733d24f16cf8b5146d6f171ccf8e7143e1843344
[deliverable/linux.git] / include / linux / netdevice.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
15 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27
28 #include <linux/pm_qos.h>
29 #include <linux/timer.h>
30 #include <linux/bug.h>
31 #include <linux/delay.h>
32 #include <linux/atomic.h>
33 #include <linux/prefetch.h>
34 #include <asm/cache.h>
35 #include <asm/byteorder.h>
36
37 #include <linux/percpu.h>
38 #include <linux/rculist.h>
39 #include <linux/dmaengine.h>
40 #include <linux/workqueue.h>
41 #include <linux/dynamic_queue_limits.h>
42
43 #include <linux/ethtool.h>
44 #include <net/net_namespace.h>
45 #include <net/dsa.h>
46 #ifdef CONFIG_DCB
47 #include <net/dcbnl.h>
48 #endif
49 #include <net/netprio_cgroup.h>
50
51 #include <linux/netdev_features.h>
52 #include <linux/neighbour.h>
53 #include <uapi/linux/netdevice.h>
54 #include <uapi/linux/if_bonding.h>
55
56 struct netpoll_info;
57 struct device;
58 struct phy_device;
59 /* 802.11 specific */
60 struct wireless_dev;
61 /* 802.15.4 specific */
62 struct wpan_dev;
63
64 void netdev_set_default_ethtool_ops(struct net_device *dev,
65 const struct ethtool_ops *ops);
66
67 /* Backlog congestion levels */
68 #define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
69 #define NET_RX_DROP 1 /* packet dropped */
70
71 /*
72 * Transmit return codes: transmit return codes originate from three different
73 * namespaces:
74 *
75 * - qdisc return codes
76 * - driver transmit return codes
77 * - errno values
78 *
79 * Drivers are allowed to return any one of those in their hard_start_xmit()
80 * function. Real network devices commonly used with qdiscs should only return
81 * the driver transmit return codes though - when qdiscs are used, the actual
82 * transmission happens asynchronously, so the value is not propagated to
83 * higher layers. Virtual network devices transmit synchronously, in this case
84 * the driver transmit return codes are consumed by dev_queue_xmit(), all
85 * others are propagated to higher layers.
86 */
87
88 /* qdisc ->enqueue() return codes. */
89 #define NET_XMIT_SUCCESS 0x00
90 #define NET_XMIT_DROP 0x01 /* skb dropped */
91 #define NET_XMIT_CN 0x02 /* congestion notification */
92 #define NET_XMIT_POLICED 0x03 /* skb is shot by police */
93 #define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
94
95 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
96 * indicates that the device will soon be dropping packets, or already drops
97 * some packets of the same priority; prompting us to send less aggressively. */
98 #define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
99 #define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
100
101 /* Driver transmit return codes */
102 #define NETDEV_TX_MASK 0xf0
103
104 enum netdev_tx {
105 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
106 NETDEV_TX_OK = 0x00, /* driver took care of packet */
107 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
108 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
109 };
110 typedef enum netdev_tx netdev_tx_t;
111
112 /*
113 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115 */
116 static inline bool dev_xmit_complete(int rc)
117 {
118 /*
119 * Positive cases with an skb consumed by a driver:
120 * - successful transmission (rc == NETDEV_TX_OK)
121 * - error while transmitting (rc < 0)
122 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 */
124 if (likely(rc < NET_XMIT_MASK))
125 return true;
126
127 return false;
128 }
129
130 /*
131 * Compute the worst case header length according to the protocols
132 * used.
133 */
134
135 #if defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
136 # if defined(CONFIG_MAC80211_MESH)
137 # define LL_MAX_HEADER 128
138 # else
139 # define LL_MAX_HEADER 96
140 # endif
141 #else
142 # define LL_MAX_HEADER 32
143 #endif
144
145 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
146 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
147 #define MAX_HEADER LL_MAX_HEADER
148 #else
149 #define MAX_HEADER (LL_MAX_HEADER + 48)
150 #endif
151
152 /*
153 * Old network device statistics. Fields are native words
154 * (unsigned long) so they can be read and written atomically.
155 */
156
157 struct net_device_stats {
158 unsigned long rx_packets;
159 unsigned long tx_packets;
160 unsigned long rx_bytes;
161 unsigned long tx_bytes;
162 unsigned long rx_errors;
163 unsigned long tx_errors;
164 unsigned long rx_dropped;
165 unsigned long tx_dropped;
166 unsigned long multicast;
167 unsigned long collisions;
168 unsigned long rx_length_errors;
169 unsigned long rx_over_errors;
170 unsigned long rx_crc_errors;
171 unsigned long rx_frame_errors;
172 unsigned long rx_fifo_errors;
173 unsigned long rx_missed_errors;
174 unsigned long tx_aborted_errors;
175 unsigned long tx_carrier_errors;
176 unsigned long tx_fifo_errors;
177 unsigned long tx_heartbeat_errors;
178 unsigned long tx_window_errors;
179 unsigned long rx_compressed;
180 unsigned long tx_compressed;
181 };
182
183
184 #include <linux/cache.h>
185 #include <linux/skbuff.h>
186
187 #ifdef CONFIG_RPS
188 #include <linux/static_key.h>
189 extern struct static_key rps_needed;
190 #endif
191
192 struct neighbour;
193 struct neigh_parms;
194 struct sk_buff;
195
196 struct netdev_hw_addr {
197 struct list_head list;
198 unsigned char addr[MAX_ADDR_LEN];
199 unsigned char type;
200 #define NETDEV_HW_ADDR_T_LAN 1
201 #define NETDEV_HW_ADDR_T_SAN 2
202 #define NETDEV_HW_ADDR_T_SLAVE 3
203 #define NETDEV_HW_ADDR_T_UNICAST 4
204 #define NETDEV_HW_ADDR_T_MULTICAST 5
205 bool global_use;
206 int sync_cnt;
207 int refcount;
208 int synced;
209 struct rcu_head rcu_head;
210 };
211
212 struct netdev_hw_addr_list {
213 struct list_head list;
214 int count;
215 };
216
217 #define netdev_hw_addr_list_count(l) ((l)->count)
218 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
219 #define netdev_hw_addr_list_for_each(ha, l) \
220 list_for_each_entry(ha, &(l)->list, list)
221
222 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
223 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
224 #define netdev_for_each_uc_addr(ha, dev) \
225 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
226
227 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
228 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
229 #define netdev_for_each_mc_addr(ha, dev) \
230 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
231
232 struct hh_cache {
233 u16 hh_len;
234 u16 __pad;
235 seqlock_t hh_lock;
236
237 /* cached hardware header; allow for machine alignment needs. */
238 #define HH_DATA_MOD 16
239 #define HH_DATA_OFF(__len) \
240 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
241 #define HH_DATA_ALIGN(__len) \
242 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
243 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
244 };
245
246 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
247 * Alternative is:
248 * dev->hard_header_len ? (dev->hard_header_len +
249 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
250 *
251 * We could use other alignment values, but we must maintain the
252 * relationship HH alignment <= LL alignment.
253 */
254 #define LL_RESERVED_SPACE(dev) \
255 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
256 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
257 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
258
259 struct header_ops {
260 int (*create) (struct sk_buff *skb, struct net_device *dev,
261 unsigned short type, const void *daddr,
262 const void *saddr, unsigned int len);
263 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
264 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
265 void (*cache_update)(struct hh_cache *hh,
266 const struct net_device *dev,
267 const unsigned char *haddr);
268 };
269
270 /* These flag bits are private to the generic network queueing
271 * layer, they may not be explicitly referenced by any other
272 * code.
273 */
274
275 enum netdev_state_t {
276 __LINK_STATE_START,
277 __LINK_STATE_PRESENT,
278 __LINK_STATE_NOCARRIER,
279 __LINK_STATE_LINKWATCH_PENDING,
280 __LINK_STATE_DORMANT,
281 };
282
283
284 /*
285 * This structure holds at boot time configured netdevice settings. They
286 * are then used in the device probing.
287 */
288 struct netdev_boot_setup {
289 char name[IFNAMSIZ];
290 struct ifmap map;
291 };
292 #define NETDEV_BOOT_SETUP_MAX 8
293
294 int __init netdev_boot_setup(char *str);
295
296 /*
297 * Structure for NAPI scheduling similar to tasklet but with weighting
298 */
299 struct napi_struct {
300 /* The poll_list must only be managed by the entity which
301 * changes the state of the NAPI_STATE_SCHED bit. This means
302 * whoever atomically sets that bit can add this napi_struct
303 * to the per-cpu poll_list, and whoever clears that bit
304 * can remove from the list right before clearing the bit.
305 */
306 struct list_head poll_list;
307
308 unsigned long state;
309 int weight;
310 unsigned int gro_count;
311 int (*poll)(struct napi_struct *, int);
312 #ifdef CONFIG_NETPOLL
313 spinlock_t poll_lock;
314 int poll_owner;
315 #endif
316 struct net_device *dev;
317 struct sk_buff *gro_list;
318 struct sk_buff *skb;
319 struct hrtimer timer;
320 struct list_head dev_list;
321 struct hlist_node napi_hash_node;
322 unsigned int napi_id;
323 };
324
325 enum {
326 NAPI_STATE_SCHED, /* Poll is scheduled */
327 NAPI_STATE_DISABLE, /* Disable pending */
328 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
329 NAPI_STATE_HASHED, /* In NAPI hash */
330 };
331
332 enum gro_result {
333 GRO_MERGED,
334 GRO_MERGED_FREE,
335 GRO_HELD,
336 GRO_NORMAL,
337 GRO_DROP,
338 };
339 typedef enum gro_result gro_result_t;
340
341 /*
342 * enum rx_handler_result - Possible return values for rx_handlers.
343 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
344 * further.
345 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
346 * case skb->dev was changed by rx_handler.
347 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
348 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
349 *
350 * rx_handlers are functions called from inside __netif_receive_skb(), to do
351 * special processing of the skb, prior to delivery to protocol handlers.
352 *
353 * Currently, a net_device can only have a single rx_handler registered. Trying
354 * to register a second rx_handler will return -EBUSY.
355 *
356 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
357 * To unregister a rx_handler on a net_device, use
358 * netdev_rx_handler_unregister().
359 *
360 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
361 * do with the skb.
362 *
363 * If the rx_handler consumed to skb in some way, it should return
364 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
365 * the skb to be delivered in some other ways.
366 *
367 * If the rx_handler changed skb->dev, to divert the skb to another
368 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
369 * new device will be called if it exists.
370 *
371 * If the rx_handler consider the skb should be ignored, it should return
372 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
373 * are registered on exact device (ptype->dev == skb->dev).
374 *
375 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
376 * delivered, it should return RX_HANDLER_PASS.
377 *
378 * A device without a registered rx_handler will behave as if rx_handler
379 * returned RX_HANDLER_PASS.
380 */
381
382 enum rx_handler_result {
383 RX_HANDLER_CONSUMED,
384 RX_HANDLER_ANOTHER,
385 RX_HANDLER_EXACT,
386 RX_HANDLER_PASS,
387 };
388 typedef enum rx_handler_result rx_handler_result_t;
389 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
390
391 void __napi_schedule(struct napi_struct *n);
392 void __napi_schedule_irqoff(struct napi_struct *n);
393
394 static inline bool napi_disable_pending(struct napi_struct *n)
395 {
396 return test_bit(NAPI_STATE_DISABLE, &n->state);
397 }
398
399 /**
400 * napi_schedule_prep - check if napi can be scheduled
401 * @n: napi context
402 *
403 * Test if NAPI routine is already running, and if not mark
404 * it as running. This is used as a condition variable
405 * insure only one NAPI poll instance runs. We also make
406 * sure there is no pending NAPI disable.
407 */
408 static inline bool napi_schedule_prep(struct napi_struct *n)
409 {
410 return !napi_disable_pending(n) &&
411 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
412 }
413
414 /**
415 * napi_schedule - schedule NAPI poll
416 * @n: napi context
417 *
418 * Schedule NAPI poll routine to be called if it is not already
419 * running.
420 */
421 static inline void napi_schedule(struct napi_struct *n)
422 {
423 if (napi_schedule_prep(n))
424 __napi_schedule(n);
425 }
426
427 /**
428 * napi_schedule_irqoff - schedule NAPI poll
429 * @n: napi context
430 *
431 * Variant of napi_schedule(), assuming hard irqs are masked.
432 */
433 static inline void napi_schedule_irqoff(struct napi_struct *n)
434 {
435 if (napi_schedule_prep(n))
436 __napi_schedule_irqoff(n);
437 }
438
439 /* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
440 static inline bool napi_reschedule(struct napi_struct *napi)
441 {
442 if (napi_schedule_prep(napi)) {
443 __napi_schedule(napi);
444 return true;
445 }
446 return false;
447 }
448
449 void __napi_complete(struct napi_struct *n);
450 void napi_complete_done(struct napi_struct *n, int work_done);
451 /**
452 * napi_complete - NAPI processing complete
453 * @n: napi context
454 *
455 * Mark NAPI processing as complete.
456 * Consider using napi_complete_done() instead.
457 */
458 static inline void napi_complete(struct napi_struct *n)
459 {
460 return napi_complete_done(n, 0);
461 }
462
463 /**
464 * napi_by_id - lookup a NAPI by napi_id
465 * @napi_id: hashed napi_id
466 *
467 * lookup @napi_id in napi_hash table
468 * must be called under rcu_read_lock()
469 */
470 struct napi_struct *napi_by_id(unsigned int napi_id);
471
472 /**
473 * napi_hash_add - add a NAPI to global hashtable
474 * @napi: napi context
475 *
476 * generate a new napi_id and store a @napi under it in napi_hash
477 */
478 void napi_hash_add(struct napi_struct *napi);
479
480 /**
481 * napi_hash_del - remove a NAPI from global table
482 * @napi: napi context
483 *
484 * Warning: caller must observe rcu grace period
485 * before freeing memory containing @napi
486 */
487 void napi_hash_del(struct napi_struct *napi);
488
489 /**
490 * napi_disable - prevent NAPI from scheduling
491 * @n: napi context
492 *
493 * Stop NAPI from being scheduled on this context.
494 * Waits till any outstanding processing completes.
495 */
496 void napi_disable(struct napi_struct *n);
497
498 /**
499 * napi_enable - enable NAPI scheduling
500 * @n: napi context
501 *
502 * Resume NAPI from being scheduled on this context.
503 * Must be paired with napi_disable.
504 */
505 static inline void napi_enable(struct napi_struct *n)
506 {
507 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
508 smp_mb__before_atomic();
509 clear_bit(NAPI_STATE_SCHED, &n->state);
510 }
511
512 #ifdef CONFIG_SMP
513 /**
514 * napi_synchronize - wait until NAPI is not running
515 * @n: napi context
516 *
517 * Wait until NAPI is done being scheduled on this context.
518 * Waits till any outstanding processing completes but
519 * does not disable future activations.
