bridge: fdb dumping takes a filter device
[deliverable/linux.git] / include / linux / netdevice.h
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/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.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
54 struct netpoll_info;
55 struct device;
56 struct phy_device;
57 /* 802.11 specific */
58 struct wireless_dev;
59
60 void netdev_set_default_ethtool_ops(struct net_device *dev,
61 const struct ethtool_ops *ops);
62
63 /* Backlog congestion levels */
64 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
65 #define NET_RX_DROP 1 /* packet dropped */
66
67 /*
68 * Transmit return codes: transmit return codes originate from three different
69 * namespaces:
70 *
71 * - qdisc return codes
72 * - driver transmit return codes
73 * - errno values
74 *
75 * Drivers are allowed to return any one of those in their hard_start_xmit()
76 * function. Real network devices commonly used with qdiscs should only return
77 * the driver transmit return codes though - when qdiscs are used, the actual
78 * transmission happens asynchronously, so the value is not propagated to
79 * higher layers. Virtual network devices transmit synchronously, in this case
80 * the driver transmit return codes are consumed by dev_queue_xmit(), all
81 * others are propagated to higher layers.
82 */
83
84 /* qdisc ->enqueue() return codes. */
85 #define NET_XMIT_SUCCESS 0x00
86 #define NET_XMIT_DROP 0x01 /* skb dropped */
87 #define NET_XMIT_CN 0x02 /* congestion notification */
88 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
89 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
90
91 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
92 * indicates that the device will soon be dropping packets, or already drops
93 * some packets of the same priority; prompting us to send less aggressively. */
94 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
95 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
96
97 /* Driver transmit return codes */
98 #define NETDEV_TX_MASK 0xf0
99
100 enum netdev_tx {
101 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
102 NETDEV_TX_OK = 0x00, /* driver took care of packet */
103 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
104 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
105 };
106 typedef enum netdev_tx netdev_tx_t;
107
108 /*
109 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
110 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
111 */
112 static inline bool dev_xmit_complete(int rc)
113 {
114 /*
115 * Positive cases with an skb consumed by a driver:
116 * - successful transmission (rc == NETDEV_TX_OK)
117 * - error while transmitting (rc < 0)
118 * - error while queueing to a different device (rc & NET_XMIT_MASK)
119 */
120 if (likely(rc < NET_XMIT_MASK))
121 return true;
122
123 return false;
124 }
125
126 /*
127 * Compute the worst case header length according to the protocols
128 * used.
129 */
130
131 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
132 # if defined(CONFIG_MAC80211_MESH)
133 # define LL_MAX_HEADER 128
134 # else
135 # define LL_MAX_HEADER 96
136 # endif
137 #else
138 # define LL_MAX_HEADER 32
139 #endif
140
141 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
142 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
143 #define MAX_HEADER LL_MAX_HEADER
144 #else
145 #define MAX_HEADER (LL_MAX_HEADER + 48)
146 #endif
147
148 /*
149 * Old network device statistics. Fields are native words
150 * (unsigned long) so they can be read and written atomically.
151 */
152
153 struct net_device_stats {
154 unsigned long rx_packets;
155 unsigned long tx_packets;
156 unsigned long rx_bytes;
157 unsigned long tx_bytes;
158 unsigned long rx_errors;
159 unsigned long tx_errors;
160 unsigned long rx_dropped;
161 unsigned long tx_dropped;
162 unsigned long multicast;
163 unsigned long collisions;
164 unsigned long rx_length_errors;
165 unsigned long rx_over_errors;
166 unsigned long rx_crc_errors;
167 unsigned long rx_frame_errors;
168 unsigned long rx_fifo_errors;
169 unsigned long rx_missed_errors;
170 unsigned long tx_aborted_errors;
171 unsigned long tx_carrier_errors;
172 unsigned long tx_fifo_errors;
173 unsigned long tx_heartbeat_errors;
174 unsigned long tx_window_errors;
175 unsigned long rx_compressed;
176 unsigned long tx_compressed;
177 };
178
179
180 #include <linux/cache.h>
181 #include <linux/skbuff.h>
182
183 #ifdef CONFIG_RPS
184 #include <linux/static_key.h>
185 extern struct static_key rps_needed;
186 #endif
187
188 struct neighbour;
189 struct neigh_parms;
190 struct sk_buff;
191
192 struct netdev_hw_addr {
193 struct list_head list;
194 unsigned char addr[MAX_ADDR_LEN];
195 unsigned char type;
196 #define NETDEV_HW_ADDR_T_LAN 1
197 #define NETDEV_HW_ADDR_T_SAN 2
198 #define NETDEV_HW_ADDR_T_SLAVE 3
199 #define NETDEV_HW_ADDR_T_UNICAST 4
200 #define NETDEV_HW_ADDR_T_MULTICAST 5
201 bool global_use;
202 int sync_cnt;
203 int refcount;
204 int synced;
205 struct rcu_head rcu_head;
206 };
207
208 struct netdev_hw_addr_list {
209 struct list_head list;
210 int count;
211 };
212
213 #define netdev_hw_addr_list_count(l) ((l)->count)
214 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
215 #define netdev_hw_addr_list_for_each(ha, l) \
216 list_for_each_entry(ha, &(l)->list, list)
217
218 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
219 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
220 #define netdev_for_each_uc_addr(ha, dev) \
221 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
222
223 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
224 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
225 #define netdev_for_each_mc_addr(ha, dev) \
226 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
227
228 struct hh_cache {
229 u16 hh_len;
230 u16 __pad;
231 seqlock_t hh_lock;
232
233 /* cached hardware header; allow for machine alignment needs. */
234 #define HH_DATA_MOD 16
235 #define HH_DATA_OFF(__len) \
236 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
237 #define HH_DATA_ALIGN(__len) \
238 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
239 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
240 };
241
242 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
243 * Alternative is:
244 * dev->hard_header_len ? (dev->hard_header_len +
245 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
246 *
247 * We could use other alignment values, but we must maintain the
248 * relationship HH alignment <= LL alignment.
249 */
250 #define LL_RESERVED_SPACE(dev) \
251 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
252 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
253 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
254
255 struct header_ops {
256 int (*create) (struct sk_buff *skb, struct net_device *dev,
257 unsigned short type, const void *daddr,
258 const void *saddr, unsigned int len);
259 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
260 int (*rebuild)(struct sk_buff *skb);
261 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
262 void (*cache_update)(struct hh_cache *hh,
263 const struct net_device *dev,
264 const unsigned char *haddr);
265 };
266
267 /* These flag bits are private to the generic network queueing
268 * layer, they may not be explicitly referenced by any other
269 * code.
270 */
271
272 enum netdev_state_t {
273 __LINK_STATE_START,
274 __LINK_STATE_PRESENT,
275 __LINK_STATE_NOCARRIER,
276 __LINK_STATE_LINKWATCH_PENDING,
277 __LINK_STATE_DORMANT,
278 };
279
280
281 /*
282 * This structure holds at boot time configured netdevice settings. They
283 * are then used in the device probing.
284 */
285 struct netdev_boot_setup {
286 char name[IFNAMSIZ];
287 struct ifmap map;
288 };
289 #define NETDEV_BOOT_SETUP_MAX 8
290
291 int __init netdev_boot_setup(char *str);
292
293 /*
294 * Structure for NAPI scheduling similar to tasklet but with weighting
295 */
296 struct napi_struct {
297 /* The poll_list must only be managed by the entity which
298 * changes the state of the NAPI_STATE_SCHED bit. This means
299 * whoever atomically sets that bit can add this napi_struct
300 * to the per-cpu poll_list, and whoever clears that bit
301 * can remove from the list right before clearing the bit.
302 */
303 struct list_head poll_list;
304
305 unsigned long state;
306 int weight;
307 unsigned int gro_count;
308 int (*poll)(struct napi_struct *, int);
309 #ifdef CONFIG_NETPOLL
310 spinlock_t poll_lock;
311 int poll_owner;
312 #endif
313 struct net_device *dev;
314 struct sk_buff *gro_list;
315 struct sk_buff *skb;
316 struct list_head dev_list;
317 struct hlist_node napi_hash_node;
318 unsigned int napi_id;
319 };
320
321 enum {
322 NAPI_STATE_SCHED, /* Poll is scheduled */
323 NAPI_STATE_DISABLE, /* Disable pending */
324 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
325 NAPI_STATE_HASHED, /* In NAPI hash */
326 };
327
328 enum gro_result {
329 GRO_MERGED,
330 GRO_MERGED_FREE,
331 GRO_HELD,
332 GRO_NORMAL,
333 GRO_DROP,
334 };
335 typedef enum gro_result gro_result_t;
336
337 /*
338 * enum rx_handler_result - Possible return values for rx_handlers.
339 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
340 * further.
341 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
342 * case skb->dev was changed by rx_handler.
343 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
344 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
345 *
346 * rx_handlers are functions called from inside __netif_receive_skb(), to do
347 * special processing of the skb, prior to delivery to protocol handlers.
348 *
349 * Currently, a net_device can only have a single rx_handler registered. Trying
350 * to register a second rx_handler will return -EBUSY.
351 *
352 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
353 * To unregister a rx_handler on a net_device, use
354 * netdev_rx_handler_unregister().
355 *
356 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
357 * do with the skb.
358 *
359 * If the rx_handler consumed to skb in some way, it should return
360 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
361 * the skb to be delivered in some other ways.
362 *
363 * If the rx_handler changed skb->dev, to divert the skb to another
364 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
365 * new device will be called if it exists.
366 *
367 * If the rx_handler consider the skb should be ignored, it should return
368 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
369 * are registered on exact device (ptype->dev == skb->dev).
370 *
371 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
372 * delivered, it should return RX_HANDLER_PASS.
373 *
374 * A device without a registered rx_handler will behave as if rx_handler
375 * returned RX_HANDLER_PASS.
376 */
377
378 enum rx_handler_result {
379 RX_HANDLER_CONSUMED,
380 RX_HANDLER_ANOTHER,
381 RX_HANDLER_EXACT,
382 RX_HANDLER_PASS,
383 };
384 typedef enum rx_handler_result rx_handler_result_t;
385 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
386
387 void __napi_schedule(struct napi_struct *n);
388
389 static inline bool napi_disable_pending(struct napi_struct *n)
390 {
391 return test_bit(NAPI_STATE_DISABLE, &n->state);
392 }
393
394 /**
395 * napi_schedule_prep - check if napi can be scheduled
396 * @n: napi context
397 *
398 * Test if NAPI routine is already running, and if not mark
399 * it as running. This is used as a condition variable
400 * insure only one NAPI poll instance runs. We also make
401 * sure there is no pending NAPI disable.
402 */
403 static inline bool napi_schedule_prep(struct napi_struct *n)
404 {
405 return !napi_disable_pending(n) &&
406 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
407 }
408
409 /**
410 * napi_schedule - schedule NAPI poll
411 * @n: napi context
412 *
413 * Schedule NAPI poll routine to be called if it is not already
414 * running.
415 */
416 static inline void napi_schedule(struct napi_struct *n)
417 {
418 if (napi_schedule_prep(n))
419 __napi_schedule(n);
420 }
421
422 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
423 static inline bool napi_reschedule(struct napi_struct *napi)
424 {
425 if (napi_schedule_prep(napi)) {
426 __napi_schedule(napi);
427 return true;
428 }
429 return false;
430 }
431
432 /**
433 * napi_complete - NAPI processing complete
434 * @n: napi context
435 *
436 * Mark NAPI processing as complete.
437 */
438 void __napi_complete(struct napi_struct *n);
439 void napi_complete(struct napi_struct *n);
440
441 /**
442 * napi_by_id - lookup a NAPI by napi_id
443 * @napi_id: hashed napi_id
444 *
445 * lookup @napi_id in napi_hash table
446 * must be called under rcu_read_lock()
447 */
448 struct napi_struct *napi_by_id(unsigned int napi_id);
449
450 /**
451 * napi_hash_add - add a NAPI to global hashtable
452 * @napi: napi context
453 *
454 * generate a new napi_id and store a @napi under it in napi_hash
455 */
456 void napi_hash_add(struct napi_struct *napi);
457
458 /**
459 * napi_hash_del - remove a NAPI from global table
460 * @napi: napi context
461 *
462 * Warning: caller must observe rcu grace period
463 * before freeing memory containing @napi
464 */
465 void napi_hash_del(struct napi_struct *napi);
466
467 /**
468 * napi_disable - prevent NAPI from scheduling
469 * @n: napi context
470 *
471 * Stop NAPI from being scheduled on this context.
472 * Waits till any outstanding processing completes.
473 */
474 static inline void napi_disable(struct napi_struct *n)
475 {
476 might_sleep();
477 set_bit(NAPI_STATE_DISABLE, &n->state);
478 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
479 msleep(1);
480 clear_bit(NAPI_STATE_DISABLE, &n->state);
481 }
482
483 /**
484 * napi_enable - enable NAPI scheduling
485 * @n: napi context
486 *
487 * Resume NAPI from being scheduled on this context.
488 * Must be paired with napi_disable.
489 */
490 static inline void napi_enable(struct napi_struct *n)
491 {
492 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
493 smp_mb__before_atomic();
494 clear_bit(NAPI_STATE_SCHED, &n->state);
495 }
496
497 #ifdef CONFIG_SMP
498 /**
499 * napi_synchronize - wait until NAPI is not running
500 * @n: napi context
501 *
502 * Wait until NAPI is done being scheduled on this context.
503 * Waits till any outstanding processing completes but
504 * does not disable future activations.
