]>
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 | |
dc41c4a9 BS |
49 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
50 | ||
cc69837f | 51 | #include <linux/ethtool.h> |
b6459415 | 52 | #include <linux/filter.h> |
1da177e4 | 53 | #include <linux/types.h> |
1da177e4 | 54 | #include <linux/mm.h> |
4fc268d2 | 55 | #include <linux/capability.h> |
1da177e4 LT |
56 | #include <linux/fcntl.h> |
57 | #include <linux/socket.h> | |
58 | #include <linux/in.h> | |
59 | #include <linux/inet.h> | |
60 | #include <linux/netdevice.h> | |
61 | #include <linux/if_packet.h> | |
62 | #include <linux/wireless.h> | |
ffbc6111 | 63 | #include <linux/kernel.h> |
1da177e4 | 64 | #include <linux/kmod.h> |
5a0e3ad6 | 65 | #include <linux/slab.h> |
0e3125c7 | 66 | #include <linux/vmalloc.h> |
457c4cbc | 67 | #include <net/net_namespace.h> |
1da177e4 LT |
68 | #include <net/ip.h> |
69 | #include <net/protocol.h> | |
70 | #include <linux/skbuff.h> | |
71 | #include <net/sock.h> | |
72 | #include <linux/errno.h> | |
73 | #include <linux/timer.h> | |
7c0f6ba6 | 74 | #include <linux/uaccess.h> |
1da177e4 LT |
75 | #include <asm/ioctls.h> |
76 | #include <asm/page.h> | |
a1f8e7f7 | 77 | #include <asm/cacheflush.h> |
1da177e4 LT |
78 | #include <asm/io.h> |
79 | #include <linux/proc_fs.h> | |
80 | #include <linux/seq_file.h> | |
81 | #include <linux/poll.h> | |
82 | #include <linux/module.h> | |
83 | #include <linux/init.h> | |
905db440 | 84 | #include <linux/mutex.h> |
05423b24 | 85 | #include <linux/if_vlan.h> |
bfd5f4a3 | 86 | #include <linux/virtio_net.h> |
ed85b565 | 87 | #include <linux/errqueue.h> |
614f60fa | 88 | #include <linux/net_tstamp.h> |
b0138408 | 89 | #include <linux/percpu.h> |
1da177e4 LT |
90 | #ifdef CONFIG_INET |
91 | #include <net/inet_common.h> | |
92 | #endif | |
47dceb8e | 93 | #include <linux/bpf.h> |
719c44d3 | 94 | #include <net/compat.h> |
0d7308c0 | 95 | #include <linux/netfilter_netdev.h> |
1da177e4 | 96 | |
2787b04b PE |
97 | #include "internal.h" |
98 | ||
1da177e4 LT |
99 | /* |
100 | Assumptions: | |
d5496990 EB |
101 | - If the device has no dev->header_ops->create, there is no LL header |
102 | visible above the device. In this case, its hard_header_len should be 0. | |
b4c58814 XH |
103 | The device may prepend its own header internally. In this case, its |
104 | needed_headroom should be set to the space needed for it to add its | |
105 | internal header. | |
106 | For example, a WiFi driver pretending to be an Ethernet driver should | |
107 | set its hard_header_len to be the Ethernet header length, and set its | |
108 | needed_headroom to be (the real WiFi header length - the fake Ethernet | |
109 | header length). | |
1da177e4 LT |
110 | - packet socket receives packets with pulled ll header, |
111 | so that SOCK_RAW should push it back. | |
112 | ||
113 | On receive: | |
114 | ----------- | |
115 | ||
d5496990 | 116 | Incoming, dev_has_header(dev) == true |
b0e380b1 ACM |
117 | mac_header -> ll header |
118 | data -> data | |
1da177e4 | 119 | |
d5496990 | 120 | Outgoing, dev_has_header(dev) == true |
b0e380b1 ACM |
121 | mac_header -> ll header |
122 | data -> ll header | |
1da177e4 | 123 | |
d5496990 | 124 | Incoming, dev_has_header(dev) == false |
b79a80bd XH |
125 | mac_header -> data |
126 | However drivers often make it point to the ll header. | |
127 | This is incorrect because the ll header should be invisible to us. | |
b0e380b1 | 128 | data -> data |
1da177e4 | 129 | |
d5496990 | 130 | Outgoing, dev_has_header(dev) == false |
b79a80bd | 131 | mac_header -> data. ll header is invisible to us. |
b0e380b1 | 132 | data -> data |
1da177e4 LT |
133 | |
134 | Resume | |
d5496990 | 135 | If dev_has_header(dev) == false we are unable to restore the ll header, |
b79a80bd | 136 | because it is invisible to us. |
1da177e4 LT |
137 | |
138 | ||
139 | On transmit: | |
140 | ------------ | |
141 | ||
21c85974 | 142 | dev_has_header(dev) == true |
b0e380b1 ACM |
143 | mac_header -> ll header |
144 | data -> ll header | |
1da177e4 | 145 | |
21c85974 | 146 | dev_has_header(dev) == false (ll header is invisible to us) |
b0e380b1 ACM |
147 | mac_header -> data |
148 | data -> data | |
1da177e4 | 149 | |
09599729 | 150 | We should set network_header on output to the correct position, |
1da177e4 LT |
151 | packet classifier depends on it. |
152 | */ | |
153 | ||
1da177e4 LT |
154 | /* Private packet socket structures. */ |
155 | ||
0fb375fb EB |
156 | /* identical to struct packet_mreq except it has |
157 | * a longer address field. | |
158 | */ | |
40d4e3df | 159 | struct packet_mreq_max { |
0fb375fb EB |
160 | int mr_ifindex; |
161 | unsigned short mr_type; | |
162 | unsigned short mr_alen; | |
163 | unsigned char mr_address[MAX_ADDR_LEN]; | |
1da177e4 | 164 | }; |
a2efcfa0 | 165 | |
184f489e DB |
166 | union tpacket_uhdr { |
167 | struct tpacket_hdr *h1; | |
168 | struct tpacket2_hdr *h2; | |
169 | struct tpacket3_hdr *h3; | |
170 | void *raw; | |
171 | }; | |
172 | ||
f6fb8f10 | 173 | static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u, |
69e3c75f JB |
174 | int closing, int tx_ring); |
175 | ||
f6fb8f10 | 176 | #define V3_ALIGNMENT (8) |
177 | ||
bc59ba39 | 178 | #define BLK_HDR_LEN (ALIGN(sizeof(struct tpacket_block_desc), V3_ALIGNMENT)) |
f6fb8f10 | 179 | |
180 | #define BLK_PLUS_PRIV(sz_of_priv) \ | |
181 | (BLK_HDR_LEN + ALIGN((sz_of_priv), V3_ALIGNMENT)) | |
182 | ||
f6fb8f10 | 183 | #define BLOCK_STATUS(x) ((x)->hdr.bh1.block_status) |
184 | #define BLOCK_NUM_PKTS(x) ((x)->hdr.bh1.num_pkts) | |
185 | #define BLOCK_O2FP(x) ((x)->hdr.bh1.offset_to_first_pkt) | |
186 | #define BLOCK_LEN(x) ((x)->hdr.bh1.blk_len) | |
187 | #define BLOCK_SNUM(x) ((x)->hdr.bh1.seq_num) | |
188 | #define BLOCK_O2PRIV(x) ((x)->offset_to_priv) | |
f6fb8f10 | 189 | |
69e3c75f | 190 | struct packet_sock; |
77f65ebd WB |
191 | static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, |
192 | struct packet_type *pt, struct net_device *orig_dev); | |
1da177e4 | 193 | |
f6fb8f10 | 194 | static void *packet_previous_frame(struct packet_sock *po, |
195 | struct packet_ring_buffer *rb, | |
196 | int status); | |
197 | static void packet_increment_head(struct packet_ring_buffer *buff); | |
878cd3ba | 198 | static int prb_curr_blk_in_use(struct tpacket_block_desc *); |
bc59ba39 | 199 | static void *prb_dispatch_next_block(struct tpacket_kbdq_core *, |
f6fb8f10 | 200 | struct packet_sock *); |
bc59ba39 | 201 | static void prb_retire_current_block(struct tpacket_kbdq_core *, |
f6fb8f10 | 202 | struct packet_sock *, unsigned int status); |
bc59ba39 | 203 | static int prb_queue_frozen(struct tpacket_kbdq_core *); |
204 | static void prb_open_block(struct tpacket_kbdq_core *, | |
205 | struct tpacket_block_desc *); | |
17bfd8c8 | 206 | static void prb_retire_rx_blk_timer_expired(struct timer_list *); |
bc59ba39 | 207 | static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *); |
bc59ba39 | 208 | static void prb_fill_rxhash(struct tpacket_kbdq_core *, struct tpacket3_hdr *); |
209 | static void prb_clear_rxhash(struct tpacket_kbdq_core *, | |
210 | struct tpacket3_hdr *); | |
211 | static void prb_fill_vlan_info(struct tpacket_kbdq_core *, | |
212 | struct tpacket3_hdr *); | |
1da177e4 | 213 | static void packet_flush_mclist(struct sock *sk); |
865b03f2 | 214 | static u16 packet_pick_tx_queue(struct sk_buff *skb); |
1da177e4 | 215 | |
ffbc6111 | 216 | struct packet_skb_cb { |
ffbc6111 HX |
217 | union { |
218 | struct sockaddr_pkt pkt; | |
2472d761 EB |
219 | union { |
220 | /* Trick: alias skb original length with | |
221 | * ll.sll_family and ll.protocol in order | |
222 | * to save room. | |
223 | */ | |
224 | unsigned int origlen; | |
225 | struct sockaddr_ll ll; | |
226 | }; | |
ffbc6111 HX |
227 | } sa; |
228 | }; | |
229 | ||
d3869efe DW |
230 | #define vio_le() virtio_legacy_is_little_endian() |
231 | ||
ffbc6111 | 232 | #define PACKET_SKB_CB(__skb) ((struct packet_skb_cb *)((__skb)->cb)) |
8dc41944 | 233 | |
bc59ba39 | 234 | #define GET_PBDQC_FROM_RB(x) ((struct tpacket_kbdq_core *)(&(x)->prb_bdqc)) |
f6fb8f10 | 235 | #define GET_PBLOCK_DESC(x, bid) \ |
bc59ba39 | 236 | ((struct tpacket_block_desc *)((x)->pkbdq[(bid)].buffer)) |
f6fb8f10 | 237 | #define GET_CURR_PBLOCK_DESC_FROM_CORE(x) \ |
bc59ba39 | 238 | ((struct tpacket_block_desc *)((x)->pkbdq[(x)->kactive_blk_num].buffer)) |
f6fb8f10 | 239 | #define GET_NEXT_PRB_BLK_NUM(x) \ |
240 | (((x)->kactive_blk_num < ((x)->knum_blocks-1)) ? \ | |
241 | ((x)->kactive_blk_num+1) : 0) | |
242 | ||
dc99f600 DM |
243 | static void __fanout_unlink(struct sock *sk, struct packet_sock *po); |
244 | static void __fanout_link(struct sock *sk, struct packet_sock *po); | |
245 | ||
0d7308c0 PNA |
246 | #ifdef CONFIG_NETFILTER_EGRESS |
247 | static noinline struct sk_buff *nf_hook_direct_egress(struct sk_buff *skb) | |
248 | { | |
249 | struct sk_buff *next, *head = NULL, *tail; | |
250 | int rc; | |
251 | ||
252 | rcu_read_lock(); | |
253 | for (; skb != NULL; skb = next) { | |
254 | next = skb->next; | |
255 | skb_mark_not_on_list(skb); | |
256 | ||
257 | if (!nf_hook_egress(skb, &rc, skb->dev)) | |
258 | continue; | |
259 | ||
260 | if (!head) | |
261 | head = skb; | |
262 | else | |
263 | tail->next = skb; | |
264 | ||
265 | tail = skb; | |
266 | } | |
267 | rcu_read_unlock(); | |
268 | ||
269 | return head; | |
270 | } | |
271 | #endif | |
272 | ||
105a201e | 273 | static int packet_xmit(const struct packet_sock *po, struct sk_buff *skb) |
d346a3fa | 274 | { |
105a201e ED |
275 | if (!packet_sock_flag(po, PACKET_SOCK_QDISC_BYPASS)) |
276 | return dev_queue_xmit(skb); | |
277 | ||
0d7308c0 PNA |
278 | #ifdef CONFIG_NETFILTER_EGRESS |
279 | if (nf_hook_egress_active()) { | |
280 | skb = nf_hook_direct_egress(skb); | |
281 | if (!skb) | |
282 | return NET_XMIT_DROP; | |
283 | } | |
284 | #endif | |
865b03f2 | 285 | return dev_direct_xmit(skb, packet_pick_tx_queue(skb)); |
d346a3fa DB |
286 | } |
287 | ||
66e56cd4 DB |
288 | static struct net_device *packet_cached_dev_get(struct packet_sock *po) |
289 | { | |
290 | struct net_device *dev; | |
291 | ||
292 | rcu_read_lock(); | |
293 | dev = rcu_dereference(po->cached_dev); | |
1160dfa1 | 294 | dev_hold(dev); |
66e56cd4 DB |
295 | rcu_read_unlock(); |
296 | ||
297 | return dev; | |
298 | } | |
299 | ||
300 | static void packet_cached_dev_assign(struct packet_sock *po, | |
301 | struct net_device *dev) | |
302 | { | |
303 | rcu_assign_pointer(po->cached_dev, dev); | |
304 | } | |
305 | ||
306 | static void packet_cached_dev_reset(struct packet_sock *po) | |
307 | { | |
308 | RCU_INIT_POINTER(po->cached_dev, NULL); | |
309 | } | |
310 | ||
865b03f2 | 311 | static u16 packet_pick_tx_queue(struct sk_buff *skb) |
0fd5d57b | 312 | { |
865b03f2 | 313 | struct net_device *dev = skb->dev; |
0fd5d57b | 314 | const struct net_device_ops *ops = dev->netdev_ops; |
b71b5837 | 315 | int cpu = raw_smp_processor_id(); |
0fd5d57b DB |
316 | u16 queue_index; |
317 | ||
b71b5837 PA |
318 | #ifdef CONFIG_XPS |
319 | skb->sender_cpu = cpu + 1; | |
320 | #endif | |
321 | skb_record_rx_queue(skb, cpu % dev->real_num_tx_queues); | |
0fd5d57b | 322 | if (ops->ndo_select_queue) { |
a350ecce | 323 | queue_index = ops->ndo_select_queue(dev, skb, NULL); |
0fd5d57b DB |
324 | queue_index = netdev_cap_txqueue(dev, queue_index); |
325 | } else { | |
b71b5837 | 326 | queue_index = netdev_pick_tx(dev, skb, NULL); |
0fd5d57b DB |
327 | } |
328 | ||
865b03f2 | 329 | return queue_index; |
0fd5d57b DB |
330 | } |
331 | ||
a6361f0c | 332 | /* __register_prot_hook must be invoked through register_prot_hook |
ce06b03e DM |
333 | * or from a context in which asynchronous accesses to the packet |
334 | * socket is not possible (packet_create()). | |
335 | */ | |
a6361f0c | 336 | static void __register_prot_hook(struct sock *sk) |
ce06b03e DM |
337 | { |
338 | struct packet_sock *po = pkt_sk(sk); | |
e40526cb | 339 | |
61edf479 | 340 | if (!packet_sock_flag(po, PACKET_SOCK_RUNNING)) { |
66e56cd4 | 341 | if (po->fanout) |
dc99f600 | 342 | __fanout_link(sk, po); |
66e56cd4 | 343 | else |
dc99f600 | 344 | dev_add_pack(&po->prot_hook); |
e40526cb | 345 | |
ce06b03e | 346 | sock_hold(sk); |
61edf479 | 347 | packet_sock_flag_set(po, PACKET_SOCK_RUNNING, 1); |
ce06b03e DM |
348 | } |
349 | } | |
350 | ||
a6361f0c WB |
351 | static void register_prot_hook(struct sock *sk) |
352 | { | |
353 | lockdep_assert_held_once(&pkt_sk(sk)->bind_lock); | |
354 | __register_prot_hook(sk); | |
355 | } | |
356 | ||
357 | /* If the sync parameter is true, we will temporarily drop | |
ce06b03e DM |
358 | * the po->bind_lock and do a synchronize_net to make sure no |
359 | * asynchronous packet processing paths still refer to the elements | |
360 | * of po->prot_hook. If the sync parameter is false, it is the | |
361 | * callers responsibility to take care of this. | |
362 | */ | |
363 | static void __unregister_prot_hook(struct sock *sk, bool sync) | |
364 | { | |
365 | struct packet_sock *po = pkt_sk(sk); | |
366 | ||
a6361f0c WB |
367 | lockdep_assert_held_once(&po->bind_lock); |
368 | ||
61edf479 | 369 | packet_sock_flag_set(po, PACKET_SOCK_RUNNING, 0); |
66e56cd4 DB |
370 | |
371 | if (po->fanout) | |
dc99f600 | 372 | __fanout_unlink(sk, po); |
66e56cd4 | 373 | else |
dc99f600 | 374 | __dev_remove_pack(&po->prot_hook); |
e40526cb | 375 | |
ce06b03e DM |
376 | __sock_put(sk); |
377 | ||
378 | if (sync) { | |
379 | spin_unlock(&po->bind_lock); | |
380 | synchronize_net(); | |
381 | spin_lock(&po->bind_lock); | |
382 | } | |
383 | } | |
384 | ||
385 | static void unregister_prot_hook(struct sock *sk, bool sync) | |
386 | { | |
387 | struct packet_sock *po = pkt_sk(sk); | |
388 | ||
61edf479 | 389 | if (packet_sock_flag(po, PACKET_SOCK_RUNNING)) |
ce06b03e DM |
390 | __unregister_prot_hook(sk, sync); |
391 | } | |
392 | ||
6e58040b | 393 | static inline struct page * __pure pgv_to_page(void *addr) |
0af55bb5 CG |
394 | { |
395 | if (is_vmalloc_addr(addr)) | |
396 | return vmalloc_to_page(addr); | |
397 | return virt_to_page(addr); | |
398 | } | |
399 | ||
69e3c75f | 400 | static void __packet_set_status(struct packet_sock *po, void *frame, int status) |
1da177e4 | 401 | { |
184f489e | 402 | union tpacket_uhdr h; |
1da177e4 | 403 | |
8a989617 ED |
404 | /* WRITE_ONCE() are paired with READ_ONCE() in __packet_get_status */ |
405 | ||
69e3c75f | 406 | h.raw = frame; |
bbd6ef87 PM |
407 | switch (po->tp_version) { |
408 | case TPACKET_V1: | |
8a989617 | 409 | WRITE_ONCE(h.h1->tp_status, status); |
0af55bb5 | 410 | flush_dcache_page(pgv_to_page(&h.h1->tp_status)); |
bbd6ef87 PM |
411 | break; |
412 | case TPACKET_V2: | |
8a989617 | 413 | WRITE_ONCE(h.h2->tp_status, status); |
0af55bb5 | 414 | flush_dcache_page(pgv_to_page(&h.h2->tp_status)); |
bbd6ef87 | 415 | break; |
f6fb8f10 | 416 | case TPACKET_V3: |
8a989617 | 417 | WRITE_ONCE(h.h3->tp_status, status); |
7f953ab2 SV |
418 | flush_dcache_page(pgv_to_page(&h.h3->tp_status)); |
419 | break; | |
69e3c75f | 420 | default: |
f6fb8f10 | 421 | WARN(1, "TPACKET version not supported.\n"); |
69e3c75f | 422 | BUG(); |
bbd6ef87 | 423 | } |
69e3c75f JB |
424 | |
425 | smp_wmb(); | |
bbd6ef87 PM |
426 | } |
427 | ||
96f657e6 | 428 | static int __packet_get_status(const struct packet_sock *po, void *frame) |
bbd6ef87 | 429 | { |
184f489e | 430 | union tpacket_uhdr h; |
bbd6ef87 | 431 | |
69e3c75f JB |
432 | smp_rmb(); |
433 | ||
8a989617 ED |
434 | /* READ_ONCE() are paired with WRITE_ONCE() in __packet_set_status */ |
435 | ||
bbd6ef87 PM |
436 | h.raw = frame; |
437 | switch (po->tp_version) { | |
438 | case TPACKET_V1: | |
0af55bb5 | 439 | flush_dcache_page(pgv_to_page(&h.h1->tp_status)); |
8a989617 | 440 | return READ_ONCE(h.h1->tp_status); |
bbd6ef87 | 441 | case TPACKET_V2: |
0af55bb5 | 442 | flush_dcache_page(pgv_to_page(&h.h2->tp_status)); |
8a989617 | 443 | return READ_ONCE(h.h2->tp_status); |
f6fb8f10 | 444 | case TPACKET_V3: |
7f953ab2 | 445 | flush_dcache_page(pgv_to_page(&h.h3->tp_status)); |
8a989617 | 446 | return READ_ONCE(h.h3->tp_status); |
69e3c75f | 447 | default: |
f6fb8f10 | 448 | WARN(1, "TPACKET version not supported.\n"); |
69e3c75f JB |
449 | BUG(); |
450 | return 0; | |
bbd6ef87 | 451 | } |
1da177e4 | 452 | } |
69e3c75f | 453 | |
d413fcb4 | 454 | static __u32 tpacket_get_timestamp(struct sk_buff *skb, struct timespec64 *ts, |
b9c32fb2 | 455 | unsigned int flags) |
7a51384c DB |
456 | { |
457 | struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb); | |
458 | ||
68a360e8 WB |
459 | if (shhwtstamps && |
460 | (flags & SOF_TIMESTAMPING_RAW_HARDWARE) && | |
d413fcb4 | 461 | ktime_to_timespec64_cond(shhwtstamps->hwtstamp, ts)) |
68a360e8 | 462 | return TP_STATUS_TS_RAW_HARDWARE; |
7a51384c | 463 | |
171c3b15 | 464 | if ((flags & SOF_TIMESTAMPING_SOFTWARE) && |
27942a15 | 465 | ktime_to_timespec64_cond(skb_tstamp(skb), ts)) |
b9c32fb2 | 466 | return TP_STATUS_TS_SOFTWARE; |
7a51384c | 467 | |
b9c32fb2 | 468 | return 0; |
7a51384c DB |
469 | } |
470 | ||
b9c32fb2 DB |
471 | static __u32 __packet_set_timestamp(struct packet_sock *po, void *frame, |
472 | struct sk_buff *skb) | |
2e31396f WB |
473 | { |
474 | union tpacket_uhdr h; | |
d413fcb4 | 475 | struct timespec64 ts; |
b9c32fb2 | 476 | __u32 ts_status; |
2e31396f | 477 | |
1051ce4a | 478 | if (!(ts_status = tpacket_get_timestamp(skb, &ts, READ_ONCE(po->tp_tstamp)))) |
b9c32fb2 | 479 | return 0; |
2e31396f WB |
480 | |
481 | h.raw = frame; | |
d413fcb4 AB |
482 | /* |
483 | * versions 1 through 3 overflow the timestamps in y2106, since they | |
484 | * all store the seconds in a 32-bit unsigned integer. | |
485 | * If we create a version 4, that should have a 64-bit timestamp, | |
486 | * either 64-bit seconds + 32-bit nanoseconds, or just 64-bit | |
487 | * nanoseconds. | |
488 | */ | |
2e31396f WB |
489 | switch (po->tp_version) { |
490 | case TPACKET_V1: | |
491 | h.h1->tp_sec = ts.tv_sec; | |
492 | h.h1->tp_usec = ts.tv_nsec / NSEC_PER_USEC; | |
493 | break; | |
494 | case TPACKET_V2: | |
495 | h.h2->tp_sec = ts.tv_sec; | |
496 | h.h2->tp_nsec = ts.tv_nsec; | |
497 | break; | |
498 | case TPACKET_V3: | |
57ea884b DB |
499 | h.h3->tp_sec = ts.tv_sec; |
500 | h.h3->tp_nsec = ts.tv_nsec; | |
501 | break; | |
2e31396f WB |
502 | default: |
503 | WARN(1, "TPACKET version not supported.\n"); | |
504 | BUG(); | |
505 | } | |
506 | ||
507 | /* one flush is safe, as both fields always lie on the same cacheline */ | |
508 | flush_dcache_page(pgv_to_page(&h.h1->tp_sec)); | |
509 | smp_wmb(); | |
b9c32fb2 DB |
510 | |
511 | return ts_status; | |
2e31396f WB |
512 | } |
513 | ||
d4b5bd98 ED |
514 | static void *packet_lookup_frame(const struct packet_sock *po, |
515 | const struct packet_ring_buffer *rb, | |
516 | unsigned int position, | |
517 | int status) | |
69e3c75f JB |
518 | { |
519 | unsigned int pg_vec_pos, frame_offset; | |
184f489e | 520 | union tpacket_uhdr h; |
69e3c75f JB |
521 | |
522 | pg_vec_pos = position / rb->frames_per_block; | |
523 | frame_offset = position % rb->frames_per_block; | |
524 | ||
0e3125c7 NH |
525 | h.raw = rb->pg_vec[pg_vec_pos].buffer + |
526 | (frame_offset * rb->frame_size); | |
69e3c75f JB |
527 | |
528 | if (status != __packet_get_status(po, h.raw)) | |
529 | return NULL; | |
530 | ||
531 | return h.raw; | |
532 | } | |
533 | ||
eea49cc9 | 534 | static void *packet_current_frame(struct packet_sock *po, |
69e3c75f JB |
535 | struct packet_ring_buffer *rb, |
536 | int status) | |
537 | { | |
538 | return packet_lookup_frame(po, rb, rb->head, status); | |
539 | } | |
540 | ||
bc59ba39 | 541 | static void prb_del_retire_blk_timer(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 542 | { |
543 | del_timer_sync(&pkc->retire_blk_timer); | |
544 | } | |
545 | ||
546 | static void prb_shutdown_retire_blk_timer(struct packet_sock *po, | |
f6fb8f10 | 547 | struct sk_buff_head *rb_queue) |
548 | { | |
bc59ba39 | 549 | struct tpacket_kbdq_core *pkc; |
f6fb8f10 | 550 | |
73d0fcf2 | 551 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 552 | |
ec6f809f | 553 | spin_lock_bh(&rb_queue->lock); |
f6fb8f10 | 554 | pkc->delete_blk_timer = 1; |
ec6f809f | 555 | spin_unlock_bh(&rb_queue->lock); |
f6fb8f10 | 556 | |
557 | prb_del_retire_blk_timer(pkc); | |
558 | } | |
559 | ||
e8e85cc5 | 560 | static void prb_setup_retire_blk_timer(struct packet_sock *po) |
f6fb8f10 | 561 | { |
bc59ba39 | 562 | struct tpacket_kbdq_core *pkc; |
f6fb8f10 | 563 | |
e8e85cc5 | 564 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
17bfd8c8 KC |
565 | timer_setup(&pkc->retire_blk_timer, prb_retire_rx_blk_timer_expired, |
566 | 0); | |
567 | pkc->retire_blk_timer.expires = jiffies; | |
f6fb8f10 | 568 | } |
569 | ||
570 | static int prb_calc_retire_blk_tmo(struct packet_sock *po, | |
571 | int blk_size_in_bytes) | |
572 | { | |
573 | struct net_device *dev; | |
0914d2bb | 574 | unsigned int mbits, div; |
7cad1bac | 575 | struct ethtool_link_ksettings ecmd; |
4bc71cb9 | 576 | int err; |
f6fb8f10 | 577 | |
4bc71cb9 JP |
578 | rtnl_lock(); |
579 | dev = __dev_get_by_index(sock_net(&po->sk), po->ifindex); | |
580 | if (unlikely(!dev)) { | |
581 | rtnl_unlock(); | |
f6fb8f10 | 582 | return DEFAULT_PRB_RETIRE_TOV; |
4bc71cb9 | 583 | } |
7cad1bac | 584 | err = __ethtool_get_link_ksettings(dev, &ecmd); |
4bc71cb9 | 585 | rtnl_unlock(); |
0914d2bb | 586 | if (err) |
b43d1f9f | 587 | return DEFAULT_PRB_RETIRE_TOV; |
f6fb8f10 | 588 | |
0914d2bb MW |
589 | /* If the link speed is so slow you don't really |
590 | * need to worry about perf anyways | |
591 | */ | |
592 | if (ecmd.base.speed < SPEED_1000 || | |
593 | ecmd.base.speed == SPEED_UNKNOWN) | |
594 | return DEFAULT_PRB_RETIRE_TOV; | |
595 | ||
596 | div = ecmd.base.speed / 1000; | |
f6fb8f10 | 597 | mbits = (blk_size_in_bytes * 8) / (1024 * 1024); |
598 | ||
599 | if (div) | |
600 | mbits /= div; | |
601 | ||
f6fb8f10 | 602 | if (div) |
0914d2bb MW |
603 | return mbits + 1; |
604 | return mbits; | |
f6fb8f10 | 605 | } |
606 | ||
bc59ba39 | 607 | static void prb_init_ft_ops(struct tpacket_kbdq_core *p1, |
f6fb8f10 | 608 | union tpacket_req_u *req_u) |
609 | { | |
610 | p1->feature_req_word = req_u->req3.tp_feature_req_word; | |
611 | } | |
612 | ||
613 | static void init_prb_bdqc(struct packet_sock *po, | |
614 | struct packet_ring_buffer *rb, | |
615 | struct pgv *pg_vec, | |
e8e85cc5 | 616 | union tpacket_req_u *req_u) |
f6fb8f10 | 617 | { |
