dev: introduce dev_get_iflink()
[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 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
799 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
800 * Required can not be NULL.
801 *
802 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
803 * void *accel_priv, select_queue_fallback_t fallback);
804 * Called to decide which queue to when device supports multiple
805 * transmit queues.
806 *
807 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
808 * This function is called to allow device receiver to make
809 * changes to configuration when multicast or promiscious is enabled.
810 *
811 * void (*ndo_set_rx_mode)(struct net_device *dev);
812 * This function is called device changes address list filtering.
813 * If driver handles unicast address filtering, it should set
814 * IFF_UNICAST_FLT to its priv_flags.
815 *
816 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
817 * This function is called when the Media Access Control address
818 * needs to be changed. If this interface is not defined, the
819 * mac address can not be changed.
820 *
821 * int (*ndo_validate_addr)(struct net_device *dev);
822 * Test if Media Access Control address is valid for the device.
823 *
824 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
825 * Called when a user request an ioctl which can't be handled by
826 * the generic interface code. If not defined ioctl's return
827 * not supported error code.
828 *
829 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
830 * Used to set network devices bus interface parameters. This interface
831 * is retained for legacy reason, new devices should use the bus
832 * interface (PCI) for low level management.
833 *
834 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
835 * Called when a user wants to change the Maximum Transfer Unit
836 * of a device. If not defined, any request to change MTU will
837 * will return an error.
838 *
839 * void (*ndo_tx_timeout)(struct net_device *dev);
840 * Callback uses when the transmitter has not made any progress
841 * for dev->watchdog ticks.
842 *
843 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
844 * struct rtnl_link_stats64 *storage);
845 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
846 * Called when a user wants to get the network device usage
847 * statistics. Drivers must do one of the following:
848 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
849 * rtnl_link_stats64 structure passed by the caller.
850 * 2. Define @ndo_get_stats to update a net_device_stats structure
851 * (which should normally be dev->stats) and return a pointer to
852 * it. The structure may be changed asynchronously only if each
853 * field is written atomically.
854 * 3. Update dev->stats asynchronously and atomically, and define
855 * neither operation.
856 *
857 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
858 * If device support VLAN filtering this function is called when a
859 * VLAN id is registered.
860 *
861 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
862 * If device support VLAN filtering this function is called when a
863 * VLAN id is unregistered.
864 *
865 * void (*ndo_poll_controller)(struct net_device *dev);
866 *
867 * SR-IOV management functions.
868 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
869 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
870 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
871 * int max_tx_rate);
872 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
873 * int (*ndo_get_vf_config)(struct net_device *dev,
874 * int vf, struct ifla_vf_info *ivf);
875 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
876 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
877 * struct nlattr *port[]);
878 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
879 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
880 * Called to setup 'tc' number of traffic classes in the net device. This
881 * is always called from the stack with the rtnl lock held and netif tx
882 * queues stopped. This allows the netdevice to perform queue management
883 * safely.
884 *
885 * Fiber Channel over Ethernet (FCoE) offload functions.
886 * int (*ndo_fcoe_enable)(struct net_device *dev);
887 * Called when the FCoE protocol stack wants to start using LLD for FCoE
888 * so the underlying device can perform whatever needed configuration or
889 * initialization to support acceleration of FCoE traffic.
890 *
891 * int (*ndo_fcoe_disable)(struct net_device *dev);
892 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
893 * so the underlying device can perform whatever needed clean-ups to
894 * stop supporting acceleration of FCoE traffic.
895 *
896 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
897 * struct scatterlist *sgl, unsigned int sgc);
898 * Called when the FCoE Initiator wants to initialize an I/O that
899 * is a possible candidate for Direct Data Placement (DDP). The LLD can
900 * perform necessary setup and returns 1 to indicate the device is set up
901 * successfully to perform DDP on this I/O, otherwise this returns 0.
902 *
903 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
904 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
905 * indicated by the FC exchange id 'xid', so the underlying device can
906 * clean up and reuse resources for later DDP requests.
907 *
908 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
909 * struct scatterlist *sgl, unsigned int sgc);
910 * Called when the FCoE Target wants to initialize an I/O that
911 * is a possible candidate for Direct Data Placement (DDP). The LLD can
912 * perform necessary setup and returns 1 to indicate the device is set up
913 * successfully to perform DDP on this I/O, otherwise this returns 0.
914 *
915 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
916 * struct netdev_fcoe_hbainfo *hbainfo);
917 * Called when the FCoE Protocol stack wants information on the underlying
918 * device. This information is utilized by the FCoE protocol stack to
919 * register attributes with Fiber Channel management service as per the
920 * FC-GS Fabric Device Management Information(FDMI) specification.
921 *
922 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
923 * Called when the underlying device wants to override default World Wide
924 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
925 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
926 * protocol stack to use.
927 *
928 * RFS acceleration.
929 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
930 * u16 rxq_index, u32 flow_id);
931 * Set hardware filter for RFS. rxq_index is the target queue index;
932 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
933 * Return the filter ID on success, or a negative error code.
934 *
935 * Slave management functions (for bridge, bonding, etc).
936 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
937 * Called to make another netdev an underling.
938 *
939 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
940 * Called to release previously enslaved netdev.
941 *
942 * Feature/offload setting functions.
943 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
944 * netdev_features_t features);
945 * Adjusts the requested feature flags according to device-specific
946 * constraints, and returns the resulting flags. Must not modify
947 * the device state.
948 *
949 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
950 * Called to update device configuration to new features. Passed
951 * feature set might be less than what was returned by ndo_fix_features()).
952 * Must return >0 or -errno if it changed dev->features itself.
953 *
954 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
955 * struct net_device *dev,
956 * const unsigned char *addr, u16 vid, u16 flags)
957 * Adds an FDB entry to dev for addr.
958 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
959 * struct net_device *dev,
960 * const unsigned char *addr, u16 vid)
961 * Deletes the FDB entry from dev coresponding to addr.
962 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
963 * struct net_device *dev, struct net_device *filter_dev,
964 * int idx)
965 * Used to add FDB entries to dump requests. Implementers should add
966 * entries to skb and update idx with the number of entries.
967 *
968 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
969 * u16 flags)
970 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
971 * struct net_device *dev, u32 filter_mask)
972 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
973 * u16 flags);
974 *
975 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
976 * Called to change device carrier. Soft-devices (like dummy, team, etc)
977 * which do not represent real hardware may define this to allow their
978 * userspace components to manage their virtual carrier state. Devices
979 * that determine carrier state from physical hardware properties (eg
980 * network cables) or protocol-dependent mechanisms (eg
981 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
982 *
983 * int (*ndo_get_phys_port_id)(struct net_device *dev,
984 * struct netdev_phys_item_id *ppid);
985 * Called to get ID of physical port of this device. If driver does
986 * not implement this, it is assumed that the hw is not able to have
987 * multiple net devices on single physical port.
988 *
989 * void (*ndo_add_vxlan_port)(struct net_device *dev,
990 * sa_family_t sa_family, __be16 port);
991 * Called by vxlan to notiy a driver about the UDP port and socket
992 * address family that vxlan is listnening to. It is called only when
993 * a new port starts listening. The operation is protected by the
994 * vxlan_net->sock_lock.
995 *
996 * void (*ndo_del_vxlan_port)(struct net_device *dev,
997 * sa_family_t sa_family, __be16 port);
998 * Called by vxlan to notify the driver about a UDP port and socket
999 * address family that vxlan is not listening to anymore. The operation
1000 * is protected by the vxlan_net->sock_lock.
1001 *
1002 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1003 * struct net_device *dev)
1004 * Called by upper layer devices to accelerate switching or other
1005 * station functionality into hardware. 'pdev is the lowerdev
1006 * to use for the offload and 'dev' is the net device that will
1007 * back the offload. Returns a pointer to the private structure
1008 * the upper layer will maintain.
1009 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1010 * Called by upper layer device to delete the station created
1011 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1012 * the station and priv is the structure returned by the add
1013 * operation.
1014 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1015 * struct net_device *dev,
1016 * void *priv);
1017 * Callback to use for xmit over the accelerated station. This
1018 * is used in place of ndo_start_xmit on accelerated net
1019 * devices.
1020 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1021 * struct net_device *dev
1022 * netdev_features_t features);
1023 * Called by core transmit path to determine if device is capable of
1024 * performing offload operations on a given packet. This is to give
1025 * the device an opportunity to implement any restrictions that cannot
1026 * be otherwise expressed by feature flags. The check is called with
1027 * the set of features that the stack has calculated and it returns
1028 * those the driver believes to be appropriate.
1029 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1030 * int queue_index, u32 maxrate);
1031 * Called when a user wants to set a max-rate limitation of specific
1032 * TX queue.
1033 * int (*ndo_get_iflink)(const struct net_device *dev);
1034 * Called to get the iflink value of this device.
1035 */
1036 struct net_device_ops {
1037 int (*ndo_init)(struct net_device *dev);
1038 void (*ndo_uninit)(struct net_device *dev);
1039 int (*ndo_open)(struct net_device *dev);
1040 int (*ndo_stop)(struct net_device *dev);
1041 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1042 struct net_device *dev);
1043 u16 (*ndo_select_queue)(struct net_device *dev,
1044 struct sk_buff *skb,
1045 void *accel_priv,
1046 select_queue_fallback_t fallback);
1047 void (*ndo_change_rx_flags)(struct net_device *dev,
1048 int flags);
1049 void (*ndo_set_rx_mode)(struct net_device *dev);
1050 int (*ndo_set_mac_address)(struct net_device *dev,
1051 void *addr);
1052 int (*ndo_validate_addr)(struct net_device *dev);
1053 int (*ndo_do_ioctl)(struct net_device *dev,
1054 struct ifreq *ifr, int cmd);
1055 int (*ndo_set_config)(struct net_device *dev,
1056 struct ifmap *map);
1057 int (*ndo_change_mtu)(struct net_device *dev,
1058 int new_mtu);
1059 int (*ndo_neigh_setup)(struct net_device *dev,
1060 struct neigh_parms *);
1061 void (*ndo_tx_timeout) (struct net_device *dev);
1062
1063 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1064 struct rtnl_link_stats64 *storage);
1065 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1066
1067 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1068 __be16 proto, u16 vid);
1069 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1070 __be16 proto, u16 vid);
1071 #ifdef CONFIG_NET_POLL_CONTROLLER
1072 void (*ndo_poll_controller)(struct net_device *dev);
1073 int (*ndo_netpoll_setup)(struct net_device *dev,
1074 struct netpoll_info *info);
1075 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1076 #endif
1077 #ifdef CONFIG_NET_RX_BUSY_POLL
1078 int (*ndo_busy_poll)(struct napi_struct *dev);
1079 #endif
1080 int (*ndo_set_vf_mac)(struct net_device *dev,
1081 int queue, u8 *mac);
1082 int (*ndo_set_vf_vlan)(struct net_device *dev,
1083 int queue, u16 vlan, u8 qos);
1084 int (*ndo_set_vf_rate)(struct net_device *dev,
1085 int vf, int min_tx_rate,
1086 int max_tx_rate);
1087 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1088 int vf, bool setting);
1089 int (*ndo_get_vf_config)(struct net_device *dev,
1090 int vf,
1091 struct ifla_vf_info *ivf);
1092 int (*ndo_set_vf_link_state)(struct net_device *dev,
1093 int vf, int link_state);
1094 int (*ndo_set_vf_port)(struct net_device *dev,
1095 int vf,
1096 struct nlattr *port[]);
1097 int (*ndo_get_vf_port)(struct net_device *dev,
1098 int vf, struct sk_buff *skb);
1099 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1100 #if IS_ENABLED(CONFIG_FCOE)
1101 int (*ndo_fcoe_enable)(struct net_device *dev);
1102 int (*ndo_fcoe_disable)(struct net_device *dev);
1103 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1104 u16 xid,
1105 struct scatterlist *sgl,
1106 unsigned int sgc);
1107 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1108 u16 xid);
1109 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1110 u16 xid,
1111 struct scatterlist *sgl,
1112 unsigned int sgc);
1113 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1114 struct netdev_fcoe_hbainfo *hbainfo);
1115 #endif
1116
1117 #if IS_ENABLED(CONFIG_LIBFCOE)
1118 #define NETDEV_FCOE_WWNN 0
1119 #define NETDEV_FCOE_WWPN 1
1120 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1121 u64 *wwn, int type);
1122 #endif
1123
1124 #ifdef CONFIG_RFS_ACCEL
1125 int (*ndo_rx_flow_steer)(struct net_device *dev,
1126 const struct sk_buff *skb,
1127 u16 rxq_index,
1128 u32 flow_id);
1129 #endif
1130 int (*ndo_add_slave)(struct net_device *dev,
1131 struct net_device *slave_dev);
1132 int (*ndo_del_slave)(struct net_device *dev,
1133 struct net_device *slave_dev);
1134 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1135 netdev_features_t features);
1136 int (*ndo_set_features)(struct net_device *dev,
1137 netdev_features_t features);
1138 int (*ndo_neigh_construct)(struct neighbour *n);
1139 void (*ndo_neigh_destroy)(struct neighbour *n);
1140
1141 int (*ndo_fdb_add)(struct ndmsg *ndm,
1142 struct nlattr *tb[],
1143 struct net_device *dev,
1144 const unsigned char *addr,
1145 u16 vid,
1146 u16 flags);
1147 int (*ndo_fdb_del)(struct ndmsg *ndm,
1148 struct nlattr *tb[],
1149 struct net_device *dev,
1150 const unsigned char *addr,
1151 u16 vid);
1152 int (*ndo_fdb_dump)(struct sk_buff *skb,
1153 struct netlink_callback *cb,
1154 struct net_device *dev,
1155 struct net_device *filter_dev,
1156 int idx);
1157
1158 int (*ndo_bridge_setlink)(struct net_device *dev,
1159 struct nlmsghdr *nlh,
1160 u16 flags);
1161 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1162 u32 pid, u32 seq,
1163 struct net_device *dev,
1164 u32 filter_mask);
1165 int (*ndo_bridge_dellink)(struct net_device *dev,
1166 struct nlmsghdr *nlh,
1167 u16 flags);
1168 int (*ndo_change_carrier)(struct net_device *dev,
1169 bool new_carrier);
1170 int (*ndo_get_phys_port_id)(struct net_device *dev,
1171 struct netdev_phys_item_id *ppid);
1172 int (*ndo_get_phys_port_name)(struct net_device *dev,
1173 char *name, size_t len);
1174 void (*ndo_add_vxlan_port)(struct net_device *dev,
1175 sa_family_t sa_family,
1176 __be16 port);
1177 void (*ndo_del_vxlan_port)(struct net_device *dev,
1178 sa_family_t sa_family,
1179 __be16 port);
1180
1181 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1182 struct net_device *dev);
1183 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1184 void *priv);
1185
1186 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1187 struct net_device *dev,
1188 void *priv);
1189 int (*ndo_get_lock_subclass)(struct net_device *dev);
1190 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1191 struct net_device *dev,
1192 netdev_features_t features);
1193 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1194 int queue_index,
1195 u32 maxrate);
1196 int (*ndo_get_iflink)(const struct net_device *dev);
1197 };
1198
1199 /**
1200 * enum net_device_priv_flags - &struct net_device priv_flags
1201 *
1202 * These are the &struct net_device, they are only set internally
1203 * by drivers and used in the kernel. These flags are invisible to
1204 * userspace, this means that the order of these flags can change
1205 * during any kernel release.
