vlan_proto); |
2351 | status |= TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
a3bcc23e BG |
2352 | } else { |
2353 | h.h2->tp_vlan_tci = 0; | |
a0cdfcf3 | 2354 | h.h2->tp_vlan_tpid = 0; |
a3bcc23e | 2355 | } |
e4d26f4b | 2356 | memset(h.h2->tp_padding, 0, sizeof(h.h2->tp_padding)); |
bbd6ef87 PM |
2357 | hdrlen = sizeof(*h.h2); |
2358 | break; | |
f6fb8f10 | 2359 | case TPACKET_V3: |
2360 | /* tp_nxt_offset,vlan are already populated above. | |
2361 | * So DONT clear those fields here | |
2362 | */ | |
2363 | h.h3->tp_status |= status; | |
2364 | h.h3->tp_len = skb->len; | |
2365 | h.h3->tp_snaplen = snaplen; | |
2366 | h.h3->tp_mac = macoff; | |
2367 | h.h3->tp_net = netoff; | |
f6fb8f10 | 2368 | h.h3->tp_sec = ts.tv_sec; |
2369 | h.h3->tp_nsec = ts.tv_nsec; | |
e4d26f4b | 2370 | memset(h.h3->tp_padding, 0, sizeof(h.h3->tp_padding)); |
f6fb8f10 | 2371 | hdrlen = sizeof(*h.h3); |
2372 | break; | |
bbd6ef87 PM |
2373 | default: |
2374 | BUG(); | |
2375 | } | |
1da177e4 | 2376 | |
bbd6ef87 | 2377 | sll = h.raw + TPACKET_ALIGN(hdrlen); |
b95cce35 | 2378 | sll->sll_halen = dev_parse_header(skb, sll->sll_addr); |
1da177e4 LT |
2379 | sll->sll_family = AF_PACKET; |
2380 | sll->sll_hatype = dev->type; | |
2381 | sll->sll_protocol = skb->protocol; | |
2382 | sll->sll_pkttype = skb->pkt_type; | |
8032b464 | 2383 | if (unlikely(po->origdev)) |
80feaacb PWJ |
2384 | sll->sll_ifindex = orig_dev->ifindex; |
2385 | else | |
2386 | sll->sll_ifindex = dev->ifindex; | |
1da177e4 | 2387 | |
e16aa207 | 2388 | smp_mb(); |
f0d4eb29 | 2389 | |
f6dafa95 | 2390 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 |
f0d4eb29 | 2391 | if (po->tp_version <= TPACKET_V2) { |
0af55bb5 CG |
2392 | u8 *start, *end; |
2393 | ||
f0d4eb29 DB |
2394 | end = (u8 *) PAGE_ALIGN((unsigned long) h.raw + |
2395 | macoff + snaplen); | |
2396 | ||
2397 | for (start = h.raw; start < end; start += PAGE_SIZE) | |
2398 | flush_dcache_page(pgv_to_page(start)); | |
1da177e4 | 2399 | } |
f0d4eb29 | 2400 | smp_wmb(); |
f6dafa95 | 2401 | #endif |
f0d4eb29 | 2402 | |
da413eec | 2403 | if (po->tp_version <= TPACKET_V2) { |
61fad681 | 2404 | spin_lock(&sk->sk_receive_queue.lock); |
f6fb8f10 | 2405 | __packet_set_status(po, h.raw, status); |
61fad681 WB |
2406 | __clear_bit(slot_id, po->rx_ring.rx_owner_map); |
2407 | spin_unlock(&sk->sk_receive_queue.lock); | |
da413eec | 2408 | sk->sk_data_ready(sk); |
88fd1cb8 | 2409 | } else if (po->tp_version == TPACKET_V3) { |
f6fb8f10 | 2410 | prb_clear_blk_fill_status(&po->rx_ring); |
da413eec | 2411 | } |
1da177e4 LT |
2412 | |
2413 | drop_n_restore: | |
2414 | if (skb_head != skb->data && skb_shared(skb)) { | |
2415 | skb->data = skb_head; | |
2416 | skb->len = skb_len; | |
2417 | } | |
2418 | drop: | |
da37845f WJ |
2419 | if (!is_drop_n_account) |
2420 | consume_skb(skb); | |
2421 | else | |
2422 | kfree_skb(skb); | |
1da177e4 LT |
2423 | return 0; |
2424 | ||
58d19b19 | 2425 | drop_n_account: |
1da177e4 | 2426 | spin_unlock(&sk->sk_receive_queue.lock); |
8e8e2951 ED |
2427 | atomic_inc(&po->tp_drops); |
2428 | is_drop_n_account = true; | |
1da177e4 | 2429 | |
676d2369 | 2430 | sk->sk_data_ready(sk); |
acb5d75b | 2431 | kfree_skb(copy_skb); |
1da177e4 LT |
2432 | goto drop_n_restore; |
2433 | } | |
2434 | ||
69e3c75f JB |
2435 | static void tpacket_destruct_skb(struct sk_buff *skb) |
2436 | { | |
2437 | struct packet_sock *po = pkt_sk(skb->sk); | |
1da177e4 | 2438 | |
69e3c75f | 2439 | if (likely(po->tx_ring.pg_vec)) { |
f0d4eb29 | 2440 | void *ph; |
b9c32fb2 DB |
2441 | __u32 ts; |
2442 | ||
5cd8d46e | 2443 | ph = skb_zcopy_get_nouarg(skb); |
b0138408 | 2444 | packet_dec_pending(&po->tx_ring); |
b9c32fb2 DB |
2445 | |
2446 | ts = __packet_set_timestamp(po, ph, skb); | |
2447 | __packet_set_status(po, ph, TP_STATUS_AVAILABLE | ts); | |
89ed5b51 NH |
2448 | |
2449 | if (!packet_read_pending(&po->tx_ring)) | |
2450 | complete(&po->skb_completion); | |
69e3c75f JB |
2451 | } |
2452 | ||
2453 | sock_wfree(skb); | |
2454 | } | |
2455 | ||
16cc1400 WB |
2456 | static int __packet_snd_vnet_parse(struct virtio_net_hdr *vnet_hdr, size_t len) |
2457 | { | |
16cc1400 WB |
2458 | if ((vnet_hdr->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
2459 | (__virtio16_to_cpu(vio_le(), vnet_hdr->csum_start) + | |
2460 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_offset) + 2 > | |
2461 | __virtio16_to_cpu(vio_le(), vnet_hdr->hdr_len))) | |
2462 | vnet_hdr->hdr_len = __cpu_to_virtio16(vio_le(), | |
2463 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_start) + | |
2464 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_offset) + 2); | |
2465 | ||
2466 | if (__virtio16_to_cpu(vio_le(), vnet_hdr->hdr_len) > len) | |
2467 | return -EINVAL; | |
2468 | ||
16cc1400 WB |
2469 | return 0; |
2470 | } | |
2471 | ||
2472 | static int packet_snd_vnet_parse(struct msghdr *msg, size_t *len, | |
2473 | struct virtio_net_hdr *vnet_hdr) | |
2474 | { | |
16cc1400 WB |
2475 | if (*len < sizeof(*vnet_hdr)) |
2476 | return -EINVAL; | |
2477 | *len -= sizeof(*vnet_hdr); | |
2478 | ||
cbbd26b8 | 2479 | if (!copy_from_iter_full(vnet_hdr, sizeof(*vnet_hdr), &msg->msg_iter)) |
16cc1400 WB |
2480 | return -EFAULT; |
2481 | ||
2482 | return __packet_snd_vnet_parse(vnet_hdr, *len); | |
2483 | } | |
2484 | ||
40d4e3df | 2485 | static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, |
8d39b4a6 | 2486 | void *frame, struct net_device *dev, void *data, int tp_len, |
c14ac945 SHY |
2487 | __be16 proto, unsigned char *addr, int hlen, int copylen, |
2488 | const struct sockcm_cookie *sockc) | |
69e3c75f | 2489 | { |
184f489e | 2490 | union tpacket_uhdr ph; |
8d39b4a6 | 2491 | int to_write, offset, len, nr_frags, len_max; |
69e3c75f JB |
2492 | struct socket *sock = po->sk.sk_socket; |
2493 | struct page *page; | |
69e3c75f JB |
2494 | int err; |
2495 | ||
2496 | ph.raw = frame; | |
2497 | ||
2498 | skb->protocol = proto; | |
2499 | skb->dev = dev; | |
2500 | skb->priority = po->sk.sk_priority; | |
2d37a186 | 2501 | skb->mark = po->sk.sk_mark; |
3d0ba8c0 | 2502 | skb->tstamp = sockc->transmit_time; |
8f932f76 | 2503 | skb_setup_tx_timestamp(skb, sockc->tsflags); |
5cd8d46e | 2504 | skb_zcopy_set_nouarg(skb, ph.raw); |
69e3c75f | 2505 | |
ae641949 | 2506 | skb_reserve(skb, hlen); |
69e3c75f | 2507 | skb_reset_network_header(skb); |
c1aad275 | 2508 | |
69e3c75f JB |
2509 | to_write = tp_len; |
2510 | ||
2511 | if (sock->type == SOCK_DGRAM) { | |
2512 | err = dev_hard_header(skb, dev, ntohs(proto), addr, | |
2513 | NULL, tp_len); | |
2514 | if (unlikely(err < 0)) | |
2515 | return -EINVAL; | |
1d036d25 | 2516 | } else if (copylen) { |
9ed988cd WB |
2517 | int hdrlen = min_t(int, copylen, tp_len); |
2518 | ||
69e3c75f | 2519 | skb_push(skb, dev->hard_header_len); |
1d036d25 | 2520 | skb_put(skb, copylen - dev->hard_header_len); |
9ed988cd | 2521 | err = skb_store_bits(skb, 0, data, hdrlen); |
69e3c75f JB |
2522 | if (unlikely(err)) |
2523 | return err; | |
9ed988cd WB |
2524 | if (!dev_validate_header(dev, skb->data, hdrlen)) |
2525 | return -EINVAL; | |
69e3c75f | 2526 | |
9ed988cd WB |
2527 | data += hdrlen; |
2528 | to_write -= hdrlen; | |
69e3c75f JB |
2529 | } |
2530 | ||
69e3c75f JB |
2531 | offset = offset_in_page(data); |
2532 | len_max = PAGE_SIZE - offset; | |
2533 | len = ((to_write > len_max) ? len_max : to_write); | |
2534 | ||
2535 | skb->data_len = to_write; | |
2536 | skb->len += to_write; | |
2537 | skb->truesize += to_write; | |
14afee4b | 2538 | refcount_add(to_write, &po->sk.sk_wmem_alloc); |
69e3c75f JB |
2539 | |
2540 | while (likely(to_write)) { | |
2541 | nr_frags = skb_shinfo(skb)->nr_frags; | |
2542 | ||
2543 | if (unlikely(nr_frags >= MAX_SKB_FRAGS)) { | |
40d4e3df ED |
2544 | pr_err("Packet exceed the number of skb frags(%lu)\n", |
2545 | MAX_SKB_FRAGS); | |
69e3c75f JB |
2546 | return -EFAULT; |
2547 | } | |
2548 | ||
0af55bb5 CG |
2549 | page = pgv_to_page(data); |
2550 | data += len; | |
69e3c75f JB |
2551 | flush_dcache_page(page); |
2552 | get_page(page); | |
0af55bb5 | 2553 | skb_fill_page_desc(skb, nr_frags, page, offset, len); |
69e3c75f JB |
2554 | to_write -= len; |
2555 | offset = 0; | |
2556 | len_max = PAGE_SIZE; | |
2557 | len = ((to_write > len_max) ? len_max : to_write); | |
2558 | } | |
2559 | ||
75c65772 | 2560 | packet_parse_headers(skb, sock); |
efdfa2f7 | 2561 | |
69e3c75f JB |
2562 | return tp_len; |
2563 | } | |
2564 | ||
8d39b4a6 WB |
2565 | static int tpacket_parse_header(struct packet_sock *po, void *frame, |
2566 | int size_max, void **data) | |
2567 | { | |
2568 | union tpacket_uhdr ph; | |
2569 | int tp_len, off; | |
2570 | ||
2571 | ph.raw = frame; | |
2572 | ||
2573 | switch (po->tp_version) { | |
7f953ab2 SV |
2574 | case TPACKET_V3: |
2575 | if (ph.h3->tp_next_offset != 0) { | |
2576 | pr_warn_once("variable sized slot not supported"); | |
2577 | return -EINVAL; | |
2578 | } | |
2579 | tp_len = ph.h3->tp_len; | |
2580 | break; | |
8d39b4a6 WB |
2581 | case TPACKET_V2: |
2582 | tp_len = ph.h2->tp_len; | |
2583 | break; | |
2584 | default: | |
2585 | tp_len = ph.h1->tp_len; | |
2586 | break; | |
2587 | } | |
2588 | if (unlikely(tp_len > size_max)) { | |
2589 | pr_err("packet size is too long (%d > %d)\n", tp_len, size_max); | |
2590 | return -EMSGSIZE; | |
2591 | } | |
2592 | ||
2593 | if (unlikely(po->tp_tx_has_off)) { | |
2594 | int off_min, off_max; | |
2595 | ||
2596 | off_min = po->tp_hdrlen - sizeof(struct sockaddr_ll); | |
2597 | off_max = po->tx_ring.frame_size - tp_len; | |
2598 | if (po->sk.sk_type == SOCK_DGRAM) { | |
2599 | switch (po->tp_version) { | |
7f953ab2 SV |
2600 | case TPACKET_V3: |
2601 | off = ph.h3->tp_net; | |
2602 | break; | |
8d39b4a6 WB |
2603 | case TPACKET_V2: |
2604 | off = ph.h2->tp_net; | |
2605 | break; | |
2606 | default: | |
2607 | off = ph.h1->tp_net; | |
2608 | break; | |
2609 | } | |
2610 | } else { | |
2611 | switch (po->tp_version) { | |
7f953ab2 SV |
2612 | case TPACKET_V3: |
2613 | off = ph.h3->tp_mac; | |
2614 | break; | |
8d39b4a6 WB |
2615 | case TPACKET_V2: |
2616 | off = ph.h2->tp_mac; | |
2617 | break; | |
2618 | default: | |
2619 | off = ph.h1->tp_mac; | |
2620 | break; | |
2621 | } | |
2622 | } | |
2623 | if (unlikely((off < off_min) || (off_max < off))) | |
2624 | return -EINVAL; | |
2625 | } else { | |
2626 | off = po->tp_hdrlen - sizeof(struct sockaddr_ll); | |
2627 | } | |
2628 | ||
2629 | *data = frame + off; | |
2630 | return tp_len; | |
2631 | } | |
2632 | ||
69e3c75f JB |