520 */
521 static inline void napi_synchronize(const struct napi_struct *n)
522 {
523 while (test_bit(NAPI_STATE_SCHED, &n->state))
524 msleep(1);
525 }
526 #else
527 # define napi_synchronize(n) barrier()
528 #endif
529
530 enum netdev_queue_state_t {
531 __QUEUE_STATE_DRV_XOFF,
532 __QUEUE_STATE_STACK_XOFF,
533 __QUEUE_STATE_FROZEN,
534 };
535
536 #define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
537 #define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
538 #define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
539
540 #define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
542 QUEUE_STATE_FROZEN)
543 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
544 QUEUE_STATE_FROZEN)
545
546 /*
547 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
548 * netif_tx_* functions below are used to manipulate this flag. The
549 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
550 * queue independently. The netif_xmit_*stopped functions below are called
551 * to check if the queue has been stopped by the driver or stack (either
552 * of the XOFF bits are set in the state). Drivers should not need to call
553 * netif_xmit*stopped functions, they should only be using netif_tx_*.
554 */
555
556 struct netdev_queue {
557 /*
558 * read mostly part
559 */
560 struct net_device *dev;
561 struct Qdisc __rcu *qdisc;
562 struct Qdisc *qdisc_sleeping;
563 #ifdef CONFIG_SYSFS
564 struct kobject kobj;
565 #endif
566 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
567 int numa_node;
568 #endif
569 /*
570 * write mostly part
571 */
572 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
573 int xmit_lock_owner;
574 /*
575 * please use this field instead of dev->trans_start
576 */
577 unsigned long trans_start;
578
579 /*
580 * Number of TX timeouts for this queue
581 * (/sys/class/net/DEV/Q/trans_timeout)
582 */
583 unsigned long trans_timeout;
584
585 unsigned long state;
586
587 #ifdef CONFIG_BQL
588 struct dql dql;
589 #endif
590 unsigned long tx_maxrate;
591 } ____cacheline_aligned_in_smp;
592
593 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
594 {
595 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
596 return q->numa_node;
597 #else
598 return NUMA_NO_NODE;
599 #endif
600 }
601
602 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
603 {
604 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
605 q->numa_node = node;
606 #endif
607 }
608
609 #ifdef CONFIG_RPS
610 /*
611 * This structure holds an RPS map which can be of variable length. The
612 * map is an array of CPUs.
613 */
614 struct rps_map {
615 unsigned int len;
616 struct rcu_head rcu;
617 u16 cpus[0];
618 };
619 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
620
621 /*
622 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
623 * tail pointer for that CPU's input queue at the time of last enqueue, and
624 * a hardware filter index.
625 */
626 struct rps_dev_flow {
627 u16 cpu;
628 u16 filter;
629 unsigned int last_qtail;
630 };
631 #define RPS_NO_FILTER 0xffff
632
633 /*
634 * The rps_dev_flow_table structure contains a table of flow mappings.
635 */
636 struct rps_dev_flow_table {
637 unsigned int mask;
638 struct rcu_head rcu;
639 struct rps_dev_flow flows[0];
640 };
641 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
642 ((_num) * sizeof(struct rps_dev_flow)))
643
644 /*
645 * The rps_sock_flow_table contains mappings of flows to the last CPU
646 * on which they were processed by the application (set in recvmsg).
647 * Each entry is a 32bit value. Upper part is the high order bits
648 * of flow hash, lower part is cpu number.
649 * rps_cpu_mask is used to partition the space, depending on number of
650 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
651 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
652 * meaning we use 32-6=26 bits for the hash.
653 */
654 struct rps_sock_flow_table {
655 u32 mask;
656
657 u32 ents[0] ____cacheline_aligned_in_smp;
658 };
659 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
660
661 #define RPS_NO_CPU 0xffff
662
663 extern u32 rps_cpu_mask;
664 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
665
666 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
667 u32 hash)
668 {
669 if (table && hash) {
670 unsigned int index = hash & table->mask;
671 u32 val = hash & ~rps_cpu_mask;
672
673 /* We only give a hint, preemption can change cpu under us */
674 val |= raw_smp_processor_id();
675
676 if (table->ents[index] != val)
677 table->ents[index] = val;
678 }
679 }
680
681 #ifdef CONFIG_RFS_ACCEL
682 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
683 u16 filter_id);
684 #endif
685 #endif /* CONFIG_RPS */
686
687 /* This structure contains an instance of an RX queue. */
688 struct netdev_rx_queue {
689 #ifdef CONFIG_RPS
690 struct rps_map __rcu *rps_map;
691 struct rps_dev_flow_table __rcu *rps_flow_table;
692 #endif
693 struct kobject kobj;
694 struct net_device *dev;
695 } ____cacheline_aligned_in_smp;
696
697 /*
698 * RX queue sysfs structures and functions.
699 */
700 struct rx_queue_attribute {
701 struct attribute attr;
702 ssize_t (*show)(struct netdev_rx_queue *queue,
703 struct rx_queue_attribute *attr, char *buf);
704 ssize_t (*store)(struct netdev_rx_queue *queue,
705 struct rx_queue_attribute *attr, const char *buf, size_t len);
706 };
707
708 #ifdef CONFIG_XPS
709 /*
710 * This structure holds an XPS map which can be of variable length. The
711 * map is an array of queues.
712 */
713 struct xps_map {
714 unsigned int len;
715 unsigned int alloc_len;
716 struct rcu_head rcu;
717 u16 queues[0];
718 };
719 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
720 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
721 / sizeof(u16))
722
723 /*
724 * This structure holds all XPS maps for device. Maps are indexed by CPU.
725 */
726 struct xps_dev_maps {
727 struct rcu_head rcu;
728 struct xps_map __rcu *cpu_map[0];
729 };
730 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
731 (nr_cpu_ids * sizeof(struct xps_map *)))
732 #endif /* CONFIG_XPS */
733
734 #define TC_MAX_QUEUE 16
735 #define TC_BITMASK 15
736 /* HW offloaded queuing disciplines txq count and offset maps */
737 struct netdev_tc_txq {
738 u16 count;
739 u16 offset;
740 };
741
742 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
743 /*
744 * This structure is to hold information about the device
745 * configured to run FCoE protocol stack.
746 */
747 struct netdev_fcoe_hbainfo {
748 char manufacturer[64];
749 char serial_number[64];
750 char hardware_version[64];
751 char driver_version[64];
752 char optionrom_version[64];
753 char firmware_version[64];
754 char model[256];
755 char model_description[256];
756 };
757 #endif
758
759 #define MAX_PHYS_ITEM_ID_LEN 32
760
761 /* This structure holds a unique identifier to identify some
762 * physical item (port for example) used by a netdevice.
763 */
764 struct netdev_phys_item_id {
765 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
766 unsigned char id_len;
767 };
768
769 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
770 struct sk_buff *skb);
771
772 /*
773 * This structure defines the management hooks for network devices.
774 * The following hooks can be defined; unless noted otherwise, they are
775 * optional and can be filled with a null pointer.
776 *
777 * int (*ndo_init)(struct net_device *dev);
778 * This function is called once when network device is registered.
779 * The network device can use this to any late stage initializaton
780 * or semantic validattion. It can fail with an error code which will
781 * be propogated back to register_netdev
782 *
783 * void (*ndo_uninit)(struct net_device *dev);
784 * This function is called when device is unregistered or when registration
785 * fails. It is not called if init fails.
786 *
787 * int (*ndo_open)(struct net_device *dev);
788 * This function is called when network device transistions to the up
789 * state.
790 *
791 * int (*ndo_stop)(struct net_device *dev);
792 * This function is called when network device transistions to the down
793 * state.
794 *
795 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
796 * struct net_device *dev);
797 * Called when a packet needs to be transmitted.
798 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
799 * the queue before that can happen; it's for obsolete devices and weird
800 * corner cases, but the stack really does a non-trivial amount
801 * of useless work if you return NETDEV_TX_BUSY.
802 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
803 * Required can not be NULL.
804 *
805 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
806 * void *accel_priv, select_queue_fallback_t fallback);
807 * Called to decide which queue to when device supports multiple
808 * transmit queues.
809 *
810 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
811 * This function is called to allow device receiver to make
812 * changes to configuration when multicast or promiscious is enabled.
813 *
814 * void (*ndo_set_rx_mode)(struct net_device *dev);
815 * This function is called device changes address list filtering.
816 * If driver handles unicast address filtering, it should set
817 * IFF_UNICAST_FLT to its priv_flags.
818 *
819 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
820 * This function is called when the Media Access Control address
821 * needs to be changed. If this interface is not defined, the
822 * mac address can not be changed.
823 *
824 * int (*ndo_validate_addr)(struct net_device *dev);
825 * Test if Media Access Control address is valid for the device.
826 *
827 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
828 * Called when a user request an ioctl which can't be handled by
829 * the generic interface code. If not defined ioctl's return
830 * not supported error code.
831 *
832 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
833 * Used to set network devices bus interface parameters. This interface
834 * is retained for legacy reason, new devices should use the bus
835 * interface (PCI) for low level management.
836 *
837 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
838 * Called when a user wants to change the Maximum Transfer Unit
839 * of a device. If not defined, any request to change MTU will
840 * will return an error.
841 *
842 * void (*ndo_tx_timeout)(struct net_device *dev);
843 * Callback uses when the transmitter has not made any progress
844 * for dev->watchdog ticks.
845 *
846 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
847 * struct rtnl_link_stats64 *storage);
848 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
849 * Called when a user wants to get the network device usage
850 * statistics. Drivers must do one of the following:
851 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
852 * rtnl_link_stats64 structure passed by the caller.
853 * 2. Define @ndo_get_stats to update a net_device_stats structure
854 * (which should normally be dev->stats) and return a pointer to
855 * it. The structure may be changed asynchronously only if each
856 * field is written atomically.
857 * 3. Update dev->stats asynchronously and atomically, and define
858 * neither operation.
859 *
860 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
861 * If device support VLAN filtering this function is called when a
862 * VLAN id is registered.
863 *
864 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
865 * If device support VLAN filtering this function is called when a
866 * VLAN id is unregistered.
867 *
868 * void (*ndo_poll_controller)(struct net_device *dev);
869 *
870 * SR-IOV management functions.
871 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
872 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
873 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
874 * int max_tx_rate);
875 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
876 * int (*ndo_get_vf_config)(struct net_device *dev,
877 * int vf, struct ifla_vf_info *ivf);
878 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
879 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
880 * struct nlattr *port[]);
881 *
882 * Enable or disable the VF ability to query its RSS Redirection Table and
883 * Hash Key. This is needed since on some devices VF share this information
884 * with PF and querying it may adduce a theoretical security risk.
885 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
886 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
887 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
888 * Called to setup 'tc' number of traffic classes in the net device. This
889 * is always called from the stack with the rtnl lock held and netif tx
890 * queues stopped. This allows the netdevice to perform queue management
891 * safely.
892 *
893 * Fiber Channel over Ethernet (FCoE) offload functions.
894 * int (*ndo_fcoe_enable)(struct net_device *dev);
895 * Called when the FCoE protocol stack wants to start using LLD for FCoE
896 * so the underlying device can perform whatever needed configuration or
897 * initialization to support acceleration of FCoE traffic.
898 *
899 * int (*ndo_fcoe_disable)(struct net_device *dev);
900 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
901 * so the underlying device can perform whatever needed clean-ups to
902 * stop supporting acceleration of FCoE traffic.
903 *
904 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
905 * struct scatterlist *sgl, unsigned int sgc);
906 * Called when the FCoE Initiator wants to initialize an I/O that
907 * is a possible candidate for Direct Data Placement (DDP). The LLD can
908 * perform necessary setup and returns 1 to indicate the device is set up
909 * successfully to perform DDP on this I/O, otherwise this returns 0.
910 *
911 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
912 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
913 * indicated by the FC exchange id 'xid', so the underlying device can
914 * clean up and reuse resources for later DDP requests.
915 *
916 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
917 * struct scatterlist *sgl, unsigned int sgc);
918 * Called when the FCoE Target wants to initialize an I/O that
919 * is a possible candidate for Direct Data Placement (DDP). The LLD can
920 * perform necessary setup and returns 1 to indicate the device is set up
921 * successfully to perform DDP on this I/O, otherwise this returns 0.
922 *
923 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
924 * struct netdev_fcoe_hbainfo *hbainfo);
925 * Called when the FCoE Protocol stack wants information on the underlying
926 * device. This information is utilized by the FCoE protocol stack to
927 * register attributes with Fiber Channel management service as per the
928 * FC-GS Fabric Device Management Information(FDMI) specification.
929 *
930 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
931 * Called when the underlying device wants to override default World Wide
932 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
933 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
934 * protocol stack to use.
935 *
936 * RFS acceleration.
937 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
938 * u16 rxq_index, u32 flow_id);
939 * Set hardware filter for RFS. rxq_index is the target queue index;
940 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
941 * Return the filter ID on success, or a negative error code.
942 *
943 * Slave management functions (for bridge, bonding, etc).
944 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
945 * Called to make another netdev an underling.
946 *
947 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
948 * Called to release previously enslaved netdev.
949 *
950 * Feature/offload setting functions.
951 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
952 * netdev_features_t features);
953 * Adjusts the requested feature flags according to device-specific
954 * constraints, and returns the resulting flags. Must not modify
955 * the device state.
956 *
957 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
958 * Called to update device configuration to new features. Passed
959 * feature set might be less than what was returned by ndo_fix_features()).
960 * Must return >0 or -errno if it changed dev->features itself.
961 *
962 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
963 * struct net_device *dev,
964 * const unsigned char *addr, u16 vid, u16 flags)
965 * Adds an FDB entry to dev for addr.
966 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
967 * struct net_device *dev,
968 * const unsigned char *addr, u16 vid)
969 * Deletes the FDB entry from dev coresponding to addr.
970 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
971 * struct net_device *dev, struct net_device *filter_dev,
972 * int idx)
973 * Used to add FDB entries to dump requests. Implementers should add
974 * entries to skb and update idx with the number of entries.
975 *
976 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
977 * u16 flags)
978 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
979 * struct net_device *dev, u32 filter_mask)
980 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
981 * u16 flags);
982 *
983 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
984 * Called to change device carrier. Soft-devices (like dummy, team, etc)
985 * which do not represent real hardware may define this to allow their
986 * userspace components to manage their virtual carrier state. Devices
987 * that determine carrier state from physical hardware properties (eg
988 * network cables) or protocol-dependent mechanisms (eg
989 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
990 *
991 * int (*ndo_get_phys_port_id)(struct net_device *dev,
992 * struct netdev_phys_item_id *ppid);
993 * Called to get ID of physical port of this device. If driver does
994 * not implement this, it is assumed that the hw is not able to have
995 * multiple net devices on single physical port.
996 *
997 * void (*ndo_add_vxlan_port)(struct net_device *dev,
998 * sa_family_t sa_family, __be16 port);
999 * Called by vxlan to notiy a driver about the UDP port and socket
1000 * address family that vxlan is listnening to. It is called only when
1001 * a new port starts listening. The operation is protected by the
1002 * vxlan_net->sock_lock.
1003 *
1004 * void (*ndo_del_vxlan_port)(struct net_device *dev,
1005 * sa_family_t sa_family, __be16 port);
1006 * Called by vxlan to notify the driver about a UDP port and socket
1007 * address family that vxlan is not listening to anymore. The operation
1008 * is protected by the vxlan_net->sock_lock.
