ddab1a2a07a0bc11f840f81bd36ad38adcdd0012
[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 } ____cacheline_aligned_in_smp;
591
592 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
593 {
594 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
595 return q->numa_node;
596 #else
597 return NUMA_NO_NODE;
598 #endif
599 }
600
601 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
602 {
603 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
604 q->numa_node = node;
605 #endif
606 }
607
608 #ifdef CONFIG_RPS
609 /*
610 * This structure holds an RPS map which can be of variable length. The
611 * map is an array of CPUs.
612 */
613 struct rps_map {
614 unsigned int len;
615 struct rcu_head rcu;
616 u16 cpus[0];
617 };
618 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
619
620 /*
621 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
622 * tail pointer for that CPU's input queue at the time of last enqueue, and
623 * a hardware filter index.
624 */
625 struct rps_dev_flow {
626 u16 cpu;
627 u16 filter;
628 unsigned int last_qtail;
629 };
630 #define RPS_NO_FILTER 0xffff
631
632 /*
633 * The rps_dev_flow_table structure contains a table of flow mappings.
634 */
635 struct rps_dev_flow_table {
636 unsigned int mask;
637 struct rcu_head rcu;
638 struct rps_dev_flow flows[0];
639 };
640 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
641 ((_num) * sizeof(struct rps_dev_flow)))
642
643 /*
644 * The rps_sock_flow_table contains mappings of flows to the last CPU
645 * on which they were processed by the application (set in recvmsg).
646 * Each entry is a 32bit value. Upper part is the high order bits
647 * of flow hash, lower part is cpu number.
648 * rps_cpu_mask is used to partition the space, depending on number of
649 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
650 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
651 * meaning we use 32-6=26 bits for the hash.
652 */
653 struct rps_sock_flow_table {
654 u32 mask;
655
656 u32 ents[0] ____cacheline_aligned_in_smp;
657 };
658 #define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
659
660 #define RPS_NO_CPU 0xffff
661
662 extern u32 rps_cpu_mask;
663 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
664
665 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
666 u32 hash)
667 {
668 if (table && hash) {
669 unsigned int index = hash & table->mask;
670 u32 val = hash & ~rps_cpu_mask;
671
672 /* We only give a hint, preemption can change cpu under us */
673 val |= raw_smp_processor_id();
674
675 if (table->ents[index] != val)
676 table->ents[index] = val;
677 }
678 }
679
680 #ifdef CONFIG_RFS_ACCEL
681 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
682 u16 filter_id);
683 #endif
684 #endif /* CONFIG_RPS */
685
686 /* This structure contains an instance of an RX queue. */
687 struct netdev_rx_queue {
688 #ifdef CONFIG_RPS
689 struct rps_map __rcu *rps_map;
690 struct rps_dev_flow_table __rcu *rps_flow_table;
691 #endif
692 struct kobject kobj;
693 struct net_device *dev;
694 } ____cacheline_aligned_in_smp;
695
696 /*
697 * RX queue sysfs structures and functions.
698 */
699 struct rx_queue_attribute {
700 struct attribute attr;
701 ssize_t (*show)(struct netdev_rx_queue *queue,
702 struct rx_queue_attribute *attr, char *buf);
703 ssize_t (*store)(struct netdev_rx_queue *queue,
704 struct rx_queue_attribute *attr, const char *buf, size_t len);
705 };
706
707 #ifdef CONFIG_XPS
708 /*
709 * This structure holds an XPS map which can be of variable length. The
710 * map is an array of queues.
711 */
712 struct xps_map {
713 unsigned int len;
714 unsigned int alloc_len;
715 struct rcu_head rcu;
716 u16 queues[0];
717 };
718 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
719 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_BYTES - sizeof(struct xps_map)) \
720 / sizeof(u16))
721
722 /*
723 * This structure holds all XPS maps for device. Maps are indexed by CPU.
724 */
725 struct xps_dev_maps {
726 struct rcu_head rcu;
727 struct xps_map __rcu *cpu_map[0];
728 };
729 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
730 (nr_cpu_ids * sizeof(struct xps_map *)))
731 #endif /* CONFIG_XPS */
732
733 #define TC_MAX_QUEUE 16
734 #define TC_BITMASK 15
735 /* HW offloaded queuing disciplines txq count and offset maps */
736 struct netdev_tc_txq {
737 u16 count;
738 u16 offset;
739 };
740
741 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
742 /*
743 * This structure is to hold information about the device
744 * configured to run FCoE protocol stack.