505 */
506 static inline void napi_synchronize(const struct napi_struct *n)
507 {
508 while (test_bit(NAPI_STATE_SCHED, &n->state))
509 msleep(1);
510 }
511 #else
512 # define napi_synchronize(n) barrier()
513 #endif
514
515 enum netdev_queue_state_t {
516 __QUEUE_STATE_DRV_XOFF,
517 __QUEUE_STATE_STACK_XOFF,
518 __QUEUE_STATE_FROZEN,
519 };
520
521 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
522 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
523 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
524
525 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
526 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
527 QUEUE_STATE_FROZEN)
528 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
529 QUEUE_STATE_FROZEN)
530
531 /*
532 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
533 * netif_tx_* functions below are used to manipulate this flag. The
534 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
535 * queue independently. The netif_xmit_*stopped functions below are called
536 * to check if the queue has been stopped by the driver or stack (either
537 * of the XOFF bits are set in the state). Drivers should not need to call
538 * netif_xmit*stopped functions, they should only be using netif_tx_*.
539 */
540
541 struct netdev_queue {
542 /*
543 * read mostly part
544 */
545 struct net_device *dev;
546 struct Qdisc *qdisc;
547 struct Qdisc *qdisc_sleeping;
548 #ifdef CONFIG_SYSFS
549 struct kobject kobj;
550 #endif
551 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
552 int numa_node;
553 #endif
554 /*
555 * write mostly part
556 */
557 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
558 int xmit_lock_owner;
559 /*
560 * please use this field instead of dev->trans_start
561 */
562 unsigned long trans_start;
563
564 /*
565 * Number of TX timeouts for this queue
566 * (/sys/class/net/DEV/Q/trans_timeout)
567 */
568 unsigned long trans_timeout;
569
570 unsigned long state;
571
572 #ifdef CONFIG_BQL
573 struct dql dql;
574 #endif
575 } ____cacheline_aligned_in_smp;
576
577 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
578 {
579 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
580 return q->numa_node;
581 #else
582 return NUMA_NO_NODE;
583 #endif
584 }
585
586 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
587 {
588 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
589 q->numa_node = node;
590 #endif
591 }
592
593 #ifdef CONFIG_RPS
594 /*
595 * This structure holds an RPS map which can be of variable length. The
596 * map is an array of CPUs.
597 */
598 struct rps_map {
599 unsigned int len;
600 struct rcu_head rcu;
601 u16 cpus[0];
602 };
603 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
604
605 /*
606 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
607 * tail pointer for that CPU's input queue at the time of last enqueue, and
608 * a hardware filter index.
609 */
610 struct rps_dev_flow {
611 u16 cpu;
612 u16 filter;
613 unsigned int last_qtail;
614 };
615 #define RPS_NO_FILTER 0xffff
616
617 /*
618 * The rps_dev_flow_table structure contains a table of flow mappings.
619 */
620 struct rps_dev_flow_table {
621 unsigned int mask;
622 struct rcu_head rcu;
623 struct rps_dev_flow flows[0];
624 };
625 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
626 ((_num) * sizeof(struct rps_dev_flow)))
627
628 /*
629 * The rps_sock_flow_table contains mappings of flows to the last CPU
630 * on which they were processed by the application (set in recvmsg).
631 */
632 struct rps_sock_flow_table {
633 unsigned int mask;
634 u16 ents[0];
635 };
636 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_sock_flow_table) + \
637 ((_num) * sizeof(u16)))
638
639 #define RPS_NO_CPU 0xffff
640
641 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
642 u32 hash)
643 {
644 if (table && hash) {
645 unsigned int cpu, index = hash & table->mask;
646
647 /* We only give a hint, preemption can change cpu under us */
648 cpu = raw_smp_processor_id();
649
650 if (table->ents[index] != cpu)
651 table->ents[index] = cpu;
652 }
653 }
654
655 static inline void rps_reset_sock_flow(struct rps_sock_flow_table *table,
656 u32 hash)
657 {
658 if (table && hash)
659 table->ents[hash & table->mask] = RPS_NO_CPU;
660 }
661
662 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
663
664 #ifdef CONFIG_RFS_ACCEL
665 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
666 u16 filter_id);
667 #endif
668 #endif /* CONFIG_RPS */
669
670 /* This structure contains an instance of an RX queue. */
671 struct netdev_rx_queue {
672 #ifdef CONFIG_RPS
673 struct rps_map __rcu *rps_map;
674 struct rps_dev_flow_table __rcu *rps_flow_table;
675 #endif
676 struct kobject kobj;
677 struct net_device *dev;
678 } ____cacheline_aligned_in_smp;
679
680 /*
681 * RX queue sysfs structures and functions.
682 */
683 struct rx_queue_attribute {
684 struct attribute attr;
685 ssize_t (*show)(struct netdev_rx_queue *queue,
686 struct rx_queue_attribute *attr, char *buf);
687 ssize_t (*store)(struct netdev_rx_queue *queue,
688 struct rx_queue_attribute *attr, const char *buf, size_t len);
689 };
690
691 #ifdef CONFIG_XPS
692 /*
693 * This structure holds an XPS map which can be of variable length. The
694 * map is an array of queues.
695 */
696 struct xps_map {
697 unsigned int len;
698 unsigned int alloc_len;
699 struct rcu_head rcu;
700 u16 queues[0];
701 };
702 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
703 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
704 / sizeof(u16))
705
706 /*
707 * This structure holds all XPS maps for device. Maps are indexed by CPU.
708 */
709 struct xps_dev_maps {
710 struct rcu_head rcu;
711 struct xps_map __rcu *cpu_map[0];
712 };
713 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
714 (nr_cpu_ids * sizeof(struct xps_map *)))
715 #endif /* CONFIG_XPS */
716
717 #define TC_MAX_QUEUE 16
718 #define TC_BITMASK 15
719 /* HW offloaded queuing disciplines txq count and offset maps */
720 struct netdev_tc_txq {
721 u16 count;
722 u16 offset;
723 };
724
725 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
726 /*
727 * This structure is to hold information about the device
728 * configured to run FCoE protocol stack.
729 */
730 struct netdev_fcoe_hbainfo {
731 char manufacturer[64];
732 char serial_number[64];
733 char hardware_version[64];
734 char driver_version[64];
735 char optionrom_version[64];
736 char firmware_version[64];
737 char model[256];
738 char model_description[256];
739 };
740 #endif
741
742 #define MAX_PHYS_PORT_ID_LEN 32
743
744 /* This structure holds a unique identifier to identify the
745 * physical port used by a netdevice.
746 */
747 struct netdev_phys_port_id {
748 unsigned char id[MAX_PHYS_PORT_ID_LEN];
749 unsigned char id_len;
750 };
751
752 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
753 struct sk_buff *skb);
754
755 /*
756 * This structure defines the management hooks for network devices.
757 * The following hooks can be defined; unless noted otherwise, they are
758 * optional and can be filled with a null pointer.
759 *
760 * int (*ndo_init)(struct net_device *dev);
761 * This function is called once when network device is registered.
762 * The network device can use this to any late stage initializaton
763 * or semantic validattion. It can fail with an error code which will
764 * be propogated back to register_netdev
765 *
766 * void (*ndo_uninit)(struct net_device *dev);
767 * This function is called when device is unregistered or when registration
768 * fails. It is not called if init fails.
769 *
770 * int (*ndo_open)(struct net_device *dev);
771 * This function is called when network device transistions to the up
772 * state.
773 *
774 * int (*ndo_stop)(struct net_device *dev);
775 * This function is called when network device transistions to the down
776 * state.
777 *
778 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
779 * struct net_device *dev);
780 * Called when a packet needs to be transmitted.
781 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
782 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
783 * Required can not be NULL.
784 *
785 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
786 * void *accel_priv, select_queue_fallback_t fallback);
787 * Called to decide which queue to when device supports multiple
788 * transmit queues.
789 *
790 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
791 * This function is called to allow device receiver to make
792 * changes to configuration when multicast or promiscious is enabled.
793 *
794 * void (*ndo_set_rx_mode)(struct net_device *dev);
795 * This function is called device changes address list filtering.
796 * If driver handles unicast address filtering, it should set
797 * IFF_UNICAST_FLT to its priv_flags.
798 *
799 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
800 * This function is called when the Media Access Control address
801 * needs to be changed. If this interface is not defined, the
802 * mac address can not be changed.
803 *
804 * int (*ndo_validate_addr)(struct net_device *dev);
805 * Test if Media Access Control address is valid for the device.
806 *
807 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
808 * Called when a user request an ioctl which can't be handled by
809 * the generic interface code. If not defined ioctl's return
810 * not supported error code.
811 *
812 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
813 * Used to set network devices bus interface parameters. This interface
814 * is retained for legacy reason, new devices should use the bus
815 * interface (PCI) for low level management.
816 *
817 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
818 * Called when a user wants to change the Maximum Transfer Unit
819 * of a device. If not defined, any request to change MTU will
820 * will return an error.
821 *
822 * void (*ndo_tx_timeout)(struct net_device *dev);
823 * Callback uses when the transmitter has not made any progress
824 * for dev->watchdog ticks.
825 *
826 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
827 * struct rtnl_link_stats64 *storage);
828 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
829 * Called when a user wants to get the network device usage
830 * statistics. Drivers must do one of the following:
831 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
832 * rtnl_link_stats64 structure passed by the caller.
833 * 2. Define @ndo_get_stats to update a net_device_stats structure
834 * (which should normally be dev->stats) and return a pointer to
835 * it. The structure may be changed asynchronously only if each
836 * field is written atomically.
837 * 3. Update dev->stats asynchronously and atomically, and define
838 * neither operation.
839 *
840 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16t vid);
841 * If device support VLAN filtering this function is called when a
842 * VLAN id is registered.
843 *
844 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, unsigned short vid);
845 * If device support VLAN filtering this function is called when a
846 * VLAN id is unregistered.
847 *
848 * void (*ndo_poll_controller)(struct net_device *dev);
849 *
850 * SR-IOV management functions.
851 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
852 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
853 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
854 * int max_tx_rate);
855 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
856 * int (*ndo_get_vf_config)(struct net_device *dev,
857 * int vf, struct ifla_vf_info *ivf);
858 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
859 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
860 * struct nlattr *port[]);
861 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
862 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
863 * Called to setup 'tc' number of traffic classes in the net device. This
864 * is always called from the stack with the rtnl lock held and netif tx
865 * queues stopped. This allows the netdevice to perform queue management
866 * safely.
867 *
868 * Fiber Channel over Ethernet (FCoE) offload functions.
869 * int (*ndo_fcoe_enable)(struct net_device *dev);
870 * Called when the FCoE protocol stack wants to start using LLD for FCoE
871 * so the underlying device can perform whatever needed configuration or
872 * initialization to support acceleration of FCoE traffic.
873 *
874 * int (*ndo_fcoe_disable)(struct net_device *dev);
875 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
876 * so the underlying device can perform whatever needed clean-ups to
877 * stop supporting acceleration of FCoE traffic.
878 *
879 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
880 * struct scatterlist *sgl, unsigned int sgc);
881 * Called when the FCoE Initiator wants to initialize an I/O that
882 * is a possible candidate for Direct Data Placement (DDP). The LLD can
883 * perform necessary setup and returns 1 to indicate the device is set up
884 * successfully to perform DDP on this I/O, otherwise this returns 0.
885 *
886 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
887 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
888 * indicated by the FC exchange id 'xid', so the underlying device can
889 * clean up and reuse resources for later DDP requests.
890 *
891 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
892 * struct scatterlist *sgl, unsigned int sgc);
893 * Called when the FCoE Target wants to initialize an I/O that
894 * is a possible candidate for Direct Data Placement (DDP). The LLD can
895 * perform necessary setup and returns 1 to indicate the device is set up
896 * successfully to perform DDP on this I/O, otherwise this returns 0.
897 *
898 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
899 * struct netdev_fcoe_hbainfo *hbainfo);
900 * Called when the FCoE Protocol stack wants information on the underlying
901 * device. This information is utilized by the FCoE protocol stack to
902 * register attributes with Fiber Channel management service as per the
903 * FC-GS Fabric Device Management Information(FDMI) specification.
904 *
905 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
906 * Called when the underlying device wants to override default World Wide
907 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
908 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
909 * protocol stack to use.
910 *
911 * RFS acceleration.
912 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
913 * u16 rxq_index, u32 flow_id);
914 * Set hardware filter for RFS. rxq_index is the target queue index;
915 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
916 * Return the filter ID on success, or a negative error code.
917 *
918 * Slave management functions (for bridge, bonding, etc).
919 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
920 * Called to make another netdev an underling.
921 *
922 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
923 * Called to release previously enslaved netdev.
924 *
925 * Feature/offload setting functions.
926 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
927 * netdev_features_t features);
928 * Adjusts the requested feature flags according to device-specific
929 * constraints, and returns the resulting flags. Must not modify
930 * the device state.
931 *
932 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
933 * Called to update device configuration to new features. Passed
934 * feature set might be less than what was returned by ndo_fix_features()).
935 * Must return >0 or -errno if it changed dev->features itself.
936 *
937 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
938 * struct net_device *dev,
939 * const unsigned char *addr, u16 flags)
940 * Adds an FDB entry to dev for addr.
941 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
942 * struct net_device *dev,
943 * const unsigned char *addr)
944 * Deletes the FDB entry from dev coresponding to addr.
945 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
946 * struct net_device *dev, struct net_device *filter_dev,
947 * int idx)
948 * Used to add FDB entries to dump requests. Implementers should add
949 * entries to skb and update idx with the number of entries.