1009 *
1010 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1011 * struct net_device *dev)
1012 * Called by upper layer devices to accelerate switching or other
1013 * station functionality into hardware. 'pdev is the lowerdev
1014 * to use for the offload and 'dev' is the net device that will
1015 * back the offload. Returns a pointer to the private structure
1016 * the upper layer will maintain.
1017 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1018 * Called by upper layer device to delete the station created
1019 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1020 * the station and priv is the structure returned by the add
1021 * operation.
1022 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1023 * struct net_device *dev,
1024 * void *priv);
1025 * Callback to use for xmit over the accelerated station. This
1026 * is used in place of ndo_start_xmit on accelerated net
1027 * devices.
1028 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1029 * struct net_device *dev
1030 * netdev_features_t features);
1031 * Called by core transmit path to determine if device is capable of
1032 * performing offload operations on a given packet. This is to give
1033 * the device an opportunity to implement any restrictions that cannot
1034 * be otherwise expressed by feature flags. The check is called with
1035 * the set of features that the stack has calculated and it returns
1036 * those the driver believes to be appropriate.
1037 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1038 * int queue_index, u32 maxrate);
1039 * Called when a user wants to set a max-rate limitation of specific
1040 * TX queue.
1041 * int (*ndo_get_iflink)(const struct net_device *dev);
1042 * Called to get the iflink value of this device.
1043 */
1044 struct net_device_ops {
1045 int (*ndo_init)(struct net_device *dev);
1046 void (*ndo_uninit)(struct net_device *dev);
1047 int (*ndo_open)(struct net_device *dev);
1048 int (*ndo_stop)(struct net_device *dev);
1049 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1050 struct net_device *dev);
1051 u16 (*ndo_select_queue)(struct net_device *dev,
1052 struct sk_buff *skb,
1053 void *accel_priv,
1054 select_queue_fallback_t fallback);
1055 void (*ndo_change_rx_flags)(struct net_device *dev,
1056 int flags);
1057 void (*ndo_set_rx_mode)(struct net_device *dev);
1058 int (*ndo_set_mac_address)(struct net_device *dev,
1059 void *addr);
1060 int (*ndo_validate_addr)(struct net_device *dev);
1061 int (*ndo_do_ioctl)(struct net_device *dev,
1062 struct ifreq *ifr, int cmd);
1063 int (*ndo_set_config)(struct net_device *dev,
1064 struct ifmap *map);
1065 int (*ndo_change_mtu)(struct net_device *dev,
1066 int new_mtu);
1067 int (*ndo_neigh_setup)(struct net_device *dev,
1068 struct neigh_parms *);
1069 void (*ndo_tx_timeout) (struct net_device *dev);
1070
1071 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1072 struct rtnl_link_stats64 *storage);
1073 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1074
1075 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1076 __be16 proto, u16 vid);
1077 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1078 __be16 proto, u16 vid);
1079 #ifdef CONFIG_NET_POLL_CONTROLLER
1080 void (*ndo_poll_controller)(struct net_device *dev);
1081 int (*ndo_netpoll_setup)(struct net_device *dev,
1082 struct netpoll_info *info);
1083 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1084 #endif
1085 #ifdef CONFIG_NET_RX_BUSY_POLL
1086 int (*ndo_busy_poll)(struct napi_struct *dev);
1087 #endif
1088 int (*ndo_set_vf_mac)(struct net_device *dev,
1089 int queue, u8 *mac);
1090 int (*ndo_set_vf_vlan)(struct net_device *dev,
1091 int queue, u16 vlan, u8 qos);
1092 int (*ndo_set_vf_rate)(struct net_device *dev,
1093 int vf, int min_tx_rate,
1094 int max_tx_rate);
1095 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1096 int vf, bool setting);
1097 int (*ndo_get_vf_config)(struct net_device *dev,
1098 int vf,
1099 struct ifla_vf_info *ivf);
1100 int (*ndo_set_vf_link_state)(struct net_device *dev,
1101 int vf, int link_state);
1102 int (*ndo_set_vf_port)(struct net_device *dev,
1103 int vf,
1104 struct nlattr *port[]);
1105 int (*ndo_get_vf_port)(struct net_device *dev,
1106 int vf, struct sk_buff *skb);
1107 int (*ndo_set_vf_rss_query_en)(
1108 struct net_device *dev,
1109 int vf, bool setting);
1110 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1111 #if IS_ENABLED(CONFIG_FCOE)
1112 int (*ndo_fcoe_enable)(struct net_device *dev);
1113 int (*ndo_fcoe_disable)(struct net_device *dev);
1114 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1115 u16 xid,
1116 struct scatterlist *sgl,
1117 unsigned int sgc);
1118 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1119 u16 xid);
1120 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1121 u16 xid,
1122 struct scatterlist *sgl,
1123 unsigned int sgc);
1124 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1125 struct netdev_fcoe_hbainfo *hbainfo);
1126 #endif
1127
1128 #if IS_ENABLED(CONFIG_LIBFCOE)
1129 #define NETDEV_FCOE_WWNN 0
1130 #define NETDEV_FCOE_WWPN 1
1131 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1132 u64 *wwn, int type);
1133 #endif
1134
1135 #ifdef CONFIG_RFS_ACCEL
1136 int (*ndo_rx_flow_steer)(struct net_device *dev,
1137 const struct sk_buff *skb,
1138 u16 rxq_index,
1139 u32 flow_id);
1140 #endif
1141 int (*ndo_add_slave)(struct net_device *dev,
1142 struct net_device *slave_dev);
1143 int (*ndo_del_slave)(struct net_device *dev,
1144 struct net_device *slave_dev);
1145 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1146 netdev_features_t features);
1147 int (*ndo_set_features)(struct net_device *dev,
1148 netdev_features_t features);
1149 int (*ndo_neigh_construct)(struct neighbour *n);
1150 void (*ndo_neigh_destroy)(struct neighbour *n);
1151
1152 int (*ndo_fdb_add)(struct ndmsg *ndm,
1153 struct nlattr *tb[],
1154 struct net_device *dev,
1155 const unsigned char *addr,
1156 u16 vid,
1157 u16 flags);
1158 int (*ndo_fdb_del)(struct ndmsg *ndm,
1159 struct nlattr *tb[],
1160 struct net_device *dev,
1161 const unsigned char *addr,
1162 u16 vid);
1163 int (*ndo_fdb_dump)(struct sk_buff *skb,
1164 struct netlink_callback *cb,
1165 struct net_device *dev,
1166 struct net_device *filter_dev,
1167 int idx);
1168
1169 int (*ndo_bridge_setlink)(struct net_device *dev,
1170 struct nlmsghdr *nlh,
1171 u16 flags);
1172 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1173 u32 pid, u32 seq,
1174 struct net_device *dev,
1175 u32 filter_mask);
1176 int (*ndo_bridge_dellink)(struct net_device *dev,
1177 struct nlmsghdr *nlh,
1178 u16 flags);
1179 int (*ndo_change_carrier)(struct net_device *dev,
1180 bool new_carrier);
1181 int (*ndo_get_phys_port_id)(struct net_device *dev,
1182 struct netdev_phys_item_id *ppid);
1183 int (*ndo_get_phys_port_name)(struct net_device *dev,
1184 char *name, size_t len);
1185 void (*ndo_add_vxlan_port)(struct net_device *dev,
1186 sa_family_t sa_family,
1187 __be16 port);
1188 void (*ndo_del_vxlan_port)(struct net_device *dev,
1189 sa_family_t sa_family,
1190 __be16 port);
1191
1192 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1193 struct net_device *dev);
1194 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1195 void *priv);
1196
1197 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1198 struct net_device *dev,
1199 void *priv);
1200 int (*ndo_get_lock_subclass)(struct net_device *dev);
1201 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1202 struct net_device *dev,
1203 netdev_features_t features);
1204 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1205 int queue_index,
1206 u32 maxrate);
1207 int (*ndo_get_iflink)(const struct net_device *dev);
1208 };
1209
1210 /**
1211 * enum net_device_priv_flags - &struct net_device priv_flags
1212 *
1213 * These are the &struct net_device, they are only set internally
1214 * by drivers and used in the kernel. These flags are invisible to
1215 * userspace, this means that the order of these flags can change
1216 * during any kernel release.
1217 *
1218 * You should have a pretty good reason to be extending these flags.
1219 *
1220 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1221 * @IFF_EBRIDGE: Ethernet bridging device
1222 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1223 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1224 * @IFF_MASTER_ALB: bonding master, balance-alb
1225 * @IFF_BONDING: bonding master or slave
1226 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1227 * @IFF_ISATAP: ISATAP interface (RFC4214)
1228 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1229 * @IFF_WAN_HDLC: WAN HDLC device
1230 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1231 * release skb->dst
1232 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1233 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1234 * @IFF_MACVLAN_PORT: device used as macvlan port
1235 * @IFF_BRIDGE_PORT: device used as bridge port
1236 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1237 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1238 * @IFF_UNICAST_FLT: Supports unicast filtering
1239 * @IFF_TEAM_PORT: device used as team port
1240 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1241 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1242 * change when it's running
1243 * @IFF_MACVLAN: Macvlan device
1244 */
1245 enum netdev_priv_flags {
1246 IFF_802_1Q_VLAN = 1<<0,
1247 IFF_EBRIDGE = 1<<1,
1248 IFF_SLAVE_INACTIVE = 1<<2,
1249 IFF_MASTER_8023AD = 1<<3,
1250 IFF_MASTER_ALB = 1<<4,
1251 IFF_BONDING = 1<<5,
1252 IFF_SLAVE_NEEDARP = 1<<6,
1253 IFF_ISATAP = 1<<7,
1254 IFF_MASTER_ARPMON = 1<<8,
1255 IFF_WAN_HDLC = 1<<9,
1256 IFF_XMIT_DST_RELEASE = 1<<10,
1257 IFF_DONT_BRIDGE = 1<<11,
1258 IFF_DISABLE_NETPOLL = 1<<12,
1259 IFF_MACVLAN_PORT = 1<<13,
1260 IFF_BRIDGE_PORT = 1<<14,
1261 IFF_OVS_DATAPATH = 1<<15,
1262 IFF_TX_SKB_SHARING = 1<<16,
1263 IFF_UNICAST_FLT = 1<<17,
1264 IFF_TEAM_PORT = 1<<18,
1265 IFF_SUPP_NOFCS = 1<<19,
1266 IFF_LIVE_ADDR_CHANGE = 1<<20,
1267 IFF_MACVLAN = 1<<21,
1268 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1269 IFF_IPVLAN_MASTER = 1<<23,
1270 IFF_IPVLAN_SLAVE = 1<<24,
1271 };
1272
1273 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1274 #define IFF_EBRIDGE IFF_EBRIDGE
1275 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1276 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1277 #define IFF_MASTER_ALB IFF_MASTER_ALB
1278 #define IFF_BONDING IFF_BONDING
1279 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1280 #define IFF_ISATAP IFF_ISATAP
1281 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1282 #define IFF_WAN_HDLC IFF_WAN_HDLC
1283 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1284 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1285 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1286 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1287 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1288 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1289 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1290 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1291 #define IFF_TEAM_PORT IFF_TEAM_PORT
1292 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1293 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1294 #define IFF_MACVLAN IFF_MACVLAN
1295 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1296 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1297 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1298
1299 /**
1300 * struct net_device - The DEVICE structure.
1301 * Actually, this whole structure is a big mistake. It mixes I/O
1302 * data with strictly "high-level" data, and it has to know about
1303 * almost every data structure used in the INET module.
1304 *
1305 * @name: This is the first field of the "visible" part of this structure
1306 * (i.e. as seen by users in the "Space.c" file). It is the name
1307 * of the interface.
1308 *
1309 * @name_hlist: Device name hash chain, please keep it close to name[]
1310 * @ifalias: SNMP alias
1311 * @mem_end: Shared memory end
1312 * @mem_start: Shared memory start
1313 * @base_addr: Device I/O address
1314 * @irq: Device IRQ number
1315 *
1316 * @state: Generic network queuing layer state, see netdev_state_t
1317 * @dev_list: The global list of network devices
1318 * @napi_list: List entry, that is used for polling napi devices
1319 * @unreg_list: List entry, that is used, when we are unregistering the
1320 * device, see the function unregister_netdev
1321 * @close_list: List entry, that is used, when we are closing the device
1322 *
1323 * @adj_list: Directly linked devices, like slaves for bonding
1324 * @all_adj_list: All linked devices, *including* neighbours
1325 * @features: Currently active device features
1326 * @hw_features: User-changeable features
1327 *
1328 * @wanted_features: User-requested features
1329 * @vlan_features: Mask of features inheritable by VLAN devices
1330 *
1331 * @hw_enc_features: Mask of features inherited by encapsulating devices
1332 * This field indicates what encapsulation
1333 * offloads the hardware is capable of doing,
1334 * and drivers will need to set them appropriately.
1335 *
1336 * @mpls_features: Mask of features inheritable by MPLS
1337 *
1338 * @ifindex: interface index
1339 * @group: The group, that the device belongs to
1340 *
1341 * @stats: Statistics struct, which was left as a legacy, use
1342 * rtnl_link_stats64 instead
1343 *
1344 * @rx_dropped: Dropped packets by core network,
1345 * do not use this in drivers
1346 * @tx_dropped: Dropped packets by core network,
1347 * do not use this in drivers
1348 *
1349 * @carrier_changes: Stats to monitor carrier on<->off transitions
1350 *
1351 * @wireless_handlers: List of functions to handle Wireless Extensions,
1352 * instead of ioctl,
1353 * see <net/iw_handler.h> for details.
1354 * @wireless_data: Instance data managed by the core of wireless extensions
1355 *
1356 * @netdev_ops: Includes several pointers to callbacks,
1357 * if one wants to override the ndo_*() functions
1358 * @ethtool_ops: Management operations
1359 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1360 * of Layer 2 headers.
1361 *
1362 * @flags: Interface flags (a la BSD)
1363 * @priv_flags: Like 'flags' but invisible to userspace,
1364 * see if.h for the definitions
1365 * @gflags: Global flags ( kept as legacy )
1366 * @padded: How much padding added by alloc_netdev()
1367 * @operstate: RFC2863 operstate
1368 * @link_mode: Mapping policy to operstate
1369 * @if_port: Selectable AUI, TP, ...