1206 *
1207 * You should have a pretty good reason to be extending these flags.
1208 *
1209 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1210 * @IFF_EBRIDGE: Ethernet bridging device
1211 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1212 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1213 * @IFF_MASTER_ALB: bonding master, balance-alb
1214 * @IFF_BONDING: bonding master or slave
1215 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1216 * @IFF_ISATAP: ISATAP interface (RFC4214)
1217 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1218 * @IFF_WAN_HDLC: WAN HDLC device
1219 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1220 * release skb->dst
1221 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1222 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1223 * @IFF_MACVLAN_PORT: device used as macvlan port
1224 * @IFF_BRIDGE_PORT: device used as bridge port
1225 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1226 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1227 * @IFF_UNICAST_FLT: Supports unicast filtering
1228 * @IFF_TEAM_PORT: device used as team port
1229 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1230 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1231 * change when it's running
1232 * @IFF_MACVLAN: Macvlan device
1233 */
1234 enum netdev_priv_flags {
1235 IFF_802_1Q_VLAN = 1<<0,
1236 IFF_EBRIDGE = 1<<1,
1237 IFF_SLAVE_INACTIVE = 1<<2,
1238 IFF_MASTER_8023AD = 1<<3,
1239 IFF_MASTER_ALB = 1<<4,
1240 IFF_BONDING = 1<<5,
1241 IFF_SLAVE_NEEDARP = 1<<6,
1242 IFF_ISATAP = 1<<7,
1243 IFF_MASTER_ARPMON = 1<<8,
1244 IFF_WAN_HDLC = 1<<9,
1245 IFF_XMIT_DST_RELEASE = 1<<10,
1246 IFF_DONT_BRIDGE = 1<<11,
1247 IFF_DISABLE_NETPOLL = 1<<12,
1248 IFF_MACVLAN_PORT = 1<<13,
1249 IFF_BRIDGE_PORT = 1<<14,
1250 IFF_OVS_DATAPATH = 1<<15,
1251 IFF_TX_SKB_SHARING = 1<<16,
1252 IFF_UNICAST_FLT = 1<<17,
1253 IFF_TEAM_PORT = 1<<18,
1254 IFF_SUPP_NOFCS = 1<<19,
1255 IFF_LIVE_ADDR_CHANGE = 1<<20,
1256 IFF_MACVLAN = 1<<21,
1257 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1258 IFF_IPVLAN_MASTER = 1<<23,
1259 IFF_IPVLAN_SLAVE = 1<<24,
1260 };
1261
1262 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1263 #define IFF_EBRIDGE IFF_EBRIDGE
1264 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1265 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1266 #define IFF_MASTER_ALB IFF_MASTER_ALB
1267 #define IFF_BONDING IFF_BONDING
1268 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1269 #define IFF_ISATAP IFF_ISATAP
1270 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1271 #define IFF_WAN_HDLC IFF_WAN_HDLC
1272 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1273 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1274 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1275 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1276 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1277 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1278 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1279 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1280 #define IFF_TEAM_PORT IFF_TEAM_PORT
1281 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1282 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1283 #define IFF_MACVLAN IFF_MACVLAN
1284 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1285 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1286 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1287
1288 /**
1289 * struct net_device - The DEVICE structure.
1290 * Actually, this whole structure is a big mistake. It mixes I/O
1291 * data with strictly "high-level" data, and it has to know about
1292 * almost every data structure used in the INET module.
1293 *
1294 * @name: This is the first field of the "visible" part of this structure
1295 * (i.e. as seen by users in the "Space.c" file). It is the name
1296 * of the interface.
1297 *
1298 * @name_hlist: Device name hash chain, please keep it close to name[]
1299 * @ifalias: SNMP alias
1300 * @mem_end: Shared memory end
1301 * @mem_start: Shared memory start
1302 * @base_addr: Device I/O address
1303 * @irq: Device IRQ number
1304 *
1305 * @state: Generic network queuing layer state, see netdev_state_t
1306 * @dev_list: The global list of network devices
1307 * @napi_list: List entry, that is used for polling napi devices
1308 * @unreg_list: List entry, that is used, when we are unregistering the
1309 * device, see the function unregister_netdev
1310 * @close_list: List entry, that is used, when we are closing the device
1311 *
1312 * @adj_list: Directly linked devices, like slaves for bonding
1313 * @all_adj_list: All linked devices, *including* neighbours
1314 * @features: Currently active device features
1315 * @hw_features: User-changeable features
1316 *
1317 * @wanted_features: User-requested features
1318 * @vlan_features: Mask of features inheritable by VLAN devices
1319 *
1320 * @hw_enc_features: Mask of features inherited by encapsulating devices
1321 * This field indicates what encapsulation
1322 * offloads the hardware is capable of doing,
1323 * and drivers will need to set them appropriately.
1324 *
1325 * @mpls_features: Mask of features inheritable by MPLS
1326 *
1327 * @ifindex: interface index
1328 * @iflink: unique device identifier
1329 *
1330 * @stats: Statistics struct, which was left as a legacy, use
1331 * rtnl_link_stats64 instead
1332 *
1333 * @rx_dropped: Dropped packets by core network,
1334 * do not use this in drivers
1335 * @tx_dropped: Dropped packets by core network,
1336 * do not use this in drivers
1337 *
1338 * @carrier_changes: Stats to monitor carrier on<->off transitions
1339 *
1340 * @wireless_handlers: List of functions to handle Wireless Extensions,
1341 * instead of ioctl,
1342 * see <net/iw_handler.h> for details.
1343 * @wireless_data: Instance data managed by the core of wireless extensions
1344 *
1345 * @netdev_ops: Includes several pointers to callbacks,
1346 * if one wants to override the ndo_*() functions
1347 * @ethtool_ops: Management operations
1348 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1349 * of Layer 2 headers.
1350 *
1351 * @flags: Interface flags (a la BSD)
1352 * @priv_flags: Like 'flags' but invisible to userspace,
1353 * see if.h for the definitions
1354 * @gflags: Global flags ( kept as legacy )
1355 * @padded: How much padding added by alloc_netdev()
1356 * @operstate: RFC2863 operstate
1357 * @link_mode: Mapping policy to operstate
1358 * @if_port: Selectable AUI, TP, ...
1359 * @dma: DMA channel
1360 * @mtu: Interface MTU value
1361 * @type: Interface hardware type
1362 * @hard_header_len: Hardware header length
1363 *
1364 * @needed_headroom: Extra headroom the hardware may need, but not in all
1365 * cases can this be guaranteed
1366 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1367 * cases can this be guaranteed. Some cases also use
1368 * LL_MAX_HEADER instead to allocate the skb
1369 *
1370 * interface address info:
1371 *
1372 * @perm_addr: Permanent hw address
1373 * @addr_assign_type: Hw address assignment type
1374 * @addr_len: Hardware address length
1375 * @neigh_priv_len; Used in neigh_alloc(),
1376 * initialized only in atm/clip.c
1377 * @dev_id: Used to differentiate devices that share
1378 * the same link layer address
1379 * @dev_port: Used to differentiate devices that share
1380 * the same function
1381 * @addr_list_lock: XXX: need comments on this one
1382 * @uc: unicast mac addresses
1383 * @mc: multicast mac addresses
1384 * @dev_addrs: list of device hw addresses
1385 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1386 * @uc_promisc: Counter, that indicates, that promiscuous mode
1387 * has been enabled due to the need to listen to
1388 * additional unicast addresses in a device that
1389 * does not implement ndo_set_rx_mode()
1390 * @promiscuity: Number of times, the NIC is told to work in
1391 * Promiscuous mode, if it becomes 0 the NIC will
1392 * exit from working in Promiscuous mode
1393 * @allmulti: Counter, enables or disables allmulticast mode
1394 *
1395 * @vlan_info: VLAN info
1396 * @dsa_ptr: dsa specific data
1397 * @tipc_ptr: TIPC specific data
1398 * @atalk_ptr: AppleTalk link
1399 * @ip_ptr: IPv4 specific data
1400 * @dn_ptr: DECnet specific data
1401 * @ip6_ptr: IPv6 specific data
1402 * @ax25_ptr: AX.25 specific data
1403 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1404 *
1405 * @last_rx: Time of last Rx
1406 * @dev_addr: Hw address (before bcast,
1407 * because most packets are unicast)
1408 *
1409 * @_rx: Array of RX queues
1410 * @num_rx_queues: Number of RX queues
1411 * allocated at register_netdev() time
1412 * @real_num_rx_queues: Number of RX queues currently active in device
1413 *
1414 * @rx_handler: handler for received packets
1415 * @rx_handler_data: XXX: need comments on this one
1416 * @ingress_queue: XXX: need comments on this one
1417 * @broadcast: hw bcast address
1418 *
1419 * @_tx: Array of TX queues
1420 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1421 * @real_num_tx_queues: Number of TX queues currently active in device
1422 * @qdisc: Root qdisc from userspace point of view
1423 * @tx_queue_len: Max frames per queue allowed
1424 * @tx_global_lock: XXX: need comments on this one
1425 *
1426 * @xps_maps: XXX: need comments on this one
1427 *
1428 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1429 * indexed by RX queue number. Assigned by driver.