950 *
951 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
952 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
953 * struct net_device *dev, u32 filter_mask)
954 *
955 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
956 * Called to change device carrier. Soft-devices (like dummy, team, etc)
957 * which do not represent real hardware may define this to allow their
958 * userspace components to manage their virtual carrier state. Devices
959 * that determine carrier state from physical hardware properties (eg
960 * network cables) or protocol-dependent mechanisms (eg
961 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
962 *
963 * int (*ndo_get_phys_port_id)(struct net_device *dev,
964 * struct netdev_phys_port_id *ppid);
965 * Called to get ID of physical port of this device. If driver does
966 * not implement this, it is assumed that the hw is not able to have
967 * multiple net devices on single physical port.
968 *
969 * void (*ndo_add_vxlan_port)(struct net_device *dev,
970 * sa_family_t sa_family, __be16 port);
971 * Called by vxlan to notiy a driver about the UDP port and socket
972 * address family that vxlan is listnening to. It is called only when
973 * a new port starts listening. The operation is protected by the
974 * vxlan_net->sock_lock.
975 *
976 * void (*ndo_del_vxlan_port)(struct net_device *dev,
977 * sa_family_t sa_family, __be16 port);
978 * Called by vxlan to notify the driver about a UDP port and socket
979 * address family that vxlan is not listening to anymore. The operation
980 * is protected by the vxlan_net->sock_lock.
981 *
982 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
983 * struct net_device *dev)
984 * Called by upper layer devices to accelerate switching or other
985 * station functionality into hardware. 'pdev is the lowerdev
986 * to use for the offload and 'dev' is the net device that will
987 * back the offload. Returns a pointer to the private structure
988 * the upper layer will maintain.
989 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
990 * Called by upper layer device to delete the station created
991 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
992 * the station and priv is the structure returned by the add
993 * operation.
994 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
995 * struct net_device *dev,
996 * void *priv);
997 * Callback to use for xmit over the accelerated station. This
998 * is used in place of ndo_start_xmit on accelerated net
999 * devices.
1000 */
1001 struct net_device_ops {
1002 int (*ndo_init)(struct net_device *dev);
1003 void (*ndo_uninit)(struct net_device *dev);
1004 int (*ndo_open)(struct net_device *dev);
1005 int (*ndo_stop)(struct net_device *dev);
1006 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1007 struct net_device *dev);
1008 u16 (*ndo_select_queue)(struct net_device *dev,
1009 struct sk_buff *skb,
1010 void *accel_priv,
1011 select_queue_fallback_t fallback);
1012 void (*ndo_change_rx_flags)(struct net_device *dev,
1013 int flags);
1014 void (*ndo_set_rx_mode)(struct net_device *dev);
1015 int (*ndo_set_mac_address)(struct net_device *dev,
1016 void *addr);
1017 int (*ndo_validate_addr)(struct net_device *dev);
1018 int (*ndo_do_ioctl)(struct net_device *dev,
1019 struct ifreq *ifr, int cmd);
1020 int (*ndo_set_config)(struct net_device *dev,
1021 struct ifmap *map);
1022 int (*ndo_change_mtu)(struct net_device *dev,
1023 int new_mtu);
1024 int (*ndo_neigh_setup)(struct net_device *dev,
1025 struct neigh_parms *);
1026 void (*ndo_tx_timeout) (struct net_device *dev);
1027
1028 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1029 struct rtnl_link_stats64 *storage);
1030 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1031
1032 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1033 __be16 proto, u16 vid);
1034 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1035 __be16 proto, u16 vid);
1036 #ifdef CONFIG_NET_POLL_CONTROLLER
1037 void (*ndo_poll_controller)(struct net_device *dev);
1038 int (*ndo_netpoll_setup)(struct net_device *dev,
1039 struct netpoll_info *info);
1040 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1041 #endif
1042 #ifdef CONFIG_NET_RX_BUSY_POLL
1043 int (*ndo_busy_poll)(struct napi_struct *dev);
1044 #endif
1045 int (*ndo_set_vf_mac)(struct net_device *dev,
1046 int queue, u8 *mac);
1047 int (*ndo_set_vf_vlan)(struct net_device *dev,
1048 int queue, u16 vlan, u8 qos);
1049 int (*ndo_set_vf_rate)(struct net_device *dev,
1050 int vf, int min_tx_rate,
1051 int max_tx_rate);
1052 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1053 int vf, bool setting);
1054 int (*ndo_get_vf_config)(struct net_device *dev,
1055 int vf,
1056 struct ifla_vf_info *ivf);
1057 int (*ndo_set_vf_link_state)(struct net_device *dev,
1058 int vf, int link_state);
1059 int (*ndo_set_vf_port)(struct net_device *dev,
1060 int vf,
1061 struct nlattr *port[]);
1062 int (*ndo_get_vf_port)(struct net_device *dev,
1063 int vf, struct sk_buff *skb);
1064 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1065 #if IS_ENABLED(CONFIG_FCOE)
1066 int (*ndo_fcoe_enable)(struct net_device *dev);
1067 int (*ndo_fcoe_disable)(struct net_device *dev);
1068 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1069 u16 xid,
1070 struct scatterlist *sgl,
1071 unsigned int sgc);
1072 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1073 u16 xid);
1074 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1075 u16 xid,
1076 struct scatterlist *sgl,
1077 unsigned int sgc);
1078 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1079 struct netdev_fcoe_hbainfo *hbainfo);
1080 #endif
1081
1082 #if IS_ENABLED(CONFIG_LIBFCOE)
1083 #define NETDEV_FCOE_WWNN 0
1084 #define NETDEV_FCOE_WWPN 1
1085 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1086 u64 *wwn, int type);
1087 #endif
1088
1089 #ifdef CONFIG_RFS_ACCEL
1090 int (*ndo_rx_flow_steer)(struct net_device *dev,
1091 const struct sk_buff *skb,
1092 u16 rxq_index,
1093 u32 flow_id);
1094 #endif
1095 int (*ndo_add_slave)(struct net_device *dev,
1096 struct net_device *slave_dev);
1097 int (*ndo_del_slave)(struct net_device *dev,
1098 struct net_device *slave_dev);
1099 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1100 netdev_features_t features);
1101 int (*ndo_set_features)(struct net_device *dev,
1102 netdev_features_t features);
1103 int (*ndo_neigh_construct)(struct neighbour *n);
1104 void (*ndo_neigh_destroy)(struct neighbour *n);
1105
1106 int (*ndo_fdb_add)(struct ndmsg *ndm,
1107 struct nlattr *tb[],
1108 struct net_device *dev,
1109 const unsigned char *addr,
1110 u16 flags);
1111 int (*ndo_fdb_del)(struct ndmsg *ndm,
1112 struct nlattr *tb[],
1113 struct net_device *dev,
1114 const unsigned char *addr);
1115 int (*ndo_fdb_dump)(struct sk_buff *skb,
1116 struct netlink_callback *cb,
1117 struct net_device *dev,
1118 struct net_device *filter_dev,
1119 int idx);
1120
1121 int (*ndo_bridge_setlink)(struct net_device *dev,
1122 struct nlmsghdr *nlh);
1123 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1124 u32 pid, u32 seq,
1125 struct net_device *dev,
1126 u32 filter_mask);
1127 int (*ndo_bridge_dellink)(struct net_device *dev,
1128 struct nlmsghdr *nlh);
1129 int (*ndo_change_carrier)(struct net_device *dev,
1130 bool new_carrier);
1131 int (*ndo_get_phys_port_id)(struct net_device *dev,
1132 struct netdev_phys_port_id *ppid);
1133 void (*ndo_add_vxlan_port)(struct net_device *dev,
1134 sa_family_t sa_family,
1135 __be16 port);
1136 void (*ndo_del_vxlan_port)(struct net_device *dev,
1137 sa_family_t sa_family,
1138 __be16 port);
1139
1140 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1141 struct net_device *dev);
1142 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1143 void *priv);
1144
1145 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1146 struct net_device *dev,
1147 void *priv);
1148 int (*ndo_get_lock_subclass)(struct net_device *dev);
1149 };
1150
1151 /**
1152 * enum net_device_priv_flags - &struct net_device priv_flags
1153 *
1154 * These are the &struct net_device, they are only set internally
1155 * by drivers and used in the kernel. These flags are invisible to
1156 * userspace, this means that the order of these flags can change
1157 * during any kernel release.
1158 *
1159 * You should have a pretty good reason to be extending these flags.
1160 *
1161 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1162 * @IFF_EBRIDGE: Ethernet bridging device
1163 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1164 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1165 * @IFF_MASTER_ALB: bonding master, balance-alb
1166 * @IFF_BONDING: bonding master or slave
1167 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1168 * @IFF_ISATAP: ISATAP interface (RFC4214)
1169 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1170 * @IFF_WAN_HDLC: WAN HDLC device
1171 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1172 * release skb->dst
1173 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1174 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1175 * @IFF_MACVLAN_PORT: device used as macvlan port
1176 * @IFF_BRIDGE_PORT: device used as bridge port
1177 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1178 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1179 * @IFF_UNICAST_FLT: Supports unicast filtering
1180 * @IFF_TEAM_PORT: device used as team port
1181 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1182 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1183 * change when it's running
1184 * @IFF_MACVLAN: Macvlan device
1185 */
1186 enum netdev_priv_flags {
1187 IFF_802_1Q_VLAN = 1<<0,
1188 IFF_EBRIDGE = 1<<1,
1189 IFF_SLAVE_INACTIVE = 1<<2,
1190 IFF_MASTER_8023AD = 1<<3,
1191 IFF_MASTER_ALB = 1<<4,
1192 IFF_BONDING = 1<<5,
1193 IFF_SLAVE_NEEDARP = 1<<6,
1194 IFF_ISATAP = 1<<7,
1195 IFF_MASTER_ARPMON = 1<<8,
1196 IFF_WAN_HDLC = 1<<9,
1197 IFF_XMIT_DST_RELEASE = 1<<10,
1198 IFF_DONT_BRIDGE = 1<<11,
1199 IFF_DISABLE_NETPOLL = 1<<12,
1200 IFF_MACVLAN_PORT = 1<<13,
1201 IFF_BRIDGE_PORT = 1<<14,
1202 IFF_OVS_DATAPATH = 1<<15,
1203 IFF_TX_SKB_SHARING = 1<<16,
1204 IFF_UNICAST_FLT = 1<<17,
1205 IFF_TEAM_PORT = 1<<18,
1206 IFF_SUPP_NOFCS = 1<<19,
1207 IFF_LIVE_ADDR_CHANGE = 1<<20,
1208 IFF_MACVLAN = 1<<21,
1209 };
1210
1211 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1212 #define IFF_EBRIDGE IFF_EBRIDGE
1213 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1214 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1215 #define IFF_MASTER_ALB IFF_MASTER_ALB
1216 #define IFF_BONDING IFF_BONDING
1217 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1218 #define IFF_ISATAP IFF_ISATAP
1219 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1220 #define IFF_WAN_HDLC IFF_WAN_HDLC
1221 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1222 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1223 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1224 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1225 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1226 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1227 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1228 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1229 #define IFF_TEAM_PORT IFF_TEAM_PORT
1230 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1231 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1232 #define IFF_MACVLAN IFF_MACVLAN
1233
1234 /*
1235 * The DEVICE structure.
1236 * Actually, this whole structure is a big mistake. It mixes I/O
1237 * data with strictly "high-level" data, and it has to know about
1238 * almost every data structure used in the INET module.
1239 *
1240 * FIXME: cleanup struct net_device such that network protocol info
1241 * moves out.
1242 */
1243
1244 struct net_device {
1245
1246 /*
1247 * This is the first field of the "visible" part of this structure
1248 * (i.e. as seen by users in the "Space.c" file). It is the name
1249 * of the interface.
1250 */
1251 char name[IFNAMSIZ];
1252
1253 /* device name hash chain, please keep it close to name[] */
1254 struct hlist_node name_hlist;
1255
1256 /* snmp alias */
1257 char *ifalias;
1258
1259 /*
1260 * I/O specific fields
1261 * FIXME: Merge these and struct ifmap into one
1262 */
1263 unsigned long mem_end; /* shared mem end */
1264 unsigned long mem_start; /* shared mem start */
1265 unsigned long base_addr; /* device I/O address */
1266 int irq; /* device IRQ number */
1267
1268 /*
1269 * Some hardware also needs these fields, but they are not
1270 * part of the usual set specified in Space.c.
1271 */
1272
1273 unsigned long state;
1274
1275 struct list_head dev_list;
1276 struct list_head napi_list;
1277 struct list_head unreg_list;
1278 struct list_head close_list;
1279
1280 /* directly linked devices, like slaves for bonding */
1281 struct {
1282 struct list_head upper;
1283 struct list_head lower;
1284 } adj_list;
1285
1286 /* all linked devices, *including* neighbours */
1287 struct {
1288 struct list_head upper;
1289 struct list_head lower;
1290 } all_adj_list;
1291
1292
1293 /* currently active device features */
1294 netdev_features_t features;
1295 /* user-changeable features */
1296 netdev_features_t hw_features;
1297 /* user-requested features */
1298 netdev_features_t wanted_features;
1299 /* mask of features inheritable by VLAN devices */
1300 netdev_features_t vlan_features;
1301 /* mask of features inherited by encapsulating devices
1302 * This field indicates what encapsulation offloads
1303 * the hardware is capable of doing, and drivers will
1304 * need to set them appropriately.
1305 */
1306 netdev_features_t hw_enc_features;
1307 /* mask of fetures inheritable by MPLS */
1308 netdev_features_t mpls_features;
1309
1310 /* Interface index. Unique device identifier */
1311 int ifindex;
1312 int iflink;
1313
1314 struct net_device_stats stats;
1315
1316 /* dropped packets by core network, Do not use this in drivers */
1317 atomic_long_t rx_dropped;
1318 atomic_long_t tx_dropped;
1319
1320 /* Stats to monitor carrier on<->off transitions */
1321 atomic_t carrier_changes;
1322
1323 #ifdef CONFIG_WIRELESS_EXT
1324 /* List of functions to handle Wireless Extensions (instead of ioctl).