2633 | static int tpacket_snd(struct packet_sock *po, struct msghdr *msg) |
2634 | { | |
89ed5b51 | 2635 | struct sk_buff *skb = NULL; |
69e3c75f | 2636 | struct net_device *dev; |
1d036d25 | 2637 | struct virtio_net_hdr *vnet_hdr = NULL; |
c14ac945 | 2638 | struct sockcm_cookie sockc; |
69e3c75f | 2639 | __be16 proto; |
09effa67 | 2640 | int err, reserve = 0; |
40d4e3df | 2641 | void *ph; |
342dfc30 | 2642 | DECLARE_SOCKADDR(struct sockaddr_ll *, saddr, msg->msg_name); |
87a2fd28 | 2643 | bool need_wait = !(msg->msg_flags & MSG_DONTWAIT); |
486efdc8 | 2644 | unsigned char *addr = NULL; |
69e3c75f | 2645 | int tp_len, size_max; |
8d39b4a6 | 2646 | void *data; |
69e3c75f | 2647 | int len_sum = 0; |
9e67030a | 2648 | int status = TP_STATUS_AVAILABLE; |
1d036d25 | 2649 | int hlen, tlen, copylen = 0; |
89ed5b51 | 2650 | long timeo = 0; |
69e3c75f | 2651 | |
69e3c75f JB |
2652 | mutex_lock(&po->pg_vec_lock); |
2653 | ||
32d3182c ED |
2654 | /* packet_sendmsg() check on tx_ring.pg_vec was lockless, |
2655 | * we need to confirm it under protection of pg_vec_lock. | |
2656 | */ | |
2657 | if (unlikely(!po->tx_ring.pg_vec)) { | |
2658 | err = -EBUSY; | |
2659 | goto out; | |
2660 | } | |
66e56cd4 | 2661 | if (likely(saddr == NULL)) { |
e40526cb | 2662 | dev = packet_cached_dev_get(po); |
69e3c75f | 2663 | proto = po->num; |
69e3c75f JB |
2664 | } else { |
2665 | err = -EINVAL; | |
2666 | if (msg->msg_namelen < sizeof(struct sockaddr_ll)) | |
2667 | goto out; | |
2668 | if (msg->msg_namelen < (saddr->sll_halen | |
2669 | + offsetof(struct sockaddr_ll, | |
2670 | sll_addr))) | |
2671 | goto out; | |
69e3c75f | 2672 | proto = saddr->sll_protocol; |
827d9780 | 2673 | dev = dev_get_by_index(sock_net(&po->sk), saddr->sll_ifindex); |
486efdc8 WB |
2674 | if (po->sk.sk_socket->type == SOCK_DGRAM) { |
2675 | if (dev && msg->msg_namelen < dev->addr_len + | |
2676 | offsetof(struct sockaddr_ll, sll_addr)) | |
2677 | goto out_put; | |
2678 | addr = saddr->sll_addr; | |
2679 | } | |
69e3c75f JB |
2680 | } |
2681 | ||
69e3c75f JB |
2682 | err = -ENXIO; |
2683 | if (unlikely(dev == NULL)) | |
2684 | goto out; | |
69e3c75f JB |
2685 | err = -ENETDOWN; |
2686 | if (unlikely(!(dev->flags & IFF_UP))) | |
2687 | goto out_put; | |
2688 | ||
657a0667 | 2689 | sockcm_init(&sockc, &po->sk); |
d19b183c DCS |
2690 | if (msg->msg_controllen) { |
2691 | err = sock_cmsg_send(&po->sk, msg, &sockc); | |
2692 | if (unlikely(err)) | |
2693 | goto out_put; | |
2694 | } | |
2695 | ||
5cfb4c8d DB |
2696 | if (po->sk.sk_socket->type == SOCK_RAW) |
2697 | reserve = dev->hard_header_len; | |
69e3c75f | 2698 | size_max = po->tx_ring.frame_size |
b5dd884e | 2699 | - (po->tp_hdrlen - sizeof(struct sockaddr_ll)); |
69e3c75f | 2700 | |
1d036d25 | 2701 | if ((size_max > dev->mtu + reserve + VLAN_HLEN) && !po->has_vnet_hdr) |
5cfb4c8d | 2702 | size_max = dev->mtu + reserve + VLAN_HLEN; |
09effa67 | 2703 | |
89ed5b51 NH |
2704 | reinit_completion(&po->skb_completion); |
2705 | ||
69e3c75f JB |
2706 | do { |
2707 | ph = packet_current_frame(po, &po->tx_ring, | |
87a2fd28 | 2708 | TP_STATUS_SEND_REQUEST); |
69e3c75f | 2709 | if (unlikely(ph == NULL)) { |
89ed5b51 NH |
2710 | if (need_wait && skb) { |
2711 | timeo = sock_sndtimeo(&po->sk, msg->msg_flags & MSG_DONTWAIT); | |
2712 | timeo = wait_for_completion_interruptible_timeout(&po->skb_completion, timeo); | |
2713 | if (timeo <= 0) { | |
2714 | err = !timeo ? -ETIMEDOUT : -ERESTARTSYS; | |
2715 | goto out_put; | |
2716 | } | |
2717 | } | |
2718 | /* check for additional frames */ | |
69e3c75f JB |
2719 | continue; |
2720 | } | |
2721 | ||
8d39b4a6 WB |
2722 | skb = NULL; |
2723 | tp_len = tpacket_parse_header(po, ph, size_max, &data); | |
2724 | if (tp_len < 0) | |
2725 | goto tpacket_error; | |
2726 | ||
69e3c75f | 2727 | status = TP_STATUS_SEND_REQUEST; |
ae641949 HX |
2728 | hlen = LL_RESERVED_SPACE(dev); |
2729 | tlen = dev->needed_tailroom; | |
1d036d25 WB |
2730 | if (po->has_vnet_hdr) { |
2731 | vnet_hdr = data; | |
2732 | data += sizeof(*vnet_hdr); | |
2733 | tp_len -= sizeof(*vnet_hdr); | |
2734 | if (tp_len < 0 || | |
2735 | __packet_snd_vnet_parse(vnet_hdr, tp_len)) { | |
2736 | tp_len = -EINVAL; | |
2737 | goto tpacket_error; | |
2738 | } | |
2739 | copylen = __virtio16_to_cpu(vio_le(), | |
2740 | vnet_hdr->hdr_len); | |
2741 | } | |
9ed988cd | 2742 | copylen = max_t(int, copylen, dev->hard_header_len); |
69e3c75f | 2743 | skb = sock_alloc_send_skb(&po->sk, |
1d036d25 WB |
2744 | hlen + tlen + sizeof(struct sockaddr_ll) + |
2745 | (copylen - dev->hard_header_len), | |
fbf33a28 | 2746 | !need_wait, &err); |
69e3c75f | 2747 | |
fbf33a28 KM |
2748 | if (unlikely(skb == NULL)) { |
2749 | /* we assume the socket was initially writeable ... */ | |
2750 | if (likely(len_sum > 0)) | |
2751 | err = len_sum; | |
69e3c75f | 2752 | goto out_status; |
fbf33a28 | 2753 | } |
8d39b4a6 | 2754 | tp_len = tpacket_fill_skb(po, skb, ph, dev, data, tp_len, proto, |
c14ac945 | 2755 | addr, hlen, copylen, &sockc); |
dbd46ab4 | 2756 | if (likely(tp_len >= 0) && |
5cfb4c8d | 2757 | tp_len > dev->mtu + reserve && |
1d036d25 | 2758 | !po->has_vnet_hdr && |
3c70c132 DB |
2759 | !packet_extra_vlan_len_allowed(dev, skb)) |
2760 | tp_len = -EMSGSIZE; | |
69e3c75f JB |
2761 | |
2762 | if (unlikely(tp_len < 0)) { | |
8d39b4a6 | 2763 | tpacket_error: |
69e3c75f JB |
2764 | if (po->tp_loss) { |
2765 | __packet_set_status(po, ph, | |
2766 | TP_STATUS_AVAILABLE); | |
2767 | packet_increment_head(&po->tx_ring); | |
2768 | kfree_skb(skb); | |
2769 | continue; | |
2770 | } else { | |
2771 | status = TP_STATUS_WRONG_FORMAT; | |
2772 | err = tp_len; | |
2773 | goto out_status; | |
2774 | } | |
2775 | } | |
2776 | ||
9d2f67e4 JT |
2777 | if (po->has_vnet_hdr) { |
2778 | if (virtio_net_hdr_to_skb(skb, vnet_hdr, vio_le())) { | |
2779 | tp_len = -EINVAL; | |
2780 | goto tpacket_error; | |
2781 | } | |
2782 | virtio_net_hdr_set_proto(skb, vnet_hdr); | |
1d036d25 WB |
2783 | } |
2784 | ||
69e3c75f JB |
2785 | skb->destructor = tpacket_destruct_skb; |
2786 | __packet_set_status(po, ph, TP_STATUS_SENDING); | |
b0138408 | 2787 | packet_inc_pending(&po->tx_ring); |
69e3c75f JB |
2788 | |
2789 | status = TP_STATUS_SEND_REQUEST; | |
d346a3fa | 2790 | err = po->xmit(skb); |
eb70df13 JP |
2791 | if (unlikely(err > 0)) { |
2792 | err = net_xmit_errno(err); | |
2793 | if (err && __packet_get_status(po, ph) == | |
2794 | TP_STATUS_AVAILABLE) { | |
2795 | /* skb was destructed already */ | |
2796 | skb = NULL; | |
2797 | goto out_status; | |
2798 | } | |
2799 | /* | |
2800 | * skb was dropped but not destructed yet; | |
2801 | * let's treat it like congestion or err < 0 | |
2802 | */ | |
2803 | err = 0; | |
2804 | } | |
69e3c75f JB |
2805 | packet_increment_head(&po->tx_ring); |
2806 | len_sum += tp_len; | |
b0138408 DB |
2807 | } while (likely((ph != NULL) || |
2808 | /* Note: packet_read_pending() might be slow if we have | |
2809 | * to call it as it's per_cpu variable, but in fast-path | |
2810 | * we already short-circuit the loop with the first | |
2811 | * condition, and luckily don't have to go that path | |
2812 | * anyway. | |
2813 | */ | |
2814 | (need_wait && packet_read_pending(&po->tx_ring)))); | |
69e3c75f JB |
2815 | |
2816 | err = len_sum; | |
2817 | goto out_put; | |
2818 | ||
69e3c75f JB |
2819 | out_status: |
2820 | __packet_set_status(po, ph, status); | |
2821 | kfree_skb(skb); | |
2822 | out_put: | |
e40526cb | 2823 | dev_put(dev); |
69e3c75f JB |
2824 | out: |
2825 | mutex_unlock(&po->pg_vec_lock); | |
2826 | return err; | |
2827 | } | |
69e3c75f | 2828 | |
eea49cc9 OJ |
2829 | static struct sk_buff *packet_alloc_skb(struct sock *sk, size_t prepad, |
2830 | size_t reserve, size_t len, | |
2831 | size_t linear, int noblock, | |
2832 | int *err) | |
bfd5f4a3 SS |
2833 | { |
2834 | struct sk_buff *skb; | |
2835 | ||
2836 | /* Under a page? Don't bother with paged skb. */ | |
2837 | if (prepad + len < PAGE_SIZE || !linear) | |
2838 | linear = len; | |
2839 | ||
2840 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, | |
28d64271 | 2841 | err, 0); |
bfd5f4a3 SS |
2842 | if (!skb) |
2843 | return NULL; | |
2844 | ||
2845 | skb_reserve(skb, reserve); | |
2846 | skb_put(skb, linear); | |
2847 | skb->data_len = len - linear; | |
2848 | skb->len += len - linear; | |
2849 | ||
2850 | return skb; | |
2851 | } | |
2852 | ||
d346a3fa | 2853 | static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 LT |
2854 | { |
2855 | struct sock *sk = sock->sk; | |
342dfc30 | 2856 | DECLARE_SOCKADDR(struct sockaddr_ll *, saddr, msg->msg_name); |
1da177e4 LT |
2857 | struct sk_buff *skb; |
2858 | struct net_device *dev; | |
0e11c91e | 2859 | __be16 proto; |
486efdc8 | 2860 | unsigned char *addr = NULL; |
827d9780 | 2861 | int err, reserve = 0; |
c7d39e32 | 2862 | struct sockcm_cookie sockc; |
bfd5f4a3 SS |
2863 | struct virtio_net_hdr vnet_hdr = { 0 }; |
2864 | int offset = 0; | |
bfd5f4a3 | 2865 | struct packet_sock *po = pkt_sk(sk); |
da7c9561 | 2866 | bool has_vnet_hdr = false; |
57031eb7 | 2867 | int hlen, tlen, linear; |
3bdc0eba | 2868 | int extra_len = 0; |
1da177e4 LT |
2869 | |
2870 | /* | |
1ce4f28b | 2871 | * Get and verify the address. |
1da177e4 | 2872 | */ |
1ce4f28b | 2873 | |
66e56cd4 | 2874 | if (likely(saddr == NULL)) { |
e40526cb | 2875 | dev = packet_cached_dev_get(po); |
1da177e4 | 2876 | proto = po->num; |
1da177e4 LT |
2877 | } else { |
2878 | err = -EINVAL; | |
2879 | if (msg->msg_namelen < sizeof(struct sockaddr_ll)) | |
2880 | goto out; | |
0fb375fb EB |
2881 | if (msg->msg_namelen < (saddr->sll_halen + offsetof(struct sockaddr_ll, sll_addr))) |
2882 | goto out; | |
1da177e4 | 2883 | proto = saddr->sll_protocol; |
827d9780 | 2884 | dev = dev_get_by_index(sock_net(sk), saddr->sll_ifindex); |
486efdc8 WB |
2885 | if (sock->type == SOCK_DGRAM) { |
2886 | if (dev && msg->msg_namelen < dev->addr_len + | |
2887 | offsetof(struct sockaddr_ll, sll_addr)) | |
2888 | goto out_unlock; | |
2889 | addr = saddr->sll_addr; | |
2890 | } | |
1da177e4 LT |
2891 | } |
2892 | ||
1da177e4 | 2893 | err = -ENXIO; |
e40526cb | 2894 | if (unlikely(dev == NULL)) |
1da177e4 | 2895 | goto out_unlock; |
d5e76b0a | 2896 | err = -ENETDOWN; |
e40526cb | 2897 | if (unlikely(!(dev->flags & IFF_UP))) |
d5e76b0a DM |
2898 | goto out_unlock; |
2899 | ||
657a0667 | 2900 | sockcm_init(&sockc, sk); |
c7d39e32 EJ |
2901 | sockc.mark = sk->sk_mark; |
2902 | if (msg->msg_controllen) { | |
2903 | err = sock_cmsg_send(sk, msg, &sockc); | |
2904 | if (unlikely(err)) | |