1430 * This must only be set if the ndo_rx_flow_steer
1431 * operation is defined
1432 *
1433 * @trans_start: Time (in jiffies) of last Tx
1434 * @watchdog_timeo: Represents the timeout that is used by
1435 * the watchdog ( see dev_watchdog() )
1436 * @watchdog_timer: List of timers
1437 *
1438 * @pcpu_refcnt: Number of references to this device
1439 * @todo_list: Delayed register/unregister
1440 * @index_hlist: Device index hash chain
1441 * @link_watch_list: XXX: need comments on this one
1442 *
1443 * @reg_state: Register/unregister state machine
1444 * @dismantle: Device is going to be freed
1445 * @rtnl_link_state: This enum represents the phases of creating
1446 * a new link
1447 *
1448 * @destructor: Called from unregister,
1449 * can be used to call free_netdev
1450 * @npinfo: XXX: need comments on this one
1451 * @nd_net: Network namespace this network device is inside
1452 *
1453 * @ml_priv: Mid-layer private
1454 * @lstats: Loopback statistics
1455 * @tstats: Tunnel statistics
1456 * @dstats: Dummy statistics
1457 * @vstats: Virtual ethernet statistics
1458 *
1459 * @garp_port: GARP
1460 * @mrp_port: MRP
1461 *
1462 * @dev: Class/net/name entry
1463 * @sysfs_groups: Space for optional device, statistics and wireless
1464 * sysfs groups
1465 *
1466 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1467 * @rtnl_link_ops: Rtnl_link_ops
1468 *
1469 * @gso_max_size: Maximum size of generic segmentation offload
1470 * @gso_max_segs: Maximum number of segments that can be passed to the
1471 * NIC for GSO
1472 * @gso_min_segs: Minimum number of segments that can be passed to the
1473 * NIC for GSO
1474 *
1475 * @dcbnl_ops: Data Center Bridging netlink ops
1476 * @num_tc: Number of traffic classes in the net device
1477 * @tc_to_txq: XXX: need comments on this one
1478 * @prio_tc_map XXX: need comments on this one
1479 *
1480 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1481 *
1482 * @priomap: XXX: need comments on this one
1483 * @phydev: Physical device may attach itself
1484 * for hardware timestamping
1485 *
1486 * @qdisc_tx_busylock: XXX: need comments on this one
1487 *
1488 * @group: The group, that the device belongs to
1489 * @pm_qos_req: Power Management QoS object
1490 *
1491 * FIXME: cleanup struct net_device such that network protocol info
1492 * moves out.
1493 */
1494
1495 struct net_device {
1496 char name[IFNAMSIZ];
1497 struct hlist_node name_hlist;
1498 char *ifalias;
1499 /*
1500 * I/O specific fields
1501 * FIXME: Merge these and struct ifmap into one
1502 */
1503 unsigned long mem_end;
1504 unsigned long mem_start;
1505 unsigned long base_addr;
1506 int irq;
1507
1508 /*
1509 * Some hardware also needs these fields (state,dev_list,
1510 * napi_list,unreg_list,close_list) but they are not
1511 * part of the usual set specified in Space.c.
1512 */
1513
1514 unsigned long state;
1515
1516 struct list_head dev_list;
1517 struct list_head napi_list;
1518 struct list_head unreg_list;
1519 struct list_head close_list;
1520 struct list_head ptype_all;
1521 struct list_head ptype_specific;
1522
1523 struct {
1524 struct list_head upper;
1525 struct list_head lower;
1526 } adj_list;
1527
1528 struct {
1529 struct list_head upper;
1530 struct list_head lower;
1531 } all_adj_list;
1532
1533 netdev_features_t features;
1534 netdev_features_t hw_features;
1535 netdev_features_t wanted_features;
1536 netdev_features_t vlan_features;
1537 netdev_features_t hw_enc_features;
1538 netdev_features_t mpls_features;
1539
1540 int ifindex;
1541 int iflink;
1542
1543 struct net_device_stats stats;
1544
1545 atomic_long_t rx_dropped;
1546 atomic_long_t tx_dropped;
1547
1548 atomic_t carrier_changes;
1549
1550 #ifdef CONFIG_WIRELESS_EXT
1551 const struct iw_handler_def * wireless_handlers;
1552 struct iw_public_data * wireless_data;
1553 #endif
1554 const struct net_device_ops *netdev_ops;
1555 const struct ethtool_ops *ethtool_ops;
1556 #ifdef CONFIG_NET_SWITCHDEV
1557 const struct swdev_ops *swdev_ops;
1558 #endif
1559
1560 const struct header_ops *header_ops;
1561
1562 unsigned int flags;
1563 unsigned int priv_flags;
1564
1565 unsigned short gflags;
1566 unsigned short padded;
1567
1568 unsigned char operstate;
1569 unsigned char link_mode;
1570
1571 unsigned char if_port;
1572 unsigned char dma;
1573
1574 unsigned int mtu;
1575 unsigned short type;
1576 unsigned short hard_header_len;
1577
1578 unsigned short needed_headroom;
1579 unsigned short needed_tailroom;
1580
1581 /* Interface address info. */
1582 unsigned char perm_addr[MAX_ADDR_LEN];
1583 unsigned char addr_assign_type;
1584 unsigned char addr_len;
1585 unsigned short neigh_priv_len;
1586 unsigned short dev_id;
1587 unsigned short dev_port;
1588 spinlock_t addr_list_lock;
1589 struct netdev_hw_addr_list uc;
1590 struct netdev_hw_addr_list mc;
1591 struct netdev_hw_addr_list dev_addrs;
1592
1593 #ifdef CONFIG_SYSFS
1594 struct kset *queues_kset;
1595 #endif
1596
1597 unsigned char name_assign_type;
1598
1599 bool uc_promisc;
1600 unsigned int promiscuity;
1601 unsigned int allmulti;
1602
1603
1604 /* Protocol specific pointers */
1605
1606 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1607 struct vlan_info __rcu *vlan_info;
1608 #endif
1609 #if IS_ENABLED(CONFIG_NET_DSA)
1610 struct dsa_switch_tree *dsa_ptr;
1611 #endif
1612 #if IS_ENABLED(CONFIG_TIPC)
1613 struct tipc_bearer __rcu *tipc_ptr;
1614 #endif
1615 void *atalk_ptr;
1616 struct in_device __rcu *ip_ptr;
1617 struct dn_dev __rcu *dn_ptr;
1618 struct inet6_dev __rcu *ip6_ptr;
1619 void *ax25_ptr;
1620 struct wireless_dev *ieee80211_ptr;
1621 struct wpan_dev *ieee802154_ptr;
1622
1623 /*
1624 * Cache lines mostly used on receive path (including eth_type_trans())
1625 */
1626 unsigned long last_rx;
1627
1628 /* Interface address info used in eth_type_trans() */
1629 unsigned char *dev_addr;
1630
1631
1632 #ifdef CONFIG_SYSFS
1633 struct netdev_rx_queue *_rx;
1634
1635 unsigned int num_rx_queues;
1636 unsigned int real_num_rx_queues;
1637
1638 #endif
1639
1640 unsigned long gro_flush_timeout;
1641 rx_handler_func_t __rcu *rx_handler;
1642 void __rcu *rx_handler_data;
1643
1644 struct netdev_queue __rcu *ingress_queue;
1645 unsigned char broadcast[MAX_ADDR_LEN];
1646
1647
1648 /*
1649 * Cache lines mostly used on transmit path
1650 */
1651 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1652 unsigned int num_tx_queues;
1653 unsigned int real_num_tx_queues;
1654 struct Qdisc *qdisc;
1655 unsigned long tx_queue_len;
1656 spinlock_t tx_global_lock;
1657
1658 #ifdef CONFIG_XPS
1659 struct xps_dev_maps __rcu *xps_maps;
1660 #endif
1661 #ifdef CONFIG_RFS_ACCEL
1662 struct cpu_rmap *rx_cpu_rmap;
1663 #endif
1664
1665 /* These may be needed for future network-power-down code. */
1666
1667 /*
1668 * trans_start here is expensive for high speed devices on SMP,
1669 * please use netdev_queue->trans_start instead.
1670 */
1671 unsigned long trans_start;
1672
1673 int watchdog_timeo;
1674 struct timer_list watchdog_timer;
1675
1676 int __percpu *pcpu_refcnt;
1677 struct list_head todo_list;
1678
1679 struct hlist_node index_hlist;
1680 struct list_head link_watch_list;
1681
1682 enum { NETREG_UNINITIALIZED=0,
1683 NETREG_REGISTERED, /* completed register_netdevice */
1684 NETREG_UNREGISTERING, /* called unregister_netdevice */
1685 NETREG_UNREGISTERED, /* completed unregister todo */
1686 NETREG_RELEASED, /* called free_netdev */
1687 NETREG_DUMMY, /* dummy device for NAPI poll */
1688 } reg_state:8;
1689
1690 bool dismantle;
1691
1692 enum {
1693 RTNL_LINK_INITIALIZED,
1694 RTNL_LINK_INITIALIZING,
1695 } rtnl_link_state:16;
1696
1697 void (*destructor)(struct net_device *dev);
1698
1699 #ifdef CONFIG_NETPOLL
1700 struct netpoll_info __rcu *npinfo;
1701 #endif
1702
1703 possible_net_t nd_net;
1704
1705 /* mid-layer private */
1706 union {
1707 void *ml_priv;
1708 struct pcpu_lstats __percpu *lstats;
1709 struct pcpu_sw_netstats __percpu *tstats;
1710 struct pcpu_dstats __percpu *dstats;
1711 struct pcpu_vstats __percpu *vstats;
1712 };
1713
1714 struct garp_port __rcu *garp_port;
1715 struct mrp_port __rcu *mrp_port;
1716
1717 struct device dev;
1718 const struct attribute_group *sysfs_groups[4];
1719 const struct attribute_group *sysfs_rx_queue_group;
1720
1721 const struct rtnl_link_ops *rtnl_link_ops;
1722
1723 /* for setting kernel sock attribute on TCP connection setup */
1724 #define GSO_MAX_SIZE 65536
1725 unsigned int gso_max_size;
1726 #define GSO_MAX_SEGS 65535
1727 u16 gso_max_segs;
1728 u16 gso_min_segs;
1729 #ifdef CONFIG_DCB
1730 const struct dcbnl_rtnl_ops *dcbnl_ops;
1731 #endif
1732 u8 num_tc;
1733 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1734 u8 prio_tc_map[TC_BITMASK + 1];
1735
1736 #if IS_ENABLED(CONFIG_FCOE)
1737 unsigned int fcoe_ddp_xid;
1738 #endif
1739 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1740 struct netprio_map __rcu *priomap;
1741 #endif
1742 struct phy_device *phydev;
1743 struct lock_class_key *qdisc_tx_busylock;
1744 int group;
1745 struct pm_qos_request pm_qos_req;
1746 };
1747 #define to_net_dev(d) container_of(d, struct net_device, dev)
1748
1749 #define NETDEV_ALIGN 32
1750
1751 static inline
1752 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1753 {
1754 return dev->prio_tc_map[prio & TC_BITMASK];
1755 }
1756
1757 static inline
1758 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1759 {
1760 if (tc >= dev->num_tc)
1761 return -EINVAL;
1762
1763 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1764 return 0;
1765 }
1766
1767 static inline
1768 void netdev_reset_tc(struct net_device *dev)
1769 {
1770 dev->num_tc = 0;
1771 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1772 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1773 }
1774
1775 static inline
1776 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1777 {
1778 if (tc >= dev->num_tc)
1779 return -EINVAL;
1780
1781 dev->tc_to_txq[tc].count = count;
1782 dev->tc_to_txq[tc].offset = offset;
1783 return 0;
1784 }
1785
1786 static inline
1787 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1788 {
1789 if (num_tc > TC_MAX_QUEUE)
1790 return -EINVAL;
1791
1792 dev->num_tc = num_tc;
1793 return 0;
1794 }
1795
1796 static inline
1797 int netdev_get_num_tc(struct net_device *dev)
1798 {
1799 return dev->num_tc;
1800 }
1801
1802 static inline
1803 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1804 unsigned int index)
1805 {
1806 return &dev->_tx[index];
1807 }
1808
1809 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1810 const struct sk_buff *skb)
1811 {
1812 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1813 }
1814
1815 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1816 void (*f)(struct net_device *,
1817 struct netdev_queue *,
1818 void *),
1819 void *arg)
1820 {
1821 unsigned int i;
1822
1823 for (i = 0; i < dev->num_tx_queues; i++)
1824 f(dev, &dev->_tx[i], arg);
1825 }
1826
1827 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1828 struct sk_buff *skb,
1829 void *accel_priv);
1830
1831 /*
1832 * Net namespace inlines
1833 */
1834 static inline
1835 struct net *dev_net(const struct net_device *dev)
1836 {
1837 return read_pnet(&dev->nd_net);
1838 }
1839
1840 static inline
1841 void dev_net_set(struct net_device *dev, struct net *net)
1842 {
1843 write_pnet(&dev->nd_net, net);
1844 }
1845
1846 static inline bool netdev_uses_dsa(struct net_device *dev)
1847 {
1848 #if IS_ENABLED(CONFIG_NET_DSA)
1849 if (dev->dsa_ptr != NULL)
1850 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1851 #endif
1852 return false;
1853 }
1854
1855 /**
1856 * netdev_priv - access network device private data
1857 * @dev: network device
1858 *
1859 * Get network device private data
1860 */
1861 static inline void *netdev_priv(const struct net_device *dev)
1862 {
1863 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1864 }
1865
1866 /* Set the sysfs physical device reference for the network logical device
1867 * if set prior to registration will cause a symlink during initialization.