1325 * See <net/iw_handler.h> for details. Jean II */
1326 const struct iw_handler_def * wireless_handlers;
1327 /* Instance data managed by the core of Wireless Extensions. */
1328 struct iw_public_data * wireless_data;
1329 #endif
1330 /* Management operations */
1331 const struct net_device_ops *netdev_ops;
1332 const struct ethtool_ops *ethtool_ops;
1333 const struct forwarding_accel_ops *fwd_ops;
1334
1335 /* Hardware header description */
1336 const struct header_ops *header_ops;
1337
1338 unsigned int flags; /* interface flags (a la BSD) */
1339 unsigned int priv_flags; /* Like 'flags' but invisible to userspace.
1340 * See if.h for definitions. */
1341 unsigned short gflags;
1342 unsigned short padded; /* How much padding added by alloc_netdev() */
1343
1344 unsigned char operstate; /* RFC2863 operstate */
1345 unsigned char link_mode; /* mapping policy to operstate */
1346
1347 unsigned char if_port; /* Selectable AUI, TP,..*/
1348 unsigned char dma; /* DMA channel */
1349
1350 unsigned int mtu; /* interface MTU value */
1351 unsigned short type; /* interface hardware type */
1352 unsigned short hard_header_len; /* hardware hdr length */
1353
1354 /* extra head- and tailroom the hardware may need, but not in all cases
1355 * can this be guaranteed, especially tailroom. Some cases also use
1356 * LL_MAX_HEADER instead to allocate the skb.
1357 */
1358 unsigned short needed_headroom;
1359 unsigned short needed_tailroom;
1360
1361 /* Interface address info. */
1362 unsigned char perm_addr[MAX_ADDR_LEN]; /* permanent hw address */
1363 unsigned char addr_assign_type; /* hw address assignment type */
1364 unsigned char addr_len; /* hardware address length */
1365 unsigned short neigh_priv_len;
1366 unsigned short dev_id; /* Used to differentiate devices
1367 * that share the same link
1368 * layer address
1369 */
1370 unsigned short dev_port; /* Used to differentiate
1371 * devices that share the same
1372 * function
1373 */
1374 spinlock_t addr_list_lock;
1375 struct netdev_hw_addr_list uc; /* Unicast mac addresses */
1376 struct netdev_hw_addr_list mc; /* Multicast mac addresses */
1377 struct netdev_hw_addr_list dev_addrs; /* list of device
1378 * hw addresses
1379 */
1380 #ifdef CONFIG_SYSFS
1381 struct kset *queues_kset;
1382 #endif
1383
1384 bool uc_promisc;
1385 unsigned int promiscuity;
1386 unsigned int allmulti;
1387
1388
1389 /* Protocol specific pointers */
1390
1391 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1392 struct vlan_info __rcu *vlan_info; /* VLAN info */
1393 #endif
1394 #if IS_ENABLED(CONFIG_NET_DSA)
1395 struct dsa_switch_tree *dsa_ptr; /* dsa specific data */
1396 #endif
1397 #if IS_ENABLED(CONFIG_TIPC)
1398 struct tipc_bearer __rcu *tipc_ptr; /* TIPC specific data */
1399 #endif
1400 void *atalk_ptr; /* AppleTalk link */
1401 struct in_device __rcu *ip_ptr; /* IPv4 specific data */
1402 struct dn_dev __rcu *dn_ptr; /* DECnet specific data */
1403 struct inet6_dev __rcu *ip6_ptr; /* IPv6 specific data */
1404 void *ax25_ptr; /* AX.25 specific data */
1405 struct wireless_dev *ieee80211_ptr; /* IEEE 802.11 specific data,
1406 assign before registering */
1407
1408 /*
1409 * Cache lines mostly used on receive path (including eth_type_trans())
1410 */
1411 unsigned long last_rx; /* Time of last Rx */
1412
1413 /* Interface address info used in eth_type_trans() */
1414 unsigned char *dev_addr; /* hw address, (before bcast
1415 because most packets are
1416 unicast) */
1417
1418
1419 #ifdef CONFIG_SYSFS
1420 struct netdev_rx_queue *_rx;
1421
1422 /* Number of RX queues allocated at register_netdev() time */
1423 unsigned int num_rx_queues;
1424
1425 /* Number of RX queues currently active in device */
1426 unsigned int real_num_rx_queues;
1427
1428 #endif
1429
1430 rx_handler_func_t __rcu *rx_handler;
1431 void __rcu *rx_handler_data;
1432
1433 struct netdev_queue __rcu *ingress_queue;
1434 unsigned char broadcast[MAX_ADDR_LEN]; /* hw bcast add */
1435
1436
1437 /*
1438 * Cache lines mostly used on transmit path
1439 */
1440 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1441
1442 /* Number of TX queues allocated at alloc_netdev_mq() time */
1443 unsigned int num_tx_queues;
1444
1445 /* Number of TX queues currently active in device */
1446 unsigned int real_num_tx_queues;
1447
1448 /* root qdisc from userspace point of view */
1449 struct Qdisc *qdisc;
1450
1451 unsigned long tx_queue_len; /* Max frames per queue allowed */
1452 spinlock_t tx_global_lock;
1453
1454 #ifdef CONFIG_XPS
1455 struct xps_dev_maps __rcu *xps_maps;
1456 #endif
1457 #ifdef CONFIG_RFS_ACCEL
1458 /* CPU reverse-mapping for RX completion interrupts, indexed
1459 * by RX queue number. Assigned by driver. This must only be
1460 * set if the ndo_rx_flow_steer operation is defined. */
1461 struct cpu_rmap *rx_cpu_rmap;
1462 #endif
1463
1464 /* These may be needed for future network-power-down code. */
1465
1466 /*
1467 * trans_start here is expensive for high speed devices on SMP,
1468 * please use netdev_queue->trans_start instead.
1469 */
1470 unsigned long trans_start; /* Time (in jiffies) of last Tx */
1471
1472 int watchdog_timeo; /* used by dev_watchdog() */
1473 struct timer_list watchdog_timer;
1474
1475 /* Number of references to this device */
1476 int __percpu *pcpu_refcnt;
1477
1478 /* delayed register/unregister */
1479 struct list_head todo_list;
1480 /* device index hash chain */
1481 struct hlist_node index_hlist;
1482
1483 struct list_head link_watch_list;
1484
1485 /* register/unregister state machine */
1486 enum { NETREG_UNINITIALIZED=0,
1487 NETREG_REGISTERED, /* completed register_netdevice */
1488 NETREG_UNREGISTERING, /* called unregister_netdevice */
1489 NETREG_UNREGISTERED, /* completed unregister todo */
1490 NETREG_RELEASED, /* called free_netdev */
1491 NETREG_DUMMY, /* dummy device for NAPI poll */
1492 } reg_state:8;
1493
1494 bool dismantle; /* device is going do be freed */
1495
1496 enum {
1497 RTNL_LINK_INITIALIZED,
1498 RTNL_LINK_INITIALIZING,
1499 } rtnl_link_state:16;
1500
1501 /* Called from unregister, can be used to call free_netdev */
1502 void (*destructor)(struct net_device *dev);
1503
1504 #ifdef CONFIG_NETPOLL
1505 struct netpoll_info __rcu *npinfo;
1506 #endif
1507
1508 #ifdef CONFIG_NET_NS
1509 /* Network namespace this network device is inside */
1510 struct net *nd_net;
1511 #endif
1512
1513 /* mid-layer private */
1514 union {
1515 void *ml_priv;
1516 struct pcpu_lstats __percpu *lstats; /* loopback stats */
1517 struct pcpu_sw_netstats __percpu *tstats;
1518 struct pcpu_dstats __percpu *dstats; /* dummy stats */
1519 struct pcpu_vstats __percpu *vstats; /* veth stats */
1520 };
1521 /* GARP */
1522 struct garp_port __rcu *garp_port;
1523 /* MRP */
1524 struct mrp_port __rcu *mrp_port;
1525
1526 /* class/net/name entry */
1527 struct device dev;
1528 /* space for optional device, statistics, and wireless sysfs groups */
1529 const struct attribute_group *sysfs_groups[4];
1530 /* space for optional per-rx queue attributes */
1531 const struct attribute_group *sysfs_rx_queue_group;
1532
1533 /* rtnetlink link ops */
1534 const struct rtnl_link_ops *rtnl_link_ops;
1535
1536 /* for setting kernel sock attribute on TCP connection setup */
1537 #define GSO_MAX_SIZE 65536
1538 unsigned int gso_max_size;
1539 #define GSO_MAX_SEGS 65535
1540 u16 gso_max_segs;
1541
1542 #ifdef CONFIG_DCB
1543 /* Data Center Bridging netlink ops */
1544 const struct dcbnl_rtnl_ops *dcbnl_ops;
1545 #endif
1546 u8 num_tc;
1547 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1548 u8 prio_tc_map[TC_BITMASK + 1];
1549
1550 #if IS_ENABLED(CONFIG_FCOE)
1551 /* max exchange id for FCoE LRO by ddp */
1552 unsigned int fcoe_ddp_xid;
1553 #endif
1554 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1555 struct netprio_map __rcu *priomap;
1556 #endif
1557 /* phy device may attach itself for hardware timestamping */
1558 struct phy_device *phydev;
1559
1560 struct lock_class_key *qdisc_tx_busylock;
1561
1562 /* group the device belongs to */
1563 int group;
1564
1565 struct pm_qos_request pm_qos_req;
1566 };
1567 #define to_net_dev(d) container_of(d, struct net_device, dev)
1568
1569 #define NETDEV_ALIGN 32
1570
1571 static inline
1572 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1573 {
1574 return dev->prio_tc_map[prio & TC_BITMASK];
1575 }
1576
1577 static inline
1578 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1579 {
1580 if (tc >= dev->num_tc)
1581 return -EINVAL;
1582
1583 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1584 return 0;
1585 }
1586
1587 static inline
1588 void netdev_reset_tc(struct net_device *dev)
1589 {
1590 dev->num_tc = 0;
1591 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1592 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1593 }
1594
1595 static inline
1596 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1597 {
1598 if (tc >= dev->num_tc)
1599 return -EINVAL;
1600
1601 dev->tc_to_txq[tc].count = count;
1602 dev->tc_to_txq[tc].offset = offset;
1603 return 0;
1604 }
1605
1606 static inline
1607 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1608 {
1609 if (num_tc > TC_MAX_QUEUE)
1610 return -EINVAL;
1611
1612 dev->num_tc = num_tc;
1613 return 0;
1614 }
1615
1616 static inline
1617 int netdev_get_num_tc(struct net_device *dev)
1618 {
1619 return dev->num_tc;
1620 }
1621
1622 static inline
1623 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1624 unsigned int index)
1625 {
1626 return &dev->_tx[index];
1627 }
1628
1629 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1630 void (*f)(struct net_device *,
1631 struct netdev_queue *,
1632 void *),
1633 void *arg)
1634 {
1635 unsigned int i;
1636
1637 for (i = 0; i < dev->num_tx_queues; i++)
1638 f(dev, &dev->_tx[i], arg);
1639 }
1640
1641 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1642 struct sk_buff *skb,
1643 void *accel_priv);
1644
1645 /*
1646 * Net namespace inlines
1647 */
1648 static inline
1649 struct net *dev_net(const struct net_device *dev)
1650 {
1651 return read_pnet(&dev->nd_net);
1652 }
1653
1654 static inline
1655 void dev_net_set(struct net_device *dev, struct net *net)
1656 {
1657 #ifdef CONFIG_NET_NS
1658 release_net(dev->nd_net);
1659 dev->nd_net = hold_net(net);
1660 #endif
1661 }
1662
1663 static inline bool netdev_uses_dsa_tags(struct net_device *dev)
1664 {
1665 #ifdef CONFIG_NET_DSA_TAG_DSA
1666 if (dev->dsa_ptr != NULL)
1667 return dsa_uses_dsa_tags(dev->dsa_ptr);
1668 #endif
1669
1670 return 0;
1671 }
1672
1673 static inline bool netdev_uses_trailer_tags(struct net_device *dev)
1674 {
1675 #ifdef CONFIG_NET_DSA_TAG_TRAILER
1676 if (dev->dsa_ptr != NULL)
1677 return dsa_uses_trailer_tags(dev->dsa_ptr);
1678 #endif
1679
1680 return 0;
1681 }
1682
1683 /**
1684 * netdev_priv - access network device private data
1685 * @dev: network device
1686 *
1687 * Get network device private data
1688 */
1689 static inline void *netdev_priv(const struct net_device *dev)
1690 {
1691 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1692 }
1693
1694 /* Set the sysfs physical device reference for the network logical device
1695 * if set prior to registration will cause a symlink during initialization.
1696 */
1697 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1698
1699 /* Set the sysfs device type for the network logical device to allow
1700 * fine-grained identification of different network device types. For
1701 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1702 */
1703 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1704
1705 /* Default NAPI poll() weight
1706 * Device drivers are strongly advised to not use bigger value
1707 */
1708 #define NAPI_POLL_WEIGHT 64
1709
1710 /**
1711 * netif_napi_add - initialize a napi context
1712 * @dev: network device
1713 * @napi: napi context
1714 * @poll: polling function
1715 * @weight: default weight
1716 *
1717 * netif_napi_add() must be used to initialize a napi context prior to calling
1718 * *any* of the other napi related functions.
