]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * Definitions for the Interfaces handler. | |
7 | * | |
8 | * Version: @(#)dev.h 1.0.10 08/12/93 | |
9 | * | |
10 | * Authors: Ross Biro | |
11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> | |
12 | * Corey Minyard <wf-rch!minyard@relay.EU.net> | |
13 | * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov> | |
14 | * Alan Cox, <alan@lxorguk.ukuu.org.uk> | |
15 | * Bjorn Ekwall. <bj0rn@blox.se> | |
16 | * Pekka Riikonen <priikone@poseidon.pspt.fi> | |
17 | * | |
18 | * This program is free software; you can redistribute it and/or | |
19 | * modify it under the terms of the GNU General Public License | |
20 | * as published by the Free Software Foundation; either version | |
21 | * 2 of the License, or (at your option) any later version. | |
22 | * | |
23 | * Moved to /usr/include/linux for NET3 | |
24 | */ | |
25 | #ifndef _LINUX_NETDEVICE_H | |
26 | #define _LINUX_NETDEVICE_H | |
27 | ||
28 | #include <linux/timer.h> | |
29 | #include <linux/bug.h> | |
30 | #include <linux/delay.h> | |
31 | #include <linux/atomic.h> | |
32 | #include <linux/prefetch.h> | |
33 | #include <asm/cache.h> | |
34 | #include <asm/byteorder.h> | |
35 | ||
36 | #include <linux/percpu.h> | |
37 | #include <linux/rculist.h> | |
38 | #include <linux/dmaengine.h> | |
39 | #include <linux/workqueue.h> | |
40 | #include <linux/dynamic_queue_limits.h> | |
41 | ||
42 | #include <linux/ethtool.h> | |
43 | #include <net/net_namespace.h> | |
44 | #include <net/dsa.h> | |
45 | #ifdef CONFIG_DCB | |
46 | #include <net/dcbnl.h> | |
47 | #endif | |
48 | #include <net/netprio_cgroup.h> | |
49 | ||
50 | #include <linux/netdev_features.h> | |
51 | #include <linux/neighbour.h> | |
52 | #include <uapi/linux/netdevice.h> | |
53 | #include <uapi/linux/if_bonding.h> | |
54 | ||
55 | struct netpoll_info; | |
56 | struct device; | |
57 | struct phy_device; | |
58 | /* 802.11 specific */ | |
59 | struct wireless_dev; | |
60 | /* 802.15.4 specific */ | |
61 | struct wpan_dev; | |
62 | struct mpls_dev; | |
63 | ||
64 | void netdev_set_default_ethtool_ops(struct net_device *dev, | |
65 | const struct ethtool_ops *ops); | |
66 | ||
67 | /* Backlog congestion levels */ | |
68 | #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */ | |
69 | #define NET_RX_DROP 1 /* packet dropped */ | |
70 | ||
71 | /* | |
72 | * Transmit return codes: transmit return codes originate from three different | |
73 | * namespaces: | |
74 | * | |
75 | * - qdisc return codes | |
76 | * - driver transmit return codes | |
77 | * - errno values | |
78 | * | |
79 | * Drivers are allowed to return any one of those in their hard_start_xmit() | |
80 | * function. Real network devices commonly used with qdiscs should only return | |
81 | * the driver transmit return codes though - when qdiscs are used, the actual | |
82 | * transmission happens asynchronously, so the value is not propagated to | |
83 | * higher layers. Virtual network devices transmit synchronously, in this case | |
84 | * the driver transmit return codes are consumed by dev_queue_xmit(), all | |
85 | * others are propagated to higher layers. | |
86 | */ | |
87 | ||
88 | /* qdisc ->enqueue() return codes. */ | |
89 | #define NET_XMIT_SUCCESS 0x00 | |
90 | #define NET_XMIT_DROP 0x01 /* skb dropped */ | |
91 | #define NET_XMIT_CN 0x02 /* congestion notification */ | |
92 | #define NET_XMIT_POLICED 0x03 /* skb is shot by police */ | |
93 | #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */ | |
94 | ||
95 | /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It | |
96 | * indicates that the device will soon be dropping packets, or already drops | |
97 | * some packets of the same priority; prompting us to send less aggressively. */ | |
98 | #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e)) | |
99 | #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0) | |
100 | ||
101 | /* Driver transmit return codes */ | |
102 | #define NETDEV_TX_MASK 0xf0 | |
103 | ||
104 | enum netdev_tx { | |
105 | __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */ | |
106 | NETDEV_TX_OK = 0x00, /* driver took care of packet */ | |
107 | NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/ | |
108 | NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */ | |
109 | }; | |
110 | typedef enum netdev_tx netdev_tx_t; | |
111 | ||
112 | /* | |
113 | * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant; | |
114 | * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed. | |
115 | */ | |
116 | static inline bool dev_xmit_complete(int rc) | |
117 | { | |
118 | /* | |
119 | * Positive cases with an skb consumed by a driver: | |
120 | * - successful transmission (rc == NETDEV_TX_OK) | |
121 | * - error while transmitting (rc < 0) | |
122 | * - error while queueing to a different device (rc & NET_XMIT_MASK) | |
123 | */ | |
124 | if (likely(rc < NET_XMIT_MASK)) | |
125 | return true; | |
126 | ||
127 | return false; | |
128 | } | |
129 | ||
130 | /* | |
131 | * Compute the worst case header length according to the protocols | |
132 | * used. | |
133 | */ | |
134 | ||
135 | #if defined(CONFIG_HYPERV_NET) | |
136 | # define LL_MAX_HEADER 128 | |
137 | #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25) | |
138 | # if defined(CONFIG_MAC80211_MESH) | |
139 | # define LL_MAX_HEADER 128 | |
140 | # else | |
141 | # define LL_MAX_HEADER 96 | |
142 | # endif | |
143 | #else | |
144 | # define LL_MAX_HEADER 32 | |
145 | #endif | |
146 | ||
147 | #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \ | |
148 | !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL) | |
149 | #define MAX_HEADER LL_MAX_HEADER | |
150 | #else | |
151 | #define MAX_HEADER (LL_MAX_HEADER + 48) | |
152 | #endif | |
153 | ||
154 | /* | |
155 | * Old network device statistics. Fields are native words | |
156 | * (unsigned long) so they can be read and written atomically. | |
157 | */ | |
158 | ||
159 | struct net_device_stats { | |
160 | unsigned long rx_packets; | |
161 | unsigned long tx_packets; | |
162 | unsigned long rx_bytes; | |
163 | unsigned long tx_bytes; | |
164 | unsigned long rx_errors; | |
165 | unsigned long tx_errors; | |
166 | unsigned long rx_dropped; | |
167 | unsigned long tx_dropped; | |
168 | unsigned long multicast; | |
169 | unsigned long collisions; | |
170 | unsigned long rx_length_errors; | |
171 | unsigned long rx_over_errors; | |
172 | unsigned long rx_crc_errors; | |
173 | unsigned long rx_frame_errors; | |
174 | unsigned long rx_fifo_errors; | |
175 | unsigned long rx_missed_errors; | |
176 | unsigned long tx_aborted_errors; | |
177 | unsigned long tx_carrier_errors; | |
178 | unsigned long tx_fifo_errors; | |
179 | unsigned long tx_heartbeat_errors; | |
180 | unsigned long tx_window_errors; | |
181 | unsigned long rx_compressed; | |
182 | unsigned long tx_compressed; | |
183 | }; | |
184 | ||
185 | ||
186 | #include <linux/cache.h> | |
187 | #include <linux/skbuff.h> | |
188 | ||
189 | #ifdef CONFIG_RPS | |
190 | #include <linux/static_key.h> | |
191 | extern struct static_key rps_needed; | |
192 | #endif | |
193 | ||
194 | struct neighbour; | |
195 | struct neigh_parms; | |
196 | struct sk_buff; | |
197 | ||
198 | struct netdev_hw_addr { | |
199 | struct list_head list; | |
200 | unsigned char addr[MAX_ADDR_LEN]; | |
201 | unsigned char type; | |
202 | #define NETDEV_HW_ADDR_T_LAN 1 | |
203 | #define NETDEV_HW_ADDR_T_SAN 2 | |
204 | #define NETDEV_HW_ADDR_T_SLAVE 3 | |
205 | #define NETDEV_HW_ADDR_T_UNICAST 4 | |
206 | #define NETDEV_HW_ADDR_T_MULTICAST 5 | |
207 | bool global_use; | |
208 | int sync_cnt; | |
209 | int refcount; | |
210 | int synced; | |
211 | struct rcu_head rcu_head; | |
212 | }; | |
213 | ||
214 | struct netdev_hw_addr_list { | |
215 | struct list_head list; | |
216 | int count; | |
217 | }; | |
218 | ||
219 | #define netdev_hw_addr_list_count(l) ((l)->count) | |
220 | #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0) | |
221 | #define netdev_hw_addr_list_for_each(ha, l) \ | |
222 | list_for_each_entry(ha, &(l)->list, list) | |
223 | ||
224 | #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc) | |
225 | #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc) | |
226 | #define netdev_for_each_uc_addr(ha, dev) \ | |
227 | netdev_hw_addr_list_for_each(ha, &(dev)->uc) | |
228 | ||
229 | #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc) | |
230 | #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc) | |
231 | #define netdev_for_each_mc_addr(ha, dev) \ | |
232 | netdev_hw_addr_list_for_each(ha, &(dev)->mc) | |
233 | ||
234 | struct hh_cache { | |
235 | u16 hh_len; | |
236 | u16 __pad; | |
237 | seqlock_t hh_lock; | |
238 | ||
239 | /* cached hardware header; allow for machine alignment needs. */ | |
240 | #define HH_DATA_MOD 16 | |
241 | #define HH_DATA_OFF(__len) \ | |
242 | (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1)) | |
243 | #define HH_DATA_ALIGN(__len) \ | |
244 | (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1)) | |
245 | unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)]; | |
246 | }; | |
247 | ||
248 | /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much. | |
249 | * Alternative is: | |
250 | * dev->hard_header_len ? (dev->hard_header_len + | |
251 | * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0 | |
252 | * | |
253 | * We could use other alignment values, but we must maintain the | |
254 | * relationship HH alignment <= LL alignment. | |
255 | */ | |
256 | #define LL_RESERVED_SPACE(dev) \ | |
257 | ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) | |
258 | #define LL_RESERVED_SPACE_EXTRA(dev,extra) \ | |
259 | ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD) | |
260 | ||
261 | struct header_ops { | |
262 | int (*create) (struct sk_buff *skb, struct net_device *dev, | |
263 | unsigned short type, const void *daddr, | |
264 | const void *saddr, unsigned int len); | |
265 | int (*parse)(const struct sk_buff *skb, unsigned char *haddr); | |
266 | int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type); | |
267 | void (*cache_update)(struct hh_cache *hh, | |
268 | const struct net_device *dev, | |
269 | const unsigned char *haddr); | |
270 | }; | |
271 | ||
272 | /* These flag bits are private to the generic network queueing | |
273 | * layer, they may not be explicitly referenced by any other | |
274 | * code. | |
275 | */ | |
276 | ||
277 | enum netdev_state_t { | |
278 | __LINK_STATE_START, | |
279 | __LINK_STATE_PRESENT, | |
280 | __LINK_STATE_NOCARRIER, | |
281 | __LINK_STATE_LINKWATCH_PENDING, | |
282 | __LINK_STATE_DORMANT, | |
283 | }; | |
284 | ||
285 | ||
286 | /* | |
287 | * This structure holds at boot time configured netdevice settings. They | |
288 | * are then used in the device probing. | |
289 | */ | |
290 | struct netdev_boot_setup { | |
291 | char name[IFNAMSIZ]; | |
292 | struct ifmap map; | |
293 | }; | |
294 | #define NETDEV_BOOT_SETUP_MAX 8 | |
295 | ||
296 | int __init netdev_boot_setup(char *str); | |
297 | ||
298 | /* | |
299 | * Structure for NAPI scheduling similar to tasklet but with weighting | |
300 | */ | |
301 | struct napi_struct { | |
302 | /* The poll_list must only be managed by the entity which | |
303 | * changes the state of the NAPI_STATE_SCHED bit. This means | |
304 | * whoever atomically sets that bit can add this napi_struct | |
305 | * to the per-cpu poll_list, and whoever clears that bit | |
306 | * can remove from the list right before clearing the bit. | |
307 | */ | |
308 | struct list_head poll_list; | |
309 | ||
310 | unsigned long state; | |
311 | int weight; | |
312 | unsigned int gro_count; | |
313 | int (*poll)(struct napi_struct *, int); | |
314 | #ifdef CONFIG_NETPOLL | |
315 | spinlock_t poll_lock; | |
316 | int poll_owner; | |
317 | #endif | |
318 | struct net_device *dev; | |
319 | struct sk_buff *gro_list; | |
320 | struct sk_buff *skb; | |
321 | struct hrtimer timer; | |
322 | struct list_head dev_list; | |
323 | struct hlist_node napi_hash_node; | |
324 | unsigned int napi_id; | |
325 | }; | |
326 | ||
327 | enum { | |
328 | NAPI_STATE_SCHED, /* Poll is scheduled */ | |
329 | NAPI_STATE_DISABLE, /* Disable pending */ | |
330 | NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */ | |
331 | NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */ | |
332 | NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */ | |
333 | }; | |
334 | ||
335 | enum gro_result { | |
336 | GRO_MERGED, | |
337 | GRO_MERGED_FREE, | |
338 | GRO_HELD, | |
339 | GRO_NORMAL, | |
340 | GRO_DROP, | |
341 | }; | |
342 | typedef enum gro_result gro_result_t; | |
343 | ||
344 | /* | |
345 | * enum rx_handler_result - Possible return values for rx_handlers. | |
346 | * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it | |
347 | * further. | |
348 | * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in | |
349 | * case skb->dev was changed by rx_handler. | |
350 | * @RX_HANDLER_EXACT: Force exact delivery, no wildcard. | |
351 | * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called. | |
352 | * | |
353 | * rx_handlers are functions called from inside __netif_receive_skb(), to do | |
354 | * special processing of the skb, prior to delivery to protocol handlers. | |
355 | * | |
356 | * Currently, a net_device can only have a single rx_handler registered. Trying | |
357 | * to register a second rx_handler will return -EBUSY. | |
358 | * | |
359 | * To register a rx_handler on a net_device, use netdev_rx_handler_register(). | |
360 | * To unregister a rx_handler on a net_device, use | |
361 | * netdev_rx_handler_unregister(). | |
362 | * | |
363 | * Upon return, rx_handler is expected to tell __netif_receive_skb() what to | |
364 | * do with the skb. | |
365 | * | |
366 | * If the rx_handler consumed to skb in some way, it should return | |
367 | * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for | |
368 | * the skb to be delivered in some other ways. | |
369 | * | |
370 | * If the rx_handler changed skb->dev, to divert the skb to another | |
371 | * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the | |
372 | * new device will be called if it exists. | |
373 | * | |
374 | * If the rx_handler consider the skb should be ignored, it should return | |
375 | * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that | |
376 | * are registered on exact device (ptype->dev == skb->dev). | |
377 | * | |
378 | * If the rx_handler didn't changed skb->dev, but want the skb to be normally | |
379 | * delivered, it should return RX_HANDLER_PASS. | |
380 | * | |
381 | * A device without a registered rx_handler will behave as if rx_handler | |
382 | * returned RX_HANDLER_PASS. | |
383 | */ | |
384 | ||
385 | enum rx_handler_result { | |
386 | RX_HANDLER_CONSUMED, | |
387 | RX_HANDLER_ANOTHER, | |
388 | RX_HANDLER_EXACT, | |
389 | RX_HANDLER_PASS, | |
390 | }; | |
391 | typedef enum rx_handler_result rx_handler_result_t; | |
392 | typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb); | |
393 | ||
394 | void __napi_schedule(struct napi_struct *n); | |
395 | void __napi_schedule_irqoff(struct napi_struct *n); | |
396 | ||
397 | static inline bool napi_disable_pending(struct napi_struct *n) | |
398 | { | |
399 | return test_bit(NAPI_STATE_DISABLE, &n->state); | |
400 | } | |
401 | ||
402 | /** | |
403 | * napi_schedule_prep - check if napi can be scheduled | |
404 | * @n: napi context | |
405 | * | |
406 | * Test if NAPI routine is already running, and if not mark | |
407 | * it as running. This is used as a condition variable | |
408 | * insure only one NAPI poll instance runs. We also make | |
409 | * sure there is no pending NAPI disable. | |
410 | */ | |
411 | static inline bool napi_schedule_prep(struct napi_struct *n) | |
412 | { | |
413 | return !napi_disable_pending(n) && | |
414 | !test_and_set_bit(NAPI_STATE_SCHED, &n->state); | |
415 | } | |
416 | ||
417 | /** | |
418 | * napi_schedule - schedule NAPI poll | |
419 | * @n: napi context | |
420 | * | |
421 | * Schedule NAPI poll routine to be called if it is not already | |
422 | * running. | |
423 | */ | |
424 | static inline void napi_schedule(struct napi_struct *n) | |
425 | { | |
426 | if (napi_schedule_prep(n)) | |
427 | __napi_schedule(n); | |
428 | } | |
429 | ||
430 | /** | |
431 | * napi_schedule_irqoff - schedule NAPI poll | |
432 | * @n: napi context | |
433 | * | |
434 | * Variant of napi_schedule(), assuming hard irqs are masked. | |
435 | */ | |
436 | static inline void napi_schedule_irqoff(struct napi_struct *n) | |
437 | { | |
438 | if (napi_schedule_prep(n)) | |
439 | __napi_schedule_irqoff(n); | |
440 | } | |
441 | ||
442 | /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */ | |
443 | static inline bool napi_reschedule(struct napi_struct *napi) | |
444 | { | |
445 | if (napi_schedule_prep(napi)) { | |
446 | __napi_schedule(napi); | |
447 | return true; | |
448 | } | |
449 | return false; | |
450 | } | |
451 | ||
452 | void __napi_complete(struct napi_struct *n); | |
453 | void napi_complete_done(struct napi_struct *n, int work_done); | |
454 | /** | |
455 | * napi_complete - NAPI processing complete | |
456 | * @n: napi context | |
457 | * | |
458 | * Mark NAPI processing as complete. | |
459 | * Consider using napi_complete_done() instead. | |
460 | */ | |
461 | static inline void napi_complete(struct napi_struct *n) | |
462 | { | |
463 | return napi_complete_done(n, 0); | |
464 | } | |
465 | ||
466 | /** | |
467 | * napi_hash_add - add a NAPI to global hashtable | |
468 | * @napi: napi context | |
469 | * | |
470 | * generate a new napi_id and store a @napi under it in napi_hash | |
471 | * Used for busy polling (CONFIG_NET_RX_BUSY_POLL) | |
472 | * Note: This is normally automatically done from netif_napi_add(), | |
473 | * so might disappear in a future linux version. | |
474 | */ | |
475 | void napi_hash_add(struct napi_struct *napi); | |
476 | ||
477 | /** | |
478 | * napi_hash_del - remove a NAPI from global table | |
479 | * @napi: napi context | |
480 | * | |
481 | * Warning: caller must observe rcu grace period | |
482 | * before freeing memory containing @napi, if | |
483 | * this function returns true. | |
484 | * Note: core networking stack automatically calls it | |
485 | * from netif_napi_del() | |
486 | * Drivers might want to call this helper to combine all | |
487 | * the needed rcu grace periods into a single one. | |
488 | */ | |
489 | bool napi_hash_del(struct napi_struct *napi); | |
490 | ||
491 | /** | |
492 | * napi_disable - prevent NAPI from scheduling | |
493 | * @n: napi context | |
494 | * | |
495 | * Stop NAPI from being scheduled on this context. | |
496 | * Waits till any outstanding processing completes. | |
497 | */ | |
498 | void napi_disable(struct napi_struct *n); | |
499 | ||
500 | /** | |
501 | * napi_enable - enable NAPI scheduling | |
502 | * @n: napi context | |
503 | * | |
504 | * Resume NAPI from being scheduled on this context. | |
505 | * Must be paired with napi_disable. | |
506 | */ | |
507 | static inline void napi_enable(struct napi_struct *n) | |
508 | { | |
509 | BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state)); | |
510 | smp_mb__before_atomic(); | |
511 | clear_bit(NAPI_STATE_SCHED, &n->state); | |
512 | clear_bit(NAPI_STATE_NPSVC, &n->state); | |
513 | } | |
514 | ||
515 | #ifdef CONFIG_SMP | |
516 | /** | |
517 | * napi_synchronize - wait until NAPI is not running | |
518 | * @n: napi context | |
519 | * | |
520 | * Wait until NAPI is done being scheduled on this context. | |
521 | * Waits till any outstanding processing completes but | |
522 | * does not disable future activations. | |
523 | */ | |
524 | static inline void napi_synchronize(const struct napi_struct *n) | |
525 | { | |
526 | while (test_bit(NAPI_STATE_SCHED, &n->state)) | |
527 | msleep(1); | |
528 | } | |
529 | #else | |
530 | # define napi_synchronize(n) barrier() | |
531 | #endif | |
532 | ||
533 | enum netdev_queue_state_t { | |
534 | __QUEUE_STATE_DRV_XOFF, | |
535 | __QUEUE_STATE_STACK_XOFF, | |
536 | __QUEUE_STATE_FROZEN, | |
537 | }; | |
538 | ||
539 | #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF) | |
540 | #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF) | |
541 | #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN) | |
542 | ||
543 | #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF) | |
544 | #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \ | |
545 | QUEUE_STATE_FROZEN) | |
546 | #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \ | |
547 | QUEUE_STATE_FROZEN) | |
548 | ||
549 | /* | |
550 | * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The | |
551 | * netif_tx_* functions below are used to manipulate this flag. The | |
552 | * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit | |
553 | * queue independently. The netif_xmit_*stopped functions below are called | |
554 | * to check if the queue has been stopped by the driver or stack (either | |