1868 */
1869 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1870
1871 /* Set the sysfs device type for the network logical device to allow
1872 * fine-grained identification of different network device types. For
1873 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1874 */
1875 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1876
1877 /* Default NAPI poll() weight
1878 * Device drivers are strongly advised to not use bigger value
1879 */
1880 #define NAPI_POLL_WEIGHT 64
1881
1882 /**
1883 * netif_napi_add - initialize a napi context
1884 * @dev: network device
1885 * @napi: napi context
1886 * @poll: polling function
1887 * @weight: default weight
1888 *
1889 * netif_napi_add() must be used to initialize a napi context prior to calling
1890 * *any* of the other napi related functions.
1891 */
1892 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1893 int (*poll)(struct napi_struct *, int), int weight);
1894
1895 /**
1896 * netif_napi_del - remove a napi context
1897 * @napi: napi context
1898 *
1899 * netif_napi_del() removes a napi context from the network device napi list
1900 */
1901 void netif_napi_del(struct napi_struct *napi);
1902
1903 struct napi_gro_cb {
1904 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1905 void *frag0;
1906
1907 /* Length of frag0. */
1908 unsigned int frag0_len;
1909
1910 /* This indicates where we are processing relative to skb->data. */
1911 int data_offset;
1912
1913 /* This is non-zero if the packet cannot be merged with the new skb. */
1914 u16 flush;
1915
1916 /* Save the IP ID here and check when we get to the transport layer */
1917 u16 flush_id;
1918
1919 /* Number of segments aggregated. */
1920 u16 count;
1921
1922 /* Start offset for remote checksum offload */
1923 u16 gro_remcsum_start;
1924
1925 /* jiffies when first packet was created/queued */
1926 unsigned long age;
1927
1928 /* Used in ipv6_gro_receive() and foo-over-udp */
1929 u16 proto;
1930
1931 /* This is non-zero if the packet may be of the same flow. */
1932 u8 same_flow:1;
1933
1934 /* Used in udp_gro_receive */
1935 u8 udp_mark:1;
1936
1937 /* GRO checksum is valid */
1938 u8 csum_valid:1;
1939
1940 /* Number of checksums via CHECKSUM_UNNECESSARY */
1941 u8 csum_cnt:3;
1942
1943 /* Free the skb? */
1944 u8 free:2;
1945 #define NAPI_GRO_FREE 1
1946 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1947
1948 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1949 u8 is_ipv6:1;
1950
1951 /* 7 bit hole */
1952
1953 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1954 __wsum csum;
1955
1956 /* used in skb_gro_receive() slow path */
1957 struct sk_buff *last;
1958 };
1959
1960 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1961
1962 struct packet_type {
1963 __be16 type; /* This is really htons(ether_type). */
1964 struct net_device *dev; /* NULL is wildcarded here */
1965 int (*func) (struct sk_buff *,
1966 struct net_device *,
1967 struct packet_type *,
1968 struct net_device *);
1969 bool (*id_match)(struct packet_type *ptype,
1970 struct sock *sk);
1971 void *af_packet_priv;
1972 struct list_head list;
1973 };
1974
1975 struct offload_callbacks {
1976 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1977 netdev_features_t features);
1978 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1979 struct sk_buff *skb);
1980 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1981 };
1982
1983 struct packet_offload {
1984 __be16 type; /* This is really htons(ether_type). */
1985 struct offload_callbacks callbacks;
1986 struct list_head list;
1987 };
1988
1989 struct udp_offload;
1990
1991 struct udp_offload_callbacks {
1992 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1993 struct sk_buff *skb,
1994 struct udp_offload *uoff);
1995 int (*gro_complete)(struct sk_buff *skb,
1996 int nhoff,
1997 struct udp_offload *uoff);
1998 };
1999
2000 struct udp_offload {
2001 __be16 port;
2002 u8 ipproto;
2003 struct udp_offload_callbacks callbacks;
2004 };
2005
2006 /* often modified stats are per cpu, other are shared (netdev->stats) */
2007 struct pcpu_sw_netstats {
2008 u64 rx_packets;
2009 u64 rx_bytes;
2010 u64 tx_packets;
2011 u64 tx_bytes;
2012 struct u64_stats_sync syncp;
2013 };
2014
2015 #define netdev_alloc_pcpu_stats(type) \
2016 ({ \
2017 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2018 if (pcpu_stats) { \
2019 int i; \
2020 for_each_possible_cpu(i) { \
2021 typeof(type) *stat; \
2022 stat = per_cpu_ptr(pcpu_stats, i); \
2023 u64_stats_init(&stat->syncp); \
2024 } \
2025 } \
2026 pcpu_stats; \
2027 })
2028
2029 #include <linux/notifier.h>
2030
2031 /* netdevice notifier chain. Please remember to update the rtnetlink
2032 * notification exclusion list in rtnetlink_event() when adding new
2033 * types.
2034 */
2035 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2036 #define NETDEV_DOWN 0x0002
2037 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2038 detected a hardware crash and restarted
2039 - we can use this eg to kick tcp sessions
2040 once done */
2041 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2042 #define NETDEV_REGISTER 0x0005
2043 #define NETDEV_UNREGISTER 0x0006
2044 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2045 #define NETDEV_CHANGEADDR 0x0008
2046 #define NETDEV_GOING_DOWN 0x0009
2047 #define NETDEV_CHANGENAME 0x000A
2048 #define NETDEV_FEAT_CHANGE 0x000B
2049 #define NETDEV_BONDING_FAILOVER 0x000C
2050 #define NETDEV_PRE_UP 0x000D
2051 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2052 #define NETDEV_POST_TYPE_CHANGE 0x000F
2053 #define NETDEV_POST_INIT 0x0010
2054 #define NETDEV_UNREGISTER_FINAL 0x0011
2055 #define NETDEV_RELEASE 0x0012
2056 #define NETDEV_NOTIFY_PEERS 0x0013
2057 #define NETDEV_JOIN 0x0014
2058 #define NETDEV_CHANGEUPPER 0x0015
2059 #define NETDEV_RESEND_IGMP 0x0016
2060 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2061 #define NETDEV_CHANGEINFODATA 0x0018
2062 #define NETDEV_BONDING_INFO 0x0019
2063
2064 int register_netdevice_notifier(struct notifier_block *nb);
2065 int unregister_netdevice_notifier(struct notifier_block *nb);
2066
2067 struct netdev_notifier_info {
2068 struct net_device *dev;
2069 };
2070
2071 struct netdev_notifier_change_info {
2072 struct netdev_notifier_info info; /* must be first */
2073 unsigned int flags_changed;
2074 };
2075
2076 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2077 struct net_device *dev)
2078 {
2079 info->dev = dev;
2080 }
2081
2082 static inline struct net_device *
2083 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2084 {
2085 return info->dev;
2086 }
2087
2088 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2089
2090
2091 extern rwlock_t dev_base_lock; /* Device list lock */
2092
2093 #define for_each_netdev(net, d) \
2094 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2095 #define for_each_netdev_reverse(net, d) \
2096 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2097 #define for_each_netdev_rcu(net, d) \
2098 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2099 #define for_each_netdev_safe(net, d, n) \
2100 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2101 #define for_each_netdev_continue(net, d) \
2102 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2103 #define for_each_netdev_continue_rcu(net, d) \
2104 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2105 #define for_each_netdev_in_bond_rcu(bond, slave) \
2106 for_each_netdev_rcu(&init_net, slave) \
2107 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2108 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2109
2110 static inline struct net_device *next_net_device(struct net_device *dev)
2111 {
2112 struct list_head *lh;
2113 struct net *net;
2114
2115 net = dev_net(dev);
2116 lh = dev->dev_list.next;
2117 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2118 }
2119
2120 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2121 {
2122 struct list_head *lh;
2123 struct net *net;
2124
2125 net = dev_net(dev);
2126 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2127 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2128 }
2129
2130 static inline struct net_device *first_net_device(struct net *net)
2131 {
2132 return list_empty(&net->dev_base_head) ? NULL :
2133 net_device_entry(net->dev_base_head.next);
2134 }
2135
2136 static inline struct net_device *first_net_device_rcu(struct net *net)
2137 {
2138 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2139
2140 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2141 }
2142
2143 int netdev_boot_setup_check(struct net_device *dev);
2144 unsigned long netdev_boot_base(const char *prefix, int unit);
2145 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2146 const char *hwaddr);
2147 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2148 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2149 void dev_add_pack(struct packet_type *pt);
2150 void dev_remove_pack(struct packet_type *pt);
2151 void __dev_remove_pack(struct packet_type *pt);
2152 void dev_add_offload(struct packet_offload *po);
2153 void dev_remove_offload(struct packet_offload *po);
2154
2155 int dev_get_iflink(const struct net_device *dev);
2156 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2157 unsigned short mask);
2158 struct net_device *dev_get_by_name(struct net *net, const char *name);
2159 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2160 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2161 int dev_alloc_name(struct net_device *dev, const char *name);
2162 int dev_open(struct net_device *dev);
2163 int dev_close(struct net_device *dev);
2164 int dev_close_many(struct list_head *head, bool unlink);
2165 void dev_disable_lro(struct net_device *dev);
2166 int dev_loopback_xmit(struct sk_buff *newskb);
2167 int dev_queue_xmit(struct sk_buff *skb);
2168 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2169 int register_netdevice(struct net_device *dev);
2170 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2171 void unregister_netdevice_many(struct list_head *head);
2172 static inline void unregister_netdevice(struct net_device *dev)
2173 {
2174 unregister_netdevice_queue(dev, NULL);
2175 }
2176
2177 int netdev_refcnt_read(const struct net_device *dev);
2178 void free_netdev(struct net_device *dev);
2179 void netdev_freemem(struct net_device *dev);
2180 void synchronize_net(void);
2181 int init_dummy_netdev(struct net_device *dev);
2182
2183 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2184 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2185 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2186 int netdev_get_name(struct net *net, char *name, int ifindex);
2187 int dev_restart(struct net_device *dev);
2188 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2189
2190 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2191 {
2192 return NAPI_GRO_CB(skb)->data_offset;
2193 }
2194
2195 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2196 {
2197 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2198 }
2199
2200 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2201 {
2202 NAPI_GRO_CB(skb)->data_offset += len;
2203 }
2204
2205 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2206 unsigned int offset)
2207 {
2208 return NAPI_GRO_CB(skb)->frag0 + offset;
2209 }
2210
2211 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2212 {
2213 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2214 }
2215
2216 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2217 unsigned int offset)
2218 {
2219 if (!pskb_may_pull(skb, hlen))
2220 return NULL;
2221
2222 NAPI_GRO_CB(skb)->frag0 = NULL;
2223 NAPI_GRO_CB(skb)->frag0_len = 0;
2224 return skb->data + offset;
2225 }
2226
2227 static inline void *skb_gro_network_header(struct sk_buff *skb)
2228 {
2229 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2230 skb_network_offset(skb);
2231 }
2232
2233 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2234 const void *start, unsigned int len)
2235 {
2236 if (NAPI_GRO_CB(skb)->csum_valid)
2237 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2238 csum_partial(start, len, 0));
2239 }
2240
2241 /* GRO checksum functions. These are logical equivalents of the normal
2242 * checksum functions (in skbuff.h) except that they operate on the GRO
2243 * offsets and fields in sk_buff.