1719 */
1720 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1721 int (*poll)(struct napi_struct *, int), int weight);
1722
1723 /**
1724 * netif_napi_del - remove a napi context
1725 * @napi: napi context
1726 *
1727 * netif_napi_del() removes a napi context from the network device napi list
1728 */
1729 void netif_napi_del(struct napi_struct *napi);
1730
1731 struct napi_gro_cb {
1732 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1733 void *frag0;
1734
1735 /* Length of frag0. */
1736 unsigned int frag0_len;
1737
1738 /* This indicates where we are processing relative to skb->data. */
1739 int data_offset;
1740
1741 /* This is non-zero if the packet cannot be merged with the new skb. */
1742 u16 flush;
1743
1744 /* Save the IP ID here and check when we get to the transport layer */
1745 u16 flush_id;
1746
1747 /* Number of segments aggregated. */
1748 u16 count;
1749
1750 /* This is non-zero if the packet may be of the same flow. */
1751 u8 same_flow;
1752
1753 /* Free the skb? */
1754 u8 free;
1755 #define NAPI_GRO_FREE 1
1756 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1757
1758 /* jiffies when first packet was created/queued */
1759 unsigned long age;
1760
1761 /* Used in ipv6_gro_receive() */
1762 u16 proto;
1763
1764 /* Used in udp_gro_receive */
1765 u16 udp_mark;
1766
1767 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1768 __wsum csum;
1769
1770 /* used in skb_gro_receive() slow path */
1771 struct sk_buff *last;
1772 };
1773
1774 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1775
1776 struct packet_type {
1777 __be16 type; /* This is really htons(ether_type). */
1778 struct net_device *dev; /* NULL is wildcarded here */
1779 int (*func) (struct sk_buff *,
1780 struct net_device *,
1781 struct packet_type *,
1782 struct net_device *);
1783 bool (*id_match)(struct packet_type *ptype,
1784 struct sock *sk);
1785 void *af_packet_priv;
1786 struct list_head list;
1787 };
1788
1789 struct offload_callbacks {
1790 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1791 netdev_features_t features);
1792 int (*gso_send_check)(struct sk_buff *skb);
1793 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1794 struct sk_buff *skb);
1795 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1796 };
1797
1798 struct packet_offload {
1799 __be16 type; /* This is really htons(ether_type). */
1800 struct offload_callbacks callbacks;
1801 struct list_head list;
1802 };
1803
1804 struct udp_offload {
1805 __be16 port;
1806 struct offload_callbacks callbacks;
1807 };
1808
1809 /* often modified stats are per cpu, other are shared (netdev->stats) */
1810 struct pcpu_sw_netstats {
1811 u64 rx_packets;
1812 u64 rx_bytes;
1813 u64 tx_packets;
1814 u64 tx_bytes;
1815 struct u64_stats_sync syncp;
1816 };
1817
1818 #define netdev_alloc_pcpu_stats(type) \
1819 ({ \
1820 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
1821 if (pcpu_stats) { \
1822 int i; \
1823 for_each_possible_cpu(i) { \
1824 typeof(type) *stat; \
1825 stat = per_cpu_ptr(pcpu_stats, i); \
1826 u64_stats_init(&stat->syncp); \
1827 } \
1828 } \
1829 pcpu_stats; \
1830 })
1831
1832 #include <linux/notifier.h>
1833
1834 /* netdevice notifier chain. Please remember to update the rtnetlink
1835 * notification exclusion list in rtnetlink_event() when adding new
1836 * types.
1837 */
1838 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
1839 #define NETDEV_DOWN 0x0002
1840 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
1841 detected a hardware crash and restarted
1842 - we can use this eg to kick tcp sessions
1843 once done */
1844 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
1845 #define NETDEV_REGISTER 0x0005
1846 #define NETDEV_UNREGISTER 0x0006
1847 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
1848 #define NETDEV_CHANGEADDR 0x0008
1849 #define NETDEV_GOING_DOWN 0x0009
1850 #define NETDEV_CHANGENAME 0x000A
1851 #define NETDEV_FEAT_CHANGE 0x000B
1852 #define NETDEV_BONDING_FAILOVER 0x000C
1853 #define NETDEV_PRE_UP 0x000D
1854 #define NETDEV_PRE_TYPE_CHANGE 0x000E
1855 #define NETDEV_POST_TYPE_CHANGE 0x000F
1856 #define NETDEV_POST_INIT 0x0010
1857 #define NETDEV_UNREGISTER_FINAL 0x0011
1858 #define NETDEV_RELEASE 0x0012
1859 #define NETDEV_NOTIFY_PEERS 0x0013
1860 #define NETDEV_JOIN 0x0014
1861 #define NETDEV_CHANGEUPPER 0x0015
1862 #define NETDEV_RESEND_IGMP 0x0016
1863 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
1864
1865 int register_netdevice_notifier(struct notifier_block *nb);
1866 int unregister_netdevice_notifier(struct notifier_block *nb);
1867
1868 struct netdev_notifier_info {
1869 struct net_device *dev;
1870 };
1871
1872 struct netdev_notifier_change_info {
1873 struct netdev_notifier_info info; /* must be first */
1874 unsigned int flags_changed;
1875 };
1876
1877 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
1878 struct net_device *dev)
1879 {
1880 info->dev = dev;
1881 }
1882
1883 static inline struct net_device *
1884 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
1885 {
1886 return info->dev;
1887 }
1888
1889 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
1890
1891
1892 extern rwlock_t dev_base_lock; /* Device list lock */
1893
1894 #define for_each_netdev(net, d) \
1895 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
1896 #define for_each_netdev_reverse(net, d) \
1897 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
1898 #define for_each_netdev_rcu(net, d) \
1899 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
1900 #define for_each_netdev_safe(net, d, n) \
1901 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
1902 #define for_each_netdev_continue(net, d) \
1903 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
1904 #define for_each_netdev_continue_rcu(net, d) \
1905 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
1906 #define for_each_netdev_in_bond_rcu(bond, slave) \
1907 for_each_netdev_rcu(&init_net, slave) \
1908 if (netdev_master_upper_dev_get_rcu(slave) == bond)
1909 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
1910
1911 static inline struct net_device *next_net_device(struct net_device *dev)
1912 {
1913 struct list_head *lh;
1914 struct net *net;
1915
1916 net = dev_net(dev);
1917 lh = dev->dev_list.next;
1918 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1919 }
1920
1921 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
1922 {
1923 struct list_head *lh;
1924 struct net *net;
1925
1926 net = dev_net(dev);
1927 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
1928 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1929 }
1930
1931 static inline struct net_device *first_net_device(struct net *net)
1932 {
1933 return list_empty(&net->dev_base_head) ? NULL :
1934 net_device_entry(net->dev_base_head.next);
1935 }
1936
1937 static inline struct net_device *first_net_device_rcu(struct net *net)
1938 {
1939 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
1940
1941 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
1942 }
1943
1944 int netdev_boot_setup_check(struct net_device *dev);
1945 unsigned long netdev_boot_base(const char *prefix, int unit);
1946 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
1947 const char *hwaddr);
1948 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
1949 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
1950 void dev_add_pack(struct packet_type *pt);
1951 void dev_remove_pack(struct packet_type *pt);
1952 void __dev_remove_pack(struct packet_type *pt);
1953 void dev_add_offload(struct packet_offload *po);
1954 void dev_remove_offload(struct packet_offload *po);
1955
1956 struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short flags,
1957 unsigned short mask);
1958 struct net_device *dev_get_by_name(struct net *net, const char *name);
1959 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
1960 struct net_device *__dev_get_by_name(struct net *net, const char *name);
1961 int dev_alloc_name(struct net_device *dev, const char *name);
1962 int dev_open(struct net_device *dev);
1963 int dev_close(struct net_device *dev);
1964 void dev_disable_lro(struct net_device *dev);
1965 int dev_loopback_xmit(struct sk_buff *newskb);
1966 int dev_queue_xmit(struct sk_buff *skb);
1967 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
1968 int register_netdevice(struct net_device *dev);
1969 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
1970 void unregister_netdevice_many(struct list_head *head);
1971 static inline void unregister_netdevice(struct net_device *dev)
1972 {
1973 unregister_netdevice_queue(dev, NULL);
1974 }
1975
1976 int netdev_refcnt_read(const struct net_device *dev);
1977 void free_netdev(struct net_device *dev);
1978 void netdev_freemem(struct net_device *dev);
1979 void synchronize_net(void);
1980 int init_dummy_netdev(struct net_device *dev);
1981
1982 struct net_device *dev_get_by_index(struct net *net, int ifindex);
1983 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
1984 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
1985 int netdev_get_name(struct net *net, char *name, int ifindex);
1986 int dev_restart(struct net_device *dev);
1987 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
1988
1989 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
1990 {
1991 return NAPI_GRO_CB(skb)->data_offset;
1992 }
1993
1994 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
1995 {
1996 return skb->len - NAPI_GRO_CB(skb)->data_offset;
1997 }
1998
1999 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2000 {
2001 NAPI_GRO_CB(skb)->data_offset += len;
2002 }
2003
2004 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2005 unsigned int offset)
2006 {
2007 return NAPI_GRO_CB(skb)->frag0 + offset;
2008 }
2009
2010 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2011 {
2012 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2013 }
2014
2015 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2016 unsigned int offset)
2017 {
2018 if (!pskb_may_pull(skb, hlen))
2019 return NULL;
2020
2021 NAPI_GRO_CB(skb)->frag0 = NULL;
2022 NAPI_GRO_CB(skb)->frag0_len = 0;
2023 return skb->data + offset;
2024 }
2025
2026 static inline void *skb_gro_network_header(struct sk_buff *skb)
2027 {
2028 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2029 skb_network_offset(skb);
2030 }
2031
2032 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2033 const void *start, unsigned int len)
2034 {
2035 if (skb->ip_summed == CHECKSUM_COMPLETE)
2036 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2037 csum_partial(start, len, 0));
2038 }
2039
2040 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2041 unsigned short type,
2042 const void *daddr, const void *saddr,
2043 unsigned int len)
2044 {
2045 if (!dev->header_ops || !dev->header_ops->create)
2046 return 0;
2047
2048 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2049 }
2050
2051 static inline int dev_parse_header(const struct sk_buff *skb,
2052 unsigned char *haddr)
2053 {
2054 const struct net_device *dev = skb->dev;
2055
2056 if (!dev->header_ops || !dev->header_ops->parse)
2057 return 0;
2058 return dev->header_ops->parse(skb, haddr);
2059 }
2060
2061 static inline int dev_rebuild_header(struct sk_buff *skb)
2062 {
2063 const struct net_device *dev = skb->dev;
2064
2065 if (!dev->header_ops || !dev->header_ops->rebuild)
2066 return 0;
2067 return dev->header_ops->rebuild(skb);
2068 }
2069
2070 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2071 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2072 static inline int unregister_gifconf(unsigned int family)
2073 {
2074 return register_gifconf(family, NULL);
2075 }
2076
2077 #ifdef CONFIG_NET_FLOW_LIMIT
2078 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2079 struct sd_flow_limit {
2080 u64 count;
2081 unsigned int num_buckets;
2082 unsigned int history_head;
2083 u16 history[FLOW_LIMIT_HISTORY];
2084 u8 buckets[];
2085 };
2086
2087 extern int netdev_flow_limit_table_len;
2088 #endif /* CONFIG_NET_FLOW_LIMIT */
2089
2090 /*
2091 * Incoming packets are placed on per-cpu queues
2092 */
2093 struct softnet_data {
2094 struct Qdisc *output_queue;
2095 struct Qdisc **output_queue_tailp;
2096 struct list_head poll_list;
2097 struct sk_buff *completion_queue;
2098 struct sk_buff_head process_queue;
2099
2100 /* stats */
2101 unsigned int processed;
2102 unsigned int time_squeeze;
2103 unsigned int cpu_collision;
2104 unsigned int received_rps;
2105
2106 #ifdef CONFIG_RPS
2107 struct softnet_data *rps_ipi_list;
2108
2109 /* Elements below can be accessed between CPUs for RPS */
2110 struct call_single_data csd ____cacheline_aligned_in_smp;
2111 struct softnet_data *rps_ipi_next;
2112 unsigned int cpu;
2113 unsigned int input_queue_head;
2114 unsigned int input_queue_tail;
2115 #endif
2116 unsigned int dropped;
2117 struct sk_buff_head input_pkt_queue;
2118 struct napi_struct backlog;
2119
2120 #ifdef CONFIG_NET_FLOW_LIMIT
2121 struct sd_flow_limit __rcu *flow_limit;
2122 #endif
2123 };
2124
2125 static inline void input_queue_head_incr(struct softnet_data *sd)
2126 {
2127 #ifdef CONFIG_RPS
2128 sd->input_queue_head++;
2129 #endif
2130 }
2131
2132 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2133 unsigned int *qtail)
2134 {
2135 #ifdef CONFIG_RPS
2136 *qtail = ++sd->input_queue_tail;
2137 #endif
2138 }
2139
2140 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2141
2142 void __netif_schedule(struct Qdisc *q);
2143
2144 static inline void netif_schedule_queue(struct netdev_queue *txq)
2145 {
2146 if (!(txq->state & QUEUE_STATE_ANY_XOFF))
2147 __netif_schedule(txq->qdisc);
2148 }
2149
2150 static inline void netif_tx_schedule_all(struct net_device *dev)
2151 {
2152 unsigned int i;
2153
2154 for (i = 0; i < dev->num_tx_queues; i++)
2155 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2156 }
2157
2158 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2159 {
2160 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2161 }
2162
2163 /**
2164 * netif_start_queue - allow transmit
2165 * @dev: network device
2166 *
2167 * Allow upper layers to call the device hard_start_xmit routine.