1370 * @dma: DMA channel
1371 * @mtu: Interface MTU value
1372 * @type: Interface hardware type
1373 * @hard_header_len: Hardware header length
1374 *
1375 * @needed_headroom: Extra headroom the hardware may need, but not in all
1376 * cases can this be guaranteed
1377 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1378 * cases can this be guaranteed. Some cases also use
1379 * LL_MAX_HEADER instead to allocate the skb
1380 *
1381 * interface address info:
1382 *
1383 * @perm_addr: Permanent hw address
1384 * @addr_assign_type: Hw address assignment type
1385 * @addr_len: Hardware address length
1386 * @neigh_priv_len; Used in neigh_alloc(),
1387 * initialized only in atm/clip.c
1388 * @dev_id: Used to differentiate devices that share
1389 * the same link layer address
1390 * @dev_port: Used to differentiate devices that share
1391 * the same function
1392 * @addr_list_lock: XXX: need comments on this one
1393 * @uc: unicast mac addresses
1394 * @mc: multicast mac addresses
1395 * @dev_addrs: list of device hw addresses
1396 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1397 * @uc_promisc: Counter, that indicates, that promiscuous mode
1398 * has been enabled due to the need to listen to
1399 * additional unicast addresses in a device that
1400 * does not implement ndo_set_rx_mode()
1401 * @promiscuity: Number of times, the NIC is told to work in
1402 * Promiscuous mode, if it becomes 0 the NIC will
1403 * exit from working in Promiscuous mode
1404 * @allmulti: Counter, enables or disables allmulticast mode
1405 *
1406 * @vlan_info: VLAN info
1407 * @dsa_ptr: dsa specific data
1408 * @tipc_ptr: TIPC specific data
1409 * @atalk_ptr: AppleTalk link
1410 * @ip_ptr: IPv4 specific data
1411 * @dn_ptr: DECnet specific data
1412 * @ip6_ptr: IPv6 specific data
1413 * @ax25_ptr: AX.25 specific data
1414 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1415 *
1416 * @last_rx: Time of last Rx
1417 * @dev_addr: Hw address (before bcast,
1418 * because most packets are unicast)
1419 *
1420 * @_rx: Array of RX queues
1421 * @num_rx_queues: Number of RX queues
1422 * allocated at register_netdev() time
1423 * @real_num_rx_queues: Number of RX queues currently active in device
1424 *
1425 * @rx_handler: handler for received packets
1426 * @rx_handler_data: XXX: need comments on this one
1427 * @ingress_queue: XXX: need comments on this one
1428 * @broadcast: hw bcast address
1429 *
1430 * @_tx: Array of TX queues
1431 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1432 * @real_num_tx_queues: Number of TX queues currently active in device
1433 * @qdisc: Root qdisc from userspace point of view
1434 * @tx_queue_len: Max frames per queue allowed
1435 * @tx_global_lock: XXX: need comments on this one
1436 *
1437 * @xps_maps: XXX: need comments on this one
1438 *
1439 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1440 * indexed by RX queue number. Assigned by driver.
1441 * This must only be set if the ndo_rx_flow_steer
1442 * operation is defined
1443 *
1444 * @trans_start: Time (in jiffies) of last Tx
1445 * @watchdog_timeo: Represents the timeout that is used by
1446 * the watchdog ( see dev_watchdog() )
1447 * @watchdog_timer: List of timers
1448 *
1449 * @pcpu_refcnt: Number of references to this device
1450 * @todo_list: Delayed register/unregister
1451 * @index_hlist: Device index hash chain
1452 * @link_watch_list: XXX: need comments on this one
1453 *
1454 * @reg_state: Register/unregister state machine
1455 * @dismantle: Device is going to be freed
1456 * @rtnl_link_state: This enum represents the phases of creating
1457 * a new link
1458 *
1459 * @destructor: Called from unregister,
1460 * can be used to call free_netdev
1461 * @npinfo: XXX: need comments on this one
1462 * @nd_net: Network namespace this network device is inside
1463 *
1464 * @ml_priv: Mid-layer private
1465 * @lstats: Loopback statistics
1466 * @tstats: Tunnel statistics
1467 * @dstats: Dummy statistics
1468 * @vstats: Virtual ethernet statistics
1469 *
1470 * @garp_port: GARP
1471 * @mrp_port: MRP
1472 *
1473 * @dev: Class/net/name entry
1474 * @sysfs_groups: Space for optional device, statistics and wireless
1475 * sysfs groups
1476 *
1477 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1478 * @rtnl_link_ops: Rtnl_link_ops
1479 *
1480 * @gso_max_size: Maximum size of generic segmentation offload
1481 * @gso_max_segs: Maximum number of segments that can be passed to the
1482 * NIC for GSO
1483 * @gso_min_segs: Minimum number of segments that can be passed to the
1484 * NIC for GSO
1485 *
1486 * @dcbnl_ops: Data Center Bridging netlink ops
1487 * @num_tc: Number of traffic classes in the net device
1488 * @tc_to_txq: XXX: need comments on this one
1489 * @prio_tc_map XXX: need comments on this one
1490 *
1491 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1492 *
1493 * @priomap: XXX: need comments on this one
1494 * @phydev: Physical device may attach itself
1495 * for hardware timestamping
1496 *
1497 * @qdisc_tx_busylock: XXX: need comments on this one
1498 *
1499 * @pm_qos_req: Power Management QoS object
1500 *
1501 * FIXME: cleanup struct net_device such that network protocol info
1502 * moves out.
1503 */
1504
1505 struct net_device {
1506 char name[IFNAMSIZ];
1507 struct hlist_node name_hlist;
1508 char *ifalias;
1509 /*
1510 * I/O specific fields
1511 * FIXME: Merge these and struct ifmap into one
1512 */
1513 unsigned long mem_end;
1514 unsigned long mem_start;
1515 unsigned long base_addr;
1516 int irq;
1517
1518 /*
1519 * Some hardware also needs these fields (state,dev_list,
1520 * napi_list,unreg_list,close_list) but they are not
1521 * part of the usual set specified in Space.c.
1522 */
1523
1524 unsigned long state;
1525
1526 struct list_head dev_list;
1527 struct list_head napi_list;
1528 struct list_head unreg_list;
1529 struct list_head close_list;
1530 struct list_head ptype_all;
1531 struct list_head ptype_specific;
1532
1533 struct {
1534 struct list_head upper;
1535 struct list_head lower;
1536 } adj_list;
1537
1538 struct {
1539 struct list_head upper;
1540 struct list_head lower;
1541 } all_adj_list;
1542
1543 netdev_features_t features;
1544 netdev_features_t hw_features;
1545 netdev_features_t wanted_features;
1546 netdev_features_t vlan_features;
1547 netdev_features_t hw_enc_features;
1548 netdev_features_t mpls_features;
1549
1550 int ifindex;
1551 int group;
1552
1553 struct net_device_stats stats;
1554
1555 atomic_long_t rx_dropped;
1556 atomic_long_t tx_dropped;
1557
1558 atomic_t carrier_changes;
1559
1560 #ifdef CONFIG_WIRELESS_EXT
1561 const struct iw_handler_def * wireless_handlers;
1562 struct iw_public_data * wireless_data;
1563 #endif
1564 const struct net_device_ops *netdev_ops;
1565 const struct ethtool_ops *ethtool_ops;
1566 #ifdef CONFIG_NET_SWITCHDEV
1567 const struct swdev_ops *swdev_ops;
1568 #endif
1569
1570 const struct header_ops *header_ops;
1571
1572 unsigned int flags;
1573 unsigned int priv_flags;
1574
1575 unsigned short gflags;
1576 unsigned short padded;
1577
1578 unsigned char operstate;
1579 unsigned char link_mode;
1580
1581 unsigned char if_port;
1582 unsigned char dma;
1583
1584 unsigned int mtu;
1585 unsigned short type;
1586 unsigned short hard_header_len;
1587
1588 unsigned short needed_headroom;
1589 unsigned short needed_tailroom;
1590
1591 /* Interface address info. */
1592 unsigned char perm_addr[MAX_ADDR_LEN];
1593 unsigned char addr_assign_type;
1594 unsigned char addr_len;
1595 unsigned short neigh_priv_len;
1596 unsigned short dev_id;
1597 unsigned short dev_port;
1598 spinlock_t addr_list_lock;
1599 struct netdev_hw_addr_list uc;
1600 struct netdev_hw_addr_list mc;
1601 struct netdev_hw_addr_list dev_addrs;
1602
1603 #ifdef CONFIG_SYSFS
1604 struct kset *queues_kset;
1605 #endif
1606
1607 unsigned char name_assign_type;
1608
1609 bool uc_promisc;
1610 unsigned int promiscuity;
1611 unsigned int allmulti;
1612
1613
1614 /* Protocol specific pointers */
1615
1616 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1617 struct vlan_info __rcu *vlan_info;
1618 #endif
1619 #if IS_ENABLED(CONFIG_NET_DSA)
1620 struct dsa_switch_tree *dsa_ptr;
1621 #endif
1622 #if IS_ENABLED(CONFIG_TIPC)
1623 struct tipc_bearer __rcu *tipc_ptr;
1624 #endif
1625 void *atalk_ptr;
1626 struct in_device __rcu *ip_ptr;
1627 struct dn_dev __rcu *dn_ptr;
1628 struct inet6_dev __rcu *ip6_ptr;
1629 void *ax25_ptr;
1630 struct wireless_dev *ieee80211_ptr;
1631 struct wpan_dev *ieee802154_ptr;
1632
1633 /*
1634 * Cache lines mostly used on receive path (including eth_type_trans())
1635 */
1636 unsigned long last_rx;
1637
1638 /* Interface address info used in eth_type_trans() */
1639 unsigned char *dev_addr;
1640
1641
1642 #ifdef CONFIG_SYSFS
1643 struct netdev_rx_queue *_rx;
1644
1645 unsigned int num_rx_queues;
1646 unsigned int real_num_rx_queues;
1647
1648 #endif
1649
1650 unsigned long gro_flush_timeout;
1651 rx_handler_func_t __rcu *rx_handler;
1652 void __rcu *rx_handler_data;
1653
1654 struct netdev_queue __rcu *ingress_queue;
1655 unsigned char broadcast[MAX_ADDR_LEN];
1656
1657
1658 /*
1659 * Cache lines mostly used on transmit path
1660 */
1661 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1662 unsigned int num_tx_queues;
1663 unsigned int real_num_tx_queues;
1664 struct Qdisc *qdisc;
1665 unsigned long tx_queue_len;
1666 spinlock_t tx_global_lock;
1667
1668 #ifdef CONFIG_XPS
1669 struct xps_dev_maps __rcu *xps_maps;
1670 #endif
1671 #ifdef CONFIG_RFS_ACCEL
1672 struct cpu_rmap *rx_cpu_rmap;
1673 #endif
1674
1675 /* These may be needed for future network-power-down code. */
1676
1677 /*
1678 * trans_start here is expensive for high speed devices on SMP,
1679 * please use netdev_queue->trans_start instead.
1680 */
1681 unsigned long trans_start;
1682
1683 int watchdog_timeo;
1684 struct timer_list watchdog_timer;
1685
1686 int __percpu *pcpu_refcnt;
1687 struct list_head todo_list;
1688
1689 struct hlist_node index_hlist;
1690 struct list_head link_watch_list;
1691
1692 enum { NETREG_UNINITIALIZED=0,
1693 NETREG_REGISTERED, /* completed register_netdevice */
1694 NETREG_UNREGISTERING, /* called unregister_netdevice */
1695 NETREG_UNREGISTERED, /* completed unregister todo */
1696 NETREG_RELEASED, /* called free_netdev */
1697 NETREG_DUMMY, /* dummy device for NAPI poll */
1698 } reg_state:8;
1699
1700 bool dismantle;
1701
1702 enum {
1703 RTNL_LINK_INITIALIZED,
1704 RTNL_LINK_INITIALIZING,
1705 } rtnl_link_state:16;
1706
1707 void (*destructor)(struct net_device *dev);
1708
1709 #ifdef CONFIG_NETPOLL
1710 struct netpoll_info __rcu *npinfo;
1711 #endif
1712
1713 possible_net_t nd_net;
1714
1715 /* mid-layer private */
1716 union {
1717 void *ml_priv;
1718 struct pcpu_lstats __percpu *lstats;
1719 struct pcpu_sw_netstats __percpu *tstats;
1720 struct pcpu_dstats __percpu *dstats;
1721 struct pcpu_vstats __percpu *vstats;
1722 };
1723
1724 struct garp_port __rcu *garp_port;
1725 struct mrp_port __rcu *mrp_port;
1726
1727 struct device dev;
1728 const struct attribute_group *sysfs_groups[4];
1729 const struct attribute_group *sysfs_rx_queue_group;
1730
1731 const struct rtnl_link_ops *rtnl_link_ops;
1732
1733 /* for setting kernel sock attribute on TCP connection setup */
1734 #define GSO_MAX_SIZE 65536
1735 unsigned int gso_max_size;
1736 #define GSO_MAX_SEGS 65535
1737 u16 gso_max_segs;
1738 u16 gso_min_segs;
1739 #ifdef CONFIG_DCB
1740 const struct dcbnl_rtnl_ops *dcbnl_ops;
1741 #endif
1742 u8 num_tc;
1743 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1744 u8 prio_tc_map[TC_BITMASK + 1];
1745
1746 #if IS_ENABLED(CONFIG_FCOE)
1747 unsigned int fcoe_ddp_xid;
1748 #endif
1749 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1750 struct netprio_map __rcu *priomap;
1751 #endif
1752 struct phy_device *phydev;
1753 struct lock_class_key *qdisc_tx_busylock;
1754 };
1755 #define to_net_dev(d) container_of(d, struct net_device, dev)
1756
1757 #define NETDEV_ALIGN 32
1758
1759 static inline
1760 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1761 {
1762 return dev->prio_tc_map[prio & TC_BITMASK];
1763 }
1764
1765 static inline
1766 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1767 {
1768 if (tc >= dev->num_tc)
1769 return -EINVAL;
1770
1771 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1772 return 0;
1773 }
1774
1775 static inline
1776 void netdev_reset_tc(struct net_device *dev)
1777 {
1778 dev->num_tc = 0;
1779 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1780 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1781 }
1782
1783 static inline
1784 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1785 {
1786 if (tc >= dev->num_tc)
1787 return -EINVAL;
1788
1789 dev->tc_to_txq[tc].count = count;
1790 dev->tc_to_txq[tc].offset = offset;
1791 return 0;
1792 }
1793
1794 static inline
1795 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1796 {
1797 if (num_tc > TC_MAX_QUEUE)
1798 return -EINVAL;
1799
1800 dev->num_tc = num_tc;
1801 return 0;
1802 }
1803
1804 static inline
1805 int netdev_get_num_tc(struct net_device *dev)
1806 {
1807 return dev->num_tc;
1808 }
1809
1810 static inline
1811 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1812 unsigned int index)
1813 {
1814 return &dev->_tx[index];
1815 }
1816
1817 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1818 const struct sk_buff *skb)
1819 {
1820 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1821 }
1822
1823 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1824 void (*f)(struct net_device *,
1825 struct netdev_queue *,
1826 void *),
1827 void *arg)
1828 {
1829 unsigned int i;
1830
1831 for (i = 0; i < dev->num_tx_queues; i++)
1832 f(dev, &dev->_tx[i], arg);
1833 }
1834
1835 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1836 struct sk_buff *skb,
1837 void *accel_priv);
1838
1839 /*
1840 * Net namespace inlines
1841 */
1842 static inline
1843 struct net *dev_net(const struct net_device *dev)
1844 {
1845 return read_pnet(&dev->nd_net);
1846 }
1847
1848 static inline
1849 void dev_net_set(struct net_device *dev, struct net *net)
1850 {
1851 write_pnet(&dev->nd_net, net);
1852 }
1853
1854 static inline bool netdev_uses_dsa(struct net_device *dev)
1855 {
1856 #if IS_ENABLED(CONFIG_NET_DSA)
1857 if (dev->dsa_ptr != NULL)
1858 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1859 #endif
1860 return false;
1861 }
1862
1863 /**
1864 * netdev_priv - access network device private data
1865 * @dev: network device
1866 *
1867 * Get network device private data
1868 */
1869 static inline void *netdev_priv(const struct net_device *dev)
1870 {
1871 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1872 }
1873
1874 /* Set the sysfs physical device reference for the network logical device
1875 * if set prior to registration will cause a symlink during initialization.