555 | * of the XOFF bits are set in the state). Drivers should not need to call | |
556 | * netif_xmit*stopped functions, they should only be using netif_tx_*. | |
557 | */ | |
558 | ||
559 | struct netdev_queue { | |
560 | /* | |
561 | * read mostly part | |
562 | */ | |
563 | struct net_device *dev; | |
564 | struct Qdisc __rcu *qdisc; | |
565 | struct Qdisc *qdisc_sleeping; | |
566 | #ifdef CONFIG_SYSFS | |
567 | struct kobject kobj; | |
568 | #endif | |
569 | #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) | |
570 | int numa_node; | |
571 | #endif | |
572 | /* | |
573 | * write mostly part | |
574 | */ | |
575 | spinlock_t _xmit_lock ____cacheline_aligned_in_smp; | |
576 | int xmit_lock_owner; | |
577 | /* | |
578 | * please use this field instead of dev->trans_start | |
579 | */ | |
580 | unsigned long trans_start; | |
581 | ||
582 | /* | |
583 | * Number of TX timeouts for this queue | |
584 | * (/sys/class/net/DEV/Q/trans_timeout) | |
585 | */ | |
586 | unsigned long trans_timeout; | |
587 | ||
588 | unsigned long state; | |
589 | ||
590 | #ifdef CONFIG_BQL | |
591 | struct dql dql; | |
592 | #endif | |
593 | unsigned long tx_maxrate; | |
594 | } ____cacheline_aligned_in_smp; | |
595 | ||
596 | static inline int netdev_queue_numa_node_read(const struct netdev_queue *q) | |
597 | { | |
598 | #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) | |
599 | return q->numa_node; | |
600 | #else | |
601 | return NUMA_NO_NODE; | |
602 | #endif | |
603 | } | |
604 | ||
605 | static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node) | |
606 | { | |
607 | #if defined(CONFIG_XPS) && defined(CONFIG_NUMA) | |
608 | q->numa_node = node; | |
609 | #endif | |
610 | } | |
611 | ||
612 | #ifdef CONFIG_RPS | |
613 | /* | |
614 | * This structure holds an RPS map which can be of variable length. The | |
615 | * map is an array of CPUs. | |
616 | */ | |
617 | struct rps_map { | |
618 | unsigned int len; | |
619 | struct rcu_head rcu; | |
620 | u16 cpus[0]; | |
621 | }; | |
622 | #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16))) | |
623 | ||
624 | /* | |
625 | * The rps_dev_flow structure contains the mapping of a flow to a CPU, the | |
626 | * tail pointer for that CPU's input queue at the time of last enqueue, and | |
627 | * a hardware filter index. | |
628 | */ | |
629 | struct rps_dev_flow { | |
630 | u16 cpu; | |
631 | u16 filter; | |
632 | unsigned int last_qtail; | |
633 | }; | |
634 | #define RPS_NO_FILTER 0xffff | |
635 | ||
636 | /* | |
637 | * The rps_dev_flow_table structure contains a table of flow mappings. | |
638 | */ | |
639 | struct rps_dev_flow_table { | |
640 | unsigned int mask; | |
641 | struct rcu_head rcu; | |
642 | struct rps_dev_flow flows[0]; | |
643 | }; | |
644 | #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \ | |
645 | ((_num) * sizeof(struct rps_dev_flow))) | |
646 | ||
647 | /* | |
648 | * The rps_sock_flow_table contains mappings of flows to the last CPU | |
649 | * on which they were processed by the application (set in recvmsg). | |
650 | * Each entry is a 32bit value. Upper part is the high order bits | |
651 | * of flow hash, lower part is cpu number. | |
652 | * rps_cpu_mask is used to partition the space, depending on number of | |
653 | * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1 | |
654 | * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f, | |
655 | * meaning we use 32-6=26 bits for the hash. | |
656 | */ | |
657 | struct rps_sock_flow_table { | |
658 | u32 mask; | |
659 | ||
660 | u32 ents[0] ____cacheline_aligned_in_smp; | |
661 | }; | |
662 | #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num])) | |
663 | ||
664 | #define RPS_NO_CPU 0xffff | |
665 | ||
666 | extern u32 rps_cpu_mask; | |
667 | extern struct rps_sock_flow_table __rcu *rps_sock_flow_table; | |
668 | ||
669 | static inline void rps_record_sock_flow(struct rps_sock_flow_table *table, | |
670 | u32 hash) | |
671 | { | |
672 | if (table && hash) { | |
673 | unsigned int index = hash & table->mask; | |
674 | u32 val = hash & ~rps_cpu_mask; | |
675 | ||
676 | /* We only give a hint, preemption can change cpu under us */ | |
677 | val |= raw_smp_processor_id(); | |
678 | ||
679 | if (table->ents[index] != val) | |
680 | table->ents[index] = val; | |
681 | } | |
682 | } | |
683 | ||
684 | #ifdef CONFIG_RFS_ACCEL | |
685 | bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id, | |
686 | u16 filter_id); | |
687 | #endif | |
688 | #endif /* CONFIG_RPS */ | |
689 | ||
690 | /* This structure contains an instance of an RX queue. */ | |
691 | struct netdev_rx_queue { | |
692 | #ifdef CONFIG_RPS | |
693 | struct rps_map __rcu *rps_map; | |
694 | struct rps_dev_flow_table __rcu *rps_flow_table; | |
695 | #endif | |
696 | struct kobject kobj; | |
697 | struct net_device *dev; | |
698 | } ____cacheline_aligned_in_smp; | |
699 | ||
700 | /* | |
701 | * RX queue sysfs structures and functions. | |
702 | */ | |
703 | struct rx_queue_attribute { | |
704 | struct attribute attr; | |
705 | ssize_t (*show)(struct netdev_rx_queue *queue, | |
706 | struct rx_queue_attribute *attr, char *buf); | |
707 | ssize_t (*store)(struct netdev_rx_queue *queue, | |
708 | struct rx_queue_attribute *attr, const char *buf, size_t len); | |
709 | }; | |
710 | ||
711 | #ifdef CONFIG_XPS | |
712 | /* | |
713 | * This structure holds an XPS map which can be of variable length. The | |
714 | * map is an array of queues. | |
715 | */ | |
716 | struct xps_map { | |
717 | unsigned int len; | |
718 | unsigned int alloc_len; | |
719 | struct rcu_head rcu; | |
720 | u16 queues[0]; | |
721 | }; | |
722 | #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16))) | |
723 | #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \ | |
724 | - sizeof(struct xps_map)) / sizeof(u16)) | |
725 | ||
726 | /* | |
727 | * This structure holds all XPS maps for device. Maps are indexed by CPU. | |
728 | */ | |
729 | struct xps_dev_maps { | |
730 | struct rcu_head rcu; | |
731 | struct xps_map __rcu *cpu_map[0]; | |
732 | }; | |
733 | #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \ | |
734 | (nr_cpu_ids * sizeof(struct xps_map *))) | |
735 | #endif /* CONFIG_XPS */ | |
736 | ||
737 | #define TC_MAX_QUEUE 16 | |
738 | #define TC_BITMASK 15 | |
739 | /* HW offloaded queuing disciplines txq count and offset maps */ | |
740 | struct netdev_tc_txq { | |
741 | u16 count; | |
742 | u16 offset; | |
743 | }; | |
744 | ||
745 | #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE) | |
746 | /* | |
747 | * This structure is to hold information about the device | |
748 | * configured to run FCoE protocol stack. | |
749 | */ | |
750 | struct netdev_fcoe_hbainfo { | |
751 | char manufacturer[64]; | |
752 | char serial_number[64]; | |
753 | char hardware_version[64]; | |
754 | char driver_version[64]; | |
755 | char optionrom_version[64]; | |
756 | char firmware_version[64]; | |
757 | char model[256]; | |
758 | char model_description[256]; | |
759 | }; | |
760 | #endif | |
761 | ||
762 | #define MAX_PHYS_ITEM_ID_LEN 32 | |
763 | ||
764 | /* This structure holds a unique identifier to identify some | |
765 | * physical item (port for example) used by a netdevice. | |
766 | */ | |
767 | struct netdev_phys_item_id { | |
768 | unsigned char id[MAX_PHYS_ITEM_ID_LEN]; | |
769 | unsigned char id_len; | |
770 | }; | |
771 | ||
772 | static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a, | |
773 | struct netdev_phys_item_id *b) | |
774 | { | |
775 | return a->id_len == b->id_len && | |
776 | memcmp(a->id, b->id, a->id_len) == 0; | |
777 | } | |
778 | ||
779 | typedef u16 (*select_queue_fallback_t)(struct net_device *dev, | |
780 | struct sk_buff *skb); | |
781 | ||
782 | /* | |
783 | * This structure defines the management hooks for network devices. | |
784 | * The following hooks can be defined; unless noted otherwise, they are | |
785 | * optional and can be filled with a null pointer. | |
786 | * | |
787 | * int (*ndo_init)(struct net_device *dev); | |
788 | * This function is called once when network device is registered. | |
789 | * The network device can use this to any late stage initializaton | |
790 | * or semantic validattion. It can fail with an error code which will | |
791 | * be propogated back to register_netdev | |
792 | * | |
793 | * void (*ndo_uninit)(struct net_device *dev); | |
794 | * This function is called when device is unregistered or when registration | |
795 | * fails. It is not called if init fails. | |
796 | * | |
797 | * int (*ndo_open)(struct net_device *dev); | |
798 | * This function is called when network device transistions to the up | |
799 | * state. | |
800 | * | |
801 | * int (*ndo_stop)(struct net_device *dev); | |
802 | * This function is called when network device transistions to the down | |
803 | * state. | |
804 | * | |
805 | * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, | |
806 | * struct net_device *dev); | |
807 | * Called when a packet needs to be transmitted. | |
808 | * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop | |
809 | * the queue before that can happen; it's for obsolete devices and weird | |
810 | * corner cases, but the stack really does a non-trivial amount | |
811 | * of useless work if you return NETDEV_TX_BUSY. | |
812 | * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX) | |
813 | * Required can not be NULL. | |
814 | * | |
815 | * netdev_features_t (*ndo_fix_features)(struct net_device *dev, | |
816 | * netdev_features_t features); | |
817 | * Adjusts the requested feature flags according to device-specific | |
818 | * constraints, and returns the resulting flags. Must not modify | |
819 | * the device state. | |
820 | * | |
821 | * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb, | |
822 | * void *accel_priv, select_queue_fallback_t fallback); | |
823 | * Called to decide which queue to when device supports multiple | |
824 | * transmit queues. | |
825 | * | |
826 | * void (*ndo_change_rx_flags)(struct net_device *dev, int flags); | |
827 | * This function is called to allow device receiver to make | |
828 | * changes to configuration when multicast or promiscious is enabled. | |
829 | * | |
830 | * void (*ndo_set_rx_mode)(struct net_device *dev); | |
831 | * This function is called device changes address list filtering. | |
832 | * If driver handles unicast address filtering, it should set | |
833 | * IFF_UNICAST_FLT to its priv_flags. | |
834 | * | |
835 | * int (*ndo_set_mac_address)(struct net_device *dev, void *addr); | |
836 | * This function is called when the Media Access Control address | |
837 | * needs to be changed. If this interface is not defined, the | |
838 | * mac address can not be changed. | |
839 | * | |
840 | * int (*ndo_validate_addr)(struct net_device *dev); | |
841 | * Test if Media Access Control address is valid for the device. | |
842 | * | |
843 | * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd); | |
844 | * Called when a user request an ioctl which can't be handled by | |
845 | * the generic interface code. If not defined ioctl's return | |
846 | * not supported error code. | |
847 | * | |
848 | * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map); | |
849 | * Used to set network devices bus interface parameters. This interface | |
850 | * is retained for legacy reason, new devices should use the bus | |
851 | * interface (PCI) for low level management. | |
852 | * | |
853 | * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu); | |
854 | * Called when a user wants to change the Maximum Transfer Unit | |
855 | * of a device. If not defined, any request to change MTU will | |
856 | * will return an error. | |
857 | * | |
858 | * void (*ndo_tx_timeout)(struct net_device *dev); | |
859 | * Callback uses when the transmitter has not made any progress | |
860 | * for dev->watchdog ticks. | |
861 | * | |
862 | * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, | |
863 | * struct rtnl_link_stats64 *storage); | |
864 | * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); | |
865 | * Called when a user wants to get the network device usage | |
866 | * statistics. Drivers must do one of the following: | |
867 | * 1. Define @ndo_get_stats64 to fill in a zero-initialised | |
868 | * rtnl_link_stats64 structure passed by the caller. | |
869 | * 2. Define @ndo_get_stats to update a net_device_stats structure | |
870 | * (which should normally be dev->stats) and return a pointer to | |
871 | * it. The structure may be changed asynchronously only if each | |
872 | * field is written atomically. | |
873 | * 3. Update dev->stats asynchronously and atomically, and define | |
874 | * neither operation. | |
875 | * | |
876 | * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid); | |
877 | * If device support VLAN filtering this function is called when a | |
878 | * VLAN id is registered. | |
879 | * | |
880 | * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid); | |
881 | * If device support VLAN filtering this function is called when a | |
882 | * VLAN id is unregistered. | |
883 | * | |
884 | * void (*ndo_poll_controller)(struct net_device *dev); | |
885 | * | |
886 | * SR-IOV management functions. | |
887 | * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac); | |
888 | * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos); | |
889 | * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate, | |
890 | * int max_tx_rate); | |
891 | * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting); | |
892 | * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting); | |
893 | * int (*ndo_get_vf_config)(struct net_device *dev, | |
894 | * int vf, struct ifla_vf_info *ivf); | |
895 | * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state); | |
896 | * int (*ndo_set_vf_port)(struct net_device *dev, int vf, | |
897 | * struct nlattr *port[]); | |
898 | * | |
899 | * Enable or disable the VF ability to query its RSS Redirection Table and | |
900 | * Hash Key. This is needed since on some devices VF share this information | |
901 | * with PF and querying it may adduce a theoretical security risk. | |
902 | * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting); | |
903 | * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb); | |
904 | * int (*ndo_setup_tc)(struct net_device *dev, u8 tc) | |
905 | * Called to setup 'tc' number of traffic classes in the net device. This | |
906 | * is always called from the stack with the rtnl lock held and netif tx | |
907 | * queues stopped. This allows the netdevice to perform queue management | |
908 | * safely. | |
909 | * | |
910 | * Fiber Channel over Ethernet (FCoE) offload functions. | |
911 | * int (*ndo_fcoe_enable)(struct net_device *dev); | |
912 | * Called when the FCoE protocol stack wants to start using LLD for FCoE | |
913 | * so the underlying device can perform whatever needed configuration or | |
914 | * initialization to support acceleration of FCoE traffic. | |
915 | * | |
916 | * int (*ndo_fcoe_disable)(struct net_device *dev); | |
917 | * Called when the FCoE protocol stack wants to stop using LLD for FCoE | |
918 | * so the underlying device can perform whatever needed clean-ups to | |
919 | * stop supporting acceleration of FCoE traffic. | |
920 | * | |
921 | * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid, | |
922 | * struct scatterlist *sgl, unsigned int sgc); | |
923 | * Called when the FCoE Initiator wants to initialize an I/O that | |
924 | * is a possible candidate for Direct Data Placement (DDP). The LLD can | |
925 | * perform necessary setup and returns 1 to indicate the device is set up | |
926 | * successfully to perform DDP on this I/O, otherwise this returns 0. | |
927 | * | |
928 | * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid); | |
929 | * Called when the FCoE Initiator/Target is done with the DDPed I/O as | |
930 | * indicated by the FC exchange id 'xid', so the underlying device can | |
931 | * clean up and reuse resources for later DDP requests. | |
932 | * | |
933 | * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid, | |
934 | * struct scatterlist *sgl, unsigned int sgc); | |
935 | * Called when the FCoE Target wants to initialize an I/O that | |
936 | * is a possible candidate for Direct Data Placement (DDP). The LLD can | |
937 | * perform necessary setup and returns 1 to indicate the device is set up | |
938 | * successfully to perform DDP on this I/O, otherwise this returns 0. | |
939 | * | |
940 | * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, | |
941 | * struct netdev_fcoe_hbainfo *hbainfo); | |
942 | * Called when the FCoE Protocol stack wants information on the underlying | |
943 | * device. This information is utilized by the FCoE protocol stack to | |
944 | * register attributes with Fiber Channel management service as per the | |
945 | * FC-GS Fabric Device Management Information(FDMI) specification. | |
946 | * | |
947 | * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type); | |
948 | * Called when the underlying device wants to override default World Wide | |
949 | * Name (WWN) generation mechanism in FCoE protocol stack to pass its own | |
950 | * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE | |
951 | * protocol stack to use. | |
952 | * | |
953 | * RFS acceleration. | |
954 | * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb, | |
955 | * u16 rxq_index, u32 flow_id); | |
956 | * Set hardware filter for RFS. rxq_index is the target queue index; | |
957 | * flow_id is a flow ID to be passed to rps_may_expire_flow() later. | |
958 | * Return the filter ID on success, or a negative error code. | |
959 | * | |
960 | * Slave management functions (for bridge, bonding, etc). | |
961 | * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev); | |
962 | * Called to make another netdev an underling. | |
963 | * | |
964 | * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev); | |
965 | * Called to release previously enslaved netdev. | |
966 | * | |
967 | * Feature/offload setting functions. | |
968 | * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features); | |
969 | * Called to update device configuration to new features. Passed | |
970 | * feature set might be less than what was returned by ndo_fix_features()). | |
971 | * Must return >0 or -errno if it changed dev->features itself. | |
972 | * | |
973 | * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[], | |
974 | * struct net_device *dev, | |
975 | * const unsigned char *addr, u16 vid, u16 flags) | |
976 | * Adds an FDB entry to dev for addr. | |
977 | * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[], | |
978 | * struct net_device *dev, | |
979 | * const unsigned char *addr, u16 vid) | |
980 | * Deletes the FDB entry from dev coresponding to addr. | |
981 | * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb, | |
982 | * struct net_device *dev, struct net_device *filter_dev, | |
983 | * int idx) | |
984 | * Used to add FDB entries to dump requests. Implementers should add | |
985 | * entries to skb and update idx with the number of entries. | |
986 | * | |
987 | * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh, | |
988 | * u16 flags) | |
989 | * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq, | |
990 | * struct net_device *dev, u32 filter_mask, | |
991 | * int nlflags) | |
992 | * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh, | |
993 | * u16 flags); | |
994 | * | |
995 | * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier); | |
996 | * Called to change device carrier. Soft-devices (like dummy, team, etc) | |
997 | * which do not represent real hardware may define this to allow their | |
998 | * userspace components to manage their virtual carrier state. Devices | |
999 | * that determine carrier state from physical hardware properties (eg | |
1000 | * network cables) or protocol-dependent mechanisms (eg | |
1001 | * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function. | |
1002 | * | |
1003 | * int (*ndo_get_phys_port_id)(struct net_device *dev, | |
1004 | * struct netdev_phys_item_id *ppid); | |
1005 | * Called to get ID of physical port of this device. If driver does | |
1006 | * not implement this, it is assumed that the hw is not able to have | |
1007 | * multiple net devices on single physical port. | |
1008 | * | |
1009 | * void (*ndo_add_vxlan_port)(struct net_device *dev, | |
1010 | * sa_family_t sa_family, __be16 port); | |
1011 | * Called by vxlan to notiy a driver about the UDP port and socket | |
1012 | * address family that vxlan is listnening to. It is called only when | |
1013 | * a new port starts listening. The operation is protected by the | |
1014 | * vxlan_net->sock_lock. | |
1015 | * | |
1016 | * void (*ndo_add_geneve_port)(struct net_device *dev, | |
1017 | * sa_family_t sa_family, __be16 port); | |
1018 | * Called by geneve to notify a driver about the UDP port and socket | |
1019 | * address family that geneve is listnening to. It is called only when | |