22781a5b | 618 | struct tpacket_kbdq_core *p1 = GET_PBDQC_FROM_RB(rb); |
bc59ba39 | 619 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 620 | |
621 | memset(p1, 0x0, sizeof(*p1)); | |
622 | ||
623 | p1->knxt_seq_num = 1; | |
624 | p1->pkbdq = pg_vec; | |
bc59ba39 | 625 | pbd = (struct tpacket_block_desc *)pg_vec[0].buffer; |
e3192690 | 626 | p1->pkblk_start = pg_vec[0].buffer; |
f6fb8f10 | 627 | p1->kblk_size = req_u->req3.tp_block_size; |
628 | p1->knum_blocks = req_u->req3.tp_block_nr; | |
629 | p1->hdrlen = po->tp_hdrlen; | |
630 | p1->version = po->tp_version; | |
631 | p1->last_kactive_blk_num = 0; | |
ee80fbf3 | 632 | po->stats.stats3.tp_freeze_q_cnt = 0; |
f6fb8f10 | 633 | if (req_u->req3.tp_retire_blk_tov) |
634 | p1->retire_blk_tov = req_u->req3.tp_retire_blk_tov; | |
635 | else | |
636 | p1->retire_blk_tov = prb_calc_retire_blk_tmo(po, | |
637 | req_u->req3.tp_block_size); | |
638 | p1->tov_in_jiffies = msecs_to_jiffies(p1->retire_blk_tov); | |
639 | p1->blk_sizeof_priv = req_u->req3.tp_sizeof_priv; | |
632ca50f | 640 | rwlock_init(&p1->blk_fill_in_prog_lock); |
f6fb8f10 | 641 | |
dc808110 | 642 | p1->max_frame_len = p1->kblk_size - BLK_PLUS_PRIV(p1->blk_sizeof_priv); |
f6fb8f10 | 643 | prb_init_ft_ops(p1, req_u); |
e8e85cc5 | 644 | prb_setup_retire_blk_timer(po); |
f6fb8f10 | 645 | prb_open_block(p1, pbd); |
646 | } | |
647 | ||
648 | /* Do NOT update the last_blk_num first. | |
649 | * Assumes sk_buff_head lock is held. | |
650 | */ | |
bc59ba39 | 651 | static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 652 | { |
653 | mod_timer(&pkc->retire_blk_timer, | |
654 | jiffies + pkc->tov_in_jiffies); | |
655 | pkc->last_kactive_blk_num = pkc->kactive_blk_num; | |
656 | } | |
657 | ||
658 | /* | |
659 | * Timer logic: | |
660 | * 1) We refresh the timer only when we open a block. | |
661 | * By doing this we don't waste cycles refreshing the timer | |
662 | * on packet-by-packet basis. | |
663 | * | |
664 | * With a 1MB block-size, on a 1Gbps line, it will take | |
665 | * i) ~8 ms to fill a block + ii) memcpy etc. | |
666 | * In this cut we are not accounting for the memcpy time. | |
667 | * | |
668 | * So, if the user sets the 'tmo' to 10ms then the timer | |
669 | * will never fire while the block is still getting filled | |
670 | * (which is what we want). However, the user could choose | |
671 | * to close a block early and that's fine. | |
672 | * | |
673 | * But when the timer does fire, we check whether or not to refresh it. | |
674 | * Since the tmo granularity is in msecs, it is not too expensive | |
675 | * to refresh the timer, lets say every '8' msecs. | |
676 | * Either the user can set the 'tmo' or we can derive it based on | |
677 | * a) line-speed and b) block-size. | |
678 | * prb_calc_retire_blk_tmo() calculates the tmo. | |
679 | * | |
680 | */ | |
17bfd8c8 | 681 | static void prb_retire_rx_blk_timer_expired(struct timer_list *t) |
f6fb8f10 | 682 | { |
17bfd8c8 KC |
683 | struct packet_sock *po = |
684 | from_timer(po, t, rx_ring.prb_bdqc.retire_blk_timer); | |
22781a5b | 685 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 686 | unsigned int frozen; |
bc59ba39 | 687 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 688 | |
689 | spin_lock(&po->sk.sk_receive_queue.lock); | |
690 | ||
691 | frozen = prb_queue_frozen(pkc); | |
692 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
693 | ||
694 | if (unlikely(pkc->delete_blk_timer)) | |
695 | goto out; | |
696 | ||
697 | /* We only need to plug the race when the block is partially filled. | |
698 | * tpacket_rcv: | |
699 | * lock(); increment BLOCK_NUM_PKTS; unlock() | |
700 | * copy_bits() is in progress ... | |
701 | * timer fires on other cpu: | |
702 | * we can't retire the current block because copy_bits | |
703 | * is in progress. | |
704 | * | |
705 | */ | |
706 | if (BLOCK_NUM_PKTS(pbd)) { | |
632ca50f JO |
707 | /* Waiting for skb_copy_bits to finish... */ |
708 | write_lock(&pkc->blk_fill_in_prog_lock); | |
709 | write_unlock(&pkc->blk_fill_in_prog_lock); | |
f6fb8f10 | 710 | } |
711 | ||
712 | if (pkc->last_kactive_blk_num == pkc->kactive_blk_num) { | |
713 | if (!frozen) { | |
41a50d62 AD |
714 | if (!BLOCK_NUM_PKTS(pbd)) { |
715 | /* An empty block. Just refresh the timer. */ | |
716 | goto refresh_timer; | |
717 | } | |
f6fb8f10 | 718 | prb_retire_current_block(pkc, po, TP_STATUS_BLK_TMO); |
719 | if (!prb_dispatch_next_block(pkc, po)) | |
720 | goto refresh_timer; | |
721 | else | |
722 | goto out; | |
723 | } else { | |
724 | /* Case 1. Queue was frozen because user-space was | |
725 | * lagging behind. | |
726 | */ | |
878cd3ba | 727 | if (prb_curr_blk_in_use(pbd)) { |
f6fb8f10 | 728 | /* |
729 | * Ok, user-space is still behind. | |
730 | * So just refresh the timer. | |
731 | */ | |
732 | goto refresh_timer; | |
733 | } else { | |
734 | /* Case 2. queue was frozen,user-space caught up, | |
735 | * now the link went idle && the timer fired. | |
736 | * We don't have a block to close.So we open this | |
737 | * block and restart the timer. | |
738 | * opening a block thaws the queue,restarts timer | |
739 | * Thawing/timer-refresh is a side effect. | |
740 | */ | |
741 | prb_open_block(pkc, pbd); | |
742 | goto out; | |
743 | } | |
744 | } | |
745 | } | |
746 | ||
747 | refresh_timer: | |
748 | _prb_refresh_rx_retire_blk_timer(pkc); | |
749 | ||
750 | out: | |
751 | spin_unlock(&po->sk.sk_receive_queue.lock); | |
752 | } | |
753 | ||
eea49cc9 | 754 | static void prb_flush_block(struct tpacket_kbdq_core *pkc1, |
bc59ba39 | 755 | struct tpacket_block_desc *pbd1, __u32 status) |
f6fb8f10 | 756 | { |
757 | /* Flush everything minus the block header */ | |
758 | ||
759 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
760 | u8 *start, *end; | |
761 | ||
762 | start = (u8 *)pbd1; | |
763 | ||
764 | /* Skip the block header(we know header WILL fit in 4K) */ | |
765 | start += PAGE_SIZE; | |
766 | ||
767 | end = (u8 *)PAGE_ALIGN((unsigned long)pkc1->pkblk_end); | |
768 | for (; start < end; start += PAGE_SIZE) | |
769 | flush_dcache_page(pgv_to_page(start)); | |
770 | ||
771 | smp_wmb(); | |
772 | #endif | |
773 | ||
774 | /* Now update the block status. */ | |
775 | ||
776 | BLOCK_STATUS(pbd1) = status; | |
777 | ||
778 | /* Flush the block header */ | |
779 | ||
780 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 | |
781 | start = (u8 *)pbd1; | |
782 | flush_dcache_page(pgv_to_page(start)); | |
783 | ||
784 | smp_wmb(); | |
785 | #endif | |
786 | } | |
787 | ||
788 | /* | |
789 | * Side effect: | |
790 | * | |
791 | * 1) flush the block | |
792 | * 2) Increment active_blk_num | |
793 | * | |
794 | * Note:We DONT refresh the timer on purpose. | |
795 | * Because almost always the next block will be opened. | |
796 | */ | |
bc59ba39 | 797 | static void prb_close_block(struct tpacket_kbdq_core *pkc1, |
798 | struct tpacket_block_desc *pbd1, | |
f6fb8f10 | 799 | struct packet_sock *po, unsigned int stat) |
800 | { | |
801 | __u32 status = TP_STATUS_USER | stat; | |
802 | ||
803 | struct tpacket3_hdr *last_pkt; | |
bc59ba39 | 804 | struct tpacket_hdr_v1 *h1 = &pbd1->hdr.bh1; |
da413eec | 805 | struct sock *sk = &po->sk; |
f6fb8f10 | 806 | |
8e8e2951 | 807 | if (atomic_read(&po->tp_drops)) |
f6fb8f10 | 808 | status |= TP_STATUS_LOSING; |
809 | ||
810 | last_pkt = (struct tpacket3_hdr *)pkc1->prev; | |
811 | last_pkt->tp_next_offset = 0; | |
812 | ||
813 | /* Get the ts of the last pkt */ | |
814 | if (BLOCK_NUM_PKTS(pbd1)) { | |
815 | h1->ts_last_pkt.ts_sec = last_pkt->tp_sec; | |
816 | h1->ts_last_pkt.ts_nsec = last_pkt->tp_nsec; | |
817 | } else { | |
41a50d62 AD |
818 | /* Ok, we tmo'd - so get the current time. |
819 | * | |
820 | * It shouldn't really happen as we don't close empty | |
821 | * blocks. See prb_retire_rx_blk_timer_expired(). | |
822 | */ | |
d413fcb4 AB |
823 | struct timespec64 ts; |
824 | ktime_get_real_ts64(&ts); | |
f6fb8f10 | 825 | h1->ts_last_pkt.ts_sec = ts.tv_sec; |
826 | h1->ts_last_pkt.ts_nsec = ts.tv_nsec; | |
827 | } | |
828 | ||
829 | smp_wmb(); | |
830 | ||
831 | /* Flush the block */ | |
832 | prb_flush_block(pkc1, pbd1, status); | |
833 | ||
da413eec DC |
834 | sk->sk_data_ready(sk); |
835 | ||
f6fb8f10 | 836 | pkc1->kactive_blk_num = GET_NEXT_PRB_BLK_NUM(pkc1); |
837 | } | |
838 | ||
eea49cc9 | 839 | static void prb_thaw_queue(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 840 | { |
841 | pkc->reset_pending_on_curr_blk = 0; | |
842 | } | |
843 | ||
844 | /* | |
845 | * Side effect of opening a block: | |
846 | * | |
847 | * 1) prb_queue is thawed. | |
848 | * 2) retire_blk_timer is refreshed. | |
849 | * | |
850 | */ | |
bc59ba39 | 851 | static void prb_open_block(struct tpacket_kbdq_core *pkc1, |
852 | struct tpacket_block_desc *pbd1) | |
f6fb8f10 | 853 | { |
d413fcb4 | 854 | struct timespec64 ts; |
bc59ba39 | 855 | struct tpacket_hdr_v1 *h1 = &pbd1->hdr.bh1; |
f6fb8f10 | 856 | |
857 | smp_rmb(); | |
858 | ||
8da3056c DB |
859 | /* We could have just memset this but we will lose the |
860 | * flexibility of making the priv area sticky | |
861 | */ | |
f6fb8f10 | 862 | |
8da3056c DB |
863 | BLOCK_SNUM(pbd1) = pkc1->knxt_seq_num++; |
864 | BLOCK_NUM_PKTS(pbd1) = 0; | |
865 | BLOCK_LEN(pbd1) = BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
f6fb8f10 | 866 | |
d413fcb4 | 867 | ktime_get_real_ts64(&ts); |
8da3056c DB |
868 | |
869 | h1->ts_first_pkt.ts_sec = ts.tv_sec; | |
870 | h1->ts_first_pkt.ts_nsec = ts.tv_nsec; | |
f6fb8f10 | 871 | |
8da3056c DB |
872 | pkc1->pkblk_start = (char *)pbd1; |
873 | pkc1->nxt_offset = pkc1->pkblk_start + BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
874 | ||
875 | BLOCK_O2FP(pbd1) = (__u32)BLK_PLUS_PRIV(pkc1->blk_sizeof_priv); | |
876 | BLOCK_O2PRIV(pbd1) = BLK_HDR_LEN; | |
877 | ||
878 | pbd1->version = pkc1->version; | |
879 | pkc1->prev = pkc1->nxt_offset; | |
880 | pkc1->pkblk_end = pkc1->pkblk_start + pkc1->kblk_size; | |
881 | ||
882 | prb_thaw_queue(pkc1); | |
883 | _prb_refresh_rx_retire_blk_timer(pkc1); | |
884 | ||
885 | smp_wmb(); | |
f6fb8f10 | 886 | } |
887 | ||
888 | /* | |
889 | * Queue freeze logic: | |
890 | * 1) Assume tp_block_nr = 8 blocks. | |
891 | * 2) At time 't0', user opens Rx ring. | |
892 | * 3) Some time past 't0', kernel starts filling blocks starting from 0 .. 7 | |
893 | * 4) user-space is either sleeping or processing block '0'. | |
894 | * 5) tpacket_rcv is currently filling block '7', since there is no space left, | |
895 | * it will close block-7,loop around and try to fill block '0'. | |
896 | * call-flow: | |
897 | * __packet_lookup_frame_in_block | |
898 | * prb_retire_current_block() | |
899 | * prb_dispatch_next_block() | |
900 | * |->(BLOCK_STATUS == USER) evaluates to true | |
901 | * 5.1) Since block-0 is currently in-use, we just freeze the queue. | |
902 | * 6) Now there are two cases: | |
903 | * 6.1) Link goes idle right after the queue is frozen. | |
904 | * But remember, the last open_block() refreshed the timer. | |
905 | * When this timer expires,it will refresh itself so that we can | |
906 | * re-open block-0 in near future. | |
907 | * 6.2) Link is busy and keeps on receiving packets. This is a simple | |
908 | * case and __packet_lookup_frame_in_block will check if block-0 | |
909 | * is free and can now be re-used. | |
910 | */ | |
eea49cc9 | 911 | static void prb_freeze_queue(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 912 | struct packet_sock *po) |
913 | { | |
914 | pkc->reset_pending_on_curr_blk = 1; | |
ee80fbf3 | 915 | po->stats.stats3.tp_freeze_q_cnt++; |
f6fb8f10 | 916 | } |
917 | ||
918 | #define TOTAL_PKT_LEN_INCL_ALIGN(length) (ALIGN((length), V3_ALIGNMENT)) | |
919 | ||
920 | /* | |
921 | * If the next block is free then we will dispatch it | |
922 | * and return a good offset. | |
923 | * Else, we will freeze the queue. | |
924 | * So, caller must check the return value. | |
925 | */ | |
bc59ba39 | 926 | static void *prb_dispatch_next_block(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 927 | struct packet_sock *po) |
928 | { | |
bc59ba39 | 929 | struct tpacket_block_desc *pbd; |
f6fb8f10 | 930 | |
931 | smp_rmb(); | |
932 | ||
933 | /* 1. Get current block num */ | |
934 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
935 | ||
936 | /* 2. If this block is currently in_use then freeze the queue */ | |
937 | if (TP_STATUS_USER & BLOCK_STATUS(pbd)) { | |
938 | prb_freeze_queue(pkc, po); | |
939 | return NULL; | |
940 | } | |
941 | ||
942 | /* | |
943 | * 3. | |
944 | * open this block and return the offset where the first packet | |
945 | * needs to get stored. | |
946 | */ | |
947 | prb_open_block(pkc, pbd); | |
948 | return (void *)pkc->nxt_offset; | |
949 | } | |
950 | ||
bc59ba39 | 951 | static void prb_retire_current_block(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 952 | struct packet_sock *po, unsigned int status) |
953 | { | |
bc59ba39 | 954 | struct tpacket_block_desc *pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); |
f6fb8f10 | 955 | |
956 | /* retire/close the current block */ | |
957 | if (likely(TP_STATUS_KERNEL == BLOCK_STATUS(pbd))) { | |
958 | /* | |
959 | * Plug the case where copy_bits() is in progress on | |
960 | * cpu-0 and tpacket_rcv() got invoked on cpu-1, didn't | |
961 | * have space to copy the pkt in the current block and | |
962 | * called prb_retire_current_block() | |
963 | * | |
964 | * We don't need to worry about the TMO case because | |
965 | * the timer-handler already handled this case. | |
966 | */ | |
967 | if (!(status & TP_STATUS_BLK_TMO)) { | |
632ca50f JO |
968 | /* Waiting for skb_copy_bits to finish... */ |
969 | write_lock(&pkc->blk_fill_in_prog_lock); | |
970 | write_unlock(&pkc->blk_fill_in_prog_lock); | |
f6fb8f10 | 971 | } |
972 | prb_close_block(pkc, pbd, po, status); | |
973 | return; | |
974 | } | |
f6fb8f10 | 975 | } |
976 | ||
878cd3ba | 977 | static int prb_curr_blk_in_use(struct tpacket_block_desc *pbd) |
f6fb8f10 | 978 | { |
979 | return TP_STATUS_USER & BLOCK_STATUS(pbd); | |
980 | } | |
981 | ||
eea49cc9 | 982 | static int prb_queue_frozen(struct tpacket_kbdq_core *pkc) |
f6fb8f10 | 983 | { |
984 | return pkc->reset_pending_on_curr_blk; | |
985 | } | |
986 | ||
eea49cc9 | 987 | static void prb_clear_blk_fill_status(struct packet_ring_buffer *rb) |
88fd1cb8 | 988 | __releases(&pkc->blk_fill_in_prog_lock) |
f6fb8f10 | 989 | { |
bc59ba39 | 990 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(rb); |
632ca50f JO |
991 | |
992 | read_unlock(&pkc->blk_fill_in_prog_lock); | |
f6fb8f10 | 993 | } |
994 | ||
eea49cc9 | 995 | static void prb_fill_rxhash(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 996 | struct tpacket3_hdr *ppd) |
997 | { | |
3958afa1 | 998 | ppd->hv1.tp_rxhash = skb_get_hash(pkc->skb); |
f6fb8f10 | 999 | } |
1000 | ||
eea49cc9 | 1001 | static void prb_clear_rxhash(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 1002 | struct tpacket3_hdr *ppd) |
1003 | { | |
1004 | ppd->hv1.tp_rxhash = 0; | |
1005 | } | |
1006 | ||
eea49cc9 | 1007 | static void prb_fill_vlan_info(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 1008 | struct tpacket3_hdr *ppd) |
1009 | { | |
df8a39de JP |
1010 | if (skb_vlan_tag_present(pkc->skb)) { |
1011 | ppd->hv1.tp_vlan_tci = skb_vlan_tag_get(pkc->skb); | |
a0cdfcf3 AW |
1012 | ppd->hv1.tp_vlan_tpid = ntohs(pkc->skb->vlan_proto); |
1013 | ppd->tp_status = TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
f6fb8f10 | 1014 | } else { |
9e67030a | 1015 | ppd->hv1.tp_vlan_tci = 0; |
a0cdfcf3 | 1016 | ppd->hv1.tp_vlan_tpid = 0; |
9e67030a | 1017 | ppd->tp_status = TP_STATUS_AVAILABLE; |
f6fb8f10 | 1018 | } |
1019 | } | |
1020 | ||
bc59ba39 | 1021 | static void prb_run_all_ft_ops(struct tpacket_kbdq_core *pkc, |
f6fb8f10 | 1022 | struct tpacket3_hdr *ppd) |
1023 | { | |
a0cdfcf3 | 1024 | ppd->hv1.tp_padding = 0; |
f6fb8f10 | 1025 | prb_fill_vlan_info(pkc, ppd); |
1026 | ||
1027 | if (pkc->feature_req_word & TP_FT_REQ_FILL_RXHASH) | |
1028 | prb_fill_rxhash(pkc, ppd); | |
1029 | else | |
1030 | prb_clear_rxhash(pkc, ppd); | |
1031 | } | |
1032 | ||
eea49cc9 | 1033 | static void prb_fill_curr_block(char *curr, |
bc59ba39 | 1034 | struct tpacket_kbdq_core *pkc, |
1035 | struct tpacket_block_desc *pbd, | |
f6fb8f10 | 1036 | unsigned int len) |
88fd1cb8 | 1037 | __acquires(&pkc->blk_fill_in_prog_lock) |
f6fb8f10 | 1038 | { |
1039 | struct tpacket3_hdr *ppd; | |
1040 | ||
1041 | ppd = (struct tpacket3_hdr *)curr; | |
1042 | ppd->tp_next_offset = TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1043 | pkc->prev = curr; | |
1044 | pkc->nxt_offset += TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1045 | BLOCK_LEN(pbd) += TOTAL_PKT_LEN_INCL_ALIGN(len); | |
1046 | BLOCK_NUM_PKTS(pbd) += 1; | |
632ca50f | 1047 | read_lock(&pkc->blk_fill_in_prog_lock); |
f6fb8f10 | 1048 | prb_run_all_ft_ops(pkc, ppd); |
1049 | } | |
1050 | ||
1051 | /* Assumes caller has the sk->rx_queue.lock */ | |
1052 | static void *__packet_lookup_frame_in_block(struct packet_sock *po, | |
1053 | struct sk_buff *skb, | |
f6fb8f10 | 1054 | unsigned int len |
1055 | ) | |
1056 | { | |
bc59ba39 | 1057 | struct tpacket_kbdq_core *pkc; |
1058 | struct tpacket_block_desc *pbd; | |
f6fb8f10 | 1059 | char *curr, *end; |
1060 | ||
e3192690 | 1061 | pkc = GET_PBDQC_FROM_RB(&po->rx_ring); |
f6fb8f10 | 1062 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); |
1063 | ||
1064 | /* Queue is frozen when user space is lagging behind */ | |
1065 | if (prb_queue_frozen(pkc)) { | |
1066 | /* | |
1067 | * Check if that last block which caused the queue to freeze, | |
1068 | * is still in_use by user-space. | |
1069 | */ | |
878cd3ba | 1070 | if (prb_curr_blk_in_use(pbd)) { |
f6fb8f10 | 1071 | /* Can't record this packet */ |
1072 | return NULL; | |
1073 | } else { | |
1074 | /* | |
1075 | * Ok, the block was released by user-space. | |
1076 | * Now let's open that block. | |
1077 | * opening a block also thaws the queue. | |
1078 | * Thawing is a side effect. | |
1079 | */ | |
1080 | prb_open_block(pkc, pbd); | |
1081 | } | |
1082 | } | |
1083 | ||
1084 | smp_mb(); | |
1085 | curr = pkc->nxt_offset; | |
1086 | pkc->skb = skb; | |
e3192690 | 1087 | end = (char *)pbd + pkc->kblk_size; |
f6fb8f10 | 1088 | |
1089 | /* first try the current block */ | |
1090 | if (curr+TOTAL_PKT_LEN_INCL_ALIGN(len) < end) { | |
1091 | prb_fill_curr_block(curr, pkc, pbd, len); | |
1092 | return (void *)curr; | |
1093 | } | |
1094 | ||
1095 | /* Ok, close the current block */ | |
1096 | prb_retire_current_block(pkc, po, 0); | |
1097 | ||
1098 | /* Now, try to dispatch the next block */ | |
1099 | curr = (char *)prb_dispatch_next_block(pkc, po); | |
1100 | if (curr) { | |
1101 | pbd = GET_CURR_PBLOCK_DESC_FROM_CORE(pkc); | |
1102 | prb_fill_curr_block(curr, pkc, pbd, len); | |
1103 | return (void *)curr; | |
1104 | } | |
1105 | ||
1106 | /* | |
1107 | * No free blocks are available.user_space hasn't caught up yet. | |
1108 | * Queue was just frozen and now this packet will get dropped. | |
1109 | */ | |
1110 | return NULL; | |
1111 | } | |
1112 | ||
eea49cc9 | 1113 | static void *packet_current_rx_frame(struct packet_sock *po, |
f6fb8f10 | 1114 | struct sk_buff *skb, |
1115 | int status, unsigned int len) | |
1116 | { | |
1117 | char *curr = NULL; | |
1118 | switch (po->tp_version) { | |
1119 | case TPACKET_V1: | |
1120 | case TPACKET_V2: | |
1121 | curr = packet_lookup_frame(po, &po->rx_ring, | |
1122 | po->rx_ring.head, status); | |
1123 | return curr; | |
1124 | case TPACKET_V3: | |
46088059 | 1125 | return __packet_lookup_frame_in_block(po, skb, len); |
f6fb8f10 | 1126 | default: |
1127 | WARN(1, "TPACKET version not supported\n"); | |
1128 | BUG(); | |
99aa3473 | 1129 | return NULL; |
f6fb8f10 | 1130 | } |
1131 | } | |
1132 | ||
dcf70cef ED |
1133 | static void *prb_lookup_block(const struct packet_sock *po, |
1134 | const struct packet_ring_buffer *rb, | |
1135 | unsigned int idx, | |
1136 | int status) | |
f6fb8f10 | 1137 | { |
bc59ba39 | 1138 | struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(rb); |
77f65ebd | 1139 | struct tpacket_block_desc *pbd = GET_PBLOCK_DESC(pkc, idx); |
f6fb8f10 | 1140 | |
1141 | if (status != BLOCK_STATUS(pbd)) | |
1142 | return NULL; | |
1143 | return pbd; | |
1144 | } | |
1145 | ||
eea49cc9 | 1146 | static int prb_previous_blk_num(struct packet_ring_buffer *rb) |
f6fb8f10 | 1147 | { |
1148 | unsigned int prev; | |
1149 | if (rb->prb_bdqc.kactive_blk_num) | |
1150 | prev = rb->prb_bdqc.kactive_blk_num-1; | |
1151 | else | |
1152 | prev = rb->prb_bdqc.knum_blocks-1; | |
1153 | return prev; | |
1154 | } | |
1155 | ||
1156 | /* Assumes caller has held the rx_queue.lock */ | |
eea49cc9 | 1157 | static void *__prb_previous_block(struct packet_sock *po, |
f6fb8f10 | 1158 | struct packet_ring_buffer *rb, |
1159 | int status) | |
1160 | { | |
1161 | unsigned int previous = prb_previous_blk_num(rb); | |
1162 | return prb_lookup_block(po, rb, previous, status); | |
1163 | } | |
1164 | ||
eea49cc9 | 1165 | static void *packet_previous_rx_frame(struct packet_sock *po, |
f6fb8f10 | 1166 | struct packet_ring_buffer *rb, |
1167 | int status) | |
1168 | { | |
1169 | if (po->tp_version <= TPACKET_V2) | |
1170 | return packet_previous_frame(po, rb, status); | |
1171 | ||
1172 | return __prb_previous_block(po, rb, status); | |
1173 | } | |
1174 | ||
eea49cc9 | 1175 | static void packet_increment_rx_head(struct packet_sock *po, |
f6fb8f10 | 1176 | struct packet_ring_buffer *rb) |
1177 | { | |
1178 | switch (po->tp_version) { | |
1179 | case TPACKET_V1: | |
1180 | case TPACKET_V2: | |
1181 | return packet_increment_head(rb); | |
1182 | case TPACKET_V3: | |
1183 | default: | |
1184 | WARN(1, "TPACKET version not supported.\n"); | |
1185 | BUG(); | |
1186 | return; | |
1187 | } | |
1188 | } | |
1189 | ||
eea49cc9 | 1190 | static void *packet_previous_frame(struct packet_sock *po, |
69e3c75f JB |
1191 | struct packet_ring_buffer *rb, |
1192 | int status) | |
1193 | { | |
1194 | unsigned int previous = rb->head ? rb->head - 1 : rb->frame_max; | |
1195 | return packet_lookup_frame(po, rb, previous, status); | |
1196 | } | |
1197 | ||
eea49cc9 | 1198 | static void packet_increment_head(struct packet_ring_buffer *buff) |
69e3c75f JB |
1199 | { |
1200 | buff->head = buff->head != buff->frame_max ? buff->head+1 : 0; | |
1201 | } | |
1202 | ||
b0138408 DB |