745 */
746 struct netdev_fcoe_hbainfo {
747 char manufacturer[64];
748 char serial_number[64];
749 char hardware_version[64];
750 char driver_version[64];
751 char optionrom_version[64];
752 char firmware_version[64];
753 char model[256];
754 char model_description[256];
755 };
756 #endif
757
758 #define MAX_PHYS_ITEM_ID_LEN 32
759
760 /* This structure holds a unique identifier to identify some
761 * physical item (port for example) used by a netdevice.
762 */
763 struct netdev_phys_item_id {
764 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
765 unsigned char id_len;
766 };
767
768 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
769 struct sk_buff *skb);
770
771 struct fib_info;
772
773 /*
774 * This structure defines the management hooks for network devices.
775 * The following hooks can be defined; unless noted otherwise, they are
776 * optional and can be filled with a null pointer.
777 *
778 * int (*ndo_init)(struct net_device *dev);
779 * This function is called once when network device is registered.
780 * The network device can use this to any late stage initializaton
781 * or semantic validattion. It can fail with an error code which will
782 * be propogated back to register_netdev
783 *
784 * void (*ndo_uninit)(struct net_device *dev);
785 * This function is called when device is unregistered or when registration
786 * fails. It is not called if init fails.
787 *
788 * int (*ndo_open)(struct net_device *dev);
789 * This function is called when network device transistions to the up
790 * state.
791 *
792 * int (*ndo_stop)(struct net_device *dev);
793 * This function is called when network device transistions to the down
794 * state.
795 *
796 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
797 * struct net_device *dev);
798 * Called when a packet needs to be transmitted.
799 * Must return NETDEV_TX_OK , NETDEV_TX_BUSY.
800 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
801 * Required can not be NULL.
802 *
803 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
804 * void *accel_priv, select_queue_fallback_t fallback);
805 * Called to decide which queue to when device supports multiple
806 * transmit queues.
807 *
808 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
809 * This function is called to allow device receiver to make
810 * changes to configuration when multicast or promiscious is enabled.
811 *
812 * void (*ndo_set_rx_mode)(struct net_device *dev);
813 * This function is called device changes address list filtering.
814 * If driver handles unicast address filtering, it should set
815 * IFF_UNICAST_FLT to its priv_flags.
816 *
817 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
818 * This function is called when the Media Access Control address
819 * needs to be changed. If this interface is not defined, the
820 * mac address can not be changed.
821 *
822 * int (*ndo_validate_addr)(struct net_device *dev);
823 * Test if Media Access Control address is valid for the device.
824 *
825 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
826 * Called when a user request an ioctl which can't be handled by
827 * the generic interface code. If not defined ioctl's return
828 * not supported error code.
829 *
830 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
831 * Used to set network devices bus interface parameters. This interface
832 * is retained for legacy reason, new devices should use the bus
833 * interface (PCI) for low level management.
834 *
835 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
836 * Called when a user wants to change the Maximum Transfer Unit
837 * of a device. If not defined, any request to change MTU will
838 * will return an error.
839 *
840 * void (*ndo_tx_timeout)(struct net_device *dev);
841 * Callback uses when the transmitter has not made any progress
842 * for dev->watchdog ticks.
843 *
844 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
845 * struct rtnl_link_stats64 *storage);
846 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
847 * Called when a user wants to get the network device usage
848 * statistics. Drivers must do one of the following:
849 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
850 * rtnl_link_stats64 structure passed by the caller.
851 * 2. Define @ndo_get_stats to update a net_device_stats structure
852 * (which should normally be dev->stats) and return a pointer to
853 * it. The structure may be changed asynchronously only if each
854 * field is written atomically.
855 * 3. Update dev->stats asynchronously and atomically, and define
856 * neither operation.
857 *
858 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
859 * If device support VLAN filtering this function is called when a
860 * VLAN id is registered.
861 *
862 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
863 * If device support VLAN filtering this function is called when a
864 * VLAN id is unregistered.
865 *
866 * void (*ndo_poll_controller)(struct net_device *dev);
867 *
868 * SR-IOV management functions.
869 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
870 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
871 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
872 * int max_tx_rate);
873 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
874 * int (*ndo_get_vf_config)(struct net_device *dev,
875 * int vf, struct ifla_vf_info *ivf);
876 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
877 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
878 * struct nlattr *port[]);
879 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
880 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
881 * Called to setup 'tc' number of traffic classes in the net device. This
882 * is always called from the stack with the rtnl lock held and netif tx
883 * queues stopped. This allows the netdevice to perform queue management
884 * safely.
885 *
886 * Fiber Channel over Ethernet (FCoE) offload functions.