2905 | goto out_unlock; | |
2906 | } | |
2907 | ||
e40526cb DB |
2908 | if (sock->type == SOCK_RAW) |
2909 | reserve = dev->hard_header_len; | |
bfd5f4a3 | 2910 | if (po->has_vnet_hdr) { |
16cc1400 WB |
2911 | err = packet_snd_vnet_parse(msg, &len, &vnet_hdr); |
2912 | if (err) | |
bfd5f4a3 | 2913 | goto out_unlock; |
da7c9561 | 2914 | has_vnet_hdr = true; |
bfd5f4a3 SS |
2915 | } |
2916 | ||
3bdc0eba BG |
2917 | if (unlikely(sock_flag(sk, SOCK_NOFCS))) { |
2918 | if (!netif_supports_nofcs(dev)) { | |
2919 | err = -EPROTONOSUPPORT; | |
2920 | goto out_unlock; | |
2921 | } | |
2922 | extra_len = 4; /* We're doing our own CRC */ | |
2923 | } | |
2924 | ||
1da177e4 | 2925 | err = -EMSGSIZE; |
16cc1400 WB |
2926 | if (!vnet_hdr.gso_type && |
2927 | (len > dev->mtu + reserve + VLAN_HLEN + extra_len)) | |
1da177e4 LT |
2928 | goto out_unlock; |
2929 | ||
bfd5f4a3 | 2930 | err = -ENOBUFS; |
ae641949 HX |
2931 | hlen = LL_RESERVED_SPACE(dev); |
2932 | tlen = dev->needed_tailroom; | |
57031eb7 WB |
2933 | linear = __virtio16_to_cpu(vio_le(), vnet_hdr.hdr_len); |
2934 | linear = max(linear, min_t(int, len, dev->hard_header_len)); | |
2935 | skb = packet_alloc_skb(sk, hlen + tlen, hlen, len, linear, | |
bfd5f4a3 | 2936 | msg->msg_flags & MSG_DONTWAIT, &err); |
40d4e3df | 2937 | if (skb == NULL) |
1da177e4 LT |
2938 | goto out_unlock; |
2939 | ||
b84bbaf7 | 2940 | skb_reset_network_header(skb); |
1da177e4 | 2941 | |
0c4e8581 | 2942 | err = -EINVAL; |
9c707762 WB |
2943 | if (sock->type == SOCK_DGRAM) { |
2944 | offset = dev_hard_header(skb, dev, ntohs(proto), addr, NULL, len); | |
46d2cfb1 | 2945 | if (unlikely(offset < 0)) |
9c707762 | 2946 | goto out_free; |
b84bbaf7 | 2947 | } else if (reserve) { |
9aad13b0 | 2948 | skb_reserve(skb, -reserve); |
88a8121d ND |
2949 | if (len < reserve + sizeof(struct ipv6hdr) && |
2950 | dev->min_header_len != dev->hard_header_len) | |
993675a3 | 2951 | skb_reset_network_header(skb); |
9c707762 | 2952 | } |
1da177e4 LT |
2953 | |
2954 | /* Returns -EFAULT on error */ | |
c0371da6 | 2955 | err = skb_copy_datagram_from_iter(skb, offset, &msg->msg_iter, len); |
1da177e4 LT |
2956 | if (err) |
2957 | goto out_free; | |
bf84a010 | 2958 | |
9ed988cd WB |
2959 | if (sock->type == SOCK_RAW && |
2960 | !dev_validate_header(dev, skb->data, len)) { | |
2961 | err = -EINVAL; | |
2962 | goto out_free; | |
2963 | } | |
2964 | ||
8f932f76 | 2965 | skb_setup_tx_timestamp(skb, sockc.tsflags); |
1da177e4 | 2966 | |
16cc1400 | 2967 | if (!vnet_hdr.gso_type && (len > dev->mtu + reserve + extra_len) && |
3c70c132 DB |
2968 | !packet_extra_vlan_len_allowed(dev, skb)) { |
2969 | err = -EMSGSIZE; | |
2970 | goto out_free; | |
57f89bfa BG |
2971 | } |
2972 | ||
09effa67 DM |
2973 | skb->protocol = proto; |
2974 | skb->dev = dev; | |
1da177e4 | 2975 | skb->priority = sk->sk_priority; |
c7d39e32 | 2976 | skb->mark = sockc.mark; |
3d0ba8c0 | 2977 | skb->tstamp = sockc.transmit_time; |
0fd5d57b | 2978 | |
da7c9561 | 2979 | if (has_vnet_hdr) { |
db60eb5f | 2980 | err = virtio_net_hdr_to_skb(skb, &vnet_hdr, vio_le()); |
16cc1400 WB |
2981 | if (err) |
2982 | goto out_free; | |
2983 | len += sizeof(vnet_hdr); | |
9d2f67e4 | 2984 | virtio_net_hdr_set_proto(skb, &vnet_hdr); |
bfd5f4a3 SS |
2985 | } |
2986 | ||
75c65772 | 2987 | packet_parse_headers(skb, sock); |
8fd6c80d | 2988 | |
3bdc0eba BG |
2989 | if (unlikely(extra_len == 4)) |
2990 | skb->no_fcs = 1; | |
2991 | ||
d346a3fa | 2992 | err = po->xmit(skb); |
1da177e4 LT |
2993 | if (err > 0 && (err = net_xmit_errno(err)) != 0) |
2994 | goto out_unlock; | |
2995 | ||
e40526cb | 2996 | dev_put(dev); |
1da177e4 | 2997 | |
40d4e3df | 2998 | return len; |
1da177e4 LT |
2999 | |
3000 | out_free: | |
3001 | kfree_skb(skb); | |
3002 | out_unlock: | |
e40526cb | 3003 | if (dev) |
1da177e4 LT |
3004 | dev_put(dev); |
3005 | out: | |
3006 | return err; | |
3007 | } | |
3008 | ||
1b784140 | 3009 | static int packet_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
69e3c75f | 3010 | { |
69e3c75f JB |
3011 | struct sock *sk = sock->sk; |
3012 | struct packet_sock *po = pkt_sk(sk); | |
d346a3fa | 3013 | |
69e3c75f JB |
3014 | if (po->tx_ring.pg_vec) |
3015 | return tpacket_snd(po, msg); | |
3016 | else | |
69e3c75f JB |
3017 | return packet_snd(sock, msg, len); |
3018 | } | |
3019 | ||
1da177e4 LT |
3020 | /* |
3021 | * Close a PACKET socket. This is fairly simple. We immediately go | |
3022 | * to 'closed' state and remove our protocol entry in the device list. | |
3023 | */ | |
3024 | ||
3025 | static int packet_release(struct socket *sock) | |
3026 | { | |
3027 | struct sock *sk = sock->sk; | |
3028 | struct packet_sock *po; | |
2bd624b4 | 3029 | struct packet_fanout *f; |
d12d01d6 | 3030 | struct net *net; |
f6fb8f10 | 3031 | union tpacket_req_u req_u; |
1da177e4 LT |
3032 | |
3033 | if (!sk) | |
3034 | return 0; | |
3035 | ||
3b1e0a65 | 3036 | net = sock_net(sk); |
1da177e4 LT |
3037 | po = pkt_sk(sk); |
3038 | ||
0fa7fa98 | 3039 | mutex_lock(&net->packet.sklist_lock); |
808f5114 | 3040 | sk_del_node_init_rcu(sk); |
0fa7fa98 PE |
3041 | mutex_unlock(&net->packet.sklist_lock); |
3042 | ||
3043 | preempt_disable(); | |
920de804 | 3044 | sock_prot_inuse_add(net, sk->sk_prot, -1); |
0fa7fa98 | 3045 | preempt_enable(); |
1da177e4 | 3046 | |
808f5114 | 3047 | spin_lock(&po->bind_lock); |
ce06b03e | 3048 | unregister_prot_hook(sk, false); |
66e56cd4 DB |
3049 | packet_cached_dev_reset(po); |
3050 | ||
160ff18a BG |
3051 | if (po->prot_hook.dev) { |
3052 | dev_put(po->prot_hook.dev); | |
3053 | po->prot_hook.dev = NULL; | |
3054 | } | |
808f5114 | 3055 | spin_unlock(&po->bind_lock); |
1da177e4 | 3056 | |
1da177e4 | 3057 | packet_flush_mclist(sk); |
1da177e4 | 3058 | |
5171b37d | 3059 | lock_sock(sk); |
9665d5d6 PS |
3060 | if (po->rx_ring.pg_vec) { |
3061 | memset(&req_u, 0, sizeof(req_u)); | |
f6fb8f10 | 3062 | packet_set_ring(sk, &req_u, 1, 0); |
9665d5d6 | 3063 | } |
69e3c75f | 3064 | |
9665d5d6 PS |
3065 | if (po->tx_ring.pg_vec) { |
3066 | memset(&req_u, 0, sizeof(req_u)); | |
f6fb8f10 | 3067 | packet_set_ring(sk, &req_u, 1, 1); |
9665d5d6 | 3068 | } |
5171b37d | 3069 | release_sock(sk); |
1da177e4 | 3070 | |
2bd624b4 | 3071 | f = fanout_release(sk); |
dc99f600 | 3072 | |
808f5114 | 3073 | synchronize_net(); |
2bd624b4 | 3074 | |
afa0925c | 3075 | kfree(po->rollover); |
2bd624b4 AS |
3076 | if (f) { |
3077 | fanout_release_data(f); | |
3078 | kfree(f); | |
3079 | } | |
1da177e4 LT |
3080 | /* |
3081 | * Now the socket is dead. No more input will appear. | |
3082 | */ | |
1da177e4 LT |
3083 | sock_orphan(sk); |
3084 | sock->sk = NULL; | |
3085 | ||
3086 | /* Purge queues */ | |
3087 | ||
3088 | skb_queue_purge(&sk->sk_receive_queue); | |
b0138408 | 3089 | packet_free_pending(po); |
17ab56a2 | 3090 | sk_refcnt_debug_release(sk); |
1da177e4 LT |
3091 | |
3092 | sock_put(sk); | |
3093 | return 0; | |
3094 | } | |
3095 | ||
3096 | /* | |
3097 | * Attach a packet hook. | |
3098 | */ | |
3099 | ||
30f7ea1c FR |
3100 | static int packet_do_bind(struct sock *sk, const char *name, int ifindex, |
3101 | __be16 proto) | |
1da177e4 LT |
3102 | { |
3103 | struct packet_sock *po = pkt_sk(sk); | |
158cd4af | 3104 | struct net_device *dev_curr; |
902fefb8 DB |
3105 | __be16 proto_curr; |
3106 | bool need_rehook; | |
30f7ea1c FR |
3107 | struct net_device *dev = NULL; |
3108 | int ret = 0; | |
3109 | bool unlisted = false; | |
dc99f600 | 3110 | |
1da177e4 | 3111 | lock_sock(sk); |
1da177e4 | 3112 | spin_lock(&po->bind_lock); |
30f7ea1c FR |
3113 | rcu_read_lock(); |
3114 | ||
4971613c WB |
3115 | if (po->fanout) { |
3116 | ret = -EINVAL; | |
3117 | goto out_unlock; | |
3118 | } | |
3119 | ||
30f7ea1c FR |
3120 | if (name) { |
3121 | dev = dev_get_by_name_rcu(sock_net(sk), name); | |
3122 | if (!dev) { | |
3123 | ret = -ENODEV; | |
3124 | goto out_unlock; | |
3125 | } | |
3126 | } else if (ifindex) { | |
3127 | dev = dev_get_by_index_rcu(sock_net(sk), ifindex); | |
3128 | if (!dev) { | |
3129 | ret = -ENODEV; | |
3130 | goto out_unlock; | |
3131 | } | |
3132 | } | |
3133 | ||
3134 | if (dev) | |
3135 | dev_hold(dev); | |
66e56cd4 | 3136 | |
902fefb8 DB |
3137 | proto_curr = po->prot_hook.type; |
3138 | dev_curr = po->prot_hook.dev; | |
3139 | ||
3140 | need_rehook = proto_curr != proto || dev_curr != dev; | |
3141 | ||
3142 | if (need_rehook) { | |
30f7ea1c FR |
3143 | if (po->running) { |
3144 | rcu_read_unlock(); | |
15fe076e ED |
3145 | /* prevents packet_notifier() from calling |
3146 | * register_prot_hook() | |
3147 | */ | |
3148 | po->num = 0; | |
30f7ea1c FR |
3149 | __unregister_prot_hook(sk, true); |
3150 | rcu_read_lock(); | |
3151 | dev_curr = po->prot_hook.dev; | |
3152 | if (dev) | |
3153 | unlisted = !dev_get_by_index_rcu(sock_net(sk), | |
3154 | dev->ifindex); | |
3155 | } | |
1da177e4 | 3156 | |
15fe076e | 3157 | BUG_ON(po->running); |
902fefb8 DB |
3158 | po->num = proto; |
3159 | po->prot_hook.type = proto; | |
902fefb8 | 3160 | |
30f7ea1c FR |
3161 | if (unlikely(unlisted)) { |
3162 | dev_put(dev); | |
3163 | po->prot_hook.dev = NULL; | |
3164 | po->ifindex = -1; | |
3165 | packet_cached_dev_reset(po); | |
3166 | } else { | |
3167 | po->prot_hook.dev = dev; | |
3168 | po->ifindex = dev ? dev->ifindex : 0; | |
3169 | packet_cached_dev_assign(po, dev); | |
3170 | } | |
902fefb8 | 3171 | } |
158cd4af LW |
3172 | if (dev_curr) |
3173 | dev_put(dev_curr); | |
66e56cd4 | 3174 | |
902fefb8 | 3175 | if (proto == 0 || !need_rehook) |
1da177e4 LT |
3176 | goto out_unlock; |
3177 | ||
30f7ea1c | 3178 | if (!unlisted && (!dev || (dev->flags & IFF_UP))) { |
ce06b03e | 3179 | register_prot_hook(sk); |
be85d4ad UT |
3180 | } else { |
3181 | sk->sk_err = ENETDOWN; | |
3182 | if (!sock_flag(sk, SOCK_DEAD)) | |
3183 | sk->sk_error_report(sk); | |
1da177e4 LT |
3184 | } |
3185 | ||
3186 | out_unlock: | |
30f7ea1c | 3187 | rcu_read_unlock(); |
1da177e4 LT |
3188 | spin_unlock(&po->bind_lock); |
3189 | release_sock(sk); | |
30f7ea1c | 3190 | return ret; |
1da177e4 LT |
3191 | } |
3192 | ||
3193 | /* | |
3194 | * Bind a packet socket to a device | |
3195 | */ | |
3196 | ||
40d4e3df ED |
3197 | static int packet_bind_spkt(struct socket *sock, struct sockaddr *uaddr, |
3198 | int addr_len) | |
1da177e4 | 3199 | { |
40d4e3df | 3200 | struct sock *sk = sock->sk; |
540e2894 | 3201 | char name[sizeof(uaddr->sa_data) + 1]; |
1ce4f28b | 3202 | |
1da177e4 LT |
3203 | /* |
3204 | * Check legality | |
3205 | */ | |
1ce4f28b | 3206 | |
8ae55f04 | 3207 | if (addr_len != sizeof(struct sockaddr)) |