1203 | static void packet_inc_pending(struct packet_ring_buffer *rb) |
1204 | { | |
1205 | this_cpu_inc(*rb->pending_refcnt); | |
1206 | } | |
1207 | ||
1208 | static void packet_dec_pending(struct packet_ring_buffer *rb) | |
1209 | { | |
1210 | this_cpu_dec(*rb->pending_refcnt); | |
1211 | } | |
1212 | ||
1213 | static unsigned int packet_read_pending(const struct packet_ring_buffer *rb) | |
1214 | { | |
1215 | unsigned int refcnt = 0; | |
1216 | int cpu; | |
1217 | ||
1218 | /* We don't use pending refcount in rx_ring. */ | |
1219 | if (rb->pending_refcnt == NULL) | |
1220 | return 0; | |
1221 | ||
1222 | for_each_possible_cpu(cpu) | |
1223 | refcnt += *per_cpu_ptr(rb->pending_refcnt, cpu); | |
1224 | ||
1225 | return refcnt; | |
1226 | } | |
1227 | ||
1228 | static int packet_alloc_pending(struct packet_sock *po) | |
1229 | { | |
1230 | po->rx_ring.pending_refcnt = NULL; | |
1231 | ||
1232 | po->tx_ring.pending_refcnt = alloc_percpu(unsigned int); | |
1233 | if (unlikely(po->tx_ring.pending_refcnt == NULL)) | |
1234 | return -ENOBUFS; | |
1235 | ||
1236 | return 0; | |
1237 | } | |
1238 | ||
1239 | static void packet_free_pending(struct packet_sock *po) | |
1240 | { | |
1241 | free_percpu(po->tx_ring.pending_refcnt); | |
1242 | } | |
1243 | ||
9954729b WB |
1244 | #define ROOM_POW_OFF 2 |
1245 | #define ROOM_NONE 0x0 | |
1246 | #define ROOM_LOW 0x1 | |
1247 | #define ROOM_NORMAL 0x2 | |
1248 | ||
d4b5bd98 | 1249 | static bool __tpacket_has_room(const struct packet_sock *po, int pow_off) |
77f65ebd | 1250 | { |
9954729b WB |
1251 | int idx, len; |
1252 | ||
d4b5bd98 ED |
1253 | len = READ_ONCE(po->rx_ring.frame_max) + 1; |
1254 | idx = READ_ONCE(po->rx_ring.head); | |
9954729b WB |
1255 | if (pow_off) |
1256 | idx += len >> pow_off; | |
1257 | if (idx >= len) | |
1258 | idx -= len; | |
1259 | return packet_lookup_frame(po, &po->rx_ring, idx, TP_STATUS_KERNEL); | |
1260 | } | |
1261 | ||
dcf70cef | 1262 | static bool __tpacket_v3_has_room(const struct packet_sock *po, int pow_off) |
9954729b WB |
1263 | { |
1264 | int idx, len; | |
1265 | ||
dcf70cef ED |
1266 | len = READ_ONCE(po->rx_ring.prb_bdqc.knum_blocks); |
1267 | idx = READ_ONCE(po->rx_ring.prb_bdqc.kactive_blk_num); | |
9954729b WB |
1268 | if (pow_off) |
1269 | idx += len >> pow_off; | |
1270 | if (idx >= len) | |
1271 | idx -= len; | |
1272 | return prb_lookup_block(po, &po->rx_ring, idx, TP_STATUS_KERNEL); | |
1273 | } | |
77f65ebd | 1274 | |
0338a145 ED |
1275 | static int __packet_rcv_has_room(const struct packet_sock *po, |
1276 | const struct sk_buff *skb) | |
9954729b | 1277 | { |
0338a145 | 1278 | const struct sock *sk = &po->sk; |
9954729b WB |
1279 | int ret = ROOM_NONE; |
1280 | ||
1281 | if (po->prot_hook.func != tpacket_rcv) { | |
0338a145 ED |
1282 | int rcvbuf = READ_ONCE(sk->sk_rcvbuf); |
1283 | int avail = rcvbuf - atomic_read(&sk->sk_rmem_alloc) | |
1284 | - (skb ? skb->truesize : 0); | |
1285 | ||
1286 | if (avail > (rcvbuf >> ROOM_POW_OFF)) | |
9954729b WB |
1287 | return ROOM_NORMAL; |
1288 | else if (avail > 0) | |
1289 | return ROOM_LOW; | |
1290 | else | |
1291 | return ROOM_NONE; | |
1292 | } | |
77f65ebd | 1293 | |
9954729b WB |
1294 | if (po->tp_version == TPACKET_V3) { |
1295 | if (__tpacket_v3_has_room(po, ROOM_POW_OFF)) | |
1296 | ret = ROOM_NORMAL; | |
1297 | else if (__tpacket_v3_has_room(po, 0)) | |
1298 | ret = ROOM_LOW; | |
1299 | } else { | |
1300 | if (__tpacket_has_room(po, ROOM_POW_OFF)) | |
1301 | ret = ROOM_NORMAL; | |
1302 | else if (__tpacket_has_room(po, 0)) | |
1303 | ret = ROOM_LOW; | |
1304 | } | |
2ccdbaa6 WB |
1305 | |
1306 | return ret; | |
1307 | } | |
1308 | ||
1309 | static int packet_rcv_has_room(struct packet_sock *po, struct sk_buff *skb) | |
1310 | { | |
791a3e9f ED |
1311 | bool pressure; |
1312 | int ret; | |
2ccdbaa6 | 1313 | |
54d7c01d | 1314 | ret = __packet_rcv_has_room(po, skb); |
3a2bb84e ED |
1315 | pressure = ret != ROOM_NORMAL; |
1316 | ||
791a3e9f ED |
1317 | if (packet_sock_flag(po, PACKET_SOCK_PRESSURE) != pressure) |
1318 | packet_sock_flag_set(po, PACKET_SOCK_PRESSURE, pressure); | |
77f65ebd | 1319 | |
9954729b | 1320 | return ret; |
77f65ebd WB |
1321 | } |
1322 | ||
9bb6cd65 ED |
1323 | static void packet_rcv_try_clear_pressure(struct packet_sock *po) |
1324 | { | |
791a3e9f | 1325 | if (packet_sock_flag(po, PACKET_SOCK_PRESSURE) && |
9bb6cd65 | 1326 | __packet_rcv_has_room(po, NULL) == ROOM_NORMAL) |
791a3e9f | 1327 | packet_sock_flag_set(po, PACKET_SOCK_PRESSURE, false); |
9bb6cd65 ED |
1328 | } |
1329 | ||
1da177e4 LT |
1330 | static void packet_sock_destruct(struct sock *sk) |
1331 | { | |
ed85b565 RC |
1332 | skb_queue_purge(&sk->sk_error_queue); |
1333 | ||
547b792c | 1334 | WARN_ON(atomic_read(&sk->sk_rmem_alloc)); |
14afee4b | 1335 | WARN_ON(refcount_read(&sk->sk_wmem_alloc)); |
1da177e4 LT |
1336 | |
1337 | if (!sock_flag(sk, SOCK_DEAD)) { | |
40d4e3df | 1338 | pr_err("Attempt to release alive packet socket: %p\n", sk); |
1da177e4 LT |
1339 | return; |
1340 | } | |
1da177e4 LT |
1341 | } |
1342 | ||
3b3a5b0a WB |
1343 | static bool fanout_flow_is_huge(struct packet_sock *po, struct sk_buff *skb) |
1344 | { | |
b756ad92 ED |
1345 | u32 *history = po->rollover->history; |
1346 | u32 victim, rxhash; | |
3b3a5b0a WB |
1347 | int i, count = 0; |
1348 | ||
1349 | rxhash = skb_get_hash(skb); | |
1350 | for (i = 0; i < ROLLOVER_HLEN; i++) | |
b756ad92 | 1351 | if (READ_ONCE(history[i]) == rxhash) |
3b3a5b0a WB |
1352 | count++; |
1353 | ||
8032bf12 | 1354 | victim = get_random_u32_below(ROLLOVER_HLEN); |
b756ad92 ED |
1355 | |
1356 | /* Avoid dirtying the cache line if possible */ | |
1357 | if (READ_ONCE(history[victim]) != rxhash) | |
1358 | WRITE_ONCE(history[victim], rxhash); | |
1359 | ||
3b3a5b0a WB |
1360 | return count > (ROLLOVER_HLEN >> 1); |
1361 | } | |
1362 | ||
77f65ebd WB |
1363 | static unsigned int fanout_demux_hash(struct packet_fanout *f, |
1364 | struct sk_buff *skb, | |
1365 | unsigned int num) | |
dc99f600 | 1366 | { |
eb70db87 | 1367 | return reciprocal_scale(__skb_get_hash_symmetric(skb), num); |
dc99f600 DM |
1368 | } |
1369 | ||
77f65ebd WB |
1370 | static unsigned int fanout_demux_lb(struct packet_fanout *f, |
1371 | struct sk_buff *skb, | |
1372 | unsigned int num) | |
dc99f600 | 1373 | { |
468479e6 | 1374 | unsigned int val = atomic_inc_return(&f->rr_cur); |
dc99f600 | 1375 | |
468479e6 | 1376 | return val % num; |
77f65ebd WB |
1377 | } |
1378 | ||
1379 | static unsigned int fanout_demux_cpu(struct packet_fanout *f, | |
1380 | struct sk_buff *skb, | |
1381 | unsigned int num) | |
1382 | { | |
1383 | return smp_processor_id() % num; | |
dc99f600 DM |
1384 | } |
1385 | ||
5df0ddfb DB |
1386 | static unsigned int fanout_demux_rnd(struct packet_fanout *f, |
1387 | struct sk_buff *skb, | |
1388 | unsigned int num) | |
1389 | { | |
8032bf12 | 1390 | return get_random_u32_below(num); |
5df0ddfb DB |
1391 | } |
1392 | ||
77f65ebd WB |
1393 | static unsigned int fanout_demux_rollover(struct packet_fanout *f, |
1394 | struct sk_buff *skb, | |
ad377cab | 1395 | unsigned int idx, bool try_self, |
77f65ebd | 1396 | unsigned int num) |
95ec3eb4 | 1397 | { |
4633c9e0 | 1398 | struct packet_sock *po, *po_next, *po_skip = NULL; |
a9b63918 | 1399 | unsigned int i, j, room = ROOM_NONE; |
95ec3eb4 | 1400 | |
94f633ea | 1401 | po = pkt_sk(rcu_dereference(f->arr[idx])); |
3b3a5b0a WB |
1402 | |
1403 | if (try_self) { | |
1404 | room = packet_rcv_has_room(po, skb); | |
1405 | if (room == ROOM_NORMAL || | |
1406 | (room == ROOM_LOW && !fanout_flow_is_huge(po, skb))) | |
1407 | return idx; | |
4633c9e0 | 1408 | po_skip = po; |
3b3a5b0a | 1409 | } |
ad377cab | 1410 | |
0648ab70 | 1411 | i = j = min_t(int, po->rollover->sock, num - 1); |
77f65ebd | 1412 | do { |
94f633ea | 1413 | po_next = pkt_sk(rcu_dereference(f->arr[i])); |
791a3e9f ED |
1414 | if (po_next != po_skip && |
1415 | !packet_sock_flag(po_next, PACKET_SOCK_PRESSURE) && | |
2ccdbaa6 | 1416 | packet_rcv_has_room(po_next, skb) == ROOM_NORMAL) { |
77f65ebd | 1417 | if (i != j) |
0648ab70 | 1418 | po->rollover->sock = i; |
a9b63918 WB |
1419 | atomic_long_inc(&po->rollover->num); |
1420 | if (room == ROOM_LOW) | |
1421 | atomic_long_inc(&po->rollover->num_huge); | |
77f65ebd WB |
1422 | return i; |
1423 | } | |
ad377cab | 1424 | |
77f65ebd WB |
1425 | if (++i == num) |
1426 | i = 0; | |
1427 | } while (i != j); | |
1428 | ||
a9b63918 | 1429 | atomic_long_inc(&po->rollover->num_failed); |
77f65ebd WB |
1430 | return idx; |
1431 | } | |
1432 | ||
2d36097d NH |
1433 | static unsigned int fanout_demux_qm(struct packet_fanout *f, |
1434 | struct sk_buff *skb, | |
1435 | unsigned int num) | |
1436 | { | |
1437 | return skb_get_queue_mapping(skb) % num; | |
1438 | } | |
1439 | ||
47dceb8e WB |
1440 | static unsigned int fanout_demux_bpf(struct packet_fanout *f, |
1441 | struct sk_buff *skb, | |
1442 | unsigned int num) | |
1443 | { | |
1444 | struct bpf_prog *prog; | |
1445 | unsigned int ret = 0; | |
1446 | ||
1447 | rcu_read_lock(); | |
1448 | prog = rcu_dereference(f->bpf_prog); | |
1449 | if (prog) | |
ff936a04 | 1450 | ret = bpf_prog_run_clear_cb(prog, skb) % num; |
47dceb8e WB |
1451 | rcu_read_unlock(); |
1452 | ||
1453 | return ret; | |
1454 | } | |
1455 | ||
77f65ebd WB |
1456 | static bool fanout_has_flag(struct packet_fanout *f, u16 flag) |
1457 | { | |
1458 | return f->flags & (flag >> 8); | |
95ec3eb4 DM |
1459 | } |
1460 | ||
95ec3eb4 DM |
1461 | static int packet_rcv_fanout(struct sk_buff *skb, struct net_device *dev, |
1462 | struct packet_type *pt, struct net_device *orig_dev) | |
dc99f600 DM |
1463 | { |
1464 | struct packet_fanout *f = pt->af_packet_priv; | |
f98f4514 | 1465 | unsigned int num = READ_ONCE(f->num_members); |
19bcf9f2 | 1466 | struct net *net = read_pnet(&f->net); |
dc99f600 | 1467 | struct packet_sock *po; |
77f65ebd | 1468 | unsigned int idx; |
dc99f600 | 1469 | |
19bcf9f2 | 1470 | if (!net_eq(dev_net(dev), net) || !num) { |
dc99f600 DM |
1471 | kfree_skb(skb); |
1472 | return 0; | |
1473 | } | |
1474 | ||
3f34b24a | 1475 | if (fanout_has_flag(f, PACKET_FANOUT_FLAG_DEFRAG)) { |
19bcf9f2 | 1476 | skb = ip_check_defrag(net, skb, IP_DEFRAG_AF_PACKET); |
3f34b24a AD |
1477 | if (!skb) |
1478 | return 0; | |
1479 | } | |
95ec3eb4 DM |
1480 | switch (f->type) { |
1481 | case PACKET_FANOUT_HASH: | |
1482 | default: | |
77f65ebd | 1483 | idx = fanout_demux_hash(f, skb, num); |
95ec3eb4 DM |
1484 | break; |
1485 | case PACKET_FANOUT_LB: | |
77f65ebd | 1486 | idx = fanout_demux_lb(f, skb, num); |
95ec3eb4 DM |
1487 | break; |
1488 | case PACKET_FANOUT_CPU: | |
77f65ebd WB |
1489 | idx = fanout_demux_cpu(f, skb, num); |
1490 | break; | |
5df0ddfb DB |
1491 | case PACKET_FANOUT_RND: |
1492 | idx = fanout_demux_rnd(f, skb, num); | |
1493 | break; | |
2d36097d NH |
1494 | case PACKET_FANOUT_QM: |
1495 | idx = fanout_demux_qm(f, skb, num); | |
1496 | break; | |
77f65ebd | 1497 | case PACKET_FANOUT_ROLLOVER: |
ad377cab | 1498 | idx = fanout_demux_rollover(f, skb, 0, false, num); |
95ec3eb4 | 1499 | break; |
47dceb8e | 1500 | case PACKET_FANOUT_CBPF: |
f2e52095 | 1501 | case PACKET_FANOUT_EBPF: |
47dceb8e WB |
1502 | idx = fanout_demux_bpf(f, skb, num); |
1503 | break; | |
dc99f600 DM |
1504 | } |
1505 | ||
ad377cab WB |
1506 | if (fanout_has_flag(f, PACKET_FANOUT_FLAG_ROLLOVER)) |
1507 | idx = fanout_demux_rollover(f, skb, idx, true, num); | |
dc99f600 | 1508 | |
94f633ea | 1509 | po = pkt_sk(rcu_dereference(f->arr[idx])); |
dc99f600 DM |
1510 | return po->prot_hook.func(skb, dev, &po->prot_hook, orig_dev); |
1511 | } | |
1512 | ||
fff3321d PE |
1513 | DEFINE_MUTEX(fanout_mutex); |
1514 | EXPORT_SYMBOL_GPL(fanout_mutex); | |
dc99f600 | 1515 | static LIST_HEAD(fanout_list); |
4a69a864 | 1516 | static u16 fanout_next_id; |
dc99f600 DM |
1517 | |
1518 | static void __fanout_link(struct sock *sk, struct packet_sock *po) | |
1519 | { | |
1520 | struct packet_fanout *f = po->fanout; | |
1521 | ||
1522 | spin_lock(&f->lock); | |
94f633ea | 1523 | rcu_assign_pointer(f->arr[f->num_members], sk); |
dc99f600 DM |
1524 | smp_wmb(); |
1525 | f->num_members++; | |
2bd624b4 AS |
1526 | if (f->num_members == 1) |
1527 | dev_add_pack(&f->prot_hook); | |
dc99f600 DM |
1528 | spin_unlock(&f->lock); |
1529 | } | |
1530 | ||
1531 | static void __fanout_unlink(struct sock *sk, struct packet_sock *po) | |
1532 | { | |
1533 | struct packet_fanout *f = po->fanout; | |
1534 | int i; | |
1535 | ||
1536 | spin_lock(&f->lock); | |
1537 | for (i = 0; i < f->num_members; i++) { | |
94f633ea ED |
1538 | if (rcu_dereference_protected(f->arr[i], |
1539 | lockdep_is_held(&f->lock)) == sk) | |
dc99f600 DM |
1540 | break; |
1541 | } | |
1542 | BUG_ON(i >= f->num_members); | |
94f633ea ED |
1543 | rcu_assign_pointer(f->arr[i], |
1544 | rcu_dereference_protected(f->arr[f->num_members - 1], | |
1545 | lockdep_is_held(&f->lock))); | |
dc99f600 | 1546 | f->num_members--; |
2bd624b4 AS |
1547 | if (f->num_members == 0) |
1548 | __dev_remove_pack(&f->prot_hook); | |
dc99f600 DM |
1549 | spin_unlock(&f->lock); |
1550 | } | |
1551 | ||
d4dd8aee | 1552 | static bool match_fanout_group(struct packet_type *ptype, struct sock *sk) |
c0de08d0 | 1553 | { |
161642e2 ED |
1554 | if (sk->sk_family != PF_PACKET) |
1555 | return false; | |
c0de08d0 | 1556 | |
161642e2 | 1557 | return ptype->af_packet_priv == pkt_sk(sk)->fanout; |
c0de08d0 EL |
1558 | } |
1559 | ||
47dceb8e WB |
1560 | static void fanout_init_data(struct packet_fanout *f) |
1561 | { | |
1562 | switch (f->type) { | |
1563 | case PACKET_FANOUT_LB: | |
1564 | atomic_set(&f->rr_cur, 0); | |
1565 | break; | |
1566 | case PACKET_FANOUT_CBPF: | |
f2e52095 | 1567 | case PACKET_FANOUT_EBPF: |
47dceb8e WB |
1568 | RCU_INIT_POINTER(f->bpf_prog, NULL); |
1569 | break; | |
1570 | } | |
1571 | } | |
1572 | ||
1573 | static void __fanout_set_data_bpf(struct packet_fanout *f, struct bpf_prog *new) | |
1574 | { | |
1575 | struct bpf_prog *old; | |
1576 | ||
1577 | spin_lock(&f->lock); | |
1578 | old = rcu_dereference_protected(f->bpf_prog, lockdep_is_held(&f->lock)); | |
1579 | rcu_assign_pointer(f->bpf_prog, new); | |
1580 | spin_unlock(&f->lock); | |
1581 | ||
1582 | if (old) { | |
1583 | synchronize_net(); | |
1584 | bpf_prog_destroy(old); | |
1585 | } | |
1586 | } | |
1587 | ||
b1ea9ff6 | 1588 | static int fanout_set_data_cbpf(struct packet_sock *po, sockptr_t data, |
47dceb8e WB |
1589 | unsigned int len) |
1590 | { | |
1591 | struct bpf_prog *new; | |
1592 | struct sock_fprog fprog; | |
1593 | int ret; | |
1594 | ||
1595 | if (sock_flag(&po->sk, SOCK_FILTER_LOCKED)) | |
1596 | return -EPERM; | |
4d295e54 CH |
1597 | |
1598 | ret = copy_bpf_fprog_from_user(&fprog, data, len); | |
1599 | if (ret) | |
1600 | return ret; | |
47dceb8e | 1601 | |
bab18991 | 1602 | ret = bpf_prog_create_from_user(&new, &fprog, NULL, false); |
47dceb8e WB |
1603 | if (ret) |
1604 | return ret; | |
1605 | ||
1606 | __fanout_set_data_bpf(po->fanout, new); | |
1607 | return 0; | |
1608 | } | |
1609 | ||
a7b75c5a | 1610 | static int fanout_set_data_ebpf(struct packet_sock *po, sockptr_t data, |
f2e52095 WB |
1611 | unsigned int len) |
1612 | { | |
1613 | struct bpf_prog *new; | |
1614 | u32 fd; | |
1615 | ||
1616 | if (sock_flag(&po->sk, SOCK_FILTER_LOCKED)) | |
1617 | return -EPERM; | |
1618 | if (len != sizeof(fd)) | |
1619 | return -EINVAL; | |
a7b75c5a | 1620 | if (copy_from_sockptr(&fd, data, len)) |
f2e52095 WB |
1621 | return -EFAULT; |
1622 | ||
113214be | 1623 | new = bpf_prog_get_type(fd, BPF_PROG_TYPE_SOCKET_FILTER); |
f2e52095 WB |
1624 | if (IS_ERR(new)) |
1625 | return PTR_ERR(new); | |
f2e52095 WB |
1626 | |
1627 | __fanout_set_data_bpf(po->fanout, new); | |
1628 | return 0; | |
1629 | } | |
1630 | ||
a7b75c5a | 1631 | static int fanout_set_data(struct packet_sock *po, sockptr_t data, |
47dceb8e WB |
1632 | unsigned int len) |
1633 | { | |
1634 | switch (po->fanout->type) { | |
1635 | case PACKET_FANOUT_CBPF: | |
a7b75c5a | 1636 | return fanout_set_data_cbpf(po, data, len); |
f2e52095 WB |
1637 | case PACKET_FANOUT_EBPF: |
1638 | return fanout_set_data_ebpf(po, data, len); | |
47dceb8e WB |
1639 | default: |
1640 | return -EINVAL; | |
07d53ae4 | 1641 | } |
47dceb8e WB |
1642 | } |
1643 | ||
1644 | static void fanout_release_data(struct packet_fanout *f) | |
1645 | { | |
1646 | switch (f->type) { | |
1647 | case PACKET_FANOUT_CBPF: | |
f2e52095 | 1648 | case PACKET_FANOUT_EBPF: |
47dceb8e | 1649 | __fanout_set_data_bpf(f, NULL); |
07d53ae4 | 1650 | } |
47dceb8e WB |
1651 | } |
1652 | ||
4a69a864 MM |
1653 | static bool __fanout_id_is_free(struct sock *sk, u16 candidate_id) |
1654 | { | |
1655 | struct packet_fanout *f; | |
1656 | ||
1657 | list_for_each_entry(f, &fanout_list, list) { | |
1658 | if (f->id == candidate_id && | |
1659 | read_pnet(&f->net) == sock_net(sk)) { | |
1660 | return false; | |
1661 | } | |
1662 | } | |
1663 | return true; | |
1664 | } | |
1665 | ||
1666 | static bool fanout_find_new_id(struct sock *sk, u16 *new_id) | |
1667 | { | |
1668 | u16 id = fanout_next_id; | |
1669 | ||
1670 | do { | |
1671 | if (__fanout_id_is_free(sk, id)) { | |
1672 | *new_id = id; | |
1673 | fanout_next_id = id + 1; | |
1674 | return true; | |
1675 | } | |
1676 | ||
1677 | id++; | |
1678 | } while (id != fanout_next_id); | |
1679 | ||
1680 | return false; | |
1681 | } | |
1682 | ||
9c661b0b | 1683 | static int fanout_add(struct sock *sk, struct fanout_args *args) |
dc99f600 | 1684 | { |
d199fab6 | 1685 | struct packet_rollover *rollover = NULL; |
dc99f600 | 1686 | struct packet_sock *po = pkt_sk(sk); |
9c661b0b | 1687 | u16 type_flags = args->type_flags; |
dc99f600 | 1688 | struct packet_fanout *f, *match; |
7736d33f | 1689 | u8 type = type_flags & 0xff; |
77f65ebd | 1690 | u8 flags = type_flags >> 8; |
9c661b0b | 1691 | u16 id = args->id; |
dc99f600 DM |
1692 | int err; |
1693 | ||
1694 | switch (type) { | |
77f65ebd WB |
1695 | case PACKET_FANOUT_ROLLOVER: |
1696 | if (type_flags & PACKET_FANOUT_FLAG_ROLLOVER) | |
1697 | return -EINVAL; | |
5af5a020 | 1698 | break; |
dc99f600 DM |
1699 | case PACKET_FANOUT_HASH: |
1700 | case PACKET_FANOUT_LB: | |
95ec3eb4 | 1701 | case PACKET_FANOUT_CPU: |
5df0ddfb | 1702 | case PACKET_FANOUT_RND: |
2d36097d | 1703 | case PACKET_FANOUT_QM: |
47dceb8e | 1704 | case PACKET_FANOUT_CBPF: |
f2e52095 | 1705 | case PACKET_FANOUT_EBPF: |
dc99f600 DM |
1706 | break; |
1707 | default: | |
1708 | return -EINVAL; | |
1709 | } | |
1710 | ||
d199fab6 ED |
1711 | mutex_lock(&fanout_mutex); |
1712 | ||
d199fab6 | 1713 | err = -EALREADY; |
dc99f600 | 1714 | if (po->fanout) |
d199fab6 | 1715 | goto out; |
dc99f600 | 1716 | |
4633c9e0 WB |
1717 | if (type == PACKET_FANOUT_ROLLOVER || |
1718 | (type_flags & PACKET_FANOUT_FLAG_ROLLOVER)) { | |
d199fab6 ED |
1719 | err = -ENOMEM; |
1720 | rollover = kzalloc(sizeof(*rollover), GFP_KERNEL); | |
1721 | if (!rollover) | |
1722 | goto out; | |
1723 | atomic_long_set(&rollover->num, 0); | |
1724 | atomic_long_set(&rollover->num_huge, 0); | |
1725 | atomic_long_set(&rollover->num_failed, 0); | |
0648ab70 WB |
1726 | } |
1727 | ||
4a69a864 MM |
1728 | if (type_flags & PACKET_FANOUT_FLAG_UNIQUEID) { |
1729 | if (id != 0) { | |
1730 | err = -EINVAL; | |
1731 | goto out; | |
1732 | } | |
1733 | if (!fanout_find_new_id(sk, &id)) { | |
1734 | err = -ENOMEM; | |
1735 | goto out; | |
1736 | } | |
1737 | /* ephemeral flag for the first socket in the group: drop it */ | |
1738 | flags &= ~(PACKET_FANOUT_FLAG_UNIQUEID >> 8); | |
1739 | } | |
1740 | ||
dc99f600 DM |
1741 | match = NULL; |
1742 | list_for_each_entry(f, &fanout_list, list) { | |
1743 | if (f->id == id && | |
1744 | read_pnet(&f->net) == sock_net(sk)) { | |
1745 | match = f; | |
1746 | break; | |
1747 | } | |
1748 | } | |
afe62c68 | 1749 | err = -EINVAL; |
9c661b0b TL |
1750 | if (match) { |
1751 | if (match->flags != flags) | |
1752 | goto out; | |
1753 | if (args->max_num_members && | |
1754 | args->max_num_members != match->max_num_members) | |
1755 | goto out; | |
1756 | } else { | |
1757 | if (args->max_num_members > PACKET_FANOUT_MAX) | |
1758 | goto out; | |
1759 | if (!args->max_num_members) | |
1760 | /* legacy PACKET_FANOUT_MAX */ | |
1761 | args->max_num_members = 256; | |
afe62c68 | 1762 | err = -ENOMEM; |
9c661b0b TL |
1763 | match = kvzalloc(struct_size(match, arr, args->max_num_members), |
1764 | GFP_KERNEL); | |
afe62c68 ED |
1765 | if (!match) |
1766 | goto out; | |
1767 | write_pnet(&match->net, sock_net(sk)); | |
1768 | match->id = id; | |
1769 | match->type = type; | |
77f65ebd | 1770 | match->flags = flags; |
afe62c68 ED |
1771 | INIT_LIST_HEAD(&match->list); |
1772 | spin_lock_init(&match->lock); | |
fb5c2c17 | 1773 | refcount_set(&match->sk_ref, 0); |
47dceb8e | 1774 | fanout_init_data(match); |
afe62c68 ED |
1775 | match->prot_hook.type = po->prot_hook.type; |
1776 | match->prot_hook.dev = po->prot_hook.dev; | |
1777 | match->prot_hook.func = packet_rcv_fanout; | |
1778 | match->prot_hook.af_packet_priv = match; | |
47934e06 | 1779 | match->prot_hook.af_packet_net = read_pnet(&match->net); |
c0de08d0 | 1780 | match->prot_hook.id_match = match_fanout_group; |
9c661b0b | 1781 | match->max_num_members = args->max_num_members; |
58ba4263 | 1782 | match->prot_hook.ignore_outgoing = type_flags & PACKET_FANOUT_FLAG_IGNORE_OUTGOING; |
afe62c68 | 1783 | list_add(&match->list, &fanout_list); |
dc99f600 | 1784 | } |
afe62c68 | 1785 | err = -EINVAL; |
008ba2a1 WB |
1786 | |
1787 | spin_lock(&po->bind_lock); | |
61edf479 | 1788 | if (packet_sock_flag(po, PACKET_SOCK_RUNNING) && |
008ba2a1 | 1789 | match->type == type && |
afe62c68 ED |
1790 | match->prot_hook.type == po->prot_hook.type && |
1791 | match->prot_hook.dev == po->prot_hook.dev) { | |
1792 | err = -ENOSPC; | |
9c661b0b | 1793 | if (refcount_read(&match->sk_ref) < match->max_num_members) { |
afe62c68 | 1794 | __dev_remove_pack(&po->prot_hook); |
e42e70ad ED |
1795 | |
1796 | /* Paired with packet_setsockopt(PACKET_FANOUT_DATA) */ | |
1797 | WRITE_ONCE(po->fanout, match); | |
1798 | ||
57f015f5 MM |
1799 | po->rollover = rollover; |
1800 | rollover = NULL; | |
fb5c2c17 | 1801 | refcount_set(&match->sk_ref, refcount_read(&match->sk_ref) + 1); |
afe62c68 ED |
1802 | __fanout_link(sk, po); |
1803 | err = 0; | |
dc99f600 DM |
1804 | } |
1805 | } | |
008ba2a1 WB |
1806 | spin_unlock(&po->bind_lock); |
1807 | ||
1808 | if (err && !refcount_read(&match->sk_ref)) { | |