1020 | * a new port starts listening. The operation is protected by the | |
1021 | * geneve_net->sock_lock. | |
1022 | * | |
1023 | * void (*ndo_del_geneve_port)(struct net_device *dev, | |
1024 | * sa_family_t sa_family, __be16 port); | |
1025 | * Called by geneve to notify the driver about a UDP port and socket | |
1026 | * address family that geneve is not listening to anymore. The operation | |
1027 | * is protected by the geneve_net->sock_lock. | |
1028 | * | |
1029 | * void (*ndo_del_vxlan_port)(struct net_device *dev, | |
1030 | * sa_family_t sa_family, __be16 port); | |
1031 | * Called by vxlan to notify the driver about a UDP port and socket | |
1032 | * address family that vxlan is not listening to anymore. The operation | |
1033 | * is protected by the vxlan_net->sock_lock. | |
1034 | * | |
1035 | * void* (*ndo_dfwd_add_station)(struct net_device *pdev, | |
1036 | * struct net_device *dev) | |
1037 | * Called by upper layer devices to accelerate switching or other | |
1038 | * station functionality into hardware. 'pdev is the lowerdev | |
1039 | * to use for the offload and 'dev' is the net device that will | |
1040 | * back the offload. Returns a pointer to the private structure | |
1041 | * the upper layer will maintain. | |
1042 | * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv) | |
1043 | * Called by upper layer device to delete the station created | |
1044 | * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing | |
1045 | * the station and priv is the structure returned by the add | |
1046 | * operation. | |
1047 | * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb, | |
1048 | * struct net_device *dev, | |
1049 | * void *priv); | |
1050 | * Callback to use for xmit over the accelerated station. This | |
1051 | * is used in place of ndo_start_xmit on accelerated net | |
1052 | * devices. | |
1053 | * netdev_features_t (*ndo_features_check) (struct sk_buff *skb, | |
1054 | * struct net_device *dev | |
1055 | * netdev_features_t features); | |
1056 | * Called by core transmit path to determine if device is capable of | |
1057 | * performing offload operations on a given packet. This is to give | |
1058 | * the device an opportunity to implement any restrictions that cannot | |
1059 | * be otherwise expressed by feature flags. The check is called with | |
1060 | * the set of features that the stack has calculated and it returns | |
1061 | * those the driver believes to be appropriate. | |
1062 | * int (*ndo_set_tx_maxrate)(struct net_device *dev, | |
1063 | * int queue_index, u32 maxrate); | |
1064 | * Called when a user wants to set a max-rate limitation of specific | |
1065 | * TX queue. | |
1066 | * int (*ndo_get_iflink)(const struct net_device *dev); | |
1067 | * Called to get the iflink value of this device. | |
1068 | * void (*ndo_change_proto_down)(struct net_device *dev, | |
1069 | * bool proto_down); | |
1070 | * This function is used to pass protocol port error state information | |
1071 | * to the switch driver. The switch driver can react to the proto_down | |
1072 | * by doing a phys down on the associated switch port. | |
1073 | * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb); | |
1074 | * This function is used to get egress tunnel information for given skb. | |
1075 | * This is useful for retrieving outer tunnel header parameters while | |
1076 | * sampling packet. | |
1077 | * | |
1078 | */ | |
1079 | struct net_device_ops { | |
1080 | int (*ndo_init)(struct net_device *dev); | |
1081 | void (*ndo_uninit)(struct net_device *dev); | |
1082 | int (*ndo_open)(struct net_device *dev); | |
1083 | int (*ndo_stop)(struct net_device *dev); | |
1084 | netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, | |
1085 | struct net_device *dev); | |
1086 | netdev_features_t (*ndo_features_check)(struct sk_buff *skb, | |
1087 | struct net_device *dev, | |
1088 | netdev_features_t features); | |
1089 | u16 (*ndo_select_queue)(struct net_device *dev, | |
1090 | struct sk_buff *skb, | |
1091 | void *accel_priv, | |
1092 | select_queue_fallback_t fallback); | |
1093 | void (*ndo_change_rx_flags)(struct net_device *dev, | |
1094 | int flags); | |
1095 | void (*ndo_set_rx_mode)(struct net_device *dev); | |
1096 | int (*ndo_set_mac_address)(struct net_device *dev, | |
1097 | void *addr); | |
1098 | int (*ndo_validate_addr)(struct net_device *dev); | |
1099 | int (*ndo_do_ioctl)(struct net_device *dev, | |
1100 | struct ifreq *ifr, int cmd); | |
1101 | int (*ndo_set_config)(struct net_device *dev, | |
1102 | struct ifmap *map); | |
1103 | int (*ndo_change_mtu)(struct net_device *dev, | |
1104 | int new_mtu); | |
1105 | int (*ndo_neigh_setup)(struct net_device *dev, | |
1106 | struct neigh_parms *); | |
1107 | void (*ndo_tx_timeout) (struct net_device *dev); | |
1108 | ||
1109 | struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev, | |
1110 | struct rtnl_link_stats64 *storage); | |
1111 | struct net_device_stats* (*ndo_get_stats)(struct net_device *dev); | |
1112 | ||
1113 | int (*ndo_vlan_rx_add_vid)(struct net_device *dev, | |
1114 | __be16 proto, u16 vid); | |
1115 | int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, | |
1116 | __be16 proto, u16 vid); | |
1117 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
1118 | void (*ndo_poll_controller)(struct net_device *dev); | |
1119 | int (*ndo_netpoll_setup)(struct net_device *dev, | |
1120 | struct netpoll_info *info); | |
1121 | void (*ndo_netpoll_cleanup)(struct net_device *dev); | |
1122 | #endif | |
1123 | #ifdef CONFIG_NET_RX_BUSY_POLL | |
1124 | int (*ndo_busy_poll)(struct napi_struct *dev); | |
1125 | #endif | |
1126 | int (*ndo_set_vf_mac)(struct net_device *dev, | |
1127 | int queue, u8 *mac); | |
1128 | int (*ndo_set_vf_vlan)(struct net_device *dev, | |
1129 | int queue, u16 vlan, u8 qos); | |
1130 | int (*ndo_set_vf_rate)(struct net_device *dev, | |
1131 | int vf, int min_tx_rate, | |
1132 | int max_tx_rate); | |
1133 | int (*ndo_set_vf_spoofchk)(struct net_device *dev, | |
1134 | int vf, bool setting); | |
1135 | int (*ndo_set_vf_trust)(struct net_device *dev, | |
1136 | int vf, bool setting); | |
1137 | int (*ndo_get_vf_config)(struct net_device *dev, | |
1138 | int vf, | |
1139 | struct ifla_vf_info *ivf); | |
1140 | int (*ndo_set_vf_link_state)(struct net_device *dev, | |
1141 | int vf, int link_state); | |
1142 | int (*ndo_get_vf_stats)(struct net_device *dev, | |
1143 | int vf, | |
1144 | struct ifla_vf_stats | |
1145 | *vf_stats); | |
1146 | int (*ndo_set_vf_port)(struct net_device *dev, | |
1147 | int vf, | |
1148 | struct nlattr *port[]); | |
1149 | int (*ndo_get_vf_port)(struct net_device *dev, | |
1150 | int vf, struct sk_buff *skb); | |
1151 | int (*ndo_set_vf_rss_query_en)( | |
1152 | struct net_device *dev, | |
1153 | int vf, bool setting); | |
1154 | int (*ndo_setup_tc)(struct net_device *dev, u8 tc); | |
1155 | #if IS_ENABLED(CONFIG_FCOE) | |
1156 | int (*ndo_fcoe_enable)(struct net_device *dev); | |
1157 | int (*ndo_fcoe_disable)(struct net_device *dev); | |
1158 | int (*ndo_fcoe_ddp_setup)(struct net_device *dev, | |
1159 | u16 xid, | |
1160 | struct scatterlist *sgl, | |
1161 | unsigned int sgc); | |
1162 | int (*ndo_fcoe_ddp_done)(struct net_device *dev, | |
1163 | u16 xid); | |
1164 | int (*ndo_fcoe_ddp_target)(struct net_device *dev, | |
1165 | u16 xid, | |
1166 | struct scatterlist *sgl, | |
1167 | unsigned int sgc); | |
1168 | int (*ndo_fcoe_get_hbainfo)(struct net_device *dev, | |
1169 | struct netdev_fcoe_hbainfo *hbainfo); | |
1170 | #endif | |
1171 | ||
1172 | #if IS_ENABLED(CONFIG_LIBFCOE) | |
1173 | #define NETDEV_FCOE_WWNN 0 | |
1174 | #define NETDEV_FCOE_WWPN 1 | |
1175 | int (*ndo_fcoe_get_wwn)(struct net_device *dev, | |
1176 | u64 *wwn, int type); | |
1177 | #endif | |
1178 | ||
1179 | #ifdef CONFIG_RFS_ACCEL | |
1180 | int (*ndo_rx_flow_steer)(struct net_device *dev, | |
1181 | const struct sk_buff *skb, | |
1182 | u16 rxq_index, | |
1183 | u32 flow_id); | |
1184 | #endif | |
1185 | int (*ndo_add_slave)(struct net_device *dev, | |
1186 | struct net_device *slave_dev); | |
1187 | int (*ndo_del_slave)(struct net_device *dev, | |
1188 | struct net_device *slave_dev); | |
1189 | netdev_features_t (*ndo_fix_features)(struct net_device *dev, | |
1190 | netdev_features_t features); | |
1191 | int (*ndo_set_features)(struct net_device *dev, | |
1192 | netdev_features_t features); | |
1193 | int (*ndo_neigh_construct)(struct neighbour *n); | |
1194 | void (*ndo_neigh_destroy)(struct neighbour *n); | |
1195 | ||
1196 | int (*ndo_fdb_add)(struct ndmsg *ndm, | |
1197 | struct nlattr *tb[], | |
1198 | struct net_device *dev, | |
1199 | const unsigned char *addr, | |
1200 | u16 vid, | |
1201 | u16 flags); | |
1202 | int (*ndo_fdb_del)(struct ndmsg *ndm, | |
1203 | struct nlattr *tb[], | |
1204 | struct net_device *dev, | |
1205 | const unsigned char *addr, | |
1206 | u16 vid); | |
1207 | int (*ndo_fdb_dump)(struct sk_buff *skb, | |
1208 | struct netlink_callback *cb, | |
1209 | struct net_device *dev, | |
1210 | struct net_device *filter_dev, | |
1211 | int idx); | |
1212 | ||
1213 | int (*ndo_bridge_setlink)(struct net_device *dev, | |
1214 | struct nlmsghdr *nlh, | |
1215 | u16 flags); | |
1216 | int (*ndo_bridge_getlink)(struct sk_buff *skb, | |
1217 | u32 pid, u32 seq, | |
1218 | struct net_device *dev, | |
1219 | u32 filter_mask, | |
1220 | int nlflags); | |
1221 | int (*ndo_bridge_dellink)(struct net_device *dev, | |
1222 | struct nlmsghdr *nlh, | |
1223 | u16 flags); | |
1224 | int (*ndo_change_carrier)(struct net_device *dev, | |
1225 | bool new_carrier); | |
1226 | int (*ndo_get_phys_port_id)(struct net_device *dev, | |
1227 | struct netdev_phys_item_id *ppid); | |
1228 | int (*ndo_get_phys_port_name)(struct net_device *dev, | |
1229 | char *name, size_t len); | |
1230 | void (*ndo_add_vxlan_port)(struct net_device *dev, | |
1231 | sa_family_t sa_family, | |
1232 | __be16 port); | |
1233 | void (*ndo_del_vxlan_port)(struct net_device *dev, | |
1234 | sa_family_t sa_family, | |
1235 | __be16 port); | |
1236 | void (*ndo_add_geneve_port)(struct net_device *dev, | |
1237 | sa_family_t sa_family, | |
1238 | __be16 port); | |
1239 | void (*ndo_del_geneve_port)(struct net_device *dev, | |
1240 | sa_family_t sa_family, | |
1241 | __be16 port); | |
1242 | void* (*ndo_dfwd_add_station)(struct net_device *pdev, | |
1243 | struct net_device *dev); | |
1244 | void (*ndo_dfwd_del_station)(struct net_device *pdev, | |
1245 | void *priv); | |
1246 | ||
1247 | netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb, | |
1248 | struct net_device *dev, | |
1249 | void *priv); | |
1250 | int (*ndo_get_lock_subclass)(struct net_device *dev); | |
1251 | int (*ndo_set_tx_maxrate)(struct net_device *dev, | |
1252 | int queue_index, | |
1253 | u32 maxrate); | |
1254 | int (*ndo_get_iflink)(const struct net_device *dev); | |
1255 | int (*ndo_change_proto_down)(struct net_device *dev, | |
1256 | bool proto_down); | |
1257 | int (*ndo_fill_metadata_dst)(struct net_device *dev, | |
1258 | struct sk_buff *skb); | |
1259 | }; | |
1260 | ||
1261 | /** | |
1262 | * enum net_device_priv_flags - &struct net_device priv_flags | |
1263 | * | |
1264 | * These are the &struct net_device, they are only set internally | |
1265 | * by drivers and used in the kernel. These flags are invisible to | |
1266 | * userspace, this means that the order of these flags can change | |
1267 | * during any kernel release. | |
1268 | * | |
1269 | * You should have a pretty good reason to be extending these flags. | |
1270 | * | |
1271 | * @IFF_802_1Q_VLAN: 802.1Q VLAN device | |
1272 | * @IFF_EBRIDGE: Ethernet bridging device | |
1273 | * @IFF_BONDING: bonding master or slave | |
1274 | * @IFF_ISATAP: ISATAP interface (RFC4214) | |
1275 | * @IFF_WAN_HDLC: WAN HDLC device | |
1276 | * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to | |
1277 | * release skb->dst | |
1278 | * @IFF_DONT_BRIDGE: disallow bridging this ether dev | |
1279 | * @IFF_DISABLE_NETPOLL: disable netpoll at run-time | |
1280 | * @IFF_MACVLAN_PORT: device used as macvlan port | |
1281 | * @IFF_BRIDGE_PORT: device used as bridge port | |
1282 | * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port | |
1283 | * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit | |
1284 | * @IFF_UNICAST_FLT: Supports unicast filtering | |
1285 | * @IFF_TEAM_PORT: device used as team port | |
1286 | * @IFF_SUPP_NOFCS: device supports sending custom FCS | |
1287 | * @IFF_LIVE_ADDR_CHANGE: device supports hardware address | |
1288 | * change when it's running | |
1289 | * @IFF_MACVLAN: Macvlan device | |
1290 | * @IFF_L3MDEV_MASTER: device is an L3 master device | |
1291 | * @IFF_NO_QUEUE: device can run without qdisc attached | |
1292 | * @IFF_OPENVSWITCH: device is a Open vSwitch master | |
1293 | * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device | |
1294 | * @IFF_TEAM: device is a team device | |
1295 | */ | |
1296 | enum netdev_priv_flags { | |
1297 | IFF_802_1Q_VLAN = 1<<0, | |
1298 | IFF_EBRIDGE = 1<<1, | |
1299 | IFF_BONDING = 1<<2, | |
1300 | IFF_ISATAP = 1<<3, | |
1301 | IFF_WAN_HDLC = 1<<4, | |
1302 | IFF_XMIT_DST_RELEASE = 1<<5, | |
1303 | IFF_DONT_BRIDGE = 1<<6, | |
1304 | IFF_DISABLE_NETPOLL = 1<<7, | |
1305 | IFF_MACVLAN_PORT = 1<<8, | |
1306 | IFF_BRIDGE_PORT = 1<<9, | |
1307 | IFF_OVS_DATAPATH = 1<<10, | |
1308 | IFF_TX_SKB_SHARING = 1<<11, | |
1309 | IFF_UNICAST_FLT = 1<<12, | |
1310 | IFF_TEAM_PORT = 1<<13, | |
1311 | IFF_SUPP_NOFCS = 1<<14, | |
1312 | IFF_LIVE_ADDR_CHANGE = 1<<15, | |
1313 | IFF_MACVLAN = 1<<16, | |
1314 | IFF_XMIT_DST_RELEASE_PERM = 1<<17, | |
1315 | IFF_IPVLAN_MASTER = 1<<18, | |
1316 | IFF_IPVLAN_SLAVE = 1<<19, | |
1317 | IFF_L3MDEV_MASTER = 1<<20, | |
1318 | IFF_NO_QUEUE = 1<<21, | |
1319 | IFF_OPENVSWITCH = 1<<22, | |
1320 | IFF_L3MDEV_SLAVE = 1<<23, | |
1321 | IFF_TEAM = 1<<24, | |
1322 | }; | |
1323 | ||
1324 | #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN | |
1325 | #define IFF_EBRIDGE IFF_EBRIDGE | |
1326 | #define IFF_BONDING IFF_BONDING | |
1327 | #define IFF_ISATAP IFF_ISATAP | |
1328 | #define IFF_WAN_HDLC IFF_WAN_HDLC | |
1329 | #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE | |
1330 | #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE | |
1331 | #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL | |
1332 | #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT | |
1333 | #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT | |
1334 | #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH | |
1335 | #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING | |
1336 | #define IFF_UNICAST_FLT IFF_UNICAST_FLT | |
1337 | #define IFF_TEAM_PORT IFF_TEAM_PORT | |
1338 | #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS | |
1339 | #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE | |
1340 | #define IFF_MACVLAN IFF_MACVLAN | |
1341 | #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM | |
1342 | #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER | |
1343 | #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE | |
1344 | #define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER | |
1345 | #define IFF_NO_QUEUE IFF_NO_QUEUE | |
1346 | #define IFF_OPENVSWITCH IFF_OPENVSWITCH | |
1347 | #define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE | |
1348 | #define IFF_TEAM IFF_TEAM | |
1349 | ||
1350 | /** | |
1351 | * struct net_device - The DEVICE structure. | |
1352 | * Actually, this whole structure is a big mistake. It mixes I/O | |
1353 | * data with strictly "high-level" data, and it has to know about | |
1354 | * almost every data structure used in the INET module. | |
1355 | * | |
1356 | * @name: This is the first field of the "visible" part of this structure | |
1357 | * (i.e. as seen by users in the "Space.c" file). It is the name | |
1358 | * of the interface. | |
1359 | * | |
1360 | * @name_hlist: Device name hash chain, please keep it close to name[] | |
1361 | * @ifalias: SNMP alias | |
1362 | * @mem_end: Shared memory end | |
1363 | * @mem_start: Shared memory start | |
1364 | * @base_addr: Device I/O address | |
1365 | * @irq: Device IRQ number | |
1366 | * | |
1367 | * @carrier_changes: Stats to monitor carrier on<->off transitions | |
1368 | * | |
1369 | * @state: Generic network queuing layer state, see netdev_state_t | |
1370 | * @dev_list: The global list of network devices | |
1371 | * @napi_list: List entry, that is used for polling napi devices | |
1372 | * @unreg_list: List entry, that is used, when we are unregistering the | |
1373 | * device, see the function unregister_netdev | |
1374 | * @close_list: List entry, that is used, when we are closing the device | |
1375 | * | |
1376 | * @adj_list: Directly linked devices, like slaves for bonding | |
1377 | * @all_adj_list: All linked devices, *including* neighbours | |
1378 | * @features: Currently active device features | |
1379 | * @hw_features: User-changeable features | |
1380 | * | |
1381 | * @wanted_features: User-requested features | |
1382 | * @vlan_features: Mask of features inheritable by VLAN devices | |
1383 | * | |
1384 | * @hw_enc_features: Mask of features inherited by encapsulating devices | |
1385 | * This field indicates what encapsulation | |
1386 | * offloads the hardware is capable of doing, | |
1387 | * and drivers will need to set them appropriately. | |
1388 | * | |
1389 | * @mpls_features: Mask of features inheritable by MPLS | |
1390 | * | |
1391 | * @ifindex: interface index | |
1392 | * @group: The group, that the device belongs to | |
1393 | * | |
1394 | * @stats: Statistics struct, which was left as a legacy, use | |
1395 | * rtnl_link_stats64 instead | |
1396 | * | |
1397 | * @rx_dropped: Dropped packets by core network, | |
1398 | * do not use this in drivers | |
1399 | * @tx_dropped: Dropped packets by core network, | |
1400 | * do not use this in drivers | |
1401 | * | |
1402 | * @wireless_handlers: List of functions to handle Wireless Extensions, | |
1403 | * instead of ioctl, | |
1404 | * see <net/iw_handler.h> for details. | |
1405 | * @wireless_data: Instance data managed by the core of wireless extensions | |
1406 | * | |
1407 | * @netdev_ops: Includes several pointers to callbacks, | |
1408 | * if one wants to override the ndo_*() functions | |
1409 | * @ethtool_ops: Management operations | |
1410 | * @header_ops: Includes callbacks for creating,parsing,caching,etc | |
1411 | * of Layer 2 headers. | |
1412 | * | |
1413 | * @flags: Interface flags (a la BSD) | |
1414 | * @priv_flags: Like 'flags' but invisible to userspace, | |
1415 | * see if.h for the definitions | |
1416 | * @gflags: Global flags ( kept as legacy ) | |
1417 | * @padded: How much padding added by alloc_netdev() | |
1418 | * @operstate: RFC2863 operstate | |
1419 | * @link_mode: Mapping policy to operstate | |
1420 | * @if_port: Selectable AUI, TP, ... | |
1421 | * @dma: DMA channel | |
1422 | * @mtu: Interface MTU value | |
1423 | * @type: Interface hardware type | |
1424 | * @hard_header_len: Hardware header length, which means that this is the | |
1425 | * minimum size of a packet. | |
1426 | * | |
1427 | * @needed_headroom: Extra headroom the hardware may need, but not in all | |
1428 | * cases can this be guaranteed | |
1429 | * @needed_tailroom: Extra tailroom the hardware may need, but not in all | |
1430 | * cases can this be guaranteed. Some cases also use | |
1431 | * LL_MAX_HEADER instead to allocate the skb | |
1432 | * | |
1433 | * interface address info: | |
1434 | * | |
1435 | * @perm_addr: Permanent hw address | |
1436 | * @addr_assign_type: Hw address assignment type | |
1437 | * @addr_len: Hardware address length | |
1438 | * @neigh_priv_len; Used in neigh_alloc(), | |
1439 | * initialized only in atm/clip.c | |
1440 | * @dev_id: Used to differentiate devices that share | |
1441 | * the same link layer address | |
1442 | * @dev_port: Used to differentiate devices that share | |
1443 | * the same function | |
1444 | * @addr_list_lock: XXX: need comments on this one | |
1445 | * @uc_promisc: Counter, that indicates, that promiscuous mode | |
1446 | * has been enabled due to the need to listen to | |
1447 | * additional unicast addresses in a device that | |
1448 | * does not implement ndo_set_rx_mode() | |
1449 | * @uc: unicast mac addresses | |
1450 | * @mc: multicast mac addresses | |
1451 | * @dev_addrs: list of device hw addresses | |
1452 | * @queues_kset: Group of all Kobjects in the Tx and RX queues | |
1453 | * @promiscuity: Number of times, the NIC is told to work in | |
1454 | * Promiscuous mode, if it becomes 0 the NIC will | |
1455 | * exit from working in Promiscuous mode | |
1456 | * @allmulti: Counter, enables or disables allmulticast mode | |
1457 | * | |
1458 | * @vlan_info: VLAN info | |
1459 | * @dsa_ptr: dsa specific data | |
1460 | * @tipc_ptr: TIPC specific data | |
1461 | * @atalk_ptr: AppleTalk link | |
1462 | * @ip_ptr: IPv4 specific data | |