1876 */
1877 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1878
1879 /* Set the sysfs device type for the network logical device to allow
1880 * fine-grained identification of different network device types. For
1881 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1882 */
1883 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1884
1885 /* Default NAPI poll() weight
1886 * Device drivers are strongly advised to not use bigger value
1887 */
1888 #define NAPI_POLL_WEIGHT 64
1889
1890 /**
1891 * netif_napi_add - initialize a napi context
1892 * @dev: network device
1893 * @napi: napi context
1894 * @poll: polling function
1895 * @weight: default weight
1896 *
1897 * netif_napi_add() must be used to initialize a napi context prior to calling
1898 * *any* of the other napi related functions.
1899 */
1900 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1901 int (*poll)(struct napi_struct *, int), int weight);
1902
1903 /**
1904 * netif_napi_del - remove a napi context
1905 * @napi: napi context
1906 *
1907 * netif_napi_del() removes a napi context from the network device napi list
1908 */
1909 void netif_napi_del(struct napi_struct *napi);
1910
1911 struct napi_gro_cb {
1912 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1913 void *frag0;
1914
1915 /* Length of frag0. */
1916 unsigned int frag0_len;
1917
1918 /* This indicates where we are processing relative to skb->data. */
1919 int data_offset;
1920
1921 /* This is non-zero if the packet cannot be merged with the new skb. */
1922 u16 flush;
1923
1924 /* Save the IP ID here and check when we get to the transport layer */
1925 u16 flush_id;
1926
1927 /* Number of segments aggregated. */
1928 u16 count;
1929
1930 /* Start offset for remote checksum offload */
1931 u16 gro_remcsum_start;
1932
1933 /* jiffies when first packet was created/queued */
1934 unsigned long age;
1935
1936 /* Used in ipv6_gro_receive() and foo-over-udp */
1937 u16 proto;
1938
1939 /* This is non-zero if the packet may be of the same flow. */
1940 u8 same_flow:1;
1941
1942 /* Used in udp_gro_receive */
1943 u8 udp_mark:1;
1944
1945 /* GRO checksum is valid */
1946 u8 csum_valid:1;
1947
1948 /* Number of checksums via CHECKSUM_UNNECESSARY */
1949 u8 csum_cnt:3;
1950
1951 /* Free the skb? */
1952 u8 free:2;
1953 #define NAPI_GRO_FREE 1
1954 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1955
1956 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1957 u8 is_ipv6:1;
1958
1959 /* 7 bit hole */
1960
1961 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1962 __wsum csum;
1963
1964 /* used in skb_gro_receive() slow path */
1965 struct sk_buff *last;
1966 };
1967
1968 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1969
1970 struct packet_type {
1971 __be16 type; /* This is really htons(ether_type). */
1972 struct net_device *dev; /* NULL is wildcarded here */
1973 int (*func) (struct sk_buff *,
1974 struct net_device *,
1975 struct packet_type *,
1976 struct net_device *);
1977 bool (*id_match)(struct packet_type *ptype,
1978 struct sock *sk);
1979 void *af_packet_priv;
1980 struct list_head list;
1981 };
1982
1983 struct offload_callbacks {
1984 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1985 netdev_features_t features);
1986 struct sk_buff **(*gro_receive)(struct sk_buff **head,
1987 struct sk_buff *skb);
1988 int (*gro_complete)(struct sk_buff *skb, int nhoff);
1989 };
1990
1991 struct packet_offload {
1992 __be16 type; /* This is really htons(ether_type). */
1993 struct offload_callbacks callbacks;
1994 struct list_head list;
1995 };
1996
1997 struct udp_offload;
1998
1999 struct udp_offload_callbacks {
2000 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2001 struct sk_buff *skb,
2002 struct udp_offload *uoff);
2003 int (*gro_complete)(struct sk_buff *skb,
2004 int nhoff,
2005 struct udp_offload *uoff);
2006 };
2007
2008 struct udp_offload {
2009 __be16 port;
2010 u8 ipproto;
2011 struct udp_offload_callbacks callbacks;
2012 };
2013
2014 /* often modified stats are per cpu, other are shared (netdev->stats) */
2015 struct pcpu_sw_netstats {
2016 u64 rx_packets;
2017 u64 rx_bytes;
2018 u64 tx_packets;
2019 u64 tx_bytes;
2020 struct u64_stats_sync syncp;
2021 };
2022
2023 #define netdev_alloc_pcpu_stats(type) \
2024 ({ \
2025 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2026 if (pcpu_stats) { \
2027 int i; \
2028 for_each_possible_cpu(i) { \
2029 typeof(type) *stat; \
2030 stat = per_cpu_ptr(pcpu_stats, i); \
2031 u64_stats_init(&stat->syncp); \
2032 } \
2033 } \
2034 pcpu_stats; \
2035 })
2036
2037 #include <linux/notifier.h>
2038
2039 /* netdevice notifier chain. Please remember to update the rtnetlink
2040 * notification exclusion list in rtnetlink_event() when adding new
2041 * types.
2042 */
2043 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2044 #define NETDEV_DOWN 0x0002
2045 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2046 detected a hardware crash and restarted
2047 - we can use this eg to kick tcp sessions
2048 once done */
2049 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2050 #define NETDEV_REGISTER 0x0005
2051 #define NETDEV_UNREGISTER 0x0006
2052 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2053 #define NETDEV_CHANGEADDR 0x0008
2054 #define NETDEV_GOING_DOWN 0x0009
2055 #define NETDEV_CHANGENAME 0x000A
2056 #define NETDEV_FEAT_CHANGE 0x000B
2057 #define NETDEV_BONDING_FAILOVER 0x000C
2058 #define NETDEV_PRE_UP 0x000D
2059 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2060 #define NETDEV_POST_TYPE_CHANGE 0x000F
2061 #define NETDEV_POST_INIT 0x0010
2062 #define NETDEV_UNREGISTER_FINAL 0x0011
2063 #define NETDEV_RELEASE 0x0012
2064 #define NETDEV_NOTIFY_PEERS 0x0013
2065 #define NETDEV_JOIN 0x0014
2066 #define NETDEV_CHANGEUPPER 0x0015
2067 #define NETDEV_RESEND_IGMP 0x0016
2068 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2069 #define NETDEV_CHANGEINFODATA 0x0018
2070 #define NETDEV_BONDING_INFO 0x0019
2071
2072 int register_netdevice_notifier(struct notifier_block *nb);
2073 int unregister_netdevice_notifier(struct notifier_block *nb);
2074
2075 struct netdev_notifier_info {
2076 struct net_device *dev;
2077 };
2078
2079 struct netdev_notifier_change_info {
2080 struct netdev_notifier_info info; /* must be first */
2081 unsigned int flags_changed;
2082 };
2083
2084 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2085 struct net_device *dev)
2086 {
2087 info->dev = dev;
2088 }
2089
2090 static inline struct net_device *
2091 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2092 {
2093 return info->dev;
2094 }
2095
2096 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2097
2098
2099 extern rwlock_t dev_base_lock; /* Device list lock */
2100
2101 #define for_each_netdev(net, d) \
2102 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2103 #define for_each_netdev_reverse(net, d) \
2104 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2105 #define for_each_netdev_rcu(net, d) \
2106 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2107 #define for_each_netdev_safe(net, d, n) \
2108 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2109 #define for_each_netdev_continue(net, d) \
2110 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2111 #define for_each_netdev_continue_rcu(net, d) \
2112 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2113 #define for_each_netdev_in_bond_rcu(bond, slave) \
2114 for_each_netdev_rcu(&init_net, slave) \
2115 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2116 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2117
2118 static inline struct net_device *next_net_device(struct net_device *dev)
2119 {
2120 struct list_head *lh;
2121 struct net *net;
2122
2123 net = dev_net(dev);
2124 lh = dev->dev_list.next;
2125 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2126 }
2127
2128 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2129 {
2130 struct list_head *lh;
2131 struct net *net;
2132
2133 net = dev_net(dev);
2134 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2135 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2136 }
2137
2138 static inline struct net_device *first_net_device(struct net *net)
2139 {
2140 return list_empty(&net->dev_base_head) ? NULL :
2141 net_device_entry(net->dev_base_head.next);
2142 }
2143
2144 static inline struct net_device *first_net_device_rcu(struct net *net)
2145 {
2146 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2147
2148 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2149 }
2150
2151 int netdev_boot_setup_check(struct net_device *dev);
2152 unsigned long netdev_boot_base(const char *prefix, int unit);
2153 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2154 const char *hwaddr);
2155 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2156 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2157 void dev_add_pack(struct packet_type *pt);
2158 void dev_remove_pack(struct packet_type *pt);
2159 void __dev_remove_pack(struct packet_type *pt);
2160 void dev_add_offload(struct packet_offload *po);
2161 void dev_remove_offload(struct packet_offload *po);
2162
2163 int dev_get_iflink(const struct net_device *dev);
2164 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2165 unsigned short mask);
2166 struct net_device *dev_get_by_name(struct net *net, const char *name);
2167 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2168 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2169 int dev_alloc_name(struct net_device *dev, const char *name);
2170 int dev_open(struct net_device *dev);
2171 int dev_close(struct net_device *dev);
2172 int dev_close_many(struct list_head *head, bool unlink);
2173 void dev_disable_lro(struct net_device *dev);
2174 int dev_loopback_xmit(struct sock *sk, struct sk_buff *newskb);
2175 int dev_queue_xmit_sk(struct sock *sk, struct sk_buff *skb);
2176 static inline int dev_queue_xmit(struct sk_buff *skb)
2177 {
2178 return dev_queue_xmit_sk(skb->sk, skb);
2179 }
2180 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2181 int register_netdevice(struct net_device *dev);
2182 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2183 void unregister_netdevice_many(struct list_head *head);
2184 static inline void unregister_netdevice(struct net_device *dev)
2185 {
2186 unregister_netdevice_queue(dev, NULL);
2187 }
2188
2189 int netdev_refcnt_read(const struct net_device *dev);
2190 void free_netdev(struct net_device *dev);
2191 void netdev_freemem(struct net_device *dev);
2192 void synchronize_net(void);
2193 int init_dummy_netdev(struct net_device *dev);
2194
2195 DECLARE_PER_CPU(int, xmit_recursion);
2196 static inline int dev_recursion_level(void)
2197 {
2198 return this_cpu_read(xmit_recursion);
2199 }
2200
2201 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2202 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2203 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2204 int netdev_get_name(struct net *net, char *name, int ifindex);
2205 int dev_restart(struct net_device *dev);
2206 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2207
2208 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2209 {
2210 return NAPI_GRO_CB(skb)->data_offset;
2211 }
2212
2213 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2214 {
2215 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2216 }
2217
2218 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2219 {
2220 NAPI_GRO_CB(skb)->data_offset += len;
2221 }
2222
2223 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2224 unsigned int offset)
2225 {
2226 return NAPI_GRO_CB(skb)->frag0 + offset;
2227 }
2228
2229 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2230 {
2231 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2232 }
2233
2234 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2235 unsigned int offset)
2236 {
2237 if (!pskb_may_pull(skb, hlen))
2238 return NULL;
2239
2240 NAPI_GRO_CB(skb)->frag0 = NULL;
2241 NAPI_GRO_CB(skb)->frag0_len = 0;
2242 return skb->data + offset;
2243 }
2244
2245 static inline void *skb_gro_network_header(struct sk_buff *skb)
2246 {
2247 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2248 skb_network_offset(skb);
2249 }
2250
2251 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2252 const void *start, unsigned int len)
2253 {
2254 if (NAPI_GRO_CB(skb)->csum_valid)
2255 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2256 csum_partial(start, len, 0));
2257 }
2258
2259 /* GRO checksum functions. These are logical equivalents of the normal
2260 * checksum functions (in skbuff.h) except that they operate on the GRO
2261 * offsets and fields in sk_buff.
2262 */
2263
2264 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2265
2266 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2267 {
2268 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2269 skb_gro_offset(skb));
2270 }
2271
2272 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2273 bool zero_okay,
2274 __sum16 check)
2275 {
2276 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2277 skb_checksum_start_offset(skb) <
2278 skb_gro_offset(skb)) &&
2279 !skb_at_gro_remcsum_start(skb) &&
2280 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2281 (!zero_okay || check));
2282 }
2283
2284 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2285 __wsum psum)
2286 {
2287 if (NAPI_GRO_CB(skb)->csum_valid &&
2288 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2289 return 0;
2290
2291 NAPI_GRO_CB(skb)->csum = psum;
2292
2293 return __skb_gro_checksum_complete(skb);
2294 }
2295
2296 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2297 {
2298 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2299 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2300 NAPI_GRO_CB(skb)->csum_cnt--;
2301 } else {
2302 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2303 * verified a new top level checksum or an encapsulated one
2304 * during GRO. This saves work if we fallback to normal path.