1809 | list_del(&match->list); | |
9c661b0b | 1810 | kvfree(match); |
008ba2a1 WB |
1811 | } |
1812 | ||
afe62c68 | 1813 | out: |
57f015f5 | 1814 | kfree(rollover); |
d199fab6 | 1815 | mutex_unlock(&fanout_mutex); |
dc99f600 DM |
1816 | return err; |
1817 | } | |
1818 | ||
2bd624b4 AS |
1819 | /* If pkt_sk(sk)->fanout->sk_ref is zero, this function removes |
1820 | * pkt_sk(sk)->fanout from fanout_list and returns pkt_sk(sk)->fanout. | |
1821 | * It is the responsibility of the caller to call fanout_release_data() and | |
1822 | * free the returned packet_fanout (after synchronize_net()) | |
1823 | */ | |
1824 | static struct packet_fanout *fanout_release(struct sock *sk) | |
dc99f600 DM |
1825 | { |
1826 | struct packet_sock *po = pkt_sk(sk); | |
1827 | struct packet_fanout *f; | |
1828 | ||
fff3321d | 1829 | mutex_lock(&fanout_mutex); |
d199fab6 ED |
1830 | f = po->fanout; |
1831 | if (f) { | |
1832 | po->fanout = NULL; | |
1833 | ||
fb5c2c17 | 1834 | if (refcount_dec_and_test(&f->sk_ref)) |
d199fab6 | 1835 | list_del(&f->list); |
2bd624b4 AS |
1836 | else |
1837 | f = NULL; | |
dc99f600 DM |
1838 | } |
1839 | mutex_unlock(&fanout_mutex); | |
2bd624b4 AS |
1840 | |
1841 | return f; | |
dc99f600 | 1842 | } |
1da177e4 | 1843 | |
3c70c132 DB |
1844 | static bool packet_extra_vlan_len_allowed(const struct net_device *dev, |
1845 | struct sk_buff *skb) | |
1846 | { | |
1847 | /* Earlier code assumed this would be a VLAN pkt, double-check | |
1848 | * this now that we have the actual packet in hand. We can only | |
1849 | * do this check on Ethernet devices. | |
1850 | */ | |
1851 | if (unlikely(dev->type != ARPHRD_ETHER)) | |
1852 | return false; | |
1853 | ||
1854 | skb_reset_mac_header(skb); | |
1855 | return likely(eth_hdr(skb)->h_proto == htons(ETH_P_8021Q)); | |
1856 | } | |
1857 | ||
90ddc4f0 | 1858 | static const struct proto_ops packet_ops; |
1da177e4 | 1859 | |
90ddc4f0 | 1860 | static const struct proto_ops packet_ops_spkt; |
1da177e4 | 1861 | |
40d4e3df ED |
1862 | static int packet_rcv_spkt(struct sk_buff *skb, struct net_device *dev, |
1863 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
1864 | { |
1865 | struct sock *sk; | |
1866 | struct sockaddr_pkt *spkt; | |
1867 | ||
1868 | /* | |
1869 | * When we registered the protocol we saved the socket in the data | |
1870 | * field for just this event. | |
1871 | */ | |
1872 | ||
1873 | sk = pt->af_packet_priv; | |
1ce4f28b | 1874 | |
1da177e4 LT |
1875 | /* |
1876 | * Yank back the headers [hope the device set this | |
1877 | * right or kerboom...] | |
1878 | * | |
1879 | * Incoming packets have ll header pulled, | |
1880 | * push it back. | |
1881 | * | |
98e399f8 | 1882 | * For outgoing ones skb->data == skb_mac_header(skb) |
1da177e4 LT |
1883 | * so that this procedure is noop. |
1884 | */ | |
1885 | ||
1886 | if (skb->pkt_type == PACKET_LOOPBACK) | |
1887 | goto out; | |
1888 | ||
09ad9bc7 | 1889 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
1890 | goto out; |
1891 | ||
40d4e3df ED |
1892 | skb = skb_share_check(skb, GFP_ATOMIC); |
1893 | if (skb == NULL) | |
1da177e4 LT |
1894 | goto oom; |
1895 | ||
1896 | /* drop any routing info */ | |
adf30907 | 1897 | skb_dst_drop(skb); |
1da177e4 | 1898 | |
84531c24 | 1899 | /* drop conntrack reference */ |
895b5c9f | 1900 | nf_reset_ct(skb); |
84531c24 | 1901 | |
ffbc6111 | 1902 | spkt = &PACKET_SKB_CB(skb)->sa.pkt; |
1da177e4 | 1903 | |
98e399f8 | 1904 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
1905 | |
1906 | /* | |
1907 | * The SOCK_PACKET socket receives _all_ frames. | |
1908 | */ | |
1909 | ||
1910 | spkt->spkt_family = dev->type; | |
8fc9d51e | 1911 | strscpy(spkt->spkt_device, dev->name, sizeof(spkt->spkt_device)); |
1da177e4 LT |
1912 | spkt->spkt_protocol = skb->protocol; |
1913 | ||
1914 | /* | |
1915 | * Charge the memory to the socket. This is done specifically | |
1916 | * to prevent sockets using all the memory up. | |
1917 | */ | |
1918 | ||
40d4e3df | 1919 | if (sock_queue_rcv_skb(sk, skb) == 0) |
1da177e4 LT |
1920 | return 0; |
1921 | ||
1922 | out: | |
1923 | kfree_skb(skb); | |
1924 | oom: | |
1925 | return 0; | |
1926 | } | |
1927 | ||
75c65772 MM |
1928 | static void packet_parse_headers(struct sk_buff *skb, struct socket *sock) |
1929 | { | |
dfed913e HL |
1930 | int depth; |
1931 | ||
18bed891 YK |
1932 | if ((!skb->protocol || skb->protocol == htons(ETH_P_ALL)) && |
1933 | sock->type == SOCK_RAW) { | |
75c65772 MM |
1934 | skb_reset_mac_header(skb); |
1935 | skb->protocol = dev_parse_header_protocol(skb); | |
1936 | } | |
1937 | ||
dfed913e HL |
1938 | /* Move network header to the right position for VLAN tagged packets */ |
1939 | if (likely(skb->dev->type == ARPHRD_ETHER) && | |
1940 | eth_type_vlan(skb->protocol) && | |
4063384e ED |
1941 | vlan_get_protocol_and_depth(skb, skb->protocol, &depth) != 0) |
1942 | skb_set_network_header(skb, depth); | |
dfed913e | 1943 | |
75c65772 MM |
1944 | skb_probe_transport_header(skb); |
1945 | } | |
1da177e4 LT |
1946 | |
1947 | /* | |
1948 | * Output a raw packet to a device layer. This bypasses all the other | |
1949 | * protocol layers and you must therefore supply it with a complete frame | |
1950 | */ | |
1ce4f28b | 1951 | |
1b784140 YX |
1952 | static int packet_sendmsg_spkt(struct socket *sock, struct msghdr *msg, |
1953 | size_t len) | |
1da177e4 LT |
1954 | { |
1955 | struct sock *sk = sock->sk; | |
342dfc30 | 1956 | DECLARE_SOCKADDR(struct sockaddr_pkt *, saddr, msg->msg_name); |
1a35ca80 | 1957 | struct sk_buff *skb = NULL; |
1da177e4 | 1958 | struct net_device *dev; |
c14ac945 | 1959 | struct sockcm_cookie sockc; |
40d4e3df | 1960 | __be16 proto = 0; |
1da177e4 | 1961 | int err; |
3bdc0eba | 1962 | int extra_len = 0; |
1ce4f28b | 1963 | |
1da177e4 | 1964 | /* |
1ce4f28b | 1965 | * Get and verify the address. |
1da177e4 LT |
1966 | */ |
1967 | ||
40d4e3df | 1968 | if (saddr) { |
1da177e4 | 1969 | if (msg->msg_namelen < sizeof(struct sockaddr)) |
40d4e3df ED |
1970 | return -EINVAL; |
1971 | if (msg->msg_namelen == sizeof(struct sockaddr_pkt)) | |
1972 | proto = saddr->spkt_protocol; | |
1973 | } else | |
1974 | return -ENOTCONN; /* SOCK_PACKET must be sent giving an address */ | |
1da177e4 LT |
1975 | |
1976 | /* | |
1ce4f28b | 1977 | * Find the device first to size check it |
1da177e4 LT |
1978 | */ |
1979 | ||
de74e92a | 1980 | saddr->spkt_device[sizeof(saddr->spkt_device) - 1] = 0; |
1a35ca80 | 1981 | retry: |
654d1f8a ED |
1982 | rcu_read_lock(); |
1983 | dev = dev_get_by_name_rcu(sock_net(sk), saddr->spkt_device); | |
1da177e4 LT |
1984 | err = -ENODEV; |
1985 | if (dev == NULL) | |
1986 | goto out_unlock; | |
1ce4f28b | 1987 | |
d5e76b0a DM |
1988 | err = -ENETDOWN; |
1989 | if (!(dev->flags & IFF_UP)) | |
1990 | goto out_unlock; | |
1991 | ||
1da177e4 | 1992 | /* |
40d4e3df ED |
1993 | * You may not queue a frame bigger than the mtu. This is the lowest level |
1994 | * raw protocol and you must do your own fragmentation at this level. | |
1da177e4 | 1995 | */ |
1ce4f28b | 1996 | |
3bdc0eba BG |
1997 | if (unlikely(sock_flag(sk, SOCK_NOFCS))) { |
1998 | if (!netif_supports_nofcs(dev)) { | |
1999 | err = -EPROTONOSUPPORT; | |
2000 | goto out_unlock; | |
2001 | } | |
2002 | extra_len = 4; /* We're doing our own CRC */ | |
2003 | } | |
2004 | ||
1da177e4 | 2005 | err = -EMSGSIZE; |
3bdc0eba | 2006 | if (len > dev->mtu + dev->hard_header_len + VLAN_HLEN + extra_len) |
1da177e4 LT |
2007 | goto out_unlock; |
2008 | ||
1a35ca80 ED |
2009 | if (!skb) { |
2010 | size_t reserved = LL_RESERVED_SPACE(dev); | |
4ce40912 | 2011 | int tlen = dev->needed_tailroom; |
1a35ca80 ED |
2012 | unsigned int hhlen = dev->header_ops ? dev->hard_header_len : 0; |
2013 | ||
2014 | rcu_read_unlock(); | |
4ce40912 | 2015 | skb = sock_wmalloc(sk, len + reserved + tlen, 0, GFP_KERNEL); |
1a35ca80 ED |
2016 | if (skb == NULL) |
2017 | return -ENOBUFS; | |
2018 | /* FIXME: Save some space for broken drivers that write a hard | |
2019 | * header at transmission time by themselves. PPP is the notable | |
2020 | * one here. This should really be fixed at the driver level. | |
2021 | */ | |
2022 | skb_reserve(skb, reserved); | |
2023 | skb_reset_network_header(skb); | |
2024 | ||
2025 | /* Try to align data part correctly */ | |
2026 | if (hhlen) { | |
2027 | skb->data -= hhlen; | |
2028 | skb->tail -= hhlen; | |
2029 | if (len < hhlen) | |
2030 | skb_reset_network_header(skb); | |
2031 | } | |
6ce8e9ce | 2032 | err = memcpy_from_msg(skb_put(skb, len), msg, len); |
1a35ca80 ED |
2033 | if (err) |
2034 | goto out_free; | |
2035 | goto retry; | |
1da177e4 LT |
2036 | } |
2037 | ||
6a341729 | 2038 | if (!dev_validate_header(dev, skb->data, len) || !skb->len) { |
9ed988cd WB |
2039 | err = -EINVAL; |
2040 | goto out_unlock; | |
2041 | } | |
3c70c132 DB |
2042 | if (len > (dev->mtu + dev->hard_header_len + extra_len) && |
2043 | !packet_extra_vlan_len_allowed(dev, skb)) { | |
2044 | err = -EMSGSIZE; | |
2045 | goto out_unlock; | |
57f89bfa | 2046 | } |
1a35ca80 | 2047 | |
657a0667 | 2048 | sockcm_init(&sockc, sk); |
c14ac945 SHY |
2049 | if (msg->msg_controllen) { |
2050 | err = sock_cmsg_send(sk, msg, &sockc); | |
f8e7718c | 2051 | if (unlikely(err)) |
c14ac945 | 2052 | goto out_unlock; |
c14ac945 SHY |
2053 | } |
2054 | ||
1da177e4 LT |
2055 | skb->protocol = proto; |
2056 | skb->dev = dev; | |
8bf43be7 | 2057 | skb->priority = READ_ONCE(sk->sk_priority); |
3c5b4d69 | 2058 | skb->mark = READ_ONCE(sk->sk_mark); |
3d0ba8c0 | 2059 | skb->tstamp = sockc.transmit_time; |
bf84a010 | 2060 | |
8f932f76 | 2061 | skb_setup_tx_timestamp(skb, sockc.tsflags); |
1da177e4 | 2062 | |
3bdc0eba BG |
2063 | if (unlikely(extra_len == 4)) |
2064 | skb->no_fcs = 1; | |
2065 | ||
75c65772 | 2066 | packet_parse_headers(skb, sock); |
c1aad275 | 2067 | |
1da177e4 | 2068 | dev_queue_xmit(skb); |
654d1f8a | 2069 | rcu_read_unlock(); |
40d4e3df | 2070 | return len; |
1da177e4 | 2071 | |
1da177e4 | 2072 | out_unlock: |
654d1f8a | 2073 | rcu_read_unlock(); |
1a35ca80 ED |
2074 | out_free: |
2075 | kfree_skb(skb); | |
1da177e4 LT |
2076 | return err; |
2077 | } | |
1da177e4 | 2078 | |
ff936a04 AS |
2079 | static unsigned int run_filter(struct sk_buff *skb, |
2080 | const struct sock *sk, | |
2081 | unsigned int res) | |
1da177e4 LT |
2082 | { |
2083 | struct sk_filter *filter; | |
fda9ef5d | 2084 | |
80f8f102 ED |
2085 | rcu_read_lock(); |
2086 | filter = rcu_dereference(sk->sk_filter); | |
dbcb5855 | 2087 | if (filter != NULL) |
ff936a04 | 2088 | res = bpf_prog_run_clear_cb(filter->prog, skb); |
80f8f102 | 2089 | rcu_read_unlock(); |
1da177e4 | 2090 | |
dbcb5855 | 2091 | return res; |
1da177e4 LT |
2092 | } |
2093 | ||
16cc1400 | 2094 | static int packet_rcv_vnet(struct msghdr *msg, const struct sk_buff *skb, |
dfc39d40 | 2095 | size_t *len, int vnet_hdr_sz) |
16cc1400 | 2096 | { |
dfc39d40 | 2097 | struct virtio_net_hdr_mrg_rxbuf vnet_hdr = { .num_buffers = 0 }; |
16cc1400 | 2098 | |
dfc39d40 | 2099 | if (*len < vnet_hdr_sz) |
16cc1400 | 2100 | return -EINVAL; |
dfc39d40 | 2101 | *len -= vnet_hdr_sz; |
16cc1400 | 2102 | |
dfc39d40 | 2103 | if (virtio_net_hdr_from_skb(skb, (struct virtio_net_hdr *)&vnet_hdr, vio_le(), true, 0)) |
16cc1400 WB |
2104 | return -EINVAL; |
2105 | ||
dfc39d40 | 2106 | return memcpy_to_msg(msg, (void *)&vnet_hdr, vnet_hdr_sz); |
16cc1400 WB |
2107 | } |
2108 | ||
1da177e4 | 2109 | /* |
62ab0812 ED |
2110 | * This function makes lazy skb cloning in hope that most of packets |
2111 | * are discarded by BPF. | |
2112 | * | |
2113 | * Note tricky part: we DO mangle shared skb! skb->data, skb->len | |
2114 | * and skb->cb are mangled. It works because (and until) packets | |
2115 | * falling here are owned by current CPU. Output packets are cloned | |
2116 | * by dev_queue_xmit_nit(), input packets are processed by net_bh | |
0e4161d0 | 2117 | * sequentially, so that if we return skb to original state on exit, |
62ab0812 | 2118 | * we will not harm anyone. |
1da177e4 LT |
2119 | */ |
2120 | ||
40d4e3df ED |
2121 | static int packet_rcv(struct sk_buff *skb, struct net_device *dev, |
2122 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
2123 | { |
2124 | struct sock *sk; | |
2125 | struct sockaddr_ll *sll; | |
2126 | struct packet_sock *po; | |
40d4e3df | 2127 | u8 *skb_head = skb->data; |
1da177e4 | 2128 | int skb_len = skb->len; |
dbcb5855 | 2129 | unsigned int snaplen, res; |
da37845f | 2130 | bool is_drop_n_account = false; |
1da177e4 LT |
2131 | |
2132 | if (skb->pkt_type == PACKET_LOOPBACK) | |
2133 | goto drop; | |
2134 | ||
2135 | sk = pt->af_packet_priv; | |
2136 | po = pkt_sk(sk); | |
2137 | ||
09ad9bc7 | 2138 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
2139 | goto drop; |
2140 | ||
1da177e4 LT |
2141 | skb->dev = dev; |
2142 | ||
d5496990 | 2143 | if (dev_has_header(dev)) { |
1da177e4 | 2144 | /* The device has an explicit notion of ll header, |
62ab0812 ED |
2145 | * exported to higher levels. |
2146 | * | |
2147 | * Otherwise, the device hides details of its frame | |
2148 | * structure, so that corresponding packet head is | |
2149 | * never delivered to user. | |
1da177e4 LT |
2150 | */ |
2151 | if (sk->sk_type != SOCK_DGRAM) | |
98e399f8 | 2152 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
2153 | else if (skb->pkt_type == PACKET_OUTGOING) { |
2154 | /* Special case: outgoing packets have ll header at head */ | |
bbe735e4 | 2155 | skb_pull(skb, skb_network_offset(skb)); |
1da177e4 LT |
2156 | } |
2157 | } | |
2158 | ||
2159 | snaplen = skb->len; | |
2160 | ||
dbcb5855 DM |
2161 | res = run_filter(skb, sk, snaplen); |
2162 | if (!res) | |
fda9ef5d | 2163 | goto drop_n_restore; |
dbcb5855 DM |
2164 | if (snaplen > res) |
2165 | snaplen = res; | |
1da177e4 | 2166 | |
0fd7bac6 | 2167 | if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) |
1da177e4 LT |
2168 | goto drop_n_acct; |
2169 | ||
2170 | if (skb_shared(skb)) { | |
2171 | struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC); | |
2172 | if (nskb == NULL) | |
2173 | goto drop_n_acct; | |
2174 | ||
2175 | if (skb_head != skb->data) { | |
2176 | skb->data = skb_head; | |
2177 | skb->len = skb_len; | |
2178 | } | |
abc4e4fa | 2179 | consume_skb(skb); |
1da177e4 LT |
2180 | skb = nskb; |
2181 | } | |
2182 | ||
b4772ef8 | 2183 | sock_skb_cb_check_size(sizeof(*PACKET_SKB_CB(skb)) + MAX_ADDR_LEN - 8); |
ffbc6111 HX |
2184 | |
2185 | sll = &PACKET_SKB_CB(skb)->sa.ll; | |
1da177e4 | 2186 | sll->sll_hatype = dev->type; |
1da177e4 | 2187 | sll->sll_pkttype = skb->pkt_type; |
ee5675ec | 2188 | if (unlikely(packet_sock_flag(po, PACKET_SOCK_ORIGDEV))) |
80feaacb PWJ |
2189 | sll->sll_ifindex = orig_dev->ifindex; |
2190 | else | |
2191 | sll->sll_ifindex = dev->ifindex; | |
1da177e4 | 2192 | |
b95cce35 | 2193 | sll->sll_halen = dev_parse_header(skb, sll->sll_addr); |
1da177e4 | 2194 | |
2472d761 EB |
2195 | /* sll->sll_family and sll->sll_protocol are set in packet_recvmsg(). |
2196 | * Use their space for storing the original skb length. | |
2197 | */ | |
2198 | PACKET_SKB_CB(skb)->sa.origlen = skb->len; | |
8dc41944 | 2199 | |
1da177e4 LT |
2200 | if (pskb_trim(skb, snaplen)) |
2201 | goto drop_n_acct; | |
2202 | ||
2203 | skb_set_owner_r(skb, sk); | |
2204 | skb->dev = NULL; | |
adf30907 | 2205 | skb_dst_drop(skb); |
1da177e4 | 2206 | |
84531c24 | 2207 | /* drop conntrack reference */ |
895b5c9f | 2208 | nf_reset_ct(skb); |
84531c24 | 2209 | |
1da177e4 | 2210 | spin_lock(&sk->sk_receive_queue.lock); |
ee80fbf3 | 2211 | po->stats.stats1.tp_packets++; |
3bc3b96f | 2212 | sock_skb_set_dropcount(sk, skb); |
27942a15 | 2213 | skb_clear_delivery_time(skb); |
1da177e4 LT |
2214 | __skb_queue_tail(&sk->sk_receive_queue, skb); |
2215 | spin_unlock(&sk->sk_receive_queue.lock); | |
676d2369 | 2216 | sk->sk_data_ready(sk); |
1da177e4 LT |
2217 | return 0; |
2218 | ||
2219 | drop_n_acct: | |
da37845f | 2220 | is_drop_n_account = true; |
8e8e2951 | 2221 | atomic_inc(&po->tp_drops); |
7091fbd8 | 2222 | atomic_inc(&sk->sk_drops); |
1da177e4 LT |
2223 | |
2224 | drop_n_restore: | |
2225 | if (skb_head != skb->data && skb_shared(skb)) { | |
2226 | skb->data = skb_head; | |
2227 | skb->len = skb_len; | |
2228 | } | |
2229 | drop: | |
da37845f WJ |
2230 | if (!is_drop_n_account) |
2231 | consume_skb(skb); | |
2232 | else | |
2233 | kfree_skb(skb); | |
1da177e4 LT |
2234 | return 0; |
2235 | } | |
2236 | ||
40d4e3df ED |
2237 | static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev, |
2238 | struct packet_type *pt, struct net_device *orig_dev) | |
1da177e4 LT |
2239 | { |
2240 | struct sock *sk; | |
2241 | struct packet_sock *po; | |
2242 | struct sockaddr_ll *sll; | |
184f489e | 2243 | union tpacket_uhdr h; |
40d4e3df | 2244 | u8 *skb_head = skb->data; |
1da177e4 | 2245 | int skb_len = skb->len; |
dbcb5855 | 2246 | unsigned int snaplen, res; |
f6fb8f10 | 2247 | unsigned long status = TP_STATUS_USER; |
acf69c94 OC |
2248 | unsigned short macoff, hdrlen; |
2249 | unsigned int netoff; | |
1da177e4 | 2250 | struct sk_buff *copy_skb = NULL; |
d413fcb4 | 2251 | struct timespec64 ts; |
b9c32fb2 | 2252 | __u32 ts_status; |
da37845f | 2253 | bool is_drop_n_account = false; |
61fad681 | 2254 | unsigned int slot_id = 0; |
dfc39d40 | 2255 | int vnet_hdr_sz = 0; |
1da177e4 | 2256 | |
51846355 AW |
2257 | /* struct tpacket{2,3}_hdr is aligned to a multiple of TPACKET_ALIGNMENT. |
2258 | * We may add members to them until current aligned size without forcing | |
2259 | * userspace to call getsockopt(..., PACKET_HDRLEN, ...). | |
2260 | */ | |
2261 | BUILD_BUG_ON(TPACKET_ALIGN(sizeof(*h.h2)) != 32); | |
2262 | BUILD_BUG_ON(TPACKET_ALIGN(sizeof(*h.h3)) != 48); | |
2263 | ||
1da177e4 LT |
2264 | if (skb->pkt_type == PACKET_LOOPBACK) |
2265 | goto drop; | |
2266 | ||
2267 | sk = pt->af_packet_priv; | |
2268 | po = pkt_sk(sk); | |
2269 | ||
09ad9bc7 | 2270 | if (!net_eq(dev_net(dev), sock_net(sk))) |
d12d01d6 DL |
2271 | goto drop; |
2272 | ||
d5496990 | 2273 | if (dev_has_header(dev)) { |
1da177e4 | 2274 | if (sk->sk_type != SOCK_DGRAM) |
98e399f8 | 2275 | skb_push(skb, skb->data - skb_mac_header(skb)); |
1da177e4 LT |
2276 | else if (skb->pkt_type == PACKET_OUTGOING) { |
2277 | /* Special case: outgoing packets have ll header at head */ | |
bbe735e4 | 2278 | skb_pull(skb, skb_network_offset(skb)); |
1da177e4 LT |
2279 | } |
2280 | } | |
2281 | ||
2282 | snaplen = skb->len; | |
2283 | ||
dbcb5855 DM |
2284 | res = run_filter(skb, sk, snaplen); |
2285 | if (!res) | |
fda9ef5d | 2286 | goto drop_n_restore; |
68c2e5de | 2287 | |
2c51c627 ED |
2288 | /* If we are flooded, just give up */ |
2289 | if (__packet_rcv_has_room(po, skb) == ROOM_NONE) { | |
2290 | atomic_inc(&po->tp_drops); | |
2291 | goto drop_n_restore; | |
2292 | } | |
2293 | ||
68c2e5de AD |
2294 | if (skb->ip_summed == CHECKSUM_PARTIAL) |
2295 | status |= TP_STATUS_CSUMNOTREADY; | |
682f048b | 2296 | else if (skb->pkt_type != PACKET_OUTGOING && |
b85f628a | 2297 | skb_csum_unnecessary(skb)) |
682f048b | 2298 | status |= TP_STATUS_CSUM_VALID; |
8e08bb75 XL |
2299 | if (skb_is_gso(skb) && skb_is_gso_tcp(skb)) |
2300 | status |= TP_STATUS_GSO_TCP; | |
68c2e5de | 2301 | |
dbcb5855 DM |
2302 | if (snaplen > res) |
2303 | snaplen = res; | |
1da177e4 LT |
2304 | |
2305 | if (sk->sk_type == SOCK_DGRAM) { | |
8913336a PM |
2306 | macoff = netoff = TPACKET_ALIGN(po->tp_hdrlen) + 16 + |
2307 | po->tp_reserve; | |
1da177e4 | 2308 | } else { |
95c96174 | 2309 | unsigned int maclen = skb_network_offset(skb); |
bbd6ef87 | 2310 | netoff = TPACKET_ALIGN(po->tp_hdrlen + |
8913336a | 2311 | (maclen < 16 ? 16 : maclen)) + |
58d19b19 | 2312 | po->tp_reserve; |
dfc39d40 JT |
2313 | vnet_hdr_sz = READ_ONCE(po->vnet_hdr_sz); |
2314 | if (vnet_hdr_sz) | |
2315 | netoff += vnet_hdr_sz; | |
1da177e4 LT |
2316 | macoff = netoff - maclen; |
2317 | } | |
acf69c94 OC |
2318 | if (netoff > USHRT_MAX) { |
2319 | atomic_inc(&po->tp_drops); | |
2320 | goto drop_n_restore; | |
2321 | } | |
f6fb8f10 | 2322 | if (po->tp_version <= TPACKET_V2) { |
2323 | if (macoff + snaplen > po->rx_ring.frame_size) { | |
2324 | if (po->copy_thresh && | |
0fd7bac6 | 2325 | atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) { |
f6fb8f10 | 2326 | if (skb_shared(skb)) { |
2327 | copy_skb = skb_clone(skb, GFP_ATOMIC); | |
2328 | } else { | |
2329 | copy_skb = skb_get(skb); | |
2330 | skb_head = skb->data; | |
2331 | } | |
c700525f ED |
2332 | if (copy_skb) { |
2333 | memset(&PACKET_SKB_CB(copy_skb)->sa.ll, 0, | |
2334 | sizeof(PACKET_SKB_CB(copy_skb)->sa.ll)); | |
f6fb8f10 | 2335 | skb_set_owner_r(copy_skb, sk); |
c700525f | 2336 | } |
1da177e4 | 2337 | } |
f6fb8f10 | 2338 | snaplen = po->rx_ring.frame_size - macoff; |
edbd58be | 2339 | if ((int)snaplen < 0) { |
f6fb8f10 | 2340 | snaplen = 0; |
dfc39d40 | 2341 | vnet_hdr_sz = 0; |
edbd58be | 2342 | } |
1da177e4 | 2343 | } |
dc808110 ED |
2344 | } else if (unlikely(macoff + snaplen > |
2345 | GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len)) { | |
2346 | u32 nval; | |
2347 | ||
2348 | nval = GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len - macoff; | |
2349 | pr_err_once("tpacket_rcv: packet too big, clamped from %u to %u. macoff=%u\n", | |
2350 | snaplen, nval, macoff); | |
2351 | snaplen = nval; | |
2352 | if (unlikely((int)snaplen < 0)) { | |
2353 | snaplen = 0; | |
2354 | macoff = GET_PBDQC_FROM_RB(&po->rx_ring)->max_frame_len; | |
dfc39d40 | 2355 | vnet_hdr_sz = 0; |
dc808110 | 2356 | } |
1da177e4 | 2357 | } |
1da177e4 | 2358 | spin_lock(&sk->sk_receive_queue.lock); |
f6fb8f10 | 2359 | h.raw = packet_current_rx_frame(po, skb, |
2360 | TP_STATUS_KERNEL, (macoff+snaplen)); | |
bbd6ef87 | 2361 | if (!h.raw) |
58d19b19 | 2362 | goto drop_n_account; |
46e4c421 | 2363 | |
61fad681 WB |
2364 | if (po->tp_version <= TPACKET_V2) { |
2365 | slot_id = po->rx_ring.head; | |
2366 | if (test_bit(slot_id, po->rx_ring.rx_owner_map)) | |
2367 | goto drop_n_account; | |
2368 | __set_bit(slot_id, po->rx_ring.rx_owner_map); | |
2369 | } | |
2370 | ||
dfc39d40 | 2371 | if (vnet_hdr_sz && |
46e4c421 WB |
2372 | virtio_net_hdr_from_skb(skb, h.raw + macoff - |
2373 | sizeof(struct virtio_net_hdr), | |
88fd1cb8 JO |
2374 | vio_le(), true, 0)) { |
2375 | if (po->tp_version == TPACKET_V3) | |
2376 | prb_clear_blk_fill_status(&po->rx_ring); | |
46e4c421 | 2377 | goto drop_n_account; |
88fd1cb8 | 2378 | } |
46e4c421 | 2379 | |
f6fb8f10 | 2380 | if (po->tp_version <= TPACKET_V2) { |