1463 | * @dn_ptr: DECnet specific data | |
1464 | * @ip6_ptr: IPv6 specific data | |
1465 | * @ax25_ptr: AX.25 specific data | |
1466 | * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering | |
1467 | * | |
1468 | * @last_rx: Time of last Rx | |
1469 | * @dev_addr: Hw address (before bcast, | |
1470 | * because most packets are unicast) | |
1471 | * | |
1472 | * @_rx: Array of RX queues | |
1473 | * @num_rx_queues: Number of RX queues | |
1474 | * allocated at register_netdev() time | |
1475 | * @real_num_rx_queues: Number of RX queues currently active in device | |
1476 | * | |
1477 | * @rx_handler: handler for received packets | |
1478 | * @rx_handler_data: XXX: need comments on this one | |
1479 | * @ingress_queue: XXX: need comments on this one | |
1480 | * @broadcast: hw bcast address | |
1481 | * | |
1482 | * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts, | |
1483 | * indexed by RX queue number. Assigned by driver. | |
1484 | * This must only be set if the ndo_rx_flow_steer | |
1485 | * operation is defined | |
1486 | * @index_hlist: Device index hash chain | |
1487 | * | |
1488 | * @_tx: Array of TX queues | |
1489 | * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time | |
1490 | * @real_num_tx_queues: Number of TX queues currently active in device | |
1491 | * @qdisc: Root qdisc from userspace point of view | |
1492 | * @tx_queue_len: Max frames per queue allowed | |
1493 | * @tx_global_lock: XXX: need comments on this one | |
1494 | * | |
1495 | * @xps_maps: XXX: need comments on this one | |
1496 | * | |
1497 | * @offload_fwd_mark: Offload device fwding mark | |
1498 | * | |
1499 | * @trans_start: Time (in jiffies) of last Tx | |
1500 | * @watchdog_timeo: Represents the timeout that is used by | |
1501 | * the watchdog ( see dev_watchdog() ) | |
1502 | * @watchdog_timer: List of timers | |
1503 | * | |
1504 | * @pcpu_refcnt: Number of references to this device | |
1505 | * @todo_list: Delayed register/unregister | |
1506 | * @link_watch_list: XXX: need comments on this one | |
1507 | * | |
1508 | * @reg_state: Register/unregister state machine | |
1509 | * @dismantle: Device is going to be freed | |
1510 | * @rtnl_link_state: This enum represents the phases of creating | |
1511 | * a new link | |
1512 | * | |
1513 | * @destructor: Called from unregister, | |
1514 | * can be used to call free_netdev | |
1515 | * @npinfo: XXX: need comments on this one | |
1516 | * @nd_net: Network namespace this network device is inside | |
1517 | * | |
1518 | * @ml_priv: Mid-layer private | |
1519 | * @lstats: Loopback statistics | |
1520 | * @tstats: Tunnel statistics | |
1521 | * @dstats: Dummy statistics | |
1522 | * @vstats: Virtual ethernet statistics | |
1523 | * | |
1524 | * @garp_port: GARP | |
1525 | * @mrp_port: MRP | |
1526 | * | |
1527 | * @dev: Class/net/name entry | |
1528 | * @sysfs_groups: Space for optional device, statistics and wireless | |
1529 | * sysfs groups | |
1530 | * | |
1531 | * @sysfs_rx_queue_group: Space for optional per-rx queue attributes | |
1532 | * @rtnl_link_ops: Rtnl_link_ops | |
1533 | * | |
1534 | * @gso_max_size: Maximum size of generic segmentation offload | |
1535 | * @gso_max_segs: Maximum number of segments that can be passed to the | |
1536 | * NIC for GSO | |
1537 | * @gso_min_segs: Minimum number of segments that can be passed to the | |
1538 | * NIC for GSO | |
1539 | * | |
1540 | * @dcbnl_ops: Data Center Bridging netlink ops | |
1541 | * @num_tc: Number of traffic classes in the net device | |
1542 | * @tc_to_txq: XXX: need comments on this one | |
1543 | * @prio_tc_map XXX: need comments on this one | |
1544 | * | |
1545 | * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp | |
1546 | * | |
1547 | * @priomap: XXX: need comments on this one | |
1548 | * @phydev: Physical device may attach itself | |
1549 | * for hardware timestamping | |
1550 | * | |
1551 | * @qdisc_tx_busylock: XXX: need comments on this one | |
1552 | * | |
1553 | * @proto_down: protocol port state information can be sent to the | |
1554 | * switch driver and used to set the phys state of the | |
1555 | * switch port. | |
1556 | * | |
1557 | * FIXME: cleanup struct net_device such that network protocol info | |
1558 | * moves out. | |
1559 | */ | |
1560 | ||
1561 | struct net_device { | |
1562 | char name[IFNAMSIZ]; | |
1563 | struct hlist_node name_hlist; | |
1564 | char *ifalias; | |
1565 | /* | |
1566 | * I/O specific fields | |
1567 | * FIXME: Merge these and struct ifmap into one | |
1568 | */ | |
1569 | unsigned long mem_end; | |
1570 | unsigned long mem_start; | |
1571 | unsigned long base_addr; | |
1572 | int irq; | |
1573 | ||
1574 | atomic_t carrier_changes; | |
1575 | ||
1576 | /* | |
1577 | * Some hardware also needs these fields (state,dev_list, | |
1578 | * napi_list,unreg_list,close_list) but they are not | |
1579 | * part of the usual set specified in Space.c. | |
1580 | */ | |
1581 | ||
1582 | unsigned long state; | |
1583 | ||
1584 | struct list_head dev_list; | |
1585 | struct list_head napi_list; | |
1586 | struct list_head unreg_list; | |
1587 | struct list_head close_list; | |
1588 | struct list_head ptype_all; | |
1589 | struct list_head ptype_specific; | |
1590 | ||
1591 | struct { | |
1592 | struct list_head upper; | |
1593 | struct list_head lower; | |
1594 | } adj_list; | |
1595 | ||
1596 | struct { | |
1597 | struct list_head upper; | |
1598 | struct list_head lower; | |
1599 | } all_adj_list; | |
1600 | ||
1601 | netdev_features_t features; | |
1602 | netdev_features_t hw_features; | |
1603 | netdev_features_t wanted_features; | |
1604 | netdev_features_t vlan_features; | |
1605 | netdev_features_t hw_enc_features; | |
1606 | netdev_features_t mpls_features; | |
1607 | ||
1608 | int ifindex; | |
1609 | int group; | |
1610 | ||
1611 | struct net_device_stats stats; | |
1612 | ||
1613 | atomic_long_t rx_dropped; | |
1614 | atomic_long_t tx_dropped; | |
1615 | ||
1616 | #ifdef CONFIG_WIRELESS_EXT | |
1617 | const struct iw_handler_def * wireless_handlers; | |
1618 | struct iw_public_data * wireless_data; | |
1619 | #endif | |
1620 | const struct net_device_ops *netdev_ops; | |
1621 | const struct ethtool_ops *ethtool_ops; | |
1622 | #ifdef CONFIG_NET_SWITCHDEV | |
1623 | const struct switchdev_ops *switchdev_ops; | |
1624 | #endif | |
1625 | #ifdef CONFIG_NET_L3_MASTER_DEV | |
1626 | const struct l3mdev_ops *l3mdev_ops; | |
1627 | #endif | |
1628 | ||
1629 | const struct header_ops *header_ops; | |
1630 | ||
1631 | unsigned int flags; | |
1632 | unsigned int priv_flags; | |
1633 | ||
1634 | unsigned short gflags; | |
1635 | unsigned short padded; | |
1636 | ||
1637 | unsigned char operstate; | |
1638 | unsigned char link_mode; | |
1639 | ||
1640 | unsigned char if_port; | |
1641 | unsigned char dma; | |
1642 | ||
1643 | unsigned int mtu; | |
1644 | unsigned short type; | |
1645 | unsigned short hard_header_len; | |
1646 | ||
1647 | unsigned short needed_headroom; | |
1648 | unsigned short needed_tailroom; | |
1649 | ||
1650 | /* Interface address info. */ | |
1651 | unsigned char perm_addr[MAX_ADDR_LEN]; | |
1652 | unsigned char addr_assign_type; | |
1653 | unsigned char addr_len; | |
1654 | unsigned short neigh_priv_len; | |
1655 | unsigned short dev_id; | |
1656 | unsigned short dev_port; | |
1657 | spinlock_t addr_list_lock; | |
1658 | unsigned char name_assign_type; | |
1659 | bool uc_promisc; | |
1660 | struct netdev_hw_addr_list uc; | |
1661 | struct netdev_hw_addr_list mc; | |
1662 | struct netdev_hw_addr_list dev_addrs; | |
1663 | ||
1664 | #ifdef CONFIG_SYSFS | |
1665 | struct kset *queues_kset; | |
1666 | #endif | |
1667 | unsigned int promiscuity; | |
1668 | unsigned int allmulti; | |
1669 | ||
1670 | ||
1671 | /* Protocol specific pointers */ | |
1672 | ||
1673 | #if IS_ENABLED(CONFIG_VLAN_8021Q) | |
1674 | struct vlan_info __rcu *vlan_info; | |
1675 | #endif | |
1676 | #if IS_ENABLED(CONFIG_NET_DSA) | |
1677 | struct dsa_switch_tree *dsa_ptr; | |
1678 | #endif | |
1679 | #if IS_ENABLED(CONFIG_TIPC) | |
1680 | struct tipc_bearer __rcu *tipc_ptr; | |
1681 | #endif | |
1682 | void *atalk_ptr; | |
1683 | struct in_device __rcu *ip_ptr; | |
1684 | struct dn_dev __rcu *dn_ptr; | |
1685 | struct inet6_dev __rcu *ip6_ptr; | |
1686 | void *ax25_ptr; | |
1687 | struct wireless_dev *ieee80211_ptr; | |
1688 | struct wpan_dev *ieee802154_ptr; | |
1689 | #if IS_ENABLED(CONFIG_MPLS_ROUTING) | |
1690 | struct mpls_dev __rcu *mpls_ptr; | |
1691 | #endif | |
1692 | ||
1693 | /* | |
1694 | * Cache lines mostly used on receive path (including eth_type_trans()) | |
1695 | */ | |
1696 | unsigned long last_rx; | |
1697 | ||
1698 | /* Interface address info used in eth_type_trans() */ | |
1699 | unsigned char *dev_addr; | |
1700 | ||
1701 | ||
1702 | #ifdef CONFIG_SYSFS | |
1703 | struct netdev_rx_queue *_rx; | |
1704 | ||
1705 | unsigned int num_rx_queues; | |
1706 | unsigned int real_num_rx_queues; | |
1707 | ||
1708 | #endif | |
1709 | ||
1710 | unsigned long gro_flush_timeout; | |
1711 | rx_handler_func_t __rcu *rx_handler; | |
1712 | void __rcu *rx_handler_data; | |
1713 | ||
1714 | #ifdef CONFIG_NET_CLS_ACT | |
1715 | struct tcf_proto __rcu *ingress_cl_list; | |
1716 | #endif | |
1717 | struct netdev_queue __rcu *ingress_queue; | |
1718 | #ifdef CONFIG_NETFILTER_INGRESS | |
1719 | struct list_head nf_hooks_ingress; | |
1720 | #endif | |
1721 | ||
1722 | unsigned char broadcast[MAX_ADDR_LEN]; | |
1723 | #ifdef CONFIG_RFS_ACCEL | |
1724 | struct cpu_rmap *rx_cpu_rmap; | |
1725 | #endif | |
1726 | struct hlist_node index_hlist; | |
1727 | ||
1728 | /* | |
1729 | * Cache lines mostly used on transmit path | |
1730 | */ | |
1731 | struct netdev_queue *_tx ____cacheline_aligned_in_smp; | |
1732 | unsigned int num_tx_queues; | |
1733 | unsigned int real_num_tx_queues; | |
1734 | struct Qdisc *qdisc; | |
1735 | unsigned long tx_queue_len; | |
1736 | spinlock_t tx_global_lock; | |
1737 | int watchdog_timeo; | |
1738 | ||
1739 | #ifdef CONFIG_XPS | |
1740 | struct xps_dev_maps __rcu *xps_maps; | |
1741 | #endif | |
1742 | #ifdef CONFIG_NET_CLS_ACT | |
1743 | struct tcf_proto __rcu *egress_cl_list; | |
1744 | #endif | |
1745 | #ifdef CONFIG_NET_SWITCHDEV | |
1746 | u32 offload_fwd_mark; | |
1747 | #endif | |
1748 | ||
1749 | /* These may be needed for future network-power-down code. */ | |
1750 | ||
1751 | /* | |
1752 | * trans_start here is expensive for high speed devices on SMP, | |
1753 | * please use netdev_queue->trans_start instead. | |
1754 | */ | |
1755 | unsigned long trans_start; | |
1756 | ||
1757 | struct timer_list watchdog_timer; | |
1758 | ||
1759 | int __percpu *pcpu_refcnt; | |
1760 | struct list_head todo_list; | |
1761 | ||
1762 | struct list_head link_watch_list; | |
1763 | ||
1764 | enum { NETREG_UNINITIALIZED=0, | |
1765 | NETREG_REGISTERED, /* completed register_netdevice */ | |
1766 | NETREG_UNREGISTERING, /* called unregister_netdevice */ | |
1767 | NETREG_UNREGISTERED, /* completed unregister todo */ | |
1768 | NETREG_RELEASED, /* called free_netdev */ | |
1769 | NETREG_DUMMY, /* dummy device for NAPI poll */ | |
1770 | } reg_state:8; | |
1771 | ||
1772 | bool dismantle; | |
1773 | ||
1774 | enum { | |
1775 | RTNL_LINK_INITIALIZED, | |
1776 | RTNL_LINK_INITIALIZING, | |
1777 | } rtnl_link_state:16; | |
1778 | ||
1779 | void (*destructor)(struct net_device *dev); | |
1780 | ||
1781 | #ifdef CONFIG_NETPOLL | |
1782 | struct netpoll_info __rcu *npinfo; | |
1783 | #endif | |
1784 | ||
1785 | possible_net_t nd_net; | |
1786 | ||
1787 | /* mid-layer private */ | |
1788 | union { | |
1789 | void *ml_priv; | |
1790 | struct pcpu_lstats __percpu *lstats; | |
1791 | struct pcpu_sw_netstats __percpu *tstats; | |
1792 | struct pcpu_dstats __percpu *dstats; | |
1793 | struct pcpu_vstats __percpu *vstats; | |
1794 | }; | |
1795 | ||
1796 | struct garp_port __rcu *garp_port; | |
1797 | struct mrp_port __rcu *mrp_port; | |
1798 | ||
1799 | struct device dev; | |
1800 | const struct attribute_group *sysfs_groups[4]; | |
1801 | const struct attribute_group *sysfs_rx_queue_group; | |
1802 | ||
1803 | const struct rtnl_link_ops *rtnl_link_ops; | |
1804 | ||
1805 | /* for setting kernel sock attribute on TCP connection setup */ | |
1806 | #define GSO_MAX_SIZE 65536 | |
1807 | unsigned int gso_max_size; | |
1808 | #define GSO_MAX_SEGS 65535 | |
1809 | u16 gso_max_segs; | |
1810 | u16 gso_min_segs; | |
1811 | #ifdef CONFIG_DCB | |
1812 | const struct dcbnl_rtnl_ops *dcbnl_ops; | |
1813 | #endif | |
1814 | u8 num_tc; | |
1815 | struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE]; | |
1816 | u8 prio_tc_map[TC_BITMASK + 1]; | |
1817 | ||
1818 | #if IS_ENABLED(CONFIG_FCOE) | |
1819 | unsigned int fcoe_ddp_xid; | |
1820 | #endif | |
1821 | #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) | |
1822 | struct netprio_map __rcu *priomap; | |
1823 | #endif | |
1824 | struct phy_device *phydev; | |
1825 | struct lock_class_key *qdisc_tx_busylock; | |
1826 | bool proto_down; | |
1827 | }; | |
1828 | #define to_net_dev(d) container_of(d, struct net_device, dev) | |
1829 | ||
1830 | #define NETDEV_ALIGN 32 | |
1831 | ||
1832 | static inline | |
1833 | int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio) | |
1834 | { | |
1835 | return dev->prio_tc_map[prio & TC_BITMASK]; | |
1836 | } | |
1837 | ||
1838 | static inline | |
1839 | int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc) | |
1840 | { | |
1841 | if (tc >= dev->num_tc) | |
1842 | return -EINVAL; | |
1843 | ||
1844 | dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK; | |
1845 | return 0; | |
1846 | } | |
1847 | ||
1848 | static inline | |
1849 | void netdev_reset_tc(struct net_device *dev) | |
1850 | { | |
1851 | dev->num_tc = 0; | |
1852 | memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq)); | |
1853 | memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map)); | |
1854 | } | |
1855 | ||
1856 | static inline | |
1857 | int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset) | |
1858 | { | |
1859 | if (tc >= dev->num_tc) | |
1860 | return -EINVAL; | |
1861 | ||
1862 | dev->tc_to_txq[tc].count = count; | |
1863 | dev->tc_to_txq[tc].offset = offset; | |
1864 | return 0; | |
1865 | } | |
1866 | ||
1867 | static inline | |
1868 | int netdev_set_num_tc(struct net_device *dev, u8 num_tc) | |
1869 | { | |
1870 | if (num_tc > TC_MAX_QUEUE) | |
1871 | return -EINVAL; | |
1872 | ||
1873 | dev->num_tc = num_tc; | |
1874 | return 0; | |
1875 | } | |
1876 | ||
1877 | static inline | |
1878 | int netdev_get_num_tc(struct net_device *dev) | |
1879 | { | |
1880 | return dev->num_tc; | |
1881 | } | |
1882 | ||
1883 | static inline | |
1884 | struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev, | |
1885 | unsigned int index) | |
1886 | { | |
1887 | return &dev->_tx[index]; | |
1888 | } | |
1889 | ||
1890 | static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev, | |
1891 | const struct sk_buff *skb) | |
1892 | { | |
1893 | return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb)); | |
1894 | } | |
1895 | ||
1896 | static inline void netdev_for_each_tx_queue(struct net_device *dev, | |
1897 | void (*f)(struct net_device *, | |
1898 | struct netdev_queue *, | |
1899 | void *), | |
1900 | void *arg) | |
1901 | { | |
1902 | unsigned int i; | |
1903 | ||
1904 | for (i = 0; i < dev->num_tx_queues; i++) | |
1905 | f(dev, &dev->_tx[i], arg); | |
1906 | } | |
1907 | ||
1908 | struct netdev_queue *netdev_pick_tx(struct net_device *dev, | |
1909 | struct sk_buff *skb, | |
1910 | void *accel_priv); | |
1911 | ||
1912 | /* | |
1913 | * Net namespace inlines | |
1914 | */ | |
1915 | static inline | |
1916 | struct net *dev_net(const struct net_device *dev) | |
1917 | { | |
1918 | return read_pnet(&dev->nd_net); | |
1919 | } | |
1920 | ||
1921 | static inline | |
1922 | void dev_net_set(struct net_device *dev, struct net *net) | |
1923 | { | |
1924 | write_pnet(&dev->nd_net, net); | |
1925 | } | |
1926 | ||
1927 | static inline bool netdev_uses_dsa(struct net_device *dev) | |
1928 | { | |
1929 | #if IS_ENABLED(CONFIG_NET_DSA) | |
1930 | if (dev->dsa_ptr != NULL) | |
1931 | return dsa_uses_tagged_protocol(dev->dsa_ptr); | |
1932 | #endif | |
1933 | return false; | |
1934 | } | |
1935 | ||
1936 | /** | |
1937 | * netdev_priv - access network device private data | |
1938 | * @dev: network device | |
1939 | * | |
1940 | * Get network device private data | |
1941 | */ | |
1942 | static inline void *netdev_priv(const struct net_device *dev) | |
1943 | { | |
1944 | return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN); | |
1945 | } | |
1946 | ||
1947 | /* Set the sysfs physical device reference for the network logical device | |
1948 | * if set prior to registration will cause a symlink during initialization. | |
1949 | */ | |
1950 | #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) | |
1951 | ||
1952 | /* Set the sysfs device type for the network logical device to allow | |
1953 | * fine-grained identification of different network device types. For | |
1954 | * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc. | |
1955 | */ | |
1956 | #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype)) | |
1957 | ||
1958 | /* Default NAPI poll() weight | |
1959 | * Device drivers are strongly advised to not use bigger value | |
1960 | */ | |
1961 | #define NAPI_POLL_WEIGHT 64 | |
1962 | ||
1963 | /** | |
1964 | * netif_napi_add - initialize a napi context | |
1965 | * @dev: network device | |
1966 | * @napi: napi context | |
1967 | * @poll: polling function | |
1968 | * @weight: default weight | |
1969 | * | |
1970 | * netif_napi_add() must be used to initialize a napi context prior to calling | |
1971 | * *any* of the other napi related functions. | |
1972 | */ | |
1973 | void netif_napi_add(struct net_device *dev, struct napi_struct *napi, | |
1974 | int (*poll)(struct napi_struct *, int), int weight); | |
1975 | ||
1976 | /** | |
1977 | * netif_tx_napi_add - initialize a napi context | |
1978 | * @dev: network device | |
1979 | * @napi: napi context | |
1980 | * @poll: polling function | |
1981 | * @weight: default weight | |
1982 | * | |
1983 | * This variant of netif_napi_add() should be used from drivers using NAPI | |
1984 | * to exclusively poll a TX queue. | |
1985 | * This will avoid we add it into napi_hash[], thus polluting this hash table. | |
1986 | */ | |
1987 | static inline void netif_tx_napi_add(struct net_device *dev, | |
1988 | struct napi_struct *napi, | |
1989 | int (*poll)(struct napi_struct *, int), | |
1990 | int weight) | |
1991 | { | |
1992 | set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state); | |
1993 | netif_napi_add(dev, napi, poll, weight); | |
1994 | } | |
1995 | ||
1996 | /** | |
1997 | * netif_napi_del - remove a napi context | |
1998 | * @napi: napi context | |
1999 | * | |
2000 | * netif_napi_del() removes a napi context from the network device napi list | |
2001 | */ | |
2002 | void netif_napi_del(struct napi_struct *napi); | |
2003 | ||
2004 | struct napi_gro_cb { | |
2005 | /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */ | |
2006 | void *frag0; | |
2007 | ||
2008 | /* Length of frag0. */ | |
2009 | unsigned int frag0_len; | |
2010 | ||
2011 | /* This indicates where we are processing relative to skb->data. */ | |
2012 | int data_offset; | |
2013 | ||
2014 | /* This is non-zero if the packet cannot be merged with the new skb. */ | |
2015 | u16 flush; | |
2016 | ||
2017 | /* Save the IP ID here and check when we get to the transport layer */ | |
2018 | u16 flush_id; | |
2019 | ||
2020 | /* Number of segments aggregated. */ | |
2021 | u16 count; | |
2022 | ||
2023 | /* Start offset for remote checksum offload */ | |
2024 | u16 gro_remcsum_start; | |
2025 | ||
2026 | /* jiffies when first packet was created/queued */ | |
2027 | unsigned long age; | |
2028 | ||
2029 | /* Used in ipv6_gro_receive() and foo-over-udp */ | |
2030 | u16 proto; | |
2031 | ||
2032 | /* This is non-zero if the packet may be of the same flow. */ | |
2033 | u8 same_flow:1; | |
2034 | ||
2035 | /* Used in udp_gro_receive */ | |
2036 | u8 udp_mark:1; | |
2037 | ||
2038 | /* GRO checksum is valid */ | |
2039 | u8 csum_valid:1; | |
2040 | ||
2041 | /* Number of checksums via CHECKSUM_UNNECESSARY */ | |
2042 | u8 csum_cnt:3; | |
2043 | ||
2044 | /* Free the skb? */ | |
2045 | u8 free:2; | |
2046 | #define NAPI_GRO_FREE 1 | |
2047 | #define NAPI_GRO_FREE_STOLEN_HEAD 2 | |
2048 | ||
2049 | /* Used in foo-over-udp, set in udp[46]_gro_receive */ | |
2050 | u8 is_ipv6:1; | |
2051 | ||
2052 | /* 7 bit hole */ | |
2053 | ||
2054 | /* used to support CHECKSUM_COMPLETE for tunneling protocols */ | |
2055 | __wsum csum; | |
2056 | ||
2057 | /* used in skb_gro_receive() slow path */ | |
2058 | struct sk_buff *last; | |
2059 | }; | |
2060 | ||
2061 | #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb) | |
2062 | ||
2063 | struct packet_type { | |
2064 | __be16 type; /* This is really htons(ether_type). */ | |