2168 */
2169 static inline void netif_start_queue(struct net_device *dev)
2170 {
2171 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2172 }
2173
2174 static inline void netif_tx_start_all_queues(struct net_device *dev)
2175 {
2176 unsigned int i;
2177
2178 for (i = 0; i < dev->num_tx_queues; i++) {
2179 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2180 netif_tx_start_queue(txq);
2181 }
2182 }
2183
2184 static inline void netif_tx_wake_queue(struct netdev_queue *dev_queue)
2185 {
2186 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state))
2187 __netif_schedule(dev_queue->qdisc);
2188 }
2189
2190 /**
2191 * netif_wake_queue - restart transmit
2192 * @dev: network device
2193 *
2194 * Allow upper layers to call the device hard_start_xmit routine.
2195 * Used for flow control when transmit resources are available.
2196 */
2197 static inline void netif_wake_queue(struct net_device *dev)
2198 {
2199 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2200 }
2201
2202 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2203 {
2204 unsigned int i;
2205
2206 for (i = 0; i < dev->num_tx_queues; i++) {
2207 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2208 netif_tx_wake_queue(txq);
2209 }
2210 }
2211
2212 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2213 {
2214 if (WARN_ON(!dev_queue)) {
2215 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2216 return;
2217 }
2218 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2219 }
2220
2221 /**
2222 * netif_stop_queue - stop transmitted packets
2223 * @dev: network device
2224 *
2225 * Stop upper layers calling the device hard_start_xmit routine.
2226 * Used for flow control when transmit resources are unavailable.
2227 */
2228 static inline void netif_stop_queue(struct net_device *dev)
2229 {
2230 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2231 }
2232
2233 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2234 {
2235 unsigned int i;
2236
2237 for (i = 0; i < dev->num_tx_queues; i++) {
2238 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2239 netif_tx_stop_queue(txq);
2240 }
2241 }
2242
2243 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2244 {
2245 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2246 }
2247
2248 /**
2249 * netif_queue_stopped - test if transmit queue is flowblocked
2250 * @dev: network device
2251 *
2252 * Test if transmit queue on device is currently unable to send.
2253 */
2254 static inline bool netif_queue_stopped(const struct net_device *dev)
2255 {
2256 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2257 }
2258
2259 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2260 {
2261 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2262 }
2263
2264 static inline bool
2265 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2266 {
2267 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2268 }
2269
2270 static inline bool
2271 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2272 {
2273 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2274 }
2275
2276 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2277 unsigned int bytes)
2278 {
2279 #ifdef CONFIG_BQL
2280 dql_queued(&dev_queue->dql, bytes);
2281
2282 if (likely(dql_avail(&dev_queue->dql) >= 0))
2283 return;
2284
2285 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2286
2287 /*
2288 * The XOFF flag must be set before checking the dql_avail below,
2289 * because in netdev_tx_completed_queue we update the dql_completed
2290 * before checking the XOFF flag.
2291 */
2292 smp_mb();
2293
2294 /* check again in case another CPU has just made room avail */
2295 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2296 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2297 #endif
2298 }
2299
2300 /**
2301 * netdev_sent_queue - report the number of bytes queued to hardware
2302 * @dev: network device
2303 * @bytes: number of bytes queued to the hardware device queue
2304 *
2305 * Report the number of bytes queued for sending/completion to the network
2306 * device hardware queue. @bytes should be a good approximation and should
2307 * exactly match netdev_completed_queue() @bytes
2308 */
2309 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2310 {
2311 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2312 }
2313
2314 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2315 unsigned int pkts, unsigned int bytes)
2316 {
2317 #ifdef CONFIG_BQL
2318 if (unlikely(!bytes))
2319 return;
2320
2321 dql_completed(&dev_queue->dql, bytes);
2322
2323 /*
2324 * Without the memory barrier there is a small possiblity that
2325 * netdev_tx_sent_queue will miss the update and cause the queue to
2326 * be stopped forever
2327 */
2328 smp_mb();
2329
2330 if (dql_avail(&dev_queue->dql) < 0)
2331 return;
2332
2333 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2334 netif_schedule_queue(dev_queue);
2335 #endif
2336 }
2337
2338 /**
2339 * netdev_completed_queue - report bytes and packets completed by device
2340 * @dev: network device
2341 * @pkts: actual number of packets sent over the medium
2342 * @bytes: actual number of bytes sent over the medium
2343 *
2344 * Report the number of bytes and packets transmitted by the network device
2345 * hardware queue over the physical medium, @bytes must exactly match the
2346 * @bytes amount passed to netdev_sent_queue()
2347 */
2348 static inline void netdev_completed_queue(struct net_device *dev,
2349 unsigned int pkts, unsigned int bytes)
2350 {
2351 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2352 }
2353
2354 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2355 {
2356 #ifdef CONFIG_BQL
2357 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2358 dql_reset(&q->dql);
2359 #endif
2360 }
2361
2362 /**
2363 * netdev_reset_queue - reset the packets and bytes count of a network device
2364 * @dev_queue: network device
2365 *
2366 * Reset the bytes and packet count of a network device and clear the
2367 * software flow control OFF bit for this network device
2368 */
2369 static inline void netdev_reset_queue(struct net_device *dev_queue)
2370 {
2371 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2372 }
2373
2374 /**
2375 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2376 * @dev: network device
2377 * @queue_index: given tx queue index
2378 *
2379 * Returns 0 if given tx queue index >= number of device tx queues,
2380 * otherwise returns the originally passed tx queue index.
2381 */
2382 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2383 {
2384 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2385 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2386 dev->name, queue_index,
2387 dev->real_num_tx_queues);
2388 return 0;
2389 }
2390
2391 return queue_index;
2392 }
2393
2394 /**
2395 * netif_running - test if up
2396 * @dev: network device
2397 *
2398 * Test if the device has been brought up.
2399 */
2400 static inline bool netif_running(const struct net_device *dev)
2401 {
2402 return test_bit(__LINK_STATE_START, &dev->state);
2403 }
2404
2405 /*
2406 * Routines to manage the subqueues on a device. We only need start
2407 * stop, and a check if it's stopped. All other device management is
2408 * done at the overall netdevice level.
2409 * Also test the device if we're multiqueue.
2410 */
2411
2412 /**
2413 * netif_start_subqueue - allow sending packets on subqueue
2414 * @dev: network device
2415 * @queue_index: sub queue index
2416 *
2417 * Start individual transmit queue of a device with multiple transmit queues.
2418 */
2419 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2420 {
2421 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2422
2423 netif_tx_start_queue(txq);
2424 }
2425
2426 /**
2427 * netif_stop_subqueue - stop sending packets on subqueue
2428 * @dev: network device
2429 * @queue_index: sub queue index
2430 *
2431 * Stop individual transmit queue of a device with multiple transmit queues.
2432 */
2433 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2434 {
2435 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2436 netif_tx_stop_queue(txq);
2437 }
2438
2439 /**
2440 * netif_subqueue_stopped - test status of subqueue
2441 * @dev: network device
2442 * @queue_index: sub queue index
2443 *
2444 * Check individual transmit queue of a device with multiple transmit queues.
2445 */
2446 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2447 u16 queue_index)
2448 {
2449 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2450
2451 return netif_tx_queue_stopped(txq);
2452 }
2453
2454 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2455 struct sk_buff *skb)
2456 {
2457 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2458 }
2459
2460 /**
2461 * netif_wake_subqueue - allow sending packets on subqueue
2462 * @dev: network device
2463 * @queue_index: sub queue index
2464 *
2465 * Resume individual transmit queue of a device with multiple transmit queues.
2466 */
2467 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
2468 {
2469 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2470 if (test_and_clear_bit(__QUEUE_STATE_DRV_XOFF, &txq->state))
2471 __netif_schedule(txq->qdisc);
2472 }
2473
2474 #ifdef CONFIG_XPS
2475 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2476 u16 index);
2477 #else
2478 static inline int netif_set_xps_queue(struct net_device *dev,
2479 const struct cpumask *mask,
2480 u16 index)
2481 {
2482 return 0;
2483 }
2484 #endif
2485
2486 /*
2487 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2488 * as a distribution range limit for the returned value.
2489 */
2490 static inline u16 skb_tx_hash(const struct net_device *dev,
2491 struct sk_buff *skb)
2492 {
2493 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2494 }
2495
2496 /**
2497 * netif_is_multiqueue - test if device has multiple transmit queues
2498 * @dev: network device
2499 *
2500 * Check if device has multiple transmit queues
2501 */
2502 static inline bool netif_is_multiqueue(const struct net_device *dev)
2503 {
2504 return dev->num_tx_queues > 1;
2505 }
2506
2507 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2508
2509 #ifdef CONFIG_SYSFS
2510 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2511 #else
2512 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2513 unsigned int rxq)
2514 {
2515 return 0;
2516 }
2517 #endif
2518
2519 static inline int netif_copy_real_num_queues(struct net_device *to_dev,
2520 const struct net_device *from_dev)
2521 {
2522 int err;
2523
2524 err = netif_set_real_num_tx_queues(to_dev,
2525 from_dev->real_num_tx_queues);
2526 if (err)
2527 return err;
2528 #ifdef CONFIG_SYSFS
2529 return netif_set_real_num_rx_queues(to_dev,
2530 from_dev->real_num_rx_queues);
2531 #else
2532 return 0;
2533 #endif
2534 }
2535
2536 #ifdef CONFIG_SYSFS
2537 static inline unsigned int get_netdev_rx_queue_index(
2538 struct netdev_rx_queue *queue)
2539 {
2540 struct net_device *dev = queue->dev;
2541 int index = queue - dev->_rx;
2542
2543 BUG_ON(index >= dev->num_rx_queues);
2544 return index;
2545 }
2546 #endif
2547
2548 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2549 int netif_get_num_default_rss_queues(void);
2550
2551 enum skb_free_reason {
2552 SKB_REASON_CONSUMED,
2553 SKB_REASON_DROPPED,
2554 };
2555
2556 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2557 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2558
2559 /*
2560 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2561 * interrupt context or with hardware interrupts being disabled.
2562 * (in_irq() || irqs_disabled())
2563 *
2564 * We provide four helpers that can be used in following contexts :
2565 *
2566 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2567 * replacing kfree_skb(skb)
2568 *
2569 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2570 * Typically used in place of consume_skb(skb) in TX completion path
2571 *
2572 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2573 * replacing kfree_skb(skb)
2574 *
2575 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2576 * and consumed a packet. Used in place of consume_skb(skb)
2577 */
2578 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2579 {
2580 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2581 }
2582
2583 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2584 {
2585 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2586 }
2587
2588 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2589 {
2590 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2591 }
2592
2593 static inline void dev_consume_skb_any(struct sk_buff *skb)
2594 {
2595 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2596 }
2597
2598 int netif_rx(struct sk_buff *skb);
2599 int netif_rx_ni(struct sk_buff *skb);
2600 int netif_receive_skb(struct sk_buff *skb);
2601 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2602 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2603 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2604 gro_result_t napi_gro_frags(struct napi_struct *napi);
2605 struct packet_offload *gro_find_receive_by_type(__be16 type);
2606 struct packet_offload *gro_find_complete_by_type(__be16 type);
2607
2608 static inline void napi_free_frags(struct napi_struct *napi)
2609 {
2610 kfree_skb(napi->skb);
2611 napi->skb = NULL;
2612 }
2613
2614 int netdev_rx_handler_register(struct net_device *dev,
2615 rx_handler_func_t *rx_handler,
2616 void *rx_handler_data);
2617 void netdev_rx_handler_unregister(struct net_device *dev);
2618
2619 bool dev_valid_name(const char *name);
2620 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2621 int dev_ethtool(struct net *net, struct ifreq *);
2622 unsigned int dev_get_flags(const struct net_device *);
2623 int __dev_change_flags(struct net_device *, unsigned int flags);
2624 int dev_change_flags(struct net_device *, unsigned int);
2625 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2626 unsigned int gchanges);
2627 int dev_change_name(struct net_device *, const char *);
2628 int dev_set_alias(struct net_device *, const char *, size_t);
2629 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2630 int dev_set_mtu(struct net_device *, int);
2631 void dev_set_group(struct net_device *, int);
2632 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2633 int dev_change_carrier(struct net_device *, bool new_carrier);
2634 int dev_get_phys_port_id(struct net_device *dev,
2635 struct netdev_phys_port_id *ppid);
2636 int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2637 struct netdev_queue *txq);
2638 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2639 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2640 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2641
2642 extern int netdev_budget;
2643
2644 /* Called by rtnetlink.c:rtnl_unlock() */
2645 void netdev_run_todo(void);
2646
2647 /**
2648 * dev_put - release reference to device
2649 * @dev: network device
2650 *
2651 * Release reference to device to allow it to be freed.
2652 */
2653 static inline void dev_put(struct net_device *dev)
2654 {
2655 this_cpu_dec(*dev->pcpu_refcnt);
2656 }
2657
2658 /**
2659 * dev_hold - get reference to device
2660 * @dev: network device
2661 *
2662 * Hold reference to device to keep it from being freed.
2663 */
2664 static inline void dev_hold(struct net_device *dev)
2665 {
2666 this_cpu_inc(*dev->pcpu_refcnt);
2667 }
2668
2669 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2670 * and _off may be called from IRQ context, but it is caller
2671 * who is responsible for serialization of these calls.
2672 *
2673 * The name carrier is inappropriate, these functions should really be
2674 * called netif_lowerlayer_*() because they represent the state of any
2675 * kind of lower layer not just hardware media.