2244 */
2245
2246 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2247
2248 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2249 {
2250 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2251 skb_gro_offset(skb));
2252 }
2253
2254 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2255 bool zero_okay,
2256 __sum16 check)
2257 {
2258 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2259 skb_checksum_start_offset(skb) <
2260 skb_gro_offset(skb)) &&
2261 !skb_at_gro_remcsum_start(skb) &&
2262 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2263 (!zero_okay || check));
2264 }
2265
2266 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2267 __wsum psum)
2268 {
2269 if (NAPI_GRO_CB(skb)->csum_valid &&
2270 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2271 return 0;
2272
2273 NAPI_GRO_CB(skb)->csum = psum;
2274
2275 return __skb_gro_checksum_complete(skb);
2276 }
2277
2278 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2279 {
2280 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2281 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2282 NAPI_GRO_CB(skb)->csum_cnt--;
2283 } else {
2284 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2285 * verified a new top level checksum or an encapsulated one
2286 * during GRO. This saves work if we fallback to normal path.
2287 */
2288 __skb_incr_checksum_unnecessary(skb);
2289 }
2290 }
2291
2292 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2293 compute_pseudo) \
2294 ({ \
2295 __sum16 __ret = 0; \
2296 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2297 __ret = __skb_gro_checksum_validate_complete(skb, \
2298 compute_pseudo(skb, proto)); \
2299 if (__ret) \
2300 __skb_mark_checksum_bad(skb); \
2301 else \
2302 skb_gro_incr_csum_unnecessary(skb); \
2303 __ret; \
2304 })
2305
2306 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2307 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2308
2309 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2310 compute_pseudo) \
2311 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2312
2313 #define skb_gro_checksum_simple_validate(skb) \
2314 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2315
2316 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2317 {
2318 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2319 !NAPI_GRO_CB(skb)->csum_valid);
2320 }
2321
2322 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2323 __sum16 check, __wsum pseudo)
2324 {
2325 NAPI_GRO_CB(skb)->csum = ~pseudo;
2326 NAPI_GRO_CB(skb)->csum_valid = 1;
2327 }
2328
2329 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2330 do { \
2331 if (__skb_gro_checksum_convert_check(skb)) \
2332 __skb_gro_checksum_convert(skb, check, \
2333 compute_pseudo(skb, proto)); \
2334 } while (0)
2335
2336 struct gro_remcsum {
2337 int offset;
2338 __wsum delta;
2339 };
2340
2341 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2342 {
2343 grc->offset = 0;
2344 grc->delta = 0;
2345 }
2346
2347 static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2348 int start, int offset,
2349 struct gro_remcsum *grc,
2350 bool nopartial)
2351 {
2352 __wsum delta;
2353
2354 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2355
2356 if (!nopartial) {
2357 NAPI_GRO_CB(skb)->gro_remcsum_start =
2358 ((unsigned char *)ptr + start) - skb->head;
2359 return;
2360 }
2361
2362 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2363
2364 /* Adjust skb->csum since we changed the packet */
2365 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2366
2367 grc->offset = (ptr + offset) - (void *)skb->head;
2368 grc->delta = delta;
2369 }
2370
2371 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2372 struct gro_remcsum *grc)
2373 {
2374 if (!grc->delta)
2375 return;
2376
2377 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2378 }
2379
2380 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2381 unsigned short type,
2382 const void *daddr, const void *saddr,
2383 unsigned int len)
2384 {
2385 if (!dev->header_ops || !dev->header_ops->create)
2386 return 0;
2387
2388 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2389 }
2390
2391 static inline int dev_parse_header(const struct sk_buff *skb,
2392 unsigned char *haddr)
2393 {
2394 const struct net_device *dev = skb->dev;
2395
2396 if (!dev->header_ops || !dev->header_ops->parse)
2397 return 0;
2398 return dev->header_ops->parse(skb, haddr);
2399 }
2400
2401 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2402 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2403 static inline int unregister_gifconf(unsigned int family)
2404 {
2405 return register_gifconf(family, NULL);
2406 }
2407
2408 #ifdef CONFIG_NET_FLOW_LIMIT
2409 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2410 struct sd_flow_limit {
2411 u64 count;
2412 unsigned int num_buckets;
2413 unsigned int history_head;
2414 u16 history[FLOW_LIMIT_HISTORY];
2415 u8 buckets[];
2416 };
2417
2418 extern int netdev_flow_limit_table_len;
2419 #endif /* CONFIG_NET_FLOW_LIMIT */
2420
2421 /*
2422 * Incoming packets are placed on per-cpu queues
2423 */
2424 struct softnet_data {
2425 struct list_head poll_list;
2426 struct sk_buff_head process_queue;
2427
2428 /* stats */
2429 unsigned int processed;
2430 unsigned int time_squeeze;
2431 unsigned int cpu_collision;
2432 unsigned int received_rps;
2433 #ifdef CONFIG_RPS
2434 struct softnet_data *rps_ipi_list;
2435 #endif
2436 #ifdef CONFIG_NET_FLOW_LIMIT
2437 struct sd_flow_limit __rcu *flow_limit;
2438 #endif
2439 struct Qdisc *output_queue;
2440 struct Qdisc **output_queue_tailp;
2441 struct sk_buff *completion_queue;
2442
2443 #ifdef CONFIG_RPS
2444 /* Elements below can be accessed between CPUs for RPS */
2445 struct call_single_data csd ____cacheline_aligned_in_smp;
2446 struct softnet_data *rps_ipi_next;
2447 unsigned int cpu;
2448 unsigned int input_queue_head;
2449 unsigned int input_queue_tail;
2450 #endif
2451 unsigned int dropped;
2452 struct sk_buff_head input_pkt_queue;
2453 struct napi_struct backlog;
2454
2455 };
2456
2457 static inline void input_queue_head_incr(struct softnet_data *sd)
2458 {
2459 #ifdef CONFIG_RPS
2460 sd->input_queue_head++;
2461 #endif
2462 }
2463
2464 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2465 unsigned int *qtail)
2466 {
2467 #ifdef CONFIG_RPS
2468 *qtail = ++sd->input_queue_tail;
2469 #endif
2470 }
2471
2472 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2473
2474 void __netif_schedule(struct Qdisc *q);
2475 void netif_schedule_queue(struct netdev_queue *txq);
2476
2477 static inline void netif_tx_schedule_all(struct net_device *dev)
2478 {
2479 unsigned int i;
2480
2481 for (i = 0; i < dev->num_tx_queues; i++)
2482 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2483 }
2484
2485 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2486 {
2487 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2488 }
2489
2490 /**
2491 * netif_start_queue - allow transmit
2492 * @dev: network device
2493 *
2494 * Allow upper layers to call the device hard_start_xmit routine.
2495 */
2496 static inline void netif_start_queue(struct net_device *dev)
2497 {
2498 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2499 }
2500
2501 static inline void netif_tx_start_all_queues(struct net_device *dev)
2502 {
2503 unsigned int i;
2504
2505 for (i = 0; i < dev->num_tx_queues; i++) {
2506 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2507 netif_tx_start_queue(txq);
2508 }
2509 }
2510
2511 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2512
2513 /**
2514 * netif_wake_queue - restart transmit
2515 * @dev: network device
2516 *
2517 * Allow upper layers to call the device hard_start_xmit routine.
2518 * Used for flow control when transmit resources are available.
2519 */
2520 static inline void netif_wake_queue(struct net_device *dev)
2521 {
2522 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2523 }
2524
2525 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2526 {
2527 unsigned int i;
2528
2529 for (i = 0; i < dev->num_tx_queues; i++) {
2530 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2531 netif_tx_wake_queue(txq);
2532 }
2533 }
2534
2535 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2536 {
2537 if (WARN_ON(!dev_queue)) {
2538 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2539 return;
2540 }
2541 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2542 }
2543
2544 /**
2545 * netif_stop_queue - stop transmitted packets
2546 * @dev: network device
2547 *
2548 * Stop upper layers calling the device hard_start_xmit routine.
2549 * Used for flow control when transmit resources are unavailable.
2550 */
2551 static inline void netif_stop_queue(struct net_device *dev)
2552 {
2553 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2554 }
2555
2556 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2557 {
2558 unsigned int i;
2559
2560 for (i = 0; i < dev->num_tx_queues; i++) {
2561 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2562 netif_tx_stop_queue(txq);
2563 }
2564 }
2565
2566 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2567 {
2568 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2569 }
2570
2571 /**
2572 * netif_queue_stopped - test if transmit queue is flowblocked
2573 * @dev: network device
2574 *
2575 * Test if transmit queue on device is currently unable to send.
2576 */
2577 static inline bool netif_queue_stopped(const struct net_device *dev)
2578 {
2579 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2580 }
2581
2582 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2583 {
2584 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2585 }
2586
2587 static inline bool
2588 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2589 {
2590 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2591 }
2592
2593 static inline bool
2594 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2595 {
2596 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2597 }
2598
2599 /**
2600 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2601 * @dev_queue: pointer to transmit queue
2602 *
2603 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2604 * to give appropriate hint to the cpu.
2605 */
2606 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2607 {
2608 #ifdef CONFIG_BQL
2609 prefetchw(&dev_queue->dql.num_queued);
2610 #endif
2611 }
2612
2613 /**
2614 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2615 * @dev_queue: pointer to transmit queue
2616 *
2617 * BQL enabled drivers might use this helper in their TX completion path,
2618 * to give appropriate hint to the cpu.
2619 */
2620 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2621 {
2622 #ifdef CONFIG_BQL
2623 prefetchw(&dev_queue->dql.limit);
2624 #endif
2625 }
2626
2627 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2628 unsigned int bytes)
2629 {
2630 #ifdef CONFIG_BQL
2631 dql_queued(&dev_queue->dql, bytes);
2632
2633 if (likely(dql_avail(&dev_queue->dql) >= 0))
2634 return;
2635
2636 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2637
2638 /*
2639 * The XOFF flag must be set before checking the dql_avail below,
2640 * because in netdev_tx_completed_queue we update the dql_completed
2641 * before checking the XOFF flag.
2642 */
2643 smp_mb();
2644
2645 /* check again in case another CPU has just made room avail */
2646 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2647 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2648 #endif
2649 }
2650
2651 /**
2652 * netdev_sent_queue - report the number of bytes queued to hardware
2653 * @dev: network device
2654 * @bytes: number of bytes queued to the hardware device queue
2655 *
2656 * Report the number of bytes queued for sending/completion to the network
2657 * device hardware queue. @bytes should be a good approximation and should
2658 * exactly match netdev_completed_queue() @bytes
2659 */
2660 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2661 {
2662 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2663 }
2664
2665 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2666 unsigned int pkts, unsigned int bytes)
2667 {
2668 #ifdef CONFIG_BQL
2669 if (unlikely(!bytes))
2670 return;
2671
2672 dql_completed(&dev_queue->dql, bytes);
2673
2674 /*
2675 * Without the memory barrier there is a small possiblity that
2676 * netdev_tx_sent_queue will miss the update and cause the queue to
2677 * be stopped forever
2678 */
2679 smp_mb();
2680
2681 if (dql_avail(&dev_queue->dql) < 0)
2682 return;
2683
2684 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2685 netif_schedule_queue(dev_queue);
2686 #endif
2687 }
2688
2689 /**
2690 * netdev_completed_queue - report bytes and packets completed by device
2691 * @dev: network device
2692 * @pkts: actual number of packets sent over the medium
2693 * @bytes: actual number of bytes sent over the medium
2694 *
2695 * Report the number of bytes and packets transmitted by the network device
2696 * hardware queue over the physical medium, @bytes must exactly match the
2697 * @bytes amount passed to netdev_sent_queue()
2698 */
2699 static inline void netdev_completed_queue(struct net_device *dev,
2700 unsigned int pkts, unsigned int bytes)
2701 {
2702 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2703 }
2704
2705 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2706 {
2707 #ifdef CONFIG_BQL
2708 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2709 dql_reset(&q->dql);
2710 #endif
2711 }
2712
2713 /**
2714 * netdev_reset_queue - reset the packets and bytes count of a network device
2715 * @dev_queue: network device
2716 *
2717 * Reset the bytes and packet count of a network device and clear the
2718 * software flow control OFF bit for this network device
2719 */
2720 static inline void netdev_reset_queue(struct net_device *dev_queue)
2721 {
2722 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2723 }
2724
2725 /**
2726 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2727 * @dev: network device
2728 * @queue_index: given tx queue index
2729 *
2730 * Returns 0 if given tx queue index >= number of device tx queues,
2731 * otherwise returns the originally passed tx queue index.
