Merge branch 'netns-next'
[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 <linux/prefetch.h>
34 #include <asm/cache.h>
35 #include <asm/byteorder.h>
36
37 #include <linux/percpu.h>
38 #include <linux/rculist.h>
39 #include <linux/dmaengine.h>
40 #include <linux/workqueue.h>
41 #include <linux/dynamic_queue_limits.h>
42
43 #include <linux/ethtool.h>
44 #include <net/net_namespace.h>
45 #include <net/dsa.h>
46 #ifdef CONFIG_DCB
47 #include <net/dcbnl.h>
48 #endif
49 #include <net/netprio_cgroup.h>
50
51 #include <linux/netdev_features.h>
52 #include <linux/neighbour.h>
53 #include <uapi/linux/netdevice.h>
54 #include <uapi/linux/if_bonding.h>
55
56 struct netpoll_info;
57 struct device;
58 struct phy_device;
59 /* 802.11 specific */
60 struct wireless_dev;
61 /* 802.15.4 specific */
62 struct wpan_dev;
63
64 void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
66
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
70
71 /*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
94
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
103
104 enum netdev_tx {
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
109 };
110 typedef enum netdev_tx netdev_tx_t;
111
112 /*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116 static inline bool dev_xmit_complete(int rc)
117 {
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128 }
129
130 /*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
134
135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
136 # if defined(CONFIG_MAC80211_MESH)
137 # define LL_MAX_HEADER 128
138 # else
139 # define LL_MAX_HEADER 96
140 # endif
141 #else
142 # define LL_MAX_HEADER 32
143 #endif
144
145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
147 #define MAX_HEADER LL_MAX_HEADER
148 #else
149 #define MAX_HEADER (LL_MAX_HEADER + 48)
150 #endif
151
152 /*
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
155 */
156
157 struct net_device_stats {
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
167 unsigned long collisions;
168 unsigned long rx_length_errors;
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181 };
182
183
184 #include <linux/cache.h>
185 #include <linux/skbuff.h>
186
187 #ifdef CONFIG_RPS
188 #include <linux/static_key.h>
189 extern struct static_key rps_needed;
190 #endif
191
192 struct neighbour;
193 struct neigh_parms;
194 struct sk_buff;
195
196 struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
200 #define NETDEV_HW_ADDR_T_LAN 1
201 #define NETDEV_HW_ADDR_T_SAN 2
202 #define NETDEV_HW_ADDR_T_SLAVE 3
203 #define NETDEV_HW_ADDR_T_UNICAST 4
204 #define NETDEV_HW_ADDR_T_MULTICAST 5
205 bool global_use;
206 int sync_cnt;
207 int refcount;
208 int synced;
209 struct rcu_head rcu_head;
210 };
211
212 struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215 };
216
217 #define netdev_hw_addr_list_count(l) ((l)->count)
218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219 #define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
221
222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224 #define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
226
227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
229 #define netdev_for_each_mc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
231
232 struct hh_cache {
233 u16 hh_len;
234 u16 __pad;
235 seqlock_t hh_lock;
236
237 /* cached hardware header; allow for machine alignment needs. */
238 #define HH_DATA_MOD 16
239 #define HH_DATA_OFF(__len) \
240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
241 #define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244 };
245
246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254 #define LL_RESERVED_SPACE(dev) \
255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
258
259 struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
262 const void *saddr, unsigned int len);
263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268 };
269
270 /* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
272 * code.
273 */
274
275 enum netdev_state_t {
276 __LINK_STATE_START,
277 __LINK_STATE_PRESENT,
278 __LINK_STATE_NOCARRIER,
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
281 };
282
283
284 /*
285 * This structure holds at boot time configured netdevice settings. They
286 * are then used in the device probing.
287 */
288 struct netdev_boot_setup {
289 char name[IFNAMSIZ];
290 struct ifmap map;
291 };
292 #define NETDEV_BOOT_SETUP_MAX 8
293
294 int __init netdev_boot_setup(char *str);
295
296 /*
297 * Structure for NAPI scheduling similar to tasklet but with weighting
298 */
299 struct napi_struct {
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
305 */
306 struct list_head poll_list;
307
308 unsigned long state;
309 int weight;
310 unsigned int gro_count;
311 int (*poll)(struct napi_struct *, int);
312 #ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
314 int poll_owner;
315 #endif
316 struct net_device *dev;
317 struct sk_buff *gro_list;
318 struct sk_buff *skb;
319 struct hrtimer timer;
320 struct list_head dev_list;
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
323 };
324
325 enum {
326 NAPI_STATE_SCHED, /* Poll is scheduled */
327 NAPI_STATE_DISABLE, /* Disable pending */
328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
329 NAPI_STATE_HASHED, /* In NAPI hash */
330 };
331
332 enum gro_result {
333 GRO_MERGED,
334 GRO_MERGED_FREE,
335 GRO_HELD,
336 GRO_NORMAL,
337 GRO_DROP,
338 };
339 typedef enum gro_result gro_result_t;
340
341 /*
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * further.
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 *
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
352 *
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
355 *
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
359 *
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * do with the skb.
362 *
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
366 *
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
370 *
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
373 * are registered on exact device (ptype->dev == skb->dev).
374 *
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
377 *
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
380 */
381
382 enum rx_handler_result {
383 RX_HANDLER_CONSUMED,
384 RX_HANDLER_ANOTHER,
385 RX_HANDLER_EXACT,
386 RX_HANDLER_PASS,
387 };
388 typedef enum rx_handler_result rx_handler_result_t;
389 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
390
391 void __napi_schedule(struct napi_struct *n);
392 void __napi_schedule_irqoff(struct napi_struct *n);
393
394 static inline bool napi_disable_pending(struct napi_struct *n)
395 {
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
397 }
398
399 /**
400 * napi_schedule_prep - check if napi can be scheduled
401 * @n: napi context
402 *
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
407 */
408 static inline bool napi_schedule_prep(struct napi_struct *n)
409 {
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
412 }
413
414 /**
415 * napi_schedule - schedule NAPI poll
416 * @n: napi context
417 *
418 * Schedule NAPI poll routine to be called if it is not already
419 * running.
420 */
421 static inline void napi_schedule(struct napi_struct *n)
422 {
423 if (napi_schedule_prep(n))
424 __napi_schedule(n);
425 }
426
427 /**
428 * napi_schedule_irqoff - schedule NAPI poll
429 * @n: napi context
430 *
431 * Variant of napi_schedule(), assuming hard irqs are masked.
432 */
433 static inline void napi_schedule_irqoff(struct napi_struct *n)
434 {
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
437 }
438
439 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
440 static inline bool napi_reschedule(struct napi_struct *napi)
441 {
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
444 return true;
445 }
446 return false;
447 }
448
449 void __napi_complete(struct napi_struct *n);
450 void napi_complete_done(struct napi_struct *n, int work_done);
451 /**
452 * napi_complete - NAPI processing complete
453 * @n: napi context
454 *
455 * Mark NAPI processing as complete.
456 * Consider using napi_complete_done() instead.
457 */
458 static inline void napi_complete(struct napi_struct *n)
459 {
460 return napi_complete_done(n, 0);
461 }
462
463 /**
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
466 *
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
469 */
470 struct napi_struct *napi_by_id(unsigned int napi_id);
471
472 /**
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
475 *
476 * generate a new napi_id and store a @napi under it in napi_hash
477 */
478 void napi_hash_add(struct napi_struct *napi);
479
480 /**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
483 *
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
486 */
487 void napi_hash_del(struct napi_struct *napi);
488
489 /**
490 * napi_disable - prevent NAPI from scheduling
491 * @n: napi context
492 *
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
495 */
496 void napi_disable(struct napi_struct *n);
497
498 /**
499 * napi_enable - enable NAPI scheduling
500 * @n: napi context
501 *
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
504 */
505 static inline void napi_enable(struct napi_struct *n)
506 {
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
508 smp_mb__before_atomic();
509 clear_bit(NAPI_STATE_SCHED, &n->state);
510 }
511
512 #ifdef CONFIG_SMP
513 /**
514 * napi_synchronize - wait until NAPI is not running
515 * @n: napi context
516 *
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
520 */
521 static inline void napi_synchronize(const struct napi_struct *n)
522 {
523 while (test_bit(NAPI_STATE_SCHED, &n->state))
524 msleep(1);
525 }
526 #else
527 # define napi_synchronize(n) barrier()
528 #endif
529
530 enum netdev_queue_state_t {
531 __QUEUE_STATE_DRV_XOFF,
532 __QUEUE_STATE_STACK_XOFF,
533 __QUEUE_STATE_FROZEN,
534 };
535
536 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
539
540 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
542 QUEUE_STATE_FROZEN)
543 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
544 QUEUE_STATE_FROZEN)
545
546 /*
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 */
555
556 struct netdev_queue {
557 /*
558 * read mostly part
559 */
560 struct net_device *dev;
561 struct Qdisc __rcu *qdisc;
562 struct Qdisc *qdisc_sleeping;
563 #ifdef CONFIG_SYSFS
564 struct kobject kobj;
565 #endif
566 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 int numa_node;
568 #endif
569 /*
570 * write mostly part
571 */
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
573 int xmit_lock_owner;
574 /*
575 * please use this field instead of dev->trans_start
576 */
577 unsigned long trans_start;
578
579 /*
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
582 */
583 unsigned long trans_timeout;
584
585 unsigned long state;
586
587 #ifdef CONFIG_BQL
588 struct dql dql;
589 #endif
590 unsigned long tx_maxrate;
591 } ____cacheline_aligned_in_smp;
592
593 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
594 {
595 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 return q->numa_node;
597 #else
598 return NUMA_NO_NODE;
599 #endif
600 }
601
602 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
603 {
604 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 q->numa_node = node;
606 #endif
607 }
608
609 #ifdef CONFIG_RPS
610 /*
611 * This structure holds an RPS map which can be of variable length. The
612 * map is an array of CPUs.
613 */
614 struct rps_map {
615 unsigned int len;
616 struct rcu_head rcu;
617 u16 cpus[0];
618 };
619 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
620
621 /*
622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
623 * tail pointer for that CPU's input queue at the time of last enqueue, and
624 * a hardware filter index.
625 */
626 struct rps_dev_flow {
627 u16 cpu;
628 u16 filter;
629 unsigned int last_qtail;
630 };
631 #define RPS_NO_FILTER 0xffff
632
633 /*
634 * The rps_dev_flow_table structure contains a table of flow mappings.
635 */
636 struct rps_dev_flow_table {
637 unsigned int mask;
638 struct rcu_head rcu;
639 struct rps_dev_flow flows[0];
640 };
641 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
642 ((_num) * sizeof(struct rps_dev_flow)))
643
644 /*
645 * The rps_sock_flow_table contains mappings of flows to the last CPU
646 * on which they were processed by the application (set in recvmsg).
647 * Each entry is a 32bit value. Upper part is the high order bits
648 * of flow hash, lower part is cpu number.
649 * rps_cpu_mask is used to partition the space, depending on number of
650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
652 * meaning we use 32-6=26 bits for the hash.
653 */
654 struct rps_sock_flow_table {
655 u32 mask;
656
657 u32 ents[0] ____cacheline_aligned_in_smp;
658 };
659 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
660
661 #define RPS_NO_CPU 0xffff
662
663 extern u32 rps_cpu_mask;
664 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665
666 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
667 u32 hash)
668 {
669 if (table && hash) {
670 unsigned int index = hash & table->mask;
671 u32 val = hash & ~rps_cpu_mask;
672
673 /* We only give a hint, preemption can change cpu under us */
674 val |= raw_smp_processor_id();
675
676 if (table->ents[index] != val)
677 table->ents[index] = val;
678 }
679 }
680
681 #ifdef CONFIG_RFS_ACCEL
682 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
683 u16 filter_id);
684 #endif
685 #endif /* CONFIG_RPS */
686
687 /* This structure contains an instance of an RX queue. */
688 struct netdev_rx_queue {
689 #ifdef CONFIG_RPS
690 struct rps_map __rcu *rps_map;
691 struct rps_dev_flow_table __rcu *rps_flow_table;
692 #endif
693 struct kobject kobj;
694 struct net_device *dev;
695 } ____cacheline_aligned_in_smp;
696
697 /*
698 * RX queue sysfs structures and functions.
699 */
700 struct rx_queue_attribute {
701 struct attribute attr;
702 ssize_t (*show)(struct netdev_rx_queue *queue,
703 struct rx_queue_attribute *attr, char *buf);
704 ssize_t (*store)(struct netdev_rx_queue *queue,
705 struct rx_queue_attribute *attr, const char *buf, size_t len);
706 };
707
708 #ifdef CONFIG_XPS
709 /*
710 * This structure holds an XPS map which can be of variable length. The
711 * map is an array of queues.
712 */
713 struct xps_map {
714 unsigned int len;
715 unsigned int alloc_len;
716 struct rcu_head rcu;
717 u16 queues[0];
718 };
719 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
720 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
721 / sizeof(u16))
722
723 /*
724 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 */
726 struct xps_dev_maps {
727 struct rcu_head rcu;
728 struct xps_map __rcu *cpu_map[0];
729 };
730 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
731 (nr_cpu_ids * sizeof(struct xps_map *)))
732 #endif /* CONFIG_XPS */
733
734 #define TC_MAX_QUEUE 16
735 #define TC_BITMASK 15
736 /* HW offloaded queuing disciplines txq count and offset maps */
737 struct netdev_tc_txq {
738 u16 count;
739 u16 offset;
740 };
741
742 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743 /*
744 * This structure is to hold information about the device
745 * configured to run FCoE protocol stack.
746 */
747 struct netdev_fcoe_hbainfo {
748 char manufacturer[64];
749 char serial_number[64];
750 char hardware_version[64];
751 char driver_version[64];
752 char optionrom_version[64];
753 char firmware_version[64];
754 char model[256];
755 char model_description[256];
756 };
757 #endif
758
759 #define MAX_PHYS_ITEM_ID_LEN 32
760
761 /* This structure holds a unique identifier to identify some
762 * physical item (port for example) used by a netdevice.
763 */
764 struct netdev_phys_item_id {
765 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
766 unsigned char id_len;
767 };
768
769 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 struct sk_buff *skb);
771
772 /*
773 * This structure defines the management hooks for network devices.