2381 | packet_increment_rx_head(po, &po->rx_ring); | |
2382 | /* | |
2383 | * LOSING will be reported till you read the stats, | |
2384 | * because it's COR - Clear On Read. | |
2385 | * Anyways, moving it for V1/V2 only as V3 doesn't need this | |
2386 | * at packet level. | |
2387 | */ | |
8e8e2951 | 2388 | if (atomic_read(&po->tp_drops)) |
f6fb8f10 | 2389 | status |= TP_STATUS_LOSING; |
2390 | } | |
945d015e | 2391 | |
ee80fbf3 | 2392 | po->stats.stats1.tp_packets++; |
1da177e4 LT |
2393 | if (copy_skb) { |
2394 | status |= TP_STATUS_COPY; | |
27942a15 | 2395 | skb_clear_delivery_time(copy_skb); |
1da177e4 LT |
2396 | __skb_queue_tail(&sk->sk_receive_queue, copy_skb); |
2397 | } | |
1da177e4 LT |
2398 | spin_unlock(&sk->sk_receive_queue.lock); |
2399 | ||
bbd6ef87 | 2400 | skb_copy_bits(skb, 0, h.raw + macoff, snaplen); |
b9c32fb2 | 2401 | |
171c3b15 RS |
2402 | /* Always timestamp; prefer an existing software timestamp taken |
2403 | * closer to the time of capture. | |
2404 | */ | |
2405 | ts_status = tpacket_get_timestamp(skb, &ts, | |
1051ce4a ED |
2406 | READ_ONCE(po->tp_tstamp) | |
2407 | SOF_TIMESTAMPING_SOFTWARE); | |
171c3b15 | 2408 | if (!ts_status) |
d413fcb4 | 2409 | ktime_get_real_ts64(&ts); |
1da177e4 | 2410 | |
b9c32fb2 DB |
2411 | status |= ts_status; |
2412 | ||
bbd6ef87 PM |
2413 | switch (po->tp_version) { |
2414 | case TPACKET_V1: | |
2415 | h.h1->tp_len = skb->len; | |
2416 | h.h1->tp_snaplen = snaplen; | |
2417 | h.h1->tp_mac = macoff; | |
2418 | h.h1->tp_net = netoff; | |
4b457bdf DB |
2419 | h.h1->tp_sec = ts.tv_sec; |
2420 | h.h1->tp_usec = ts.tv_nsec / NSEC_PER_USEC; | |
bbd6ef87 PM |
2421 | hdrlen = sizeof(*h.h1); |
2422 | break; | |
2423 | case TPACKET_V2: | |
2424 | h.h2->tp_len = skb->len; | |
2425 | h.h2->tp_snaplen = snaplen; | |
2426 | h.h2->tp_mac = macoff; | |
2427 | h.h2->tp_net = netoff; | |
bbd6ef87 PM |
2428 | h.h2->tp_sec = ts.tv_sec; |
2429 | h.h2->tp_nsec = ts.tv_nsec; | |
df8a39de JP |
2430 | if (skb_vlan_tag_present(skb)) { |
2431 | h.h2->tp_vlan_tci = skb_vlan_tag_get(skb); | |
a0cdfcf3 AW |
2432 | h.h2->tp_vlan_tpid = ntohs(skb->vlan_proto); |
2433 | status |= TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
a3bcc23e BG |
2434 | } else { |
2435 | h.h2->tp_vlan_tci = 0; | |
a0cdfcf3 | 2436 | h.h2->tp_vlan_tpid = 0; |
a3bcc23e | 2437 | } |
e4d26f4b | 2438 | memset(h.h2->tp_padding, 0, sizeof(h.h2->tp_padding)); |
bbd6ef87 PM |
2439 | hdrlen = sizeof(*h.h2); |
2440 | break; | |
f6fb8f10 | 2441 | case TPACKET_V3: |
2442 | /* tp_nxt_offset,vlan are already populated above. | |
2443 | * So DONT clear those fields here | |
2444 | */ | |
2445 | h.h3->tp_status |= status; | |
2446 | h.h3->tp_len = skb->len; | |
2447 | h.h3->tp_snaplen = snaplen; | |
2448 | h.h3->tp_mac = macoff; | |
2449 | h.h3->tp_net = netoff; | |
f6fb8f10 | 2450 | h.h3->tp_sec = ts.tv_sec; |
2451 | h.h3->tp_nsec = ts.tv_nsec; | |
e4d26f4b | 2452 | memset(h.h3->tp_padding, 0, sizeof(h.h3->tp_padding)); |
f6fb8f10 | 2453 | hdrlen = sizeof(*h.h3); |
2454 | break; | |
bbd6ef87 PM |
2455 | default: |
2456 | BUG(); | |
2457 | } | |
1da177e4 | 2458 | |
bbd6ef87 | 2459 | sll = h.raw + TPACKET_ALIGN(hdrlen); |
b95cce35 | 2460 | sll->sll_halen = dev_parse_header(skb, sll->sll_addr); |
1da177e4 LT |
2461 | sll->sll_family = AF_PACKET; |
2462 | sll->sll_hatype = dev->type; | |
2463 | sll->sll_protocol = skb->protocol; | |
2464 | sll->sll_pkttype = skb->pkt_type; | |
ee5675ec | 2465 | if (unlikely(packet_sock_flag(po, PACKET_SOCK_ORIGDEV))) |
80feaacb PWJ |
2466 | sll->sll_ifindex = orig_dev->ifindex; |
2467 | else | |
2468 | sll->sll_ifindex = dev->ifindex; | |
1da177e4 | 2469 | |
e16aa207 | 2470 | smp_mb(); |
f0d4eb29 | 2471 | |
f6dafa95 | 2472 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1 |
f0d4eb29 | 2473 | if (po->tp_version <= TPACKET_V2) { |
0af55bb5 CG |
2474 | u8 *start, *end; |
2475 | ||
f0d4eb29 DB |
2476 | end = (u8 *) PAGE_ALIGN((unsigned long) h.raw + |
2477 | macoff + snaplen); | |
2478 | ||
2479 | for (start = h.raw; start < end; start += PAGE_SIZE) | |
2480 | flush_dcache_page(pgv_to_page(start)); | |
1da177e4 | 2481 | } |
f0d4eb29 | 2482 | smp_wmb(); |
f6dafa95 | 2483 | #endif |
f0d4eb29 | 2484 | |
da413eec | 2485 | if (po->tp_version <= TPACKET_V2) { |
61fad681 | 2486 | spin_lock(&sk->sk_receive_queue.lock); |
f6fb8f10 | 2487 | __packet_set_status(po, h.raw, status); |
61fad681 WB |
2488 | __clear_bit(slot_id, po->rx_ring.rx_owner_map); |
2489 | spin_unlock(&sk->sk_receive_queue.lock); | |
da413eec | 2490 | sk->sk_data_ready(sk); |
88fd1cb8 | 2491 | } else if (po->tp_version == TPACKET_V3) { |
f6fb8f10 | 2492 | prb_clear_blk_fill_status(&po->rx_ring); |
da413eec | 2493 | } |
1da177e4 LT |
2494 | |
2495 | drop_n_restore: | |
2496 | if (skb_head != skb->data && skb_shared(skb)) { | |
2497 | skb->data = skb_head; | |
2498 | skb->len = skb_len; | |
2499 | } | |
2500 | drop: | |
da37845f WJ |
2501 | if (!is_drop_n_account) |
2502 | consume_skb(skb); | |
2503 | else | |
2504 | kfree_skb(skb); | |
1da177e4 LT |
2505 | return 0; |
2506 | ||
58d19b19 | 2507 | drop_n_account: |
1da177e4 | 2508 | spin_unlock(&sk->sk_receive_queue.lock); |
8e8e2951 ED |
2509 | atomic_inc(&po->tp_drops); |
2510 | is_drop_n_account = true; | |
1da177e4 | 2511 | |
676d2369 | 2512 | sk->sk_data_ready(sk); |
acb5d75b | 2513 | kfree_skb(copy_skb); |
1da177e4 LT |
2514 | goto drop_n_restore; |
2515 | } | |
2516 | ||
69e3c75f JB |
2517 | static void tpacket_destruct_skb(struct sk_buff *skb) |
2518 | { | |
2519 | struct packet_sock *po = pkt_sk(skb->sk); | |
1da177e4 | 2520 | |
69e3c75f | 2521 | if (likely(po->tx_ring.pg_vec)) { |
f0d4eb29 | 2522 | void *ph; |
b9c32fb2 DB |
2523 | __u32 ts; |
2524 | ||
5cd8d46e | 2525 | ph = skb_zcopy_get_nouarg(skb); |
b0138408 | 2526 | packet_dec_pending(&po->tx_ring); |
b9c32fb2 DB |
2527 | |
2528 | ts = __packet_set_timestamp(po, ph, skb); | |
2529 | __packet_set_status(po, ph, TP_STATUS_AVAILABLE | ts); | |
89ed5b51 NH |
2530 | |
2531 | if (!packet_read_pending(&po->tx_ring)) | |
2532 | complete(&po->skb_completion); | |
69e3c75f JB |
2533 | } |
2534 | ||
2535 | sock_wfree(skb); | |
2536 | } | |
2537 | ||
16cc1400 WB |
2538 | static int __packet_snd_vnet_parse(struct virtio_net_hdr *vnet_hdr, size_t len) |
2539 | { | |
16cc1400 WB |
2540 | if ((vnet_hdr->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && |
2541 | (__virtio16_to_cpu(vio_le(), vnet_hdr->csum_start) + | |
2542 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_offset) + 2 > | |
2543 | __virtio16_to_cpu(vio_le(), vnet_hdr->hdr_len))) | |
2544 | vnet_hdr->hdr_len = __cpu_to_virtio16(vio_le(), | |
2545 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_start) + | |
2546 | __virtio16_to_cpu(vio_le(), vnet_hdr->csum_offset) + 2); | |
2547 | ||
2548 | if (__virtio16_to_cpu(vio_le(), vnet_hdr->hdr_len) > len) | |
2549 | return -EINVAL; | |
2550 | ||
16cc1400 WB |
2551 | return 0; |
2552 | } | |
2553 | ||
2554 | static int packet_snd_vnet_parse(struct msghdr *msg, size_t *len, | |
dfc39d40 | 2555 | struct virtio_net_hdr *vnet_hdr, int vnet_hdr_sz) |
16cc1400 | 2556 | { |
dfc39d40 JT |
2557 | int ret; |
2558 | ||
2559 | if (*len < vnet_hdr_sz) | |
16cc1400 | 2560 | return -EINVAL; |
dfc39d40 | 2561 | *len -= vnet_hdr_sz; |
16cc1400 | 2562 | |
cbbd26b8 | 2563 | if (!copy_from_iter_full(vnet_hdr, sizeof(*vnet_hdr), &msg->msg_iter)) |
16cc1400 WB |
2564 | return -EFAULT; |
2565 | ||
dfc39d40 JT |
2566 | ret = __packet_snd_vnet_parse(vnet_hdr, *len); |
2567 | if (ret) | |
2568 | return ret; | |
2569 | ||
2570 | /* move iter to point to the start of mac header */ | |
2571 | if (vnet_hdr_sz != sizeof(struct virtio_net_hdr)) | |
2572 | iov_iter_advance(&msg->msg_iter, vnet_hdr_sz - sizeof(struct virtio_net_hdr)); | |
2573 | ||
2574 | return 0; | |
16cc1400 WB |
2575 | } |
2576 | ||
40d4e3df | 2577 | static int tpacket_fill_skb(struct packet_sock *po, struct sk_buff *skb, |
8d39b4a6 | 2578 | void *frame, struct net_device *dev, void *data, int tp_len, |
c14ac945 SHY |
2579 | __be16 proto, unsigned char *addr, int hlen, int copylen, |
2580 | const struct sockcm_cookie *sockc) | |
69e3c75f | 2581 | { |
184f489e | 2582 | union tpacket_uhdr ph; |
8d39b4a6 | 2583 | int to_write, offset, len, nr_frags, len_max; |
69e3c75f JB |
2584 | struct socket *sock = po->sk.sk_socket; |
2585 | struct page *page; | |
69e3c75f JB |
2586 | int err; |
2587 | ||
2588 | ph.raw = frame; | |
2589 | ||
2590 | skb->protocol = proto; | |
2591 | skb->dev = dev; | |
8bf43be7 | 2592 | skb->priority = READ_ONCE(po->sk.sk_priority); |
3c5b4d69 | 2593 | skb->mark = READ_ONCE(po->sk.sk_mark); |
3d0ba8c0 | 2594 | skb->tstamp = sockc->transmit_time; |
8f932f76 | 2595 | skb_setup_tx_timestamp(skb, sockc->tsflags); |
5cd8d46e | 2596 | skb_zcopy_set_nouarg(skb, ph.raw); |
69e3c75f | 2597 | |
ae641949 | 2598 | skb_reserve(skb, hlen); |
69e3c75f | 2599 | skb_reset_network_header(skb); |
c1aad275 | 2600 | |
69e3c75f JB |
2601 | to_write = tp_len; |
2602 | ||
2603 | if (sock->type == SOCK_DGRAM) { | |
2604 | err = dev_hard_header(skb, dev, ntohs(proto), addr, | |
2605 | NULL, tp_len); | |
2606 | if (unlikely(err < 0)) | |
2607 | return -EINVAL; | |
1d036d25 | 2608 | } else if (copylen) { |
9ed988cd WB |
2609 | int hdrlen = min_t(int, copylen, tp_len); |
2610 | ||
69e3c75f | 2611 | skb_push(skb, dev->hard_header_len); |
1d036d25 | 2612 | skb_put(skb, copylen - dev->hard_header_len); |
9ed988cd | 2613 | err = skb_store_bits(skb, 0, data, hdrlen); |
69e3c75f JB |
2614 | if (unlikely(err)) |
2615 | return err; | |
9ed988cd WB |
2616 | if (!dev_validate_header(dev, skb->data, hdrlen)) |
2617 | return -EINVAL; | |
69e3c75f | 2618 | |
9ed988cd WB |
2619 | data += hdrlen; |
2620 | to_write -= hdrlen; | |
69e3c75f JB |
2621 | } |
2622 | ||
69e3c75f JB |
2623 | offset = offset_in_page(data); |
2624 | len_max = PAGE_SIZE - offset; | |
2625 | len = ((to_write > len_max) ? len_max : to_write); | |
2626 | ||
2627 | skb->data_len = to_write; | |
2628 | skb->len += to_write; | |
2629 | skb->truesize += to_write; | |
14afee4b | 2630 | refcount_add(to_write, &po->sk.sk_wmem_alloc); |
69e3c75f JB |
2631 | |
2632 | while (likely(to_write)) { | |
2633 | nr_frags = skb_shinfo(skb)->nr_frags; | |
2634 | ||
2635 | if (unlikely(nr_frags >= MAX_SKB_FRAGS)) { | |
3948b059 ED |
2636 | pr_err("Packet exceed the number of skb frags(%u)\n", |
2637 | (unsigned int)MAX_SKB_FRAGS); | |
69e3c75f JB |
2638 | return -EFAULT; |
2639 | } | |
2640 | ||
0af55bb5 CG |
2641 | page = pgv_to_page(data); |
2642 | data += len; | |
69e3c75f JB |
2643 | flush_dcache_page(page); |
2644 | get_page(page); | |
0af55bb5 | 2645 | skb_fill_page_desc(skb, nr_frags, page, offset, len); |
69e3c75f JB |
2646 | to_write -= len; |
2647 | offset = 0; | |
2648 | len_max = PAGE_SIZE; | |
2649 | len = ((to_write > len_max) ? len_max : to_write); | |
2650 | } | |
2651 | ||
75c65772 | 2652 | packet_parse_headers(skb, sock); |
efdfa2f7 | 2653 | |
69e3c75f JB |
2654 | return tp_len; |
2655 | } | |
2656 | ||
8d39b4a6 WB |
2657 | static int tpacket_parse_header(struct packet_sock *po, void *frame, |
2658 | int size_max, void **data) | |
2659 | { | |
2660 | union tpacket_uhdr ph; | |
2661 | int tp_len, off; | |
2662 | ||
2663 | ph.raw = frame; | |
2664 | ||
2665 | switch (po->tp_version) { | |
7f953ab2 SV |
2666 | case TPACKET_V3: |
2667 | if (ph.h3->tp_next_offset != 0) { | |
2668 | pr_warn_once("variable sized slot not supported"); | |
2669 | return -EINVAL; | |
2670 | } | |
2671 | tp_len = ph.h3->tp_len; | |
2672 | break; | |
8d39b4a6 WB |
2673 | case TPACKET_V2: |
2674 | tp_len = ph.h2->tp_len; | |
2675 | break; | |
2676 | default: | |
2677 | tp_len = ph.h1->tp_len; | |
2678 | break; | |
2679 | } | |
2680 | if (unlikely(tp_len > size_max)) { | |
2681 | pr_err("packet size is too long (%d > %d)\n", tp_len, size_max); | |
2682 | return -EMSGSIZE; | |
2683 | } | |
2684 | ||
74383446 | 2685 | if (unlikely(packet_sock_flag(po, PACKET_SOCK_TX_HAS_OFF))) { |
8d39b4a6 WB |
2686 | int off_min, off_max; |
2687 | ||
2688 | off_min = po->tp_hdrlen - sizeof(struct sockaddr_ll); | |
2689 | off_max = po->tx_ring.frame_size - tp_len; | |
2690 | if (po->sk.sk_type == SOCK_DGRAM) { | |
2691 | switch (po->tp_version) { | |
7f953ab2 SV |
2692 | case TPACKET_V3: |
2693 | off = ph.h3->tp_net; | |
2694 | break; | |
8d39b4a6 WB |
2695 | case TPACKET_V2: |
2696 | off = ph.h2->tp_net; | |
2697 | break; | |
2698 | default: | |
2699 | off = ph.h1->tp_net; | |
2700 | break; | |
2701 | } | |
2702 | } else { | |
2703 | switch (po->tp_version) { | |
7f953ab2 SV |
2704 | case TPACKET_V3: |
2705 | off = ph.h3->tp_mac; | |
2706 | break; | |
8d39b4a6 WB |
2707 | case TPACKET_V2: |
2708 | off = ph.h2->tp_mac; | |
2709 | break; | |
2710 | default: | |
2711 | off = ph.h1->tp_mac; | |
2712 | break; | |
2713 | } | |
2714 | } | |
2715 | if (unlikely((off < off_min) || (off_max < off))) | |
2716 | return -EINVAL; | |
2717 | } else { | |
2718 | off = po->tp_hdrlen - sizeof(struct sockaddr_ll); | |
2719 | } | |
2720 | ||
2721 | *data = frame + off; | |
2722 | return tp_len; | |
2723 | } | |
2724 | ||
69e3c75f JB |
2725 | static int tpacket_snd(struct packet_sock *po, struct msghdr *msg) |
2726 | { | |
89ed5b51 | 2727 | struct sk_buff *skb = NULL; |
69e3c75f | 2728 | struct net_device *dev; |
1d036d25 | 2729 | struct virtio_net_hdr *vnet_hdr = NULL; |
c14ac945 | 2730 | struct sockcm_cookie sockc; |
69e3c75f | 2731 | __be16 proto; |
09effa67 | 2732 | int err, reserve = 0; |
40d4e3df | 2733 | void *ph; |
342dfc30 | 2734 | DECLARE_SOCKADDR(struct sockaddr_ll *, saddr, msg->msg_name); |
87a2fd28 | 2735 | bool need_wait = !(msg->msg_flags & MSG_DONTWAIT); |
dfc39d40 | 2736 | int vnet_hdr_sz = READ_ONCE(po->vnet_hdr_sz); |
486efdc8 | 2737 | unsigned char *addr = NULL; |
69e3c75f | 2738 | int tp_len, size_max; |
8d39b4a6 | 2739 | void *data; |
69e3c75f | 2740 | int len_sum = 0; |
9e67030a | 2741 | int status = TP_STATUS_AVAILABLE; |
1d036d25 | 2742 | int hlen, tlen, copylen = 0; |
89ed5b51 | 2743 | long timeo = 0; |
69e3c75f | 2744 | |
69e3c75f JB |
2745 | mutex_lock(&po->pg_vec_lock); |
2746 | ||
32d3182c ED |
2747 | /* packet_sendmsg() check on tx_ring.pg_vec was lockless, |
2748 | * we need to confirm it under protection of pg_vec_lock. | |
2749 | */ | |
2750 | if (unlikely(!po->tx_ring.pg_vec)) { | |
2751 | err = -EBUSY; | |
2752 | goto out; | |
2753 | } | |
66e56cd4 | 2754 | if (likely(saddr == NULL)) { |
e40526cb | 2755 | dev = packet_cached_dev_get(po); |
c7d2ef5d | 2756 | proto = READ_ONCE(po->num); |
69e3c75f JB |
2757 | } else { |
2758 | err = -EINVAL; | |
2759 | if (msg->msg_namelen < sizeof(struct sockaddr_ll)) | |
2760 | goto out; | |
2761 | if (msg->msg_namelen < (saddr->sll_halen | |
2762 | + offsetof(struct sockaddr_ll, | |
2763 | sll_addr))) | |
2764 | goto out; | |
69e3c75f | 2765 | proto = saddr->sll_protocol; |
827d9780 | 2766 | dev = dev_get_by_index(sock_net(&po->sk), saddr->sll_ifindex); |
486efdc8 WB |
2767 | if (po->sk.sk_socket->type == SOCK_DGRAM) { |
2768 | if (dev && msg->msg_namelen < dev->addr_len + | |
2769 | offsetof(struct sockaddr_ll, sll_addr)) | |
2770 | goto out_put; | |
2771 | addr = saddr->sll_addr; | |
2772 | } | |
69e3c75f JB |
2773 | } |
2774 | ||
69e3c75f JB |
2775 | err = -ENXIO; |
2776 | if (unlikely(dev == NULL)) | |
2777 | goto out; | |
69e3c75f JB |
2778 | err = -ENETDOWN; |
2779 | if (unlikely(!(dev->flags & IFF_UP))) | |
2780 | goto out_put; | |
2781 | ||
657a0667 | 2782 | sockcm_init(&sockc, &po->sk); |
d19b183c DCS |
2783 | if (msg->msg_controllen) { |
2784 | err = sock_cmsg_send(&po->sk, msg, &sockc); | |
2785 | if (unlikely(err)) | |
2786 | goto out_put; | |
2787 | } | |
2788 | ||
5cfb4c8d DB |
2789 | if (po->sk.sk_socket->type == SOCK_RAW) |
2790 | reserve = dev->hard_header_len; | |
69e3c75f | 2791 | size_max = po->tx_ring.frame_size |
b5dd884e | 2792 | - (po->tp_hdrlen - sizeof(struct sockaddr_ll)); |
69e3c75f | 2793 | |
dfc39d40 | 2794 | if ((size_max > dev->mtu + reserve + VLAN_HLEN) && !vnet_hdr_sz) |
5cfb4c8d | 2795 | size_max = dev->mtu + reserve + VLAN_HLEN; |
09effa67 | 2796 | |
89ed5b51 NH |
2797 | reinit_completion(&po->skb_completion); |
2798 | ||
69e3c75f JB |
2799 | do { |
2800 | ph = packet_current_frame(po, &po->tx_ring, | |
87a2fd28 | 2801 | TP_STATUS_SEND_REQUEST); |
69e3c75f | 2802 | if (unlikely(ph == NULL)) { |
89ed5b51 NH |
2803 | if (need_wait && skb) { |
2804 | timeo = sock_sndtimeo(&po->sk, msg->msg_flags & MSG_DONTWAIT); | |
2805 | timeo = wait_for_completion_interruptible_timeout(&po->skb_completion, timeo); | |
2806 | if (timeo <= 0) { | |
2807 | err = !timeo ? -ETIMEDOUT : -ERESTARTSYS; | |
2808 | goto out_put; | |
2809 | } | |
2810 | } | |
2811 | /* check for additional frames */ | |
69e3c75f JB |
2812 | continue; |
2813 | } | |
2814 | ||
8d39b4a6 WB |
2815 | skb = NULL; |
2816 | tp_len = tpacket_parse_header(po, ph, size_max, &data); | |
2817 | if (tp_len < 0) | |
2818 | goto tpacket_error; | |
2819 | ||
69e3c75f | 2820 | status = TP_STATUS_SEND_REQUEST; |
ae641949 HX |
2821 | hlen = LL_RESERVED_SPACE(dev); |
2822 | tlen = dev->needed_tailroom; | |
dfc39d40 | 2823 | if (vnet_hdr_sz) { |
1d036d25 | 2824 | vnet_hdr = data; |
dfc39d40 JT |
2825 | data += vnet_hdr_sz; |
2826 | tp_len -= vnet_hdr_sz; | |
1d036d25 WB |
2827 | if (tp_len < 0 || |
2828 | __packet_snd_vnet_parse(vnet_hdr, tp_len)) { | |
2829 | tp_len = -EINVAL; | |
2830 | goto tpacket_error; | |
2831 | } | |
2832 | copylen = __virtio16_to_cpu(vio_le(), | |
2833 | vnet_hdr->hdr_len); | |
2834 | } | |
9ed988cd | 2835 | copylen = max_t(int, copylen, dev->hard_header_len); |
69e3c75f | 2836 | skb = sock_alloc_send_skb(&po->sk, |
1d036d25 WB |
2837 | hlen + tlen + sizeof(struct sockaddr_ll) + |
2838 | (copylen - dev->hard_header_len), | |
fbf33a28 | 2839 | !need_wait, &err); |
69e3c75f | 2840 | |
fbf33a28 KM |
2841 | if (unlikely(skb == NULL)) { |
2842 | /* we assume the socket was initially writeable ... */ | |
2843 | if (likely(len_sum > 0)) | |
2844 | err = len_sum; | |
69e3c75f | 2845 | goto out_status; |
fbf33a28 | 2846 | } |
8d39b4a6 | 2847 | tp_len = tpacket_fill_skb(po, skb, ph, dev, data, tp_len, proto, |
c14ac945 | 2848 | addr, hlen, copylen, &sockc); |
dbd46ab4 | 2849 | if (likely(tp_len >= 0) && |
5cfb4c8d | 2850 | tp_len > dev->mtu + reserve && |
dfc39d40 | 2851 | !vnet_hdr_sz && |
3c70c132 DB |
2852 | !packet_extra_vlan_len_allowed(dev, skb)) |
2853 | tp_len = -EMSGSIZE; | |
69e3c75f JB |
2854 | |
2855 | if (unlikely(tp_len < 0)) { | |
8d39b4a6 | 2856 | tpacket_error: |
164bddac | 2857 | if (packet_sock_flag(po, PACKET_SOCK_TP_LOSS)) { |
69e3c75f JB |
2858 | __packet_set_status(po, ph, |
2859 | TP_STATUS_AVAILABLE); | |
2860 | packet_increment_head(&po->tx_ring); | |
2861 | kfree_skb(skb); | |
2862 | continue; | |
2863 | } else { | |
2864 | status = TP_STATUS_WRONG_FORMAT; | |
2865 | err = tp_len; | |
2866 | goto out_status; | |
2867 | } | |
2868 | } | |
2869 | ||
dfc39d40 | 2870 | if (vnet_hdr_sz) { |
9d2f67e4 JT |
2871 | if (virtio_net_hdr_to_skb(skb, vnet_hdr, vio_le())) { |
2872 | tp_len = -EINVAL; | |
2873 | goto tpacket_error; | |
2874 | } | |
2875 | virtio_net_hdr_set_proto(skb, vnet_hdr); | |
1d036d25 WB |
2876 | } |
2877 | ||
69e3c75f JB |
2878 | skb->destructor = tpacket_destruct_skb; |
2879 | __packet_set_status(po, ph, TP_STATUS_SENDING); | |
b0138408 | 2880 | packet_inc_pending(&po->tx_ring); |
69e3c75f JB |
2881 | |
2882 | status = TP_STATUS_SEND_REQUEST; | |
105a201e | 2883 | err = packet_xmit(po, skb); |
29e8e659 HL |
2884 | if (unlikely(err != 0)) { |
2885 | if (err > 0) | |
2886 | err = net_xmit_errno(err); | |
eb70df13 JP |
2887 | if (err && __packet_get_status(po, ph) == |
2888 | TP_STATUS_AVAILABLE) { | |
2889 | /* skb was destructed already */ | |
2890 | skb = NULL; | |
2891 | goto out_status; | |
2892 | } | |
2893 | /* | |
2894 | * skb was dropped but not destructed yet; | |
2895 | * let's treat it like congestion or err < 0 | |
2896 | */ | |
2897 | err = 0; | |
2898 | } | |
69e3c75f JB |
2899 | packet_increment_head(&po->tx_ring); |
2900 | len_sum += tp_len; | |
b0138408 DB |
2901 | } while (likely((ph != NULL) || |
2902 | /* Note: packet_read_pending() might be slow if we have | |
2903 | * to call it as it's per_cpu variable, but in fast-path | |
2904 | * we already short-circuit the loop with the first | |
2905 | * condition, and luckily don't have to go that path | |
2906 | * anyway. | |
2907 | */ | |
2908 | (need_wait && packet_read_pending(&po->tx_ring)))); | |
69e3c75f JB |
2909 | |
2910 | err = len_sum; | |
2911 | goto out_put; | |
2912 | ||
69e3c75f JB |
2913 | out_status: |
2914 | __packet_set_status(po, ph, status); | |
2915 | kfree_skb(skb); | |
2916 | out_put: | |
e40526cb | 2917 | dev_put(dev); |
69e3c75f JB |
2918 | out: |
2919 | mutex_unlock(&po->pg_vec_lock); | |
2920 | return err; | |
2921 | } | |
69e3c75f | 2922 | |
eea49cc9 OJ |
2923 | static struct sk_buff *packet_alloc_skb(struct sock *sk, size_t prepad, |
2924 | size_t reserve, size_t len, | |
2925 | size_t linear, int noblock, | |
2926 | int *err) | |
bfd5f4a3 SS |
2927 | { |
2928 | struct sk_buff *skb; | |
2929 | ||
2930 | /* Under a page? Don't bother with paged skb. */ | |
2931 | if (prepad + len < PAGE_SIZE || !linear) | |
2932 | linear = len; | |
2933 | ||
ae6db08f ED |
2934 | if (len - linear > MAX_SKB_FRAGS * (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) |
2935 | linear = len - MAX_SKB_FRAGS * (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER); | |
bfd5f4a3 | 2936 | skb = sock_alloc_send_pskb(sk, prepad + linear, len - linear, noblock, |
ae6db08f | 2937 | err, PAGE_ALLOC_COSTLY_ORDER); |
bfd5f4a3 SS |
2938 | if (!skb) |
2939 | return NULL; | |
2940 | ||
2941 | skb_reserve(skb, reserve); | |
2942 | skb_put(skb, linear); | |
2943 | skb->data_len = len - linear; | |
2944 | skb->len += len - linear; | |
2945 | ||
2946 | return skb; | |
2947 | } | |
2948 | ||
d346a3fa | 2949 | static int packet_snd(struct socket *sock, struct msghdr *msg, size_t len) |
1da177e4 LT |
2950 | { |
2951 | struct sock *sk = sock->sk; | |
342dfc30 | 2952 | DECLARE_SOCKADDR(struct sockaddr_ll *, saddr, msg->msg_name); |
1da177e4 LT |
2953 | struct sk_buff *skb; |
2954 | struct net_device *dev; | |
0e11c91e | 2955 | __be16 proto; |
486efdc8 | 2956 | unsigned char *addr = NULL; |
827d9780 | 2957 | int err, reserve = 0; |