2732 */
2733 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2734 {
2735 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2736 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2737 dev->name, queue_index,
2738 dev->real_num_tx_queues);
2739 return 0;
2740 }
2741
2742 return queue_index;
2743 }
2744
2745 /**
2746 * netif_running - test if up
2747 * @dev: network device
2748 *
2749 * Test if the device has been brought up.
2750 */
2751 static inline bool netif_running(const struct net_device *dev)
2752 {
2753 return test_bit(__LINK_STATE_START, &dev->state);
2754 }
2755
2756 /*
2757 * Routines to manage the subqueues on a device. We only need start
2758 * stop, and a check if it's stopped. All other device management is
2759 * done at the overall netdevice level.
2760 * Also test the device if we're multiqueue.
2761 */
2762
2763 /**
2764 * netif_start_subqueue - allow sending packets on subqueue
2765 * @dev: network device
2766 * @queue_index: sub queue index
2767 *
2768 * Start individual transmit queue of a device with multiple transmit queues.
2769 */
2770 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2771 {
2772 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2773
2774 netif_tx_start_queue(txq);
2775 }
2776
2777 /**
2778 * netif_stop_subqueue - stop sending packets on subqueue
2779 * @dev: network device
2780 * @queue_index: sub queue index
2781 *
2782 * Stop individual transmit queue of a device with multiple transmit queues.
2783 */
2784 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2785 {
2786 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2787 netif_tx_stop_queue(txq);
2788 }
2789
2790 /**
2791 * netif_subqueue_stopped - test status of subqueue
2792 * @dev: network device
2793 * @queue_index: sub queue index
2794 *
2795 * Check individual transmit queue of a device with multiple transmit queues.
2796 */
2797 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2798 u16 queue_index)
2799 {
2800 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2801
2802 return netif_tx_queue_stopped(txq);
2803 }
2804
2805 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2806 struct sk_buff *skb)
2807 {
2808 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2809 }
2810
2811 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2812
2813 #ifdef CONFIG_XPS
2814 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2815 u16 index);
2816 #else
2817 static inline int netif_set_xps_queue(struct net_device *dev,
2818 const struct cpumask *mask,
2819 u16 index)
2820 {
2821 return 0;
2822 }
2823 #endif
2824
2825 /*
2826 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2827 * as a distribution range limit for the returned value.
2828 */
2829 static inline u16 skb_tx_hash(const struct net_device *dev,
2830 struct sk_buff *skb)
2831 {
2832 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2833 }
2834
2835 /**
2836 * netif_is_multiqueue - test if device has multiple transmit queues
2837 * @dev: network device
2838 *
2839 * Check if device has multiple transmit queues
2840 */
2841 static inline bool netif_is_multiqueue(const struct net_device *dev)
2842 {
2843 return dev->num_tx_queues > 1;
2844 }
2845
2846 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2847
2848 #ifdef CONFIG_SYSFS
2849 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2850 #else
2851 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2852 unsigned int rxq)
2853 {
2854 return 0;
2855 }
2856 #endif
2857
2858 #ifdef CONFIG_SYSFS
2859 static inline unsigned int get_netdev_rx_queue_index(
2860 struct netdev_rx_queue *queue)
2861 {
2862 struct net_device *dev = queue->dev;
2863 int index = queue - dev->_rx;
2864
2865 BUG_ON(index >= dev->num_rx_queues);
2866 return index;
2867 }
2868 #endif
2869
2870 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2871 int netif_get_num_default_rss_queues(void);
2872
2873 enum skb_free_reason {
2874 SKB_REASON_CONSUMED,
2875 SKB_REASON_DROPPED,
2876 };
2877
2878 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2879 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2880
2881 /*
2882 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2883 * interrupt context or with hardware interrupts being disabled.
2884 * (in_irq() || irqs_disabled())
2885 *
2886 * We provide four helpers that can be used in following contexts :
2887 *
2888 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2889 * replacing kfree_skb(skb)
2890 *
2891 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2892 * Typically used in place of consume_skb(skb) in TX completion path
2893 *
2894 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2895 * replacing kfree_skb(skb)
2896 *
2897 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2898 * and consumed a packet. Used in place of consume_skb(skb)
2899 */
2900 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2901 {
2902 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2903 }
2904
2905 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2906 {
2907 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2908 }
2909
2910 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2911 {
2912 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2913 }
2914
2915 static inline void dev_consume_skb_any(struct sk_buff *skb)
2916 {
2917 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2918 }
2919
2920 int netif_rx(struct sk_buff *skb);
2921 int netif_rx_ni(struct sk_buff *skb);
2922 int netif_receive_skb(struct sk_buff *skb);
2923 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2924 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2925 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2926 gro_result_t napi_gro_frags(struct napi_struct *napi);
2927 struct packet_offload *gro_find_receive_by_type(__be16 type);
2928 struct packet_offload *gro_find_complete_by_type(__be16 type);
2929
2930 static inline void napi_free_frags(struct napi_struct *napi)
2931 {
2932 kfree_skb(napi->skb);
2933 napi->skb = NULL;
2934 }
2935
2936 int netdev_rx_handler_register(struct net_device *dev,
2937 rx_handler_func_t *rx_handler,
2938 void *rx_handler_data);
2939 void netdev_rx_handler_unregister(struct net_device *dev);
2940
2941 bool dev_valid_name(const char *name);
2942 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2943 int dev_ethtool(struct net *net, struct ifreq *);
2944 unsigned int dev_get_flags(const struct net_device *);
2945 int __dev_change_flags(struct net_device *, unsigned int flags);
2946 int dev_change_flags(struct net_device *, unsigned int);
2947 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2948 unsigned int gchanges);
2949 int dev_change_name(struct net_device *, const char *);
2950 int dev_set_alias(struct net_device *, const char *, size_t);
2951 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2952 int dev_set_mtu(struct net_device *, int);
2953 void dev_set_group(struct net_device *, int);
2954 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2955 int dev_change_carrier(struct net_device *, bool new_carrier);
2956 int dev_get_phys_port_id(struct net_device *dev,
2957 struct netdev_phys_item_id *ppid);
2958 int dev_get_phys_port_name(struct net_device *dev,
2959 char *name, size_t len);
2960 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2961 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2962 struct netdev_queue *txq, int *ret);
2963 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2964 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2965 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2966
2967 extern int netdev_budget;
2968
2969 /* Called by rtnetlink.c:rtnl_unlock() */
2970 void netdev_run_todo(void);
2971
2972 /**
2973 * dev_put - release reference to device
2974 * @dev: network device
2975 *
2976 * Release reference to device to allow it to be freed.
2977 */
2978 static inline void dev_put(struct net_device *dev)
2979 {
2980 this_cpu_dec(*dev->pcpu_refcnt);
2981 }
2982
2983 /**
2984 * dev_hold - get reference to device
2985 * @dev: network device
2986 *
2987 * Hold reference to device to keep it from being freed.
2988 */
2989 static inline void dev_hold(struct net_device *dev)
2990 {
2991 this_cpu_inc(*dev->pcpu_refcnt);
2992 }
2993
2994 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
2995 * and _off may be called from IRQ context, but it is caller
2996 * who is responsible for serialization of these calls.
2997 *
2998 * The name carrier is inappropriate, these functions should really be
2999 * called netif_lowerlayer_*() because they represent the state of any
3000 * kind of lower layer not just hardware media.
3001 */
3002
3003 void linkwatch_init_dev(struct net_device *dev);
3004 void linkwatch_fire_event(struct net_device *dev);
3005 void linkwatch_forget_dev(struct net_device *dev);
3006
3007 /**
3008 * netif_carrier_ok - test if carrier present
3009 * @dev: network device
3010 *
3011 * Check if carrier is present on device
3012 */
3013 static inline bool netif_carrier_ok(const struct net_device *dev)
3014 {
3015 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3016 }
3017
3018 unsigned long dev_trans_start(struct net_device *dev);
3019
3020 void __netdev_watchdog_up(struct net_device *dev);
3021
3022 void netif_carrier_on(struct net_device *dev);
3023
3024 void netif_carrier_off(struct net_device *dev);
3025
3026 /**
3027 * netif_dormant_on - mark device as dormant.
3028 * @dev: network device
3029 *
3030 * Mark device as dormant (as per RFC2863).
3031 *
3032 * The dormant state indicates that the relevant interface is not
3033 * actually in a condition to pass packets (i.e., it is not 'up') but is
3034 * in a "pending" state, waiting for some external event. For "on-
3035 * demand" interfaces, this new state identifies the situation where the
3036 * interface is waiting for events to place it in the up state.
3037 *
3038 */
3039 static inline void netif_dormant_on(struct net_device *dev)
3040 {
3041 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3042 linkwatch_fire_event(dev);
3043 }
3044
3045 /**
3046 * netif_dormant_off - set device as not dormant.
3047 * @dev: network device
3048 *
3049 * Device is not in dormant state.
3050 */
3051 static inline void netif_dormant_off(struct net_device *dev)
3052 {
3053 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3054 linkwatch_fire_event(dev);
3055 }
3056
3057 /**
3058 * netif_dormant - test if carrier present
3059 * @dev: network device
3060 *
3061 * Check if carrier is present on device
3062 */
3063 static inline bool netif_dormant(const struct net_device *dev)
3064 {
3065 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3066 }
3067
3068
3069 /**
3070 * netif_oper_up - test if device is operational
3071 * @dev: network device
3072 *
3073 * Check if carrier is operational
3074 */
3075 static inline bool netif_oper_up(const struct net_device *dev)
3076 {
3077 return (dev->operstate == IF_OPER_UP ||
3078 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3079 }
3080
3081 /**
3082 * netif_device_present - is device available or removed
3083 * @dev: network device
3084 *
3085 * Check if device has not been removed from system.
3086 */
3087 static inline bool netif_device_present(struct net_device *dev)
3088 {
3089 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3090 }
3091
3092 void netif_device_detach(struct net_device *dev);
3093
3094 void netif_device_attach(struct net_device *dev);
3095
3096 /*
3097 * Network interface message level settings
3098 */
3099
3100 enum {
3101 NETIF_MSG_DRV = 0x0001,
3102 NETIF_MSG_PROBE = 0x0002,
3103 NETIF_MSG_LINK = 0x0004,
3104 NETIF_MSG_TIMER = 0x0008,
3105 NETIF_MSG_IFDOWN = 0x0010,
3106 NETIF_MSG_IFUP = 0x0020,
3107 NETIF_MSG_RX_ERR = 0x0040,
3108 NETIF_MSG_TX_ERR = 0x0080,
3109 NETIF_MSG_TX_QUEUED = 0x0100,
3110 NETIF_MSG_INTR = 0x0200,
3111 NETIF_MSG_TX_DONE = 0x0400,
3112 NETIF_MSG_RX_STATUS = 0x0800,
3113 NETIF_MSG_PKTDATA = 0x1000,
3114 NETIF_MSG_HW = 0x2000,
3115 NETIF_MSG_WOL = 0x4000,
3116 };
3117
3118 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3119 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3120 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3121 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3122 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3123 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3124 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3125 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3126 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3127 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3128 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3129 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3130 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3131 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3132 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3133
3134 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3135 {
3136 /* use default */
3137 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3138 return default_msg_enable_bits;
3139 if (debug_value == 0) /* no output */
3140 return 0;
3141 /* set low N bits */
3142 return (1 << debug_value) - 1;
3143 }
3144
3145 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3146 {
3147 spin_lock(&txq->_xmit_lock);
3148 txq->xmit_lock_owner = cpu;
3149 }
3150
3151 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3152 {
3153 spin_lock_bh(&txq->_xmit_lock);
3154 txq->xmit_lock_owner = smp_processor_id();
3155 }
3156
3157 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3158 {
3159 bool ok = spin_trylock(&txq->_xmit_lock);
3160 if (likely(ok))
3161 txq->xmit_lock_owner = smp_processor_id();
3162 return ok;
3163 }
3164
3165 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3166 {
3167 txq->xmit_lock_owner = -1;
3168 spin_unlock(&txq->_xmit_lock);
3169 }
3170
3171 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3172 {
3173 txq->xmit_lock_owner = -1;
3174 spin_unlock_bh(&txq->_xmit_lock);
3175 }
3176
3177 static inline void txq_trans_update(struct netdev_queue *txq)
3178 {
3179 if (txq->xmit_lock_owner != -1)
3180 txq->trans_start = jiffies;
3181 }
3182
3183 /**
3184 * netif_tx_lock - grab network device transmit lock
3185 * @dev: network device
3186 *
3187 * Get network device transmit lock
3188 */
3189 static inline void netif_tx_lock(struct net_device *dev)
3190 {
3191 unsigned int i;
3192 int cpu;
3193
3194 spin_lock(&dev->tx_global_lock);
3195 cpu = smp_processor_id();
3196 for (i = 0; i < dev->num_tx_queues; i++) {
3197 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3198
3199 /* We are the only thread of execution doing a
3200 * freeze, but we have to grab the _xmit_lock in
3201 * order to synchronize with threads which are in
3202 * the ->hard_start_xmit() handler and already
3203 * checked the frozen bit.