774 * The following hooks can be defined; unless noted otherwise, they are
775 * optional and can be filled with a null pointer.
776 *
777 * int (*ndo_init)(struct net_device *dev);
778 * This function is called once when network device is registered.
779 * The network device can use this to any late stage initializaton
780 * or semantic validattion. It can fail with an error code which will
781 * be propogated back to register_netdev
782 *
783 * void (*ndo_uninit)(struct net_device *dev);
784 * This function is called when device is unregistered or when registration
785 * fails. It is not called if init fails.
786 *
787 * int (*ndo_open)(struct net_device *dev);
788 * This function is called when network device transistions to the up
789 * state.
790 *
791 * int (*ndo_stop)(struct net_device *dev);
792 * This function is called when network device transistions to the down
793 * state.
794 *
795 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
796 * struct net_device *dev);
797 * Called when a packet needs to be transmitted.
798 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
799 * the queue before that can happen; it's for obsolete devices and weird
800 * corner cases, but the stack really does a non-trivial amount
801 * of useless work if you return NETDEV_TX_BUSY.
802 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
803 * Required can not be NULL.
804 *
805 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
806 * void *accel_priv, select_queue_fallback_t fallback);
807 * Called to decide which queue to when device supports multiple
808 * transmit queues.
809 *
810 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
811 * This function is called to allow device receiver to make
812 * changes to configuration when multicast or promiscious is enabled.
813 *
814 * void (*ndo_set_rx_mode)(struct net_device *dev);
815 * This function is called device changes address list filtering.
816 * If driver handles unicast address filtering, it should set
817 * IFF_UNICAST_FLT to its priv_flags.
818 *
819 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
820 * This function is called when the Media Access Control address
821 * needs to be changed. If this interface is not defined, the
822 * mac address can not be changed.
823 *
824 * int (*ndo_validate_addr)(struct net_device *dev);
825 * Test if Media Access Control address is valid for the device.
826 *
827 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
828 * Called when a user request an ioctl which can't be handled by
829 * the generic interface code. If not defined ioctl's return
830 * not supported error code.
831 *
832 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
833 * Used to set network devices bus interface parameters. This interface
834 * is retained for legacy reason, new devices should use the bus
835 * interface (PCI) for low level management.
836 *
837 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
838 * Called when a user wants to change the Maximum Transfer Unit
839 * of a device. If not defined, any request to change MTU will
840 * will return an error.
841 *
842 * void (*ndo_tx_timeout)(struct net_device *dev);
843 * Callback uses when the transmitter has not made any progress
844 * for dev->watchdog ticks.
845 *
846 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
847 * struct rtnl_link_stats64 *storage);
848 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
849 * Called when a user wants to get the network device usage
850 * statistics. Drivers must do one of the following:
851 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
852 * rtnl_link_stats64 structure passed by the caller.
853 * 2. Define @ndo_get_stats to update a net_device_stats structure
854 * (which should normally be dev->stats) and return a pointer to
855 * it. The structure may be changed asynchronously only if each
856 * field is written atomically.
857 * 3. Update dev->stats asynchronously and atomically, and define
858 * neither operation.
859 *
860 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
861 * If device support VLAN filtering this function is called when a
862 * VLAN id is registered.
863 *
864 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
865 * If device support VLAN filtering this function is called when a
866 * VLAN id is unregistered.
867 *
868 * void (*ndo_poll_controller)(struct net_device *dev);
869 *
870 * SR-IOV management functions.
871 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
872 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
873 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
874 * int max_tx_rate);
875 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
876 * int (*ndo_get_vf_config)(struct net_device *dev,
877 * int vf, struct ifla_vf_info *ivf);
878 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
879 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
880 * struct nlattr *port[]);
881 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
882 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
883 * Called to setup 'tc' number of traffic classes in the net device. This
884 * is always called from the stack with the rtnl lock held and netif tx
885 * queues stopped. This allows the netdevice to perform queue management
886 * safely.
887 *
888 * Fiber Channel over Ethernet (FCoE) offload functions.
889 * int (*ndo_fcoe_enable)(struct net_device *dev);
890 * Called when the FCoE protocol stack wants to start using LLD for FCoE
891 * so the underlying device can perform whatever needed configuration or
892 * initialization to support acceleration of FCoE traffic.
893 *
894 * int (*ndo_fcoe_disable)(struct net_device *dev);
895 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
896 * so the underlying device can perform whatever needed clean-ups to
897 * stop supporting acceleration of FCoE traffic.
898 *
899 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
900 * struct scatterlist *sgl, unsigned int sgc);
901 * Called when the FCoE Initiator wants to initialize an I/O that
902 * is a possible candidate for Direct Data Placement (DDP). The LLD can
903 * perform necessary setup and returns 1 to indicate the device is set up
904 * successfully to perform DDP on this I/O, otherwise this returns 0.
905 *
906 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
907 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
908 * indicated by the FC exchange id 'xid', so the underlying device can
909 * clean up and reuse resources for later DDP requests.
910 *
911 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
912 * struct scatterlist *sgl, unsigned int sgc);
913 * Called when the FCoE Target wants to initialize an I/O that
914 * is a possible candidate for Direct Data Placement (DDP). The LLD can
915 * perform necessary setup and returns 1 to indicate the device is set up
916 * successfully to perform DDP on this I/O, otherwise this returns 0.
917 *
918 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
919 * struct netdev_fcoe_hbainfo *hbainfo);
920 * Called when the FCoE Protocol stack wants information on the underlying
921 * device. This information is utilized by the FCoE protocol stack to
922 * register attributes with Fiber Channel management service as per the
923 * FC-GS Fabric Device Management Information(FDMI) specification.
924 *
925 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
926 * Called when the underlying device wants to override default World Wide
927 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
928 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
929 * protocol stack to use.
930 *
931 * RFS acceleration.
932 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
933 * u16 rxq_index, u32 flow_id);
934 * Set hardware filter for RFS. rxq_index is the target queue index;
935 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
936 * Return the filter ID on success, or a negative error code.
937 *
938 * Slave management functions (for bridge, bonding, etc).
939 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
940 * Called to make another netdev an underling.
941 *
942 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
943 * Called to release previously enslaved netdev.
944 *
945 * Feature/offload setting functions.
946 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
947 * netdev_features_t features);
948 * Adjusts the requested feature flags according to device-specific
949 * constraints, and returns the resulting flags. Must not modify
950 * the device state.
951 *
952 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
953 * Called to update device configuration to new features. Passed
954 * feature set might be less than what was returned by ndo_fix_features()).
955 * Must return >0 or -errno if it changed dev->features itself.
956 *
957 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
958 * struct net_device *dev,
959 * const unsigned char *addr, u16 vid, u16 flags)
960 * Adds an FDB entry to dev for addr.
961 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
962 * struct net_device *dev,
963 * const unsigned char *addr, u16 vid)
964 * Deletes the FDB entry from dev coresponding to addr.
965 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
966 * struct net_device *dev, struct net_device *filter_dev,
967 * int idx)
968 * Used to add FDB entries to dump requests. Implementers should add
969 * entries to skb and update idx with the number of entries.
970 *
971 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
972 * u16 flags)
973 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
974 * struct net_device *dev, u32 filter_mask)
975 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
976 * u16 flags);
977 *
978 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
979 * Called to change device carrier. Soft-devices (like dummy, team, etc)
980 * which do not represent real hardware may define this to allow their
981 * userspace components to manage their virtual carrier state. Devices
982 * that determine carrier state from physical hardware properties (eg
983 * network cables) or protocol-dependent mechanisms (eg
984 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
985 *
986 * int (*ndo_get_phys_port_id)(struct net_device *dev,
987 * struct netdev_phys_item_id *ppid);
988 * Called to get ID of physical port of this device. If driver does
989 * not implement this, it is assumed that the hw is not able to have
990 * multiple net devices on single physical port.
991 *
992 * void (*ndo_add_vxlan_port)(struct net_device *dev,
993 * sa_family_t sa_family, __be16 port);
994 * Called by vxlan to notiy a driver about the UDP port and socket
995 * address family that vxlan is listnening to. It is called only when
996 * a new port starts listening. The operation is protected by the
997 * vxlan_net->sock_lock.
998 *
999 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1000 * sa_family_t sa_family, __be16 port);
1001 * Called by vxlan to notify the driver about a UDP port and socket
1002 * address family that vxlan is not listening to anymore. The operation
1003 * is protected by the vxlan_net->sock_lock.
1004 *
1005 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1006 * struct net_device *dev)
1007 * Called by upper layer devices to accelerate switching or other
1008 * station functionality into hardware. 'pdev is the lowerdev
1009 * to use for the offload and 'dev' is the net device that will
1010 * back the offload. Returns a pointer to the private structure
1011 * the upper layer will maintain.
1012 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1013 * Called by upper layer device to delete the station created
1014 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1015 * the station and priv is the structure returned by the add
1016 * operation.
1017 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1018 * struct net_device *dev,
1019 * void *priv);
1020 * Callback to use for xmit over the accelerated station. This
1021 * is used in place of ndo_start_xmit on accelerated net
1022 * devices.
1023 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1024 * struct net_device *dev
1025 * netdev_features_t features);
1026 * Called by core transmit path to determine if device is capable of
1027 * performing offload operations on a given packet. This is to give
1028 * the device an opportunity to implement any restrictions that cannot
1029 * be otherwise expressed by feature flags. The check is called with
1030 * the set of features that the stack has calculated and it returns
1031 * those the driver believes to be appropriate.
1032 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1033 * int queue_index, u32 maxrate);
1034 * Called when a user wants to set a max-rate limitation of specific
1035 * TX queue.
1036 * int (*ndo_get_iflink)(const struct net_device *dev);
1037 * Called to get the iflink value of this device.
1038 */
1039 struct net_device_ops {
1040 int (*ndo_init)(struct net_device *dev);
1041 void (*ndo_uninit)(struct net_device *dev);
1042 int (*ndo_open)(struct net_device *dev);
1043 int (*ndo_stop)(struct net_device *dev);
1044 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1045 struct net_device *dev);
1046 u16 (*ndo_select_queue)(struct net_device *dev,
1047 struct sk_buff *skb,
1048 void *accel_priv,
1049 select_queue_fallback_t fallback);
1050 void (*ndo_change_rx_flags)(struct net_device *dev,
1051 int flags);
1052 void (*ndo_set_rx_mode)(struct net_device *dev);
1053 int (*ndo_set_mac_address)(struct net_device *dev,
1054 void *addr);
1055 int (*ndo_validate_addr)(struct net_device *dev);
1056 int (*ndo_do_ioctl)(struct net_device *dev,
1057 struct ifreq *ifr, int cmd);
1058 int (*ndo_set_config)(struct net_device *dev,
1059 struct ifmap *map);
1060 int (*ndo_change_mtu)(struct net_device *dev,
1061 int new_mtu);
1062 int (*ndo_neigh_setup)(struct net_device *dev,
1063 struct neigh_parms *);
1064 void (*ndo_tx_timeout) (struct net_device *dev);
1065
1066 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1067 struct rtnl_link_stats64 *storage);
1068 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1069
1070 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1071 __be16 proto, u16 vid);
1072 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1073 __be16 proto, u16 vid);
1074 #ifdef CONFIG_NET_POLL_CONTROLLER
1075 void (*ndo_poll_controller)(struct net_device *dev);
1076 int (*ndo_netpoll_setup)(struct net_device *dev,
1077 struct netpoll_info *info);
1078 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1079 #endif
1080 #ifdef CONFIG_NET_RX_BUSY_POLL
1081 int (*ndo_busy_poll)(struct napi_struct *dev);
1082 #endif
1083 int (*ndo_set_vf_mac)(struct net_device *dev,
1084 int queue, u8 *mac);
1085 int (*ndo_set_vf_vlan)(struct net_device *dev,
1086 int queue, u16 vlan, u8 qos);
1087 int (*ndo_set_vf_rate)(struct net_device *dev,
1088 int vf, int min_tx_rate,
1089 int max_tx_rate);
1090 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1091 int vf, bool setting);
1092 int (*ndo_get_vf_config)(struct net_device *dev,
1093 int vf,
1094 struct ifla_vf_info *ivf);
1095 int (*ndo_set_vf_link_state)(struct net_device *dev,
1096 int vf, int link_state);
1097 int (*ndo_set_vf_port)(struct net_device *dev,
1098 int vf,
1099 struct nlattr *port[]);
1100 int (*ndo_get_vf_port)(struct net_device *dev,
1101 int vf, struct sk_buff *skb);
1102 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1103 #if IS_ENABLED(CONFIG_FCOE)
1104 int (*ndo_fcoe_enable)(struct net_device *dev);
1105 int (*ndo_fcoe_disable)(struct net_device *dev);
1106 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1107 u16 xid,
1108 struct scatterlist *sgl,
1109 unsigned int sgc);
1110 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1111 u16 xid);
1112 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1113 u16 xid,
1114 struct scatterlist *sgl,
1115 unsigned int sgc);
1116 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1117 struct netdev_fcoe_hbainfo *hbainfo);
1118 #endif
1119
1120 #if IS_ENABLED(CONFIG_LIBFCOE)
1121 #define NETDEV_FCOE_WWNN 0
1122 #define NETDEV_FCOE_WWPN 1
1123 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1124 u64 *wwn, int type);
1125 #endif
1126
1127 #ifdef CONFIG_RFS_ACCEL
1128 int (*ndo_rx_flow_steer)(struct net_device *dev,
1129 const struct sk_buff *skb,
1130 u16 rxq_index,
1131 u32 flow_id);
1132 #endif
1133 int (*ndo_add_slave)(struct net_device *dev,
1134 struct net_device *slave_dev);
1135 int (*ndo_del_slave)(struct net_device *dev,
1136 struct net_device *slave_dev);
1137 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1138 netdev_features_t features);
1139 int (*ndo_set_features)(struct net_device *dev,
1140 netdev_features_t features);
1141 int (*ndo_neigh_construct)(struct neighbour *n);
1142 void (*ndo_neigh_destroy)(struct neighbour *n);
1143
1144 int (*ndo_fdb_add)(struct ndmsg *ndm,
1145 struct nlattr *tb[],
1146 struct net_device *dev,
1147 const unsigned char *addr,
1148 u16 vid,
1149 u16 flags);
1150 int (*ndo_fdb_del)(struct ndmsg *ndm,
1151 struct nlattr *tb[],
1152 struct net_device *dev,
1153 const unsigned char *addr,
1154 u16 vid);
1155 int (*ndo_fdb_dump)(struct sk_buff *skb,
1156 struct netlink_callback *cb,
1157 struct net_device *dev,
1158 struct net_device *filter_dev,
1159 int idx);
1160
1161 int (*ndo_bridge_setlink)(struct net_device *dev,
1162 struct nlmsghdr *nlh,
1163 u16 flags);
1164 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1165 u32 pid, u32 seq,
1166 struct net_device *dev,
1167 u32 filter_mask);
1168 int (*ndo_bridge_dellink)(struct net_device *dev,
1169 struct nlmsghdr *nlh,
1170 u16 flags);
1171 int (*ndo_change_carrier)(struct net_device *dev,
1172 bool new_carrier);
1173 int (*ndo_get_phys_port_id)(struct net_device *dev,
1174 struct netdev_phys_item_id *ppid);
1175 int (*ndo_get_phys_port_name)(struct net_device *dev,
1176 char *name, size_t len);
1177 void (*ndo_add_vxlan_port)(struct net_device *dev,
1178 sa_family_t sa_family,
1179 __be16 port);
1180 void (*ndo_del_vxlan_port)(struct net_device *dev,
1181 sa_family_t sa_family,
1182 __be16 port);
1183
1184 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1185 struct net_device *dev);
1186 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1187 void *priv);
1188
1189 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1190 struct net_device *dev,
1191 void *priv);
1192 int (*ndo_get_lock_subclass)(struct net_device *dev);
1193 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1194 struct net_device *dev,
1195 netdev_features_t features);
1196 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1197 int queue_index,
1198 u32 maxrate);
1199 int (*ndo_get_iflink)(const struct net_device *dev);
1200 };
1201
1202 /**
1203 * enum net_device_priv_flags - &struct net_device priv_flags
1204 *
1205 * These are the &struct net_device, they are only set internally
1206 * by drivers and used in the kernel. These flags are invisible to
1207 * userspace, this means that the order of these flags can change
1208 * during any kernel release.