1da177e4 | 3208 | return -EINVAL; |
540e2894 AP |
3209 | /* uaddr->sa_data comes from the userspace, it's not guaranteed to be |
3210 | * zero-terminated. | |
3211 | */ | |
3212 | memcpy(name, uaddr->sa_data, sizeof(uaddr->sa_data)); | |
3213 | name[sizeof(uaddr->sa_data)] = 0; | |
1da177e4 | 3214 | |
30f7ea1c | 3215 | return packet_do_bind(sk, name, 0, pkt_sk(sk)->num); |
1da177e4 | 3216 | } |
1da177e4 LT |
3217 | |
3218 | static int packet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
3219 | { | |
40d4e3df ED |
3220 | struct sockaddr_ll *sll = (struct sockaddr_ll *)uaddr; |
3221 | struct sock *sk = sock->sk; | |
1da177e4 LT |
3222 | |
3223 | /* | |
3224 | * Check legality | |
3225 | */ | |
1ce4f28b | 3226 | |
1da177e4 LT |
3227 | if (addr_len < sizeof(struct sockaddr_ll)) |
3228 | return -EINVAL; | |
3229 | if (sll->sll_family != AF_PACKET) | |
3230 | return -EINVAL; | |
3231 | ||
30f7ea1c FR |
3232 | return packet_do_bind(sk, NULL, sll->sll_ifindex, |
3233 | sll->sll_protocol ? : pkt_sk(sk)->num); | |
1da177e4 LT |
3234 | } |
3235 | ||
3236 | static struct proto packet_proto = { | |
3237 | .name = "PACKET", | |
3238 | .owner = THIS_MODULE, | |
3239 | .obj_size = sizeof(struct packet_sock), | |
3240 | }; | |
3241 | ||
3242 | /* | |
1ce4f28b | 3243 | * Create a packet of type SOCK_PACKET. |
1da177e4 LT |
3244 | */ |
3245 | ||
3f378b68 EP |
3246 | static int packet_create(struct net *net, struct socket *sock, int protocol, |
3247 | int kern) | |
1da177e4 LT |
3248 | { |
3249 | struct sock *sk; | |
3250 | struct packet_sock *po; | |
0e11c91e | 3251 | __be16 proto = (__force __be16)protocol; /* weird, but documented */ |
1da177e4 LT |
3252 | int err; |
3253 | ||
df008c91 | 3254 | if (!ns_capable(net->user_ns, CAP_NET_RAW)) |
1da177e4 | 3255 | return -EPERM; |
be02097c DM |
3256 | if (sock->type != SOCK_DGRAM && sock->type != SOCK_RAW && |
3257 | sock->type != SOCK_PACKET) | |
1da177e4 LT |
3258 | return -ESOCKTNOSUPPORT; |
3259 | ||
3260 | sock->state = SS_UNCONNECTED; | |
3261 | ||
3262 | err = -ENOBUFS; | |
11aa9c28 | 3263 | sk = sk_alloc(net, PF_PACKET, GFP_KERNEL, &packet_proto, kern); |
1da177e4 LT |
3264 | if (sk == NULL) |
3265 | goto out; | |
3266 | ||
3267 | sock->ops = &packet_ops; | |
1da177e4 LT |
3268 | if (sock->type == SOCK_PACKET) |
3269 | sock->ops = &packet_ops_spkt; | |
be02097c | 3270 | |
1da177e4 LT |
3271 | sock_init_data(sock, sk); |
3272 | ||
3273 | po = pkt_sk(sk); | |
89ed5b51 | 3274 | init_completion(&po->skb_completion); |
1da177e4 | 3275 | sk->sk_family = PF_PACKET; |
0e11c91e | 3276 | po->num = proto; |
d346a3fa | 3277 | po->xmit = dev_queue_xmit; |
66e56cd4 | 3278 | |
b0138408 DB |
3279 | err = packet_alloc_pending(po); |
3280 | if (err) | |
3281 | goto out2; | |
3282 | ||
66e56cd4 | 3283 | packet_cached_dev_reset(po); |
1da177e4 LT |
3284 | |
3285 | sk->sk_destruct = packet_sock_destruct; | |
17ab56a2 | 3286 | sk_refcnt_debug_inc(sk); |
1da177e4 LT |
3287 | |
3288 | /* | |
3289 | * Attach a protocol block | |
3290 | */ | |
3291 | ||
3292 | spin_lock_init(&po->bind_lock); | |
905db440 | 3293 | mutex_init(&po->pg_vec_lock); |
0648ab70 | 3294 | po->rollover = NULL; |
1da177e4 | 3295 | po->prot_hook.func = packet_rcv; |
be02097c | 3296 | |
1da177e4 LT |
3297 | if (sock->type == SOCK_PACKET) |
3298 | po->prot_hook.func = packet_rcv_spkt; | |
be02097c | 3299 | |
1da177e4 LT |
3300 | po->prot_hook.af_packet_priv = sk; |
3301 | ||
0e11c91e AV |
3302 | if (proto) { |
3303 | po->prot_hook.type = proto; | |
a6361f0c | 3304 | __register_prot_hook(sk); |
1da177e4 LT |
3305 | } |
3306 | ||
0fa7fa98 | 3307 | mutex_lock(&net->packet.sklist_lock); |
a4dc6a49 | 3308 | sk_add_node_tail_rcu(sk, &net->packet.sklist); |
0fa7fa98 PE |
3309 | mutex_unlock(&net->packet.sklist_lock); |
3310 | ||
3311 | preempt_disable(); | |
3680453c | 3312 | sock_prot_inuse_add(net, &packet_proto, 1); |
0fa7fa98 | 3313 | preempt_enable(); |
808f5114 | 3314 | |
40d4e3df | 3315 | return 0; |
b0138408 DB |
3316 | out2: |
3317 | sk_free(sk); | |
1da177e4 LT |
3318 | out: |
3319 | return err; | |
3320 | } | |
3321 | ||
3322 | /* | |
3323 | * Pull a packet from our receive queue and hand it to the user. | |
3324 | * If necessary we block. | |
3325 | */ | |
3326 | ||
1b784140 YX |
3327 | static int packet_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
3328 | int flags) | |
1da177e4 LT |
3329 | { |
3330 | struct sock *sk = sock->sk; | |
3331 | struct sk_buff *skb; | |
3332 | int copied, err; | |
bfd5f4a3 | 3333 | int vnet_hdr_len = 0; |
2472d761 | 3334 | unsigned int origlen = 0; |
1da177e4 LT |
3335 | |
3336 | err = -EINVAL; | |
ed85b565 | 3337 | if (flags & ~(MSG_PEEK|MSG_DONTWAIT|MSG_TRUNC|MSG_CMSG_COMPAT|MSG_ERRQUEUE)) |
1da177e4 LT |
3338 | goto out; |
3339 | ||
3340 | #if 0 | |
3341 | /* What error should we return now? EUNATTACH? */ | |
3342 | if (pkt_sk(sk)->ifindex < 0) | |
3343 | return -ENODEV; | |
3344 | #endif | |
3345 | ||
ed85b565 | 3346 | if (flags & MSG_ERRQUEUE) { |
cb820f8e RC |
3347 | err = sock_recv_errqueue(sk, msg, len, |
3348 | SOL_PACKET, PACKET_TX_TIMESTAMP); | |
ed85b565 RC |
3349 | goto out; |
3350 | } | |
3351 | ||
1da177e4 LT |
3352 | /* |
3353 | * Call the generic datagram receiver. This handles all sorts | |
3354 | * of horrible races and re-entrancy so we can forget about it | |
3355 | * in the protocol layers. | |
3356 | * | |
3357 | * Now it will return ENETDOWN, if device have just gone down, | |
3358 | * but then it will block. | |
3359 | */ | |
3360 | ||
40d4e3df | 3361 | skb = skb_recv_datagram(sk, flags, flags & MSG_DONTWAIT, &err); |
1da177e4 LT |
3362 | |
3363 | /* | |
1ce4f28b | 3364 | * An error occurred so return it. Because skb_recv_datagram() |
1da177e4 LT |
3365 | * handles the blocking we don't see and worry about blocking |
3366 | * retries. | |
3367 | */ | |
3368 | ||
8ae55f04 | 3369 | if (skb == NULL) |
1da177e4 LT |
3370 | goto out; |
3371 | ||
9bb6cd65 | 3372 | packet_rcv_try_clear_pressure(pkt_sk(sk)); |
2ccdbaa6 | 3373 | |
bfd5f4a3 | 3374 | if (pkt_sk(sk)->has_vnet_hdr) { |
16cc1400 WB |
3375 | err = packet_rcv_vnet(msg, skb, &len); |
3376 | if (err) | |
bfd5f4a3 | 3377 | goto out_free; |
16cc1400 | 3378 | vnet_hdr_len = sizeof(struct virtio_net_hdr); |
bfd5f4a3 SS |
3379 | } |
3380 | ||
f3d33426 HFS |
3381 | /* You lose any data beyond the buffer you gave. If it worries |
3382 | * a user program they can ask the device for its MTU | |
3383 | * anyway. | |
1da177e4 | 3384 | */ |
1da177e4 | 3385 | copied = skb->len; |
40d4e3df ED |
3386 | if (copied > len) { |
3387 | copied = len; | |
3388 | msg->msg_flags |= MSG_TRUNC; | |
1da177e4 LT |
3389 | } |
3390 | ||
51f3d02b | 3391 | err = skb_copy_datagram_msg(skb, 0, msg, copied); |
1da177e4 LT |
3392 | if (err) |
3393 | goto out_free; | |
3394 | ||
2472d761 EB |
3395 | if (sock->type != SOCK_PACKET) { |
3396 | struct sockaddr_ll *sll = &PACKET_SKB_CB(skb)->sa.ll; | |
3397 | ||
3398 | /* Original length was stored in sockaddr_ll fields */ | |
3399 | origlen = PACKET_SKB_CB(skb)->sa.origlen; | |
3400 | sll->sll_family = AF_PACKET; | |
3401 | sll->sll_protocol = skb->protocol; | |
3402 | } | |
3403 | ||
3b885787 | 3404 | sock_recv_ts_and_drops(msg, sk, skb); |
1da177e4 | 3405 | |
f3d33426 | 3406 | if (msg->msg_name) { |
b2cf86e1 WB |
3407 | int copy_len; |
3408 | ||
f3d33426 HFS |
3409 | /* If the address length field is there to be filled |
3410 | * in, we fill it in now. | |
3411 | */ | |
3412 | if (sock->type == SOCK_PACKET) { | |
342dfc30 | 3413 | __sockaddr_check_size(sizeof(struct sockaddr_pkt)); |
f3d33426 | 3414 | msg->msg_namelen = sizeof(struct sockaddr_pkt); |
b2cf86e1 | 3415 | copy_len = msg->msg_namelen; |
f3d33426 HFS |
3416 | } else { |
3417 | struct sockaddr_ll *sll = &PACKET_SKB_CB(skb)->sa.ll; | |
2472d761 | 3418 | |
f3d33426 HFS |
3419 | msg->msg_namelen = sll->sll_halen + |
3420 | offsetof(struct sockaddr_ll, sll_addr); | |
b2cf86e1 WB |
3421 | copy_len = msg->msg_namelen; |
3422 | if (msg->msg_namelen < sizeof(struct sockaddr_ll)) { | |
3423 | memset(msg->msg_name + | |
3424 | offsetof(struct sockaddr_ll, sll_addr), | |
3425 | 0, sizeof(sll->sll_addr)); | |
3426 | msg->msg_namelen = sizeof(struct sockaddr_ll); | |
3427 | } | |
f3d33426 | 3428 | } |
b2cf86e1 | 3429 | memcpy(msg->msg_name, &PACKET_SKB_CB(skb)->sa, copy_len); |
f3d33426 | 3430 | } |
1da177e4 | 3431 | |
8dc41944 | 3432 | if (pkt_sk(sk)->auxdata) { |
ffbc6111 HX |
3433 | struct tpacket_auxdata aux; |
3434 | ||
3435 | aux.tp_status = TP_STATUS_USER; | |
3436 | if (skb->ip_summed == CHECKSUM_PARTIAL) | |
3437 | aux.tp_status |= TP_STATUS_CSUMNOTREADY; | |
682f048b AD |
3438 | else if (skb->pkt_type != PACKET_OUTGOING && |
3439 | (skb->ip_summed == CHECKSUM_COMPLETE || | |
3440 | skb_csum_unnecessary(skb))) | |
3441 | aux.tp_status |= TP_STATUS_CSUM_VALID; | |
3442 | ||
2472d761 | 3443 | aux.tp_len = origlen; |
ffbc6111 HX |
3444 | aux.tp_snaplen = skb->len; |
3445 | aux.tp_mac = 0; | |
bbe735e4 | 3446 | aux.tp_net = skb_network_offset(skb); |
df8a39de JP |
3447 | if (skb_vlan_tag_present(skb)) { |
3448 | aux.tp_vlan_tci = skb_vlan_tag_get(skb); | |
a0cdfcf3 AW |
3449 | aux.tp_vlan_tpid = ntohs(skb->vlan_proto); |
3450 | aux.tp_status |= TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
a3bcc23e BG |
3451 | } else { |
3452 | aux.tp_vlan_tci = 0; | |
a0cdfcf3 | 3453 | aux.tp_vlan_tpid = 0; |
a3bcc23e | 3454 | } |
ffbc6111 | 3455 | put_cmsg(msg, SOL_PACKET, PACKET_AUXDATA, sizeof(aux), &aux); |
8dc41944 HX |
3456 | } |
3457 | ||
1da177e4 LT |
3458 | /* |
3459 | * Free or return the buffer as appropriate. Again this | |
3460 | * hides all the races and re-entrancy issues from us. | |
3461 | */ | |
bfd5f4a3 | 3462 | err = vnet_hdr_len + ((flags&MSG_TRUNC) ? skb->len : copied); |
1da177e4 LT |
3463 | |
3464 | out_free: | |
3465 | skb_free_datagram(sk, skb); | |
3466 | out: | |
3467 | return err; | |
3468 | } | |
3469 | ||
1da177e4 | 3470 | static int packet_getname_spkt(struct socket *sock, struct sockaddr *uaddr, |
9b2c45d4 | 3471 | int peer) |
1da177e4 LT |
3472 | { |
3473 | struct net_device *dev; | |
3474 | struct sock *sk = sock->sk; | |
3475 | ||
3476 | if (peer) | |
3477 | return -EOPNOTSUPP; | |
3478 | ||
3479 | uaddr->sa_family = AF_PACKET; | |
2dc85bf3 | 3480 | memset(uaddr->sa_data, 0, sizeof(uaddr->sa_data)); |
654d1f8a ED |
3481 | rcu_read_lock(); |
3482 | dev = dev_get_by_index_rcu(sock_net(sk), pkt_sk(sk)->ifindex); | |
3483 | if (dev) | |