2305 */
2306 __skb_incr_checksum_unnecessary(skb);
2307 }
2308 }
2309
2310 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2311 compute_pseudo) \
2312 ({ \
2313 __sum16 __ret = 0; \
2314 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2315 __ret = __skb_gro_checksum_validate_complete(skb, \
2316 compute_pseudo(skb, proto)); \
2317 if (__ret) \
2318 __skb_mark_checksum_bad(skb); \
2319 else \
2320 skb_gro_incr_csum_unnecessary(skb); \
2321 __ret; \
2322 })
2323
2324 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2325 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2326
2327 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2328 compute_pseudo) \
2329 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2330
2331 #define skb_gro_checksum_simple_validate(skb) \
2332 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2333
2334 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2335 {
2336 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2337 !NAPI_GRO_CB(skb)->csum_valid);
2338 }
2339
2340 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2341 __sum16 check, __wsum pseudo)
2342 {
2343 NAPI_GRO_CB(skb)->csum = ~pseudo;
2344 NAPI_GRO_CB(skb)->csum_valid = 1;
2345 }
2346
2347 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2348 do { \
2349 if (__skb_gro_checksum_convert_check(skb)) \
2350 __skb_gro_checksum_convert(skb, check, \
2351 compute_pseudo(skb, proto)); \
2352 } while (0)
2353
2354 struct gro_remcsum {
2355 int offset;
2356 __wsum delta;
2357 };
2358
2359 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2360 {
2361 grc->offset = 0;
2362 grc->delta = 0;
2363 }
2364
2365 static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2366 int start, int offset,
2367 struct gro_remcsum *grc,
2368 bool nopartial)
2369 {
2370 __wsum delta;
2371
2372 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2373
2374 if (!nopartial) {
2375 NAPI_GRO_CB(skb)->gro_remcsum_start =
2376 ((unsigned char *)ptr + start) - skb->head;
2377 return;
2378 }
2379
2380 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2381
2382 /* Adjust skb->csum since we changed the packet */
2383 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2384
2385 grc->offset = (ptr + offset) - (void *)skb->head;
2386 grc->delta = delta;
2387 }
2388
2389 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2390 struct gro_remcsum *grc)
2391 {
2392 if (!grc->delta)
2393 return;
2394
2395 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2396 }
2397
2398 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2399 unsigned short type,
2400 const void *daddr, const void *saddr,
2401 unsigned int len)
2402 {
2403 if (!dev->header_ops || !dev->header_ops->create)
2404 return 0;
2405
2406 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2407 }
2408
2409 static inline int dev_parse_header(const struct sk_buff *skb,
2410 unsigned char *haddr)
2411 {
2412 const struct net_device *dev = skb->dev;
2413
2414 if (!dev->header_ops || !dev->header_ops->parse)
2415 return 0;
2416 return dev->header_ops->parse(skb, haddr);
2417 }
2418
2419 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2420 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2421 static inline int unregister_gifconf(unsigned int family)
2422 {
2423 return register_gifconf(family, NULL);
2424 }
2425
2426 #ifdef CONFIG_NET_FLOW_LIMIT
2427 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2428 struct sd_flow_limit {
2429 u64 count;
2430 unsigned int num_buckets;
2431 unsigned int history_head;
2432 u16 history[FLOW_LIMIT_HISTORY];
2433 u8 buckets[];
2434 };
2435
2436 extern int netdev_flow_limit_table_len;
2437 #endif /* CONFIG_NET_FLOW_LIMIT */
2438
2439 /*
2440 * Incoming packets are placed on per-cpu queues
2441 */
2442 struct softnet_data {
2443 struct list_head poll_list;
2444 struct sk_buff_head process_queue;
2445
2446 /* stats */
2447 unsigned int processed;
2448 unsigned int time_squeeze;
2449 unsigned int cpu_collision;
2450 unsigned int received_rps;
2451 #ifdef CONFIG_RPS
2452 struct softnet_data *rps_ipi_list;
2453 #endif
2454 #ifdef CONFIG_NET_FLOW_LIMIT
2455 struct sd_flow_limit __rcu *flow_limit;
2456 #endif
2457 struct Qdisc *output_queue;
2458 struct Qdisc **output_queue_tailp;
2459 struct sk_buff *completion_queue;
2460
2461 #ifdef CONFIG_RPS
2462 /* Elements below can be accessed between CPUs for RPS */
2463 struct call_single_data csd ____cacheline_aligned_in_smp;
2464 struct softnet_data *rps_ipi_next;
2465 unsigned int cpu;
2466 unsigned int input_queue_head;
2467 unsigned int input_queue_tail;
2468 #endif
2469 unsigned int dropped;
2470 struct sk_buff_head input_pkt_queue;
2471 struct napi_struct backlog;
2472
2473 };
2474
2475 static inline void input_queue_head_incr(struct softnet_data *sd)
2476 {
2477 #ifdef CONFIG_RPS
2478 sd->input_queue_head++;
2479 #endif
2480 }
2481
2482 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2483 unsigned int *qtail)
2484 {
2485 #ifdef CONFIG_RPS
2486 *qtail = ++sd->input_queue_tail;
2487 #endif
2488 }
2489
2490 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2491
2492 void __netif_schedule(struct Qdisc *q);
2493 void netif_schedule_queue(struct netdev_queue *txq);
2494
2495 static inline void netif_tx_schedule_all(struct net_device *dev)
2496 {
2497 unsigned int i;
2498
2499 for (i = 0; i < dev->num_tx_queues; i++)
2500 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2501 }
2502
2503 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2504 {
2505 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2506 }
2507
2508 /**
2509 * netif_start_queue - allow transmit
2510 * @dev: network device
2511 *
2512 * Allow upper layers to call the device hard_start_xmit routine.
2513 */
2514 static inline void netif_start_queue(struct net_device *dev)
2515 {
2516 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2517 }
2518
2519 static inline void netif_tx_start_all_queues(struct net_device *dev)
2520 {
2521 unsigned int i;
2522
2523 for (i = 0; i < dev->num_tx_queues; i++) {
2524 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2525 netif_tx_start_queue(txq);
2526 }
2527 }
2528
2529 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2530
2531 /**
2532 * netif_wake_queue - restart transmit
2533 * @dev: network device
2534 *
2535 * Allow upper layers to call the device hard_start_xmit routine.
2536 * Used for flow control when transmit resources are available.
2537 */
2538 static inline void netif_wake_queue(struct net_device *dev)
2539 {
2540 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2541 }
2542
2543 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2544 {
2545 unsigned int i;
2546
2547 for (i = 0; i < dev->num_tx_queues; i++) {
2548 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2549 netif_tx_wake_queue(txq);
2550 }
2551 }
2552
2553 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2554 {
2555 if (WARN_ON(!dev_queue)) {
2556 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2557 return;
2558 }
2559 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2560 }
2561
2562 /**
2563 * netif_stop_queue - stop transmitted packets
2564 * @dev: network device
2565 *
2566 * Stop upper layers calling the device hard_start_xmit routine.
2567 * Used for flow control when transmit resources are unavailable.
2568 */
2569 static inline void netif_stop_queue(struct net_device *dev)
2570 {
2571 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2572 }
2573
2574 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2575 {
2576 unsigned int i;
2577
2578 for (i = 0; i < dev->num_tx_queues; i++) {
2579 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2580 netif_tx_stop_queue(txq);
2581 }
2582 }
2583
2584 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2585 {
2586 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2587 }
2588
2589 /**
2590 * netif_queue_stopped - test if transmit queue is flowblocked
2591 * @dev: network device
2592 *
2593 * Test if transmit queue on device is currently unable to send.
2594 */
2595 static inline bool netif_queue_stopped(const struct net_device *dev)
2596 {
2597 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2598 }
2599
2600 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2601 {
2602 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2603 }
2604
2605 static inline bool
2606 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2607 {
2608 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2609 }
2610
2611 static inline bool
2612 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2613 {
2614 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2615 }
2616
2617 /**
2618 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2619 * @dev_queue: pointer to transmit queue
2620 *
2621 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2622 * to give appropriate hint to the cpu.
2623 */
2624 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2625 {
2626 #ifdef CONFIG_BQL
2627 prefetchw(&dev_queue->dql.num_queued);
2628 #endif
2629 }
2630
2631 /**
2632 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2633 * @dev_queue: pointer to transmit queue
2634 *
2635 * BQL enabled drivers might use this helper in their TX completion path,
2636 * to give appropriate hint to the cpu.
2637 */
2638 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2639 {
2640 #ifdef CONFIG_BQL
2641 prefetchw(&dev_queue->dql.limit);
2642 #endif
2643 }
2644
2645 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2646 unsigned int bytes)
2647 {
2648 #ifdef CONFIG_BQL
2649 dql_queued(&dev_queue->dql, bytes);
2650
2651 if (likely(dql_avail(&dev_queue->dql) >= 0))
2652 return;
2653
2654 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2655
2656 /*
2657 * The XOFF flag must be set before checking the dql_avail below,
2658 * because in netdev_tx_completed_queue we update the dql_completed
2659 * before checking the XOFF flag.
2660 */
2661 smp_mb();
2662
2663 /* check again in case another CPU has just made room avail */
2664 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2665 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2666 #endif
2667 }
2668
2669 /**
2670 * netdev_sent_queue - report the number of bytes queued to hardware
2671 * @dev: network device
2672 * @bytes: number of bytes queued to the hardware device queue
2673 *
2674 * Report the number of bytes queued for sending/completion to the network
2675 * device hardware queue. @bytes should be a good approximation and should
2676 * exactly match netdev_completed_queue() @bytes
2677 */
2678 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2679 {
2680 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2681 }
2682
2683 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2684 unsigned int pkts, unsigned int bytes)
2685 {
2686 #ifdef CONFIG_BQL
2687 if (unlikely(!bytes))
2688 return;
2689
2690 dql_completed(&dev_queue->dql, bytes);
2691
2692 /*
2693 * Without the memory barrier there is a small possiblity that
2694 * netdev_tx_sent_queue will miss the update and cause the queue to
2695 * be stopped forever
2696 */
2697 smp_mb();
2698
2699 if (dql_avail(&dev_queue->dql) < 0)
2700 return;
2701
2702 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2703 netif_schedule_queue(dev_queue);
2704 #endif
2705 }
2706
2707 /**
2708 * netdev_completed_queue - report bytes and packets completed by device
2709 * @dev: network device
2710 * @pkts: actual number of packets sent over the medium
2711 * @bytes: actual number of bytes sent over the medium
2712 *
2713 * Report the number of bytes and packets transmitted by the network device
2714 * hardware queue over the physical medium, @bytes must exactly match the
2715 * @bytes amount passed to netdev_sent_queue()
2716 */
2717 static inline void netdev_completed_queue(struct net_device *dev,
2718 unsigned int pkts, unsigned int bytes)
2719 {
2720 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2721 }
2722
2723 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2724 {
2725 #ifdef CONFIG_BQL
2726 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2727 dql_reset(&q->dql);
2728 #endif
2729 }
2730
2731 /**
2732 * netdev_reset_queue - reset the packets and bytes count of a network device
2733 * @dev_queue: network device
2734 *
2735 * Reset the bytes and packet count of a network device and clear the
2736 * software flow control OFF bit for this network device
2737 */
2738 static inline void netdev_reset_queue(struct net_device *dev_queue)
2739 {
2740 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2741 }
2742
2743 /**
2744 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2745 * @dev: network device
2746 * @queue_index: given tx queue index
2747 *
2748 * Returns 0 if given tx queue index >= number of device tx queues,
2749 * otherwise returns the originally passed tx queue index.
2750 */
2751 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2752 {
2753 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2754 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2755 dev->name, queue_index,
2756 dev->real_num_tx_queues);
2757 return 0;
2758 }
2759
2760 return queue_index;
2761 }
2762
2763 /**
2764 * netif_running - test if up
2765 * @dev: network device
2766 *
2767 * Test if the device has been brought up.
2768 */
2769 static inline bool netif_running(const struct net_device *dev)
2770 {
2771 return test_bit(__LINK_STATE_START, &dev->state);
2772 }
2773
2774 /*
2775 * Routines to manage the subqueues on a device. We only need start
2776 * stop, and a check if it's stopped. All other device management is
2777 * done at the overall netdevice level.
2778 * Also test the device if we're multiqueue.
2779 */
2780
2781 /**
2782 * netif_start_subqueue - allow sending packets on subqueue
2783 * @dev: network device
2784 * @queue_index: sub queue index
2785 *
2786 * Start individual transmit queue of a device with multiple transmit queues.
2787 */
2788 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2789 {
2790 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2791
2792 netif_tx_start_queue(txq);
2793 }
2794
2795 /**
2796 * netif_stop_subqueue - stop sending packets on subqueue
2797 * @dev: network device
2798 * @queue_index: sub queue index
2799 *
2800 * Stop individual transmit queue of a device with multiple transmit queues.
2801 */
2802 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2803 {
2804 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2805 netif_tx_stop_queue(txq);
2806 }
2807
2808 /**
2809 * netif_subqueue_stopped - test status of subqueue
2810 * @dev: network device
2811 * @queue_index: sub queue index
2812 *
2813 * Check individual transmit queue of a device with multiple transmit queues.
2814 */
2815 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2816 u16 queue_index)
2817 {
2818 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2819
2820 return netif_tx_queue_stopped(txq);
2821 }
2822
2823 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2824 struct sk_buff *skb)
2825 {
2826 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2827 }
2828
2829 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2830
2831 #ifdef CONFIG_XPS
2832 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2833 u16 index);
2834 #else
2835 static inline int netif_set_xps_queue(struct net_device *dev,
2836 const struct cpumask *mask,
2837 u16 index)
2838 {
2839 return 0;
2840 }
2841 #endif
2842
2843 /*
2844 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2845 * as a distribution range limit for the returned value.
2846 */
2847 static inline u16 skb_tx_hash(const struct net_device *dev,
2848 struct sk_buff *skb)
2849 {
2850 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2851 }
2852
2853 /**
2854 * netif_is_multiqueue - test if device has multiple transmit queues
2855 * @dev: network device
2856 *
2857 * Check if device has multiple transmit queues
2858 */
2859 static inline bool netif_is_multiqueue(const struct net_device *dev)
2860 {
2861 return dev->num_tx_queues > 1;
2862 }
2863
2864 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2865
2866 #ifdef CONFIG_SYSFS
2867 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2868 #else
2869 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2870 unsigned int rxq)
2871 {
2872 return 0;
2873 }
2874 #endif
2875
2876 #ifdef CONFIG_SYSFS
2877 static inline unsigned int get_netdev_rx_queue_index(
2878 struct netdev_rx_queue *queue)
2879 {
2880 struct net_device *dev = queue->dev;
2881 int index = queue - dev->_rx;
2882
2883 BUG_ON(index >= dev->num_rx_queues);
2884 return index;
2885 }
2886 #endif
2887
2888 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2889 int netif_get_num_default_rss_queues(void);
2890
2891 enum skb_free_reason {
2892 SKB_REASON_CONSUMED,
2893 SKB_REASON_DROPPED,
2894 };
2895
2896 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2897 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2898
2899 /*
2900 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2901 * interrupt context or with hardware interrupts being disabled.
2902 * (in_irq() || irqs_disabled())
2903 *
2904 * We provide four helpers that can be used in following contexts :
2905 *
2906 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2907 * replacing kfree_skb(skb)
2908 *
2909 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2910 * Typically used in place of consume_skb(skb) in TX completion path
2911 *
2912 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2913 * replacing kfree_skb(skb)
2914 *
2915 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2916 * and consumed a packet. Used in place of consume_skb(skb)
2917 */
2918 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2919 {
2920 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2921 }
2922
2923 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2924 {
2925 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2926 }
2927
2928 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2929 {
2930 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2931 }
2932
2933 static inline void dev_consume_skb_any(struct sk_buff *skb)
2934 {
2935 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2936 }
2937
2938 int netif_rx(struct sk_buff *skb);
2939 int netif_rx_ni(struct sk_buff *skb);
2940 int netif_receive_skb_sk(struct sock *sk, struct sk_buff *skb);
2941 static inline int netif_receive_skb(struct sk_buff *skb)
2942 {
2943 return netif_receive_skb_sk(skb->sk, skb);
2944 }
2945 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2946 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2947 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2948 gro_result_t napi_gro_frags(struct napi_struct *napi);
2949 struct packet_offload *gro_find_receive_by_type(__be16 type);
2950 struct packet_offload *gro_find_complete_by_type(__be16 type);
2951
2952 static inline void napi_free_frags(struct napi_struct *napi)
2953 {
2954 kfree_skb(napi->skb);
2955 napi->skb = NULL;
2956 }
2957
2958 int netdev_rx_handler_register(struct net_device *dev,
2959 rx_handler_func_t *rx_handler,
2960 void *rx_handler_data);
2961 void netdev_rx_handler_unregister(struct net_device *dev);
2962
2963 bool dev_valid_name(const char *name);
2964 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2965 int dev_ethtool(struct net *net, struct ifreq *);
2966 unsigned int dev_get_flags(const struct net_device *);
2967 int __dev_change_flags(struct net_device *, unsigned int flags);
2968 int dev_change_flags(struct net_device *, unsigned int);
2969 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2970 unsigned int gchanges);
2971 int dev_change_name(struct net_device *, const char *);
2972 int dev_set_alias(struct net_device *, const char *, size_t);
2973 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2974 int dev_set_mtu(struct net_device *, int);
2975 void dev_set_group(struct net_device *, int);
2976 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2977 int dev_change_carrier(struct net_device *, bool new_carrier);
2978 int dev_get_phys_port_id(struct net_device *dev,
2979 struct netdev_phys_item_id *ppid);
2980 int dev_get_phys_port_name(struct net_device *dev,
2981 char *name, size_t len);
2982 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2983 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2984 struct netdev_queue *txq, int *ret);
2985 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2986 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2987 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2988
2989 extern int netdev_budget;
2990
2991 /* Called by rtnetlink.c:rtnl_unlock() */
2992 void netdev_run_todo(void);
2993
2994 /**
2995 * dev_put - release reference to device
2996 * @dev: network device
2997 *
2998 * Release reference to device to allow it to be freed.