3204 */
3205 __netif_tx_lock(txq, cpu);
3206 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3207 __netif_tx_unlock(txq);
3208 }
3209 }
3210
3211 static inline void netif_tx_lock_bh(struct net_device *dev)
3212 {
3213 local_bh_disable();
3214 netif_tx_lock(dev);
3215 }
3216
3217 static inline void netif_tx_unlock(struct net_device *dev)
3218 {
3219 unsigned int i;
3220
3221 for (i = 0; i < dev->num_tx_queues; i++) {
3222 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3223
3224 /* No need to grab the _xmit_lock here. If the
3225 * queue is not stopped for another reason, we
3226 * force a schedule.
3227 */
3228 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3229 netif_schedule_queue(txq);
3230 }
3231 spin_unlock(&dev->tx_global_lock);
3232 }
3233
3234 static inline void netif_tx_unlock_bh(struct net_device *dev)
3235 {
3236 netif_tx_unlock(dev);
3237 local_bh_enable();
3238 }
3239
3240 #define HARD_TX_LOCK(dev, txq, cpu) { \
3241 if ((dev->features & NETIF_F_LLTX) == 0) { \
3242 __netif_tx_lock(txq, cpu); \
3243 } \
3244 }
3245
3246 #define HARD_TX_TRYLOCK(dev, txq) \
3247 (((dev->features & NETIF_F_LLTX) == 0) ? \
3248 __netif_tx_trylock(txq) : \
3249 true )
3250
3251 #define HARD_TX_UNLOCK(dev, txq) { \
3252 if ((dev->features & NETIF_F_LLTX) == 0) { \
3253 __netif_tx_unlock(txq); \
3254 } \
3255 }
3256
3257 static inline void netif_tx_disable(struct net_device *dev)
3258 {
3259 unsigned int i;
3260 int cpu;
3261
3262 local_bh_disable();
3263 cpu = smp_processor_id();
3264 for (i = 0; i < dev->num_tx_queues; i++) {
3265 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3266
3267 __netif_tx_lock(txq, cpu);
3268 netif_tx_stop_queue(txq);
3269 __netif_tx_unlock(txq);
3270 }
3271 local_bh_enable();
3272 }
3273
3274 static inline void netif_addr_lock(struct net_device *dev)
3275 {
3276 spin_lock(&dev->addr_list_lock);
3277 }
3278
3279 static inline void netif_addr_lock_nested(struct net_device *dev)
3280 {
3281 int subclass = SINGLE_DEPTH_NESTING;
3282
3283 if (dev->netdev_ops->ndo_get_lock_subclass)
3284 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3285
3286 spin_lock_nested(&dev->addr_list_lock, subclass);
3287 }
3288
3289 static inline void netif_addr_lock_bh(struct net_device *dev)
3290 {
3291 spin_lock_bh(&dev->addr_list_lock);
3292 }
3293
3294 static inline void netif_addr_unlock(struct net_device *dev)
3295 {
3296 spin_unlock(&dev->addr_list_lock);
3297 }
3298
3299 static inline void netif_addr_unlock_bh(struct net_device *dev)
3300 {
3301 spin_unlock_bh(&dev->addr_list_lock);
3302 }
3303
3304 /*
3305 * dev_addrs walker. Should be used only for read access. Call with
3306 * rcu_read_lock held.
3307 */
3308 #define for_each_dev_addr(dev, ha) \
3309 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3310
3311 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3312
3313 void ether_setup(struct net_device *dev);
3314
3315 /* Support for loadable net-drivers */
3316 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3317 unsigned char name_assign_type,
3318 void (*setup)(struct net_device *),
3319 unsigned int txqs, unsigned int rxqs);
3320 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3321 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3322
3323 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3324 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3325 count)
3326
3327 int register_netdev(struct net_device *dev);
3328 void unregister_netdev(struct net_device *dev);
3329
3330 /* General hardware address lists handling functions */
3331 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3332 struct netdev_hw_addr_list *from_list, int addr_len);
3333 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3334 struct netdev_hw_addr_list *from_list, int addr_len);
3335 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3336 struct net_device *dev,
3337 int (*sync)(struct net_device *, const unsigned char *),
3338 int (*unsync)(struct net_device *,
3339 const unsigned char *));
3340 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3341 struct net_device *dev,
3342 int (*unsync)(struct net_device *,
3343 const unsigned char *));
3344 void __hw_addr_init(struct netdev_hw_addr_list *list);
3345
3346 /* Functions used for device addresses handling */
3347 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3348 unsigned char addr_type);
3349 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3350 unsigned char addr_type);
3351 void dev_addr_flush(struct net_device *dev);
3352 int dev_addr_init(struct net_device *dev);
3353
3354 /* Functions used for unicast addresses handling */
3355 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3356 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3357 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3358 int dev_uc_sync(struct net_device *to, struct net_device *from);
3359 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3360 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3361 void dev_uc_flush(struct net_device *dev);
3362 void dev_uc_init(struct net_device *dev);
3363
3364 /**
3365 * __dev_uc_sync - Synchonize device's unicast list
3366 * @dev: device to sync
3367 * @sync: function to call if address should be added
3368 * @unsync: function to call if address should be removed
3369 *
3370 * Add newly added addresses to the interface, and release
3371 * addresses that have been deleted.
3372 **/
3373 static inline int __dev_uc_sync(struct net_device *dev,
3374 int (*sync)(struct net_device *,
3375 const unsigned char *),
3376 int (*unsync)(struct net_device *,
3377 const unsigned char *))
3378 {
3379 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3380 }
3381
3382 /**
3383 * __dev_uc_unsync - Remove synchronized addresses from device
3384 * @dev: device to sync
3385 * @unsync: function to call if address should be removed
3386 *
3387 * Remove all addresses that were added to the device by dev_uc_sync().
3388 **/
3389 static inline void __dev_uc_unsync(struct net_device *dev,
3390 int (*unsync)(struct net_device *,
3391 const unsigned char *))
3392 {
3393 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3394 }
3395
3396 /* Functions used for multicast addresses handling */
3397 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3398 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3399 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3400 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3401 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3402 int dev_mc_sync(struct net_device *to, struct net_device *from);
3403 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3404 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3405 void dev_mc_flush(struct net_device *dev);
3406 void dev_mc_init(struct net_device *dev);
3407
3408 /**
3409 * __dev_mc_sync - Synchonize device's multicast list
3410 * @dev: device to sync
3411 * @sync: function to call if address should be added
3412 * @unsync: function to call if address should be removed
3413 *
3414 * Add newly added addresses to the interface, and release
3415 * addresses that have been deleted.
3416 **/
3417 static inline int __dev_mc_sync(struct net_device *dev,
3418 int (*sync)(struct net_device *,
3419 const unsigned char *),
3420 int (*unsync)(struct net_device *,
3421 const unsigned char *))
3422 {
3423 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3424 }
3425
3426 /**
3427 * __dev_mc_unsync - Remove synchronized addresses from device
3428 * @dev: device to sync
3429 * @unsync: function to call if address should be removed
3430 *
3431 * Remove all addresses that were added to the device by dev_mc_sync().
3432 **/
3433 static inline void __dev_mc_unsync(struct net_device *dev,
3434 int (*unsync)(struct net_device *,
3435 const unsigned char *))
3436 {
3437 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3438 }
3439
3440 /* Functions used for secondary unicast and multicast support */
3441 void dev_set_rx_mode(struct net_device *dev);
3442 void __dev_set_rx_mode(struct net_device *dev);
3443 int dev_set_promiscuity(struct net_device *dev, int inc);
3444 int dev_set_allmulti(struct net_device *dev, int inc);
3445 void netdev_state_change(struct net_device *dev);
3446 void netdev_notify_peers(struct net_device *dev);
3447 void netdev_features_change(struct net_device *dev);
3448 /* Load a device via the kmod */
3449 void dev_load(struct net *net, const char *name);
3450 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3451 struct rtnl_link_stats64 *storage);
3452 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3453 const struct net_device_stats *netdev_stats);
3454
3455 extern int netdev_max_backlog;
3456 extern int netdev_tstamp_prequeue;
3457 extern int weight_p;
3458 extern int bpf_jit_enable;
3459
3460 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3461 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3462 struct list_head **iter);
3463 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3464 struct list_head **iter);
3465
3466 /* iterate through upper list, must be called under RCU read lock */
3467 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3468 for (iter = &(dev)->adj_list.upper, \
3469 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3470 updev; \
3471 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3472
3473 /* iterate through upper list, must be called under RCU read lock */
3474 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3475 for (iter = &(dev)->all_adj_list.upper, \
3476 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3477 updev; \
3478 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3479
3480 void *netdev_lower_get_next_private(struct net_device *dev,
3481 struct list_head **iter);
3482 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3483 struct list_head **iter);
3484
3485 #define netdev_for_each_lower_private(dev, priv, iter) \
3486 for (iter = (dev)->adj_list.lower.next, \
3487 priv = netdev_lower_get_next_private(dev, &(iter)); \
3488 priv; \
3489 priv = netdev_lower_get_next_private(dev, &(iter)))
3490
3491 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3492 for (iter = &(dev)->adj_list.lower, \
3493 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3494 priv; \
3495 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3496
3497 void *netdev_lower_get_next(struct net_device *dev,
3498 struct list_head **iter);
3499 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3500 for (iter = &(dev)->adj_list.lower, \
3501 ldev = netdev_lower_get_next(dev, &(iter)); \
3502 ldev; \
3503 ldev = netdev_lower_get_next(dev, &(iter)))
3504
3505 void *netdev_adjacent_get_private(struct list_head *adj_list);
3506 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3507 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3508 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3509 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3510 int netdev_master_upper_dev_link(struct net_device *dev,
3511 struct net_device *upper_dev);
3512 int netdev_master_upper_dev_link_private(struct net_device *dev,
3513 struct net_device *upper_dev,
3514 void *private);
3515 void netdev_upper_dev_unlink(struct net_device *dev,
3516 struct net_device *upper_dev);
3517 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3518 void *netdev_lower_dev_get_private(struct net_device *dev,
3519 struct net_device *lower_dev);
3520
3521 /* RSS keys are 40 or 52 bytes long */
3522 #define NETDEV_RSS_KEY_LEN 52
3523 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3524 void netdev_rss_key_fill(void *buffer, size_t len);
3525
3526 int dev_get_nest_level(struct net_device *dev,
3527 bool (*type_check)(struct net_device *dev));
3528 int skb_checksum_help(struct sk_buff *skb);
3529 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3530 netdev_features_t features, bool tx_path);
3531 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3532 netdev_features_t features);
3533
3534 struct netdev_bonding_info {
3535 ifslave slave;
3536 ifbond master;
3537 };
3538
3539 struct netdev_notifier_bonding_info {
3540 struct netdev_notifier_info info; /* must be first */
3541 struct netdev_bonding_info bonding_info;
3542 };
3543
3544 void netdev_bonding_info_change(struct net_device *dev,
3545 struct netdev_bonding_info *bonding_info);
3546
3547 static inline
3548 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3549 {
3550 return __skb_gso_segment(skb, features, true);
3551 }
3552 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3553
3554 static inline bool can_checksum_protocol(netdev_features_t features,
3555 __be16 protocol)
3556 {
3557 return ((features & NETIF_F_GEN_CSUM) ||
3558 ((features & NETIF_F_V4_CSUM) &&
3559 protocol == htons(ETH_P_IP)) ||
3560 ((features & NETIF_F_V6_CSUM) &&
3561 protocol == htons(ETH_P_IPV6)) ||
3562 ((features & NETIF_F_FCOE_CRC) &&
3563 protocol == htons(ETH_P_FCOE)));
3564 }
3565
3566 #ifdef CONFIG_BUG
3567 void netdev_rx_csum_fault(struct net_device *dev);
3568 #else
3569 static inline void netdev_rx_csum_fault(struct net_device *dev)
3570 {
3571 }
3572 #endif
3573 /* rx skb timestamps */
3574 void net_enable_timestamp(void);
3575 void net_disable_timestamp(void);
3576
3577 #ifdef CONFIG_PROC_FS
3578 int __init dev_proc_init(void);
3579 #else
3580 #define dev_proc_init() 0
3581 #endif
3582
3583 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3584 struct sk_buff *skb, struct net_device *dev,
3585 bool more)
3586 {
3587 skb->xmit_more = more ? 1 : 0;
3588 return ops->ndo_start_xmit(skb, dev);
3589 }
3590
3591 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3592 struct netdev_queue *txq, bool more)
3593 {
3594 const struct net_device_ops *ops = dev->netdev_ops;
3595 int rc;
3596
3597 rc = __netdev_start_xmit(ops, skb, dev, more);
3598 if (rc == NETDEV_TX_OK)
3599 txq_trans_update(txq);
3600
3601 return rc;
3602 }
3603
3604 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3605 const void *ns);
3606 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3607 const void *ns);
3608
3609 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3610 {
3611 return netdev_class_create_file_ns(class_attr, NULL);
3612 }
3613
3614 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3615 {
3616 netdev_class_remove_file_ns(class_attr, NULL);
3617 }
3618
3619 extern struct kobj_ns_type_operations net_ns_type_operations;
3620
3621 const char *netdev_drivername(const struct net_device *dev);
3622
3623 void linkwatch_run_queue(void);
3624
3625 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3626 netdev_features_t f2)
3627 {
3628 if (f1 & NETIF_F_GEN_CSUM)
3629 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3630 if (f2 & NETIF_F_GEN_CSUM)
3631 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3632 f1 &= f2;
3633 if (f1 & NETIF_F_GEN_CSUM)
3634 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3635
3636 return f1;
3637 }
3638
3639 static inline netdev_features_t netdev_get_wanted_features(
3640 struct net_device *dev)
3641 {
3642 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3643 }
3644 netdev_features_t netdev_increment_features(netdev_features_t all,
3645 netdev_features_t one, netdev_features_t mask);
3646
3647 /* Allow TSO being used on stacked device :
3648 * Performing the GSO segmentation before last device
3649 * is a performance improvement.