2dc85bf3 | 3484 | strlcpy(uaddr->sa_data, dev->name, sizeof(uaddr->sa_data)); |
654d1f8a | 3485 | rcu_read_unlock(); |
1da177e4 | 3486 | |
9b2c45d4 | 3487 | return sizeof(*uaddr); |
1da177e4 | 3488 | } |
1da177e4 LT |
3489 | |
3490 | static int packet_getname(struct socket *sock, struct sockaddr *uaddr, | |
9b2c45d4 | 3491 | int peer) |
1da177e4 LT |
3492 | { |
3493 | struct net_device *dev; | |
3494 | struct sock *sk = sock->sk; | |
3495 | struct packet_sock *po = pkt_sk(sk); | |
13cfa97b | 3496 | DECLARE_SOCKADDR(struct sockaddr_ll *, sll, uaddr); |
1da177e4 LT |
3497 | |
3498 | if (peer) | |
3499 | return -EOPNOTSUPP; | |
3500 | ||
3501 | sll->sll_family = AF_PACKET; | |
3502 | sll->sll_ifindex = po->ifindex; | |
3503 | sll->sll_protocol = po->num; | |
67286640 | 3504 | sll->sll_pkttype = 0; |
654d1f8a ED |
3505 | rcu_read_lock(); |
3506 | dev = dev_get_by_index_rcu(sock_net(sk), po->ifindex); | |
1da177e4 LT |
3507 | if (dev) { |
3508 | sll->sll_hatype = dev->type; | |
3509 | sll->sll_halen = dev->addr_len; | |
3510 | memcpy(sll->sll_addr, dev->dev_addr, dev->addr_len); | |
1da177e4 LT |
3511 | } else { |
3512 | sll->sll_hatype = 0; /* Bad: we have no ARPHRD_UNSPEC */ | |
3513 | sll->sll_halen = 0; | |
3514 | } | |
654d1f8a | 3515 | rcu_read_unlock(); |
1da177e4 | 3516 | |
9b2c45d4 | 3517 | return offsetof(struct sockaddr_ll, sll_addr) + sll->sll_halen; |
1da177e4 LT |
3518 | } |
3519 | ||
2aeb0b88 WC |
3520 | static int packet_dev_mc(struct net_device *dev, struct packet_mclist *i, |
3521 | int what) | |
1da177e4 LT |
3522 | { |
3523 | switch (i->type) { | |
3524 | case PACKET_MR_MULTICAST: | |
1162563f JP |
3525 | if (i->alen != dev->addr_len) |
3526 | return -EINVAL; | |
1da177e4 | 3527 | if (what > 0) |
22bedad3 | 3528 | return dev_mc_add(dev, i->addr); |
1da177e4 | 3529 | else |
22bedad3 | 3530 | return dev_mc_del(dev, i->addr); |
1da177e4 LT |
3531 | break; |
3532 | case PACKET_MR_PROMISC: | |
2aeb0b88 | 3533 | return dev_set_promiscuity(dev, what); |
1da177e4 | 3534 | case PACKET_MR_ALLMULTI: |
2aeb0b88 | 3535 | return dev_set_allmulti(dev, what); |
d95ed927 | 3536 | case PACKET_MR_UNICAST: |
1162563f JP |
3537 | if (i->alen != dev->addr_len) |
3538 | return -EINVAL; | |
d95ed927 | 3539 | if (what > 0) |
a748ee24 | 3540 | return dev_uc_add(dev, i->addr); |
d95ed927 | 3541 | else |
a748ee24 | 3542 | return dev_uc_del(dev, i->addr); |
d95ed927 | 3543 | break; |
40d4e3df ED |
3544 | default: |
3545 | break; | |
1da177e4 | 3546 | } |
2aeb0b88 | 3547 | return 0; |
1da177e4 LT |
3548 | } |
3549 | ||
82f17091 FR |
3550 | static void packet_dev_mclist_delete(struct net_device *dev, |
3551 | struct packet_mclist **mlp) | |
1da177e4 | 3552 | { |
82f17091 FR |
3553 | struct packet_mclist *ml; |
3554 | ||
3555 | while ((ml = *mlp) != NULL) { | |
3556 | if (ml->ifindex == dev->ifindex) { | |
3557 | packet_dev_mc(dev, ml, -1); | |
3558 | *mlp = ml->next; | |
3559 | kfree(ml); | |
3560 | } else | |
3561 | mlp = &ml->next; | |
1da177e4 LT |
3562 | } |
3563 | } | |
3564 | ||
0fb375fb | 3565 | static int packet_mc_add(struct sock *sk, struct packet_mreq_max *mreq) |
1da177e4 LT |
3566 | { |
3567 | struct packet_sock *po = pkt_sk(sk); | |
3568 | struct packet_mclist *ml, *i; | |
3569 | struct net_device *dev; | |
3570 | int err; | |
3571 | ||
3572 | rtnl_lock(); | |
3573 | ||
3574 | err = -ENODEV; | |
3b1e0a65 | 3575 | dev = __dev_get_by_index(sock_net(sk), mreq->mr_ifindex); |
1da177e4 LT |
3576 | if (!dev) |
3577 | goto done; | |
3578 | ||
3579 | err = -EINVAL; | |
1162563f | 3580 | if (mreq->mr_alen > dev->addr_len) |
1da177e4 LT |
3581 | goto done; |
3582 | ||
3583 | err = -ENOBUFS; | |
8b3a7005 | 3584 | i = kmalloc(sizeof(*i), GFP_KERNEL); |
1da177e4 LT |
3585 | if (i == NULL) |
3586 | goto done; | |
3587 | ||
3588 | err = 0; | |
3589 | for (ml = po->mclist; ml; ml = ml->next) { | |
3590 | if (ml->ifindex == mreq->mr_ifindex && | |
3591 | ml->type == mreq->mr_type && | |
3592 | ml->alen == mreq->mr_alen && | |
3593 | memcmp(ml->addr, mreq->mr_address, ml->alen) == 0) { | |
3594 | ml->count++; | |
3595 | /* Free the new element ... */ | |
3596 | kfree(i); | |
3597 | goto done; | |
3598 | } | |
3599 | } | |
3600 | ||
3601 | i->type = mreq->mr_type; | |
3602 | i->ifindex = mreq->mr_ifindex; | |
3603 | i->alen = mreq->mr_alen; | |
3604 | memcpy(i->addr, mreq->mr_address, i->alen); | |
309cf37f | 3605 | memset(i->addr + i->alen, 0, sizeof(i->addr) - i->alen); |
1da177e4 LT |
3606 | i->count = 1; |
3607 | i->next = po->mclist; | |
3608 | po->mclist = i; | |
2aeb0b88 WC |
3609 | err = packet_dev_mc(dev, i, 1); |
3610 | if (err) { | |
3611 | po->mclist = i->next; | |
3612 | kfree(i); | |
3613 | } | |
1da177e4 LT |
3614 | |
3615 | done: | |
3616 | rtnl_unlock(); | |
3617 | return err; | |
3618 | } | |
3619 | ||
0fb375fb | 3620 | static int packet_mc_drop(struct sock *sk, struct packet_mreq_max *mreq) |
1da177e4 LT |
3621 | { |
3622 | struct packet_mclist *ml, **mlp; | |
3623 | ||
3624 | rtnl_lock(); | |
3625 | ||
3626 | for (mlp = &pkt_sk(sk)->mclist; (ml = *mlp) != NULL; mlp = &ml->next) { | |
3627 | if (ml->ifindex == mreq->mr_ifindex && | |
3628 | ml->type == mreq->mr_type && | |
3629 | ml->alen == mreq->mr_alen && | |
3630 | memcmp(ml->addr, mreq->mr_address, ml->alen) == 0) { | |
3631 | if (--ml->count == 0) { | |
3632 | struct net_device *dev; | |
3633 | *mlp = ml->next; | |
ad959e76 ED |
3634 | dev = __dev_get_by_index(sock_net(sk), ml->ifindex); |
3635 | if (dev) | |
1da177e4 | 3636 | packet_dev_mc(dev, ml, -1); |
1da177e4 LT |
3637 | kfree(ml); |
3638 | } | |
82f17091 | 3639 | break; |
1da177e4 LT |
3640 | } |
3641 | } | |
3642 | rtnl_unlock(); | |
82f17091 | 3643 | return 0; |
1da177e4 LT |
3644 | } |
3645 | ||
3646 | static void packet_flush_mclist(struct sock *sk) | |
3647 | { | |
3648 | struct packet_sock *po = pkt_sk(sk); | |
3649 | struct packet_mclist *ml; | |
3650 | ||
3651 | if (!po->mclist) | |
3652 | return; | |
3653 | ||
3654 | rtnl_lock(); | |
3655 | while ((ml = po->mclist) != NULL) { | |
3656 | struct net_device *dev; | |
3657 | ||
3658 | po->mclist = ml->next; | |
ad959e76 ED |
3659 | dev = __dev_get_by_index(sock_net(sk), ml->ifindex); |
3660 | if (dev != NULL) | |
1da177e4 | 3661 | packet_dev_mc(dev, ml, -1); |
1da177e4 LT |
3662 | kfree(ml); |
3663 | } | |
3664 | rtnl_unlock(); | |
3665 | } | |
1da177e4 LT |
3666 | |
3667 | static int | |
a7b75c5a CH |
3668 | packet_setsockopt(struct socket *sock, int level, int optname, sockptr_t optval, |
3669 | unsigned int optlen) | |
1da177e4 LT |
3670 | { |
3671 | struct sock *sk = sock->sk; | |
8dc41944 | 3672 | struct packet_sock *po = pkt_sk(sk); |
1da177e4 LT |
3673 | int ret; |
3674 | ||
3675 | if (level != SOL_PACKET) | |
3676 | return -ENOPROTOOPT; | |
3677 | ||
69e3c75f | 3678 | switch (optname) { |
1ce4f28b | 3679 | case PACKET_ADD_MEMBERSHIP: |
1da177e4 LT |
3680 | case PACKET_DROP_MEMBERSHIP: |
3681 | { | |
0fb375fb EB |
3682 | struct packet_mreq_max mreq; |
3683 | int len = optlen; | |
3684 | memset(&mreq, 0, sizeof(mreq)); | |
3685 | if (len < sizeof(struct packet_mreq)) | |
1da177e4 | 3686 | return -EINVAL; |
0fb375fb EB |
3687 | if (len > sizeof(mreq)) |
3688 | len = sizeof(mreq); | |
a7b75c5a | 3689 | if (copy_from_sockptr(&mreq, optval, len)) |
1da177e4 | 3690 | return -EFAULT; |
0fb375fb EB |
3691 | if (len < (mreq.mr_alen + offsetof(struct packet_mreq, mr_address))) |
3692 | return -EINVAL; | |
1da177e4 LT |
3693 | if (optname == PACKET_ADD_MEMBERSHIP) |
3694 | ret = packet_mc_add(sk, &mreq); | |
3695 | else | |
3696 | ret = packet_mc_drop(sk, &mreq); | |
3697 | return ret; | |
3698 | } | |
a2efcfa0 | 3699 | |
1da177e4 | 3700 | case PACKET_RX_RING: |
69e3c75f | 3701 | case PACKET_TX_RING: |
1da177e4 | 3702 | { |
f6fb8f10 | 3703 | union tpacket_req_u req_u; |
3704 | int len; | |
1da177e4 | 3705 | |
5171b37d | 3706 | lock_sock(sk); |
f6fb8f10 | 3707 | switch (po->tp_version) { |
3708 | case TPACKET_V1: | |
3709 | case TPACKET_V2: | |
3710 | len = sizeof(req_u.req); | |
3711 | break; | |
3712 | case TPACKET_V3: | |
3713 | default: | |
3714 | len = sizeof(req_u.req3); | |
3715 | break; | |
3716 | } | |
5171b37d ED |
3717 | if (optlen < len) { |
3718 | ret = -EINVAL; | |
3719 | } else { | |
a7b75c5a | 3720 | if (copy_from_sockptr(&req_u.req, optval, len)) |
5171b37d ED |
3721 | ret = -EFAULT; |
3722 | else | |
3723 | ret = packet_set_ring(sk, &req_u, 0, | |
3724 | optname == PACKET_TX_RING); | |
3725 | } | |
3726 | release_sock(sk); | |
3727 | return ret; | |
1da177e4 LT |
3728 | } |
3729 | case PACKET_COPY_THRESH: | |
3730 | { | |
3731 | int val; | |
3732 | ||
40d4e3df | 3733 | if (optlen != sizeof(val)) |
1da177e4 | 3734 | return -EINVAL; |
a7b75c5a | 3735 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
1da177e4 LT |
3736 | return -EFAULT; |
3737 | ||
3738 | pkt_sk(sk)->copy_thresh = val; | |
3739 | return 0; | |
3740 | } | |
bbd6ef87 PM |
3741 | case PACKET_VERSION: |
3742 | { | |
3743 | int val; | |
3744 | ||
3745 | if (optlen != sizeof(val)) | |
3746 | return -EINVAL; | |
a7b75c5a | 3747 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
bbd6ef87 PM |
3748 | return -EFAULT; |
3749 | switch (val) { | |
3750 | case TPACKET_V1: | |
3751 | case TPACKET_V2: | |
f6fb8f10 | 3752 | case TPACKET_V3: |
84ac7260 | 3753 | break; |
bbd6ef87 PM |
3754 | default: |
3755 | return -EINVAL; | |
3756 | } | |
84ac7260 PP |
3757 | lock_sock(sk); |
3758 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3759 | ret = -EBUSY; | |
3760 | } else { | |
3761 | po->tp_version = val; | |
3762 | ret = 0; | |
3763 | } | |
3764 | release_sock(sk); | |
3765 | return ret; | |
bbd6ef87 | 3766 | } |
8913336a PM |
3767 | case PACKET_RESERVE: |
3768 | { | |
3769 | unsigned int val; | |
3770 | ||
3771 | if (optlen != sizeof(val)) | |
3772 | return -EINVAL; | |
a7b75c5a | 3773 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
8913336a | 3774 | return -EFAULT; |
bcc5364b AK |
3775 | if (val > INT_MAX) |
3776 | return -EINVAL; | |
c27927e3 WB |
3777 | lock_sock(sk); |
3778 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3779 | ret = -EBUSY; | |
3780 | } else { | |
3781 | po->tp_reserve = val; | |
3782 | ret = 0; | |
3783 | } | |
3784 | release_sock(sk); | |
3785 | return ret; | |
8913336a | 3786 | } |
69e3c75f JB |
3787 | case PACKET_LOSS: |
3788 | { | |
3789 | unsigned int val; | |
3790 | ||
3791 | if (optlen != sizeof(val)) | |
3792 | return -EINVAL; | |
a7b75c5a | 3793 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
69e3c75f | 3794 | return -EFAULT; |