c7d39e32 | 2958 | struct sockcm_cookie sockc; |
bfd5f4a3 SS |
2959 | struct virtio_net_hdr vnet_hdr = { 0 }; |
2960 | int offset = 0; | |
bfd5f4a3 | 2961 | struct packet_sock *po = pkt_sk(sk); |
dfc39d40 | 2962 | int vnet_hdr_sz = READ_ONCE(po->vnet_hdr_sz); |
57031eb7 | 2963 | int hlen, tlen, linear; |
3bdc0eba | 2964 | int extra_len = 0; |
1da177e4 LT |
2965 | |
2966 | /* | |
1ce4f28b | 2967 | * Get and verify the address. |
1da177e4 | 2968 | */ |
1ce4f28b | 2969 | |
66e56cd4 | 2970 | if (likely(saddr == NULL)) { |
e40526cb | 2971 | dev = packet_cached_dev_get(po); |
c7d2ef5d | 2972 | proto = READ_ONCE(po->num); |
1da177e4 LT |
2973 | } else { |
2974 | err = -EINVAL; | |
2975 | if (msg->msg_namelen < sizeof(struct sockaddr_ll)) | |
2976 | goto out; | |
0fb375fb EB |
2977 | if (msg->msg_namelen < (saddr->sll_halen + offsetof(struct sockaddr_ll, sll_addr))) |
2978 | goto out; | |
1da177e4 | 2979 | proto = saddr->sll_protocol; |
827d9780 | 2980 | dev = dev_get_by_index(sock_net(sk), saddr->sll_ifindex); |
486efdc8 WB |
2981 | if (sock->type == SOCK_DGRAM) { |
2982 | if (dev && msg->msg_namelen < dev->addr_len + | |
2983 | offsetof(struct sockaddr_ll, sll_addr)) | |
2984 | goto out_unlock; | |
2985 | addr = saddr->sll_addr; | |
2986 | } | |
1da177e4 LT |
2987 | } |
2988 | ||
1da177e4 | 2989 | err = -ENXIO; |
e40526cb | 2990 | if (unlikely(dev == NULL)) |
1da177e4 | 2991 | goto out_unlock; |
d5e76b0a | 2992 | err = -ENETDOWN; |
e40526cb | 2993 | if (unlikely(!(dev->flags & IFF_UP))) |
d5e76b0a DM |
2994 | goto out_unlock; |
2995 | ||
657a0667 | 2996 | sockcm_init(&sockc, sk); |
3c5b4d69 | 2997 | sockc.mark = READ_ONCE(sk->sk_mark); |
c7d39e32 EJ |
2998 | if (msg->msg_controllen) { |
2999 | err = sock_cmsg_send(sk, msg, &sockc); | |
3000 | if (unlikely(err)) | |
3001 | goto out_unlock; | |
3002 | } | |
3003 | ||
e40526cb DB |
3004 | if (sock->type == SOCK_RAW) |
3005 | reserve = dev->hard_header_len; | |
dfc39d40 JT |
3006 | if (vnet_hdr_sz) { |
3007 | err = packet_snd_vnet_parse(msg, &len, &vnet_hdr, vnet_hdr_sz); | |
16cc1400 | 3008 | if (err) |
bfd5f4a3 | 3009 | goto out_unlock; |
bfd5f4a3 SS |
3010 | } |
3011 | ||
3bdc0eba BG |
3012 | if (unlikely(sock_flag(sk, SOCK_NOFCS))) { |
3013 | if (!netif_supports_nofcs(dev)) { | |
3014 | err = -EPROTONOSUPPORT; | |
3015 | goto out_unlock; | |
3016 | } | |
3017 | extra_len = 4; /* We're doing our own CRC */ | |
3018 | } | |
3019 | ||
1da177e4 | 3020 | err = -EMSGSIZE; |
16cc1400 WB |
3021 | if (!vnet_hdr.gso_type && |
3022 | (len > dev->mtu + reserve + VLAN_HLEN + extra_len)) | |
1da177e4 LT |
3023 | goto out_unlock; |
3024 | ||
bfd5f4a3 | 3025 | err = -ENOBUFS; |
ae641949 HX |
3026 | hlen = LL_RESERVED_SPACE(dev); |
3027 | tlen = dev->needed_tailroom; | |
57031eb7 WB |
3028 | linear = __virtio16_to_cpu(vio_le(), vnet_hdr.hdr_len); |
3029 | linear = max(linear, min_t(int, len, dev->hard_header_len)); | |
3030 | skb = packet_alloc_skb(sk, hlen + tlen, hlen, len, linear, | |
bfd5f4a3 | 3031 | msg->msg_flags & MSG_DONTWAIT, &err); |
40d4e3df | 3032 | if (skb == NULL) |
1da177e4 LT |
3033 | goto out_unlock; |
3034 | ||
b84bbaf7 | 3035 | skb_reset_network_header(skb); |
1da177e4 | 3036 | |
0c4e8581 | 3037 | err = -EINVAL; |
9c707762 WB |
3038 | if (sock->type == SOCK_DGRAM) { |
3039 | offset = dev_hard_header(skb, dev, ntohs(proto), addr, NULL, len); | |
46d2cfb1 | 3040 | if (unlikely(offset < 0)) |
9c707762 | 3041 | goto out_free; |
b84bbaf7 | 3042 | } else if (reserve) { |
9aad13b0 | 3043 | skb_reserve(skb, -reserve); |
88a8121d ND |
3044 | if (len < reserve + sizeof(struct ipv6hdr) && |
3045 | dev->min_header_len != dev->hard_header_len) | |
993675a3 | 3046 | skb_reset_network_header(skb); |
9c707762 | 3047 | } |
1da177e4 LT |
3048 | |
3049 | /* Returns -EFAULT on error */ | |
c0371da6 | 3050 | err = skb_copy_datagram_from_iter(skb, offset, &msg->msg_iter, len); |
1da177e4 LT |
3051 | if (err) |
3052 | goto out_free; | |
bf84a010 | 3053 | |
dc633700 ZS |
3054 | if ((sock->type == SOCK_RAW && |
3055 | !dev_validate_header(dev, skb->data, len)) || !skb->len) { | |
9ed988cd WB |
3056 | err = -EINVAL; |
3057 | goto out_free; | |
3058 | } | |
3059 | ||
8f932f76 | 3060 | skb_setup_tx_timestamp(skb, sockc.tsflags); |
1da177e4 | 3061 | |
16cc1400 | 3062 | if (!vnet_hdr.gso_type && (len > dev->mtu + reserve + extra_len) && |
3c70c132 DB |
3063 | !packet_extra_vlan_len_allowed(dev, skb)) { |
3064 | err = -EMSGSIZE; | |
3065 | goto out_free; | |
57f89bfa BG |
3066 | } |
3067 | ||
09effa67 DM |
3068 | skb->protocol = proto; |
3069 | skb->dev = dev; | |
8bf43be7 | 3070 | skb->priority = READ_ONCE(sk->sk_priority); |
c7d39e32 | 3071 | skb->mark = sockc.mark; |
3d0ba8c0 | 3072 | skb->tstamp = sockc.transmit_time; |
0fd5d57b | 3073 | |
dfed913e HL |
3074 | if (unlikely(extra_len == 4)) |
3075 | skb->no_fcs = 1; | |
3076 | ||
3077 | packet_parse_headers(skb, sock); | |
3078 | ||
dfc39d40 | 3079 | if (vnet_hdr_sz) { |
db60eb5f | 3080 | err = virtio_net_hdr_to_skb(skb, &vnet_hdr, vio_le()); |
16cc1400 WB |
3081 | if (err) |
3082 | goto out_free; | |
dfc39d40 | 3083 | len += vnet_hdr_sz; |
9d2f67e4 | 3084 | virtio_net_hdr_set_proto(skb, &vnet_hdr); |
bfd5f4a3 SS |
3085 | } |
3086 | ||
105a201e ED |
3087 | err = packet_xmit(po, skb); |
3088 | ||
29e8e659 HL |
3089 | if (unlikely(err != 0)) { |
3090 | if (err > 0) | |
3091 | err = net_xmit_errno(err); | |
3092 | if (err) | |
3093 | goto out_unlock; | |
3094 | } | |
1da177e4 | 3095 | |
e40526cb | 3096 | dev_put(dev); |
1da177e4 | 3097 | |
40d4e3df | 3098 | return len; |
1da177e4 LT |
3099 | |
3100 | out_free: | |
3101 | kfree_skb(skb); | |
3102 | out_unlock: | |
1160dfa1 | 3103 | dev_put(dev); |
1da177e4 LT |
3104 | out: |
3105 | return err; | |
3106 | } | |
3107 | ||
1b784140 | 3108 | static int packet_sendmsg(struct socket *sock, struct msghdr *msg, size_t len) |
69e3c75f | 3109 | { |
69e3c75f JB |
3110 | struct sock *sk = sock->sk; |
3111 | struct packet_sock *po = pkt_sk(sk); | |
d346a3fa | 3112 | |
d1b5bee4 ED |
3113 | /* Reading tx_ring.pg_vec without holding pg_vec_lock is racy. |
3114 | * tpacket_snd() will redo the check safely. | |
3115 | */ | |
3116 | if (data_race(po->tx_ring.pg_vec)) | |
69e3c75f | 3117 | return tpacket_snd(po, msg); |
d1b5bee4 ED |
3118 | |
3119 | return packet_snd(sock, msg, len); | |
69e3c75f JB |
3120 | } |
3121 | ||
1da177e4 LT |
3122 | /* |
3123 | * Close a PACKET socket. This is fairly simple. We immediately go | |
3124 | * to 'closed' state and remove our protocol entry in the device list. | |
3125 | */ | |
3126 | ||
3127 | static int packet_release(struct socket *sock) | |
3128 | { | |
3129 | struct sock *sk = sock->sk; | |
3130 | struct packet_sock *po; | |
2bd624b4 | 3131 | struct packet_fanout *f; |
d12d01d6 | 3132 | struct net *net; |
f6fb8f10 | 3133 | union tpacket_req_u req_u; |
1da177e4 LT |
3134 | |
3135 | if (!sk) | |
3136 | return 0; | |
3137 | ||
3b1e0a65 | 3138 | net = sock_net(sk); |
1da177e4 LT |
3139 | po = pkt_sk(sk); |
3140 | ||
0fa7fa98 | 3141 | mutex_lock(&net->packet.sklist_lock); |
808f5114 | 3142 | sk_del_node_init_rcu(sk); |
0fa7fa98 PE |
3143 | mutex_unlock(&net->packet.sklist_lock); |
3144 | ||
920de804 | 3145 | sock_prot_inuse_add(net, sk->sk_prot, -1); |
1da177e4 | 3146 | |
808f5114 | 3147 | spin_lock(&po->bind_lock); |
ce06b03e | 3148 | unregister_prot_hook(sk, false); |
66e56cd4 DB |
3149 | packet_cached_dev_reset(po); |
3150 | ||
160ff18a | 3151 | if (po->prot_hook.dev) { |
d62607c3 | 3152 | netdev_put(po->prot_hook.dev, &po->prot_hook.dev_tracker); |
160ff18a BG |
3153 | po->prot_hook.dev = NULL; |
3154 | } | |
808f5114 | 3155 | spin_unlock(&po->bind_lock); |
1da177e4 | 3156 | |
1da177e4 | 3157 | packet_flush_mclist(sk); |
1da177e4 | 3158 | |
5171b37d | 3159 | lock_sock(sk); |
9665d5d6 PS |
3160 | if (po->rx_ring.pg_vec) { |
3161 | memset(&req_u, 0, sizeof(req_u)); | |
f6fb8f10 | 3162 | packet_set_ring(sk, &req_u, 1, 0); |
9665d5d6 | 3163 | } |
69e3c75f | 3164 | |
9665d5d6 PS |
3165 | if (po->tx_ring.pg_vec) { |
3166 | memset(&req_u, 0, sizeof(req_u)); | |
f6fb8f10 | 3167 | packet_set_ring(sk, &req_u, 1, 1); |
9665d5d6 | 3168 | } |
5171b37d | 3169 | release_sock(sk); |
1da177e4 | 3170 | |
2bd624b4 | 3171 | f = fanout_release(sk); |
dc99f600 | 3172 | |
808f5114 | 3173 | synchronize_net(); |
2bd624b4 | 3174 | |
afa0925c | 3175 | kfree(po->rollover); |
2bd624b4 AS |
3176 | if (f) { |
3177 | fanout_release_data(f); | |
9c661b0b | 3178 | kvfree(f); |
2bd624b4 | 3179 | } |
1da177e4 LT |
3180 | /* |
3181 | * Now the socket is dead. No more input will appear. | |
3182 | */ | |
1da177e4 LT |
3183 | sock_orphan(sk); |
3184 | sock->sk = NULL; | |
3185 | ||
3186 | /* Purge queues */ | |
3187 | ||
3188 | skb_queue_purge(&sk->sk_receive_queue); | |
b0138408 | 3189 | packet_free_pending(po); |
1da177e4 LT |
3190 | |
3191 | sock_put(sk); | |
3192 | return 0; | |
3193 | } | |
3194 | ||
3195 | /* | |
3196 | * Attach a packet hook. | |
3197 | */ | |
3198 | ||
30f7ea1c FR |
3199 | static int packet_do_bind(struct sock *sk, const char *name, int ifindex, |
3200 | __be16 proto) | |
1da177e4 LT |
3201 | { |
3202 | struct packet_sock *po = pkt_sk(sk); | |
30f7ea1c | 3203 | struct net_device *dev = NULL; |
30f7ea1c | 3204 | bool unlisted = false; |
bf44077c ED |
3205 | bool need_rehook; |
3206 | int ret = 0; | |
dc99f600 | 3207 | |
1da177e4 | 3208 | lock_sock(sk); |
1da177e4 | 3209 | spin_lock(&po->bind_lock); |
6ffc57ea ED |
3210 | if (!proto) |
3211 | proto = po->num; | |
3212 | ||
30f7ea1c FR |
3213 | rcu_read_lock(); |
3214 | ||
4971613c WB |
3215 | if (po->fanout) { |
3216 | ret = -EINVAL; | |
3217 | goto out_unlock; | |
3218 | } | |
3219 | ||
30f7ea1c FR |
3220 | if (name) { |
3221 | dev = dev_get_by_name_rcu(sock_net(sk), name); | |
3222 | if (!dev) { | |
3223 | ret = -ENODEV; | |
3224 | goto out_unlock; | |
3225 | } | |
3226 | } else if (ifindex) { | |
3227 | dev = dev_get_by_index_rcu(sock_net(sk), ifindex); | |
3228 | if (!dev) { | |
3229 | ret = -ENODEV; | |
3230 | goto out_unlock; | |
3231 | } | |
3232 | } | |
3233 | ||
bf44077c | 3234 | need_rehook = po->prot_hook.type != proto || po->prot_hook.dev != dev; |
902fefb8 DB |
3235 | |
3236 | if (need_rehook) { | |
bf44077c | 3237 | dev_hold(dev); |
61edf479 | 3238 | if (packet_sock_flag(po, PACKET_SOCK_RUNNING)) { |
30f7ea1c | 3239 | rcu_read_unlock(); |
15fe076e ED |
3240 | /* prevents packet_notifier() from calling |
3241 | * register_prot_hook() | |
3242 | */ | |
c7d2ef5d | 3243 | WRITE_ONCE(po->num, 0); |
30f7ea1c FR |
3244 | __unregister_prot_hook(sk, true); |
3245 | rcu_read_lock(); | |
30f7ea1c FR |
3246 | if (dev) |
3247 | unlisted = !dev_get_by_index_rcu(sock_net(sk), | |
3248 | dev->ifindex); | |
3249 | } | |
1da177e4 | 3250 | |
61edf479 | 3251 | BUG_ON(packet_sock_flag(po, PACKET_SOCK_RUNNING)); |
c7d2ef5d | 3252 | WRITE_ONCE(po->num, proto); |
902fefb8 | 3253 | po->prot_hook.type = proto; |
902fefb8 | 3254 | |
d62607c3 | 3255 | netdev_put(po->prot_hook.dev, &po->prot_hook.dev_tracker); |
f1d9268e | 3256 | |
30f7ea1c | 3257 | if (unlikely(unlisted)) { |
30f7ea1c | 3258 | po->prot_hook.dev = NULL; |
e032f7c9 | 3259 | WRITE_ONCE(po->ifindex, -1); |
30f7ea1c FR |
3260 | packet_cached_dev_reset(po); |
3261 | } else { | |
d62607c3 JK |
3262 | netdev_hold(dev, &po->prot_hook.dev_tracker, |
3263 | GFP_ATOMIC); | |
30f7ea1c | 3264 | po->prot_hook.dev = dev; |
e032f7c9 | 3265 | WRITE_ONCE(po->ifindex, dev ? dev->ifindex : 0); |
30f7ea1c FR |
3266 | packet_cached_dev_assign(po, dev); |
3267 | } | |
bf44077c | 3268 | dev_put(dev); |
902fefb8 | 3269 | } |
66e56cd4 | 3270 | |
902fefb8 | 3271 | if (proto == 0 || !need_rehook) |
1da177e4 LT |
3272 | goto out_unlock; |
3273 | ||
30f7ea1c | 3274 | if (!unlisted && (!dev || (dev->flags & IFF_UP))) { |
ce06b03e | 3275 | register_prot_hook(sk); |
be85d4ad UT |
3276 | } else { |
3277 | sk->sk_err = ENETDOWN; | |
3278 | if (!sock_flag(sk, SOCK_DEAD)) | |
e3ae2365 | 3279 | sk_error_report(sk); |
1da177e4 LT |
3280 | } |
3281 | ||
3282 | out_unlock: | |
30f7ea1c | 3283 | rcu_read_unlock(); |
1da177e4 LT |
3284 | spin_unlock(&po->bind_lock); |
3285 | release_sock(sk); | |
30f7ea1c | 3286 | return ret; |
1da177e4 LT |
3287 | } |
3288 | ||
3289 | /* | |
3290 | * Bind a packet socket to a device | |
3291 | */ | |
3292 | ||
40d4e3df ED |
3293 | static int packet_bind_spkt(struct socket *sock, struct sockaddr *uaddr, |
3294 | int addr_len) | |
1da177e4 | 3295 | { |
40d4e3df | 3296 | struct sock *sk = sock->sk; |
b5f0de6d | 3297 | char name[sizeof(uaddr->sa_data_min) + 1]; |
1ce4f28b | 3298 | |
1da177e4 LT |
3299 | /* |
3300 | * Check legality | |
3301 | */ | |
1ce4f28b | 3302 | |
8ae55f04 | 3303 | if (addr_len != sizeof(struct sockaddr)) |
1da177e4 | 3304 | return -EINVAL; |
540e2894 AP |
3305 | /* uaddr->sa_data comes from the userspace, it's not guaranteed to be |
3306 | * zero-terminated. | |
3307 | */ | |
b5f0de6d KC |
3308 | memcpy(name, uaddr->sa_data, sizeof(uaddr->sa_data_min)); |
3309 | name[sizeof(uaddr->sa_data_min)] = 0; | |
1da177e4 | 3310 | |
6ffc57ea | 3311 | return packet_do_bind(sk, name, 0, 0); |
1da177e4 | 3312 | } |
1da177e4 LT |
3313 | |
3314 | static int packet_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) | |
3315 | { | |
40d4e3df ED |
3316 | struct sockaddr_ll *sll = (struct sockaddr_ll *)uaddr; |
3317 | struct sock *sk = sock->sk; | |
1da177e4 LT |
3318 | |
3319 | /* | |
3320 | * Check legality | |
3321 | */ | |
1ce4f28b | 3322 | |
1da177e4 LT |
3323 | if (addr_len < sizeof(struct sockaddr_ll)) |
3324 | return -EINVAL; | |
3325 | if (sll->sll_family != AF_PACKET) | |
3326 | return -EINVAL; | |
3327 | ||
6ffc57ea | 3328 | return packet_do_bind(sk, NULL, sll->sll_ifindex, sll->sll_protocol); |
1da177e4 LT |
3329 | } |
3330 | ||
3331 | static struct proto packet_proto = { | |
3332 | .name = "PACKET", | |
3333 | .owner = THIS_MODULE, | |
3334 | .obj_size = sizeof(struct packet_sock), | |
3335 | }; | |
3336 | ||
3337 | /* | |
1ce4f28b | 3338 | * Create a packet of type SOCK_PACKET. |
1da177e4 LT |
3339 | */ |
3340 | ||
3f378b68 EP |
3341 | static int packet_create(struct net *net, struct socket *sock, int protocol, |
3342 | int kern) | |
1da177e4 LT |
3343 | { |
3344 | struct sock *sk; | |
3345 | struct packet_sock *po; | |
0e11c91e | 3346 | __be16 proto = (__force __be16)protocol; /* weird, but documented */ |
1da177e4 LT |
3347 | int err; |
3348 | ||
df008c91 | 3349 | if (!ns_capable(net->user_ns, CAP_NET_RAW)) |
1da177e4 | 3350 | return -EPERM; |
be02097c DM |
3351 | if (sock->type != SOCK_DGRAM && sock->type != SOCK_RAW && |
3352 | sock->type != SOCK_PACKET) | |
1da177e4 LT |
3353 | return -ESOCKTNOSUPPORT; |
3354 | ||
3355 | sock->state = SS_UNCONNECTED; | |
3356 | ||
3357 | err = -ENOBUFS; | |
11aa9c28 | 3358 | sk = sk_alloc(net, PF_PACKET, GFP_KERNEL, &packet_proto, kern); |
1da177e4 LT |
3359 | if (sk == NULL) |
3360 | goto out; | |
3361 | ||
3362 | sock->ops = &packet_ops; | |
1da177e4 LT |
3363 | if (sock->type == SOCK_PACKET) |
3364 | sock->ops = &packet_ops_spkt; | |
be02097c | 3365 | |
1da177e4 LT |
3366 | sock_init_data(sock, sk); |
3367 | ||
3368 | po = pkt_sk(sk); | |
89ed5b51 | 3369 | init_completion(&po->skb_completion); |
1da177e4 | 3370 | sk->sk_family = PF_PACKET; |
0e11c91e | 3371 | po->num = proto; |
66e56cd4 | 3372 | |
b0138408 DB |
3373 | err = packet_alloc_pending(po); |
3374 | if (err) | |
3375 | goto out2; | |
3376 | ||
66e56cd4 | 3377 | packet_cached_dev_reset(po); |
1da177e4 LT |
3378 | |
3379 | sk->sk_destruct = packet_sock_destruct; | |
1da177e4 LT |
3380 | |
3381 | /* | |
3382 | * Attach a protocol block | |
3383 | */ | |
3384 | ||
3385 | spin_lock_init(&po->bind_lock); | |
905db440 | 3386 | mutex_init(&po->pg_vec_lock); |
0648ab70 | 3387 | po->rollover = NULL; |
1da177e4 | 3388 | po->prot_hook.func = packet_rcv; |
be02097c | 3389 | |
1da177e4 LT |
3390 | if (sock->type == SOCK_PACKET) |
3391 | po->prot_hook.func = packet_rcv_spkt; | |
be02097c | 3392 | |
1da177e4 | 3393 | po->prot_hook.af_packet_priv = sk; |
47934e06 | 3394 | po->prot_hook.af_packet_net = sock_net(sk); |
1da177e4 | 3395 | |
0e11c91e AV |
3396 | if (proto) { |
3397 | po->prot_hook.type = proto; | |
a6361f0c | 3398 | __register_prot_hook(sk); |
1da177e4 LT |
3399 | } |
3400 | ||
0fa7fa98 | 3401 | mutex_lock(&net->packet.sklist_lock); |
a4dc6a49 | 3402 | sk_add_node_tail_rcu(sk, &net->packet.sklist); |
0fa7fa98 PE |
3403 | mutex_unlock(&net->packet.sklist_lock); |
3404 | ||
3680453c | 3405 | sock_prot_inuse_add(net, &packet_proto, 1); |
808f5114 | 3406 | |
40d4e3df | 3407 | return 0; |
b0138408 DB |
3408 | out2: |
3409 | sk_free(sk); | |
1da177e4 LT |
3410 | out: |
3411 | return err; | |
3412 | } | |
3413 | ||
3414 | /* | |
3415 | * Pull a packet from our receive queue and hand it to the user. | |
3416 | * If necessary we block. | |
3417 | */ | |
3418 | ||
1b784140 YX |
3419 | static int packet_recvmsg(struct socket *sock, struct msghdr *msg, size_t len, |
3420 | int flags) | |
1da177e4 LT |
3421 | { |
3422 | struct sock *sk = sock->sk; | |
3423 | struct sk_buff *skb; | |
3424 | int copied, err; | |
dfc39d40 | 3425 | int vnet_hdr_len = READ_ONCE(pkt_sk(sk)->vnet_hdr_sz); |
2472d761 | 3426 | unsigned int origlen = 0; |
1da177e4 LT |
3427 | |
3428 | err = -EINVAL; | |
ed85b565 | 3429 | if (flags & ~(MSG_PEEK|MSG_DONTWAIT|MSG_TRUNC|MSG_CMSG_COMPAT|MSG_ERRQUEUE)) |
1da177e4 LT |
3430 | goto out; |
3431 | ||
3432 | #if 0 | |
3433 | /* What error should we return now? EUNATTACH? */ | |
3434 | if (pkt_sk(sk)->ifindex < 0) | |
3435 | return -ENODEV; | |
3436 | #endif | |
3437 | ||
ed85b565 | 3438 | if (flags & MSG_ERRQUEUE) { |
cb820f8e RC |
3439 | err = sock_recv_errqueue(sk, msg, len, |
3440 | SOL_PACKET, PACKET_TX_TIMESTAMP); | |
ed85b565 RC |
3441 | goto out; |
3442 | } | |
3443 | ||
1da177e4 LT |
3444 | /* |
3445 | * Call the generic datagram receiver. This handles all sorts | |
3446 | * of horrible races and re-entrancy so we can forget about it | |
3447 | * in the protocol layers. | |
3448 | * | |
3449 | * Now it will return ENETDOWN, if device have just gone down, | |
3450 | * but then it will block. | |
3451 | */ | |
3452 | ||
f4b41f06 | 3453 | skb = skb_recv_datagram(sk, flags, &err); |
1da177e4 LT |
3454 | |
3455 | /* | |
1ce4f28b | 3456 | * An error occurred so return it. Because skb_recv_datagram() |
1da177e4 LT |
3457 | * handles the blocking we don't see and worry about blocking |
3458 | * retries. | |
3459 | */ | |
3460 | ||
8ae55f04 | 3461 | if (skb == NULL) |
1da177e4 LT |
3462 | goto out; |
3463 | ||
9bb6cd65 | 3464 | packet_rcv_try_clear_pressure(pkt_sk(sk)); |
2ccdbaa6 | 3465 | |
dfc39d40 JT |
3466 | if (vnet_hdr_len) { |
3467 | err = packet_rcv_vnet(msg, skb, &len, vnet_hdr_len); | |
16cc1400 | 3468 | if (err) |
bfd5f4a3 SS |
3469 | goto out_free; |
3470 | } | |
3471 | ||
f3d33426 HFS |
3472 | /* You lose any data beyond the buffer you gave. If it worries |
3473 | * a user program they can ask the device for its MTU | |
3474 | * anyway. | |
1da177e4 | 3475 | */ |
1da177e4 | 3476 | copied = skb->len; |
40d4e3df ED |
3477 | if (copied > len) { |
3478 | copied = len; | |
3479 | msg->msg_flags |= MSG_TRUNC; | |
1da177e4 LT |
3480 | } |
3481 | ||
51f3d02b | 3482 | err = skb_copy_datagram_msg(skb, 0, msg, copied); |
1da177e4 LT |
3483 | if (err) |
3484 | goto out_free; | |
3485 | ||
2472d761 EB |
3486 | if (sock->type != SOCK_PACKET) { |
3487 | struct sockaddr_ll *sll = &PACKET_SKB_CB(skb)->sa.ll; | |
3488 | ||
3489 | /* Original length was stored in sockaddr_ll fields */ | |
3490 | origlen = PACKET_SKB_CB(skb)->sa.origlen; | |
3491 | sll->sll_family = AF_PACKET; | |
3492 | sll->sll_protocol = skb->protocol; | |
3493 | } | |
3494 | ||
6fd1d51c | 3495 | sock_recv_cmsgs(msg, sk, skb); |
1da177e4 | 3496 | |
f3d33426 | 3497 | if (msg->msg_name) { |
c700525f ED |
3498 | const size_t max_len = min(sizeof(skb->cb), |
3499 | sizeof(struct sockaddr_storage)); | |
b2cf86e1 WB |
3500 | int copy_len; |
3501 | ||
f3d33426 HFS |
3502 | /* If the address length field is there to be filled |
3503 | * in, we fill it in now. | |
3504 | */ | |
3505 | if (sock->type == SOCK_PACKET) { | |
342dfc30 | 3506 | __sockaddr_check_size(sizeof(struct sockaddr_pkt)); |
f3d33426 | 3507 | msg->msg_namelen = sizeof(struct sockaddr_pkt); |
b2cf86e1 | 3508 | copy_len = msg->msg_namelen; |
f3d33426 HFS |
3509 | } else { |
3510 | struct sockaddr_ll *sll = &PACKET_SKB_CB(skb)->sa.ll; | |
2472d761 | 3511 | |
f3d33426 HFS |
3512 | msg->msg_namelen = sll->sll_halen + |
3513 | offsetof(struct sockaddr_ll, sll_addr); | |
b2cf86e1 WB |
3514 | copy_len = msg->msg_namelen; |
3515 | if (msg->msg_namelen < sizeof(struct sockaddr_ll)) { | |
3516 | memset(msg->msg_name + | |
3517 | offsetof(struct sockaddr_ll, sll_addr), | |
3518 | 0, sizeof(sll->sll_addr)); | |
3519 | msg->msg_namelen = sizeof(struct sockaddr_ll); | |
3520 | } | |
f3d33426 | 3521 | } |
c700525f ED |
3522 | if (WARN_ON_ONCE(copy_len > max_len)) { |
3523 | copy_len = max_len; | |
3524 | msg->msg_namelen = copy_len; | |
3525 | } | |
b2cf86e1 | 3526 | memcpy(msg->msg_name, &PACKET_SKB_CB(skb)->sa, copy_len); |
f3d33426 | 3527 | } |
1da177e4 | 3528 | |
fd53c297 | 3529 | if (packet_sock_flag(pkt_sk(sk), PACKET_SOCK_AUXDATA)) { |
ffbc6111 HX |
3530 | struct tpacket_auxdata aux; |
3531 | ||
3532 | aux.tp_status = TP_STATUS_USER; | |
3533 | if (skb->ip_summed == CHECKSUM_PARTIAL) | |
3534 | aux.tp_status |= TP_STATUS_CSUMNOTREADY; | |
682f048b | 3535 | else if (skb->pkt_type != PACKET_OUTGOING && |
b85f628a | 3536 | skb_csum_unnecessary(skb)) |
682f048b | 3537 | aux.tp_status |= TP_STATUS_CSUM_VALID; |
8e08bb75 XL |
3538 | if (skb_is_gso(skb) && skb_is_gso_tcp(skb)) |
3539 | aux.tp_status |= TP_STATUS_GSO_TCP; | |
682f048b | 3540 | |
2472d761 | 3541 | aux.tp_len = origlen; |
ffbc6111 HX |
3542 | aux.tp_snaplen = skb->len; |
3543 | aux.tp_mac = 0; | |
bbe735e4 | 3544 | aux.tp_net = skb_network_offset(skb); |
df8a39de JP |
3545 | if (skb_vlan_tag_present(skb)) { |
3546 | aux.tp_vlan_tci = skb_vlan_tag_get(skb); | |
a0cdfcf3 AW |
3547 | aux.tp_vlan_tpid = ntohs(skb->vlan_proto); |
3548 | aux.tp_status |= TP_STATUS_VLAN_VALID | TP_STATUS_VLAN_TPID_VALID; | |
a3bcc23e BG |
3549 | } else { |
3550 | aux.tp_vlan_tci = 0; | |
a0cdfcf3 | 3551 | aux.tp_vlan_tpid = 0; |
a3bcc23e | 3552 | } |
ffbc6111 | 3553 | put_cmsg(msg, SOL_PACKET, PACKET_AUXDATA, sizeof(aux), &aux); |
8dc41944 HX |
3554 | } |
3555 | ||