887 * int (*ndo_fcoe_enable)(struct net_device *dev);
888 * Called when the FCoE protocol stack wants to start using LLD for FCoE
889 * so the underlying device can perform whatever needed configuration or
890 * initialization to support acceleration of FCoE traffic.
891 *
892 * int (*ndo_fcoe_disable)(struct net_device *dev);
893 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
894 * so the underlying device can perform whatever needed clean-ups to
895 * stop supporting acceleration of FCoE traffic.
896 *
897 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
898 * struct scatterlist *sgl, unsigned int sgc);
899 * Called when the FCoE Initiator wants to initialize an I/O that
900 * is a possible candidate for Direct Data Placement (DDP). The LLD can
901 * perform necessary setup and returns 1 to indicate the device is set up
902 * successfully to perform DDP on this I/O, otherwise this returns 0.
903 *
904 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
905 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
906 * indicated by the FC exchange id 'xid', so the underlying device can
907 * clean up and reuse resources for later DDP requests.
908 *
909 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
910 * struct scatterlist *sgl, unsigned int sgc);
911 * Called when the FCoE Target wants to initialize an I/O that
912 * is a possible candidate for Direct Data Placement (DDP). The LLD can
913 * perform necessary setup and returns 1 to indicate the device is set up
914 * successfully to perform DDP on this I/O, otherwise this returns 0.
915 *
916 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
917 * struct netdev_fcoe_hbainfo *hbainfo);
918 * Called when the FCoE Protocol stack wants information on the underlying
919 * device. This information is utilized by the FCoE protocol stack to
920 * register attributes with Fiber Channel management service as per the
921 * FC-GS Fabric Device Management Information(FDMI) specification.
922 *
923 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
924 * Called when the underlying device wants to override default World Wide
925 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
926 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
927 * protocol stack to use.
928 *
929 * RFS acceleration.
930 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
931 * u16 rxq_index, u32 flow_id);
932 * Set hardware filter for RFS. rxq_index is the target queue index;
933 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
934 * Return the filter ID on success, or a negative error code.
935 *
936 * Slave management functions (for bridge, bonding, etc).
937 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
938 * Called to make another netdev an underling.
939 *
940 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
941 * Called to release previously enslaved netdev.
942 *
943 * Feature/offload setting functions.
944 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
945 * netdev_features_t features);
946 * Adjusts the requested feature flags according to device-specific
947 * constraints, and returns the resulting flags. Must not modify
948 * the device state.
949 *
950 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
951 * Called to update device configuration to new features. Passed
952 * feature set might be less than what was returned by ndo_fix_features()).
953 * Must return >0 or -errno if it changed dev->features itself.
954 *
955 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
956 * struct net_device *dev,
957 * const unsigned char *addr, u16 vid, u16 flags)
958 * Adds an FDB entry to dev for addr.
959 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
960 * struct net_device *dev,
961 * const unsigned char *addr, u16 vid)
962 * Deletes the FDB entry from dev coresponding to addr.
963 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
964 * struct net_device *dev, struct net_device *filter_dev,
965 * int idx)
966 * Used to add FDB entries to dump requests. Implementers should add
967 * entries to skb and update idx with the number of entries.
968 *
969 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh)
970 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
971 * struct net_device *dev, u32 filter_mask)
972 *
973 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
974 * Called to change device carrier. Soft-devices (like dummy, team, etc)
975 * which do not represent real hardware may define this to allow their
976 * userspace components to manage their virtual carrier state. Devices
977 * that determine carrier state from physical hardware properties (eg
978 * network cables) or protocol-dependent mechanisms (eg
979 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
980 *
981 * int (*ndo_get_phys_port_id)(struct net_device *dev,
982 * struct netdev_phys_item_id *ppid);
983 * Called to get ID of physical port of this device. If driver does
984 * not implement this, it is assumed that the hw is not able to have
985 * multiple net devices on single physical port.
986 *
987 * void (*ndo_add_vxlan_port)(struct net_device *dev,
988 * sa_family_t sa_family, __be16 port);
989 * Called by vxlan to notiy a driver about the UDP port and socket
990 * address family that vxlan is listnening to. It is called only when
991 * a new port starts listening. The operation is protected by the
992 * vxlan_net->sock_lock.
993 *
994 * void (*ndo_del_vxlan_port)(struct net_device *dev,
995 * sa_family_t sa_family, __be16 port);
996 * Called by vxlan to notify the driver about a UDP port and socket
997 * address family that vxlan is not listening to anymore. The operation
998 * is protected by the vxlan_net->sock_lock.