1209 *
1210 * You should have a pretty good reason to be extending these flags.
1211 *
1212 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1213 * @IFF_EBRIDGE: Ethernet bridging device
1214 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1215 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1216 * @IFF_MASTER_ALB: bonding master, balance-alb
1217 * @IFF_BONDING: bonding master or slave
1218 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1219 * @IFF_ISATAP: ISATAP interface (RFC4214)
1220 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1221 * @IFF_WAN_HDLC: WAN HDLC device
1222 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1223 * release skb->dst
1224 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1225 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1226 * @IFF_MACVLAN_PORT: device used as macvlan port
1227 * @IFF_BRIDGE_PORT: device used as bridge port
1228 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1229 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1230 * @IFF_UNICAST_FLT: Supports unicast filtering
1231 * @IFF_TEAM_PORT: device used as team port
1232 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1233 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1234 * change when it's running
1235 * @IFF_MACVLAN: Macvlan device
1236 */
1237 enum netdev_priv_flags {
1238 IFF_802_1Q_VLAN = 1<<0,
1239 IFF_EBRIDGE = 1<<1,
1240 IFF_SLAVE_INACTIVE = 1<<2,
1241 IFF_MASTER_8023AD = 1<<3,
1242 IFF_MASTER_ALB = 1<<4,
1243 IFF_BONDING = 1<<5,
1244 IFF_SLAVE_NEEDARP = 1<<6,
1245 IFF_ISATAP = 1<<7,
1246 IFF_MASTER_ARPMON = 1<<8,
1247 IFF_WAN_HDLC = 1<<9,
1248 IFF_XMIT_DST_RELEASE = 1<<10,
1249 IFF_DONT_BRIDGE = 1<<11,
1250 IFF_DISABLE_NETPOLL = 1<<12,
1251 IFF_MACVLAN_PORT = 1<<13,
1252 IFF_BRIDGE_PORT = 1<<14,
1253 IFF_OVS_DATAPATH = 1<<15,
1254 IFF_TX_SKB_SHARING = 1<<16,
1255 IFF_UNICAST_FLT = 1<<17,
1256 IFF_TEAM_PORT = 1<<18,
1257 IFF_SUPP_NOFCS = 1<<19,
1258 IFF_LIVE_ADDR_CHANGE = 1<<20,
1259 IFF_MACVLAN = 1<<21,
1260 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1261 IFF_IPVLAN_MASTER = 1<<23,
1262 IFF_IPVLAN_SLAVE = 1<<24,
1263 };
1264
1265 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1266 #define IFF_EBRIDGE IFF_EBRIDGE
1267 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1268 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1269 #define IFF_MASTER_ALB IFF_MASTER_ALB
1270 #define IFF_BONDING IFF_BONDING
1271 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1272 #define IFF_ISATAP IFF_ISATAP
1273 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1274 #define IFF_WAN_HDLC IFF_WAN_HDLC
1275 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1276 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1277 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1278 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1279 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1280 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1281 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1282 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1283 #define IFF_TEAM_PORT IFF_TEAM_PORT
1284 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1285 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1286 #define IFF_MACVLAN IFF_MACVLAN
1287 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1288 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1289 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1290
1291 /**
1292 * struct net_device - The DEVICE structure.
1293 * Actually, this whole structure is a big mistake. It mixes I/O
1294 * data with strictly "high-level" data, and it has to know about
1295 * almost every data structure used in the INET module.
1296 *
1297 * @name: This is the first field of the "visible" part of this structure
1298 * (i.e. as seen by users in the "Space.c" file). It is the name
1299 * of the interface.
1300 *
1301 * @name_hlist: Device name hash chain, please keep it close to name[]
1302 * @ifalias: SNMP alias
1303 * @mem_end: Shared memory end
1304 * @mem_start: Shared memory start
1305 * @base_addr: Device I/O address
1306 * @irq: Device IRQ number
1307 *
1308 * @state: Generic network queuing layer state, see netdev_state_t
1309 * @dev_list: The global list of network devices
1310 * @napi_list: List entry, that is used for polling napi devices
1311 * @unreg_list: List entry, that is used, when we are unregistering the
1312 * device, see the function unregister_netdev
1313 * @close_list: List entry, that is used, when we are closing the device
1314 *
1315 * @adj_list: Directly linked devices, like slaves for bonding
1316 * @all_adj_list: All linked devices, *including* neighbours
1317 * @features: Currently active device features
1318 * @hw_features: User-changeable features
1319 *
1320 * @wanted_features: User-requested features
1321 * @vlan_features: Mask of features inheritable by VLAN devices
1322 *
1323 * @hw_enc_features: Mask of features inherited by encapsulating devices
1324 * This field indicates what encapsulation
1325 * offloads the hardware is capable of doing,
1326 * and drivers will need to set them appropriately.
1327 *
1328 * @mpls_features: Mask of features inheritable by MPLS
1329 *
1330 * @ifindex: interface index
1331 * @iflink: unique device identifier
1332 *
1333 * @stats: Statistics struct, which was left as a legacy, use
1334 * rtnl_link_stats64 instead
1335 *
1336 * @rx_dropped: Dropped packets by core network,
1337 * do not use this in drivers
1338 * @tx_dropped: Dropped packets by core network,
1339 * do not use this in drivers
1340 *
1341 * @carrier_changes: Stats to monitor carrier on<->off transitions
1342 *
1343 * @wireless_handlers: List of functions to handle Wireless Extensions,
1344 * instead of ioctl,
1345 * see <net/iw_handler.h> for details.
1346 * @wireless_data: Instance data managed by the core of wireless extensions
1347 *
1348 * @netdev_ops: Includes several pointers to callbacks,
1349 * if one wants to override the ndo_*() functions
1350 * @ethtool_ops: Management operations
1351 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1352 * of Layer 2 headers.
1353 *
1354 * @flags: Interface flags (a la BSD)
1355 * @priv_flags: Like 'flags' but invisible to userspace,
1356 * see if.h for the definitions
1357 * @gflags: Global flags ( kept as legacy )
1358 * @padded: How much padding added by alloc_netdev()
1359 * @operstate: RFC2863 operstate
1360 * @link_mode: Mapping policy to operstate
1361 * @if_port: Selectable AUI, TP, ...
1362 * @dma: DMA channel
1363 * @mtu: Interface MTU value
1364 * @type: Interface hardware type
1365 * @hard_header_len: Hardware header length
1366 *
1367 * @needed_headroom: Extra headroom the hardware may need, but not in all
1368 * cases can this be guaranteed
1369 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1370 * cases can this be guaranteed. Some cases also use
1371 * LL_MAX_HEADER instead to allocate the skb
1372 *
1373 * interface address info:
1374 *
1375 * @perm_addr: Permanent hw address
1376 * @addr_assign_type: Hw address assignment type
1377 * @addr_len: Hardware address length
1378 * @neigh_priv_len; Used in neigh_alloc(),
1379 * initialized only in atm/clip.c
1380 * @dev_id: Used to differentiate devices that share
1381 * the same link layer address
1382 * @dev_port: Used to differentiate devices that share
1383 * the same function
1384 * @addr_list_lock: XXX: need comments on this one
1385 * @uc: unicast mac addresses
1386 * @mc: multicast mac addresses
1387 * @dev_addrs: list of device hw addresses
1388 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1389 * @uc_promisc: Counter, that indicates, that promiscuous mode
1390 * has been enabled due to the need to listen to
1391 * additional unicast addresses in a device that
1392 * does not implement ndo_set_rx_mode()
1393 * @promiscuity: Number of times, the NIC is told to work in
1394 * Promiscuous mode, if it becomes 0 the NIC will
1395 * exit from working in Promiscuous mode
1396 * @allmulti: Counter, enables or disables allmulticast mode
1397 *
1398 * @vlan_info: VLAN info
1399 * @dsa_ptr: dsa specific data
1400 * @tipc_ptr: TIPC specific data
1401 * @atalk_ptr: AppleTalk link
1402 * @ip_ptr: IPv4 specific data
1403 * @dn_ptr: DECnet specific data
1404 * @ip6_ptr: IPv6 specific data
1405 * @ax25_ptr: AX.25 specific data
1406 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1407 *
1408 * @last_rx: Time of last Rx
1409 * @dev_addr: Hw address (before bcast,
1410 * because most packets are unicast)
1411 *
1412 * @_rx: Array of RX queues
1413 * @num_rx_queues: Number of RX queues
1414 * allocated at register_netdev() time
1415 * @real_num_rx_queues: Number of RX queues currently active in device
1416 *
1417 * @rx_handler: handler for received packets
1418 * @rx_handler_data: XXX: need comments on this one
1419 * @ingress_queue: XXX: need comments on this one
1420 * @broadcast: hw bcast address
1421 *
1422 * @_tx: Array of TX queues
1423 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1424 * @real_num_tx_queues: Number of TX queues currently active in device
1425 * @qdisc: Root qdisc from userspace point of view
1426 * @tx_queue_len: Max frames per queue allowed
1427 * @tx_global_lock: XXX: need comments on this one
1428 *
1429 * @xps_maps: XXX: need comments on this one
1430 *
1431 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1432 * indexed by RX queue number. Assigned by driver.
1433 * This must only be set if the ndo_rx_flow_steer
1434 * operation is defined
1435 *
1436 * @trans_start: Time (in jiffies) of last Tx
1437 * @watchdog_timeo: Represents the timeout that is used by
1438 * the watchdog ( see dev_watchdog() )
1439 * @watchdog_timer: List of timers
1440 *
1441 * @pcpu_refcnt: Number of references to this device
1442 * @todo_list: Delayed register/unregister
1443 * @index_hlist: Device index hash chain
1444 * @link_watch_list: XXX: need comments on this one
1445 *
1446 * @reg_state: Register/unregister state machine
1447 * @dismantle: Device is going to be freed
1448 * @rtnl_link_state: This enum represents the phases of creating
1449 * a new link
1450 *
1451 * @destructor: Called from unregister,
1452 * can be used to call free_netdev
1453 * @npinfo: XXX: need comments on this one
1454 * @nd_net: Network namespace this network device is inside
1455 *
1456 * @ml_priv: Mid-layer private
1457 * @lstats: Loopback statistics
1458 * @tstats: Tunnel statistics
1459 * @dstats: Dummy statistics
1460 * @vstats: Virtual ethernet statistics
1461 *
1462 * @garp_port: GARP
1463 * @mrp_port: MRP
1464 *
1465 * @dev: Class/net/name entry
1466 * @sysfs_groups: Space for optional device, statistics and wireless
1467 * sysfs groups
1468 *
1469 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1470 * @rtnl_link_ops: Rtnl_link_ops
1471 *
1472 * @gso_max_size: Maximum size of generic segmentation offload
1473 * @gso_max_segs: Maximum number of segments that can be passed to the
1474 * NIC for GSO
1475 * @gso_min_segs: Minimum number of segments that can be passed to the
1476 * NIC for GSO
1477 *
1478 * @dcbnl_ops: Data Center Bridging netlink ops
1479 * @num_tc: Number of traffic classes in the net device
1480 * @tc_to_txq: XXX: need comments on this one
1481 * @prio_tc_map XXX: need comments on this one
1482 *
1483 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1484 *
1485 * @priomap: XXX: need comments on this one
1486 * @phydev: Physical device may attach itself
1487 * for hardware timestamping
1488 *
1489 * @qdisc_tx_busylock: XXX: need comments on this one
1490 *
1491 * @group: The group, that the device belongs to
1492 * @pm_qos_req: Power Management QoS object
1493 *
1494 * FIXME: cleanup struct net_device such that network protocol info
1495 * moves out.
1496 */
1497
1498 struct net_device {
1499 char name[IFNAMSIZ];
1500 struct hlist_node name_hlist;
1501 char *ifalias;
1502 /*
1503 * I/O specific fields
1504 * FIXME: Merge these and struct ifmap into one
1505 */
1506 unsigned long mem_end;
1507 unsigned long mem_start;
1508 unsigned long base_addr;
1509 int irq;
1510
1511 /*
1512 * Some hardware also needs these fields (state,dev_list,
1513 * napi_list,unreg_list,close_list) but they are not
1514 * part of the usual set specified in Space.c.