2999 */
3000 static inline void dev_put(struct net_device *dev)
3001 {
3002 this_cpu_dec(*dev->pcpu_refcnt);
3003 }
3004
3005 /**
3006 * dev_hold - get reference to device
3007 * @dev: network device
3008 *
3009 * Hold reference to device to keep it from being freed.
3010 */
3011 static inline void dev_hold(struct net_device *dev)
3012 {
3013 this_cpu_inc(*dev->pcpu_refcnt);
3014 }
3015
3016 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3017 * and _off may be called from IRQ context, but it is caller
3018 * who is responsible for serialization of these calls.
3019 *
3020 * The name carrier is inappropriate, these functions should really be
3021 * called netif_lowerlayer_*() because they represent the state of any
3022 * kind of lower layer not just hardware media.
3023 */
3024
3025 void linkwatch_init_dev(struct net_device *dev);
3026 void linkwatch_fire_event(struct net_device *dev);
3027 void linkwatch_forget_dev(struct net_device *dev);
3028
3029 /**
3030 * netif_carrier_ok - test if carrier present
3031 * @dev: network device
3032 *
3033 * Check if carrier is present on device
3034 */
3035 static inline bool netif_carrier_ok(const struct net_device *dev)
3036 {
3037 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3038 }
3039
3040 unsigned long dev_trans_start(struct net_device *dev);
3041
3042 void __netdev_watchdog_up(struct net_device *dev);
3043
3044 void netif_carrier_on(struct net_device *dev);
3045
3046 void netif_carrier_off(struct net_device *dev);
3047
3048 /**
3049 * netif_dormant_on - mark device as dormant.
3050 * @dev: network device
3051 *
3052 * Mark device as dormant (as per RFC2863).
3053 *
3054 * The dormant state indicates that the relevant interface is not
3055 * actually in a condition to pass packets (i.e., it is not 'up') but is
3056 * in a "pending" state, waiting for some external event. For "on-
3057 * demand" interfaces, this new state identifies the situation where the
3058 * interface is waiting for events to place it in the up state.
3059 *
3060 */
3061 static inline void netif_dormant_on(struct net_device *dev)
3062 {
3063 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3064 linkwatch_fire_event(dev);
3065 }
3066
3067 /**
3068 * netif_dormant_off - set device as not dormant.
3069 * @dev: network device
3070 *
3071 * Device is not in dormant state.
3072 */
3073 static inline void netif_dormant_off(struct net_device *dev)
3074 {
3075 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3076 linkwatch_fire_event(dev);
3077 }
3078
3079 /**
3080 * netif_dormant - test if carrier present
3081 * @dev: network device
3082 *
3083 * Check if carrier is present on device
3084 */
3085 static inline bool netif_dormant(const struct net_device *dev)
3086 {
3087 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3088 }
3089
3090
3091 /**
3092 * netif_oper_up - test if device is operational
3093 * @dev: network device
3094 *
3095 * Check if carrier is operational
3096 */
3097 static inline bool netif_oper_up(const struct net_device *dev)
3098 {
3099 return (dev->operstate == IF_OPER_UP ||
3100 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3101 }
3102
3103 /**
3104 * netif_device_present - is device available or removed
3105 * @dev: network device
3106 *
3107 * Check if device has not been removed from system.
3108 */
3109 static inline bool netif_device_present(struct net_device *dev)
3110 {
3111 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3112 }
3113
3114 void netif_device_detach(struct net_device *dev);
3115
3116 void netif_device_attach(struct net_device *dev);
3117
3118 /*
3119 * Network interface message level settings
3120 */
3121
3122 enum {
3123 NETIF_MSG_DRV = 0x0001,
3124 NETIF_MSG_PROBE = 0x0002,
3125 NETIF_MSG_LINK = 0x0004,
3126 NETIF_MSG_TIMER = 0x0008,
3127 NETIF_MSG_IFDOWN = 0x0010,
3128 NETIF_MSG_IFUP = 0x0020,
3129 NETIF_MSG_RX_ERR = 0x0040,
3130 NETIF_MSG_TX_ERR = 0x0080,
3131 NETIF_MSG_TX_QUEUED = 0x0100,
3132 NETIF_MSG_INTR = 0x0200,
3133 NETIF_MSG_TX_DONE = 0x0400,
3134 NETIF_MSG_RX_STATUS = 0x0800,
3135 NETIF_MSG_PKTDATA = 0x1000,
3136 NETIF_MSG_HW = 0x2000,
3137 NETIF_MSG_WOL = 0x4000,
3138 };
3139
3140 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3141 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3142 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3143 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3144 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3145 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3146 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3147 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3148 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3149 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3150 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3151 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3152 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3153 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3154 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3155
3156 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3157 {
3158 /* use default */
3159 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3160 return default_msg_enable_bits;
3161 if (debug_value == 0) /* no output */
3162 return 0;
3163 /* set low N bits */
3164 return (1 << debug_value) - 1;
3165 }
3166
3167 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3168 {
3169 spin_lock(&txq->_xmit_lock);
3170 txq->xmit_lock_owner = cpu;
3171 }
3172
3173 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3174 {
3175 spin_lock_bh(&txq->_xmit_lock);
3176 txq->xmit_lock_owner = smp_processor_id();
3177 }
3178
3179 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3180 {
3181 bool ok = spin_trylock(&txq->_xmit_lock);
3182 if (likely(ok))
3183 txq->xmit_lock_owner = smp_processor_id();
3184 return ok;
3185 }
3186
3187 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3188 {
3189 txq->xmit_lock_owner = -1;
3190 spin_unlock(&txq->_xmit_lock);
3191 }
3192
3193 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3194 {
3195 txq->xmit_lock_owner = -1;
3196 spin_unlock_bh(&txq->_xmit_lock);
3197 }
3198
3199 static inline void txq_trans_update(struct netdev_queue *txq)
3200 {
3201 if (txq->xmit_lock_owner != -1)
3202 txq->trans_start = jiffies;
3203 }
3204
3205 /**
3206 * netif_tx_lock - grab network device transmit lock
3207 * @dev: network device
3208 *
3209 * Get network device transmit lock
3210 */
3211 static inline void netif_tx_lock(struct net_device *dev)
3212 {
3213 unsigned int i;
3214 int cpu;
3215
3216 spin_lock(&dev->tx_global_lock);
3217 cpu = smp_processor_id();
3218 for (i = 0; i < dev->num_tx_queues; i++) {
3219 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3220
3221 /* We are the only thread of execution doing a
3222 * freeze, but we have to grab the _xmit_lock in
3223 * order to synchronize with threads which are in
3224 * the ->hard_start_xmit() handler and already
3225 * checked the frozen bit.
3226 */
3227 __netif_tx_lock(txq, cpu);
3228 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3229 __netif_tx_unlock(txq);
3230 }
3231 }
3232
3233 static inline void netif_tx_lock_bh(struct net_device *dev)
3234 {
3235 local_bh_disable();
3236 netif_tx_lock(dev);
3237 }
3238
3239 static inline void netif_tx_unlock(struct net_device *dev)
3240 {
3241 unsigned int i;
3242
3243 for (i = 0; i < dev->num_tx_queues; i++) {
3244 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3245
3246 /* No need to grab the _xmit_lock here. If the
3247 * queue is not stopped for another reason, we
3248 * force a schedule.
3249 */
3250 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3251 netif_schedule_queue(txq);
3252 }
3253 spin_unlock(&dev->tx_global_lock);
3254 }
3255
3256 static inline void netif_tx_unlock_bh(struct net_device *dev)
3257 {
3258 netif_tx_unlock(dev);
3259 local_bh_enable();
3260 }
3261
3262 #define HARD_TX_LOCK(dev, txq, cpu) { \
3263 if ((dev->features & NETIF_F_LLTX) == 0) { \
3264 __netif_tx_lock(txq, cpu); \
3265 } \
3266 }
3267
3268 #define HARD_TX_TRYLOCK(dev, txq) \
3269 (((dev->features & NETIF_F_LLTX) == 0) ? \
3270 __netif_tx_trylock(txq) : \
3271 true )
3272
3273 #define HARD_TX_UNLOCK(dev, txq) { \
3274 if ((dev->features & NETIF_F_LLTX) == 0) { \
3275 __netif_tx_unlock(txq); \
3276 } \
3277 }
3278
3279 static inline void netif_tx_disable(struct net_device *dev)
3280 {
3281 unsigned int i;
3282 int cpu;
3283
3284 local_bh_disable();
3285 cpu = smp_processor_id();
3286 for (i = 0; i < dev->num_tx_queues; i++) {
3287 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3288
3289 __netif_tx_lock(txq, cpu);
3290 netif_tx_stop_queue(txq);
3291 __netif_tx_unlock(txq);
3292 }
3293 local_bh_enable();
3294 }
3295
3296 static inline void netif_addr_lock(struct net_device *dev)
3297 {
3298 spin_lock(&dev->addr_list_lock);
3299 }
3300
3301 static inline void netif_addr_lock_nested(struct net_device *dev)
3302 {
3303 int subclass = SINGLE_DEPTH_NESTING;
3304
3305 if (dev->netdev_ops->ndo_get_lock_subclass)
3306 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3307
3308 spin_lock_nested(&dev->addr_list_lock, subclass);
3309 }
3310
3311 static inline void netif_addr_lock_bh(struct net_device *dev)
3312 {
3313 spin_lock_bh(&dev->addr_list_lock);
3314 }
3315
3316 static inline void netif_addr_unlock(struct net_device *dev)
3317 {
3318 spin_unlock(&dev->addr_list_lock);
3319 }
3320
3321 static inline void netif_addr_unlock_bh(struct net_device *dev)
3322 {
3323 spin_unlock_bh(&dev->addr_list_lock);
3324 }
3325
3326 /*
3327 * dev_addrs walker. Should be used only for read access. Call with
3328 * rcu_read_lock held.
3329 */
3330 #define for_each_dev_addr(dev, ha) \
3331 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3332
3333 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3334
3335 void ether_setup(struct net_device *dev);
3336
3337 /* Support for loadable net-drivers */
3338 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3339 unsigned char name_assign_type,
3340 void (*setup)(struct net_device *),
3341 unsigned int txqs, unsigned int rxqs);
3342 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3343 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3344
3345 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3346 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3347 count)
3348
3349 int register_netdev(struct net_device *dev);
3350 void unregister_netdev(struct net_device *dev);
3351
3352 /* General hardware address lists handling functions */
3353 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3354 struct netdev_hw_addr_list *from_list, int addr_len);
3355 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3356 struct netdev_hw_addr_list *from_list, int addr_len);
3357 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3358 struct net_device *dev,
3359 int (*sync)(struct net_device *, const unsigned char *),
3360 int (*unsync)(struct net_device *,
3361 const unsigned char *));
3362 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3363 struct net_device *dev,
3364 int (*unsync)(struct net_device *,
3365 const unsigned char *));
3366 void __hw_addr_init(struct netdev_hw_addr_list *list);
3367
3368 /* Functions used for device addresses handling */
3369 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3370 unsigned char addr_type);
3371 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3372 unsigned char addr_type);
3373 void dev_addr_flush(struct net_device *dev);
3374 int dev_addr_init(struct net_device *dev);
3375
3376 /* Functions used for unicast addresses handling */
3377 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3378 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3379 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3380 int dev_uc_sync(struct net_device *to, struct net_device *from);
3381 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3382 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3383 void dev_uc_flush(struct net_device *dev);
3384 void dev_uc_init(struct net_device *dev);
3385
3386 /**
3387 * __dev_uc_sync - Synchonize device's unicast list
3388 * @dev: device to sync
3389 * @sync: function to call if address should be added
3390 * @unsync: function to call if address should be removed
3391 *
3392 * Add newly added addresses to the interface, and release
3393 * addresses that have been deleted.
3394 **/
3395 static inline int __dev_uc_sync(struct net_device *dev,
3396 int (*sync)(struct net_device *,
3397 const unsigned char *),
3398 int (*unsync)(struct net_device *,
3399 const unsigned char *))
3400 {
3401 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3402 }
3403
3404 /**
3405 * __dev_uc_unsync - Remove synchronized addresses from device
3406 * @dev: device to sync
3407 * @unsync: function to call if address should be removed
3408 *
3409 * Remove all addresses that were added to the device by dev_uc_sync().
3410 **/
3411 static inline void __dev_uc_unsync(struct net_device *dev,
3412 int (*unsync)(struct net_device *,
3413 const unsigned char *))
3414 {
3415 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3416 }
3417
3418 /* Functions used for multicast addresses handling */
3419 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3420 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3421 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3422 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3423 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3424 int dev_mc_sync(struct net_device *to, struct net_device *from);
3425 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3426 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3427 void dev_mc_flush(struct net_device *dev);
3428 void dev_mc_init(struct net_device *dev);
3429
3430 /**
3431 * __dev_mc_sync - Synchonize device's multicast list
3432 * @dev: device to sync
3433 * @sync: function to call if address should be added
3434 * @unsync: function to call if address should be removed
3435 *
3436 * Add newly added addresses to the interface, and release
3437 * addresses that have been deleted.
3438 **/
3439 static inline int __dev_mc_sync(struct net_device *dev,
3440 int (*sync)(struct net_device *,
3441 const unsigned char *),
3442 int (*unsync)(struct net_device *,
3443 const unsigned char *))
3444 {
3445 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3446 }
3447
3448 /**
3449 * __dev_mc_unsync - Remove synchronized addresses from device
3450 * @dev: device to sync
3451 * @unsync: function to call if address should be removed
3452 *
3453 * Remove all addresses that were added to the device by dev_mc_sync().