a6361f0c WB |
3795 | |
3796 | lock_sock(sk); | |
3797 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3798 | ret = -EBUSY; | |
3799 | } else { | |
3800 | po->tp_loss = !!val; | |
3801 | ret = 0; | |
3802 | } | |
3803 | release_sock(sk); | |
3804 | return ret; | |
69e3c75f | 3805 | } |
8dc41944 HX |
3806 | case PACKET_AUXDATA: |
3807 | { | |
3808 | int val; | |
3809 | ||
3810 | if (optlen < sizeof(val)) | |
3811 | return -EINVAL; | |
a7b75c5a | 3812 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
8dc41944 HX |
3813 | return -EFAULT; |
3814 | ||
a6361f0c | 3815 | lock_sock(sk); |
8dc41944 | 3816 | po->auxdata = !!val; |
a6361f0c | 3817 | release_sock(sk); |
8dc41944 HX |
3818 | return 0; |
3819 | } | |
80feaacb PWJ |
3820 | case PACKET_ORIGDEV: |
3821 | { | |
3822 | int val; | |
3823 | ||
3824 | if (optlen < sizeof(val)) | |
3825 | return -EINVAL; | |
a7b75c5a | 3826 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
80feaacb PWJ |
3827 | return -EFAULT; |
3828 | ||
a6361f0c | 3829 | lock_sock(sk); |
80feaacb | 3830 | po->origdev = !!val; |
a6361f0c | 3831 | release_sock(sk); |
80feaacb PWJ |
3832 | return 0; |
3833 | } | |
bfd5f4a3 SS |
3834 | case PACKET_VNET_HDR: |
3835 | { | |
3836 | int val; | |
3837 | ||
3838 | if (sock->type != SOCK_RAW) | |
3839 | return -EINVAL; | |
bfd5f4a3 SS |
3840 | if (optlen < sizeof(val)) |
3841 | return -EINVAL; | |
a7b75c5a | 3842 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
bfd5f4a3 SS |
3843 | return -EFAULT; |
3844 | ||
a6361f0c WB |
3845 | lock_sock(sk); |
3846 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3847 | ret = -EBUSY; | |
3848 | } else { | |
3849 | po->has_vnet_hdr = !!val; | |
3850 | ret = 0; | |
3851 | } | |
3852 | release_sock(sk); | |
3853 | return ret; | |
bfd5f4a3 | 3854 | } |
614f60fa SM |
3855 | case PACKET_TIMESTAMP: |
3856 | { | |
3857 | int val; | |
3858 | ||
3859 | if (optlen != sizeof(val)) | |
3860 | return -EINVAL; | |
a7b75c5a | 3861 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
614f60fa SM |
3862 | return -EFAULT; |
3863 | ||
3864 | po->tp_tstamp = val; | |
3865 | return 0; | |
3866 | } | |
dc99f600 DM |
3867 | case PACKET_FANOUT: |
3868 | { | |
3869 | int val; | |
3870 | ||
3871 | if (optlen != sizeof(val)) | |
3872 | return -EINVAL; | |
a7b75c5a | 3873 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
dc99f600 DM |
3874 | return -EFAULT; |
3875 | ||
3876 | return fanout_add(sk, val & 0xffff, val >> 16); | |
3877 | } | |
47dceb8e WB |
3878 | case PACKET_FANOUT_DATA: |
3879 | { | |
3880 | if (!po->fanout) | |
3881 | return -EINVAL; | |
3882 | ||
3883 | return fanout_set_data(po, optval, optlen); | |
3884 | } | |
fa788d98 VW |
3885 | case PACKET_IGNORE_OUTGOING: |
3886 | { | |
3887 | int val; | |
3888 | ||
3889 | if (optlen != sizeof(val)) | |
3890 | return -EINVAL; | |
a7b75c5a | 3891 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
fa788d98 VW |
3892 | return -EFAULT; |
3893 | if (val < 0 || val > 1) | |
3894 | return -EINVAL; | |
3895 | ||
3896 | po->prot_hook.ignore_outgoing = !!val; | |
3897 | return 0; | |
3898 | } | |
5920cd3a PC |
3899 | case PACKET_TX_HAS_OFF: |
3900 | { | |
3901 | unsigned int val; | |
3902 | ||
3903 | if (optlen != sizeof(val)) | |
3904 | return -EINVAL; | |
a7b75c5a | 3905 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
5920cd3a | 3906 | return -EFAULT; |
a6361f0c WB |
3907 | |
3908 | lock_sock(sk); | |
3909 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3910 | ret = -EBUSY; | |
3911 | } else { | |
3912 | po->tp_tx_has_off = !!val; | |
3913 | ret = 0; | |
3914 | } | |
3915 | release_sock(sk); | |
5920cd3a PC |
3916 | return 0; |
3917 | } | |
d346a3fa DB |
3918 | case PACKET_QDISC_BYPASS: |
3919 | { | |
3920 | int val; | |
3921 | ||
3922 | if (optlen != sizeof(val)) | |
3923 | return -EINVAL; | |
a7b75c5a | 3924 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
d346a3fa DB |
3925 | return -EFAULT; |
3926 | ||
3927 | po->xmit = val ? packet_direct_xmit : dev_queue_xmit; | |
3928 | return 0; | |
3929 | } | |
1da177e4 LT |
3930 | default: |
3931 | return -ENOPROTOOPT; | |
3932 | } | |
3933 | } | |
3934 | ||
3935 | static int packet_getsockopt(struct socket *sock, int level, int optname, | |
3936 | char __user *optval, int __user *optlen) | |
3937 | { | |
3938 | int len; | |
c06fff6e | 3939 | int val, lv = sizeof(val); |
1da177e4 LT |
3940 | struct sock *sk = sock->sk; |
3941 | struct packet_sock *po = pkt_sk(sk); | |
c06fff6e | 3942 | void *data = &val; |
ee80fbf3 | 3943 | union tpacket_stats_u st; |
a9b63918 | 3944 | struct tpacket_rollover_stats rstats; |
8e8e2951 | 3945 | int drops; |
1da177e4 LT |
3946 | |
3947 | if (level != SOL_PACKET) | |
3948 | return -ENOPROTOOPT; | |
3949 | ||
8ae55f04 KK |
3950 | if (get_user(len, optlen)) |
3951 | return -EFAULT; | |
1da177e4 LT |
3952 | |
3953 | if (len < 0) | |
3954 | return -EINVAL; | |
1ce4f28b | 3955 | |
69e3c75f | 3956 | switch (optname) { |
1da177e4 | 3957 | case PACKET_STATISTICS: |
1da177e4 | 3958 | spin_lock_bh(&sk->sk_receive_queue.lock); |
ee80fbf3 DB |
3959 | memcpy(&st, &po->stats, sizeof(st)); |
3960 | memset(&po->stats, 0, sizeof(po->stats)); | |
3961 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
8e8e2951 | 3962 | drops = atomic_xchg(&po->tp_drops, 0); |
ee80fbf3 | 3963 | |
f6fb8f10 | 3964 | if (po->tp_version == TPACKET_V3) { |
c06fff6e | 3965 | lv = sizeof(struct tpacket_stats_v3); |
8e8e2951 ED |
3966 | st.stats3.tp_drops = drops; |
3967 | st.stats3.tp_packets += drops; | |
ee80fbf3 | 3968 | data = &st.stats3; |
f6fb8f10 | 3969 | } else { |
c06fff6e | 3970 | lv = sizeof(struct tpacket_stats); |
8e8e2951 ED |
3971 | st.stats1.tp_drops = drops; |
3972 | st.stats1.tp_packets += drops; | |
ee80fbf3 | 3973 | data = &st.stats1; |
f6fb8f10 | 3974 | } |
ee80fbf3 | 3975 | |
8dc41944 HX |
3976 | break; |
3977 | case PACKET_AUXDATA: | |
8dc41944 | 3978 | val = po->auxdata; |
80feaacb PWJ |
3979 | break; |
3980 | case PACKET_ORIGDEV: | |
80feaacb | 3981 | val = po->origdev; |
bfd5f4a3 SS |
3982 | break; |
3983 | case PACKET_VNET_HDR: | |
bfd5f4a3 | 3984 | val = po->has_vnet_hdr; |
1da177e4 | 3985 | break; |
bbd6ef87 | 3986 | case PACKET_VERSION: |
bbd6ef87 | 3987 | val = po->tp_version; |
bbd6ef87 PM |
3988 | break; |
3989 | case PACKET_HDRLEN: | |
3990 | if (len > sizeof(int)) | |
3991 | len = sizeof(int); | |
fd2c83b3 AP |
3992 | if (len < sizeof(int)) |
3993 | return -EINVAL; | |
bbd6ef87 PM |
3994 | if (copy_from_user(&val, optval, len)) |
3995 | return -EFAULT; | |
3996 | switch (val) { | |
3997 | case TPACKET_V1: | |
3998 | val = sizeof(struct tpacket_hdr); | |
3999 | break; | |
4000 | case TPACKET_V2: | |
4001 | val = sizeof(struct tpacket2_hdr); | |
4002 | break; | |
f6fb8f10 | 4003 | case TPACKET_V3: |
4004 | val = sizeof(struct tpacket3_hdr); | |
4005 | break; | |
bbd6ef87 PM |
4006 | default: |
4007 | return -EINVAL; | |
4008 | } | |
bbd6ef87 | 4009 | break; |
8913336a | 4010 | case PACKET_RESERVE: |
8913336a | 4011 | val = po->tp_reserve; |
8913336a | 4012 | break; |
69e3c75f | 4013 | case PACKET_LOSS: |
69e3c75f | 4014 | val = po->tp_loss; |
69e3c75f | 4015 | break; |
614f60fa | 4016 | case PACKET_TIMESTAMP: |
614f60fa | 4017 | val = po->tp_tstamp; |
614f60fa | 4018 | break; |
dc99f600 | 4019 | case PACKET_FANOUT: |
dc99f600 DM |
4020 | val = (po->fanout ? |
4021 | ((u32)po->fanout->id | | |
77f65ebd WB |
4022 | ((u32)po->fanout->type << 16) | |
4023 | ((u32)po->fanout->flags << 24)) : | |
dc99f600 | 4024 | 0); |
dc99f600 | 4025 | break; |
fa788d98 VW |
4026 | case PACKET_IGNORE_OUTGOING: |
4027 | val = po->prot_hook.ignore_outgoing; | |
4028 | break; | |
a9b63918 | 4029 | case PACKET_ROLLOVER_STATS: |
57f015f5 | 4030 | if (!po->rollover) |
a9b63918 | 4031 | return -EINVAL; |
57f015f5 MM |
4032 | rstats.tp_all = atomic_long_read(&po->rollover->num); |
4033 | rstats.tp_huge = atomic_long_read(&po->rollover->num_huge); | |
4034 | rstats.tp_failed = atomic_long_read(&po->rollover->num_failed); | |
4035 | data = &rstats; | |
4036 | lv = sizeof(rstats); | |
a9b63918 | 4037 | break; |
5920cd3a PC |
4038 | case PACKET_TX_HAS_OFF: |
4039 | val = po->tp_tx_has_off; | |
4040 | break; | |
d346a3fa DB |
4041 | case PACKET_QDISC_BYPASS: |
4042 | val = packet_use_direct_xmit(po); | |
4043 | break; | |
1da177e4 LT |
4044 | default: |
4045 | return -ENOPROTOOPT; | |
4046 | } | |
4047 | ||
c06fff6e ED |
4048 | if (len > lv) |
4049 | len = lv; | |
8ae55f04 KK |
4050 | if (put_user(len, optlen)) |
4051 | return -EFAULT; | |
8dc41944 HX |
4052 | if (copy_to_user(optval, data, len)) |
4053 | return -EFAULT; | |
8ae55f04 | 4054 | return 0; |
1da177e4 LT |
4055 | } |
4056 | ||
351638e7 JP |
4057 | static int packet_notifier(struct notifier_block *this, |
4058 | unsigned long msg, void *ptr) | |
1da177e4 LT |
4059 | { |
4060 | struct sock *sk; | |
351638e7 | 4061 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); |
c346dca1 | 4062 | struct net *net = dev_net(dev); |
1da177e4 | 4063 | |
808f5114 | 4064 | rcu_read_lock(); |
b67bfe0d | 4065 | sk_for_each_rcu(sk, &net->packet.sklist) { |
1da177e4 LT |
4066 | struct packet_sock *po = pkt_sk(sk); |
4067 | ||
4068 | switch (msg) { | |
4069 | case NETDEV_UNREGISTER: | |
1da177e4 | 4070 | if (po->mclist) |
82f17091 | 4071 | packet_dev_mclist_delete(dev, &po->mclist); |
df561f66 | 4072 | fallthrough; |
a2efcfa0 | 4073 | |
1da177e4 LT |
4074 | case NETDEV_DOWN: |
4075 | if (dev->ifindex == po->ifindex) { | |
4076 | spin_lock(&po->bind_lock); | |
4077 | if (po->running) { | |
ce06b03e | 4078 | __unregister_prot_hook(sk, false); |
1da177e4 LT |
4079 | sk->sk_err = ENETDOWN; |
4080 | if (!sock_flag(sk, SOCK_DEAD)) | |
4081 | sk->sk_error_report(sk); | |
4082 | } | |
4083 | if (msg == NETDEV_UNREGISTER) { | |
66e56cd4 | 4084 | packet_cached_dev_reset(po); |
1da177e4 | 4085 | po->ifindex = -1; |
160ff18a BG |
4086 | if (po->prot_hook.dev) |
4087 | dev_put(po->prot_hook.dev); | |
1da177e4 LT |
4088 | po->prot_hook.dev = NULL; |
4089 | } | |
4090 | spin_unlock(&po->bind_lock); | |
4091 | } | |
4092 | break; | |
4093 | case NETDEV_UP: | |
808f5114 | 4094 | if (dev->ifindex == po->ifindex) { |
4095 | spin_lock(&po->bind_lock); | |
ce06b03e DM |
4096 | if (po->num) |
4097 | register_prot_hook(sk); | |
808f5114 | 4098 | spin_unlock(&po->bind_lock); |
1da177e4 | 4099 | } |
1da177e4 LT |
4100 | break; |
4101 | } | |
4102 | } | |
808f5114 | 4103 | rcu_read_unlock(); |
1da177e4 LT |
4104 | return NOTIFY_DONE; |
4105 | } | |
4106 | ||