1515 */
1516
1517 unsigned long state;
1518
1519 struct list_head dev_list;
1520 struct list_head napi_list;
1521 struct list_head unreg_list;
1522 struct list_head close_list;
1523 struct list_head ptype_all;
1524 struct list_head ptype_specific;
1525
1526 struct {
1527 struct list_head upper;
1528 struct list_head lower;
1529 } adj_list;
1530
1531 struct {
1532 struct list_head upper;
1533 struct list_head lower;
1534 } all_adj_list;
1535
1536 netdev_features_t features;
1537 netdev_features_t hw_features;
1538 netdev_features_t wanted_features;
1539 netdev_features_t vlan_features;
1540 netdev_features_t hw_enc_features;
1541 netdev_features_t mpls_features;
1542
1543 int ifindex;
1544 int group;
1545
1546 struct net_device_stats stats;
1547
1548 atomic_long_t rx_dropped;
1549 atomic_long_t tx_dropped;
1550
1551 atomic_t carrier_changes;
1552
1553 #ifdef CONFIG_WIRELESS_EXT
1554 const struct iw_handler_def * wireless_handlers;
1555 struct iw_public_data * wireless_data;
1556 #endif
1557 const struct net_device_ops *netdev_ops;
1558 const struct ethtool_ops *ethtool_ops;
1559 #ifdef CONFIG_NET_SWITCHDEV
1560 const struct swdev_ops *swdev_ops;
1561 #endif
1562
1563 const struct header_ops *header_ops;
1564
1565 unsigned int flags;
1566 unsigned int priv_flags;
1567
1568 unsigned short gflags;
1569 unsigned short padded;
1570
1571 unsigned char operstate;
1572 unsigned char link_mode;
1573
1574 unsigned char if_port;
1575 unsigned char dma;
1576
1577 unsigned int mtu;
1578 unsigned short type;
1579 unsigned short hard_header_len;
1580
1581 unsigned short needed_headroom;
1582 unsigned short needed_tailroom;
1583
1584 /* Interface address info. */
1585 unsigned char perm_addr[MAX_ADDR_LEN];
1586 unsigned char addr_assign_type;
1587 unsigned char addr_len;
1588 unsigned short neigh_priv_len;
1589 unsigned short dev_id;
1590 unsigned short dev_port;
1591 spinlock_t addr_list_lock;
1592 struct netdev_hw_addr_list uc;
1593 struct netdev_hw_addr_list mc;
1594 struct netdev_hw_addr_list dev_addrs;
1595
1596 #ifdef CONFIG_SYSFS
1597 struct kset *queues_kset;
1598 #endif
1599
1600 unsigned char name_assign_type;
1601
1602 bool uc_promisc;
1603 unsigned int promiscuity;
1604 unsigned int allmulti;
1605
1606
1607 /* Protocol specific pointers */
1608
1609 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1610 struct vlan_info __rcu *vlan_info;
1611 #endif
1612 #if IS_ENABLED(CONFIG_NET_DSA)
1613 struct dsa_switch_tree *dsa_ptr;
1614 #endif
1615 #if IS_ENABLED(CONFIG_TIPC)
1616 struct tipc_bearer __rcu *tipc_ptr;
1617 #endif
1618 void *atalk_ptr;
1619 struct in_device __rcu *ip_ptr;
1620 struct dn_dev __rcu *dn_ptr;
1621 struct inet6_dev __rcu *ip6_ptr;
1622 void *ax25_ptr;
1623 struct wireless_dev *ieee80211_ptr;
1624 struct wpan_dev *ieee802154_ptr;
1625
1626 /*
1627 * Cache lines mostly used on receive path (including eth_type_trans())
1628 */
1629 unsigned long last_rx;
1630
1631 /* Interface address info used in eth_type_trans() */
1632 unsigned char *dev_addr;
1633
1634
1635 #ifdef CONFIG_SYSFS
1636 struct netdev_rx_queue *_rx;
1637
1638 unsigned int num_rx_queues;
1639 unsigned int real_num_rx_queues;
1640
1641 #endif
1642
1643 unsigned long gro_flush_timeout;
1644 rx_handler_func_t __rcu *rx_handler;
1645 void __rcu *rx_handler_data;
1646
1647 struct netdev_queue __rcu *ingress_queue;
1648 unsigned char broadcast[MAX_ADDR_LEN];
1649
1650
1651 /*
1652 * Cache lines mostly used on transmit path
1653 */
1654 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1655 unsigned int num_tx_queues;
1656 unsigned int real_num_tx_queues;
1657 struct Qdisc *qdisc;
1658 unsigned long tx_queue_len;
1659 spinlock_t tx_global_lock;
1660
1661 #ifdef CONFIG_XPS
1662 struct xps_dev_maps __rcu *xps_maps;
1663 #endif
1664 #ifdef CONFIG_RFS_ACCEL
1665 struct cpu_rmap *rx_cpu_rmap;
1666 #endif
1667
1668 /* These may be needed for future network-power-down code. */
1669
1670 /*
1671 * trans_start here is expensive for high speed devices on SMP,
1672 * please use netdev_queue->trans_start instead.
1673 */
1674 unsigned long trans_start;
1675
1676 int watchdog_timeo;
1677 struct timer_list watchdog_timer;
1678
1679 int __percpu *pcpu_refcnt;
1680 struct list_head todo_list;
1681
1682 struct hlist_node index_hlist;
1683 struct list_head link_watch_list;
1684
1685 enum { NETREG_UNINITIALIZED=0,
1686 NETREG_REGISTERED, /* completed register_netdevice */
1687 NETREG_UNREGISTERING, /* called unregister_netdevice */
1688 NETREG_UNREGISTERED, /* completed unregister todo */
1689 NETREG_RELEASED, /* called free_netdev */
1690 NETREG_DUMMY, /* dummy device for NAPI poll */
1691 } reg_state:8;
1692
1693 bool dismantle;
1694
1695 enum {
1696 RTNL_LINK_INITIALIZED,
1697 RTNL_LINK_INITIALIZING,
1698 } rtnl_link_state:16;
1699
1700 void (*destructor)(struct net_device *dev);
1701
1702 #ifdef CONFIG_NETPOLL
1703 struct netpoll_info __rcu *npinfo;
1704 #endif
1705
1706 possible_net_t nd_net;
1707
1708 /* mid-layer private */
1709 union {
1710 void *ml_priv;
1711 struct pcpu_lstats __percpu *lstats;
1712 struct pcpu_sw_netstats __percpu *tstats;
1713 struct pcpu_dstats __percpu *dstats;
1714 struct pcpu_vstats __percpu *vstats;
1715 };
1716
1717 struct garp_port __rcu *garp_port;
1718 struct mrp_port __rcu *mrp_port;
1719
1720 struct device dev;
1721 const struct attribute_group *sysfs_groups[4];
1722 const struct attribute_group *sysfs_rx_queue_group;
1723
1724 const struct rtnl_link_ops *rtnl_link_ops;
1725
1726 /* for setting kernel sock attribute on TCP connection setup */
1727 #define GSO_MAX_SIZE 65536
1728 unsigned int gso_max_size;
1729 #define GSO_MAX_SEGS 65535
1730 u16 gso_max_segs;
1731 u16 gso_min_segs;
1732 #ifdef CONFIG_DCB
1733 const struct dcbnl_rtnl_ops *dcbnl_ops;
1734 #endif
1735 u8 num_tc;
1736 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1737 u8 prio_tc_map[TC_BITMASK + 1];
1738
1739 #if IS_ENABLED(CONFIG_FCOE)
1740 unsigned int fcoe_ddp_xid;
1741 #endif
1742 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1743 struct netprio_map __rcu *priomap;
1744 #endif
1745 struct phy_device *phydev;
1746 struct lock_class_key *qdisc_tx_busylock;
1747 struct pm_qos_request pm_qos_req;
1748 };
1749 #define to_net_dev(d) container_of(d, struct net_device, dev)
1750
1751 #define NETDEV_ALIGN 32
1752
1753 static inline
1754 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1755 {
1756 return dev->prio_tc_map[prio & TC_BITMASK];
1757 }
1758
1759 static inline
1760 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1761 {
1762 if (tc >= dev->num_tc)
1763 return -EINVAL;
1764
1765 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1766 return 0;
1767 }
1768
1769 static inline
1770 void netdev_reset_tc(struct net_device *dev)
1771 {
1772 dev->num_tc = 0;
1773 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1774 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1775 }
1776
1777 static inline
1778 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1779 {
1780 if (tc >= dev->num_tc)
1781 return -EINVAL;
1782
1783 dev->tc_to_txq[tc].count = count;
1784 dev->tc_to_txq[tc].offset = offset;
1785 return 0;
1786 }
1787
1788 static inline
1789 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1790 {
1791 if (num_tc > TC_MAX_QUEUE)
1792 return -EINVAL;
1793
1794 dev->num_tc = num_tc;
1795 return 0;
1796 }
1797
1798 static inline
1799 int netdev_get_num_tc(struct net_device *dev)
1800 {
1801 return dev->num_tc;
1802 }
1803
1804 static inline
1805 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1806 unsigned int index)
1807 {
1808 return &dev->_tx[index];
1809 }
1810
1811 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1812 const struct sk_buff *skb)
1813 {
1814 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1815 }
1816
1817 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1818 void (*f)(struct net_device *,
1819 struct netdev_queue *,
1820 void *),
1821 void *arg)
1822 {
1823 unsigned int i;
1824
1825 for (i = 0; i < dev->num_tx_queues; i++)
1826 f(dev, &dev->_tx[i], arg);
1827 }
1828
1829 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1830 struct sk_buff *skb,
1831 void *accel_priv);
1832
1833 /*
1834 * Net namespace inlines
1835 */
1836 static inline
1837 struct net *dev_net(const struct net_device *dev)
1838 {
1839 return read_pnet(&dev->nd_net);
1840 }
1841
1842 static inline
1843 void dev_net_set(struct net_device *dev, struct net *net)
1844 {
1845 write_pnet(&dev->nd_net, net);
1846 }
1847
1848 static inline bool netdev_uses_dsa(struct net_device *dev)
1849 {
1850 #if IS_ENABLED(CONFIG_NET_DSA)
1851 if (dev->dsa_ptr != NULL)
1852 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1853 #endif
1854 return false;
1855 }
1856
1857 /**
1858 * netdev_priv - access network device private data
1859 * @dev: network device
1860 *
1861 * Get network device private data
1862 */
1863 static inline void *netdev_priv(const struct net_device *dev)
1864 {
1865 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1866 }
1867
1868 /* Set the sysfs physical device reference for the network logical device
1869 * if set prior to registration will cause a symlink during initialization.
1870 */
1871 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1872
1873 /* Set the sysfs device type for the network logical device to allow
1874 * fine-grained identification of different network device types. For
1875 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1876 */
1877 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1878
1879 /* Default NAPI poll() weight
1880 * Device drivers are strongly advised to not use bigger value
1881 */
1882 #define NAPI_POLL_WEIGHT 64
1883
1884 /**
1885 * netif_napi_add - initialize a napi context
1886 * @dev: network device
1887 * @napi: napi context
1888 * @poll: polling function
1889 * @weight: default weight
1890 *
1891 * netif_napi_add() must be used to initialize a napi context prior to calling
1892 * *any* of the other napi related functions.
1893 */
1894 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1895 int (*poll)(struct napi_struct *, int), int weight);
1896
1897 /**
1898 * netif_napi_del - remove a napi context
1899 * @napi: napi context
1900 *
1901 * netif_napi_del() removes a napi context from the network device napi list
1902 */
1903 void netif_napi_del(struct napi_struct *napi);
1904
1905 struct napi_gro_cb {
1906 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1907 void *frag0;
1908
1909 /* Length of frag0. */
1910 unsigned int frag0_len;
1911
1912 /* This indicates where we are processing relative to skb->data. */
1913 int data_offset;
1914
1915 /* This is non-zero if the packet cannot be merged with the new skb. */
1916 u16 flush;
1917
1918 /* Save the IP ID here and check when we get to the transport layer */
1919 u16 flush_id;
1920
1921 /* Number of segments aggregated. */
1922 u16 count;
1923
1924 /* Start offset for remote checksum offload */
1925 u16 gro_remcsum_start;
1926
1927 /* jiffies when first packet was created/queued */
1928 unsigned long age;
1929
1930 /* Used in ipv6_gro_receive() and foo-over-udp */
1931 u16 proto;
1932
1933 /* This is non-zero if the packet may be of the same flow. */
1934 u8 same_flow:1;
1935
1936 /* Used in udp_gro_receive */
1937 u8 udp_mark:1;
1938
1939 /* GRO checksum is valid */
1940 u8 csum_valid:1;
1941
1942 /* Number of checksums via CHECKSUM_UNNECESSARY */
1943 u8 csum_cnt:3;
1944
1945 /* Free the skb? */
1946 u8 free:2;
1947 #define NAPI_GRO_FREE 1
1948 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1949
1950 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1951 u8 is_ipv6:1;
1952
1953 /* 7 bit hole */
1954
1955 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1956 __wsum csum;
1957
1958 /* used in skb_gro_receive() slow path */
1959 struct sk_buff *last;
1960 };
1961
1962 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1963
1964 struct packet_type {
1965 __be16 type; /* This is really htons(ether_type). */
1966 struct net_device *dev; /* NULL is wildcarded here */
1967 int (*func) (struct sk_buff *,
1968 struct net_device *,
1969 struct packet_type *,
1970 struct net_device *);
1971 bool (*id_match)(struct packet_type *ptype,
1972 struct sock *sk);
1973 void *af_packet_priv;
1974 struct list_head list;
1975 };
1976
1977 struct offload_callbacks {
1978 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1979 netdev_features_t features);
1980 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1981 struct sk_buff *skb);
1982 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1983 };
1984
1985 struct packet_offload {
1986 __be16 type; /* This is really htons(ether_type). */
1987 struct offload_callbacks callbacks;
1988 struct list_head list;
1989 };
1990
1991 struct udp_offload;
1992
1993 struct udp_offload_callbacks {
1994 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1995 struct sk_buff *skb,
1996 struct udp_offload *uoff);
1997 int (*gro_complete)(struct sk_buff *skb,
1998 int nhoff,
1999 struct udp_offload *uoff);
2000 };
2001
2002 struct udp_offload {
2003 __be16 port;
2004 u8 ipproto;
2005 struct udp_offload_callbacks callbacks;
2006 };
2007
2008 /* often modified stats are per cpu, other are shared (netdev->stats) */
2009 struct pcpu_sw_netstats {
2010 u64 rx_packets;
2011 u64 rx_bytes;
2012 u64 tx_packets;
2013 u64 tx_bytes;
2014 struct u64_stats_sync syncp;
2015 };
2016
2017 #define netdev_alloc_pcpu_stats(type) \
2018 ({ \
2019 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2020 if (pcpu_stats) { \
2021 int i; \
2022 for_each_possible_cpu(i) { \
2023 typeof(type) *stat; \
2024 stat = per_cpu_ptr(pcpu_stats, i); \
2025 u64_stats_init(&stat->syncp); \
2026 } \
2027 } \
2028 pcpu_stats; \
2029 })
2030
2031 #include <linux/notifier.h>
2032
2033 /* netdevice notifier chain. Please remember to update the rtnetlink
2034 * notification exclusion list in rtnetlink_event() when adding new
2035 * types.