999 *
1000 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1001 * struct net_device *dev)
1002 * Called by upper layer devices to accelerate switching or other
1003 * station functionality into hardware. 'pdev is the lowerdev
1004 * to use for the offload and 'dev' is the net device that will
1005 * back the offload. Returns a pointer to the private structure
1006 * the upper layer will maintain.
1007 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1008 * Called by upper layer device to delete the station created
1009 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1010 * the station and priv is the structure returned by the add
1011 * operation.
1012 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1013 * struct net_device *dev,
1014 * void *priv);
1015 * Callback to use for xmit over the accelerated station. This
1016 * is used in place of ndo_start_xmit on accelerated net
1017 * devices.
1018 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1019 * struct net_device *dev
1020 * netdev_features_t features);
1021 * Called by core transmit path to determine if device is capable of
1022 * performing offload operations on a given packet. This is to give
1023 * the device an opportunity to implement any restrictions that cannot
1024 * be otherwise expressed by feature flags. The check is called with
1025 * the set of features that the stack has calculated and it returns
1026 * those the driver believes to be appropriate.
1027 *
1028 * int (*ndo_switch_parent_id_get)(struct net_device *dev,
1029 * struct netdev_phys_item_id *psid);
1030 * Called to get an ID of the switch chip this port is part of.
1031 * If driver implements this, it indicates that it represents a port
1032 * of a switch chip.
1033 * int (*ndo_switch_port_stp_update)(struct net_device *dev, u8 state);
1034 * Called to notify switch device port of bridge port STP
1035 * state change.
1036 * int (*ndo_sw_parent_fib_ipv4_add)(struct net_device *dev, __be32 dst,
1037 * int dst_len, struct fib_info *fi,
1038 * u8 tos, u8 type, u32 nlflags, u32 tb_id);
1039 * Called to add/modify IPv4 route to switch device.
1040 * int (*ndo_sw_parent_fib_ipv4_del)(struct net_device *dev, __be32 dst,
1041 * int dst_len, struct fib_info *fi,
1042 * u8 tos, u8 type, u32 tb_id);
1043 * Called to delete IPv4 route from switch device.
1044 */
1045 struct net_device_ops {
1046 int (*ndo_init)(struct net_device *dev);
1047 void (*ndo_uninit)(struct net_device *dev);
1048 int (*ndo_open)(struct net_device *dev);
1049 int (*ndo_stop)(struct net_device *dev);
1050 netdev_tx_t (*ndo_start_xmit) (struct sk_buff *skb,
1051 struct net_device *dev);
1052 u16 (*ndo_select_queue)(struct net_device *dev,
1053 struct sk_buff *skb,
1054 void *accel_priv,
1055 select_queue_fallback_t fallback);
1056 void (*ndo_change_rx_flags)(struct net_device *dev,
1057 int flags);
1058 void (*ndo_set_rx_mode)(struct net_device *dev);
1059 int (*ndo_set_mac_address)(struct net_device *dev,
1060 void *addr);
1061 int (*ndo_validate_addr)(struct net_device *dev);
1062 int (*ndo_do_ioctl)(struct net_device *dev,
1063 struct ifreq *ifr, int cmd);
1064 int (*ndo_set_config)(struct net_device *dev,
1065 struct ifmap *map);
1066 int (*ndo_change_mtu)(struct net_device *dev,
1067 int new_mtu);
1068 int (*ndo_neigh_setup)(struct net_device *dev,
1069 struct neigh_parms *);
1070 void (*ndo_tx_timeout) (struct net_device *dev);
1071
1072 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1073 struct rtnl_link_stats64 *storage);
1074 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1075
1076 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
1077 __be16 proto, u16 vid);
1078 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1079 __be16 proto, u16 vid);
1080 #ifdef CONFIG_NET_POLL_CONTROLLER
1081 void (*ndo_poll_controller)(struct net_device *dev);
1082 int (*ndo_netpoll_setup)(struct net_device *dev,
1083 struct netpoll_info *info);
1084 void (*ndo_netpoll_cleanup)(struct net_device *dev);
1085 #endif
1086 #ifdef CONFIG_NET_RX_BUSY_POLL
1087 int (*ndo_busy_poll)(struct napi_struct *dev);
1088 #endif
1089 int (*ndo_set_vf_mac)(struct net_device *dev,
1090 int queue, u8 *mac);