3650 */
3651 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3652 netdev_features_t mask)
3653 {
3654 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3655 }
3656
3657 int __netdev_update_features(struct net_device *dev);
3658 void netdev_update_features(struct net_device *dev);
3659 void netdev_change_features(struct net_device *dev);
3660
3661 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3662 struct net_device *dev);
3663
3664 netdev_features_t passthru_features_check(struct sk_buff *skb,
3665 struct net_device *dev,
3666 netdev_features_t features);
3667 netdev_features_t netif_skb_features(struct sk_buff *skb);
3668
3669 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3670 {
3671 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3672
3673 /* check flags correspondence */
3674 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3675 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3676 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3677 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3678 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3679 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3680 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3681 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3682 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3683 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3684 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3685 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3686 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3687
3688 return (features & feature) == feature;
3689 }
3690
3691 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3692 {
3693 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3694 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3695 }
3696
3697 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3698 netdev_features_t features)
3699 {
3700 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3701 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3702 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3703 }
3704
3705 static inline void netif_set_gso_max_size(struct net_device *dev,
3706 unsigned int size)
3707 {
3708 dev->gso_max_size = size;
3709 }
3710
3711 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3712 int pulled_hlen, u16 mac_offset,
3713 int mac_len)
3714 {
3715 skb->protocol = protocol;
3716 skb->encapsulation = 1;
3717 skb_push(skb, pulled_hlen);
3718 skb_reset_transport_header(skb);
3719 skb->mac_header = mac_offset;
3720 skb->network_header = skb->mac_header + mac_len;
3721 skb->mac_len = mac_len;
3722 }
3723
3724 static inline bool netif_is_macvlan(struct net_device *dev)
3725 {
3726 return dev->priv_flags & IFF_MACVLAN;
3727 }
3728
3729 static inline bool netif_is_macvlan_port(struct net_device *dev)
3730 {
3731 return dev->priv_flags & IFF_MACVLAN_PORT;
3732 }
3733
3734 static inline bool netif_is_ipvlan(struct net_device *dev)
3735 {
3736 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3737 }
3738
3739 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3740 {
3741 return dev->priv_flags & IFF_IPVLAN_MASTER;
3742 }
3743
3744 static inline bool netif_is_bond_master(struct net_device *dev)
3745 {
3746 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3747 }
3748
3749 static inline bool netif_is_bond_slave(struct net_device *dev)
3750 {
3751 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3752 }
3753
3754 static inline bool netif_supports_nofcs(struct net_device *dev)
3755 {
3756 return dev->priv_flags & IFF_SUPP_NOFCS;
3757 }
3758
3759 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3760 static inline void netif_keep_dst(struct net_device *dev)
3761 {
3762 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3763 }
3764
3765 extern struct pernet_operations __net_initdata loopback_net_ops;
3766
3767 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3768
3769 /* netdev_printk helpers, similar to dev_printk */
3770
3771 static inline const char *netdev_name(const struct net_device *dev)
3772 {
3773 if (!dev->name[0] || strchr(dev->name, '%'))
3774 return "(unnamed net_device)";
3775 return dev->name;
3776 }
3777
3778 static inline const char *netdev_reg_state(const struct net_device *dev)
3779 {
3780 switch (dev->reg_state) {
3781 case NETREG_UNINITIALIZED: return " (uninitialized)";
3782 case NETREG_REGISTERED: return "";
3783 case NETREG_UNREGISTERING: return " (unregistering)";
3784 case NETREG_UNREGISTERED: return " (unregistered)";
3785 case NETREG_RELEASED: return " (released)";
3786 case NETREG_DUMMY: return " (dummy)";
3787 }
3788
3789 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3790 return " (unknown)";
3791 }
3792
3793 __printf(3, 4)
3794 void netdev_printk(const char *level, const struct net_device *dev,
3795 const char *format, ...);
3796 __printf(2, 3)
3797 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3798 __printf(2, 3)
3799 void netdev_alert(const struct net_device *dev, const char *format, ...);
3800 __printf(2, 3)
3801 void netdev_crit(const struct net_device *dev, const char *format, ...);
3802 __printf(2, 3)
3803 void netdev_err(const struct net_device *dev, const char *format, ...);
3804 __printf(2, 3)
3805 void netdev_warn(const struct net_device *dev, const char *format, ...);
3806 __printf(2, 3)
3807 void netdev_notice(const struct net_device *dev, const char *format, ...);
3808 __printf(2, 3)
3809 void netdev_info(const struct net_device *dev, const char *format, ...);
3810
3811 #define MODULE_ALIAS_NETDEV(device) \
3812 MODULE_ALIAS("netdev-" device)
3813
3814 #if defined(CONFIG_DYNAMIC_DEBUG)
3815 #define netdev_dbg(__dev, format, args...) \
3816 do { \
3817 dynamic_netdev_dbg(__dev, format, ##args); \
3818 } while (0)
3819 #elif defined(DEBUG)
3820 #define netdev_dbg(__dev, format, args...) \
3821 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3822 #else
3823 #define netdev_dbg(__dev, format, args...) \
3824 ({ \
3825 if (0) \
3826 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3827 })
3828 #endif
3829
3830 #if defined(VERBOSE_DEBUG)
3831 #define netdev_vdbg netdev_dbg
3832 #else
3833
3834 #define netdev_vdbg(dev, format, args...) \
3835 ({ \
3836 if (0) \
3837 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3838 0; \
3839 })
3840 #endif
3841
3842 /*
3843 * netdev_WARN() acts like dev_printk(), but with the key difference
3844 * of using a WARN/WARN_ON to get the message out, including the
3845 * file/line information and a backtrace.
3846 */
3847 #define netdev_WARN(dev, format, args...) \
3848 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3849 netdev_reg_state(dev), ##args)
3850
3851 /* netif printk helpers, similar to netdev_printk */
3852
3853 #define netif_printk(priv, type, level, dev, fmt, args...) \
3854 do { \
3855 if (netif_msg_##type(priv)) \
3856 netdev_printk(level, (dev), fmt, ##args); \
3857 } while (0)
3858
3859 #define netif_level(level, priv, type, dev, fmt, args...) \
3860 do { \
3861 if (netif_msg_##type(priv)) \
3862 netdev_##level(dev, fmt, ##args); \
3863 } while (0)
3864
3865 #define netif_emerg(priv, type, dev, fmt, args...) \
3866 netif_level(emerg, priv, type, dev, fmt, ##args)
3867 #define netif_alert(priv, type, dev, fmt, args...) \
3868 netif_level(alert, priv, type, dev, fmt, ##args)
3869 #define netif_crit(priv, type, dev, fmt, args...) \
3870 netif_level(crit, priv, type, dev, fmt, ##args)
3871 #define netif_err(priv, type, dev, fmt, args...) \
3872 netif_level(err, priv, type, dev, fmt, ##args)
3873 #define netif_warn(priv, type, dev, fmt, args...) \
3874 netif_level(warn, priv, type, dev, fmt, ##args)
3875 #define netif_notice(priv, type, dev, fmt, args...) \
3876 netif_level(notice, priv, type, dev, fmt, ##args)
3877 #define netif_info(priv, type, dev, fmt, args...) \
3878 netif_level(info, priv, type, dev, fmt, ##args)
3879
3880 #if defined(CONFIG_DYNAMIC_DEBUG)
3881 #define netif_dbg(priv, type, netdev, format, args...) \
3882 do { \
3883 if (netif_msg_##type(priv)) \
3884 dynamic_netdev_dbg(netdev, format, ##args); \
3885 } while (0)
3886 #elif defined(DEBUG)
3887 #define netif_dbg(priv, type, dev, format, args...) \
3888 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3889 #else
3890 #define netif_dbg(priv, type, dev, format, args...) \
3891 ({ \
3892 if (0) \
3893 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3894 0; \
3895 })
3896 #endif
3897
3898 #if defined(VERBOSE_DEBUG)
3899 #define netif_vdbg netif_dbg
3900 #else
3901 #define netif_vdbg(priv, type, dev, format, args...) \
3902 ({ \
3903 if (0) \
3904 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3905 0; \
3906 })
3907 #endif
3908
3909 /*
3910 * The list of packet types we will receive (as opposed to discard)
3911 * and the routines to invoke.
3912 *
3913 * Why 16. Because with 16 the only overlap we get on a hash of the
3914 * low nibble of the protocol value is RARP/SNAP/X.25.
3915 *
3916 * NOTE: That is no longer true with the addition of VLAN tags. Not
3917 * sure which should go first, but I bet it won't make much
3918 * difference if we are running VLANs. The good news is that
3919 * this protocol won't be in the list unless compiled in, so
3920 * the average user (w/out VLANs) will not be adversely affected.
3921 * --BLG
3922 *
3923 * 0800 IP
3924 * 8100 802.1Q VLAN
3925 * 0001 802.3
3926 * 0002 AX.25
3927 * 0004 802.2
3928 * 8035 RARP
3929 * 0005 SNAP
3930 * 0805 X.25
3931 * 0806 ARP
3932 * 8137 IPX
3933 * 0009 Localtalk
3934 * 86DD IPv6
3935 */
3936 #define PTYPE_HASH_SIZE (16)
3937 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3938
3939 #endif /* _LINUX_NETDEVICE_H */
This page took 0.139023 seconds and 5 git commands to generate.