1da177e4 LT |
3556 | /* |
3557 | * Free or return the buffer as appropriate. Again this | |
3558 | * hides all the races and re-entrancy issues from us. | |
3559 | */ | |
bfd5f4a3 | 3560 | err = vnet_hdr_len + ((flags&MSG_TRUNC) ? skb->len : copied); |
1da177e4 LT |
3561 | |
3562 | out_free: | |
3563 | skb_free_datagram(sk, skb); | |
3564 | out: | |
3565 | return err; | |
3566 | } | |
3567 | ||
1da177e4 | 3568 | static int packet_getname_spkt(struct socket *sock, struct sockaddr *uaddr, |
9b2c45d4 | 3569 | int peer) |
1da177e4 LT |
3570 | { |
3571 | struct net_device *dev; | |
3572 | struct sock *sk = sock->sk; | |
3573 | ||
3574 | if (peer) | |
3575 | return -EOPNOTSUPP; | |
3576 | ||
3577 | uaddr->sa_family = AF_PACKET; | |
b5f0de6d | 3578 | memset(uaddr->sa_data, 0, sizeof(uaddr->sa_data_min)); |
654d1f8a | 3579 | rcu_read_lock(); |
e032f7c9 | 3580 | dev = dev_get_by_index_rcu(sock_net(sk), READ_ONCE(pkt_sk(sk)->ifindex)); |
654d1f8a | 3581 | if (dev) |
b5f0de6d | 3582 | strscpy(uaddr->sa_data, dev->name, sizeof(uaddr->sa_data_min)); |
654d1f8a | 3583 | rcu_read_unlock(); |
1da177e4 | 3584 | |
9b2c45d4 | 3585 | return sizeof(*uaddr); |
1da177e4 | 3586 | } |
1da177e4 LT |
3587 | |
3588 | static int packet_getname(struct socket *sock, struct sockaddr *uaddr, | |
9b2c45d4 | 3589 | int peer) |
1da177e4 LT |
3590 | { |
3591 | struct net_device *dev; | |
3592 | struct sock *sk = sock->sk; | |
3593 | struct packet_sock *po = pkt_sk(sk); | |
13cfa97b | 3594 | DECLARE_SOCKADDR(struct sockaddr_ll *, sll, uaddr); |
e032f7c9 | 3595 | int ifindex; |
1da177e4 LT |
3596 | |
3597 | if (peer) | |
3598 | return -EOPNOTSUPP; | |
3599 | ||
e032f7c9 | 3600 | ifindex = READ_ONCE(po->ifindex); |
1da177e4 | 3601 | sll->sll_family = AF_PACKET; |
e032f7c9 | 3602 | sll->sll_ifindex = ifindex; |
c7d2ef5d | 3603 | sll->sll_protocol = READ_ONCE(po->num); |
67286640 | 3604 | sll->sll_pkttype = 0; |
654d1f8a | 3605 | rcu_read_lock(); |
e032f7c9 | 3606 | dev = dev_get_by_index_rcu(sock_net(sk), ifindex); |
1da177e4 LT |
3607 | if (dev) { |
3608 | sll->sll_hatype = dev->type; | |
3609 | sll->sll_halen = dev->addr_len; | |
e2bca487 KI |
3610 | |
3611 | /* Let __fortify_memcpy_chk() know the actual buffer size. */ | |
3612 | memcpy(((struct sockaddr_storage *)sll)->__data + | |
3613 | offsetof(struct sockaddr_ll, sll_addr) - | |
3614 | offsetofend(struct sockaddr_ll, sll_family), | |
3615 | dev->dev_addr, dev->addr_len); | |
1da177e4 LT |
3616 | } else { |
3617 | sll->sll_hatype = 0; /* Bad: we have no ARPHRD_UNSPEC */ | |
3618 | sll->sll_halen = 0; | |
3619 | } | |
654d1f8a | 3620 | rcu_read_unlock(); |
1da177e4 | 3621 | |
9b2c45d4 | 3622 | return offsetof(struct sockaddr_ll, sll_addr) + sll->sll_halen; |
1da177e4 LT |
3623 | } |
3624 | ||
2aeb0b88 WC |
3625 | static int packet_dev_mc(struct net_device *dev, struct packet_mclist *i, |
3626 | int what) | |
1da177e4 LT |
3627 | { |
3628 | switch (i->type) { | |
3629 | case PACKET_MR_MULTICAST: | |
1162563f JP |
3630 | if (i->alen != dev->addr_len) |
3631 | return -EINVAL; | |
1da177e4 | 3632 | if (what > 0) |
22bedad3 | 3633 | return dev_mc_add(dev, i->addr); |
1da177e4 | 3634 | else |
22bedad3 | 3635 | return dev_mc_del(dev, i->addr); |
1da177e4 LT |
3636 | break; |
3637 | case PACKET_MR_PROMISC: | |
2aeb0b88 | 3638 | return dev_set_promiscuity(dev, what); |
1da177e4 | 3639 | case PACKET_MR_ALLMULTI: |
2aeb0b88 | 3640 | return dev_set_allmulti(dev, what); |
d95ed927 | 3641 | case PACKET_MR_UNICAST: |
1162563f JP |
3642 | if (i->alen != dev->addr_len) |
3643 | return -EINVAL; | |
d95ed927 | 3644 | if (what > 0) |
a748ee24 | 3645 | return dev_uc_add(dev, i->addr); |
d95ed927 | 3646 | else |
a748ee24 | 3647 | return dev_uc_del(dev, i->addr); |
d95ed927 | 3648 | break; |
40d4e3df ED |
3649 | default: |
3650 | break; | |
1da177e4 | 3651 | } |
2aeb0b88 | 3652 | return 0; |
1da177e4 LT |
3653 | } |
3654 | ||
82f17091 FR |
3655 | static void packet_dev_mclist_delete(struct net_device *dev, |
3656 | struct packet_mclist **mlp) | |
1da177e4 | 3657 | { |
82f17091 FR |
3658 | struct packet_mclist *ml; |
3659 | ||
3660 | while ((ml = *mlp) != NULL) { | |
3661 | if (ml->ifindex == dev->ifindex) { | |
3662 | packet_dev_mc(dev, ml, -1); | |
3663 | *mlp = ml->next; | |
3664 | kfree(ml); | |
3665 | } else | |
3666 | mlp = &ml->next; | |
1da177e4 LT |
3667 | } |
3668 | } | |
3669 | ||
0fb375fb | 3670 | static int packet_mc_add(struct sock *sk, struct packet_mreq_max *mreq) |
1da177e4 LT |
3671 | { |
3672 | struct packet_sock *po = pkt_sk(sk); | |
3673 | struct packet_mclist *ml, *i; | |
3674 | struct net_device *dev; | |
3675 | int err; | |
3676 | ||
3677 | rtnl_lock(); | |
3678 | ||
3679 | err = -ENODEV; | |
3b1e0a65 | 3680 | dev = __dev_get_by_index(sock_net(sk), mreq->mr_ifindex); |
1da177e4 LT |
3681 | if (!dev) |
3682 | goto done; | |
3683 | ||
3684 | err = -EINVAL; | |
1162563f | 3685 | if (mreq->mr_alen > dev->addr_len) |
1da177e4 LT |
3686 | goto done; |
3687 | ||
3688 | err = -ENOBUFS; | |
8b3a7005 | 3689 | i = kmalloc(sizeof(*i), GFP_KERNEL); |
1da177e4 LT |
3690 | if (i == NULL) |
3691 | goto done; | |
3692 | ||
3693 | err = 0; | |
3694 | for (ml = po->mclist; ml; ml = ml->next) { | |
3695 | if (ml->ifindex == mreq->mr_ifindex && | |
3696 | ml->type == mreq->mr_type && | |
3697 | ml->alen == mreq->mr_alen && | |
3698 | memcmp(ml->addr, mreq->mr_address, ml->alen) == 0) { | |
3699 | ml->count++; | |
3700 | /* Free the new element ... */ | |
3701 | kfree(i); | |
3702 | goto done; | |
3703 | } | |
3704 | } | |
3705 | ||
3706 | i->type = mreq->mr_type; | |
3707 | i->ifindex = mreq->mr_ifindex; | |
3708 | i->alen = mreq->mr_alen; | |
3709 | memcpy(i->addr, mreq->mr_address, i->alen); | |
309cf37f | 3710 | memset(i->addr + i->alen, 0, sizeof(i->addr) - i->alen); |
1da177e4 LT |
3711 | i->count = 1; |
3712 | i->next = po->mclist; | |
3713 | po->mclist = i; | |
2aeb0b88 WC |
3714 | err = packet_dev_mc(dev, i, 1); |
3715 | if (err) { | |
3716 | po->mclist = i->next; | |
3717 | kfree(i); | |
3718 | } | |
1da177e4 LT |
3719 | |
3720 | done: | |
3721 | rtnl_unlock(); | |
3722 | return err; | |
3723 | } | |
3724 | ||
0fb375fb | 3725 | static int packet_mc_drop(struct sock *sk, struct packet_mreq_max *mreq) |
1da177e4 LT |
3726 | { |
3727 | struct packet_mclist *ml, **mlp; | |
3728 | ||
3729 | rtnl_lock(); | |
3730 | ||
3731 | for (mlp = &pkt_sk(sk)->mclist; (ml = *mlp) != NULL; mlp = &ml->next) { | |
3732 | if (ml->ifindex == mreq->mr_ifindex && | |
3733 | ml->type == mreq->mr_type && | |
3734 | ml->alen == mreq->mr_alen && | |
3735 | memcmp(ml->addr, mreq->mr_address, ml->alen) == 0) { | |
3736 | if (--ml->count == 0) { | |
3737 | struct net_device *dev; | |
3738 | *mlp = ml->next; | |
ad959e76 ED |
3739 | dev = __dev_get_by_index(sock_net(sk), ml->ifindex); |
3740 | if (dev) | |
1da177e4 | 3741 | packet_dev_mc(dev, ml, -1); |
1da177e4 LT |
3742 | kfree(ml); |
3743 | } | |
82f17091 | 3744 | break; |
1da177e4 LT |
3745 | } |
3746 | } | |
3747 | rtnl_unlock(); | |
82f17091 | 3748 | return 0; |
1da177e4 LT |
3749 | } |
3750 | ||
3751 | static void packet_flush_mclist(struct sock *sk) | |
3752 | { | |
3753 | struct packet_sock *po = pkt_sk(sk); | |
3754 | struct packet_mclist *ml; | |
3755 | ||
3756 | if (!po->mclist) | |
3757 | return; | |
3758 | ||
3759 | rtnl_lock(); | |
3760 | while ((ml = po->mclist) != NULL) { | |
3761 | struct net_device *dev; | |
3762 | ||
3763 | po->mclist = ml->next; | |
ad959e76 ED |
3764 | dev = __dev_get_by_index(sock_net(sk), ml->ifindex); |
3765 | if (dev != NULL) | |
1da177e4 | 3766 | packet_dev_mc(dev, ml, -1); |
1da177e4 LT |
3767 | kfree(ml); |
3768 | } | |
3769 | rtnl_unlock(); | |
3770 | } | |
1da177e4 LT |
3771 | |
3772 | static int | |
a7b75c5a CH |
3773 | packet_setsockopt(struct socket *sock, int level, int optname, sockptr_t optval, |
3774 | unsigned int optlen) | |
1da177e4 LT |
3775 | { |
3776 | struct sock *sk = sock->sk; | |
8dc41944 | 3777 | struct packet_sock *po = pkt_sk(sk); |
1da177e4 LT |
3778 | int ret; |
3779 | ||
3780 | if (level != SOL_PACKET) | |
3781 | return -ENOPROTOOPT; | |
3782 | ||
69e3c75f | 3783 | switch (optname) { |
1ce4f28b | 3784 | case PACKET_ADD_MEMBERSHIP: |
1da177e4 LT |
3785 | case PACKET_DROP_MEMBERSHIP: |
3786 | { | |
0fb375fb EB |
3787 | struct packet_mreq_max mreq; |
3788 | int len = optlen; | |
3789 | memset(&mreq, 0, sizeof(mreq)); | |
3790 | if (len < sizeof(struct packet_mreq)) | |
1da177e4 | 3791 | return -EINVAL; |
0fb375fb EB |
3792 | if (len > sizeof(mreq)) |
3793 | len = sizeof(mreq); | |
a7b75c5a | 3794 | if (copy_from_sockptr(&mreq, optval, len)) |
1da177e4 | 3795 | return -EFAULT; |
0fb375fb EB |
3796 | if (len < (mreq.mr_alen + offsetof(struct packet_mreq, mr_address))) |
3797 | return -EINVAL; | |
1da177e4 LT |
3798 | if (optname == PACKET_ADD_MEMBERSHIP) |
3799 | ret = packet_mc_add(sk, &mreq); | |
3800 | else | |
3801 | ret = packet_mc_drop(sk, &mreq); | |
3802 | return ret; | |
3803 | } | |
a2efcfa0 | 3804 | |
1da177e4 | 3805 | case PACKET_RX_RING: |
69e3c75f | 3806 | case PACKET_TX_RING: |
1da177e4 | 3807 | { |
f6fb8f10 | 3808 | union tpacket_req_u req_u; |
3809 | int len; | |
1da177e4 | 3810 | |
5171b37d | 3811 | lock_sock(sk); |
f6fb8f10 | 3812 | switch (po->tp_version) { |
3813 | case TPACKET_V1: | |
3814 | case TPACKET_V2: | |
3815 | len = sizeof(req_u.req); | |
3816 | break; | |
3817 | case TPACKET_V3: | |
3818 | default: | |
3819 | len = sizeof(req_u.req3); | |
3820 | break; | |
3821 | } | |
5171b37d ED |
3822 | if (optlen < len) { |
3823 | ret = -EINVAL; | |
3824 | } else { | |
a7b75c5a | 3825 | if (copy_from_sockptr(&req_u.req, optval, len)) |
5171b37d ED |
3826 | ret = -EFAULT; |
3827 | else | |
3828 | ret = packet_set_ring(sk, &req_u, 0, | |
3829 | optname == PACKET_TX_RING); | |
3830 | } | |
3831 | release_sock(sk); | |
3832 | return ret; | |
1da177e4 LT |
3833 | } |
3834 | case PACKET_COPY_THRESH: | |
3835 | { | |
3836 | int val; | |
3837 | ||
40d4e3df | 3838 | if (optlen != sizeof(val)) |
1da177e4 | 3839 | return -EINVAL; |
a7b75c5a | 3840 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
1da177e4 LT |
3841 | return -EFAULT; |
3842 | ||
3843 | pkt_sk(sk)->copy_thresh = val; | |
3844 | return 0; | |
3845 | } | |
bbd6ef87 PM |
3846 | case PACKET_VERSION: |
3847 | { | |
3848 | int val; | |
3849 | ||
3850 | if (optlen != sizeof(val)) | |
3851 | return -EINVAL; | |
a7b75c5a | 3852 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
bbd6ef87 PM |
3853 | return -EFAULT; |
3854 | switch (val) { | |
3855 | case TPACKET_V1: | |
3856 | case TPACKET_V2: | |
f6fb8f10 | 3857 | case TPACKET_V3: |
84ac7260 | 3858 | break; |
bbd6ef87 PM |
3859 | default: |
3860 | return -EINVAL; | |
3861 | } | |
84ac7260 PP |
3862 | lock_sock(sk); |
3863 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3864 | ret = -EBUSY; | |
3865 | } else { | |
3866 | po->tp_version = val; | |
3867 | ret = 0; | |
3868 | } | |
3869 | release_sock(sk); | |
3870 | return ret; | |
bbd6ef87 | 3871 | } |
8913336a PM |
3872 | case PACKET_RESERVE: |
3873 | { | |
3874 | unsigned int val; | |
3875 | ||
3876 | if (optlen != sizeof(val)) | |
3877 | return -EINVAL; | |
a7b75c5a | 3878 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
8913336a | 3879 | return -EFAULT; |
bcc5364b AK |
3880 | if (val > INT_MAX) |
3881 | return -EINVAL; | |
c27927e3 WB |
3882 | lock_sock(sk); |
3883 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3884 | ret = -EBUSY; | |
3885 | } else { | |
3886 | po->tp_reserve = val; | |
3887 | ret = 0; | |
3888 | } | |
3889 | release_sock(sk); | |
3890 | return ret; | |
8913336a | 3891 | } |
69e3c75f JB |
3892 | case PACKET_LOSS: |
3893 | { | |
3894 | unsigned int val; | |
3895 | ||
3896 | if (optlen != sizeof(val)) | |
3897 | return -EINVAL; | |
a7b75c5a | 3898 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
69e3c75f | 3899 | return -EFAULT; |
a6361f0c WB |
3900 | |
3901 | lock_sock(sk); | |
3902 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3903 | ret = -EBUSY; | |
3904 | } else { | |
164bddac | 3905 | packet_sock_flag_set(po, PACKET_SOCK_TP_LOSS, val); |
a6361f0c WB |
3906 | ret = 0; |
3907 | } | |
3908 | release_sock(sk); | |
3909 | return ret; | |
69e3c75f | 3910 | } |
8dc41944 HX |
3911 | case PACKET_AUXDATA: |
3912 | { | |
3913 | int val; | |
3914 | ||
3915 | if (optlen < sizeof(val)) | |
3916 | return -EINVAL; | |
a7b75c5a | 3917 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
8dc41944 HX |
3918 | return -EFAULT; |
3919 | ||
fd53c297 | 3920 | packet_sock_flag_set(po, PACKET_SOCK_AUXDATA, val); |
8dc41944 HX |
3921 | return 0; |
3922 | } | |
80feaacb PWJ |
3923 | case PACKET_ORIGDEV: |
3924 | { | |
3925 | int val; | |
3926 | ||
3927 | if (optlen < sizeof(val)) | |
3928 | return -EINVAL; | |
a7b75c5a | 3929 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
80feaacb PWJ |
3930 | return -EFAULT; |
3931 | ||
ee5675ec | 3932 | packet_sock_flag_set(po, PACKET_SOCK_ORIGDEV, val); |
80feaacb PWJ |
3933 | return 0; |
3934 | } | |
bfd5f4a3 | 3935 | case PACKET_VNET_HDR: |
dfc39d40 | 3936 | case PACKET_VNET_HDR_SZ: |
bfd5f4a3 | 3937 | { |
dfc39d40 | 3938 | int val, hdr_len; |
bfd5f4a3 SS |
3939 | |
3940 | if (sock->type != SOCK_RAW) | |
3941 | return -EINVAL; | |
bfd5f4a3 SS |
3942 | if (optlen < sizeof(val)) |
3943 | return -EINVAL; | |
a7b75c5a | 3944 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
bfd5f4a3 SS |
3945 | return -EFAULT; |
3946 | ||
dfc39d40 JT |
3947 | if (optname == PACKET_VNET_HDR_SZ) { |
3948 | if (val && val != sizeof(struct virtio_net_hdr) && | |
3949 | val != sizeof(struct virtio_net_hdr_mrg_rxbuf)) | |
3950 | return -EINVAL; | |
3951 | hdr_len = val; | |
3952 | } else { | |
3953 | hdr_len = val ? sizeof(struct virtio_net_hdr) : 0; | |
3954 | } | |
a6361f0c WB |
3955 | lock_sock(sk); |
3956 | if (po->rx_ring.pg_vec || po->tx_ring.pg_vec) { | |
3957 | ret = -EBUSY; | |
3958 | } else { | |
dfc39d40 | 3959 | WRITE_ONCE(po->vnet_hdr_sz, hdr_len); |
a6361f0c WB |
3960 | ret = 0; |
3961 | } | |
3962 | release_sock(sk); | |
3963 | return ret; | |
bfd5f4a3 | 3964 | } |
614f60fa SM |
3965 | case PACKET_TIMESTAMP: |
3966 | { | |
3967 | int val; | |
3968 | ||
3969 | if (optlen != sizeof(val)) | |
3970 | return -EINVAL; | |
a7b75c5a | 3971 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
614f60fa SM |
3972 | return -EFAULT; |
3973 | ||
1051ce4a | 3974 | WRITE_ONCE(po->tp_tstamp, val); |
614f60fa SM |
3975 | return 0; |
3976 | } | |
dc99f600 DM |
3977 | case PACKET_FANOUT: |
3978 | { | |
9c661b0b | 3979 | struct fanout_args args = { 0 }; |
dc99f600 | 3980 | |
9c661b0b | 3981 | if (optlen != sizeof(int) && optlen != sizeof(args)) |
dc99f600 | 3982 | return -EINVAL; |
9c661b0b | 3983 | if (copy_from_sockptr(&args, optval, optlen)) |
dc99f600 DM |
3984 | return -EFAULT; |
3985 | ||
9c661b0b | 3986 | return fanout_add(sk, &args); |
dc99f600 | 3987 | } |
47dceb8e WB |
3988 | case PACKET_FANOUT_DATA: |
3989 | { | |
e42e70ad ED |
3990 | /* Paired with the WRITE_ONCE() in fanout_add() */ |
3991 | if (!READ_ONCE(po->fanout)) | |
47dceb8e WB |
3992 | return -EINVAL; |
3993 | ||
3994 | return fanout_set_data(po, optval, optlen); | |
3995 | } | |
fa788d98 VW |
3996 | case PACKET_IGNORE_OUTGOING: |
3997 | { | |
3998 | int val; | |
3999 | ||
4000 | if (optlen != sizeof(val)) | |
4001 | return -EINVAL; | |
a7b75c5a | 4002 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
fa788d98 VW |
4003 | return -EFAULT; |
4004 | if (val < 0 || val > 1) | |
4005 | return -EINVAL; | |
4006 | ||
4007 | po->prot_hook.ignore_outgoing = !!val; | |
4008 | return 0; | |
4009 | } | |
5920cd3a PC |
4010 | case PACKET_TX_HAS_OFF: |
4011 | { | |
4012 | unsigned int val; | |
4013 | ||
4014 | if (optlen != sizeof(val)) | |
4015 | return -EINVAL; | |
a7b75c5a | 4016 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
5920cd3a | 4017 | return -EFAULT; |
a6361f0c WB |
4018 | |
4019 | lock_sock(sk); | |
25c55b38 | 4020 | if (!po->rx_ring.pg_vec && !po->tx_ring.pg_vec) |
74383446 | 4021 | packet_sock_flag_set(po, PACKET_SOCK_TX_HAS_OFF, val); |
25c55b38 | 4022 | |
a6361f0c | 4023 | release_sock(sk); |
5920cd3a PC |
4024 | return 0; |
4025 | } | |
d346a3fa DB |
4026 | case PACKET_QDISC_BYPASS: |
4027 | { | |
4028 | int val; | |
4029 | ||
4030 | if (optlen != sizeof(val)) | |
4031 | return -EINVAL; | |
a7b75c5a | 4032 | if (copy_from_sockptr(&val, optval, sizeof(val))) |
d346a3fa DB |
4033 | return -EFAULT; |
4034 | ||
105a201e | 4035 | packet_sock_flag_set(po, PACKET_SOCK_QDISC_BYPASS, val); |
d346a3fa DB |
4036 | return 0; |
4037 | } | |
1da177e4 LT |
4038 | default: |
4039 | return -ENOPROTOOPT; | |
4040 | } | |
4041 | } | |
4042 | ||
4043 | static int packet_getsockopt(struct socket *sock, int level, int optname, | |
4044 | char __user *optval, int __user *optlen) | |
4045 | { | |
4046 | int len; | |
c06fff6e | 4047 | int val, lv = sizeof(val); |
1da177e4 LT |
4048 | struct sock *sk = sock->sk; |
4049 | struct packet_sock *po = pkt_sk(sk); | |
c06fff6e | 4050 | void *data = &val; |
ee80fbf3 | 4051 | union tpacket_stats_u st; |
a9b63918 | 4052 | struct tpacket_rollover_stats rstats; |
8e8e2951 | 4053 | int drops; |
1da177e4 LT |
4054 | |
4055 | if (level != SOL_PACKET) | |
4056 | return -ENOPROTOOPT; | |
4057 | ||
8ae55f04 KK |
4058 | if (get_user(len, optlen)) |
4059 | return -EFAULT; | |
1da177e4 LT |
4060 | |
4061 | if (len < 0) | |
4062 | return -EINVAL; | |
1ce4f28b | 4063 | |
69e3c75f | 4064 | switch (optname) { |
1da177e4 | 4065 | case PACKET_STATISTICS: |
1da177e4 | 4066 | spin_lock_bh(&sk->sk_receive_queue.lock); |
ee80fbf3 DB |
4067 | memcpy(&st, &po->stats, sizeof(st)); |
4068 | memset(&po->stats, 0, sizeof(po->stats)); | |
4069 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
8e8e2951 | 4070 | drops = atomic_xchg(&po->tp_drops, 0); |
ee80fbf3 | 4071 | |
f6fb8f10 | 4072 | if (po->tp_version == TPACKET_V3) { |
c06fff6e | 4073 | lv = sizeof(struct tpacket_stats_v3); |
8e8e2951 ED |
4074 | st.stats3.tp_drops = drops; |
4075 | st.stats3.tp_packets += drops; | |
ee80fbf3 | 4076 | data = &st.stats3; |
f6fb8f10 | 4077 | } else { |
c06fff6e | 4078 | lv = sizeof(struct tpacket_stats); |
8e8e2951 ED |
4079 | st.stats1.tp_drops = drops; |
4080 | st.stats1.tp_packets += drops; | |
ee80fbf3 | 4081 | data = &st.stats1; |
f6fb8f10 | 4082 | } |
ee80fbf3 | 4083 | |
8dc41944 HX |
4084 | break; |
4085 | case PACKET_AUXDATA: | |
fd53c297 | 4086 | val = packet_sock_flag(po, PACKET_SOCK_AUXDATA); |
80feaacb PWJ |
4087 | break; |
4088 | case PACKET_ORIGDEV: | |
ee5675ec | 4089 | val = packet_sock_flag(po, PACKET_SOCK_ORIGDEV); |
bfd5f4a3 SS |
4090 | break; |
4091 | case PACKET_VNET_HDR: | |
dfc39d40 JT |
4092 | val = !!READ_ONCE(po->vnet_hdr_sz); |
4093 | break; | |
4094 | case PACKET_VNET_HDR_SZ: | |
4095 | val = READ_ONCE(po->vnet_hdr_sz); | |
1da177e4 | 4096 | break; |
bbd6ef87 | 4097 | case PACKET_VERSION: |
bbd6ef87 | 4098 | val = po->tp_version; |
bbd6ef87 PM |
4099 | break; |
4100 | case PACKET_HDRLEN: | |
4101 | if (len > sizeof(int)) | |
4102 | len = sizeof(int); | |
fd2c83b3 AP |
4103 | if (len < sizeof(int)) |
4104 | return -EINVAL; | |
bbd6ef87 PM |
4105 | if (copy_from_user(&val, optval, len)) |
4106 | return -EFAULT; | |
4107 | switch (val) { | |
4108 | case TPACKET_V1: | |
4109 | val = sizeof(struct tpacket_hdr); | |
4110 | break; | |
4111 | case TPACKET_V2: | |
4112 | val = sizeof(struct tpacket2_hdr); | |
4113 | break; | |
f6fb8f10 | 4114 | case TPACKET_V3: |
4115 | val = sizeof(struct tpacket3_hdr); | |
4116 | break; | |
bbd6ef87 PM |
4117 | default: |
4118 | return -EINVAL; | |
4119 | } | |
bbd6ef87 | 4120 | break; |
8913336a | 4121 | case PACKET_RESERVE: |
8913336a | 4122 | val = po->tp_reserve; |
8913336a | 4123 | break; |
69e3c75f | 4124 | case PACKET_LOSS: |
164bddac | 4125 | val = packet_sock_flag(po, PACKET_SOCK_TP_LOSS); |
69e3c75f | 4126 | break; |
614f60fa | 4127 | case PACKET_TIMESTAMP: |
1051ce4a | 4128 | val = READ_ONCE(po->tp_tstamp); |
614f60fa | 4129 | break; |
dc99f600 | 4130 | case PACKET_FANOUT: |
dc99f600 DM |
4131 | val = (po->fanout ? |
4132 | ((u32)po->fanout->id | | |
77f65ebd WB |
4133 | ((u32)po->fanout->type << 16) | |
4134 | ((u32)po->fanout->flags << 24)) : | |
dc99f600 | 4135 | 0); |
dc99f600 | 4136 | break; |
fa788d98 VW |
4137 | case PACKET_IGNORE_OUTGOING: |
4138 | val = po->prot_hook.ignore_outgoing; | |
4139 | break; | |
a9b63918 | 4140 | case PACKET_ROLLOVER_STATS: |
57f015f5 | 4141 | if (!po->rollover) |
a9b63918 | 4142 | return -EINVAL; |
57f015f5 MM |
4143 | rstats.tp_all = atomic_long_read(&po->rollover->num); |
4144 | rstats.tp_huge = atomic_long_read(&po->rollover->num_huge); | |
4145 | rstats.tp_failed = atomic_long_read(&po->rollover->num_failed); | |
4146 | data = &rstats; | |
4147 | lv = sizeof(rstats); | |
a9b63918 | 4148 | break; |
5920cd3a | 4149 | case PACKET_TX_HAS_OFF: |
74383446 | 4150 | val = packet_sock_flag(po, PACKET_SOCK_TX_HAS_OFF); |
5920cd3a | 4151 | break; |
d346a3fa | 4152 | case PACKET_QDISC_BYPASS: |
105a201e | 4153 | val = packet_sock_flag(po, PACKET_SOCK_QDISC_BYPASS); |
d346a3fa | 4154 | break; |
1da177e4 LT |
4155 | default: |
4156 | return -ENOPROTOOPT; | |
4157 | } | |
4158 | ||
c06fff6e ED |
4159 | if (len > lv) |
4160 | len = lv; | |
8ae55f04 KK |
4161 | if (put_user(len, optlen)) |
4162 | return -EFAULT; | |
8dc41944 HX |
4163 | if (copy_to_user(optval, data, len)) |
4164 | return -EFAULT; | |
8ae55f04 | 4165 | return 0; |
1da177e4 LT |
4166 | } |
4167 | ||
351638e7 JP |
4168 | static int packet_notifier(struct notifier_block *this, |
4169 | unsigned long msg, void *ptr) | |
1da177e4 LT |
4170 | { |
4171 | struct sock *sk; | |
351638e7 | 4172 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); |
c346dca1 | 4173 | struct net *net = dev_net(dev); |
1da177e4 | 4174 | |
808f5114 | 4175 | rcu_read_lock(); |
b67bfe0d | 4176 | sk_for_each_rcu(sk, &net->packet.sklist) { |
1da177e4 LT |
4177 | struct packet_sock *po = pkt_sk(sk); |
4178 | ||
4179 | switch (msg) { | |
4180 | case NETDEV_UNREGISTER: | |
1da177e4 | 4181 | if (po->mclist) |
82f17091 | 4182 | packet_dev_mclist_delete(dev, &po->mclist); |
df561f66 | 4183 | fallthrough; |
a2efcfa0 | 4184 | |
1da177e4 LT |
4185 | case NETDEV_DOWN: |
4186 | if (dev->ifindex == po->ifindex) { | |
4187 | spin_lock(&po->bind_lock); | |
61edf479 | 4188 | if (packet_sock_flag(po, PACKET_SOCK_RUNNING)) { |