2036 */
2037 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2038 #define NETDEV_DOWN 0x0002
2039 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2040 detected a hardware crash and restarted
2041 - we can use this eg to kick tcp sessions
2042 once done */
2043 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2044 #define NETDEV_REGISTER 0x0005
2045 #define NETDEV_UNREGISTER 0x0006
2046 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2047 #define NETDEV_CHANGEADDR 0x0008
2048 #define NETDEV_GOING_DOWN 0x0009
2049 #define NETDEV_CHANGENAME 0x000A
2050 #define NETDEV_FEAT_CHANGE 0x000B
2051 #define NETDEV_BONDING_FAILOVER 0x000C
2052 #define NETDEV_PRE_UP 0x000D
2053 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2054 #define NETDEV_POST_TYPE_CHANGE 0x000F
2055 #define NETDEV_POST_INIT 0x0010
2056 #define NETDEV_UNREGISTER_FINAL 0x0011
2057 #define NETDEV_RELEASE 0x0012
2058 #define NETDEV_NOTIFY_PEERS 0x0013
2059 #define NETDEV_JOIN 0x0014
2060 #define NETDEV_CHANGEUPPER 0x0015
2061 #define NETDEV_RESEND_IGMP 0x0016
2062 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2063 #define NETDEV_CHANGEINFODATA 0x0018
2064 #define NETDEV_BONDING_INFO 0x0019
2065
2066 int register_netdevice_notifier(struct notifier_block *nb);
2067 int unregister_netdevice_notifier(struct notifier_block *nb);
2068
2069 struct netdev_notifier_info {
2070 struct net_device *dev;
2071 };
2072
2073 struct netdev_notifier_change_info {
2074 struct netdev_notifier_info info; /* must be first */
2075 unsigned int flags_changed;
2076 };
2077
2078 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2079 struct net_device *dev)
2080 {
2081 info->dev = dev;
2082 }
2083
2084 static inline struct net_device *
2085 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2086 {
2087 return info->dev;
2088 }
2089
2090 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2091
2092
2093 extern rwlock_t dev_base_lock; /* Device list lock */
2094
2095 #define for_each_netdev(net, d) \
2096 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2097 #define for_each_netdev_reverse(net, d) \
2098 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2099 #define for_each_netdev_rcu(net, d) \
2100 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2101 #define for_each_netdev_safe(net, d, n) \
2102 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2103 #define for_each_netdev_continue(net, d) \
2104 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2105 #define for_each_netdev_continue_rcu(net, d) \
2106 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2107 #define for_each_netdev_in_bond_rcu(bond, slave) \
2108 for_each_netdev_rcu(&init_net, slave) \
2109 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2110 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2111
2112 static inline struct net_device *next_net_device(struct net_device *dev)
2113 {
2114 struct list_head *lh;
2115 struct net *net;
2116
2117 net = dev_net(dev);
2118 lh = dev->dev_list.next;
2119 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2120 }
2121
2122 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2123 {
2124 struct list_head *lh;
2125 struct net *net;
2126
2127 net = dev_net(dev);
2128 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2129 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2130 }
2131
2132 static inline struct net_device *first_net_device(struct net *net)
2133 {
2134 return list_empty(&net->dev_base_head) ? NULL :
2135 net_device_entry(net->dev_base_head.next);
2136 }
2137
2138 static inline struct net_device *first_net_device_rcu(struct net *net)
2139 {
2140 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2141
2142 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2143 }
2144
2145 int netdev_boot_setup_check(struct net_device *dev);
2146 unsigned long netdev_boot_base(const char *prefix, int unit);
2147 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2148 const char *hwaddr);
2149 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2150 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2151 void dev_add_pack(struct packet_type *pt);
2152 void dev_remove_pack(struct packet_type *pt);
2153 void __dev_remove_pack(struct packet_type *pt);
2154 void dev_add_offload(struct packet_offload *po);
2155 void dev_remove_offload(struct packet_offload *po);
2156
2157 int dev_get_iflink(const struct net_device *dev);
2158 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2159 unsigned short mask);
2160 struct net_device *dev_get_by_name(struct net *net, const char *name);
2161 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2162 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2163 int dev_alloc_name(struct net_device *dev, const char *name);
2164 int dev_open(struct net_device *dev);
2165 int dev_close(struct net_device *dev);
2166 int dev_close_many(struct list_head *head, bool unlink);
2167 void dev_disable_lro(struct net_device *dev);
2168 int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb);
2169 int dev_queue_xmit_sk(struct sock *sk, struct sk_buff *skb);
2170 static inline int dev_queue_xmit(struct sk_buff *skb)
2171 {
2172 return dev_queue_xmit_sk(skb->sk, skb);
2173 }
2174 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2175 int register_netdevice(struct net_device *dev);
2176 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2177 void unregister_netdevice_many(struct list_head *head);
2178 static inline void unregister_netdevice(struct net_device *dev)
2179 {
2180 unregister_netdevice_queue(dev, NULL);
2181 }
2182
2183 int netdev_refcnt_read(const struct net_device *dev);
2184 void free_netdev(struct net_device *dev);
2185 void netdev_freemem(struct net_device *dev);
2186 void synchronize_net(void);
2187 int init_dummy_netdev(struct net_device *dev);
2188
2189 DECLARE_PER_CPU(int, xmit_recursion);
2190 static inline int dev_recursion_level(void)
2191 {
2192 return this_cpu_read(xmit_recursion);
2193 }
2194
2195 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2196 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2197 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2198 int netdev_get_name(struct net *net, char *name, int ifindex);
2199 int dev_restart(struct net_device *dev);
2200 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2201
2202 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2203 {
2204 return NAPI_GRO_CB(skb)->data_offset;
2205 }
2206
2207 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2208 {
2209 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2210 }
2211
2212 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2213 {
2214 NAPI_GRO_CB(skb)->data_offset += len;
2215 }
2216
2217 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2218 unsigned int offset)
2219 {
2220 return NAPI_GRO_CB(skb)->frag0 + offset;
2221 }
2222
2223 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2224 {
2225 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2226 }
2227
2228 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2229 unsigned int offset)
2230 {
2231 if (!pskb_may_pull(skb, hlen))
2232 return NULL;
2233
2234 NAPI_GRO_CB(skb)->frag0 = NULL;
2235 NAPI_GRO_CB(skb)->frag0_len = 0;
2236 return skb->data + offset;
2237 }
2238
2239 static inline void *skb_gro_network_header(struct sk_buff *skb)
2240 {
2241 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2242 skb_network_offset(skb);
2243 }
2244
2245 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2246 const void *start, unsigned int len)
2247 {
2248 if (NAPI_GRO_CB(skb)->csum_valid)
2249 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2250 csum_partial(start, len, 0));
2251 }
2252
2253 /* GRO checksum functions. These are logical equivalents of the normal
2254 * checksum functions (in skbuff.h) except that they operate on the GRO
2255 * offsets and fields in sk_buff.
2256 */
2257
2258 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2259
2260 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2261 {
2262 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2263 skb_gro_offset(skb));
2264 }
2265
2266 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2267 bool zero_okay,
2268 __sum16 check)
2269 {
2270 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2271 skb_checksum_start_offset(skb) <
2272 skb_gro_offset(skb)) &&
2273 !skb_at_gro_remcsum_start(skb) &&
2274 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2275 (!zero_okay || check));
2276 }
2277
2278 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2279 __wsum psum)
2280 {
2281 if (NAPI_GRO_CB(skb)->csum_valid &&
2282 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2283 return 0;
2284
2285 NAPI_GRO_CB(skb)->csum = psum;
2286
2287 return __skb_gro_checksum_complete(skb);
2288 }
2289
2290 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2291 {
2292 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2293 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2294 NAPI_GRO_CB(skb)->csum_cnt--;
2295 } else {
2296 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2297 * verified a new top level checksum or an encapsulated one
2298 * during GRO. This saves work if we fallback to normal path.
2299 */
2300 __skb_incr_checksum_unnecessary(skb);
2301 }
2302 }
2303
2304 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2305 compute_pseudo) \
2306 ({ \
2307 __sum16 __ret = 0; \
2308 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2309 __ret = __skb_gro_checksum_validate_complete(skb, \
2310 compute_pseudo(skb, proto)); \
2311 if (__ret) \
2312 __skb_mark_checksum_bad(skb); \
2313 else \
2314 skb_gro_incr_csum_unnecessary(skb); \
2315 __ret; \
2316 })
2317
2318 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2319 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2320
2321 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2322 compute_pseudo) \
2323 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2324
2325 #define skb_gro_checksum_simple_validate(skb) \
2326 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2327
2328 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2329 {
2330 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2331 !NAPI_GRO_CB(skb)->csum_valid);
2332 }
2333
2334 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2335 __sum16 check, __wsum pseudo)
2336 {
2337 NAPI_GRO_CB(skb)->csum = ~pseudo;
2338 NAPI_GRO_CB(skb)->csum_valid = 1;
2339 }
2340
2341 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2342 do { \
2343 if (__skb_gro_checksum_convert_check(skb)) \
2344 __skb_gro_checksum_convert(skb, check, \
2345 compute_pseudo(skb, proto)); \
2346 } while (0)
2347
2348 struct gro_remcsum {
2349 int offset;
2350 __wsum delta;
2351 };
2352
2353 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2354 {
2355 grc->offset = 0;
2356 grc->delta = 0;
2357 }
2358
2359 static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2360 int start, int offset,
2361 struct gro_remcsum *grc,
2362 bool nopartial)
2363 {
2364 __wsum delta;
2365
2366 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2367
2368 if (!nopartial) {
2369 NAPI_GRO_CB(skb)->gro_remcsum_start =
2370 ((unsigned char *)ptr + start) - skb->head;
2371 return;
2372 }
2373
2374 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2375
2376 /* Adjust skb->csum since we changed the packet */
2377 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2378
2379 grc->offset = (ptr + offset) - (void *)skb->head;
2380 grc->delta = delta;
2381 }
2382
2383 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2384 struct gro_remcsum *grc)
2385 {
2386 if (!grc->delta)
2387 return;
2388
2389 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2390 }
2391
2392 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2393 unsigned short type,
2394 const void *daddr, const void *saddr,
2395 unsigned int len)
2396 {
2397 if (!dev->header_ops || !dev->header_ops->create)
2398 return 0;
2399
2400 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2401 }
2402
2403 static inline int dev_parse_header(const struct sk_buff *skb,
2404 unsigned char *haddr)
2405 {
2406 const struct net_device *dev = skb->dev;
2407
2408 if (!dev->header_ops || !dev->header_ops->parse)
2409 return 0;
2410 return dev->header_ops->parse(skb, haddr);
2411 }
2412
2413 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2414 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2415 static inline int unregister_gifconf(unsigned int family)
2416 {
2417 return register_gifconf(family, NULL);
2418 }
2419
2420 #ifdef CONFIG_NET_FLOW_LIMIT
2421 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2422 struct sd_flow_limit {
2423 u64 count;
2424 unsigned int num_buckets;
2425 unsigned int history_head;
2426 u16 history[FLOW_LIMIT_HISTORY];
2427 u8 buckets[];
2428 };
2429
2430 extern int netdev_flow_limit_table_len;
2431 #endif /* CONFIG_NET_FLOW_LIMIT */
2432
2433 /*
2434 * Incoming packets are placed on per-cpu queues
2435 */
2436 struct softnet_data {
2437 struct list_head poll_list;
2438 struct sk_buff_head process_queue;
2439
2440 /* stats */
2441 unsigned int processed;
2442 unsigned int time_squeeze;
2443 unsigned int cpu_collision;
2444 unsigned int received_rps;
2445 #ifdef CONFIG_RPS
2446 struct softnet_data *rps_ipi_list;
2447 #endif
2448 #ifdef CONFIG_NET_FLOW_LIMIT
2449 struct sd_flow_limit __rcu *flow_limit;
2450 #endif
2451 struct Qdisc *output_queue;
2452 struct Qdisc **output_queue_tailp;
2453 struct sk_buff *completion_queue;
2454
2455 #ifdef CONFIG_RPS
2456 /* Elements below can be accessed between CPUs for RPS */
2457 struct call_single_data csd ____cacheline_aligned_in_smp;
2458 struct softnet_data *rps_ipi_next;
2459 unsigned int cpu;
2460 unsigned int input_queue_head;
2461 unsigned int input_queue_tail;
2462 #endif
2463 unsigned int dropped;
2464 struct sk_buff_head input_pkt_queue;
2465 struct napi_struct backlog;
2466
2467 };
2468
2469 static inline void input_queue_head_incr(struct softnet_data *sd)
2470 {
2471 #ifdef CONFIG_RPS
2472 sd->input_queue_head++;
2473 #endif
2474 }
2475
2476 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2477 unsigned int *qtail)
2478 {
2479 #ifdef CONFIG_RPS
2480 *qtail = ++sd->input_queue_tail;
2481 #endif
2482 }
2483
2484 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2485
2486 void __netif_schedule(struct Qdisc *q);
2487 void netif_schedule_queue(struct netdev_queue *txq);
2488
2489 static inline void netif_tx_schedule_all(struct net_device *dev)
2490 {
2491 unsigned int i;
2492
2493 for (i = 0; i < dev->num_tx_queues; i++)
2494 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2495 }
2496
2497 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2498 {
2499 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2500 }
2501
2502 /**
2503 * netif_start_queue - allow transmit
2504 * @dev: network device
2505 *
2506 * Allow upper layers to call the device hard_start_xmit routine.