2065 | struct net_device *dev; /* NULL is wildcarded here */ | |
2066 | int (*func) (struct sk_buff *, | |
2067 | struct net_device *, | |
2068 | struct packet_type *, | |
2069 | struct net_device *); | |
2070 | bool (*id_match)(struct packet_type *ptype, | |
2071 | struct sock *sk); | |
2072 | void *af_packet_priv; | |
2073 | struct list_head list; | |
2074 | }; | |
2075 | ||
2076 | struct offload_callbacks { | |
2077 | struct sk_buff *(*gso_segment)(struct sk_buff *skb, | |
2078 | netdev_features_t features); | |
2079 | struct sk_buff **(*gro_receive)(struct sk_buff **head, | |
2080 | struct sk_buff *skb); | |
2081 | int (*gro_complete)(struct sk_buff *skb, int nhoff); | |
2082 | }; | |
2083 | ||
2084 | struct packet_offload { | |
2085 | __be16 type; /* This is really htons(ether_type). */ | |
2086 | u16 priority; | |
2087 | struct offload_callbacks callbacks; | |
2088 | struct list_head list; | |
2089 | }; | |
2090 | ||
2091 | struct udp_offload; | |
2092 | ||
2093 | struct udp_offload_callbacks { | |
2094 | struct sk_buff **(*gro_receive)(struct sk_buff **head, | |
2095 | struct sk_buff *skb, | |
2096 | struct udp_offload *uoff); | |
2097 | int (*gro_complete)(struct sk_buff *skb, | |
2098 | int nhoff, | |
2099 | struct udp_offload *uoff); | |
2100 | }; | |
2101 | ||
2102 | struct udp_offload { | |
2103 | __be16 port; | |
2104 | u8 ipproto; | |
2105 | struct udp_offload_callbacks callbacks; | |
2106 | }; | |
2107 | ||
2108 | /* often modified stats are per cpu, other are shared (netdev->stats) */ | |
2109 | struct pcpu_sw_netstats { | |
2110 | u64 rx_packets; | |
2111 | u64 rx_bytes; | |
2112 | u64 tx_packets; | |
2113 | u64 tx_bytes; | |
2114 | struct u64_stats_sync syncp; | |
2115 | }; | |
2116 | ||
2117 | #define __netdev_alloc_pcpu_stats(type, gfp) \ | |
2118 | ({ \ | |
2119 | typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\ | |
2120 | if (pcpu_stats) { \ | |
2121 | int __cpu; \ | |
2122 | for_each_possible_cpu(__cpu) { \ | |
2123 | typeof(type) *stat; \ | |
2124 | stat = per_cpu_ptr(pcpu_stats, __cpu); \ | |
2125 | u64_stats_init(&stat->syncp); \ | |
2126 | } \ | |
2127 | } \ | |
2128 | pcpu_stats; \ | |
2129 | }) | |
2130 | ||
2131 | #define netdev_alloc_pcpu_stats(type) \ | |
2132 | __netdev_alloc_pcpu_stats(type, GFP_KERNEL) | |
2133 | ||
2134 | enum netdev_lag_tx_type { | |
2135 | NETDEV_LAG_TX_TYPE_UNKNOWN, | |
2136 | NETDEV_LAG_TX_TYPE_RANDOM, | |
2137 | NETDEV_LAG_TX_TYPE_BROADCAST, | |
2138 | NETDEV_LAG_TX_TYPE_ROUNDROBIN, | |
2139 | NETDEV_LAG_TX_TYPE_ACTIVEBACKUP, | |
2140 | NETDEV_LAG_TX_TYPE_HASH, | |
2141 | }; | |
2142 | ||
2143 | struct netdev_lag_upper_info { | |
2144 | enum netdev_lag_tx_type tx_type; | |
2145 | }; | |
2146 | ||
2147 | struct netdev_lag_lower_state_info { | |
2148 | u8 link_up : 1, | |
2149 | tx_enabled : 1; | |
2150 | }; | |
2151 | ||
2152 | #include <linux/notifier.h> | |
2153 | ||
2154 | /* netdevice notifier chain. Please remember to update the rtnetlink | |
2155 | * notification exclusion list in rtnetlink_event() when adding new | |
2156 | * types. | |
2157 | */ | |
2158 | #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */ | |
2159 | #define NETDEV_DOWN 0x0002 | |
2160 | #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface | |
2161 | detected a hardware crash and restarted | |
2162 | - we can use this eg to kick tcp sessions | |
2163 | once done */ | |
2164 | #define NETDEV_CHANGE 0x0004 /* Notify device state change */ | |
2165 | #define NETDEV_REGISTER 0x0005 | |
2166 | #define NETDEV_UNREGISTER 0x0006 | |
2167 | #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */ | |
2168 | #define NETDEV_CHANGEADDR 0x0008 | |
2169 | #define NETDEV_GOING_DOWN 0x0009 | |
2170 | #define NETDEV_CHANGENAME 0x000A | |
2171 | #define NETDEV_FEAT_CHANGE 0x000B | |
2172 | #define NETDEV_BONDING_FAILOVER 0x000C | |
2173 | #define NETDEV_PRE_UP 0x000D | |
2174 | #define NETDEV_PRE_TYPE_CHANGE 0x000E | |
2175 | #define NETDEV_POST_TYPE_CHANGE 0x000F | |
2176 | #define NETDEV_POST_INIT 0x0010 | |
2177 | #define NETDEV_UNREGISTER_FINAL 0x0011 | |
2178 | #define NETDEV_RELEASE 0x0012 | |
2179 | #define NETDEV_NOTIFY_PEERS 0x0013 | |
2180 | #define NETDEV_JOIN 0x0014 | |
2181 | #define NETDEV_CHANGEUPPER 0x0015 | |
2182 | #define NETDEV_RESEND_IGMP 0x0016 | |
2183 | #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */ | |
2184 | #define NETDEV_CHANGEINFODATA 0x0018 | |
2185 | #define NETDEV_BONDING_INFO 0x0019 | |
2186 | #define NETDEV_PRECHANGEUPPER 0x001A | |
2187 | #define NETDEV_CHANGELOWERSTATE 0x001B | |
2188 | ||
2189 | int register_netdevice_notifier(struct notifier_block *nb); | |
2190 | int unregister_netdevice_notifier(struct notifier_block *nb); | |
2191 | ||
2192 | struct netdev_notifier_info { | |
2193 | struct net_device *dev; | |
2194 | }; | |
2195 | ||
2196 | struct netdev_notifier_change_info { | |
2197 | struct netdev_notifier_info info; /* must be first */ | |
2198 | unsigned int flags_changed; | |
2199 | }; | |
2200 | ||
2201 | struct netdev_notifier_changeupper_info { | |
2202 | struct netdev_notifier_info info; /* must be first */ | |
2203 | struct net_device *upper_dev; /* new upper dev */ | |
2204 | bool master; /* is upper dev master */ | |
2205 | bool linking; /* is the nofication for link or unlink */ | |
2206 | void *upper_info; /* upper dev info */ | |
2207 | }; | |
2208 | ||
2209 | struct netdev_notifier_changelowerstate_info { | |
2210 | struct netdev_notifier_info info; /* must be first */ | |
2211 | void *lower_state_info; /* is lower dev state */ | |
2212 | }; | |
2213 | ||
2214 | static inline void netdev_notifier_info_init(struct netdev_notifier_info *info, | |
2215 | struct net_device *dev) | |
2216 | { | |
2217 | info->dev = dev; | |
2218 | } | |
2219 | ||
2220 | static inline struct net_device * | |
2221 | netdev_notifier_info_to_dev(const struct netdev_notifier_info *info) | |
2222 | { | |
2223 | return info->dev; | |
2224 | } | |
2225 | ||
2226 | int call_netdevice_notifiers(unsigned long val, struct net_device *dev); | |
2227 | ||
2228 | ||
2229 | extern rwlock_t dev_base_lock; /* Device list lock */ | |
2230 | ||
2231 | #define for_each_netdev(net, d) \ | |
2232 | list_for_each_entry(d, &(net)->dev_base_head, dev_list) | |
2233 | #define for_each_netdev_reverse(net, d) \ | |
2234 | list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list) | |
2235 | #define for_each_netdev_rcu(net, d) \ | |
2236 | list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list) | |
2237 | #define for_each_netdev_safe(net, d, n) \ | |
2238 | list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list) | |
2239 | #define for_each_netdev_continue(net, d) \ | |
2240 | list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list) | |
2241 | #define for_each_netdev_continue_rcu(net, d) \ | |
2242 | list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list) | |
2243 | #define for_each_netdev_in_bond_rcu(bond, slave) \ | |
2244 | for_each_netdev_rcu(&init_net, slave) \ | |
2245 | if (netdev_master_upper_dev_get_rcu(slave) == (bond)) | |
2246 | #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list) | |
2247 | ||
2248 | static inline struct net_device *next_net_device(struct net_device *dev) | |
2249 | { | |
2250 | struct list_head *lh; | |
2251 | struct net *net; | |
2252 | ||
2253 | net = dev_net(dev); | |
2254 | lh = dev->dev_list.next; | |
2255 | return lh == &net->dev_base_head ? NULL : net_device_entry(lh); | |
2256 | } | |
2257 | ||
2258 | static inline struct net_device *next_net_device_rcu(struct net_device *dev) | |
2259 | { | |
2260 | struct list_head *lh; | |
2261 | struct net *net; | |
2262 | ||
2263 | net = dev_net(dev); | |
2264 | lh = rcu_dereference(list_next_rcu(&dev->dev_list)); | |
2265 | return lh == &net->dev_base_head ? NULL : net_device_entry(lh); | |
2266 | } | |
2267 | ||
2268 | static inline struct net_device *first_net_device(struct net *net) | |
2269 | { | |
2270 | return list_empty(&net->dev_base_head) ? NULL : | |
2271 | net_device_entry(net->dev_base_head.next); | |
2272 | } | |
2273 | ||
2274 | static inline struct net_device *first_net_device_rcu(struct net *net) | |
2275 | { | |
2276 | struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head)); | |
2277 | ||
2278 | return lh == &net->dev_base_head ? NULL : net_device_entry(lh); | |
2279 | } | |
2280 | ||
2281 | int netdev_boot_setup_check(struct net_device *dev); | |
2282 | unsigned long netdev_boot_base(const char *prefix, int unit); | |
2283 | struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type, | |
2284 | const char *hwaddr); | |
2285 | struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type); | |
2286 | struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type); | |
2287 | void dev_add_pack(struct packet_type *pt); | |
2288 | void dev_remove_pack(struct packet_type *pt); | |
2289 | void __dev_remove_pack(struct packet_type *pt); | |
2290 | void dev_add_offload(struct packet_offload *po); | |
2291 | void dev_remove_offload(struct packet_offload *po); | |
2292 | ||
2293 | int dev_get_iflink(const struct net_device *dev); | |
2294 | int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb); | |
2295 | struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags, | |
2296 | unsigned short mask); | |
2297 | struct net_device *dev_get_by_name(struct net *net, const char *name); | |
2298 | struct net_device *dev_get_by_name_rcu(struct net *net, const char *name); | |
2299 | struct net_device *__dev_get_by_name(struct net *net, const char *name); | |
2300 | int dev_alloc_name(struct net_device *dev, const char *name); | |
2301 | int dev_open(struct net_device *dev); | |
2302 | int dev_close(struct net_device *dev); | |
2303 | int dev_close_many(struct list_head *head, bool unlink); | |
2304 | void dev_disable_lro(struct net_device *dev); | |
2305 | int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb); | |
2306 | int dev_queue_xmit(struct sk_buff *skb); | |
2307 | int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv); | |
2308 | int register_netdevice(struct net_device *dev); | |
2309 | void unregister_netdevice_queue(struct net_device *dev, struct list_head *head); | |
2310 | void unregister_netdevice_many(struct list_head *head); | |
2311 | static inline void unregister_netdevice(struct net_device *dev) | |
2312 | { | |
2313 | unregister_netdevice_queue(dev, NULL); | |
2314 | } | |
2315 | ||
2316 | int netdev_refcnt_read(const struct net_device *dev); | |
2317 | void free_netdev(struct net_device *dev); | |
2318 | void netdev_freemem(struct net_device *dev); | |
2319 | void synchronize_net(void); | |
2320 | int init_dummy_netdev(struct net_device *dev); | |
2321 | ||
2322 | DECLARE_PER_CPU(int, xmit_recursion); | |
2323 | static inline int dev_recursion_level(void) | |
2324 | { | |
2325 | return this_cpu_read(xmit_recursion); | |
2326 | } | |
2327 | ||
2328 | struct net_device *dev_get_by_index(struct net *net, int ifindex); | |
2329 | struct net_device *__dev_get_by_index(struct net *net, int ifindex); | |
2330 | struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex); | |
2331 | int netdev_get_name(struct net *net, char *name, int ifindex); | |
2332 | int dev_restart(struct net_device *dev); | |
2333 | int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb); | |
2334 | ||
2335 | static inline unsigned int skb_gro_offset(const struct sk_buff *skb) | |
2336 | { | |
2337 | return NAPI_GRO_CB(skb)->data_offset; | |
2338 | } | |
2339 | ||
2340 | static inline unsigned int skb_gro_len(const struct sk_buff *skb) | |
2341 | { | |
2342 | return skb->len - NAPI_GRO_CB(skb)->data_offset; | |
2343 | } | |
2344 | ||
2345 | static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len) | |
2346 | { | |
2347 | NAPI_GRO_CB(skb)->data_offset += len; | |
2348 | } | |
2349 | ||
2350 | static inline void *skb_gro_header_fast(struct sk_buff *skb, | |
2351 | unsigned int offset) | |
2352 | { | |
2353 | return NAPI_GRO_CB(skb)->frag0 + offset; | |
2354 | } | |
2355 | ||
2356 | static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen) | |
2357 | { | |
2358 | return NAPI_GRO_CB(skb)->frag0_len < hlen; | |
2359 | } | |
2360 | ||
2361 | static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen, | |
2362 | unsigned int offset) | |
2363 | { | |
2364 | if (!pskb_may_pull(skb, hlen)) | |
2365 | return NULL; | |
2366 | ||
2367 | NAPI_GRO_CB(skb)->frag0 = NULL; | |
2368 | NAPI_GRO_CB(skb)->frag0_len = 0; | |
2369 | return skb->data + offset; | |
2370 | } | |
2371 | ||
2372 | static inline void *skb_gro_network_header(struct sk_buff *skb) | |
2373 | { | |
2374 | return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) + | |
2375 | skb_network_offset(skb); | |
2376 | } | |
2377 | ||
2378 | static inline void skb_gro_postpull_rcsum(struct sk_buff *skb, | |
2379 | const void *start, unsigned int len) | |
2380 | { | |
2381 | if (NAPI_GRO_CB(skb)->csum_valid) | |
2382 | NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum, | |
2383 | csum_partial(start, len, 0)); | |
2384 | } | |
2385 | ||
2386 | /* GRO checksum functions. These are logical equivalents of the normal | |
2387 | * checksum functions (in skbuff.h) except that they operate on the GRO | |
2388 | * offsets and fields in sk_buff. | |
2389 | */ | |
2390 | ||
2391 | __sum16 __skb_gro_checksum_complete(struct sk_buff *skb); | |
2392 | ||
2393 | static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb) | |
2394 | { | |
2395 | return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb)); | |
2396 | } | |
2397 | ||
2398 | static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb, | |
2399 | bool zero_okay, | |
2400 | __sum16 check) | |
2401 | { | |
2402 | return ((skb->ip_summed != CHECKSUM_PARTIAL || | |
2403 | skb_checksum_start_offset(skb) < | |
2404 | skb_gro_offset(skb)) && | |
2405 | !skb_at_gro_remcsum_start(skb) && | |
2406 | NAPI_GRO_CB(skb)->csum_cnt == 0 && | |
2407 | (!zero_okay || check)); | |
2408 | } | |
2409 | ||
2410 | static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb, | |
2411 | __wsum psum) | |
2412 | { | |
2413 | if (NAPI_GRO_CB(skb)->csum_valid && | |
2414 | !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum))) | |
2415 | return 0; | |
2416 | ||
2417 | NAPI_GRO_CB(skb)->csum = psum; | |
2418 | ||
2419 | return __skb_gro_checksum_complete(skb); | |
2420 | } | |
2421 | ||
2422 | static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb) | |
2423 | { | |
2424 | if (NAPI_GRO_CB(skb)->csum_cnt > 0) { | |
2425 | /* Consume a checksum from CHECKSUM_UNNECESSARY */ | |
2426 | NAPI_GRO_CB(skb)->csum_cnt--; | |
2427 | } else { | |
2428 | /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we | |
2429 | * verified a new top level checksum or an encapsulated one | |
2430 | * during GRO. This saves work if we fallback to normal path. | |
2431 | */ | |
2432 | __skb_incr_checksum_unnecessary(skb); | |
2433 | } | |
2434 | } | |
2435 | ||
2436 | #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \ | |
2437 | compute_pseudo) \ | |
2438 | ({ \ | |
2439 | __sum16 __ret = 0; \ | |
2440 | if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \ | |
2441 | __ret = __skb_gro_checksum_validate_complete(skb, \ | |
2442 | compute_pseudo(skb, proto)); \ | |
2443 | if (__ret) \ | |
2444 | __skb_mark_checksum_bad(skb); \ | |
2445 | else \ | |
2446 | skb_gro_incr_csum_unnecessary(skb); \ | |
2447 | __ret; \ | |
2448 | }) | |
2449 | ||
2450 | #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \ | |
2451 | __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo) | |
2452 | ||
2453 | #define skb_gro_checksum_validate_zero_check(skb, proto, check, \ | |
2454 | compute_pseudo) \ | |
2455 | __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo) | |
2456 | ||
2457 | #define skb_gro_checksum_simple_validate(skb) \ | |
2458 | __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo) | |
2459 | ||
2460 | static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb) | |
2461 | { | |
2462 | return (NAPI_GRO_CB(skb)->csum_cnt == 0 && | |
2463 | !NAPI_GRO_CB(skb)->csum_valid); | |
2464 | } | |
2465 | ||
2466 | static inline void __skb_gro_checksum_convert(struct sk_buff *skb, | |
2467 | __sum16 check, __wsum pseudo) | |
2468 | { | |
2469 | NAPI_GRO_CB(skb)->csum = ~pseudo; | |
2470 | NAPI_GRO_CB(skb)->csum_valid = 1; | |
2471 | } | |
2472 | ||
2473 | #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \ | |
2474 | do { \ | |
2475 | if (__skb_gro_checksum_convert_check(skb)) \ | |
2476 | __skb_gro_checksum_convert(skb, check, \ | |
2477 | compute_pseudo(skb, proto)); \ | |
2478 | } while (0) | |
2479 | ||
2480 | struct gro_remcsum { | |
2481 | int offset; | |
2482 | __wsum delta; | |
2483 | }; | |
2484 | ||
2485 | static inline void skb_gro_remcsum_init(struct gro_remcsum *grc) | |
2486 | { | |
2487 | grc->offset = 0; | |
2488 | grc->delta = 0; | |
2489 | } | |
2490 | ||
2491 | static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr, | |
2492 | unsigned int off, size_t hdrlen, | |
2493 | int start, int offset, | |
2494 | struct gro_remcsum *grc, | |
2495 | bool nopartial) | |
2496 | { | |
2497 | __wsum delta; | |
2498 | size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start); | |
2499 | ||
2500 | BUG_ON(!NAPI_GRO_CB(skb)->csum_valid); | |
2501 | ||
2502 | if (!nopartial) { | |
2503 | NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start; | |
2504 | return ptr; | |
2505 | } | |
2506 | ||
2507 | ptr = skb_gro_header_fast(skb, off); | |
2508 | if (skb_gro_header_hard(skb, off + plen)) { | |
2509 | ptr = skb_gro_header_slow(skb, off + plen, off); | |
2510 | if (!ptr) | |
2511 | return NULL; | |
2512 | } | |
2513 | ||
2514 | delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum, | |
2515 | start, offset); | |
2516 | ||
2517 | /* Adjust skb->csum since we changed the packet */ | |
2518 | NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta); | |
2519 | ||
2520 | grc->offset = off + hdrlen + offset; | |
2521 | grc->delta = delta; | |
2522 | ||
2523 | return ptr; | |
2524 | } | |
2525 | ||
2526 | static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb, | |
2527 | struct gro_remcsum *grc) | |
2528 | { | |
2529 | void *ptr; | |
2530 | size_t plen = grc->offset + sizeof(u16); | |
2531 | ||
2532 | if (!grc->delta) | |
2533 | return; | |
2534 | ||
2535 | ptr = skb_gro_header_fast(skb, grc->offset); | |
2536 | if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) { | |
2537 | ptr = skb_gro_header_slow(skb, plen, grc->offset); | |
2538 | if (!ptr) | |
2539 | return; | |
2540 | } | |
2541 | ||
2542 | remcsum_unadjust((__sum16 *)ptr, grc->delta); | |
2543 | } | |
2544 | ||
2545 | struct skb_csum_offl_spec { | |
2546 | __u16 ipv4_okay:1, | |
2547 | ipv6_okay:1, | |
2548 | encap_okay:1, | |
2549 | ip_options_okay:1, | |
2550 | ext_hdrs_okay:1, | |
2551 | tcp_okay:1, | |
2552 | udp_okay:1, | |
2553 | sctp_okay:1, | |
2554 | vlan_okay:1, | |
2555 | no_encapped_ipv6:1, | |
2556 | no_not_encapped:1; | |
2557 | }; | |
2558 | ||
2559 | bool __skb_csum_offload_chk(struct sk_buff *skb, | |
2560 | const struct skb_csum_offl_spec *spec, | |
2561 | bool *csum_encapped, | |
2562 | bool csum_help); | |
2563 | ||
2564 | static inline bool skb_csum_offload_chk(struct sk_buff *skb, | |
2565 | const struct skb_csum_offl_spec *spec, | |
2566 | bool *csum_encapped, | |
2567 | bool csum_help) | |
2568 | { | |
2569 | if (skb->ip_summed != CHECKSUM_PARTIAL) | |
2570 | return false; | |
2571 | ||
2572 | return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help); | |
2573 | } | |
2574 | ||
2575 | static inline bool skb_csum_offload_chk_help(struct sk_buff *skb, | |
2576 | const struct skb_csum_offl_spec *spec) | |
2577 | { | |
2578 | bool csum_encapped; | |
2579 | ||
2580 | return skb_csum_offload_chk(skb, spec, &csum_encapped, true); | |
2581 | } | |
2582 | ||
2583 | static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb) | |
2584 | { | |
2585 | static const struct skb_csum_offl_spec csum_offl_spec = { | |
2586 | .ipv4_okay = 1, | |
2587 | .ip_options_okay = 1, | |
2588 | .ipv6_okay = 1, | |
2589 | .vlan_okay = 1, | |
2590 | .tcp_okay = 1, | |
2591 | .udp_okay = 1, | |
2592 | }; | |
2593 | ||
2594 | return skb_csum_offload_chk_help(skb, &csum_offl_spec); | |
2595 | } | |
2596 | ||
2597 | static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb) | |
2598 | { | |
2599 | static const struct skb_csum_offl_spec csum_offl_spec = { | |
2600 | .ipv4_okay = 1, | |
2601 | .ip_options_okay = 1, | |
2602 | .tcp_okay = 1, | |
2603 | .udp_okay = 1, | |
2604 | .vlan_okay = 1, | |
2605 | }; | |
2606 | ||
2607 | return skb_csum_offload_chk_help(skb, &csum_offl_spec); | |
2608 | } | |
2609 | ||
2610 | static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev, | |
2611 | unsigned short type, | |
2612 | const void *daddr, const void *saddr, | |
2613 | unsigned int len) | |
2614 | { | |
2615 | if (!dev->header_ops || !dev->header_ops->create) | |
2616 | return 0; | |
2617 | ||