ce06b03e | 4189 | __unregister_prot_hook(sk, false); |
1da177e4 LT |
4190 | sk->sk_err = ENETDOWN; |
4191 | if (!sock_flag(sk, SOCK_DEAD)) | |
e3ae2365 | 4192 | sk_error_report(sk); |
1da177e4 LT |
4193 | } |
4194 | if (msg == NETDEV_UNREGISTER) { | |
66e56cd4 | 4195 | packet_cached_dev_reset(po); |
e032f7c9 | 4196 | WRITE_ONCE(po->ifindex, -1); |
d62607c3 JK |
4197 | netdev_put(po->prot_hook.dev, |
4198 | &po->prot_hook.dev_tracker); | |
1da177e4 LT |
4199 | po->prot_hook.dev = NULL; |
4200 | } | |
4201 | spin_unlock(&po->bind_lock); | |
4202 | } | |
4203 | break; | |
4204 | case NETDEV_UP: | |
808f5114 | 4205 | if (dev->ifindex == po->ifindex) { |
4206 | spin_lock(&po->bind_lock); | |
ce06b03e DM |
4207 | if (po->num) |
4208 | register_prot_hook(sk); | |
808f5114 | 4209 | spin_unlock(&po->bind_lock); |
1da177e4 | 4210 | } |
1da177e4 LT |
4211 | break; |
4212 | } | |
4213 | } | |
808f5114 | 4214 | rcu_read_unlock(); |
1da177e4 LT |
4215 | return NOTIFY_DONE; |
4216 | } | |
4217 | ||
4218 | ||
4219 | static int packet_ioctl(struct socket *sock, unsigned int cmd, | |
4220 | unsigned long arg) | |
4221 | { | |
4222 | struct sock *sk = sock->sk; | |
4223 | ||
69e3c75f | 4224 | switch (cmd) { |
40d4e3df ED |
4225 | case SIOCOUTQ: |
4226 | { | |
4227 | int amount = sk_wmem_alloc_get(sk); | |
31e6d363 | 4228 | |
40d4e3df ED |
4229 | return put_user(amount, (int __user *)arg); |
4230 | } | |
4231 | case SIOCINQ: | |
4232 | { | |
4233 | struct sk_buff *skb; | |
4234 | int amount = 0; | |
4235 | ||
4236 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
4237 | skb = skb_peek(&sk->sk_receive_queue); | |
4238 | if (skb) | |
4239 | amount = skb->len; | |
4240 | spin_unlock_bh(&sk->sk_receive_queue.lock); | |
4241 | return put_user(amount, (int __user *)arg); | |
4242 | } | |
1da177e4 | 4243 | #ifdef CONFIG_INET |
40d4e3df ED |
4244 | case SIOCADDRT: |
4245 | case SIOCDELRT: | |
4246 | case SIOCDARP: | |
4247 | case SIOCGARP: | |
4248 | case SIOCSARP: | |
4249 | case SIOCGIFADDR: | |
4250 | case SIOCSIFADDR: | |
4251 | case SIOCGIFBRDADDR: | |
4252 | case SIOCSIFBRDADDR: | |
4253 | case SIOCGIFNETMASK: | |
4254 | case SIOCSIFNETMASK: | |
4255 | case SIOCGIFDSTADDR: | |
4256 | case SIOCSIFDSTADDR: | |
4257 | case SIOCSIFFLAGS: | |
40d4e3df | 4258 | return inet_dgram_ops.ioctl(sock, cmd, arg); |
1da177e4 LT |
4259 | #endif |
4260 | ||
40d4e3df ED |
4261 | default: |
4262 | return -ENOIOCTLCMD; | |
1da177e4 LT |
4263 | } |
4264 | return 0; | |
4265 | } | |
4266 | ||
a11e1d43 LT |
4267 | static __poll_t packet_poll(struct file *file, struct socket *sock, |
4268 | poll_table *wait) | |
1da177e4 LT |
4269 | { |
4270 | struct sock *sk = sock->sk; | |
4271 | struct packet_sock *po = pkt_sk(sk); | |
a11e1d43 | 4272 | __poll_t mask = datagram_poll(file, sock, wait); |
1da177e4 LT |
4273 | |
4274 | spin_lock_bh(&sk->sk_receive_queue.lock); | |
69e3c75f | 4275 | if (po->rx_ring.pg_vec) { |
f6fb8f10 | 4276 | if (!packet_previous_rx_frame(po, &po->rx_ring, |
4277 | TP_STATUS_KERNEL)) | |
a9a08845 | 4278 | mask |= EPOLLIN | EPOLLRDNORM; |
1da177e4 | 4279 | } |
9bb6cd65 | 4280 | packet_rcv_try_clear_pressure(po); |
1da177e4 | 4281 | spin_unlock_bh(&sk->sk_receive_queue.lock); |
69e3c75f JB |
4282 | spin_lock_bh(&sk->sk_write_queue.lock); |
4283 | if (po->tx_ring.pg_vec) { | |
4284 | if (packet_current_frame(po, &po->tx_ring, TP_STATUS_AVAILABLE)) | |
a9a08845 | 4285 | mask |= EPOLLOUT | EPOLLWRNORM; |
69e3c75f JB |
4286 | } |
4287 | spin_unlock_bh(&sk->sk_write_queue.lock); | |
1da177e4 LT |
4288 | return mask; |
4289 | } | |
4290 | ||
4291 | ||
4292 | /* Dirty? Well, I still did not learn better way to account | |
4293 | * for user mmaps. | |
4294 | */ | |
4295 | ||
4296 | static void packet_mm_open(struct vm_area_struct *vma) | |
4297 | { | |
4298 | struct file *file = vma->vm_file; | |
40d4e3df | 4299 | struct socket *sock = file->private_data; |
1da177e4 | 4300 | struct sock *sk = sock->sk; |
1ce4f28b | 4301 | |
1da177e4 | 4302 | if (sk) |
db3fadac | 4303 | atomic_long_inc(&pkt_sk(sk)->mapped); |
1da177e4 LT |
4304 | } |
4305 | ||
4306 | static void packet_mm_close(struct vm_area_struct *vma) | |
4307 | { | |
4308 | struct file *file = vma->vm_file; | |
40d4e3df | 4309 | struct socket *sock = file->private_data; |
1da177e4 | 4310 | struct sock *sk = sock->sk; |
1ce4f28b | 4311 | |
1da177e4 | 4312 | if (sk) |
db3fadac | 4313 | atomic_long_dec(&pkt_sk(sk)->mapped); |
1da177e4 LT |
4314 | } |
4315 | ||
f0f37e2f | 4316 | static const struct vm_operations_struct packet_mmap_ops = { |
40d4e3df ED |
4317 | .open = packet_mm_open, |
4318 | .close = packet_mm_close, | |
1da177e4 LT |
4319 | }; |
4320 | ||
3a7ad063 ED |
4321 | static void free_pg_vec(struct pgv *pg_vec, unsigned int order, |
4322 | unsigned int len) | |
1da177e4 LT |
4323 | { |
4324 | int i; | |
4325 | ||
4ebf0ae2 | 4326 | for (i = 0; i < len; i++) { |
0e3125c7 | 4327 | if (likely(pg_vec[i].buffer)) { |
3a7ad063 ED |
4328 | if (is_vmalloc_addr(pg_vec[i].buffer)) |
4329 | vfree(pg_vec[i].buffer); | |
4330 | else | |
4331 | free_pages((unsigned long)pg_vec[i].buffer, | |
4332 | order); | |
0e3125c7 NH |
4333 | pg_vec[i].buffer = NULL; |
4334 | } | |
1da177e4 LT |
4335 | } |
4336 | kfree(pg_vec); | |
4337 | } | |
4338 | ||
3a7ad063 | 4339 | static char *alloc_one_pg_vec_page(unsigned long order) |
4ebf0ae2 | 4340 | { |
f0d4eb29 | 4341 | char *buffer; |
3a7ad063 ED |
4342 | gfp_t gfp_flags = GFP_KERNEL | __GFP_COMP | |
4343 | __GFP_ZERO | __GFP_NOWARN | __GFP_NORETRY; | |
0e3125c7 | 4344 | |
3a7ad063 | 4345 | buffer = (char *) __get_free_pages(gfp_flags, order); |
0e3125c7 NH |
4346 | if (buffer) |
4347 | return buffer; | |
4348 | ||
3a7ad063 ED |
4349 | /* __get_free_pages failed, fall back to vmalloc */ |
4350 | buffer = vzalloc(array_size((1 << order), PAGE_SIZE)); | |
4351 | if (buffer) | |
4352 | return buffer; | |
0e3125c7 | 4353 | |
3a7ad063 ED |
4354 | /* vmalloc failed, lets dig into swap here */ |
4355 | gfp_flags &= ~__GFP_NORETRY; | |
4356 | buffer = (char *) __get_free_pages(gfp_flags, order); | |
4357 | if (buffer) | |
4358 | return buffer; | |
4359 | ||
4360 | /* complete and utter failure */ | |
4361 | return NULL; | |
4ebf0ae2 DM |
4362 | } |
4363 | ||
3a7ad063 | 4364 | static struct pgv *alloc_pg_vec(struct tpacket_req *req, int order) |
4ebf0ae2 DM |
4365 | { |
4366 | unsigned int block_nr = req->tp_block_nr; | |
0e3125c7 | 4367 | struct pgv *pg_vec; |
4ebf0ae2 DM |
4368 | int i; |
4369 | ||
398f0132 | 4370 | pg_vec = kcalloc(block_nr, sizeof(struct pgv), GFP_KERNEL | __GFP_NOWARN); |
4ebf0ae2 DM |
4371 | if (unlikely(!pg_vec)) |
4372 | goto out; | |
4373 | ||
4374 | for (i = 0; i < block_nr; i++) { | |
3a7ad063 | 4375 | pg_vec[i].buffer = alloc_one_pg_vec_page(order); |
0e3125c7 | 4376 | if (unlikely(!pg_vec[i].buffer)) |
4ebf0ae2 DM |
4377 | goto out_free_pgvec; |
4378 | } | |
4379 | ||
4380 | out: | |
4381 | return pg_vec; | |
4382 | ||
4383 | out_free_pgvec: | |
3a7ad063 | 4384 | free_pg_vec(pg_vec, order, block_nr); |
4ebf0ae2 DM |
4385 | pg_vec = NULL; |
4386 | goto out; | |
4387 | } | |
1da177e4 | 4388 | |
f6fb8f10 | 4389 | static int packet_set_ring(struct sock *sk, union tpacket_req_u *req_u, |
69e3c75f | 4390 | int closing, int tx_ring) |
1da177e4 | 4391 | { |
0e3125c7 | 4392 | struct pgv *pg_vec = NULL; |
1da177e4 | 4393 | struct packet_sock *po = pkt_sk(sk); |
61fad681 | 4394 | unsigned long *rx_owner_map = NULL; |
3a7ad063 | 4395 | int was_running, order = 0; |
69e3c75f JB |
4396 | struct packet_ring_buffer *rb; |
4397 | struct sk_buff_head *rb_queue; | |
0e11c91e | 4398 | __be16 num; |
2a6d6c31 | 4399 | int err; |
f6fb8f10 | 4400 | /* Added to avoid minimal code churn */ |
4401 | struct tpacket_req *req = &req_u->req; | |
4402 | ||
69e3c75f JB |
4403 | rb = tx_ring ? &po->tx_ring : &po->rx_ring; |
4404 | rb_queue = tx_ring ? &sk->sk_write_queue : &sk->sk_receive_queue; | |
1da177e4 | 4405 | |
69e3c75f JB |
4406 | err = -EBUSY; |
4407 | if (!closing) { | |
db3fadac | 4408 | if (atomic_long_read(&po->mapped)) |
69e3c75f | 4409 | goto out; |
b0138408 | 4410 | if (packet_read_pending(rb)) |
69e3c75f JB |
4411 | goto out; |
4412 | } | |
1da177e4 | 4413 | |
69e3c75f | 4414 | if (req->tp_block_nr) { |
4576cd46 WB |
4415 | unsigned int min_frame_size; |
4416 | ||
69e3c75f JB |
4417 | /* Sanity tests and some calculations */ |
4418 | err = -EBUSY; | |
4419 | if (unlikely(rb->pg_vec)) | |
4420 | goto out; | |
1da177e4 | 4421 | |
bbd6ef87 PM |
4422 | switch (po->tp_version) { |
4423 | case TPACKET_V1: | |
4424 | po->tp_hdrlen = TPACKET_HDRLEN; | |
4425 | break; | |
4426 | case TPACKET_V2: | |
4427 | po->tp_hdrlen = TPACKET2_HDRLEN; | |
4428 | break; | |
f6fb8f10 | 4429 | case TPACKET_V3: |
4430 | po->tp_hdrlen = TPACKET3_HDRLEN; | |
4431 | break; | |
bbd6ef87 PM |
4432 | } |
4433 | ||
69e3c75f | 4434 | err = -EINVAL; |
4ebf0ae2 | 4435 | if (unlikely((int)req->tp_block_size <= 0)) |
69e3c75f | 4436 | goto out; |
90836b67 | 4437 | if (unlikely(!PAGE_ALIGNED(req->tp_block_size))) |
69e3c75f | 4438 | goto out; |
4576cd46 | 4439 | min_frame_size = po->tp_hdrlen + po->tp_reserve; |
dc808110 | 4440 | if (po->tp_version >= TPACKET_V3 && |
4576cd46 WB |
4441 | req->tp_block_size < |
4442 | BLK_PLUS_PRIV((u64)req_u->req3.tp_sizeof_priv) + min_frame_size) | |
dc808110 | 4443 | goto out; |
4576cd46 | 4444 | if (unlikely(req->tp_frame_size < min_frame_size)) |
69e3c75f | 4445 | goto out; |
4ebf0ae2 | 4446 | if (unlikely(req->tp_frame_size & (TPACKET_ALIGNMENT - 1))) |
69e3c75f | 4447 | goto out; |
1da177e4 | 4448 | |
4194b491 TK |
4449 | rb->frames_per_block = req->tp_block_size / req->tp_frame_size; |
4450 | if (unlikely(rb->frames_per_block == 0)) | |
69e3c75f | 4451 | goto out; |
fc62814d | 4452 | if (unlikely(rb->frames_per_block > UINT_MAX / req->tp_block_nr)) |
8f8d28e4 | 4453 | goto out; |
69e3c75f JB |
4454 | if (unlikely((rb->frames_per_block * req->tp_block_nr) != |
4455 | req->tp_frame_nr)) | |
4456 | goto out; | |
1da177e4 LT |
4457 | |
4458 | err = -ENOMEM; | |
3a7ad063 ED |
4459 | order = get_order(req->tp_block_size); |
4460 | pg_vec = alloc_pg_vec(req, order); | |
4ebf0ae2 | 4461 | if (unlikely(!pg_vec)) |
1da177e4 | 4462 | goto out; |
f6fb8f10 | 4463 | switch (po->tp_version) { |
4464 | case TPACKET_V3: | |
7f953ab2 SV |
4465 | /* Block transmit is not supported yet */ |
4466 | if (!tx_ring) { | |
e8e85cc5 | 4467 | init_prb_bdqc(po, rb, pg_vec, req_u); |
7f953ab2 SV |
4468 | } else { |
4469 | struct tpacket_req3 *req3 = &req_u->req3; | |
4470 | ||
4471 | if (req3->tp_retire_blk_tov || | |
4472 | req3->tp_sizeof_priv || | |
4473 | req3->tp_feature_req_word) { | |
4474 | err = -EINVAL; | |
55655e3d | 4475 | goto out_free_pg_vec; |
7f953ab2 SV |
4476 | } |
4477 | } | |
d7cf0c34 | 4478 | break; |
f6fb8f10 | 4479 | default: |
61fad681 WB |
4480 | if (!tx_ring) { |
4481 | rx_owner_map = bitmap_alloc(req->tp_frame_nr, | |
4482 | GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO); | |
4483 | if (!rx_owner_map) | |
4484 | goto out_free_pg_vec; | |
4485 | } | |
f6fb8f10 | 4486 | break; |
4487 | } | |
69e3c75f JB |
4488 | } |
4489 | /* Done */ | |
4490 | else { | |
4491 | err = -EINVAL; | |
4ebf0ae2 | 4492 | if (unlikely(req->tp_frame_nr)) |
69e3c75f | 4493 | goto out; |
1da177e4 LT |
4494 | } |
4495 | ||
1da177e4 LT |
4496 | |
4497 | /* Detach socket from network */ | |
4498 | spin_lock(&po->bind_lock); | |
61edf479 | 4499 | was_running = packet_sock_flag(po, PACKET_SOCK_RUNNING); |
1da177e4 LT |
4500 | num = po->num; |
4501 | if (was_running) { | |
c7d2ef5d | 4502 | WRITE_ONCE(po->num, 0); |
ce06b03e | 4503 | __unregister_prot_hook(sk, false); |
1da177e4 LT |
4504 | } |
4505 | spin_unlock(&po->bind_lock); | |
1ce4f28b | 4506 | |
1da177e4 LT |
4507 | synchronize_net(); |
4508 | ||
4509 | err = -EBUSY; | |
905db440 | 4510 | mutex_lock(&po->pg_vec_lock); |
db3fadac | 4511 | if (closing || atomic_long_read(&po->mapped) == 0) { |
1da177e4 | 4512 | err = 0; |
69e3c75f | 4513 | spin_lock_bh(&rb_queue->lock); |
c053fd96 | 4514 | swap(rb->pg_vec, pg_vec); |
61fad681 WB |
4515 | if (po->tp_version <= TPACKET_V2) |
4516 | swap(rb->rx_owner_map, rx_owner_map); | |
69e3c75f JB |
4517 | rb->frame_max = (req->tp_frame_nr - 1); |
4518 | rb->head = 0; | |
4519 | rb->frame_size = req->tp_frame_size; | |
4520 | spin_unlock_bh(&rb_queue->lock); | |
4521 | ||
3a7ad063 | 4522 | swap(rb->pg_vec_order, order); |
c053fd96 | 4523 | swap(rb->pg_vec_len, req->tp_block_nr); |
69e3c75f JB |
4524 | |
4525 | rb->pg_vec_pages = req->tp_block_size/PAGE_SIZE; | |
4526 | po->prot_hook.func = (po->rx_ring.pg_vec) ? | |
4527 | tpacket_rcv : packet_rcv; | |
4528 | skb_queue_purge(rb_queue); | |
db3fadac DB |
4529 | if (atomic_long_read(&po->mapped)) |
4530 | pr_err("packet_mmap: vma is busy: %ld\n", | |
4531 | atomic_long_read(&po->mapped)); | |
1da177e4 | 4532 | } |
905db440 | 4533 | mutex_unlock(&po->pg_vec_lock); |
1da177e4 LT |
4534 | |
4535 | spin_lock(&po->bind_lock); | |
ce06b03e | 4536 | if (was_running) { |
c7d2ef5d | 4537 | WRITE_ONCE(po->num, num); |
ce06b03e | 4538 | register_prot_hook(sk); |
1da177e4 LT |
4539 | } |
4540 | spin_unlock(&po->bind_lock); | |
c800aaf8 | 4541 | if (pg_vec && (po->tp_version > TPACKET_V2)) { |
f6fb8f10 | 4542 | /* Because we don't support block-based V3 on tx-ring */ |
4543 | if (!tx_ring) | |
73d0fcf2 | 4544 | prb_shutdown_retire_blk_timer(po, rb_queue); |
f6fb8f10 | 4545 | } |
1da177e4 | 4546 | |
55655e3d | 4547 | out_free_pg_vec: |
ec6af094 WB |
4548 | if (pg_vec) { |
4549 | bitmap_free(rx_owner_map); | |
3a7ad063 | 4550 | free_pg_vec(pg_vec, order, req->tp_block_nr); |
ec6af094 | 4551 | } |
1da177e4 LT |
4552 | out: |
4553 | return err; | |
4554 | } | |
4555 | ||
69e3c75f JB |
4556 | static int packet_mmap(struct file *file, struct socket *sock, |
4557 | struct vm_area_struct *vma) | |
1da177e4 LT |
4558 | { |
4559 | struct sock *sk = sock->sk; | |
4560 | struct packet_sock *po = pkt_sk(sk); | |
69e3c75f JB |
4561 | unsigned long size, expected_size; |
4562 | struct packet_ring_buffer *rb; | |
1da177e4 LT |
4563 | unsigned long start; |
4564 | int err = -EINVAL; | |
4565 | int i; | |
4566 | ||
4567 | if (vma->vm_pgoff) | |
4568 | return -EINVAL; | |
4569 | ||
905db440 | 4570 | mutex_lock(&po->pg_vec_lock); |
69e3c75f JB |
4571 | |
4572 | expected_size = 0; | |
4573 | for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) { | |
4574 | if (rb->pg_vec) { | |
4575 | expected_size += rb->pg_vec_len | |
4576 | * rb->pg_vec_pages | |
4577 | * PAGE_SIZE; | |
4578 | } | |
4579 | } | |
4580 | ||
4581 | if (expected_size == 0) | |
1da177e4 | 4582 | goto out; |
69e3c75f JB |
4583 | |
4584 | size = vma->vm_end - vma->vm_start; | |
4585 | if (size != expected_size) | |
1da177e4 LT |
4586 | goto out; |
4587 | ||
1da177e4 | 4588 | start = vma->vm_start; |
69e3c75f JB |
4589 | for (rb = &po->rx_ring; rb <= &po->tx_ring; rb++) { |
4590 | if (rb->pg_vec == NULL) | |
4591 | continue; | |
4592 | ||
4593 | for (i = 0; i < rb->pg_vec_len; i++) { | |
0e3125c7 NH |
4594 | struct page *page; |
4595 | void *kaddr = rb->pg_vec[i].buffer; | |
69e3c75f JB |
4596 | int pg_num; |
4597 | ||
c56b4d90 CG |
4598 | for (pg_num = 0; pg_num < rb->pg_vec_pages; pg_num++) { |
4599 | page = pgv_to_page(kaddr); | |
69e3c75f JB |
4600 | err = vm_insert_page(vma, start, page); |
4601 | if (unlikely(err)) | |
4602 | goto out; | |
4603 | start += PAGE_SIZE; | |
0e3125c7 | 4604 | kaddr += PAGE_SIZE; |
69e3c75f | 4605 | } |
4ebf0ae2 | 4606 | } |
1da177e4 | 4607 | } |
69e3c75f | 4608 | |
db3fadac | 4609 | atomic_long_inc(&po->mapped); |
1da177e4 LT |
4610 | vma->vm_ops = &packet_mmap_ops; |
4611 | err = 0; | |
4612 | ||
4613 | out: | |
905db440 | 4614 | mutex_unlock(&po->pg_vec_lock); |
1da177e4 LT |
4615 | return err; |
4616 | } | |
1da177e4 | 4617 | |
90ddc4f0 | 4618 | static const struct proto_ops packet_ops_spkt = { |
1da177e4 LT |
4619 | .family = PF_PACKET, |
4620 | .owner = THIS_MODULE, | |
4621 | .release = packet_release, | |
4622 | .bind = packet_bind_spkt, | |
4623 | .connect = sock_no_connect, | |
4624 | .socketpair = sock_no_socketpair, | |
4625 | .accept = sock_no_accept, | |
4626 | .getname = packet_getname_spkt, | |
a11e1d43 | 4627 | .poll = datagram_poll, |
1da177e4 | 4628 | .ioctl = packet_ioctl, |
c7cbdbf2 | 4629 | .gettstamp = sock_gettstamp, |
1da177e4 LT |
4630 | .listen = sock_no_listen, |
4631 | .shutdown = sock_no_shutdown, | |
1da177e4 LT |
4632 | .sendmsg = packet_sendmsg_spkt, |
4633 | .recvmsg = packet_recvmsg, | |
4634 | .mmap = sock_no_mmap, | |
1da177e4 | 4635 | }; |
1da177e4 | 4636 | |
90ddc4f0 | 4637 | static const struct proto_ops packet_ops = { |
1da177e4 LT |
4638 | .family = PF_PACKET, |
4639 | .owner = THIS_MODULE, | |
4640 | .release = packet_release, | |
4641 | .bind = packet_bind, | |
4642 | .connect = sock_no_connect, | |
4643 | .socketpair = sock_no_socketpair, | |
4644 | .accept = sock_no_accept, | |
1ce4f28b | 4645 | .getname = packet_getname, |
a11e1d43 | 4646 | .poll = packet_poll, |
1da177e4 | 4647 | .ioctl = packet_ioctl, |
c7cbdbf2 | 4648 | .gettstamp = sock_gettstamp, |
1da177e4 LT |
4649 | .listen = sock_no_listen, |
4650 | .shutdown = sock_no_shutdown, | |
4651 | .setsockopt = packet_setsockopt, | |
4652 | .getsockopt = packet_getsockopt, | |
4653 | .sendmsg = packet_sendmsg, | |
4654 | .recvmsg = packet_recvmsg, | |
4655 | .mmap = packet_mmap, | |
1da177e4 LT |
4656 | }; |
4657 | ||
ec1b4cf7 | 4658 | static const struct net_proto_family packet_family_ops = { |
1da177e4 LT |
4659 | .family = PF_PACKET, |
4660 | .create = packet_create, | |
4661 | .owner = THIS_MODULE, | |
4662 | }; | |
4663 | ||
4664 | static struct notifier_block packet_netdev_notifier = { | |
40d4e3df | 4665 | .notifier_call = packet_notifier, |
1da177e4 LT |
4666 | }; |
4667 | ||
4668 | #ifdef CONFIG_PROC_FS | |
1da177e4 LT |
4669 | |
4670 | static void *packet_seq_start(struct seq_file *seq, loff_t *pos) | |
808f5114 | 4671 | __acquires(RCU) |
1da177e4 | 4672 | { |
e372c414 | 4673 | struct net *net = seq_file_net(seq); |
808f5114 | 4674 | |
4675 | rcu_read_lock(); | |
4676 | return seq_hlist_start_head_rcu(&net->packet.sklist, *pos); | |
1da177e4 LT |
4677 | } |
4678 | ||
4679 | static void *packet_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
4680 | { | |
1bf40954 | 4681 | struct net *net = seq_file_net(seq); |
808f5114 | 4682 | return seq_hlist_next_rcu(v, &net->packet.sklist, pos); |
1da177e4 LT |
4683 | } |
4684 | ||
4685 | static void packet_seq_stop(struct seq_file *seq, void *v) | |
808f5114 | 4686 | __releases(RCU) |
1da177e4 | 4687 | { |
808f5114 | 4688 | rcu_read_unlock(); |
1da177e4 LT |
4689 | } |
4690 | ||
1ce4f28b | 4691 | static int packet_seq_show(struct seq_file *seq, void *v) |
1da177e4 LT |
4692 | { |
4693 | if (v == SEQ_START_TOKEN) | |
abdcd06c BS |
4694 | seq_printf(seq, |
4695 | "%*sRefCnt Type Proto Iface R Rmem User Inode\n", | |
4696 | IS_ENABLED(CONFIG_64BIT) ? -17 : -9, "sk"); | |
1da177e4 | 4697 | else { |
b7ceabd9 | 4698 | struct sock *s = sk_entry(v); |
1da177e4 LT |
4699 | const struct packet_sock *po = pkt_sk(s); |
4700 | ||
4701 | seq_printf(seq, | |
71338aa7 | 4702 | "%pK %-6d %-4d %04x %-5d %1d %-6u %-6u %-6lu\n", |
1da177e4 | 4703 | s, |
41c6d650 | 4704 | refcount_read(&s->sk_refcnt), |
1da177e4 | 4705 | s->sk_type, |
c7d2ef5d | 4706 | ntohs(READ_ONCE(po->num)), |
e032f7c9 | 4707 | READ_ONCE(po->ifindex), |
61edf479 | 4708 | packet_sock_flag(po, PACKET_SOCK_RUNNING), |
1da177e4 | 4709 | atomic_read(&s->sk_rmem_alloc), |
a7cb5a49 | 4710 | from_kuid_munged(seq_user_ns(seq), sock_i_uid(s)), |
40d4e3df | 4711 | sock_i_ino(s)); |
1da177e4 LT |
4712 | } |
4713 | ||
4714 | return 0; | |
4715 | } | |
4716 | ||
56b3d975 | 4717 | static const struct seq_operations packet_seq_ops = { |
1da177e4 LT |
4718 | .start = packet_seq_start, |
4719 | .next = packet_seq_next, | |
4720 | .stop = packet_seq_stop, | |
4721 | .show = packet_seq_show, | |
4722 | }; | |
1da177e4 LT |
4723 | #endif |
4724 | ||
2c8c1e72 | 4725 | static int __net_init packet_net_init(struct net *net) |
d12d01d6 | 4726 | { |
0fa7fa98 | 4727 | mutex_init(&net->packet.sklist_lock); |
2aaef4e4 | 4728 | INIT_HLIST_HEAD(&net->packet.sklist); |
d12d01d6 | 4729 | |
a268e0f2 | 4730 | #ifdef CONFIG_PROC_FS |
c3506372 CH |
4731 | if (!proc_create_net("packet", 0, net->proc_net, &packet_seq_ops, |
4732 | sizeof(struct seq_net_private))) | |
d12d01d6 | 4733 | return -ENOMEM; |
a268e0f2 | 4734 | #endif /* CONFIG_PROC_FS */ |
d12d01d6 DL |
4735 | |
4736 | return 0; | |
4737 | } | |
4738 | ||
2c8c1e72 | 4739 | static void __net_exit packet_net_exit(struct net *net) |
d12d01d6 | 4740 | { |
ece31ffd | 4741 | remove_proc_entry("packet", net->proc_net); |
669f8f1a | 4742 | WARN_ON_ONCE(!hlist_empty(&net->packet.sklist)); |
d12d01d6 DL |
4743 | } |
4744 | ||
4745 | static struct pernet_operations packet_net_ops = { | |
4746 | .init = packet_net_init, | |
4747 | .exit = packet_net_exit, | |
4748 | }; | |
4749 | ||
4750 | ||
1da177e4 LT |
4751 | static void __exit packet_exit(void) |
4752 | { | |
1da177e4 LT |
4753 | sock_unregister(PF_PACKET); |
4754 | proto_unregister(&packet_proto); | |
63b7c2eb ZX |
4755 | unregister_netdevice_notifier(&packet_netdev_notifier); |
4756 | unregister_pernet_subsys(&packet_net_ops); | |
1da177e4 LT |
4757 | } |
4758 | ||
4759 | static int __init packet_init(void) | |
4760 | { | |
36096f2f | 4761 | int rc; |
1da177e4 | 4762 | |
36096f2f Y |
4763 | rc = register_pernet_subsys(&packet_net_ops); |
4764 | if (rc) | |
63b7c2eb | 4765 | goto out; |
36096f2f Y |
4766 | rc = register_netdevice_notifier(&packet_netdev_notifier); |
4767 | if (rc) | |
4768 | goto out_pernet; | |
63b7c2eb ZX |
4769 | rc = proto_register(&packet_proto, 0); |
4770 | if (rc) | |
4771 | goto out_notifier; | |
4772 | rc = sock_register(&packet_family_ops); | |
4773 | if (rc) | |
4774 | goto out_proto; | |
1da177e4 | 4775 | |
36096f2f Y |
4776 | return 0; |
4777 | ||
36096f2f Y |
4778 | out_proto: |
4779 | proto_unregister(&packet_proto); | |
63b7c2eb ZX |
4780 | out_notifier: |
4781 | unregister_netdevice_notifier(&packet_netdev_notifier); | |
4782 | out_pernet: | |
4783 | unregister_pernet_subsys(&packet_net_ops); | |
1da177e4 LT |
4784 | out: |
4785 | return rc; | |
4786 | } | |
4787 | ||
4788 | module_init(packet_init); | |
4789 | module_exit(packet_exit); | |
4790 | MODULE_LICENSE("GPL"); | |
4791 | MODULE_ALIAS_NETPROTO(PF_PACKET); |