4107 | ||
4108 | static int packet_ioctl(struct socket *sock, unsigned int cmd, | |
4109 | unsigned long arg) | |
4110 | { | |
4111 | struct sock *sk = sock->sk; | |
4112 | ||
69e3c75f | 4113 | switch (cmd) { |
40d4e3df ED |
4114 | case SIOCOUTQ: |
4115 | { | |
4116 | int amount = sk_wmem_alloc_get(sk); | |
31e6d363 | 4117 | |
40d4e3df ED |
4118 | return put_user(amount, (int __user *)arg); |
4119 | } | |
4120 | case SIOCINQ: | |
4121 | { | |
4122 | struct sk_buff *skb; | |
4123 | int amount = 0; | |
4124 | ||
4125 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
4126 | skb = skb_peek(&sk->sk_receive_queue); | |
4127 | if (skb) | |
4128 | amount = skb->len; | |
4129 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
4130 | return put_user(amount, (int __user *)arg); | |
4131 | } | |
1da177e4 | 4132 | #ifdef CONFIG_INET |
40d4e3df ED |
4133 | case SIOCADDRT: |
4134 | case SIOCDELRT: | |
4135 | case SIOCDARP: | |
4136 | case SIOCGARP: | |
4137 | case SIOCSARP: | |
4138 | case SIOCGIFADDR: | |
4139 | case SIOCSIFADDR: | |
4140 | case SIOCGIFBRDADDR: | |
4141 | case SIOCSIFBRDADDR: | |
4142 | case SIOCGIFNETMASK: | |
4143 | case SIOCSIFNETMASK: | |
4144 | case SIOCGIFDSTADDR: | |
4145 | case SIOCSIFDSTADDR: | |
4146 | case SIOCSIFFLAGS: | |
40d4e3df | 4147 | return inet_dgram_ops.ioctl(sock, cmd, arg); |
1da177e4 LT |
4148 | #endif |
4149 | ||
40d4e3df ED |
4150 | default: |
4151 | return -ENOIOCTLCMD; | |
1da177e4 LT |
4152 | } |
4153 | return 0; | |
4154 | } | |
4155 | ||
a11e1d43 LT |
4156 | static __poll_t packet_poll(struct file *file, struct socket *sock, |
4157 | poll_table *wait) | |
1da177e4 LT |
4158 | { |
4159 | struct sock *sk = sock->sk; | |
4160 | struct packet_sock *po = pkt_sk(sk); | |
a11e1d43 | 4161 | __poll_t mask = datagram_poll(file, sock, wait); |
1da177e4 LT |
4162 | |
4163 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
69e3c75f | 4164 | if (po->rx_ring.pg_vec) { |
f6fb8f10 | 4165 | if (!packet_previous_rx_frame(po, &po->rx_ring, |
4166 | TP_STATUS_KERNEL)) | |
a9a08845 | 4167 | mask |= EPOLLIN | EPOLLRDNORM; |
1da177e4 | 4168 | } |
9bb6cd65 | 4169 | packet_rcv_try_clear_pressure(po); |
1da177e4 | 4170 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
69e3c75f JB |
4171 | spin_lock_bh(&sk->sk_write_queue.lock); |
4172 | if (po->tx_ring.pg_vec) { | |
4173 | if (packet_current_frame(po, &po->tx_ring, TP_STATUS_AVAILABLE)) | |
a9a08845 | 4174 | mask |= EPOLLOUT | EPOLLWRNORM; |
69e3c75f JB |
4175 | } |
4176 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
1da177e4 LT |
4177 | return mask; |
4178 | } | |
4179 | ||
4180 | ||
4181 | /* Dirty? Well, I still did not learn better way to account | |
4182 | * for user mmaps. | |
4183 | */ | |
4184 | ||
4185 | static void packet_mm_open(struct vm_area_struct *vma) | |
4186 | { | |
4187 | struct file *file = vma->vm_file; | |
40d4e3df | 4188 | struct socket *sock = file->private_data; |
1da177e4 | 4189 | struct sock *sk = sock->sk; |
1ce4f28b | 4190 | |
1da177e4 LT |
4191 | if (sk) |
4192 | atomic_inc(&pkt_sk(sk)->mapped); | |
4193 | } | |
4194 | ||
4195 | static void packet_mm_close(struct vm_area_struct *vma) | |
4196 | { | |
4197 | struct file *file = vma->vm_file; | |
40d4e3df | 4198 | struct socket *sock = file->private_data; |
1da177e4 | 4199 | struct sock *sk = sock->sk; |
1ce4f28b | 4200 | |
1da177e4 LT |
4201 | if (sk) |
4202 | atomic_dec(&pkt_sk(sk)->mapped); | |
4203 | } | |
4204 | ||
f0f37e2f | 4205 | static const struct vm_operations_struct packet_mmap_ops = { |
40d4e3df ED |
4206 | .open = packet_mm_open, |
4207 | .close = packet_mm_close, | |
1da177e4 LT |
4208 | }; |
4209 | ||
3a7ad063 ED |
4210 | static void free_pg_vec(struct pgv *pg_vec, unsigned int order, |
4211 | unsigned int len) | |
1da177e4 LT |
4212 | { |
4213 | int i; | |
4214 | ||
4ebf0ae2 | 4215 | for (i = 0; i < len; i++) { |
0e3125c7 | 4216 | if (likely(pg_vec[i].buffer)) { |
3a7ad063 ED |
4217 | if (is_vmalloc_addr(pg_vec[i].buffer)) |
4218 | vfree(pg_vec[i].buffer); | |
4219 | else | |
4220 | free_pages((unsigned long)pg_vec[i].buffer, | |
4221 | order); | |
0e3125c7 NH |
4222 | pg_vec[i].buffer = NULL; |
4223 | } | |
1da177e4 LT |
4224 | } |
4225 | kfree(pg_vec); | |
4226 | } | |
4227 | ||
3a7ad063 | 4228 | static char *alloc_one_pg_vec_page(unsigned long order) |
4ebf0ae2 | 4229 | { |
f0d4eb29 | 4230 | char *buffer; |
3a7ad063 ED |
4231 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | |
4232 | __GFP_ZERO | __GFP_NOWARN | __GFP_NORETRY; | |
0e3125c7 | 4233 | |
3a7ad063 | 4234 | buffer = (char *) __get_free_pages(gfp_flags, order); |
0e3125c7 NH |
4235 | if (buffer) |
4236 | return buffer; | |
4237 | ||
3a7ad063 ED |
4238 | /* __get_free_pages failed, fall back to vmalloc */ |
4239 | buffer = vzalloc(array_size((1 << order), PAGE_SIZE)); | |
4240 | if (buffer) | |
4241 | return buffer; | |
0e3125c7 | 4242 | |
3a7ad063 ED |
4243 | /* vmalloc failed, lets dig into swap here */ |
4244 | gfp_flags &= ~__GFP_NORETRY; | |
4245 | buffer = (char *) __get_free_pages(gfp_flags, order); | |
4246 | if (buffer) | |
4247 | return buffer; | |
4248 | ||
4249 | /* complete and utter failure */ | |
4250 | return NULL; | |
4ebf0ae2 DM |
4251 | } |
4252 | ||
3a7ad063 | 4253 | static struct pgv *alloc_pg_vec(struct tpacket_req *req, int order) |
4ebf0ae2 DM |
4254 | { |
4255 | unsigned int block_nr = req->tp_block_nr; | |
0e3125c7 | 4256 | struct pgv *pg_vec; |
4ebf0ae2 DM |
4257 | int i; |
4258 | ||
398f0132 | 4259 | pg_vec = kcalloc(block_nr, sizeof(struct pgv), GFP_KERNEL | __GFP_NOWARN); |
4ebf0ae2 DM |
4260 | if (unlikely(!pg_vec)) |
4261 | goto out; | |
4262 | ||
4263 | for (i = 0; i < block_nr; i++) { | |
3a7ad063 | 4264 | pg_vec[i].buffer = alloc_one_pg_vec_page(order); |
0e3125c7 | 4265 | if (unlikely(!pg_vec[i].buffer)) |
4ebf0ae2 DM |
4266 | goto out_free_pgvec; |
4267 | } | |
4268 | ||
4269 | out: | |
4270 | return pg_vec; | |
4271 | ||
4272 | out_free_pgvec: | |
3a7ad063 | 4273 | free_pg_vec(pg_vec, order, block_nr); |
4ebf0ae2 DM |
4274 | pg_vec = NULL; |
4275 | goto out; | |
4276 | } | |
1da177e4 | 4277 | |
f6fb8f10 | 4278 | static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u, |
69e3c75f | 4279 | int closing, int tx_ring) |
1da177e4 | 4280 | { |
0e3125c7 | 4281 | struct pgv *pg_vec = NULL; |
1da177e4 | 4282 | struct packet_sock *po = pkt_sk(sk); |
61fad681 | 4283 | unsigned long *rx_owner_map = NULL; |
3a7ad063 | 4284 | int was_running, order = 0; |
69e3c75f JB |
4285 | struct packet_ring_buffer *rb; |
4286 | struct sk_buff_head *rb_queue; | |
0e11c91e | 4287 | __be16 num; |
2a6d6c31 | 4288 | int err; |
f6fb8f10 | 4289 | /* Added to avoid minimal code churn */ |
4290 | struct tpacket_req *req = &req_u->req; | |
4291 | ||
69e3c75f JB |
4292 | rb = tx_ring ? &po->tx_ring : &po->rx_ring; |
4293 | rb_queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
1da177e4 | 4294 | |
69e3c75f JB |
4295 | err = -EBUSY; |
4296 | if (!closing) { | |
4297 | if (atomic_read(&po->mapped)) | |
4298 | goto out; | |
b0138408 | 4299 | if (packet_read_pending(rb)) |
69e3c75f JB |
4300 | goto out; |
4301 | } | |
1da177e4 | 4302 | |
69e3c75f | 4303 | if (req->tp_block_nr) { |
4576cd46 WB |
4304 | unsigned int min_frame_size; |
4305 | ||
69e3c75f JB |
4306 | /* Sanity tests and some calculations */ |
4307 | err = -EBUSY; | |
4308 | if (unlikely(rb->pg_vec)) | |
4309 | goto out; | |
1da177e4 | 4310 | |
bbd6ef87 PM |
4311 | switch (po->tp_version) { |
4312 | case TPACKET_V1: | |
4313 | po->tp_hdrlen = TPACKET_HDRLEN; | |
4314 | break; | |
4315 | case TPACKET_V2: | |
4316 | po->tp_hdrlen = TPACKET2_HDRLEN; | |
4317 | break; | |
f6fb8f10 | 4318 | case TPACKET_V3: |
4319 | po->tp_hdrlen = TPACKET3_HDRLEN; | |
4320 | break; | |
bbd6ef87 PM |
4321 | } |
4322 | ||
69e3c75f | 4323 | err = -EINVAL; |
4ebf0ae2 | 4324 | if (unlikely((int)req->tp_block_size <= 0)) |
69e3c75f | 4325 | goto out; |
90836b67 | 4326 | if (unlikely(!PAGE_ALIGNED(req->tp_block_size))) |
69e3c75f | 4327 | goto out; |
4576cd46 | 4328 | min_frame_size = po->tp_hdrlen + po->tp_reserve; |
dc808110 | 4329 | if (po->tp_version >= TPACKET_V3 && |
4576cd46 WB |
4330 | req->tp_block_size < |
4331 | BLK_PLUS_PRIV((u64)req_u->req3.tp_sizeof_priv) + min_frame_size) | |
dc808110 | 4332 | goto out; |
4576cd46 | 4333 | if (unlikely(req->tp_frame_size < min_frame_size)) |
69e3c75f | 4334 | goto out; |
4ebf0ae2 | 4335 | if (unlikely(req->tp_frame_size & (TPACKET_ALIGNMENT - 1))) |
69e3c75f | 4336 | goto out; |
1da177e4 | 4337 | |
4194b491 TK |
4338 | rb->frames_per_block = req->tp_block_size / req->tp_frame_size; |
4339 | if (unlikely(rb->frames_per_block == 0)) | |
69e3c75f | 4340 | goto out; |
fc62814d | 4341 | if (unlikely(rb->frames_per_block > UINT_MAX / req->tp_block_nr)) |
8f8d28e4 | 4342 | goto out; |
69e3c75f JB |
4343 | if (unlikely((rb->frames_per_block * req->tp_block_nr) != |
4344 | req->tp_frame_nr)) | |
4345 | goto out; | |
1da177e4 LT |
4346 | |
4347 | err = -ENOMEM; | |
3a7ad063 ED |
4348 | order = get_order(req->tp_block_size); |
4349 | pg_vec = alloc_pg_vec(req, order); | |
4ebf0ae2 | 4350 | if (unlikely(!pg_vec)) |
1da177e4 | 4351 | goto out; |
f6fb8f10 | 4352 | switch (po->tp_version) { |
4353 | case TPACKET_V3: | |
7f953ab2 SV |
4354 | /* Block transmit is not supported yet */ |
4355 | if (!tx_ring) { | |
e8e85cc5 | 4356 | init_prb_bdqc(po, rb, pg_vec, req_u); |
7f953ab2 SV |
4357 | } else { |
4358 | struct tpacket_req3 *req3 = &req_u->req3; | |
4359 | ||
4360 | if (req3->tp_retire_blk_tov || | |
4361 | req3->tp_sizeof_priv || | |
4362 | req3->tp_feature_req_word) { | |
4363 | err = -EINVAL; | |
55655e3d | 4364 | goto out_free_pg_vec; |
7f953ab2 SV |
4365 | } |
4366 | } | |
d7cf0c34 | 4367 | break; |
f6fb8f10 | 4368 | default: |
61fad681 WB |
4369 | if (!tx_ring) { |
4370 | rx_owner_map = bitmap_alloc(req->tp_frame_nr, | |
4371 | GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO); | |
4372 | if (!rx_owner_map) | |
4373 | goto out_free_pg_vec; | |
4374 | } | |
f6fb8f10 | 4375 | break; |
4376 | } | |
69e3c75f JB |
4377 | } |
4378 | /* Done */ | |
4379 | else { | |
4380 | err = -EINVAL; | |
4ebf0ae2 | 4381 | if (unlikely(req->tp_frame_nr)) |
69e3c75f | 4382 | goto out; |
1da177e4 LT |
4383 | } |
4384 | ||
1da177e4 LT |
4385 | |
4386 | /* Detach socket from network */ | |
4387 | spin_lock(&po->bind_lock); | |
4388 | was_running = po->running; | |
4389 | num = po->num; | |
4390 | if (was_running) { | |
1da177e4 | 4391 | po->num = 0; |