2507 */
2508 static inline void netif_start_queue(struct net_device *dev)
2509 {
2510 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2511 }
2512
2513 static inline void netif_tx_start_all_queues(struct net_device *dev)
2514 {
2515 unsigned int i;
2516
2517 for (i = 0; i < dev->num_tx_queues; i++) {
2518 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2519 netif_tx_start_queue(txq);
2520 }
2521 }
2522
2523 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2524
2525 /**
2526 * netif_wake_queue - restart transmit
2527 * @dev: network device
2528 *
2529 * Allow upper layers to call the device hard_start_xmit routine.
2530 * Used for flow control when transmit resources are available.
2531 */
2532 static inline void netif_wake_queue(struct net_device *dev)
2533 {
2534 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2535 }
2536
2537 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2538 {
2539 unsigned int i;
2540
2541 for (i = 0; i < dev->num_tx_queues; i++) {
2542 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2543 netif_tx_wake_queue(txq);
2544 }
2545 }
2546
2547 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2548 {
2549 if (WARN_ON(!dev_queue)) {
2550 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2551 return;
2552 }
2553 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2554 }
2555
2556 /**
2557 * netif_stop_queue - stop transmitted packets
2558 * @dev: network device
2559 *
2560 * Stop upper layers calling the device hard_start_xmit routine.
2561 * Used for flow control when transmit resources are unavailable.
2562 */
2563 static inline void netif_stop_queue(struct net_device *dev)
2564 {
2565 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2566 }
2567
2568 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2569 {
2570 unsigned int i;
2571
2572 for (i = 0; i < dev->num_tx_queues; i++) {
2573 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2574 netif_tx_stop_queue(txq);
2575 }
2576 }
2577
2578 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2579 {
2580 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2581 }
2582
2583 /**
2584 * netif_queue_stopped - test if transmit queue is flowblocked
2585 * @dev: network device
2586 *
2587 * Test if transmit queue on device is currently unable to send.
2588 */
2589 static inline bool netif_queue_stopped(const struct net_device *dev)
2590 {
2591 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2592 }
2593
2594 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2595 {
2596 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2597 }
2598
2599 static inline bool
2600 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2601 {
2602 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2603 }
2604
2605 static inline bool
2606 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2607 {
2608 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2609 }
2610
2611 /**
2612 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2613 * @dev_queue: pointer to transmit queue
2614 *
2615 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2616 * to give appropriate hint to the cpu.
2617 */
2618 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2619 {
2620 #ifdef CONFIG_BQL
2621 prefetchw(&dev_queue->dql.num_queued);
2622 #endif
2623 }
2624
2625 /**
2626 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2627 * @dev_queue: pointer to transmit queue
2628 *
2629 * BQL enabled drivers might use this helper in their TX completion path,
2630 * to give appropriate hint to the cpu.
2631 */
2632 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2633 {
2634 #ifdef CONFIG_BQL
2635 prefetchw(&dev_queue->dql.limit);
2636 #endif
2637 }
2638
2639 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2640 unsigned int bytes)
2641 {
2642 #ifdef CONFIG_BQL
2643 dql_queued(&dev_queue->dql, bytes);
2644
2645 if (likely(dql_avail(&dev_queue->dql) >= 0))
2646 return;
2647
2648 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2649
2650 /*
2651 * The XOFF flag must be set before checking the dql_avail below,
2652 * because in netdev_tx_completed_queue we update the dql_completed
2653 * before checking the XOFF flag.
2654 */
2655 smp_mb();
2656
2657 /* check again in case another CPU has just made room avail */
2658 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2659 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2660 #endif
2661 }
2662
2663 /**
2664 * netdev_sent_queue - report the number of bytes queued to hardware
2665 * @dev: network device
2666 * @bytes: number of bytes queued to the hardware device queue
2667 *
2668 * Report the number of bytes queued for sending/completion to the network
2669 * device hardware queue. @bytes should be a good approximation and should
2670 * exactly match netdev_completed_queue() @bytes
2671 */
2672 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2673 {
2674 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2675 }
2676
2677 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2678 unsigned int pkts, unsigned int bytes)
2679 {
2680 #ifdef CONFIG_BQL
2681 if (unlikely(!bytes))
2682 return;
2683
2684 dql_completed(&dev_queue->dql, bytes);
2685
2686 /*
2687 * Without the memory barrier there is a small possiblity that
2688 * netdev_tx_sent_queue will miss the update and cause the queue to
2689 * be stopped forever
2690 */
2691 smp_mb();
2692
2693 if (dql_avail(&dev_queue->dql) < 0)
2694 return;
2695
2696 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2697 netif_schedule_queue(dev_queue);
2698 #endif
2699 }
2700
2701 /**
2702 * netdev_completed_queue - report bytes and packets completed by device
2703 * @dev: network device
2704 * @pkts: actual number of packets sent over the medium
2705 * @bytes: actual number of bytes sent over the medium
2706 *
2707 * Report the number of bytes and packets transmitted by the network device
2708 * hardware queue over the physical medium, @bytes must exactly match the
2709 * @bytes amount passed to netdev_sent_queue()
2710 */
2711 static inline void netdev_completed_queue(struct net_device *dev,
2712 unsigned int pkts, unsigned int bytes)
2713 {
2714 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2715 }
2716
2717 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2718 {
2719 #ifdef CONFIG_BQL
2720 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2721 dql_reset(&q->dql);
2722 #endif
2723 }
2724
2725 /**
2726 * netdev_reset_queue - reset the packets and bytes count of a network device
2727 * @dev_queue: network device
2728 *
2729 * Reset the bytes and packet count of a network device and clear the
2730 * software flow control OFF bit for this network device
2731 */
2732 static inline void netdev_reset_queue(struct net_device *dev_queue)
2733 {
2734 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2735 }
2736
2737 /**
2738 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2739 * @dev: network device
2740 * @queue_index: given tx queue index
2741 *
2742 * Returns 0 if given tx queue index >= number of device tx queues,
2743 * otherwise returns the originally passed tx queue index.
2744 */
2745 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2746 {
2747 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2748 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2749 dev->name, queue_index,
2750 dev->real_num_tx_queues);
2751 return 0;
2752 }
2753
2754 return queue_index;
2755 }
2756
2757 /**
2758 * netif_running - test if up
2759 * @dev: network device
2760 *
2761 * Test if the device has been brought up.
2762 */
2763 static inline bool netif_running(const struct net_device *dev)
2764 {
2765 return test_bit(__LINK_STATE_START, &dev->state);
2766 }
2767
2768 /*
2769 * Routines to manage the subqueues on a device. We only need start
2770 * stop, and a check if it's stopped. All other device management is
2771 * done at the overall netdevice level.
2772 * Also test the device if we're multiqueue.
2773 */
2774
2775 /**
2776 * netif_start_subqueue - allow sending packets on subqueue
2777 * @dev: network device
2778 * @queue_index: sub queue index
2779 *
2780 * Start individual transmit queue of a device with multiple transmit queues.
2781 */
2782 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2783 {
2784 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2785
2786 netif_tx_start_queue(txq);
2787 }
2788
2789 /**
2790 * netif_stop_subqueue - stop sending packets on subqueue
2791 * @dev: network device
2792 * @queue_index: sub queue index
2793 *
2794 * Stop individual transmit queue of a device with multiple transmit queues.
2795 */
2796 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2797 {
2798 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2799 netif_tx_stop_queue(txq);
2800 }
2801
2802 /**
2803 * netif_subqueue_stopped - test status of subqueue
2804 * @dev: network device
2805 * @queue_index: sub queue index
2806 *
2807 * Check individual transmit queue of a device with multiple transmit queues.
2808 */
2809 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2810 u16 queue_index)
2811 {
2812 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2813
2814 return netif_tx_queue_stopped(txq);
2815 }
2816
2817 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2818 struct sk_buff *skb)
2819 {
2820 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2821 }
2822
2823 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2824
2825 #ifdef CONFIG_XPS
2826 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2827 u16 index);
2828 #else
2829 static inline int netif_set_xps_queue(struct net_device *dev,
2830 const struct cpumask *mask,
2831 u16 index)
2832 {
2833 return 0;
2834 }
2835 #endif
2836
2837 /*
2838 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2839 * as a distribution range limit for the returned value.
2840 */
2841 static inline u16 skb_tx_hash(const struct net_device *dev,
2842 struct sk_buff *skb)
2843 {
2844 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2845 }
2846
2847 /**
2848 * netif_is_multiqueue - test if device has multiple transmit queues
2849 * @dev: network device
2850 *
2851 * Check if device has multiple transmit queues
2852 */
2853 static inline bool netif_is_multiqueue(const struct net_device *dev)
2854 {
2855 return dev->num_tx_queues > 1;
2856 }
2857
2858 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2859
2860 #ifdef CONFIG_SYSFS
2861 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2862 #else
2863 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2864 unsigned int rxq)
2865 {
2866 return 0;
2867 }
2868 #endif
2869
2870 #ifdef CONFIG_SYSFS
2871 static inline unsigned int get_netdev_rx_queue_index(
2872 struct netdev_rx_queue *queue)
2873 {
2874 struct net_device *dev = queue->dev;
2875 int index = queue - dev->_rx;
2876
2877 BUG_ON(index >= dev->num_rx_queues);
2878 return index;
2879 }
2880 #endif
2881
2882 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2883 int netif_get_num_default_rss_queues(void);
2884
2885 enum skb_free_reason {
2886 SKB_REASON_CONSUMED,
2887 SKB_REASON_DROPPED,
2888 };
2889
2890 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2891 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2892
2893 /*
2894 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2895 * interrupt context or with hardware interrupts being disabled.
2896 * (in_irq() || irqs_disabled())
2897 *
2898 * We provide four helpers that can be used in following contexts :
2899 *
2900 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2901 * replacing kfree_skb(skb)
2902 *
2903 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2904 * Typically used in place of consume_skb(skb) in TX completion path
2905 *
2906 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2907 * replacing kfree_skb(skb)
2908 *
2909 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2910 * and consumed a packet. Used in place of consume_skb(skb)
2911 */
2912 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2913 {
2914 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2915 }
2916
2917 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2918 {
2919 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2920 }
2921
2922 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2923 {
2924 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2925 }
2926
2927 static inline void dev_consume_skb_any(struct sk_buff *skb)
2928 {
2929 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2930 }
2931
2932 int netif_rx(struct sk_buff *skb);
2933 int netif_rx_ni(struct sk_buff *skb);
2934 int netif_receive_skb_sk(struct sock *sk, struct sk_buff *skb);
2935 static inline int netif_receive_skb(struct sk_buff *skb)
2936 {
2937 return netif_receive_skb_sk(skb->sk, skb);
2938 }
2939 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2940 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2941 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2942 gro_result_t napi_gro_frags(struct napi_struct *napi);
2943 struct packet_offload *gro_find_receive_by_type(__be16 type);
2944 struct packet_offload *gro_find_complete_by_type(__be16 type);
2945
2946 static inline void napi_free_frags(struct napi_struct *napi)
2947 {
2948 kfree_skb(napi->skb);
2949 napi->skb = NULL;
2950 }
2951
2952 int netdev_rx_handler_register(struct net_device *dev,
2953 rx_handler_func_t *rx_handler,
2954 void *rx_handler_data);
2955 void netdev_rx_handler_unregister(struct net_device *dev);
2956
2957 bool dev_valid_name(const char *name);
2958 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2959 int dev_ethtool(struct net *net, struct ifreq *);
2960 unsigned int dev_get_flags(const struct net_device *);
2961 int __dev_change_flags(struct net_device *, unsigned int flags);
2962 int dev_change_flags(struct net_device *, unsigned int);
2963 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2964 unsigned int gchanges);
2965 int dev_change_name(struct net_device *, const char *);
2966 int dev_set_alias(struct net_device *, const char *, size_t);
2967 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2968 int dev_set_mtu(struct net_device *, int);
2969 void dev_set_group(struct net_device *, int);
2970 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2971 int dev_change_carrier(struct net_device *, bool new_carrier);
2972 int dev_get_phys_port_id(struct net_device *dev,
2973 struct netdev_phys_item_id *ppid);
2974 int dev_get_phys_port_name(struct net_device *dev,
2975 char *name, size_t len);
2976 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2977 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2978 struct netdev_queue *txq, int *ret);
2979 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2980 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2981 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2982
2983 extern int netdev_budget;
2984
2985 /* Called by rtnetlink.c:rtnl_unlock() */
2986 void netdev_run_todo(void);
2987
2988 /**
2989 * dev_put - release reference to device
2990 * @dev: network device
2991 *
2992 * Release reference to device to allow it to be freed.
2993 */
2994 static inline void dev_put(struct net_device *dev)
2995 {
2996 this_cpu_dec(*dev->pcpu_refcnt);
2997 }
2998
2999 /**
3000 * dev_hold - get reference to device
3001 * @dev: network device
3002 *
3003 * Hold reference to device to keep it from being freed.
3004 */
3005 static inline void dev_hold(struct net_device *dev)
3006 {
3007 this_cpu_inc(*dev->pcpu_refcnt);
3008 }
3009
3010 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3011 * and _off may be called from IRQ context, but it is caller
3012 * who is responsible for serialization of these calls.
3013 *
3014 * The name carrier is inappropriate, these functions should really be
3015 * called netif_lowerlayer_*() because they represent the state of any
3016 * kind of lower layer not just hardware media.
3017 */
3018
3019 void linkwatch_init_dev(struct net_device *dev);
3020 void linkwatch_fire_event(struct net_device *dev);
3021 void linkwatch_forget_dev(struct net_device *dev);
3022
3023 /**
3024 * netif_carrier_ok - test if carrier present
3025 * @dev: network device
3026 *
3027 * Check if carrier is present on device
3028 */
3029 static inline bool netif_carrier_ok(const struct net_device *dev)
3030 {
3031 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3032 }
3033
3034 unsigned long dev_trans_start(struct net_device *dev);
3035
3036 void __netdev_watchdog_up(struct net_device *dev);
3037
3038 void netif_carrier_on(struct net_device *dev);
3039
3040 void netif_carrier_off(struct net_device *dev);
3041
3042 /**
3043 * netif_dormant_on - mark device as dormant.