2618 | return dev->header_ops->create(skb, dev, type, daddr, saddr, len); | |
2619 | } | |
2620 | ||
2621 | static inline int dev_parse_header(const struct sk_buff *skb, | |
2622 | unsigned char *haddr) | |
2623 | { | |
2624 | const struct net_device *dev = skb->dev; | |
2625 | ||
2626 | if (!dev->header_ops || !dev->header_ops->parse) | |
2627 | return 0; | |
2628 | return dev->header_ops->parse(skb, haddr); | |
2629 | } | |
2630 | ||
2631 | typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len); | |
2632 | int register_gifconf(unsigned int family, gifconf_func_t *gifconf); | |
2633 | static inline int unregister_gifconf(unsigned int family) | |
2634 | { | |
2635 | return register_gifconf(family, NULL); | |
2636 | } | |
2637 | ||
2638 | #ifdef CONFIG_NET_FLOW_LIMIT | |
2639 | #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ | |
2640 | struct sd_flow_limit { | |
2641 | u64 count; | |
2642 | unsigned int num_buckets; | |
2643 | unsigned int history_head; | |
2644 | u16 history[FLOW_LIMIT_HISTORY]; | |
2645 | u8 buckets[]; | |
2646 | }; | |
2647 | ||
2648 | extern int netdev_flow_limit_table_len; | |
2649 | #endif /* CONFIG_NET_FLOW_LIMIT */ | |
2650 | ||
2651 | /* | |
2652 | * Incoming packets are placed on per-cpu queues | |
2653 | */ | |
2654 | struct softnet_data { | |
2655 | struct list_head poll_list; | |
2656 | struct sk_buff_head process_queue; | |
2657 | ||
2658 | /* stats */ | |
2659 | unsigned int processed; | |
2660 | unsigned int time_squeeze; | |
2661 | unsigned int cpu_collision; | |
2662 | unsigned int received_rps; | |
2663 | #ifdef CONFIG_RPS | |
2664 | struct softnet_data *rps_ipi_list; | |
2665 | #endif | |
2666 | #ifdef CONFIG_NET_FLOW_LIMIT | |
2667 | struct sd_flow_limit __rcu *flow_limit; | |
2668 | #endif | |
2669 | struct Qdisc *output_queue; | |
2670 | struct Qdisc **output_queue_tailp; | |
2671 | struct sk_buff *completion_queue; | |
2672 | ||
2673 | #ifdef CONFIG_RPS | |
2674 | /* Elements below can be accessed between CPUs for RPS */ | |
2675 | struct call_single_data csd ____cacheline_aligned_in_smp; | |
2676 | struct softnet_data *rps_ipi_next; | |
2677 | unsigned int cpu; | |
2678 | unsigned int input_queue_head; | |
2679 | unsigned int input_queue_tail; | |
2680 | #endif | |
2681 | unsigned int dropped; | |
2682 | struct sk_buff_head input_pkt_queue; | |
2683 | struct napi_struct backlog; | |
2684 | ||
2685 | }; | |
2686 | ||
2687 | static inline void input_queue_head_incr(struct softnet_data *sd) | |
2688 | { | |
2689 | #ifdef CONFIG_RPS | |
2690 | sd->input_queue_head++; | |
2691 | #endif | |
2692 | } | |
2693 | ||
2694 | static inline void input_queue_tail_incr_save(struct softnet_data *sd, | |
2695 | unsigned int *qtail) | |
2696 | { | |
2697 | #ifdef CONFIG_RPS | |
2698 | *qtail = ++sd->input_queue_tail; | |
2699 | #endif | |
2700 | } | |
2701 | ||
2702 | DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data); | |
2703 | ||
2704 | void __netif_schedule(struct Qdisc *q); | |
2705 | void netif_schedule_queue(struct netdev_queue *txq); | |
2706 | ||
2707 | static inline void netif_tx_schedule_all(struct net_device *dev) | |
2708 | { | |
2709 | unsigned int i; | |
2710 | ||
2711 | for (i = 0; i < dev->num_tx_queues; i++) | |
2712 | netif_schedule_queue(netdev_get_tx_queue(dev, i)); | |
2713 | } | |
2714 | ||
2715 | static inline void netif_tx_start_queue(struct netdev_queue *dev_queue) | |
2716 | { | |
2717 | clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); | |
2718 | } | |
2719 | ||
2720 | /** | |
2721 | * netif_start_queue - allow transmit | |
2722 | * @dev: network device | |
2723 | * | |
2724 | * Allow upper layers to call the device hard_start_xmit routine. | |
2725 | */ | |
2726 | static inline void netif_start_queue(struct net_device *dev) | |
2727 | { | |
2728 | netif_tx_start_queue(netdev_get_tx_queue(dev, 0)); | |
2729 | } | |
2730 | ||
2731 | static inline void netif_tx_start_all_queues(struct net_device *dev) | |
2732 | { | |
2733 | unsigned int i; | |
2734 | ||
2735 | for (i = 0; i < dev->num_tx_queues; i++) { | |
2736 | struct netdev_queue *txq = netdev_get_tx_queue(dev, i); | |
2737 | netif_tx_start_queue(txq); | |
2738 | } | |
2739 | } | |
2740 | ||
2741 | void netif_tx_wake_queue(struct netdev_queue *dev_queue); | |
2742 | ||
2743 | /** | |
2744 | * netif_wake_queue - restart transmit | |
2745 | * @dev: network device | |
2746 | * | |
2747 | * Allow upper layers to call the device hard_start_xmit routine. | |
2748 | * Used for flow control when transmit resources are available. | |
2749 | */ | |
2750 | static inline void netif_wake_queue(struct net_device *dev) | |
2751 | { | |
2752 | netif_tx_wake_queue(netdev_get_tx_queue(dev, 0)); | |
2753 | } | |
2754 | ||
2755 | static inline void netif_tx_wake_all_queues(struct net_device *dev) | |
2756 | { | |
2757 | unsigned int i; | |
2758 | ||
2759 | for (i = 0; i < dev->num_tx_queues; i++) { | |
2760 | struct netdev_queue *txq = netdev_get_tx_queue(dev, i); | |
2761 | netif_tx_wake_queue(txq); | |
2762 | } | |
2763 | } | |
2764 | ||
2765 | static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue) | |
2766 | { | |
2767 | set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); | |
2768 | } | |
2769 | ||
2770 | /** | |
2771 | * netif_stop_queue - stop transmitted packets | |
2772 | * @dev: network device | |
2773 | * | |
2774 | * Stop upper layers calling the device hard_start_xmit routine. | |
2775 | * Used for flow control when transmit resources are unavailable. | |
2776 | */ | |
2777 | static inline void netif_stop_queue(struct net_device *dev) | |
2778 | { | |
2779 | netif_tx_stop_queue(netdev_get_tx_queue(dev, 0)); | |
2780 | } | |
2781 | ||
2782 | void netif_tx_stop_all_queues(struct net_device *dev); | |
2783 | ||
2784 | static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue) | |
2785 | { | |
2786 | return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state); | |
2787 | } | |
2788 | ||
2789 | /** | |
2790 | * netif_queue_stopped - test if transmit queue is flowblocked | |
2791 | * @dev: network device | |
2792 | * | |
2793 | * Test if transmit queue on device is currently unable to send. | |
2794 | */ | |
2795 | static inline bool netif_queue_stopped(const struct net_device *dev) | |
2796 | { | |
2797 | return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0)); | |
2798 | } | |
2799 | ||
2800 | static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue) | |
2801 | { | |
2802 | return dev_queue->state & QUEUE_STATE_ANY_XOFF; | |
2803 | } | |
2804 | ||
2805 | static inline bool | |
2806 | netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue) | |
2807 | { | |
2808 | return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN; | |
2809 | } | |
2810 | ||
2811 | static inline bool | |
2812 | netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue) | |
2813 | { | |
2814 | return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN; | |
2815 | } | |
2816 | ||
2817 | /** | |
2818 | * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write | |
2819 | * @dev_queue: pointer to transmit queue | |
2820 | * | |
2821 | * BQL enabled drivers might use this helper in their ndo_start_xmit(), | |
2822 | * to give appropriate hint to the cpu. | |
2823 | */ | |
2824 | static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue) | |
2825 | { | |
2826 | #ifdef CONFIG_BQL | |
2827 | prefetchw(&dev_queue->dql.num_queued); | |
2828 | #endif | |
2829 | } | |
2830 | ||
2831 | /** | |
2832 | * netdev_txq_bql_complete_prefetchw - prefetch bql data for write | |
2833 | * @dev_queue: pointer to transmit queue | |
2834 | * | |
2835 | * BQL enabled drivers might use this helper in their TX completion path, | |
2836 | * to give appropriate hint to the cpu. | |
2837 | */ | |
2838 | static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue) | |
2839 | { | |
2840 | #ifdef CONFIG_BQL | |
2841 | prefetchw(&dev_queue->dql.limit); | |
2842 | #endif | |
2843 | } | |
2844 | ||
2845 | static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue, | |
2846 | unsigned int bytes) | |
2847 | { | |
2848 | #ifdef CONFIG_BQL | |
2849 | dql_queued(&dev_queue->dql, bytes); | |
2850 | ||
2851 | if (likely(dql_avail(&dev_queue->dql) >= 0)) | |
2852 | return; | |
2853 | ||
2854 | set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); | |
2855 | ||
2856 | /* | |
2857 | * The XOFF flag must be set before checking the dql_avail below, | |
2858 | * because in netdev_tx_completed_queue we update the dql_completed | |
2859 | * before checking the XOFF flag. | |
2860 | */ | |
2861 | smp_mb(); | |
2862 | ||
2863 | /* check again in case another CPU has just made room avail */ | |
2864 | if (unlikely(dql_avail(&dev_queue->dql) >= 0)) | |
2865 | clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state); | |
2866 | #endif | |
2867 | } | |
2868 | ||
2869 | /** | |
2870 | * netdev_sent_queue - report the number of bytes queued to hardware | |
2871 | * @dev: network device | |
2872 | * @bytes: number of bytes queued to the hardware device queue | |
2873 | * | |
2874 | * Report the number of bytes queued for sending/completion to the network | |
2875 | * device hardware queue. @bytes should be a good approximation and should | |
2876 | * exactly match netdev_completed_queue() @bytes | |
2877 | */ | |
2878 | static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes) | |
2879 | { | |
2880 | netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes); | |
2881 | } | |
2882 | ||
2883 | static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue, | |
2884 | unsigned int pkts, unsigned int bytes) | |
2885 | { | |
2886 | #ifdef CONFIG_BQL | |
2887 | if (unlikely(!bytes)) | |
2888 | return; | |
2889 | ||
2890 | dql_completed(&dev_queue->dql, bytes); | |
2891 | ||
2892 | /* | |
2893 | * Without the memory barrier there is a small possiblity that | |
2894 | * netdev_tx_sent_queue will miss the update and cause the queue to | |
2895 | * be stopped forever | |
2896 | */ | |
2897 | smp_mb(); | |
2898 | ||
2899 | if (dql_avail(&dev_queue->dql) < 0) | |
2900 | return; | |
2901 | ||
2902 | if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state)) | |
2903 | netif_schedule_queue(dev_queue); | |
2904 | #endif | |
2905 | } | |
2906 | ||
2907 | /** | |
2908 | * netdev_completed_queue - report bytes and packets completed by device | |
2909 | * @dev: network device | |
2910 | * @pkts: actual number of packets sent over the medium | |
2911 | * @bytes: actual number of bytes sent over the medium | |
2912 | * | |
2913 | * Report the number of bytes and packets transmitted by the network device | |
2914 | * hardware queue over the physical medium, @bytes must exactly match the | |
2915 | * @bytes amount passed to netdev_sent_queue() | |
2916 | */ | |
2917 | static inline void netdev_completed_queue(struct net_device *dev, | |
2918 | unsigned int pkts, unsigned int bytes) | |
2919 | { | |
2920 | netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes); | |
2921 | } | |
2922 | ||
2923 | static inline void netdev_tx_reset_queue(struct netdev_queue *q) | |
2924 | { | |
2925 | #ifdef CONFIG_BQL | |
2926 | clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state); | |
2927 | dql_reset(&q->dql); | |
2928 | #endif | |
2929 | } | |
2930 | ||
2931 | /** | |
2932 | * netdev_reset_queue - reset the packets and bytes count of a network device | |
2933 | * @dev_queue: network device | |
2934 | * | |
2935 | * Reset the bytes and packet count of a network device and clear the | |
2936 | * software flow control OFF bit for this network device | |
2937 | */ | |
2938 | static inline void netdev_reset_queue(struct net_device *dev_queue) | |
2939 | { | |
2940 | netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0)); | |
2941 | } | |
2942 | ||
2943 | /** | |
2944 | * netdev_cap_txqueue - check if selected tx queue exceeds device queues | |
2945 | * @dev: network device | |
2946 | * @queue_index: given tx queue index | |
2947 | * | |
2948 | * Returns 0 if given tx queue index >= number of device tx queues, | |
2949 | * otherwise returns the originally passed tx queue index. | |
2950 | */ | |
2951 | static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index) | |
2952 | { | |
2953 | if (unlikely(queue_index >= dev->real_num_tx_queues)) { | |
2954 | net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n", | |
2955 | dev->name, queue_index, | |
2956 | dev->real_num_tx_queues); | |
2957 | return 0; | |
2958 | } | |
2959 | ||
2960 | return queue_index; | |
2961 | } | |
2962 | ||
2963 | /** | |
2964 | * netif_running - test if up | |
2965 | * @dev: network device | |
2966 | * | |
2967 | * Test if the device has been brought up. | |
2968 | */ | |
2969 | static inline bool netif_running(const struct net_device *dev) | |
2970 | { | |
2971 | return test_bit(__LINK_STATE_START, &dev->state); | |
2972 | } | |
2973 | ||
2974 | /* | |
2975 | * Routines to manage the subqueues on a device. We only need start | |
2976 | * stop, and a check if it's stopped. All other device management is | |
2977 | * done at the overall netdevice level. | |
2978 | * Also test the device if we're multiqueue. | |
2979 | */ | |
2980 | ||
2981 | /** | |
2982 | * netif_start_subqueue - allow sending packets on subqueue | |
2983 | * @dev: network device | |
2984 | * @queue_index: sub queue index | |
2985 | * | |
2986 | * Start individual transmit queue of a device with multiple transmit queues. | |
2987 | */ | |
2988 | static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) | |
2989 | { | |
2990 | struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); | |
2991 | ||
2992 | netif_tx_start_queue(txq); | |
2993 | } | |
2994 | ||
2995 | /** | |
2996 | * netif_stop_subqueue - stop sending packets on subqueue | |
2997 | * @dev: network device | |
2998 | * @queue_index: sub queue index | |
2999 | * | |
3000 | * Stop individual transmit queue of a device with multiple transmit queues. | |
3001 | */ | |
3002 | static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) | |
3003 | { | |
3004 | struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); | |
3005 | netif_tx_stop_queue(txq); | |
3006 | } | |
3007 | ||
3008 | /** | |
3009 | * netif_subqueue_stopped - test status of subqueue | |
3010 | * @dev: network device | |
3011 | * @queue_index: sub queue index | |
3012 | * | |
3013 | * Check individual transmit queue of a device with multiple transmit queues. | |
3014 | */ | |
3015 | static inline bool __netif_subqueue_stopped(const struct net_device *dev, | |
3016 | u16 queue_index) | |
3017 | { | |
3018 | struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index); | |
3019 | ||
3020 | return netif_tx_queue_stopped(txq); | |
3021 | } | |
3022 | ||
3023 | static inline bool netif_subqueue_stopped(const struct net_device *dev, | |
3024 | struct sk_buff *skb) | |
3025 | { | |
3026 | return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb)); | |
3027 | } | |
3028 | ||
3029 | void netif_wake_subqueue(struct net_device *dev, u16 queue_index); | |
3030 | ||
3031 | #ifdef CONFIG_XPS | |
3032 | int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask, | |
3033 | u16 index); | |
3034 | #else | |
3035 | static inline int netif_set_xps_queue(struct net_device *dev, | |
3036 | const struct cpumask *mask, | |
3037 | u16 index) | |
3038 | { | |
3039 | return 0; | |
3040 | } | |
3041 | #endif | |
3042 | ||
3043 | u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb, | |
3044 | unsigned int num_tx_queues); | |
3045 | ||
3046 | /* | |
3047 | * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used | |
3048 | * as a distribution range limit for the returned value. | |
3049 | */ | |
3050 | static inline u16 skb_tx_hash(const struct net_device *dev, | |
3051 | struct sk_buff *skb) | |
3052 | { | |
3053 | return __skb_tx_hash(dev, skb, dev->real_num_tx_queues); | |
3054 | } | |
3055 | ||
3056 | /** | |
3057 | * netif_is_multiqueue - test if device has multiple transmit queues | |
3058 | * @dev: network device | |
3059 | * | |
3060 | * Check if device has multiple transmit queues | |
3061 | */ | |
3062 | static inline bool netif_is_multiqueue(const struct net_device *dev) | |
3063 | { | |
3064 | return dev->num_tx_queues > 1; | |
3065 | } | |
3066 | ||
3067 | int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq); | |
3068 | ||
3069 | #ifdef CONFIG_SYSFS | |
3070 | int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq); | |
3071 | #else | |
3072 | static inline int netif_set_real_num_rx_queues(struct net_device *dev, | |
3073 | unsigned int rxq) | |
3074 | { | |
3075 | return 0; | |
3076 | } | |
3077 | #endif | |
3078 | ||
3079 | #ifdef CONFIG_SYSFS | |
3080 | static inline unsigned int get_netdev_rx_queue_index( | |
3081 | struct netdev_rx_queue *queue) | |
3082 | { | |
3083 | struct net_device *dev = queue->dev; | |
3084 | int index = queue - dev->_rx; | |
3085 | ||
3086 | BUG_ON(index >= dev->num_rx_queues); | |
3087 | return index; | |
3088 | } | |
3089 | #endif | |
3090 | ||
3091 | #define DEFAULT_MAX_NUM_RSS_QUEUES (8) | |
3092 | int netif_get_num_default_rss_queues(void); | |
3093 | ||
3094 | enum skb_free_reason { | |
3095 | SKB_REASON_CONSUMED, | |
3096 | SKB_REASON_DROPPED, | |
3097 | }; | |
3098 | ||
3099 | void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason); | |
3100 | void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason); | |
3101 | ||
3102 | /* | |
3103 | * It is not allowed to call kfree_skb() or consume_skb() from hardware | |
3104 | * interrupt context or with hardware interrupts being disabled. | |
3105 | * (in_irq() || irqs_disabled()) | |
3106 | * | |
3107 | * We provide four helpers that can be used in following contexts : | |
3108 | * | |
3109 | * dev_kfree_skb_irq(skb) when caller drops a packet from irq context, | |
3110 | * replacing kfree_skb(skb) | |
3111 | * | |
3112 | * dev_consume_skb_irq(skb) when caller consumes a packet from irq context. | |
3113 | * Typically used in place of consume_skb(skb) in TX completion path | |
3114 | * | |
3115 | * dev_kfree_skb_any(skb) when caller doesn't know its current irq context, | |
3116 | * replacing kfree_skb(skb) | |
3117 | * | |
3118 | * dev_consume_skb_any(skb) when caller doesn't know its current irq context, | |
3119 | * and consumed a packet. Used in place of consume_skb(skb) | |
3120 | */ | |
3121 | static inline void dev_kfree_skb_irq(struct sk_buff *skb) | |
3122 | { | |
3123 | __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED); | |
3124 | } | |
3125 | ||
3126 | static inline void dev_consume_skb_irq(struct sk_buff *skb) | |
3127 | { | |
3128 | __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED); | |
3129 | } | |
3130 | ||
3131 | static inline void dev_kfree_skb_any(struct sk_buff *skb) | |
3132 | { | |
3133 | __dev_kfree_skb_any(skb, SKB_REASON_DROPPED); | |
3134 | } | |
3135 | ||
3136 | static inline void dev_consume_skb_any(struct sk_buff *skb) | |
3137 | { | |
3138 | __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED); | |
3139 | } | |
3140 | ||
3141 | int netif_rx(struct sk_buff *skb); | |
3142 | int netif_rx_ni(struct sk_buff *skb); | |
3143 | int netif_receive_skb(struct sk_buff *skb); | |
3144 | gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb); | |
3145 | void napi_gro_flush(struct napi_struct *napi, bool flush_old); | |
3146 | struct sk_buff *napi_get_frags(struct napi_struct *napi); | |
3147 | gro_result_t napi_gro_frags(struct napi_struct *napi); | |
3148 | struct packet_offload *gro_find_receive_by_type(__be16 type); | |
3149 | struct packet_offload *gro_find_complete_by_type(__be16 type); | |
3150 | ||
3151 | static inline void napi_free_frags(struct napi_struct *napi) | |
3152 | { | |
3153 | kfree_skb(napi->skb); | |
3154 | napi->skb = NULL; | |
3155 | } | |
3156 | ||
3157 | int netdev_rx_handler_register(struct net_device *dev, | |
3158 | rx_handler_func_t *rx_handler, | |
3159 | void *rx_handler_data); | |
3160 | void netdev_rx_handler_unregister(struct net_device *dev); | |
3161 | ||
3162 | bool dev_valid_name(const char *name); | |
3163 | int dev_ioctl(struct net *net, unsigned int cmd, void __user *); | |
3164 | int dev_ethtool(struct net *net, struct ifreq *); | |
3165 | unsigned int dev_get_flags(const struct net_device *); | |
3166 | int __dev_change_flags(struct net_device *, unsigned int flags); | |
3167 | int dev_change_flags(struct net_device *, unsigned int); | |
3168 | void __dev_notify_flags(struct net_device *, unsigned int old_flags, | |
3169 | unsigned int gchanges); | |
3170 | int dev_change_name(struct net_device *, const char *); | |
3171 | int dev_set_alias(struct net_device *, const char *, size_t); | |
3172 | int dev_change_net_namespace(struct net_device *, struct net *, const char *); | |
3173 | int dev_set_mtu(struct net_device *, int); | |
3174 | void dev_set_group(struct net_device *, int); | |
3175 | int dev_set_mac_address(struct net_device *, struct sockaddr *); | |
3176 | int dev_change_carrier(struct net_device *, bool new_carrier); | |