3454 **/
3455 static inline void __dev_mc_unsync(struct net_device *dev,
3456 int (*unsync)(struct net_device *,
3457 const unsigned char *))
3458 {
3459 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3460 }
3461
3462 /* Functions used for secondary unicast and multicast support */
3463 void dev_set_rx_mode(struct net_device *dev);
3464 void __dev_set_rx_mode(struct net_device *dev);
3465 int dev_set_promiscuity(struct net_device *dev, int inc);
3466 int dev_set_allmulti(struct net_device *dev, int inc);
3467 void netdev_state_change(struct net_device *dev);
3468 void netdev_notify_peers(struct net_device *dev);
3469 void netdev_features_change(struct net_device *dev);
3470 /* Load a device via the kmod */
3471 void dev_load(struct net *net, const char *name);
3472 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3473 struct rtnl_link_stats64 *storage);
3474 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3475 const struct net_device_stats *netdev_stats);
3476
3477 extern int netdev_max_backlog;
3478 extern int netdev_tstamp_prequeue;
3479 extern int weight_p;
3480 extern int bpf_jit_enable;
3481
3482 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3483 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3484 struct list_head **iter);
3485 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3486 struct list_head **iter);
3487
3488 /* iterate through upper list, must be called under RCU read lock */
3489 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3490 for (iter = &(dev)->adj_list.upper, \
3491 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3492 updev; \
3493 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3494
3495 /* iterate through upper list, must be called under RCU read lock */
3496 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3497 for (iter = &(dev)->all_adj_list.upper, \
3498 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3499 updev; \
3500 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3501
3502 void *netdev_lower_get_next_private(struct net_device *dev,
3503 struct list_head **iter);
3504 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3505 struct list_head **iter);
3506
3507 #define netdev_for_each_lower_private(dev, priv, iter) \
3508 for (iter = (dev)->adj_list.lower.next, \
3509 priv = netdev_lower_get_next_private(dev, &(iter)); \
3510 priv; \
3511 priv = netdev_lower_get_next_private(dev, &(iter)))
3512
3513 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3514 for (iter = &(dev)->adj_list.lower, \
3515 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3516 priv; \
3517 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3518
3519 void *netdev_lower_get_next(struct net_device *dev,
3520 struct list_head **iter);
3521 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3522 for (iter = &(dev)->adj_list.lower, \
3523 ldev = netdev_lower_get_next(dev, &(iter)); \
3524 ldev; \
3525 ldev = netdev_lower_get_next(dev, &(iter)))
3526
3527 void *netdev_adjacent_get_private(struct list_head *adj_list);
3528 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3529 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3530 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3531 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3532 int netdev_master_upper_dev_link(struct net_device *dev,
3533 struct net_device *upper_dev);
3534 int netdev_master_upper_dev_link_private(struct net_device *dev,
3535 struct net_device *upper_dev,
3536 void *private);
3537 void netdev_upper_dev_unlink(struct net_device *dev,
3538 struct net_device *upper_dev);
3539 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3540 void *netdev_lower_dev_get_private(struct net_device *dev,
3541 struct net_device *lower_dev);
3542
3543 /* RSS keys are 40 or 52 bytes long */
3544 #define NETDEV_RSS_KEY_LEN 52
3545 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3546 void netdev_rss_key_fill(void *buffer, size_t len);
3547
3548 int dev_get_nest_level(struct net_device *dev,
3549 bool (*type_check)(struct net_device *dev));
3550 int skb_checksum_help(struct sk_buff *skb);
3551 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3552 netdev_features_t features, bool tx_path);
3553 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3554 netdev_features_t features);
3555
3556 struct netdev_bonding_info {
3557 ifslave slave;
3558 ifbond master;
3559 };
3560
3561 struct netdev_notifier_bonding_info {
3562 struct netdev_notifier_info info; /* must be first */
3563 struct netdev_bonding_info bonding_info;
3564 };
3565
3566 void netdev_bonding_info_change(struct net_device *dev,
3567 struct netdev_bonding_info *bonding_info);
3568
3569 static inline
3570 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3571 {
3572 return __skb_gso_segment(skb, features, true);
3573 }
3574 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3575
3576 static inline bool can_checksum_protocol(netdev_features_t features,
3577 __be16 protocol)
3578 {
3579 return ((features & NETIF_F_GEN_CSUM) ||
3580 ((features & NETIF_F_V4_CSUM) &&
3581 protocol == htons(ETH_P_IP)) ||
3582 ((features & NETIF_F_V6_CSUM) &&
3583 protocol == htons(ETH_P_IPV6)) ||
3584 ((features & NETIF_F_FCOE_CRC) &&
3585 protocol == htons(ETH_P_FCOE)));
3586 }
3587
3588 #ifdef CONFIG_BUG
3589 void netdev_rx_csum_fault(struct net_device *dev);
3590 #else
3591 static inline void netdev_rx_csum_fault(struct net_device *dev)
3592 {
3593 }
3594 #endif
3595 /* rx skb timestamps */
3596 void net_enable_timestamp(void);
3597 void net_disable_timestamp(void);
3598
3599 #ifdef CONFIG_PROC_FS
3600 int __init dev_proc_init(void);
3601 #else
3602 #define dev_proc_init() 0
3603 #endif
3604
3605 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3606 struct sk_buff *skb, struct net_device *dev,
3607 bool more)
3608 {
3609 skb->xmit_more = more ? 1 : 0;
3610 return ops->ndo_start_xmit(skb, dev);
3611 }
3612
3613 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3614 struct netdev_queue *txq, bool more)
3615 {
3616 const struct net_device_ops *ops = dev->netdev_ops;
3617 int rc;
3618
3619 rc = __netdev_start_xmit(ops, skb, dev, more);
3620 if (rc == NETDEV_TX_OK)
3621 txq_trans_update(txq);
3622
3623 return rc;
3624 }
3625
3626 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3627 const void *ns);
3628 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3629 const void *ns);
3630
3631 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3632 {
3633 return netdev_class_create_file_ns(class_attr, NULL);
3634 }
3635
3636 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3637 {
3638 netdev_class_remove_file_ns(class_attr, NULL);
3639 }
3640
3641 extern struct kobj_ns_type_operations net_ns_type_operations;
3642
3643 const char *netdev_drivername(const struct net_device *dev);
3644
3645 void linkwatch_run_queue(void);
3646
3647 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3648 netdev_features_t f2)
3649 {
3650 if (f1 & NETIF_F_GEN_CSUM)
3651 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3652 if (f2 & NETIF_F_GEN_CSUM)
3653 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3654 f1 &= f2;
3655 if (f1 & NETIF_F_GEN_CSUM)
3656 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3657
3658 return f1;
3659 }
3660
3661 static inline netdev_features_t netdev_get_wanted_features(
3662 struct net_device *dev)
3663 {
3664 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3665 }
3666 netdev_features_t netdev_increment_features(netdev_features_t all,
3667 netdev_features_t one, netdev_features_t mask);
3668
3669 /* Allow TSO being used on stacked device :
3670 * Performing the GSO segmentation before last device
3671 * is a performance improvement.
3672 */
3673 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3674 netdev_features_t mask)
3675 {
3676 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3677 }
3678
3679 int __netdev_update_features(struct net_device *dev);
3680 void netdev_update_features(struct net_device *dev);
3681 void netdev_change_features(struct net_device *dev);
3682
3683 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3684 struct net_device *dev);
3685
3686 netdev_features_t passthru_features_check(struct sk_buff *skb,
3687 struct net_device *dev,
3688 netdev_features_t features);
3689 netdev_features_t netif_skb_features(struct sk_buff *skb);
3690
3691 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3692 {
3693 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3694
3695 /* check flags correspondence */
3696 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3697 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3698 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3699 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3700 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3701 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3702 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3703 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3704 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3705 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3706 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3707 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3708 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3709
3710 return (features & feature) == feature;
3711 }
3712
3713 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3714 {
3715 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3716 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3717 }
3718
3719 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3720 netdev_features_t features)
3721 {
3722 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3723 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3724 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3725 }
3726
3727 static inline void netif_set_gso_max_size(struct net_device *dev,
3728 unsigned int size)
3729 {
3730 dev->gso_max_size = size;
3731 }
3732
3733 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3734 int pulled_hlen, u16 mac_offset,
3735 int mac_len)
3736 {
3737 skb->protocol = protocol;
3738 skb->encapsulation = 1;
3739 skb_push(skb, pulled_hlen);
3740 skb_reset_transport_header(skb);
3741 skb->mac_header = mac_offset;
3742 skb->network_header = skb->mac_header + mac_len;
3743 skb->mac_len = mac_len;
3744 }
3745
3746 static inline bool netif_is_macvlan(struct net_device *dev)
3747 {
3748 return dev->priv_flags & IFF_MACVLAN;
3749 }
3750
3751 static inline bool netif_is_macvlan_port(struct net_device *dev)
3752 {
3753 return dev->priv_flags & IFF_MACVLAN_PORT;
3754 }
3755
3756 static inline bool netif_is_ipvlan(struct net_device *dev)
3757 {
3758 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3759 }
3760
3761 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3762 {
3763 return dev->priv_flags & IFF_IPVLAN_MASTER;
3764 }
3765
3766 static inline bool netif_is_bond_master(struct net_device *dev)
3767 {
3768 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3769 }
3770
3771 static inline bool netif_is_bond_slave(struct net_device *dev)
3772 {
3773 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3774 }
3775
3776 static inline bool netif_supports_nofcs(struct net_device *dev)
3777 {
3778 return dev->priv_flags & IFF_SUPP_NOFCS;
3779 }
3780
3781 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3782 static inline void netif_keep_dst(struct net_device *dev)
3783 {
3784 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3785 }
3786
3787 extern struct pernet_operations __net_initdata loopback_net_ops;
3788
3789 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3790
3791 /* netdev_printk helpers, similar to dev_printk */
3792
3793 static inline const char *netdev_name(const struct net_device *dev)
3794 {
3795 if (!dev->name[0] || strchr(dev->name, '%'))
3796 return "(unnamed net_device)";
3797 return dev->name;
3798 }
3799
3800 static inline const char *netdev_reg_state(const struct net_device *dev)
3801 {
3802 switch (dev->reg_state) {
3803 case NETREG_UNINITIALIZED: return " (uninitialized)";
3804 case NETREG_REGISTERED: return "";
3805 case NETREG_UNREGISTERING: return " (unregistering)";
3806 case NETREG_UNREGISTERED: return " (unregistered)";
3807 case NETREG_RELEASED: return " (released)";
3808 case NETREG_DUMMY: return " (dummy)";
3809 }
3810
3811 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3812 return " (unknown)";
3813 }
3814
3815 __printf(3, 4)
3816 void netdev_printk(const char *level, const struct net_device *dev,
3817 const char *format, ...);
3818 __printf(2, 3)
3819 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3820 __printf(2, 3)
3821 void netdev_alert(const struct net_device *dev, const char *format, ...);
3822 __printf(2, 3)
3823 void netdev_crit(const struct net_device *dev, const char *format, ...);
3824 __printf(2, 3)
3825 void netdev_err(const struct net_device *dev, const char *format, ...);
3826 __printf(2, 3)
3827 void netdev_warn(const struct net_device *dev, const char *format, ...);
3828 __printf(2, 3)
3829 void netdev_notice(const struct net_device *dev, const char *format, ...);
3830 __printf(2, 3)
3831 void netdev_info(const struct net_device *dev, const char *format, ...);
3832
3833 #define MODULE_ALIAS_NETDEV(device) \
3834 MODULE_ALIAS("netdev-" device)
3835
3836 #if defined(CONFIG_DYNAMIC_DEBUG)
3837 #define netdev_dbg(__dev, format, args...) \
3838 do { \
3839 dynamic_netdev_dbg(__dev, format, ##args); \
3840 } while (0)
3841 #elif defined(DEBUG)
3842 #define netdev_dbg(__dev, format, args...) \
3843 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3844 #else
3845 #define netdev_dbg(__dev, format, args...) \
3846 ({ \
3847 if (0) \
3848 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3849 })
3850 #endif
3851
3852 #if defined(VERBOSE_DEBUG)
3853 #define netdev_vdbg netdev_dbg
3854 #else
3855
3856 #define netdev_vdbg(dev, format, args...) \
3857 ({ \
3858 if (0) \
3859 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3860 0; \
3861 })
3862 #endif
3863
3864 /*
3865 * netdev_WARN() acts like dev_printk(), but with the key difference
3866 * of using a WARN/WARN_ON to get the message out, including the
3867 * file/line information and a backtrace.
3868 */
3869 #define netdev_WARN(dev, format, args...) \
3870 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3871 netdev_reg_state(dev), ##args)
3872
3873 /* netif printk helpers, similar to netdev_printk */
3874
3875 #define netif_printk(priv, type, level, dev, fmt, args...) \
3876 do { \
3877 if (netif_msg_##type(priv)) \
3878 netdev_printk(level, (dev), fmt, ##args); \
3879 } while (0)
3880
3881 #define netif_level(level, priv, type, dev, fmt, args...) \
3882 do { \
3883 if (netif_msg_##type(priv)) \
3884 netdev_##level(dev, fmt, ##args); \
3885 } while (0)
3886
3887 #define netif_emerg(priv, type, dev, fmt, args...) \
3888 netif_level(emerg, priv, type, dev, fmt, ##args)
3889 #define netif_alert(priv, type, dev, fmt, args...) \
3890 netif_level(alert, priv, type, dev, fmt, ##args)
3891 #define netif_crit(priv, type, dev, fmt, args...) \
3892 netif_level(crit, priv, type, dev, fmt, ##args)
3893 #define netif_err(priv, type, dev, fmt, args...) \
3894 netif_level(err, priv, type, dev, fmt, ##args)
3895 #define netif_warn(priv, type, dev, fmt, args...) \
3896 netif_level(warn, priv, type, dev, fmt, ##args)
3897 #define netif_notice(priv, type, dev, fmt, args...) \
3898 netif_level(notice, priv, type, dev, fmt, ##args)
3899 #define netif_info(priv, type, dev, fmt, args...) \
3900 netif_level(info, priv, type, dev, fmt, ##args)
3901
3902 #if defined(CONFIG_DYNAMIC_DEBUG)
3903 #define netif_dbg(priv, type, netdev, format, args...) \
3904 do { \
3905 if (netif_msg_##type(priv)) \
3906 dynamic_netdev_dbg(netdev, format, ##args); \
3907 } while (0)
3908 #elif defined(DEBUG)
3909 #define netif_dbg(priv, type, dev, format, args...) \
3910 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3911 #else
3912 #define netif_dbg(priv, type, dev, format, args...) \
3913 ({ \
3914 if (0) \
3915 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3916 0; \
3917 })
3918 #endif
3919
3920 #if defined(VERBOSE_DEBUG)
3921 #define netif_vdbg netif_dbg
3922 #else
3923 #define netif_vdbg(priv, type, dev, format, args...) \
3924 ({ \
3925 if (0) \
3926 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3927 0; \
3928 })
3929 #endif
3930
3931 /*
3932 * The list of packet types we will receive (as opposed to discard)
3933 * and the routines to invoke.
3934 *
3935 * Why 16. Because with 16 the only overlap we get on a hash of the
3936 * low nibble of the protocol value is RARP/SNAP/X.25.
3937 *
3938 * NOTE: That is no longer true with the addition of VLAN tags. Not
3939 * sure which should go first, but I bet it won't make much
3940 * difference if we are running VLANs. The good news is that
3941 * this protocol won't be in the list unless compiled in, so
3942 * the average user (w/out VLANs) will not be adversely affected.
3943 * --BLG
3944 *
3945 * 0800 IP
3946 * 8100 802.1Q VLAN
3947 * 0001 802.3
3948 * 0002 AX.25
3949 * 0004 802.2
3950 * 8035 RARP
3951 * 0005 SNAP
3952 * 0805 X.25
3953 * 0806 ARP
3954 * 8137 IPX
3955 * 0009 Localtalk
3956 * 86DD IPv6
3957 */
3958 #define PTYPE_HASH_SIZE (16)
3959 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3960
3961 #endif /* _LINUX_NETDEVICE_H */
This page took 0.112845 seconds and 4 git commands to generate.