3044 * @dev: network device
3045 *
3046 * Mark device as dormant (as per RFC2863).
3047 *
3048 * The dormant state indicates that the relevant interface is not
3049 * actually in a condition to pass packets (i.e., it is not 'up') but is
3050 * in a "pending" state, waiting for some external event. For "on-
3051 * demand" interfaces, this new state identifies the situation where the
3052 * interface is waiting for events to place it in the up state.
3053 *
3054 */
3055 static inline void netif_dormant_on(struct net_device *dev)
3056 {
3057 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3058 linkwatch_fire_event(dev);
3059 }
3060
3061 /**
3062 * netif_dormant_off - set device as not dormant.
3063 * @dev: network device
3064 *
3065 * Device is not in dormant state.
3066 */
3067 static inline void netif_dormant_off(struct net_device *dev)
3068 {
3069 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3070 linkwatch_fire_event(dev);
3071 }
3072
3073 /**
3074 * netif_dormant - test if carrier present
3075 * @dev: network device
3076 *
3077 * Check if carrier is present on device
3078 */
3079 static inline bool netif_dormant(const struct net_device *dev)
3080 {
3081 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3082 }
3083
3084
3085 /**
3086 * netif_oper_up - test if device is operational
3087 * @dev: network device
3088 *
3089 * Check if carrier is operational
3090 */
3091 static inline bool netif_oper_up(const struct net_device *dev)
3092 {
3093 return (dev->operstate == IF_OPER_UP ||
3094 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3095 }
3096
3097 /**
3098 * netif_device_present - is device available or removed
3099 * @dev: network device
3100 *
3101 * Check if device has not been removed from system.
3102 */
3103 static inline bool netif_device_present(struct net_device *dev)
3104 {
3105 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3106 }
3107
3108 void netif_device_detach(struct net_device *dev);
3109
3110 void netif_device_attach(struct net_device *dev);
3111
3112 /*
3113 * Network interface message level settings
3114 */
3115
3116 enum {
3117 NETIF_MSG_DRV = 0x0001,
3118 NETIF_MSG_PROBE = 0x0002,
3119 NETIF_MSG_LINK = 0x0004,
3120 NETIF_MSG_TIMER = 0x0008,
3121 NETIF_MSG_IFDOWN = 0x0010,
3122 NETIF_MSG_IFUP = 0x0020,
3123 NETIF_MSG_RX_ERR = 0x0040,
3124 NETIF_MSG_TX_ERR = 0x0080,
3125 NETIF_MSG_TX_QUEUED = 0x0100,
3126 NETIF_MSG_INTR = 0x0200,
3127 NETIF_MSG_TX_DONE = 0x0400,
3128 NETIF_MSG_RX_STATUS = 0x0800,
3129 NETIF_MSG_PKTDATA = 0x1000,
3130 NETIF_MSG_HW = 0x2000,
3131 NETIF_MSG_WOL = 0x4000,
3132 };
3133
3134 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3135 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3136 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3137 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3138 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3139 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3140 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3141 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3142 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3143 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3144 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3145 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3146 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3147 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3148 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3149
3150 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3151 {
3152 /* use default */
3153 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3154 return default_msg_enable_bits;
3155 if (debug_value == 0) /* no output */
3156 return 0;
3157 /* set low N bits */
3158 return (1 << debug_value) - 1;
3159 }
3160
3161 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3162 {
3163 spin_lock(&txq->_xmit_lock);
3164 txq->xmit_lock_owner = cpu;
3165 }
3166
3167 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3168 {
3169 spin_lock_bh(&txq->_xmit_lock);
3170 txq->xmit_lock_owner = smp_processor_id();
3171 }
3172
3173 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3174 {
3175 bool ok = spin_trylock(&txq->_xmit_lock);
3176 if (likely(ok))
3177 txq->xmit_lock_owner = smp_processor_id();
3178 return ok;
3179 }
3180
3181 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3182 {
3183 txq->xmit_lock_owner = -1;
3184 spin_unlock(&txq->_xmit_lock);
3185 }
3186
3187 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3188 {
3189 txq->xmit_lock_owner = -1;
3190 spin_unlock_bh(&txq->_xmit_lock);
3191 }
3192
3193 static inline void txq_trans_update(struct netdev_queue *txq)
3194 {
3195 if (txq->xmit_lock_owner != -1)
3196 txq->trans_start = jiffies;
3197 }
3198
3199 /**
3200 * netif_tx_lock - grab network device transmit lock
3201 * @dev: network device
3202 *
3203 * Get network device transmit lock
3204 */
3205 static inline void netif_tx_lock(struct net_device *dev)
3206 {
3207 unsigned int i;
3208 int cpu;
3209
3210 spin_lock(&dev->tx_global_lock);
3211 cpu = smp_processor_id();
3212 for (i = 0; i < dev->num_tx_queues; i++) {
3213 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3214
3215 /* We are the only thread of execution doing a
3216 * freeze, but we have to grab the _xmit_lock in
3217 * order to synchronize with threads which are in
3218 * the ->hard_start_xmit() handler and already
3219 * checked the frozen bit.
3220 */
3221 __netif_tx_lock(txq, cpu);
3222 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3223 __netif_tx_unlock(txq);
3224 }
3225 }
3226
3227 static inline void netif_tx_lock_bh(struct net_device *dev)
3228 {
3229 local_bh_disable();
3230 netif_tx_lock(dev);
3231 }
3232
3233 static inline void netif_tx_unlock(struct net_device *dev)
3234 {
3235 unsigned int i;
3236
3237 for (i = 0; i < dev->num_tx_queues; i++) {
3238 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3239
3240 /* No need to grab the _xmit_lock here. If the
3241 * queue is not stopped for another reason, we
3242 * force a schedule.
3243 */
3244 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3245 netif_schedule_queue(txq);
3246 }
3247 spin_unlock(&dev->tx_global_lock);
3248 }
3249
3250 static inline void netif_tx_unlock_bh(struct net_device *dev)
3251 {
3252 netif_tx_unlock(dev);
3253 local_bh_enable();
3254 }
3255
3256 #define HARD_TX_LOCK(dev, txq, cpu) { \
3257 if ((dev->features & NETIF_F_LLTX) == 0) { \
3258 __netif_tx_lock(txq, cpu); \
3259 } \
3260 }
3261
3262 #define HARD_TX_TRYLOCK(dev, txq) \
3263 (((dev->features & NETIF_F_LLTX) == 0) ? \
3264 __netif_tx_trylock(txq) : \
3265 true )
3266
3267 #define HARD_TX_UNLOCK(dev, txq) { \
3268 if ((dev->features & NETIF_F_LLTX) == 0) { \
3269 __netif_tx_unlock(txq); \
3270 } \
3271 }
3272
3273 static inline void netif_tx_disable(struct net_device *dev)
3274 {
3275 unsigned int i;
3276 int cpu;
3277
3278 local_bh_disable();
3279 cpu = smp_processor_id();
3280 for (i = 0; i < dev->num_tx_queues; i++) {
3281 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3282
3283 __netif_tx_lock(txq, cpu);
3284 netif_tx_stop_queue(txq);
3285 __netif_tx_unlock(txq);
3286 }
3287 local_bh_enable();
3288 }
3289
3290 static inline void netif_addr_lock(struct net_device *dev)
3291 {
3292 spin_lock(&dev->addr_list_lock);
3293 }
3294
3295 static inline void netif_addr_lock_nested(struct net_device *dev)
3296 {
3297 int subclass = SINGLE_DEPTH_NESTING;
3298
3299 if (dev->netdev_ops->ndo_get_lock_subclass)
3300 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3301
3302 spin_lock_nested(&dev->addr_list_lock, subclass);
3303 }
3304
3305 static inline void netif_addr_lock_bh(struct net_device *dev)
3306 {
3307 spin_lock_bh(&dev->addr_list_lock);
3308 }
3309
3310 static inline void netif_addr_unlock(struct net_device *dev)
3311 {
3312 spin_unlock(&dev->addr_list_lock);
3313 }
3314
3315 static inline void netif_addr_unlock_bh(struct net_device *dev)
3316 {
3317 spin_unlock_bh(&dev->addr_list_lock);
3318 }
3319
3320 /*
3321 * dev_addrs walker. Should be used only for read access. Call with
3322 * rcu_read_lock held.
3323 */
3324 #define for_each_dev_addr(dev, ha) \
3325 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3326
3327 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3328
3329 void ether_setup(struct net_device *dev);
3330
3331 /* Support for loadable net-drivers */
3332 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3333 unsigned char name_assign_type,
3334 void (*setup)(struct net_device *),
3335 unsigned int txqs, unsigned int rxqs);
3336 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3337 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3338
3339 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3340 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3341 count)
3342
3343 int register_netdev(struct net_device *dev);
3344 void unregister_netdev(struct net_device *dev);
3345
3346 /* General hardware address lists handling functions */
3347 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3348 struct netdev_hw_addr_list *from_list, int addr_len);
3349 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3350 struct netdev_hw_addr_list *from_list, int addr_len);
3351 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3352 struct net_device *dev,
3353 int (*sync)(struct net_device *, const unsigned char *),
3354 int (*unsync)(struct net_device *,
3355 const unsigned char *));
3356 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3357 struct net_device *dev,
3358 int (*unsync)(struct net_device *,
3359 const unsigned char *));
3360 void __hw_addr_init(struct netdev_hw_addr_list *list);
3361
3362 /* Functions used for device addresses handling */
3363 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3364 unsigned char addr_type);
3365 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3366 unsigned char addr_type);
3367 void dev_addr_flush(struct net_device *dev);
3368 int dev_addr_init(struct net_device *dev);
3369
3370 /* Functions used for unicast addresses handling */
3371 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3372 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3373 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3374 int dev_uc_sync(struct net_device *to, struct net_device *from);
3375 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3376 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3377 void dev_uc_flush(struct net_device *dev);
3378 void dev_uc_init(struct net_device *dev);
3379
3380 /**
3381 * __dev_uc_sync - Synchonize device's unicast list
3382 * @dev: device to sync
3383 * @sync: function to call if address should be added
3384 * @unsync: function to call if address should be removed
3385 *
3386 * Add newly added addresses to the interface, and release
3387 * addresses that have been deleted.
3388 **/
3389 static inline int __dev_uc_sync(struct net_device *dev,
3390 int (*sync)(struct net_device *,
3391 const unsigned char *),
3392 int (*unsync)(struct net_device *,
3393 const unsigned char *))
3394 {
3395 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3396 }
3397
3398 /**
3399 * __dev_uc_unsync - Remove synchronized addresses from device
3400 * @dev: device to sync
3401 * @unsync: function to call if address should be removed
3402 *
3403 * Remove all addresses that were added to the device by dev_uc_sync().
3404 **/
3405 static inline void __dev_uc_unsync(struct net_device *dev,
3406 int (*unsync)(struct net_device *,
3407 const unsigned char *))
3408 {
3409 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3410 }
3411
3412 /* Functions used for multicast addresses handling */
3413 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3414 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3415 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3416 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3417 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3418 int dev_mc_sync(struct net_device *to, struct net_device *from);
3419 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3420 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3421 void dev_mc_flush(struct net_device *dev);
3422 void dev_mc_init(struct net_device *dev);
3423
3424 /**
3425 * __dev_mc_sync - Synchonize device's multicast list
3426 * @dev: device to sync
3427 * @sync: function to call if address should be added
3428 * @unsync: function to call if address should be removed
3429 *
3430 * Add newly added addresses to the interface, and release
3431 * addresses that have been deleted.
3432 **/
3433 static inline int __dev_mc_sync(struct net_device *dev,
3434 int (*sync)(struct net_device *,
3435 const unsigned char *),
3436 int (*unsync)(struct net_device *,
3437 const unsigned char *))
3438 {
3439 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3440 }
3441
3442 /**
3443 * __dev_mc_unsync - Remove synchronized addresses from device
3444 * @dev: device to sync
3445 * @unsync: function to call if address should be removed
3446 *
3447 * Remove all addresses that were added to the device by dev_mc_sync().