2676 */
2677
2678 void linkwatch_init_dev(struct net_device *dev);
2679 void linkwatch_fire_event(struct net_device *dev);
2680 void linkwatch_forget_dev(struct net_device *dev);
2681
2682 /**
2683 * netif_carrier_ok - test if carrier present
2684 * @dev: network device
2685 *
2686 * Check if carrier is present on device
2687 */
2688 static inline bool netif_carrier_ok(const struct net_device *dev)
2689 {
2690 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
2691 }
2692
2693 unsigned long dev_trans_start(struct net_device *dev);
2694
2695 void __netdev_watchdog_up(struct net_device *dev);
2696
2697 void netif_carrier_on(struct net_device *dev);
2698
2699 void netif_carrier_off(struct net_device *dev);
2700
2701 /**
2702 * netif_dormant_on - mark device as dormant.
2703 * @dev: network device
2704 *
2705 * Mark device as dormant (as per RFC2863).
2706 *
2707 * The dormant state indicates that the relevant interface is not
2708 * actually in a condition to pass packets (i.e., it is not 'up') but is
2709 * in a "pending" state, waiting for some external event. For "on-
2710 * demand" interfaces, this new state identifies the situation where the
2711 * interface is waiting for events to place it in the up state.
2712 *
2713 */
2714 static inline void netif_dormant_on(struct net_device *dev)
2715 {
2716 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
2717 linkwatch_fire_event(dev);
2718 }
2719
2720 /**
2721 * netif_dormant_off - set device as not dormant.
2722 * @dev: network device
2723 *
2724 * Device is not in dormant state.
2725 */
2726 static inline void netif_dormant_off(struct net_device *dev)
2727 {
2728 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
2729 linkwatch_fire_event(dev);
2730 }
2731
2732 /**
2733 * netif_dormant - test if carrier present
2734 * @dev: network device
2735 *
2736 * Check if carrier is present on device
2737 */
2738 static inline bool netif_dormant(const struct net_device *dev)
2739 {
2740 return test_bit(__LINK_STATE_DORMANT, &dev->state);
2741 }
2742
2743
2744 /**
2745 * netif_oper_up - test if device is operational
2746 * @dev: network device
2747 *
2748 * Check if carrier is operational
2749 */
2750 static inline bool netif_oper_up(const struct net_device *dev)
2751 {
2752 return (dev->operstate == IF_OPER_UP ||
2753 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
2754 }
2755
2756 /**
2757 * netif_device_present - is device available or removed
2758 * @dev: network device
2759 *
2760 * Check if device has not been removed from system.
2761 */
2762 static inline bool netif_device_present(struct net_device *dev)
2763 {
2764 return test_bit(__LINK_STATE_PRESENT, &dev->state);
2765 }
2766
2767 void netif_device_detach(struct net_device *dev);
2768
2769 void netif_device_attach(struct net_device *dev);
2770
2771 /*
2772 * Network interface message level settings
2773 */
2774
2775 enum {
2776 NETIF_MSG_DRV = 0x0001,
2777 NETIF_MSG_PROBE = 0x0002,
2778 NETIF_MSG_LINK = 0x0004,
2779 NETIF_MSG_TIMER = 0x0008,
2780 NETIF_MSG_IFDOWN = 0x0010,
2781 NETIF_MSG_IFUP = 0x0020,
2782 NETIF_MSG_RX_ERR = 0x0040,
2783 NETIF_MSG_TX_ERR = 0x0080,
2784 NETIF_MSG_TX_QUEUED = 0x0100,
2785 NETIF_MSG_INTR = 0x0200,
2786 NETIF_MSG_TX_DONE = 0x0400,
2787 NETIF_MSG_RX_STATUS = 0x0800,
2788 NETIF_MSG_PKTDATA = 0x1000,
2789 NETIF_MSG_HW = 0x2000,
2790 NETIF_MSG_WOL = 0x4000,
2791 };
2792
2793 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
2794 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
2795 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
2796 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
2797 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
2798 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
2799 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
2800 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
2801 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
2802 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
2803 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
2804 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
2805 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
2806 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
2807 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
2808
2809 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
2810 {
2811 /* use default */
2812 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
2813 return default_msg_enable_bits;
2814 if (debug_value == 0) /* no output */
2815 return 0;
2816 /* set low N bits */
2817 return (1 << debug_value) - 1;
2818 }
2819
2820 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
2821 {
2822 spin_lock(&txq->_xmit_lock);
2823 txq->xmit_lock_owner = cpu;
2824 }
2825
2826 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
2827 {
2828 spin_lock_bh(&txq->_xmit_lock);
2829 txq->xmit_lock_owner = smp_processor_id();
2830 }
2831
2832 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
2833 {
2834 bool ok = spin_trylock(&txq->_xmit_lock);
2835 if (likely(ok))
2836 txq->xmit_lock_owner = smp_processor_id();
2837 return ok;
2838 }
2839
2840 static inline void __netif_tx_unlock(struct netdev_queue *txq)
2841 {
2842 txq->xmit_lock_owner = -1;
2843 spin_unlock(&txq->_xmit_lock);
2844 }
2845
2846 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
2847 {
2848 txq->xmit_lock_owner = -1;
2849 spin_unlock_bh(&txq->_xmit_lock);
2850 }
2851
2852 static inline void txq_trans_update(struct netdev_queue *txq)
2853 {
2854 if (txq->xmit_lock_owner != -1)
2855 txq->trans_start = jiffies;
2856 }
2857
2858 /**
2859 * netif_tx_lock - grab network device transmit lock
2860 * @dev: network device
2861 *
2862 * Get network device transmit lock
2863 */
2864 static inline void netif_tx_lock(struct net_device *dev)
2865 {
2866 unsigned int i;
2867 int cpu;
2868
2869 spin_lock(&dev->tx_global_lock);
2870 cpu = smp_processor_id();
2871 for (i = 0; i < dev->num_tx_queues; i++) {
2872 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2873
2874 /* We are the only thread of execution doing a
2875 * freeze, but we have to grab the _xmit_lock in
2876 * order to synchronize with threads which are in
2877 * the ->hard_start_xmit() handler and already
2878 * checked the frozen bit.
2879 */
2880 __netif_tx_lock(txq, cpu);
2881 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
2882 __netif_tx_unlock(txq);
2883 }
2884 }
2885
2886 static inline void netif_tx_lock_bh(struct net_device *dev)
2887 {
2888 local_bh_disable();
2889 netif_tx_lock(dev);
2890 }
2891
2892 static inline void netif_tx_unlock(struct net_device *dev)
2893 {
2894 unsigned int i;
2895
2896 for (i = 0; i < dev->num_tx_queues; i++) {
2897 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2898
2899 /* No need to grab the _xmit_lock here. If the
2900 * queue is not stopped for another reason, we
2901 * force a schedule.
2902 */
2903 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
2904 netif_schedule_queue(txq);
2905 }
2906 spin_unlock(&dev->tx_global_lock);
2907 }
2908
2909 static inline void netif_tx_unlock_bh(struct net_device *dev)
2910 {
2911 netif_tx_unlock(dev);
2912 local_bh_enable();
2913 }
2914
2915 #define HARD_TX_LOCK(dev, txq, cpu) { \
2916 if ((dev->features & NETIF_F_LLTX) == 0) { \
2917 __netif_tx_lock(txq, cpu); \
2918 } \
2919 }
2920
2921 #define HARD_TX_TRYLOCK(dev, txq) \
2922 (((dev->features & NETIF_F_LLTX) == 0) ? \
2923 __netif_tx_trylock(txq) : \
2924 true )
2925
2926 #define HARD_TX_UNLOCK(dev, txq) { \
2927 if ((dev->features & NETIF_F_LLTX) == 0) { \
2928 __netif_tx_unlock(txq); \
2929 } \
2930 }
2931
2932 static inline void netif_tx_disable(struct net_device *dev)
2933 {
2934 unsigned int i;
2935 int cpu;
2936
2937 local_bh_disable();
2938 cpu = smp_processor_id();
2939 for (i = 0; i < dev->num_tx_queues; i++) {
2940 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2941
2942 __netif_tx_lock(txq, cpu);
2943 netif_tx_stop_queue(txq);
2944 __netif_tx_unlock(txq);
2945 }
2946 local_bh_enable();
2947 }
2948
2949 static inline void netif_addr_lock(struct net_device *dev)
2950 {
2951 spin_lock(&dev->addr_list_lock);
2952 }
2953
2954 static inline void netif_addr_lock_nested(struct net_device *dev)
2955 {
2956 int subclass = SINGLE_DEPTH_NESTING;
2957
2958 if (dev->netdev_ops->ndo_get_lock_subclass)
2959 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
2960
2961 spin_lock_nested(&dev->addr_list_lock, subclass);
2962 }
2963
2964 static inline void netif_addr_lock_bh(struct net_device *dev)
2965 {
2966 spin_lock_bh(&dev->addr_list_lock);
2967 }
2968
2969 static inline void netif_addr_unlock(struct net_device *dev)
2970 {
2971 spin_unlock(&dev->addr_list_lock);
2972 }
2973
2974 static inline void netif_addr_unlock_bh(struct net_device *dev)
2975 {
2976 spin_unlock_bh(&dev->addr_list_lock);
2977 }
2978
2979 /*
2980 * dev_addrs walker. Should be used only for read access. Call with
2981 * rcu_read_lock held.
2982 */
2983 #define for_each_dev_addr(dev, ha) \
2984 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
2985
2986 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
2987
2988 void ether_setup(struct net_device *dev);
2989
2990 /* Support for loadable net-drivers */
2991 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
2992 void (*setup)(struct net_device *),
2993 unsigned int txqs, unsigned int rxqs);
2994 #define alloc_netdev(sizeof_priv, name, setup) \
2995 alloc_netdev_mqs(sizeof_priv, name, setup, 1, 1)
2996
2997 #define alloc_netdev_mq(sizeof_priv, name, setup, count) \
2998 alloc_netdev_mqs(sizeof_priv, name, setup, count, count)
2999
3000 int register_netdev(struct net_device *dev);
3001 void unregister_netdev(struct net_device *dev);
3002
3003 /* General hardware address lists handling functions */
3004 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3005 struct netdev_hw_addr_list *from_list, int addr_len);
3006 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3007 struct netdev_hw_addr_list *from_list, int addr_len);
3008 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3009 struct net_device *dev,
3010 int (*sync)(struct net_device *, const unsigned char *),
3011 int (*unsync)(struct net_device *,
3012 const unsigned char *));
3013 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3014 struct net_device *dev,
3015 int (*unsync)(struct net_device *,
3016 const unsigned char *));
3017 void __hw_addr_init(struct netdev_hw_addr_list *list);
3018
3019 /* Functions used for device addresses handling */
3020 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3021 unsigned char addr_type);
3022 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3023 unsigned char addr_type);
3024 void dev_addr_flush(struct net_device *dev);
3025 int dev_addr_init(struct net_device *dev);
3026
3027 /* Functions used for unicast addresses handling */
3028 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3029 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3030 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3031 int dev_uc_sync(struct net_device *to, struct net_device *from);
3032 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3033 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3034 void dev_uc_flush(struct net_device *dev);
3035 void dev_uc_init(struct net_device *dev);
3036
3037 /**
3038 * __dev_uc_sync - Synchonize device's unicast list
3039 * @dev: device to sync
3040 * @sync: function to call if address should be added
3041 * @unsync: function to call if address should be removed
3042 *
3043 * Add newly added addresses to the interface, and release
3044 * addresses that have been deleted.
3045 **/
3046 static inline int __dev_uc_sync(struct net_device *dev,
3047 int (*sync)(struct net_device *,
3048 const unsigned char *),
3049 int (*unsync)(struct net_device *,
3050 const unsigned char *))
3051 {
3052 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3053 }
3054
3055 /**
3056 * __dev_uc_unsync - Remove synchonized addresses from device
3057 * @dev: device to sync
3058 * @unsync: function to call if address should be removed
3059 *
3060 * Remove all addresses that were added to the device by dev_uc_sync().
3061 **/
3062 static inline void __dev_uc_unsync(struct net_device *dev,
3063 int (*unsync)(struct net_device *,
3064 const unsigned char *))
3065 {
3066 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3067 }
3068
3069 /* Functions used for multicast addresses handling */
3070 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3071 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3072 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3073 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3074 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3075 int dev_mc_sync(struct net_device *to, struct net_device *from);
3076 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3077 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3078 void dev_mc_flush(struct net_device *dev);
3079 void dev_mc_init(struct net_device *dev);
3080
3081 /**
3082 * __dev_mc_sync - Synchonize device's multicast list
3083 * @dev: device to sync
3084 * @sync: function to call if address should be added
3085 * @unsync: function to call if address should be removed
3086 *
3087 * Add newly added addresses to the interface, and release
3088 * addresses that have been deleted.
3089 **/
3090 static inline int __dev_mc_sync(struct net_device *dev,
3091 int (*sync)(struct net_device *,
3092 const unsigned char *),
3093 int (*unsync)(struct net_device *,
3094 const unsigned char *))
3095 {
3096 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3097 }
3098
3099 /**
3100 * __dev_mc_unsync - Remove synchonized addresses from device
3101 * @dev: device to sync
3102 * @unsync: function to call if address should be removed
3103 *
3104 * Remove all addresses that were added to the device by dev_mc_sync().