3177 | int dev_get_phys_port_id(struct net_device *dev, | |
3178 | struct netdev_phys_item_id *ppid); | |
3179 | int dev_get_phys_port_name(struct net_device *dev, | |
3180 | char *name, size_t len); | |
3181 | int dev_change_proto_down(struct net_device *dev, bool proto_down); | |
3182 | struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev); | |
3183 | struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev, | |
3184 | struct netdev_queue *txq, int *ret); | |
3185 | int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb); | |
3186 | int dev_forward_skb(struct net_device *dev, struct sk_buff *skb); | |
3187 | bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb); | |
3188 | ||
3189 | extern int netdev_budget; | |
3190 | ||
3191 | /* Called by rtnetlink.c:rtnl_unlock() */ | |
3192 | void netdev_run_todo(void); | |
3193 | ||
3194 | /** | |
3195 | * dev_put - release reference to device | |
3196 | * @dev: network device | |
3197 | * | |
3198 | * Release reference to device to allow it to be freed. | |
3199 | */ | |
3200 | static inline void dev_put(struct net_device *dev) | |
3201 | { | |
3202 | this_cpu_dec(*dev->pcpu_refcnt); | |
3203 | } | |
3204 | ||
3205 | /** | |
3206 | * dev_hold - get reference to device | |
3207 | * @dev: network device | |
3208 | * | |
3209 | * Hold reference to device to keep it from being freed. | |
3210 | */ | |
3211 | static inline void dev_hold(struct net_device *dev) | |
3212 | { | |
3213 | this_cpu_inc(*dev->pcpu_refcnt); | |
3214 | } | |
3215 | ||
3216 | /* Carrier loss detection, dial on demand. The functions netif_carrier_on | |
3217 | * and _off may be called from IRQ context, but it is caller | |
3218 | * who is responsible for serialization of these calls. | |
3219 | * | |
3220 | * The name carrier is inappropriate, these functions should really be | |
3221 | * called netif_lowerlayer_*() because they represent the state of any | |
3222 | * kind of lower layer not just hardware media. | |
3223 | */ | |
3224 | ||
3225 | void linkwatch_init_dev(struct net_device *dev); | |
3226 | void linkwatch_fire_event(struct net_device *dev); | |
3227 | void linkwatch_forget_dev(struct net_device *dev); | |
3228 | ||
3229 | /** | |
3230 | * netif_carrier_ok - test if carrier present | |
3231 | * @dev: network device | |
3232 | * | |
3233 | * Check if carrier is present on device | |
3234 | */ | |
3235 | static inline bool netif_carrier_ok(const struct net_device *dev) | |
3236 | { | |
3237 | return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); | |
3238 | } | |
3239 | ||
3240 | unsigned long dev_trans_start(struct net_device *dev); | |
3241 | ||
3242 | void __netdev_watchdog_up(struct net_device *dev); | |
3243 | ||
3244 | void netif_carrier_on(struct net_device *dev); | |
3245 | ||
3246 | void netif_carrier_off(struct net_device *dev); | |
3247 | ||
3248 | /** | |
3249 | * netif_dormant_on - mark device as dormant. | |
3250 | * @dev: network device | |
3251 | * | |
3252 | * Mark device as dormant (as per RFC2863). | |
3253 | * | |
3254 | * The dormant state indicates that the relevant interface is not | |
3255 | * actually in a condition to pass packets (i.e., it is not 'up') but is | |
3256 | * in a "pending" state, waiting for some external event. For "on- | |
3257 | * demand" interfaces, this new state identifies the situation where the | |
3258 | * interface is waiting for events to place it in the up state. | |
3259 | * | |
3260 | */ | |
3261 | static inline void netif_dormant_on(struct net_device *dev) | |
3262 | { | |
3263 | if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) | |
3264 | linkwatch_fire_event(dev); | |
3265 | } | |
3266 | ||
3267 | /** | |
3268 | * netif_dormant_off - set device as not dormant. | |
3269 | * @dev: network device | |
3270 | * | |
3271 | * Device is not in dormant state. | |
3272 | */ | |
3273 | static inline void netif_dormant_off(struct net_device *dev) | |
3274 | { | |
3275 | if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) | |
3276 | linkwatch_fire_event(dev); | |
3277 | } | |
3278 | ||
3279 | /** | |
3280 | * netif_dormant - test if carrier present | |
3281 | * @dev: network device | |
3282 | * | |
3283 | * Check if carrier is present on device | |
3284 | */ | |
3285 | static inline bool netif_dormant(const struct net_device *dev) | |
3286 | { | |
3287 | return test_bit(__LINK_STATE_DORMANT, &dev->state); | |
3288 | } | |
3289 | ||
3290 | ||
3291 | /** | |
3292 | * netif_oper_up - test if device is operational | |
3293 | * @dev: network device | |
3294 | * | |
3295 | * Check if carrier is operational | |
3296 | */ | |
3297 | static inline bool netif_oper_up(const struct net_device *dev) | |
3298 | { | |
3299 | return (dev->operstate == IF_OPER_UP || | |
3300 | dev->operstate == IF_OPER_UNKNOWN /* backward compat */); | |
3301 | } | |
3302 | ||
3303 | /** | |
3304 | * netif_device_present - is device available or removed | |
3305 | * @dev: network device | |
3306 | * | |
3307 | * Check if device has not been removed from system. | |
3308 | */ | |
3309 | static inline bool netif_device_present(struct net_device *dev) | |
3310 | { | |
3311 | return test_bit(__LINK_STATE_PRESENT, &dev->state); | |
3312 | } | |
3313 | ||
3314 | void netif_device_detach(struct net_device *dev); | |
3315 | ||
3316 | void netif_device_attach(struct net_device *dev); | |
3317 | ||
3318 | /* | |
3319 | * Network interface message level settings | |
3320 | */ | |
3321 | ||
3322 | enum { | |
3323 | NETIF_MSG_DRV = 0x0001, | |
3324 | NETIF_MSG_PROBE = 0x0002, | |
3325 | NETIF_MSG_LINK = 0x0004, | |
3326 | NETIF_MSG_TIMER = 0x0008, | |
3327 | NETIF_MSG_IFDOWN = 0x0010, | |
3328 | NETIF_MSG_IFUP = 0x0020, | |
3329 | NETIF_MSG_RX_ERR = 0x0040, | |
3330 | NETIF_MSG_TX_ERR = 0x0080, | |
3331 | NETIF_MSG_TX_QUEUED = 0x0100, | |
3332 | NETIF_MSG_INTR = 0x0200, | |
3333 | NETIF_MSG_TX_DONE = 0x0400, | |
3334 | NETIF_MSG_RX_STATUS = 0x0800, | |
3335 | NETIF_MSG_PKTDATA = 0x1000, | |
3336 | NETIF_MSG_HW = 0x2000, | |
3337 | NETIF_MSG_WOL = 0x4000, | |
3338 | }; | |
3339 | ||
3340 | #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV) | |
3341 | #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE) | |
3342 | #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK) | |
3343 | #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER) | |
3344 | #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN) | |
3345 | #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP) | |
3346 | #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR) | |
3347 | #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR) | |
3348 | #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED) | |
3349 | #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR) | |
3350 | #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE) | |
3351 | #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS) | |
3352 | #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA) | |
3353 | #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW) | |
3354 | #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL) | |
3355 | ||
3356 | static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits) | |
3357 | { | |
3358 | /* use default */ | |
3359 | if (debug_value < 0 || debug_value >= (sizeof(u32) * 8)) | |
3360 | return default_msg_enable_bits; | |
3361 | if (debug_value == 0) /* no output */ | |
3362 | return 0; | |
3363 | /* set low N bits */ | |
3364 | return (1 << debug_value) - 1; | |
3365 | } | |
3366 | ||
3367 | static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu) | |
3368 | { | |
3369 | spin_lock(&txq->_xmit_lock); | |
3370 | txq->xmit_lock_owner = cpu; | |
3371 | } | |
3372 | ||
3373 | static inline void __netif_tx_lock_bh(struct netdev_queue *txq) | |
3374 | { | |
3375 | spin_lock_bh(&txq->_xmit_lock); | |
3376 | txq->xmit_lock_owner = smp_processor_id(); | |
3377 | } | |
3378 | ||
3379 | static inline bool __netif_tx_trylock(struct netdev_queue *txq) | |
3380 | { | |
3381 | bool ok = spin_trylock(&txq->_xmit_lock); | |
3382 | if (likely(ok)) | |
3383 | txq->xmit_lock_owner = smp_processor_id(); | |
3384 | return ok; | |
3385 | } | |
3386 | ||
3387 | static inline void __netif_tx_unlock(struct netdev_queue *txq) | |
3388 | { | |
3389 | txq->xmit_lock_owner = -1; | |
3390 | spin_unlock(&txq->_xmit_lock); | |
3391 | } | |
3392 | ||
3393 | static inline void __netif_tx_unlock_bh(struct netdev_queue *txq) | |
3394 | { | |
3395 | txq->xmit_lock_owner = -1; | |
3396 | spin_unlock_bh(&txq->_xmit_lock); | |
3397 | } | |
3398 | ||
3399 | static inline void txq_trans_update(struct netdev_queue *txq) | |
3400 | { | |
3401 | if (txq->xmit_lock_owner != -1) | |
3402 | txq->trans_start = jiffies; | |
3403 | } | |
3404 | ||
3405 | /** | |
3406 | * netif_tx_lock - grab network device transmit lock | |
3407 | * @dev: network device | |
3408 | * | |
3409 | * Get network device transmit lock | |
3410 | */ | |
3411 | static inline void netif_tx_lock(struct net_device *dev) | |
3412 | { | |
3413 | unsigned int i; | |
3414 | int cpu; | |
3415 | ||
3416 | spin_lock(&dev->tx_global_lock); | |
3417 | cpu = smp_processor_id(); | |
3418 | for (i = 0; i < dev->num_tx_queues; i++) { | |
3419 | struct netdev_queue *txq = netdev_get_tx_queue(dev, i); | |
3420 | ||
3421 | /* We are the only thread of execution doing a | |
3422 | * freeze, but we have to grab the _xmit_lock in | |
3423 | * order to synchronize with threads which are in | |
3424 | * the ->hard_start_xmit() handler and already | |
3425 | * checked the frozen bit. | |
3426 | */ | |
3427 | __netif_tx_lock(txq, cpu); | |
3428 | set_bit(__QUEUE_STATE_FROZEN, &txq->state); | |
3429 | __netif_tx_unlock(txq); | |
3430 | } | |
3431 | } | |
3432 | ||
3433 | static inline void netif_tx_lock_bh(struct net_device *dev) | |
3434 | { | |
3435 | local_bh_disable(); | |
3436 | netif_tx_lock(dev); | |
3437 | } | |
3438 | ||
3439 | static inline void netif_tx_unlock(struct net_device *dev) | |
3440 | { | |
3441 | unsigned int i; | |
3442 | ||
3443 | for (i = 0; i < dev->num_tx_queues; i++) { | |
3444 | struct netdev_queue *txq = netdev_get_tx_queue(dev, i); | |
3445 | ||
3446 | /* No need to grab the _xmit_lock here. If the | |
3447 | * queue is not stopped for another reason, we | |
3448 | * force a schedule. | |
3449 | */ | |
3450 | clear_bit(__QUEUE_STATE_FROZEN, &txq->state); | |
3451 | netif_schedule_queue(txq); | |
3452 | } | |
3453 | spin_unlock(&dev->tx_global_lock); | |
3454 | } | |
3455 | ||
3456 | static inline void netif_tx_unlock_bh(struct net_device *dev) | |
3457 | { | |
3458 | netif_tx_unlock(dev); | |
3459 | local_bh_enable(); | |
3460 | } | |
3461 | ||
3462 | #define HARD_TX_LOCK(dev, txq, cpu) { \ | |
3463 | if ((dev->features & NETIF_F_LLTX) == 0) { \ | |
3464 | __netif_tx_lock(txq, cpu); \ | |
3465 | } \ | |
3466 | } | |
3467 | ||
3468 | #define HARD_TX_TRYLOCK(dev, txq) \ | |
3469 | (((dev->features & NETIF_F_LLTX) == 0) ? \ | |
3470 | __netif_tx_trylock(txq) : \ | |
3471 | true ) | |
3472 | ||
3473 | #define HARD_TX_UNLOCK(dev, txq) { \ | |
3474 | if ((dev->features & NETIF_F_LLTX) == 0) { \ | |
3475 | __netif_tx_unlock(txq); \ | |
3476 | } \ | |
3477 | } | |
3478 | ||
3479 | static inline void netif_tx_disable(struct net_device *dev) | |
3480 | { | |
3481 | unsigned int i; | |
3482 | int cpu; | |
3483 | ||
3484 | local_bh_disable(); | |
3485 | cpu = smp_processor_id(); | |
3486 | for (i = 0; i < dev->num_tx_queues; i++) { | |
3487 | struct netdev_queue *txq = netdev_get_tx_queue(dev, i); | |
3488 | ||
3489 | __netif_tx_lock(txq, cpu); | |
3490 | netif_tx_stop_queue(txq); | |
3491 | __netif_tx_unlock(txq); | |
3492 | } | |
3493 | local_bh_enable(); | |
3494 | } | |
3495 | ||
3496 | static inline void netif_addr_lock(struct net_device *dev) | |
3497 | { | |
3498 | spin_lock(&dev->addr_list_lock); | |
3499 | } | |
3500 | ||
3501 | static inline void netif_addr_lock_nested(struct net_device *dev) | |
3502 | { | |
3503 | int subclass = SINGLE_DEPTH_NESTING; | |
3504 | ||
3505 | if (dev->netdev_ops->ndo_get_lock_subclass) | |
3506 | subclass = dev->netdev_ops->ndo_get_lock_subclass(dev); | |
3507 | ||
3508 | spin_lock_nested(&dev->addr_list_lock, subclass); | |
3509 | } | |
3510 | ||
3511 | static inline void netif_addr_lock_bh(struct net_device *dev) | |
3512 | { | |
3513 | spin_lock_bh(&dev->addr_list_lock); | |
3514 | } | |
3515 | ||
3516 | static inline void netif_addr_unlock(struct net_device *dev) | |
3517 | { | |
3518 | spin_unlock(&dev->addr_list_lock); | |
3519 | } | |
3520 | ||
3521 | static inline void netif_addr_unlock_bh(struct net_device *dev) | |
3522 | { | |
3523 | spin_unlock_bh(&dev->addr_list_lock); | |
3524 | } | |
3525 | ||
3526 | /* | |
3527 | * dev_addrs walker. Should be used only for read access. Call with | |
3528 | * rcu_read_lock held. | |
3529 | */ | |
3530 | #define for_each_dev_addr(dev, ha) \ | |
3531 | list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list) | |
3532 | ||
3533 | /* These functions live elsewhere (drivers/net/net_init.c, but related) */ | |
3534 | ||
3535 | void ether_setup(struct net_device *dev); | |
3536 | ||
3537 | /* Support for loadable net-drivers */ | |
3538 | struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name, | |
3539 | unsigned char name_assign_type, | |
3540 | void (*setup)(struct net_device *), | |
3541 | unsigned int txqs, unsigned int rxqs); | |
3542 | #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \ | |
3543 | alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1) | |
3544 | ||
3545 | #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \ | |
3546 | alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \ | |
3547 | count) | |
3548 | ||
3549 | int register_netdev(struct net_device *dev); | |
3550 | void unregister_netdev(struct net_device *dev); | |
3551 | ||
3552 | /* General hardware address lists handling functions */ | |
3553 | int __hw_addr_sync(struct netdev_hw_addr_list *to_list, | |
3554 | struct netdev_hw_addr_list *from_list, int addr_len); | |
3555 | void __hw_addr_unsync(struct netdev_hw_addr_list *to_list, | |
3556 | struct netdev_hw_addr_list *from_list, int addr_len); | |
3557 | int __hw_addr_sync_dev(struct netdev_hw_addr_list *list, | |
3558 | struct net_device *dev, | |
3559 | int (*sync)(struct net_device *, const unsigned char *), | |
3560 | int (*unsync)(struct net_device *, | |
3561 | const unsigned char *)); | |
3562 | void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list, | |
3563 | struct net_device *dev, | |
3564 | int (*unsync)(struct net_device *, | |
3565 | const unsigned char *)); | |
3566 | void __hw_addr_init(struct netdev_hw_addr_list *list); | |
3567 | ||
3568 | /* Functions used for device addresses handling */ | |
3569 | int dev_addr_add(struct net_device *dev, const unsigned char *addr, | |
3570 | unsigned char addr_type); | |
3571 | int dev_addr_del(struct net_device *dev, const unsigned char *addr, | |
3572 | unsigned char addr_type); | |
3573 | void dev_addr_flush(struct net_device *dev); | |
3574 | int dev_addr_init(struct net_device *dev); | |
3575 | ||
3576 | /* Functions used for unicast addresses handling */ | |
3577 | int dev_uc_add(struct net_device *dev, const unsigned char *addr); | |
3578 | int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr); | |
3579 | int dev_uc_del(struct net_device *dev, const unsigned char *addr); | |
3580 | int dev_uc_sync(struct net_device *to, struct net_device *from); | |
3581 | int dev_uc_sync_multiple(struct net_device *to, struct net_device *from); | |
3582 | void dev_uc_unsync(struct net_device *to, struct net_device *from); | |
3583 | void dev_uc_flush(struct net_device *dev); | |
3584 | void dev_uc_init(struct net_device *dev); | |
3585 | ||
3586 | /** | |
3587 | * __dev_uc_sync - Synchonize device's unicast list | |
3588 | * @dev: device to sync | |
3589 | * @sync: function to call if address should be added | |
3590 | * @unsync: function to call if address should be removed | |
3591 | * | |
3592 | * Add newly added addresses to the interface, and release | |
3593 | * addresses that have been deleted. | |
3594 | **/ | |
3595 | static inline int __dev_uc_sync(struct net_device *dev, | |
3596 | int (*sync)(struct net_device *, | |
3597 | const unsigned char *), | |
3598 | int (*unsync)(struct net_device *, | |
3599 | const unsigned char *)) | |
3600 | { | |
3601 | return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync); | |
3602 | } | |
3603 | ||
3604 | /** | |
3605 | * __dev_uc_unsync - Remove synchronized addresses from device | |
3606 | * @dev: device to sync | |
3607 | * @unsync: function to call if address should be removed | |
3608 | * | |
3609 | * Remove all addresses that were added to the device by dev_uc_sync(). | |
3610 | **/ | |
3611 | static inline void __dev_uc_unsync(struct net_device *dev, | |
3612 | int (*unsync)(struct net_device *, | |
3613 | const unsigned char *)) | |
3614 | { | |
3615 | __hw_addr_unsync_dev(&dev->uc, dev, unsync); | |
3616 | } | |
3617 | ||
3618 | /* Functions used for multicast addresses handling */ | |
3619 | int dev_mc_add(struct net_device *dev, const unsigned char *addr); | |
3620 | int dev_mc_add_global(struct net_device *dev, const unsigned char *addr); | |
3621 | int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr); | |
3622 | int dev_mc_del(struct net_device *dev, const unsigned char *addr); | |
3623 | int dev_mc_del_global(struct net_device *dev, const unsigned char *addr); | |
3624 | int dev_mc_sync(struct net_device *to, struct net_device *from); | |
3625 | int dev_mc_sync_multiple(struct net_device *to, struct net_device *from); | |
3626 | void dev_mc_unsync(struct net_device *to, struct net_device *from); | |
3627 | void dev_mc_flush(struct net_device *dev); | |
3628 | void dev_mc_init(struct net_device *dev); | |
3629 | ||
3630 | /** | |
3631 | * __dev_mc_sync - Synchonize device's multicast list | |
3632 | * @dev: device to sync | |
3633 | * @sync: function to call if address should be added | |
3634 | * @unsync: function to call if address should be removed | |
3635 | * | |
3636 | * Add newly added addresses to the interface, and release | |
3637 | * addresses that have been deleted. | |
3638 | **/ | |
3639 | static inline int __dev_mc_sync(struct net_device *dev, | |
3640 | int (*sync)(struct net_device *, | |
3641 | const unsigned char *), | |
3642 | int (*unsync)(struct net_device *, | |
3643 | const unsigned char *)) | |
3644 | { | |
3645 | return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync); | |
3646 | } | |
3647 | ||
3648 | /** | |
3649 | * __dev_mc_unsync - Remove synchronized addresses from device | |
3650 | * @dev: device to sync | |
3651 | * @unsync: function to call if address should be removed | |
3652 | * | |
3653 | * Remove all addresses that were added to the device by dev_mc_sync(). | |
3654 | **/ | |
3655 | static inline void __dev_mc_unsync(struct net_device *dev, | |
3656 | int (*unsync)(struct net_device *, | |
3657 | const unsigned char *)) | |
3658 | { | |
3659 | __hw_addr_unsync_dev(&dev->mc, dev, unsync); | |
3660 | } | |
3661 | ||
3662 | /* Functions used for secondary unicast and multicast support */ | |
3663 | void dev_set_rx_mode(struct net_device *dev); | |
3664 | void __dev_set_rx_mode(struct net_device *dev); | |
3665 | int dev_set_promiscuity(struct net_device *dev, int inc); | |
3666 | int dev_set_allmulti(struct net_device *dev, int inc); | |
3667 | void netdev_state_change(struct net_device *dev); | |
3668 | void netdev_notify_peers(struct net_device *dev); | |
3669 | void netdev_features_change(struct net_device *dev); | |
3670 | /* Load a device via the kmod */ | |
3671 | void dev_load(struct net *net, const char *name); | |
3672 | struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev, | |
3673 | struct rtnl_link_stats64 *storage); | |
3674 | void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64, | |
3675 | const struct net_device_stats *netdev_stats); | |
3676 | ||
3677 | extern int netdev_max_backlog; | |
3678 | extern int netdev_tstamp_prequeue; | |
3679 | extern int weight_p; | |
3680 | extern int bpf_jit_enable; | |
3681 | ||
3682 | bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev); | |
3683 | struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev, | |
3684 | struct list_head **iter); | |
3685 | struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, | |
3686 | struct list_head **iter); | |
3687 | ||
3688 | /* iterate through upper list, must be called under RCU read lock */ | |
3689 | #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \ | |
3690 | for (iter = &(dev)->adj_list.upper, \ | |
3691 | updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ | |
3692 | updev; \ | |
3693 | updev = netdev_upper_get_next_dev_rcu(dev, &(iter))) | |
3694 | ||
3695 | /* iterate through upper list, must be called under RCU read lock */ | |
3696 | #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \ | |
3697 | for (iter = &(dev)->all_adj_list.upper, \ | |
3698 | updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ | |
3699 | updev; \ | |
3700 | updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) | |
3701 | ||
3702 | void *netdev_lower_get_next_private(struct net_device *dev, | |
3703 | struct list_head **iter); | |
3704 | void *netdev_lower_get_next_private_rcu(struct net_device *dev, | |
3705 | struct list_head **iter); | |
3706 | ||
3707 | #define netdev_for_each_lower_private(dev, priv, iter) \ | |
3708 | for (iter = (dev)->adj_list.lower.next, \ | |
3709 | priv = netdev_lower_get_next_private(dev, &(iter)); \ | |
3710 | priv; \ | |
3711 | priv = netdev_lower_get_next_private(dev, &(iter))) | |
3712 | ||
3713 | #define netdev_for_each_lower_private_rcu(dev, priv, iter) \ | |
3714 | for (iter = &(dev)->adj_list.lower, \ | |
3715 | priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ | |
3716 | priv; \ | |
3717 | priv = netdev_lower_get_next_private_rcu(dev, &(iter))) | |
3718 | ||
3719 | void *netdev_lower_get_next(struct net_device *dev, | |
3720 | struct list_head **iter); | |
3721 | #define netdev_for_each_lower_dev(dev, ldev, iter) \ | |
3722 | for (iter = &(dev)->adj_list.lower, \ | |
3723 | ldev = netdev_lower_get_next(dev, &(iter)); \ | |
3724 | ldev; \ | |
3725 | ldev = netdev_lower_get_next(dev, &(iter))) | |
3726 | ||
3727 | void *netdev_adjacent_get_private(struct list_head *adj_list); | |
3728 | void *netdev_lower_get_first_private_rcu(struct net_device *dev); | |
3729 | struct net_device *netdev_master_upper_dev_get(struct net_device *dev); | |
3730 | struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev); | |
3731 | int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev); | |
3732 | int netdev_master_upper_dev_link(struct net_device *dev, | |
3733 | struct net_device *upper_dev, | |
3734 | void *upper_priv, void *upper_info); | |
3735 | void netdev_upper_dev_unlink(struct net_device *dev, | |
3736 | struct net_device *upper_dev); | |
3737 | void netdev_adjacent_rename_links(struct net_device *dev, char *oldname); | |
3738 | void *netdev_lower_dev_get_private(struct net_device *dev, | |
3739 | struct net_device *lower_dev); | |
3740 | void netdev_lower_state_changed(struct net_device *lower_dev, | |
3741 | void *lower_state_info); | |
3742 | ||
3743 | /* RSS keys are 40 or 52 bytes long */ | |
3744 | #define NETDEV_RSS_KEY_LEN 52 | |
3745 | extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN]; | |
3746 | void netdev_rss_key_fill(void *buffer, size_t len); | |
3747 | ||
3748 | int dev_get_nest_level(struct net_device *dev, | |
3749 | bool (*type_check)(const struct net_device *dev)); | |
3750 | int skb_checksum_help(struct sk_buff *skb); | |
3751 | struct sk_buff *__skb_gso_segment(struct sk_buff *skb, | |
3752 | netdev_features_t features, bool tx_path); | |
3753 | struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb, | |
3754 | netdev_features_t features); | |
3755 | ||
3756 | struct netdev_bonding_info { | |
3757 | ifslave slave; | |
3758 | ifbond master; | |
3759 | }; | |
3760 | ||
3761 | struct netdev_notifier_bonding_info { | |
3762 | struct netdev_notifier_info info; /* must be first */ | |
3763 | struct netdev_bonding_info bonding_info; | |
3764 | }; | |
3765 | ||
3766 | void netdev_bonding_info_change(struct net_device *dev, | |
3767 | struct netdev_bonding_info *bonding_info); | |
3768 | ||
3769 | static inline | |
3770 | struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features) | |
3771 | { | |
3772 | return __skb_gso_segment(skb, features, true); | |
3773 | } | |
3774 | __be16 skb_network_protocol(struct sk_buff *skb, int *depth); | |
3775 | ||
3776 | static inline bool can_checksum_protocol(netdev_features_t features, | |
3777 | __be16 protocol) | |
3778 | { | |
3779 | if (protocol == htons(ETH_P_FCOE)) | |
3780 | return !!(features & NETIF_F_FCOE_CRC); | |
3781 | ||
3782 | /* Assume this is an IP checksum (not SCTP CRC) */ | |
3783 | ||
3784 | if (features & NETIF_F_HW_CSUM) { | |
3785 | /* Can checksum everything */ | |
3786 | return true; | |
3787 | } | |
3788 | ||
3789 | switch (protocol) { | |
3790 | case htons(ETH_P_IP): | |
3791 | return !!(features & NETIF_F_IP_CSUM); | |
3792 | case htons(ETH_P_IPV6): | |
3793 | return !!(features & NETIF_F_IPV6_CSUM); | |
3794 | default: | |
3795 | return false; | |
3796 | } | |
3797 | } | |
3798 | ||
3799 | /* Map an ethertype into IP protocol if possible */ | |
3800 | static inline int eproto_to_ipproto(int eproto) | |
3801 | { | |
3802 | switch (eproto) { | |
3803 | case htons(ETH_P_IP): | |
3804 | return IPPROTO_IP; | |
3805 | case htons(ETH_P_IPV6): | |
3806 | return IPPROTO_IPV6; | |
3807 | default: | |
3808 | return -1; | |
3809 | } | |
3810 | } | |
3811 | ||
3812 | #ifdef CONFIG_BUG | |
3813 | void netdev_rx_csum_fault(struct net_device *dev); | |
3814 | #else | |
3815 | static inline void netdev_rx_csum_fault(struct net_device *dev) | |
3816 | { | |
3817 | } | |
3818 | #endif | |
3819 | /* rx skb timestamps */ | |
3820 | void net_enable_timestamp(void); | |
3821 | void net_disable_timestamp(void); | |
3822 | ||
3823 | #ifdef CONFIG_PROC_FS | |
3824 | int __init dev_proc_init(void); | |
3825 | #else | |
3826 | #define dev_proc_init() 0 | |
3827 | #endif | |
3828 | ||
3829 | static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops, | |
3830 | struct sk_buff *skb, struct net_device *dev, | |
3831 | bool more) | |
3832 | { | |
3833 | skb->xmit_more = more ? 1 : 0; | |
3834 | return ops->ndo_start_xmit(skb, dev); | |
3835 | } | |
3836 | ||
3837 | static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev, | |
3838 | struct netdev_queue *txq, bool more) | |
3839 | { | |
3840 | const struct net_device_ops *ops = dev->netdev_ops; | |
3841 | int rc; | |
3842 | ||
3843 | rc = __netdev_start_xmit(ops, skb, dev, more); | |
3844 | if (rc == NETDEV_TX_OK) | |
3845 | txq_trans_update(txq); | |
3846 | ||
3847 | return rc; | |
3848 | } | |
3849 | ||
3850 | int netdev_class_create_file_ns(struct class_attribute *class_attr, | |
3851 | const void *ns); | |
3852 | void netdev_class_remove_file_ns(struct class_attribute *class_attr, | |
3853 | const void *ns); | |
3854 | ||
3855 | static inline int netdev_class_create_file(struct class_attribute *class_attr) | |
3856 | { | |
3857 | return netdev_class_create_file_ns(class_attr, NULL); | |
3858 | } | |
3859 | ||
3860 | static inline void netdev_class_remove_file(struct class_attribute *class_attr) | |
3861 | { | |
3862 | netdev_class_remove_file_ns(class_attr, NULL); | |
3863 | } | |
3864 | ||
3865 | extern struct kobj_ns_type_operations net_ns_type_operations; | |
3866 | ||
3867 | const char *netdev_drivername(const struct net_device *dev); | |
3868 | ||
3869 | void linkwatch_run_queue(void); | |
3870 | ||
3871 | static inline netdev_features_t netdev_intersect_features(netdev_features_t f1, | |
3872 | netdev_features_t f2) | |
3873 | { | |
3874 | if ((f1 ^ f2) & NETIF_F_HW_CSUM) { | |
3875 | if (f1 & NETIF_F_HW_CSUM) | |
3876 | f1 |= (NETIF_F_IP_CSUM|NETIF_F_IP_CSUM); | |
3877 | else | |
3878 | f2 |= (NETIF_F_IP_CSUM|NETIF_F_IP_CSUM); | |
3879 | } | |
3880 | ||
3881 | return f1 & f2; | |
3882 | } | |
3883 | ||
3884 | static inline netdev_features_t netdev_get_wanted_features( | |
3885 | struct net_device *dev) | |
3886 | { | |
3887 | return (dev->features & ~dev->hw_features) | dev->wanted_features; | |
3888 | } | |
3889 | netdev_features_t netdev_increment_features(netdev_features_t all, | |
3890 | netdev_features_t one, netdev_features_t mask); | |
3891 | ||
3892 | /* Allow TSO being used on stacked device : | |
3893 | * Performing the GSO segmentation before last device | |
3894 | * is a performance improvement. | |
3895 | */ | |
3896 | static inline netdev_features_t netdev_add_tso_features(netdev_features_t features, | |
3897 | netdev_features_t mask) | |
3898 | { | |
3899 | return netdev_increment_features(features, NETIF_F_ALL_TSO, mask); | |
3900 | } | |
3901 | ||
3902 | int __netdev_update_features(struct net_device *dev); | |
3903 | void netdev_update_features(struct net_device *dev); | |
3904 | void netdev_change_features(struct net_device *dev); | |
3905 | ||
3906 | void netif_stacked_transfer_operstate(const struct net_device *rootdev, | |
3907 | struct net_device *dev); | |
3908 | ||
3909 | netdev_features_t passthru_features_check(struct sk_buff *skb, | |
3910 | struct net_device *dev, | |
3911 | netdev_features_t features); | |
3912 | netdev_features_t netif_skb_features(struct sk_buff *skb); | |
3913 | ||
3914 | static inline bool net_gso_ok(netdev_features_t features, int gso_type) | |
3915 | { | |
3916 | netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT; | |
3917 | ||
3918 | /* check flags correspondence */ | |
3919 | BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT)); | |
3920 | BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT)); | |
3921 | BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT)); | |
3922 | BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT)); | |
3923 | BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT)); | |
3924 | BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT)); | |
3925 | BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT)); | |
3926 | BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT)); | |
3927 | BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT)); | |
3928 | BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT)); | |
3929 | BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT)); | |
3930 | BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT)); | |
3931 | BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT)); | |
3932 | ||
3933 | return (features & feature) == feature; | |
3934 | } | |
3935 | ||
3936 | static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features) | |
3937 | { | |
3938 | return net_gso_ok(features, skb_shinfo(skb)->gso_type) && | |
3939 | (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST)); | |
3940 | } | |
3941 | ||
3942 | static inline bool netif_needs_gso(struct sk_buff *skb, | |
3943 | netdev_features_t features) | |
3944 | { | |
3945 | return skb_is_gso(skb) && (!skb_gso_ok(skb, features) || | |
3946 | unlikely((skb->ip_summed != CHECKSUM_PARTIAL) && | |
3947 | (skb->ip_summed != CHECKSUM_UNNECESSARY))); | |
3948 | } | |
3949 | ||
3950 | static inline void netif_set_gso_max_size(struct net_device *dev, | |
3951 | unsigned int size) | |
3952 | { | |
3953 | dev->gso_max_size = size; | |
3954 | } | |
3955 | ||
3956 | static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol, | |
3957 | int pulled_hlen, u16 mac_offset, | |
3958 | int mac_len) | |
3959 | { | |
3960 | skb->protocol = protocol; | |
3961 | skb->encapsulation = 1; | |
3962 | skb_push(skb, pulled_hlen); | |
3963 | skb_reset_transport_header(skb); | |
3964 | skb->mac_header = mac_offset; | |
3965 | skb->network_header = skb->mac_header + mac_len; | |
3966 | skb->mac_len = mac_len; | |
3967 | } | |
3968 | ||
3969 | static inline bool netif_is_macvlan(const struct net_device *dev) | |
3970 | { | |
3971 | return dev->priv_flags & IFF_MACVLAN; | |
3972 | } | |
3973 | ||
3974 | static inline bool netif_is_macvlan_port(const struct net_device *dev) | |
3975 | { | |
3976 | return dev->priv_flags & IFF_MACVLAN_PORT; | |
3977 | } | |
3978 | ||
3979 | static inline bool netif_is_ipvlan(const struct net_device *dev) | |
3980 | { | |
3981 | return dev->priv_flags & IFF_IPVLAN_SLAVE; | |
3982 | } | |
3983 | ||
3984 | static inline bool netif_is_ipvlan_port(const struct net_device *dev) | |
3985 | { | |
3986 | return dev->priv_flags & IFF_IPVLAN_MASTER; | |
3987 | } | |
3988 | ||
3989 | static inline bool netif_is_bond_master(const struct net_device *dev) | |
3990 | { | |
3991 | return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING; | |
3992 | } | |
3993 | ||
3994 | static inline bool netif_is_bond_slave(const struct net_device *dev) | |
3995 | { | |
3996 | return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING; | |
3997 | } | |
3998 | ||
3999 | static inline bool netif_supports_nofcs(struct net_device *dev) | |
4000 | { | |
4001 | return dev->priv_flags & IFF_SUPP_NOFCS; | |
4002 | } | |
4003 | ||
4004 | static inline bool netif_is_l3_master(const struct net_device *dev) | |
4005 | { | |
4006 | return dev->priv_flags & IFF_L3MDEV_MASTER; | |
4007 | } | |
4008 | ||
4009 | static inline bool netif_is_l3_slave(const struct net_device *dev) | |
4010 | { | |
4011 | return dev->priv_flags & IFF_L3MDEV_SLAVE; | |
4012 | } | |
4013 | ||
4014 | static inline bool netif_is_bridge_master(const struct net_device *dev) | |
4015 | { | |
4016 | return dev->priv_flags & IFF_EBRIDGE; | |
4017 | } | |
4018 | ||
4019 | static inline bool netif_is_bridge_port(const struct net_device *dev) | |
4020 | { | |
4021 | return dev->priv_flags & IFF_BRIDGE_PORT; | |
4022 | } | |
4023 | ||
4024 | static inline bool netif_is_ovs_master(const struct net_device *dev) | |
4025 | { | |
4026 | return dev->priv_flags & IFF_OPENVSWITCH; | |
4027 | } | |
4028 | ||
4029 | static inline bool netif_is_team_master(const struct net_device *dev) | |
4030 | { | |
4031 | return dev->priv_flags & IFF_TEAM; | |
4032 | } | |
4033 | ||
4034 | static inline bool netif_is_team_port(const struct net_device *dev) | |
4035 | { | |
4036 | return dev->priv_flags & IFF_TEAM_PORT; | |
4037 | } | |
4038 | ||
4039 | static inline bool netif_is_lag_master(const struct net_device *dev) | |
4040 | { | |
4041 | return netif_is_bond_master(dev) || netif_is_team_master(dev); | |
4042 | } | |
4043 | ||
4044 | static inline bool netif_is_lag_port(const struct net_device *dev) | |
4045 | { | |
4046 | return netif_is_bond_slave(dev) || netif_is_team_port(dev); | |
4047 | } | |
4048 | ||
4049 | /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */ | |
4050 | static inline void netif_keep_dst(struct net_device *dev) | |
4051 | { | |
4052 | dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM); | |
4053 | } | |
4054 | ||
4055 | extern struct pernet_operations __net_initdata loopback_net_ops; | |
4056 | ||
4057 | /* Logging, debugging and troubleshooting/diagnostic helpers. */ | |
4058 | ||
4059 | /* netdev_printk helpers, similar to dev_printk */ | |
4060 | ||
4061 | static inline const char *netdev_name(const struct net_device *dev) | |
4062 | { | |
4063 | if (!dev->name[0] || strchr(dev->name, '%')) | |
4064 | return "(unnamed net_device)"; | |
4065 | return dev->name; | |
4066 | } | |
4067 | ||
4068 | static inline const char *netdev_reg_state(const struct net_device *dev) | |
4069 | { | |
4070 | switch (dev->reg_state) { | |
4071 | case NETREG_UNINITIALIZED: return " (uninitialized)"; | |
4072 | case NETREG_REGISTERED: return ""; | |
4073 | case NETREG_UNREGISTERING: return " (unregistering)"; | |
4074 | case NETREG_UNREGISTERED: return " (unregistered)"; | |
4075 | case NETREG_RELEASED: return " (released)"; | |
4076 | case NETREG_DUMMY: return " (dummy)"; | |
4077 | } | |
4078 | ||
4079 | WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state); | |
4080 | return " (unknown)"; | |
4081 | } | |
4082 | ||
4083 | __printf(3, 4) | |
4084 | void netdev_printk(const char *level, const struct net_device *dev, | |
4085 | const char *format, ...); | |
4086 | __printf(2, 3) | |
4087 | void netdev_emerg(const struct net_device *dev, const char *format, ...); | |
4088 | __printf(2, 3) | |
4089 | void netdev_alert(const struct net_device *dev, const char *format, ...); | |
4090 | __printf(2, 3) | |
4091 | void netdev_crit(const struct net_device *dev, const char *format, ...); | |
4092 | __printf(2, 3) | |
4093 | void netdev_err(const struct net_device *dev, const char *format, ...); | |
4094 | __printf(2, 3) | |
4095 | void netdev_warn(const struct net_device *dev, const char *format, ...); | |
4096 | __printf(2, 3) | |
4097 | void netdev_notice(const struct net_device *dev, const char *format, ...); | |
4098 | __printf(2, 3) | |
4099 | void netdev_info(const struct net_device *dev, const char *format, ...); | |
4100 | ||
4101 | #define MODULE_ALIAS_NETDEV(device) \ | |
4102 | MODULE_ALIAS("netdev-" device) | |
4103 | ||
4104 | #if defined(CONFIG_DYNAMIC_DEBUG) | |
4105 | #define netdev_dbg(__dev, format, args...) \ | |
4106 | do { \ | |
4107 | dynamic_netdev_dbg(__dev, format, ##args); \ | |
4108 | } while (0) | |
4109 | #elif defined(DEBUG) | |
4110 | #define netdev_dbg(__dev, format, args...) \ | |
4111 | netdev_printk(KERN_DEBUG, __dev, format, ##args) | |
4112 | #else | |
4113 | #define netdev_dbg(__dev, format, args...) \ | |
4114 | ({ \ | |
4115 | if (0) \ | |
4116 | netdev_printk(KERN_DEBUG, __dev, format, ##args); \ | |
4117 | }) | |
4118 | #endif | |
4119 | ||
4120 | #if defined(VERBOSE_DEBUG) | |
4121 | #define netdev_vdbg netdev_dbg | |
4122 | #else | |
4123 | ||
4124 | #define netdev_vdbg(dev, format, args...) \ | |
4125 | ({ \ | |
4126 | if (0) \ | |
4127 | netdev_printk(KERN_DEBUG, dev, format, ##args); \ | |
4128 | 0; \ | |
4129 | }) | |
4130 | #endif | |
4131 | ||
4132 | /* | |
4133 | * netdev_WARN() acts like dev_printk(), but with the key difference | |
4134 | * of using a WARN/WARN_ON to get the message out, including the | |
4135 | * file/line information and a backtrace. | |
4136 | */ | |
4137 | #define netdev_WARN(dev, format, args...) \ | |
4138 | WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \ | |
4139 | netdev_reg_state(dev), ##args) | |
4140 | ||
4141 | /* netif printk helpers, similar to netdev_printk */ | |
4142 | ||
4143 | #define netif_printk(priv, type, level, dev, fmt, args...) \ | |
4144 | do { \ | |
4145 | if (netif_msg_##type(priv)) \ | |
4146 | netdev_printk(level, (dev), fmt, ##args); \ | |
4147 | } while (0) | |
4148 | ||
4149 | #define netif_level(level, priv, type, dev, fmt, args...) \ | |
4150 | do { \ | |
4151 | if (netif_msg_##type(priv)) \ | |
4152 | netdev_##level(dev, fmt, ##args); \ | |
4153 | } while (0) | |
4154 | ||
4155 | #define netif_emerg(priv, type, dev, fmt, args...) \ | |
4156 | netif_level(emerg, priv, type, dev, fmt, ##args) | |
4157 | #define netif_alert(priv, type, dev, fmt, args...) \ | |
4158 | netif_level(alert, priv, type, dev, fmt, ##args) | |
4159 | #define netif_crit(priv, type, dev, fmt, args...) \ | |
4160 | netif_level(crit, priv, type, dev, fmt, ##args) | |
4161 | #define netif_err(priv, type, dev, fmt, args...) \ | |
4162 | netif_level(err, priv, type, dev, fmt, ##args) | |
4163 | #define netif_warn(priv, type, dev, fmt, args...) \ | |
4164 | netif_level(warn, priv, type, dev, fmt, ##args) | |
4165 | #define netif_notice(priv, type, dev, fmt, args...) \ | |
4166 | netif_level(notice, priv, type, dev, fmt, ##args) | |
4167 | #define netif_info(priv, type, dev, fmt, args...) \ | |
4168 | netif_level(info, priv, type, dev, fmt, ##args) | |
4169 | ||
4170 | #if defined(CONFIG_DYNAMIC_DEBUG) | |
4171 | #define netif_dbg(priv, type, netdev, format, args...) \ | |
4172 | do { \ | |
4173 | if (netif_msg_##type(priv)) \ | |
4174 | dynamic_netdev_dbg(netdev, format, ##args); \ | |
4175 | } while (0) | |
4176 | #elif defined(DEBUG) | |
4177 | #define netif_dbg(priv, type, dev, format, args...) \ | |
4178 | netif_printk(priv, type, KERN_DEBUG, dev, format, ##args) | |
4179 | #else | |
4180 | #define netif_dbg(priv, type, dev, format, args...) \ | |
4181 | ({ \ | |
4182 | if (0) \ | |
4183 | netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ | |
4184 | 0; \ | |
4185 | }) | |
4186 | #endif | |
4187 | ||
4188 | #if defined(VERBOSE_DEBUG) | |
4189 | #define netif_vdbg netif_dbg | |
4190 | #else | |
4191 | #define netif_vdbg(priv, type, dev, format, args...) \ | |
4192 | ({ \ | |
4193 | if (0) \ | |
4194 | netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \ | |
4195 | 0; \ | |
4196 | }) | |
4197 | #endif | |
4198 | ||
4199 | /* | |
4200 | * The list of packet types we will receive (as opposed to discard) | |
4201 | * and the routines to invoke. | |
4202 | * | |
4203 | * Why 16. Because with 16 the only overlap we get on a hash of the | |
4204 | * low nibble of the protocol value is RARP/SNAP/X.25. | |
4205 | * | |
4206 | * NOTE: That is no longer true with the addition of VLAN tags. Not | |
4207 | * sure which should go first, but I bet it won't make much | |
4208 | * difference if we are running VLANs. The good news is that | |
4209 | * this protocol won't be in the list unless compiled in, so | |
4210 | * the average user (w/out VLANs) will not be adversely affected. | |
4211 | * --BLG | |
4212 | * | |
4213 | * 0800 IP | |
4214 | * 8100 802.1Q VLAN | |
4215 | * 0001 802.3 | |
4216 | * 0002 AX.25 | |
4217 | * 0004 802.2 | |
4218 | * 8035 RARP | |
4219 | * 0005 SNAP | |
4220 | * 0805 X.25 | |
4221 | * 0806 ARP | |
4222 | * 8137 IPX | |
4223 | * 0009 Localtalk | |
4224 | * 86DD IPv6 | |
4225 | */ | |
4226 | #define PTYPE_HASH_SIZE (16) | |
4227 | #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1) | |
4228 | ||
4229 | #endif /* _LINUX_NETDEVICE_H */ |