3448 **/
3449 static inline void __dev_mc_unsync(struct net_device *dev,
3450 int (*unsync)(struct net_device *,
3451 const unsigned char *))
3452 {
3453 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3454 }
3455
3456 /* Functions used for secondary unicast and multicast support */
3457 void dev_set_rx_mode(struct net_device *dev);
3458 void __dev_set_rx_mode(struct net_device *dev);
3459 int dev_set_promiscuity(struct net_device *dev, int inc);
3460 int dev_set_allmulti(struct net_device *dev, int inc);
3461 void netdev_state_change(struct net_device *dev);
3462 void netdev_notify_peers(struct net_device *dev);
3463 void netdev_features_change(struct net_device *dev);
3464 /* Load a device via the kmod */
3465 void dev_load(struct net *net, const char *name);
3466 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3467 struct rtnl_link_stats64 *storage);
3468 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3469 const struct net_device_stats *netdev_stats);
3470
3471 extern int netdev_max_backlog;
3472 extern int netdev_tstamp_prequeue;
3473 extern int weight_p;
3474 extern int bpf_jit_enable;
3475
3476 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3477 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3478 struct list_head **iter);
3479 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3480 struct list_head **iter);
3481
3482 /* iterate through upper list, must be called under RCU read lock */
3483 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3484 for (iter = &(dev)->adj_list.upper, \
3485 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3486 updev; \
3487 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3488
3489 /* iterate through upper list, must be called under RCU read lock */
3490 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3491 for (iter = &(dev)->all_adj_list.upper, \
3492 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3493 updev; \
3494 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3495
3496 void *netdev_lower_get_next_private(struct net_device *dev,
3497 struct list_head **iter);
3498 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3499 struct list_head **iter);
3500
3501 #define netdev_for_each_lower_private(dev, priv, iter) \
3502 for (iter = (dev)->adj_list.lower.next, \
3503 priv = netdev_lower_get_next_private(dev, &(iter)); \
3504 priv; \
3505 priv = netdev_lower_get_next_private(dev, &(iter)))
3506
3507 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3508 for (iter = &(dev)->adj_list.lower, \
3509 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3510 priv; \
3511 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3512
3513 void *netdev_lower_get_next(struct net_device *dev,
3514 struct list_head **iter);
3515 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3516 for (iter = &(dev)->adj_list.lower, \
3517 ldev = netdev_lower_get_next(dev, &(iter)); \
3518 ldev; \
3519 ldev = netdev_lower_get_next(dev, &(iter)))
3520
3521 void *netdev_adjacent_get_private(struct list_head *adj_list);
3522 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3523 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3524 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3525 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3526 int netdev_master_upper_dev_link(struct net_device *dev,
3527 struct net_device *upper_dev);
3528 int netdev_master_upper_dev_link_private(struct net_device *dev,
3529 struct net_device *upper_dev,
3530 void *private);
3531 void netdev_upper_dev_unlink(struct net_device *dev,
3532 struct net_device *upper_dev);
3533 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3534 void *netdev_lower_dev_get_private(struct net_device *dev,
3535 struct net_device *lower_dev);
3536
3537 /* RSS keys are 40 or 52 bytes long */
3538 #define NETDEV_RSS_KEY_LEN 52
3539 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3540 void netdev_rss_key_fill(void *buffer, size_t len);
3541
3542 int dev_get_nest_level(struct net_device *dev,
3543 bool (*type_check)(struct net_device *dev));
3544 int skb_checksum_help(struct sk_buff *skb);
3545 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3546 netdev_features_t features, bool tx_path);
3547 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3548 netdev_features_t features);
3549
3550 struct netdev_bonding_info {
3551 ifslave slave;
3552 ifbond master;
3553 };
3554
3555 struct netdev_notifier_bonding_info {
3556 struct netdev_notifier_info info; /* must be first */
3557 struct netdev_bonding_info bonding_info;
3558 };
3559
3560 void netdev_bonding_info_change(struct net_device *dev,
3561 struct netdev_bonding_info *bonding_info);
3562
3563 static inline
3564 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3565 {
3566 return __skb_gso_segment(skb, features, true);
3567 }
3568 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3569
3570 static inline bool can_checksum_protocol(netdev_features_t features,
3571 __be16 protocol)
3572 {
3573 return ((features & NETIF_F_GEN_CSUM) ||
3574 ((features & NETIF_F_V4_CSUM) &&
3575 protocol == htons(ETH_P_IP)) ||
3576 ((features & NETIF_F_V6_CSUM) &&
3577 protocol == htons(ETH_P_IPV6)) ||
3578 ((features & NETIF_F_FCOE_CRC) &&
3579 protocol == htons(ETH_P_FCOE)));
3580 }
3581
3582 #ifdef CONFIG_BUG
3583 void netdev_rx_csum_fault(struct net_device *dev);
3584 #else
3585 static inline void netdev_rx_csum_fault(struct net_device *dev)
3586 {
3587 }
3588 #endif
3589 /* rx skb timestamps */
3590 void net_enable_timestamp(void);
3591 void net_disable_timestamp(void);
3592
3593 #ifdef CONFIG_PROC_FS
3594 int __init dev_proc_init(void);
3595 #else
3596 #define dev_proc_init() 0
3597 #endif
3598
3599 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3600 struct sk_buff *skb, struct net_device *dev,
3601 bool more)
3602 {
3603 skb->xmit_more = more ? 1 : 0;
3604 return ops->ndo_start_xmit(skb, dev);
3605 }
3606
3607 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3608 struct netdev_queue *txq, bool more)
3609 {
3610 const struct net_device_ops *ops = dev->netdev_ops;
3611 int rc;
3612
3613 rc = __netdev_start_xmit(ops, skb, dev, more);
3614 if (rc == NETDEV_TX_OK)
3615 txq_trans_update(txq);
3616
3617 return rc;
3618 }
3619
3620 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3621 const void *ns);
3622 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3623 const void *ns);
3624
3625 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3626 {
3627 return netdev_class_create_file_ns(class_attr, NULL);
3628 }
3629
3630 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3631 {
3632 netdev_class_remove_file_ns(class_attr, NULL);
3633 }
3634
3635 extern struct kobj_ns_type_operations net_ns_type_operations;
3636
3637 const char *netdev_drivername(const struct net_device *dev);
3638
3639 void linkwatch_run_queue(void);
3640
3641 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3642 netdev_features_t f2)
3643 {
3644 if (f1 & NETIF_F_GEN_CSUM)
3645 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3646 if (f2 & NETIF_F_GEN_CSUM)
3647 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3648 f1 &= f2;
3649 if (f1 & NETIF_F_GEN_CSUM)
3650 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3651
3652 return f1;
3653 }
3654
3655 static inline netdev_features_t netdev_get_wanted_features(
3656 struct net_device *dev)
3657 {
3658 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3659 }
3660 netdev_features_t netdev_increment_features(netdev_features_t all,
3661 netdev_features_t one, netdev_features_t mask);
3662
3663 /* Allow TSO being used on stacked device :
3664 * Performing the GSO segmentation before last device
3665 * is a performance improvement.
3666 */
3667 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3668 netdev_features_t mask)
3669 {
3670 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3671 }
3672
3673 int __netdev_update_features(struct net_device *dev);
3674 void netdev_update_features(struct net_device *dev);
3675 void netdev_change_features(struct net_device *dev);
3676
3677 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3678 struct net_device *dev);
3679
3680 netdev_features_t passthru_features_check(struct sk_buff *skb,
3681 struct net_device *dev,
3682 netdev_features_t features);
3683 netdev_features_t netif_skb_features(struct sk_buff *skb);
3684
3685 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3686 {
3687 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3688
3689 /* check flags correspondence */
3690 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3691 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3692 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3693 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3694 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3695 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3696 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3697 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3698 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3699 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3700 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3701 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3702 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3703
3704 return (features & feature) == feature;
3705 }
3706
3707 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3708 {
3709 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3710 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3711 }
3712
3713 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3714 netdev_features_t features)
3715 {
3716 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3717 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3718 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3719 }
3720
3721 static inline void netif_set_gso_max_size(struct net_device *dev,
3722 unsigned int size)
3723 {
3724 dev->gso_max_size = size;
3725 }
3726
3727 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3728 int pulled_hlen, u16 mac_offset,
3729 int mac_len)
3730 {
3731 skb->protocol = protocol;
3732 skb->encapsulation = 1;
3733 skb_push(skb, pulled_hlen);
3734 skb_reset_transport_header(skb);
3735 skb->mac_header = mac_offset;
3736 skb->network_header = skb->mac_header + mac_len;
3737 skb->mac_len = mac_len;
3738 }
3739
3740 static inline bool netif_is_macvlan(struct net_device *dev)
3741 {
3742 return dev->priv_flags & IFF_MACVLAN;
3743 }
3744
3745 static inline bool netif_is_macvlan_port(struct net_device *dev)
3746 {
3747 return dev->priv_flags & IFF_MACVLAN_PORT;
3748 }
3749
3750 static inline bool netif_is_ipvlan(struct net_device *dev)
3751 {
3752 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3753 }
3754
3755 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3756 {
3757 return dev->priv_flags & IFF_IPVLAN_MASTER;
3758 }
3759
3760 static inline bool netif_is_bond_master(struct net_device *dev)
3761 {
3762 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3763 }
3764
3765 static inline bool netif_is_bond_slave(struct net_device *dev)
3766 {
3767 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3768 }
3769
3770 static inline bool netif_supports_nofcs(struct net_device *dev)
3771 {
3772 return dev->priv_flags & IFF_SUPP_NOFCS;
3773 }
3774
3775 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3776 static inline void netif_keep_dst(struct net_device *dev)
3777 {
3778 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3779 }
3780
3781 extern struct pernet_operations __net_initdata loopback_net_ops;
3782
3783 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3784
3785 /* netdev_printk helpers, similar to dev_printk */
3786
3787 static inline const char *netdev_name(const struct net_device *dev)
3788 {
3789 if (!dev->name[0] || strchr(dev->name, '%'))
3790 return "(unnamed net_device)";
3791 return dev->name;
3792 }
3793
3794 static inline const char *netdev_reg_state(const struct net_device *dev)
3795 {
3796 switch (dev->reg_state) {
3797 case NETREG_UNINITIALIZED: return " (uninitialized)";
3798 case NETREG_REGISTERED: return "";
3799 case NETREG_UNREGISTERING: return " (unregistering)";
3800 case NETREG_UNREGISTERED: return " (unregistered)";
3801 case NETREG_RELEASED: return " (released)";
3802 case NETREG_DUMMY: return " (dummy)";
3803 }
3804
3805 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3806 return " (unknown)";
3807 }
3808
3809 __printf(3, 4)
3810 void netdev_printk(const char *level, const struct net_device *dev,
3811 const char *format, ...);
3812 __printf(2, 3)
3813 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3814 __printf(2, 3)
3815 void netdev_alert(const struct net_device *dev, const char *format, ...);
3816 __printf(2, 3)
3817 void netdev_crit(const struct net_device *dev, const char *format, ...);
3818 __printf(2, 3)
3819 void netdev_err(const struct net_device *dev, const char *format, ...);
3820 __printf(2, 3)
3821 void netdev_warn(const struct net_device *dev, const char *format, ...);
3822 __printf(2, 3)
3823 void netdev_notice(const struct net_device *dev, const char *format, ...);
3824 __printf(2, 3)
3825 void netdev_info(const struct net_device *dev, const char *format, ...);
3826
3827 #define MODULE_ALIAS_NETDEV(device) \
3828 MODULE_ALIAS("netdev-" device)
3829
3830 #if defined(CONFIG_DYNAMIC_DEBUG)
3831 #define netdev_dbg(__dev, format, args...) \
3832 do { \
3833 dynamic_netdev_dbg(__dev, format, ##args); \
3834 } while (0)
3835 #elif defined(DEBUG)
3836 #define netdev_dbg(__dev, format, args...) \
3837 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3838 #else
3839 #define netdev_dbg(__dev, format, args...) \
3840 ({ \
3841 if (0) \
3842 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3843 })
3844 #endif
3845
3846 #if defined(VERBOSE_DEBUG)
3847 #define netdev_vdbg netdev_dbg
3848 #else
3849
3850 #define netdev_vdbg(dev, format, args...) \
3851 ({ \
3852 if (0) \
3853 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3854 0; \
3855 })
3856 #endif
3857
3858 /*
3859 * netdev_WARN() acts like dev_printk(), but with the key difference
3860 * of using a WARN/WARN_ON to get the message out, including the
3861 * file/line information and a backtrace.
3862 */
3863 #define netdev_WARN(dev, format, args...) \
3864 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3865 netdev_reg_state(dev), ##args)
3866
3867 /* netif printk helpers, similar to netdev_printk */
3868
3869 #define netif_printk(priv, type, level, dev, fmt, args...) \
3870 do { \
3871 if (netif_msg_##type(priv)) \
3872 netdev_printk(level, (dev), fmt, ##args); \
3873 } while (0)
3874
3875 #define netif_level(level, priv, type, dev, fmt, args...) \
3876 do { \
3877 if (netif_msg_##type(priv)) \
3878 netdev_##level(dev, fmt, ##args); \
3879 } while (0)
3880
3881 #define netif_emerg(priv, type, dev, fmt, args...) \
3882 netif_level(emerg, priv, type, dev, fmt, ##args)
3883 #define netif_alert(priv, type, dev, fmt, args...) \
3884 netif_level(alert, priv, type, dev, fmt, ##args)
3885 #define netif_crit(priv, type, dev, fmt, args...) \
3886 netif_level(crit, priv, type, dev, fmt, ##args)
3887 #define netif_err(priv, type, dev, fmt, args...) \
3888 netif_level(err, priv, type, dev, fmt, ##args)
3889 #define netif_warn(priv, type, dev, fmt, args...) \
3890 netif_level(warn, priv, type, dev, fmt, ##args)
3891 #define netif_notice(priv, type, dev, fmt, args...) \
3892 netif_level(notice, priv, type, dev, fmt, ##args)
3893 #define netif_info(priv, type, dev, fmt, args...) \
3894 netif_level(info, priv, type, dev, fmt, ##args)
3895
3896 #if defined(CONFIG_DYNAMIC_DEBUG)
3897 #define netif_dbg(priv, type, netdev, format, args...) \
3898 do { \
3899 if (netif_msg_##type(priv)) \
3900 dynamic_netdev_dbg(netdev, format, ##args); \
3901 } while (0)
3902 #elif defined(DEBUG)
3903 #define netif_dbg(priv, type, dev, format, args...) \
3904 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3905 #else
3906 #define netif_dbg(priv, type, dev, format, args...) \
3907 ({ \
3908 if (0) \
3909 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3910 0; \
3911 })
3912 #endif
3913
3914 #if defined(VERBOSE_DEBUG)
3915 #define netif_vdbg netif_dbg
3916 #else
3917 #define netif_vdbg(priv, type, dev, format, args...) \
3918 ({ \
3919 if (0) \
3920 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3921 0; \
3922 })
3923 #endif
3924
3925 /*
3926 * The list of packet types we will receive (as opposed to discard)
3927 * and the routines to invoke.
3928 *
3929 * Why 16. Because with 16 the only overlap we get on a hash of the
3930 * low nibble of the protocol value is RARP/SNAP/X.25.
3931 *
3932 * NOTE: That is no longer true with the addition of VLAN tags. Not
3933 * sure which should go first, but I bet it won't make much
3934 * difference if we are running VLANs. The good news is that
3935 * this protocol won't be in the list unless compiled in, so
3936 * the average user (w/out VLANs) will not be adversely affected.
3937 * --BLG
3938 *
3939 * 0800 IP
3940 * 8100 802.1Q VLAN
3941 * 0001 802.3
3942 * 0002 AX.25
3943 * 0004 802.2
3944 * 8035 RARP
3945 * 0005 SNAP
3946 * 0805 X.25
3947 * 0806 ARP
3948 * 8137 IPX
3949 * 0009 Localtalk
3950 * 86DD IPv6
3951 */
3952 #define PTYPE_HASH_SIZE (16)
3953 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3954
3955 #endif /* _LINUX_NETDEVICE_H */
This page took 0.109263 seconds and 6 git commands to generate.