3105 **/
3106 static inline void __dev_mc_unsync(struct net_device *dev,
3107 int (*unsync)(struct net_device *,
3108 const unsigned char *))
3109 {
3110 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3111 }
3112
3113 /* Functions used for secondary unicast and multicast support */
3114 void dev_set_rx_mode(struct net_device *dev);
3115 void __dev_set_rx_mode(struct net_device *dev);
3116 int dev_set_promiscuity(struct net_device *dev, int inc);
3117 int dev_set_allmulti(struct net_device *dev, int inc);
3118 void netdev_state_change(struct net_device *dev);
3119 void netdev_notify_peers(struct net_device *dev);
3120 void netdev_features_change(struct net_device *dev);
3121 /* Load a device via the kmod */
3122 void dev_load(struct net *net, const char *name);
3123 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3124 struct rtnl_link_stats64 *storage);
3125 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3126 const struct net_device_stats *netdev_stats);
3127
3128 extern int netdev_max_backlog;
3129 extern int netdev_tstamp_prequeue;
3130 extern int weight_p;
3131 extern int bpf_jit_enable;
3132
3133 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3134 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3135 struct list_head **iter);
3136 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3137 struct list_head **iter);
3138
3139 /* iterate through upper list, must be called under RCU read lock */
3140 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3141 for (iter = &(dev)->adj_list.upper, \
3142 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3143 updev; \
3144 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3145
3146 /* iterate through upper list, must be called under RCU read lock */
3147 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3148 for (iter = &(dev)->all_adj_list.upper, \
3149 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3150 updev; \
3151 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3152
3153 void *netdev_lower_get_next_private(struct net_device *dev,
3154 struct list_head **iter);
3155 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3156 struct list_head **iter);
3157
3158 #define netdev_for_each_lower_private(dev, priv, iter) \
3159 for (iter = (dev)->adj_list.lower.next, \
3160 priv = netdev_lower_get_next_private(dev, &(iter)); \
3161 priv; \
3162 priv = netdev_lower_get_next_private(dev, &(iter)))
3163
3164 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3165 for (iter = &(dev)->adj_list.lower, \
3166 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3167 priv; \
3168 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3169
3170 void *netdev_lower_get_next(struct net_device *dev,
3171 struct list_head **iter);
3172 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3173 for (iter = &(dev)->adj_list.lower, \
3174 ldev = netdev_lower_get_next(dev, &(iter)); \
3175 ldev; \
3176 ldev = netdev_lower_get_next(dev, &(iter)))
3177
3178 void *netdev_adjacent_get_private(struct list_head *adj_list);
3179 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3180 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3181 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3182 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3183 int netdev_master_upper_dev_link(struct net_device *dev,
3184 struct net_device *upper_dev);
3185 int netdev_master_upper_dev_link_private(struct net_device *dev,
3186 struct net_device *upper_dev,
3187 void *private);
3188 void netdev_upper_dev_unlink(struct net_device *dev,
3189 struct net_device *upper_dev);
3190 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3191 void *netdev_lower_dev_get_private(struct net_device *dev,
3192 struct net_device *lower_dev);
3193 int dev_get_nest_level(struct net_device *dev,
3194 bool (*type_check)(struct net_device *dev));
3195 int skb_checksum_help(struct sk_buff *skb);
3196 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3197 netdev_features_t features, bool tx_path);
3198 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3199 netdev_features_t features);
3200
3201 static inline
3202 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3203 {
3204 return __skb_gso_segment(skb, features, true);
3205 }
3206 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3207
3208 static inline bool can_checksum_protocol(netdev_features_t features,
3209 __be16 protocol)
3210 {
3211 return ((features & NETIF_F_GEN_CSUM) ||
3212 ((features & NETIF_F_V4_CSUM) &&
3213 protocol == htons(ETH_P_IP)) ||
3214 ((features & NETIF_F_V6_CSUM) &&
3215 protocol == htons(ETH_P_IPV6)) ||
3216 ((features & NETIF_F_FCOE_CRC) &&
3217 protocol == htons(ETH_P_FCOE)));
3218 }
3219
3220 #ifdef CONFIG_BUG
3221 void netdev_rx_csum_fault(struct net_device *dev);
3222 #else
3223 static inline void netdev_rx_csum_fault(struct net_device *dev)
3224 {
3225 }
3226 #endif
3227 /* rx skb timestamps */
3228 void net_enable_timestamp(void);
3229 void net_disable_timestamp(void);
3230
3231 #ifdef CONFIG_PROC_FS
3232 int __init dev_proc_init(void);
3233 #else
3234 #define dev_proc_init() 0
3235 #endif
3236
3237 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3238 const void *ns);
3239 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3240 const void *ns);
3241
3242 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3243 {
3244 return netdev_class_create_file_ns(class_attr, NULL);
3245 }
3246
3247 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3248 {
3249 netdev_class_remove_file_ns(class_attr, NULL);
3250 }
3251
3252 extern struct kobj_ns_type_operations net_ns_type_operations;
3253
3254 const char *netdev_drivername(const struct net_device *dev);
3255
3256 void linkwatch_run_queue(void);
3257
3258 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3259 netdev_features_t f2)
3260 {
3261 if (f1 & NETIF_F_GEN_CSUM)
3262 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3263 if (f2 & NETIF_F_GEN_CSUM)
3264 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3265 f1 &= f2;
3266 if (f1 & NETIF_F_GEN_CSUM)
3267 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3268
3269 return f1;
3270 }
3271
3272 static inline netdev_features_t netdev_get_wanted_features(
3273 struct net_device *dev)
3274 {
3275 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3276 }
3277 netdev_features_t netdev_increment_features(netdev_features_t all,
3278 netdev_features_t one, netdev_features_t mask);
3279
3280 /* Allow TSO being used on stacked device :
3281 * Performing the GSO segmentation before last device
3282 * is a performance improvement.
3283 */
3284 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3285 netdev_features_t mask)
3286 {
3287 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3288 }
3289
3290 int __netdev_update_features(struct net_device *dev);
3291 void netdev_update_features(struct net_device *dev);
3292 void netdev_change_features(struct net_device *dev);
3293
3294 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3295 struct net_device *dev);
3296
3297 netdev_features_t netif_skb_features(struct sk_buff *skb);
3298
3299 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3300 {
3301 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3302
3303 /* check flags correspondence */
3304 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3305 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3306 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3307 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3308 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3309 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3310 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3311 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3312 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3313 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3314 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3315 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3316 BUILD_BUG_ON(SKB_GSO_MPLS != (NETIF_F_GSO_MPLS >> NETIF_F_GSO_SHIFT));
3317
3318 return (features & feature) == feature;
3319 }
3320
3321 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3322 {
3323 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3324 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3325 }
3326
3327 static inline bool netif_needs_gso(struct sk_buff *skb,
3328 netdev_features_t features)
3329 {
3330 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3331 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3332 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3333 }
3334
3335 static inline void netif_set_gso_max_size(struct net_device *dev,
3336 unsigned int size)
3337 {
3338 dev->gso_max_size = size;
3339 }
3340
3341 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3342 int pulled_hlen, u16 mac_offset,
3343 int mac_len)
3344 {
3345 skb->protocol = protocol;
3346 skb->encapsulation = 1;
3347 skb_push(skb, pulled_hlen);
3348 skb_reset_transport_header(skb);
3349 skb->mac_header = mac_offset;
3350 skb->network_header = skb->mac_header + mac_len;
3351 skb->mac_len = mac_len;
3352 }
3353
3354 static inline bool netif_is_macvlan(struct net_device *dev)
3355 {
3356 return dev->priv_flags & IFF_MACVLAN;
3357 }
3358
3359 static inline bool netif_is_bond_master(struct net_device *dev)
3360 {
3361 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3362 }
3363
3364 static inline bool netif_is_bond_slave(struct net_device *dev)
3365 {
3366 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3367 }
3368
3369 static inline bool netif_supports_nofcs(struct net_device *dev)
3370 {
3371 return dev->priv_flags & IFF_SUPP_NOFCS;
3372 }
3373
3374 extern struct pernet_operations __net_initdata loopback_net_ops;
3375
3376 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3377
3378 /* netdev_printk helpers, similar to dev_printk */
3379
3380 static inline const char *netdev_name(const struct net_device *dev)
3381 {
3382 if (dev->reg_state != NETREG_REGISTERED)
3383 return "(unregistered net_device)";
3384 return dev->name;
3385 }
3386
3387 __printf(3, 4)
3388 int netdev_printk(const char *level, const struct net_device *dev,
3389 const char *format, ...);
3390 __printf(2, 3)
3391 int netdev_emerg(const struct net_device *dev, const char *format, ...);
3392 __printf(2, 3)
3393 int netdev_alert(const struct net_device *dev, const char *format, ...);
3394 __printf(2, 3)
3395 int netdev_crit(const struct net_device *dev, const char *format, ...);
3396 __printf(2, 3)
3397 int netdev_err(const struct net_device *dev, const char *format, ...);
3398 __printf(2, 3)
3399 int netdev_warn(const struct net_device *dev, const char *format, ...);
3400 __printf(2, 3)
3401 int netdev_notice(const struct net_device *dev, const char *format, ...);
3402 __printf(2, 3)
3403 int netdev_info(const struct net_device *dev, const char *format, ...);
3404
3405 #define MODULE_ALIAS_NETDEV(device) \
3406 MODULE_ALIAS("netdev-" device)
3407
3408 #if defined(CONFIG_DYNAMIC_DEBUG)
3409 #define netdev_dbg(__dev, format, args...) \
3410 do { \
3411 dynamic_netdev_dbg(__dev, format, ##args); \
3412 } while (0)
3413 #elif defined(DEBUG)
3414 #define netdev_dbg(__dev, format, args...) \
3415 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3416 #else
3417 #define netdev_dbg(__dev, format, args...) \
3418 ({ \
3419 if (0) \
3420 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3421 0; \
3422 })
3423 #endif
3424
3425 #if defined(VERBOSE_DEBUG)
3426 #define netdev_vdbg netdev_dbg
3427 #else
3428
3429 #define netdev_vdbg(dev, format, args...) \
3430 ({ \
3431 if (0) \
3432 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3433 0; \
3434 })
3435 #endif
3436
3437 /*
3438 * netdev_WARN() acts like dev_printk(), but with the key difference
3439 * of using a WARN/WARN_ON to get the message out, including the
3440 * file/line information and a backtrace.
3441 */
3442 #define netdev_WARN(dev, format, args...) \
3443 WARN(1, "netdevice: %s\n" format, netdev_name(dev), ##args)
3444
3445 /* netif printk helpers, similar to netdev_printk */
3446
3447 #define netif_printk(priv, type, level, dev, fmt, args...) \
3448 do { \
3449 if (netif_msg_##type(priv)) \
3450 netdev_printk(level, (dev), fmt, ##args); \
3451 } while (0)
3452
3453 #define netif_level(level, priv, type, dev, fmt, args...) \
3454 do { \
3455 if (netif_msg_##type(priv)) \
3456 netdev_##level(dev, fmt, ##args); \
3457 } while (0)
3458
3459 #define netif_emerg(priv, type, dev, fmt, args...) \
3460 netif_level(emerg, priv, type, dev, fmt, ##args)
3461 #define netif_alert(priv, type, dev, fmt, args...) \
3462 netif_level(alert, priv, type, dev, fmt, ##args)
3463 #define netif_crit(priv, type, dev, fmt, args...) \
3464 netif_level(crit, priv, type, dev, fmt, ##args)
3465 #define netif_err(priv, type, dev, fmt, args...) \
3466 netif_level(err, priv, type, dev, fmt, ##args)
3467 #define netif_warn(priv, type, dev, fmt, args...) \
3468 netif_level(warn, priv, type, dev, fmt, ##args)
3469 #define netif_notice(priv, type, dev, fmt, args...) \
3470 netif_level(notice, priv, type, dev, fmt, ##args)
3471 #define netif_info(priv, type, dev, fmt, args...) \
3472 netif_level(info, priv, type, dev, fmt, ##args)
3473
3474 #if defined(CONFIG_DYNAMIC_DEBUG)
3475 #define netif_dbg(priv, type, netdev, format, args...) \
3476 do { \
3477 if (netif_msg_##type(priv)) \
3478 dynamic_netdev_dbg(netdev, format, ##args); \
3479 } while (0)
3480 #elif defined(DEBUG)
3481 #define netif_dbg(priv, type, dev, format, args...) \
3482 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3483 #else
3484 #define netif_dbg(priv, type, dev, format, args...) \
3485 ({ \
3486 if (0) \
3487 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3488 0; \
3489 })
3490 #endif
3491
3492 #if defined(VERBOSE_DEBUG)
3493 #define netif_vdbg netif_dbg
3494 #else
3495 #define netif_vdbg(priv, type, dev, format, args...) \
3496 ({ \
3497 if (0) \
3498 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3499 0; \
3500 })
3501 #endif
3502
3503 /*
3504 * The list of packet types we will receive (as opposed to discard)
3505 * and the routines to invoke.
3506 *
3507 * Why 16. Because with 16 the only overlap we get on a hash of the
3508 * low nibble of the protocol value is RARP/SNAP/X.25.
3509 *
3510 * NOTE: That is no longer true with the addition of VLAN tags. Not
3511 * sure which should go first, but I bet it won't make much
3512 * difference if we are running VLANs. The good news is that
3513 * this protocol won't be in the list unless compiled in, so
3514 * the average user (w/out VLANs) will not be adversely affected.
3515 * --BLG
3516 *
3517 * 0800 IP
3518 * 8100 802.1Q VLAN
3519 * 0001 802.3
3520 * 0002 AX.25
3521 * 0004 802.2
3522 * 8035 RARP
3523 * 0005 SNAP
3524 * 0805 X.25
3525 * 0806 ARP
3526 * 8137 IPX
3527 * 0009 Localtalk
3528 * 86DD IPv6
3529 */
3530 #define PTYPE_HASH_SIZE (16)
3531 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3532
3533 #endif /* _LINUX_NETDEVICE_H */
This page took 0.18023 seconds and 5 git commands to generate.