ce06b03e | 4392 | __unregister_prot_hook(sk, false); |
1da177e4 LT |
4393 | } |
4394 | spin_unlock(&po->bind_lock); | |
1ce4f28b | 4395 | |
1da177e4 LT |
4396 | synchronize_net(); |
4397 | ||
4398 | err = -EBUSY; | |
905db440 | 4399 | mutex_lock(&po->pg_vec_lock); |
1da177e4 LT |
4400 | if (closing || atomic_read(&po->mapped) == 0) { |
4401 | err = 0; | |
69e3c75f | 4402 | spin_lock_bh(&rb_queue->lock); |
c053fd96 | 4403 | swap(rb->pg_vec, pg_vec); |
61fad681 WB |
4404 | if (po->tp_version <= TPACKET_V2) |
4405 | swap(rb->rx_owner_map, rx_owner_map); | |
69e3c75f JB |
4406 | rb->frame_max = (req->tp_frame_nr - 1); |
4407 | rb->head = 0; | |
4408 | rb->frame_size = req->tp_frame_size; | |
4409 | spin_unlock_bh(&rb_queue->lock); | |
4410 | ||
3a7ad063 | 4411 | swap(rb->pg_vec_order, order); |
c053fd96 | 4412 | swap(rb->pg_vec_len, req->tp_block_nr); |
69e3c75f JB |
4413 | |
4414 | rb->pg_vec_pages = req->tp_block_size/PAGE_SIZE; | |
4415 | po->prot_hook.func = (po->rx_ring.pg_vec) ? | |
4416 | tpacket_rcv : packet_rcv; | |
4417 | skb_queue_purge(rb_queue); | |
1da177e4 | 4418 | if (atomic_read(&po->mapped)) |
40d4e3df ED |
4419 | pr_err("packet_mmap: vma is busy: %d\n", |
4420 | atomic_read(&po->mapped)); | |
1da177e4 | 4421 | } |
905db440 | 4422 | mutex_unlock(&po->pg_vec_lock); |
1da177e4 LT |
4423 | |
4424 | spin_lock(&po->bind_lock); | |
ce06b03e | 4425 | if (was_running) { |
1da177e4 | 4426 | po->num = num; |
ce06b03e | 4427 | register_prot_hook(sk); |
1da177e4 LT |
4428 | } |
4429 | spin_unlock(&po->bind_lock); | |
c800aaf8 | 4430 | if (pg_vec && (po->tp_version > TPACKET_V2)) { |
f6fb8f10 | 4431 | /* Because we don't support block-based V3 on tx-ring */ |
4432 | if (!tx_ring) | |
73d0fcf2 | 4433 | prb_shutdown_retire_blk_timer(po, rb_queue); |
f6fb8f10 | 4434 | } |
1da177e4 | 4435 | |
55655e3d | 4436 | out_free_pg_vec: |
61fad681 | 4437 | bitmap_free(rx_owner_map); |
1da177e4 | 4438 | if (pg_vec) |
3a7ad063 | 4439 | free_pg_vec(pg_vec, order, req->tp_block_nr); |
1da177e4 LT |
4440 | out: |
4441 | return err; | |
4442 | } | |
4443 | ||
69e3c75f JB |
4444 | static int packet_mmap(struct file *file, struct socket *sock, |
4445 | struct vm_area_struct *vma) | |
1da177e4 LT |
4446 | { |
4447 | struct sock *sk = sock->sk; | |
4448 | struct packet_sock *po = pkt_sk(sk); | |
69e3c75f JB |
4449 | unsigned long size, expected_size; |
4450 | struct packet_ring_buffer *rb; | |
1da177e4 LT |
4451 | unsigned long start; |
4452 | int err = -EINVAL; | |
4453 | int i; | |
4454 | ||
4455 | if (vma->vm_pgoff) | |
4456 | return -EINVAL; | |
4457 | ||
905db440 | 4458 | mutex_lock(&po->pg_vec_lock); |
69e3c75f JB |
4459 | |
4460 | expected_size = 0; | |
4461 | for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) { | |
4462 | if (rb->pg_vec) { | |
4463 | expected_size += rb->pg_vec_len | |
4464 | * rb->pg_vec_pages | |
4465 | * PAGE_SIZE; | |
4466 | } | |
4467 | } | |
4468 | ||
4469 | if (expected_size == 0) | |
1da177e4 | 4470 | goto out; |
69e3c75f JB |
4471 | |
4472 | size = vma->vm_end - vma->vm_start; | |
4473 | if (size != expected_size) | |
1da177e4 LT |
4474 | goto out; |
4475 | ||
1da177e4 | 4476 | start = vma->vm_start; |
69e3c75f JB |
4477 | for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) { |
4478 | if (rb->pg_vec == NULL) | |
4479 | continue; | |
4480 | ||
4481 | for (i = 0; i < rb->pg_vec_len; i++) { | |
0e3125c7 NH |
4482 | struct page *page; |
4483 | void *kaddr = rb->pg_vec[i].buffer; | |
69e3c75f JB |
4484 | int pg_num; |
4485 | ||
c56b4d90 CG |
4486 | for (pg_num = 0; pg_num < rb->pg_vec_pages; pg_num++) { |
4487 | page = pgv_to_page(kaddr); | |
69e3c75f JB |
4488 | err = vm_insert_page(vma, start, page); |
4489 | if (unlikely(err)) | |
4490 | goto out; | |
4491 | start += PAGE_SIZE; | |
0e3125c7 | 4492 | kaddr += PAGE_SIZE; |
69e3c75f | 4493 | } |
4ebf0ae2 | 4494 | } |
1da177e4 | 4495 | } |
69e3c75f | 4496 | |
4ebf0ae2 | 4497 | atomic_inc(&po->mapped); |
1da177e4 LT |
4498 | vma->vm_ops = &packet_mmap_ops; |
4499 | err = 0; | |
4500 | ||
4501 | out: | |
905db440 | 4502 | mutex_unlock(&po->pg_vec_lock); |
1da177e4 LT |
4503 | return err; |
4504 | } | |
1da177e4 | 4505 | |
90ddc4f0 | 4506 | static const struct proto_ops packet_ops_spkt = { |
1da177e4 LT |
4507 | .family = PF_PACKET, |
4508 | .owner = THIS_MODULE, | |
4509 | .release = packet_release, | |
4510 | .bind = packet_bind_spkt, | |
4511 | .connect = sock_no_connect, | |
4512 | .socketpair = sock_no_socketpair, | |
4513 | .accept = sock_no_accept, | |
4514 | .getname = packet_getname_spkt, | |
a11e1d43 | 4515 | .poll = datagram_poll, |
1da177e4 | 4516 | .ioctl = packet_ioctl, |
c7cbdbf2 | 4517 | .gettstamp = sock_gettstamp, |
1da177e4 LT |
4518 | .listen = sock_no_listen, |
4519 | .shutdown = sock_no_shutdown, | |
1da177e4 LT |
4520 | .sendmsg = packet_sendmsg_spkt, |
4521 | .recvmsg = packet_recvmsg, | |
4522 | .mmap = sock_no_mmap, | |
4523 | .sendpage = sock_no_sendpage, | |
4524 | }; | |
1da177e4 | 4525 | |
90ddc4f0 | 4526 | static const struct proto_ops packet_ops = { |
1da177e4 LT |
4527 | .family = PF_PACKET, |
4528 | .owner = THIS_MODULE, | |
4529 | .release = packet_release, | |
4530 | .bind = packet_bind, | |
4531 | .connect = sock_no_connect, | |
4532 | .socketpair = sock_no_socketpair, | |
4533 | .accept = sock_no_accept, | |
1ce4f28b | 4534 | .getname = packet_getname, |
a11e1d43 | 4535 | .poll = packet_poll, |
1da177e4 | 4536 | .ioctl = packet_ioctl, |
c7cbdbf2 | 4537 | .gettstamp = sock_gettstamp, |
1da177e4 LT |
4538 | .listen = sock_no_listen, |
4539 | .shutdown = sock_no_shutdown, | |
4540 | .setsockopt = packet_setsockopt, | |
4541 | .getsockopt = packet_getsockopt, | |
4542 | .sendmsg = packet_sendmsg, | |
4543 | .recvmsg = packet_recvmsg, | |
4544 | .mmap = packet_mmap, | |
4545 | .sendpage = sock_no_sendpage, | |
4546 | }; | |
4547 | ||
ec1b4cf7 | 4548 | static const struct net_proto_family packet_family_ops = { |
1da177e4 LT |
4549 | .family = PF_PACKET, |
4550 | .create = packet_create, | |
4551 | .owner = THIS_MODULE, | |
4552 | }; | |
4553 | ||
4554 | static struct notifier_block packet_netdev_notifier = { | |
40d4e3df | 4555 | .notifier_call = packet_notifier, |
1da177e4 LT |
4556 | }; |
4557 | ||
4558 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
4559 | |
4560 | static void *packet_seq_start(struct seq_file *seq, loff_t *pos) | |
808f5114 | 4561 | __acquires(RCU) |
1da177e4 | 4562 | { |
e372c414 | 4563 | struct net *net = seq_file_net(seq); |
808f5114 | 4564 | |
4565 | rcu_read_lock(); | |
4566 | return seq_hlist_start_head_rcu(&net->packet.sklist, *pos); | |
1da177e4 LT |
4567 | } |
4568 | ||
4569 | static void *packet_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
4570 | { | |
1bf40954 | 4571 | struct net *net = seq_file_net(seq); |
808f5114 | 4572 | return seq_hlist_next_rcu(v, &net->packet.sklist, pos); |
1da177e4 LT |
4573 | } |
4574 | ||
4575 | static void packet_seq_stop(struct seq_file *seq, void *v) | |
808f5114 | 4576 | __releases(RCU) |
1da177e4 | 4577 | { |
808f5114 | 4578 | rcu_read_unlock(); |
1da177e4 LT |
4579 | } |
4580 | ||
1ce4f28b | 4581 | static int packet_seq_show(struct seq_file *seq, void *v) |
1da177e4 LT |
4582 | { |
4583 | if (v == SEQ_START_TOKEN) | |
4584 | seq_puts(seq, "sk RefCnt Type Proto Iface R Rmem User Inode\n"); | |
4585 | else { | |
b7ceabd9 | 4586 | struct sock *s = sk_entry(v); |
1da177e4 LT |
4587 | const struct packet_sock *po = pkt_sk(s); |
4588 | ||
4589 | seq_printf(seq, | |
71338aa7 | 4590 | "%pK %-6d %-4d %04x %-5d %1d %-6u %-6u %-6lu\n", |
1da177e4 | 4591 | s, |
41c6d650 | 4592 | refcount_read(&s->sk_refcnt), |
1da177e4 LT |
4593 | s->sk_type, |
4594 | ntohs(po->num), | |
4595 | po->ifindex, | |
4596 | po->running, | |
4597 | atomic_read(&s->sk_rmem_alloc), | |
a7cb5a49 | 4598 | from_kuid_munged(seq_user_ns(seq), sock_i_uid(s)), |
40d4e3df | 4599 | sock_i_ino(s)); |
1da177e4 LT |
4600 | } |
4601 | ||
4602 | return 0; | |
4603 | } | |
4604 | ||
56b3d975 | 4605 | static const struct seq_operations packet_seq_ops = { |
1da177e4 LT |
4606 | .start = packet_seq_start, |
4607 | .next = packet_seq_next, | |
4608 | .stop = packet_seq_stop, | |
4609 | .show = packet_seq_show, | |
4610 | }; | |
1da177e4 LT |
4611 | #endif |
4612 | ||
2c8c1e72 | 4613 | static int __net_init packet_net_init(struct net *net) |
d12d01d6 | 4614 | { |
0fa7fa98 | 4615 | mutex_init(&net->packet.sklist_lock); |
2aaef4e4 | 4616 | INIT_HLIST_HEAD(&net->packet.sklist); |
d12d01d6 | 4617 | |
c3506372 CH |
4618 | if (!proc_create_net("packet", 0, net->proc_net, &packet_seq_ops, |
4619 | sizeof(struct seq_net_private))) | |
d12d01d6 DL |
4620 | return -ENOMEM; |
4621 | ||
4622 | return 0; | |
4623 | } | |
4624 | ||
2c8c1e72 | 4625 | static void __net_exit packet_net_exit(struct net *net) |
d12d01d6 | 4626 | { |
ece31ffd | 4627 | remove_proc_entry("packet", net->proc_net); |
669f8f1a | 4628 | WARN_ON_ONCE(!hlist_empty(&net->packet.sklist)); |
d12d01d6 DL |
4629 | } |
4630 | ||
4631 | static struct pernet_operations packet_net_ops = { | |
4632 | .init = packet_net_init, | |
4633 | .exit = packet_net_exit, | |
4634 | }; | |
4635 | ||
4636 | ||
1da177e4 LT |
4637 | static void __exit packet_exit(void) |
4638 | { | |
1da177e4 | 4639 | unregister_netdevice_notifier(&packet_netdev_notifier); |
d12d01d6 | 4640 | unregister_pernet_subsys(&packet_net_ops); |
1da177e4 LT |
4641 | sock_unregister(PF_PACKET); |
4642 | proto_unregister(&packet_proto); | |
4643 | } | |
4644 | ||
4645 | static int __init packet_init(void) | |
4646 | { | |
36096f2f | 4647 | int rc; |
1da177e4 | 4648 | |
36096f2f Y |
4649 | rc = proto_register(&packet_proto, 0); |
4650 | if (rc) | |
1da177e4 | 4651 | goto out; |
36096f2f Y |
4652 | rc = sock_register(&packet_family_ops); |
4653 | if (rc) | |
4654 | goto out_proto; | |
4655 | rc = register_pernet_subsys(&packet_net_ops); | |
4656 | if (rc) | |
4657 | goto out_sock; | |
4658 | rc = register_netdevice_notifier(&packet_netdev_notifier); | |
4659 | if (rc) | |
4660 | goto out_pernet; | |
1da177e4 | 4661 | |
36096f2f Y |
4662 | return 0; |
4663 | ||
4664 | out_pernet: | |
4665 | unregister_pernet_subsys(&packet_net_ops); | |
4666 | out_sock: | |
4667 | sock_unregister(PF_PACKET); | |
4668 | out_proto: | |
4669 | proto_unregister(&packet_proto); | |
1da177e4 LT |
4670 | out: |
4671 | return rc; | |
4672 | } | |
4673 | ||
4674 | module_init(packet_init); | |
4675 | module_exit(packet_exit); | |
4676 | MODULE_LICENSE("GPL"); | |
4677 | MODULE_ALIAS_NETPROTO(PF_PACKET); |