1091 int (*ndo_set_vf_vlan)(struct net_device *dev,
1092 int queue, u16 vlan, u8 qos);
1093 int (*ndo_set_vf_rate)(struct net_device *dev,
1094 int vf, int min_tx_rate,
1095 int max_tx_rate);
1096 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1097 int vf, bool setting);
1098 int (*ndo_get_vf_config)(struct net_device *dev,
1099 int vf,
1100 struct ifla_vf_info *ivf);
1101 int (*ndo_set_vf_link_state)(struct net_device *dev,
1102 int vf, int link_state);
1103 int (*ndo_set_vf_port)(struct net_device *dev,
1104 int vf,
1105 struct nlattr *port[]);
1106 int (*ndo_get_vf_port)(struct net_device *dev,
1107 int vf, struct sk_buff *skb);
1108 int (*ndo_setup_tc)(struct net_device *dev, u8 tc);
1109 #if IS_ENABLED(CONFIG_FCOE)
1110 int (*ndo_fcoe_enable)(struct net_device *dev);
1111 int (*ndo_fcoe_disable)(struct net_device *dev);
1112 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1113 u16 xid,
1114 struct scatterlist *sgl,
1115 unsigned int sgc);
1116 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1117 u16 xid);
1118 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1119 u16 xid,
1120 struct scatterlist *sgl,
1121 unsigned int sgc);
1122 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1123 struct netdev_fcoe_hbainfo *hbainfo);
1124 #endif
1125
1126 #if IS_ENABLED(CONFIG_LIBFCOE)
1127 #define NETDEV_FCOE_WWNN 0
1128 #define NETDEV_FCOE_WWPN 1
1129 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1130 u64 *wwn, int type);
1131 #endif
1132
1133 #ifdef CONFIG_RFS_ACCEL
1134 int (*ndo_rx_flow_steer)(struct net_device *dev,
1135 const struct sk_buff *skb,
1136 u16 rxq_index,
1137 u32 flow_id);
1138 #endif
1139 int (*ndo_add_slave)(struct net_device *dev,
1140 struct net_device *slave_dev);
1141 int (*ndo_del_slave)(struct net_device *dev,
1142 struct net_device *slave_dev);
1143 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1144 netdev_features_t features);
1145 int (*ndo_set_features)(struct net_device *dev,
1146 netdev_features_t features);
1147 int (*ndo_neigh_construct)(struct neighbour *n);
1148 void (*ndo_neigh_destroy)(struct neighbour *n);
1149
1150 int (*ndo_fdb_add)(struct ndmsg *ndm,
1151 struct nlattr *tb[],
1152 struct net_device *dev,
1153 const unsigned char *addr,
1154 u16 vid,
1155 u16 flags);
1156 int (*ndo_fdb_del)(struct ndmsg *ndm,
1157 struct nlattr *tb[],
1158 struct net_device *dev,
1159 const unsigned char *addr,
1160 u16 vid);
1161 int (*ndo_fdb_dump)(struct sk_buff *skb,
1162 struct netlink_callback *cb,
1163 struct net_device *dev,
1164 struct net_device *filter_dev,
1165 int idx);
1166
1167 int (*ndo_bridge_setlink)(struct net_device *dev,
1168 struct nlmsghdr *nlh,
1169 u16 flags);
1170 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1171 u32 pid, u32 seq,
1172 struct net_device *dev,
1173 u32 filter_mask);
1174 int (*ndo_bridge_dellink)(struct net_device *dev,
1175 struct nlmsghdr *nlh,
1176 u16 flags);
1177 int (*ndo_change_carrier)(struct net_device *dev,
1178 bool new_carrier);
1179 int (*ndo_get_phys_port_id)(struct net_device *dev,
1180 struct netdev_phys_item_id *ppid);
1181 void (*ndo_add_vxlan_port)(struct net_device *dev,
1182 sa_family_t sa_family,
1183 __be16 port);
1184 void (*ndo_del_vxlan_port)(struct net_device *dev,
1185 sa_family_t sa_family,
1186 __be16 port);
1187
1188 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1189 struct net_device *dev);
1190 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1191 void *priv);
1192
1193 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1194 struct net_device *dev,
1195 void *priv);
1196 int (*ndo_get_lock_subclass)(struct net_device *dev);
1197 netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1198 struct net_device *dev,
1199 netdev_features_t features);
1200 #ifdef CONFIG_NET_SWITCHDEV
1201 int (*ndo_switch_parent_id_get)(struct net_device *dev,
1202 struct netdev_phys_item_id *psid);
1203 int (*ndo_switch_port_stp_update)(struct net_device *dev,
1204 u8 state);
1205 int (*ndo_switch_fib_ipv4_add)(struct net_device *dev,
1206 __be32 dst,
1207 int dst_len,
1208 struct fib_info *fi,
1209 u8 tos, u8 type,
1210 u32 nlflags,
1211 u32 tb_id);
1212 int (*ndo_switch_fib_ipv4_del)(struct net_device *dev,
1213 __be32 dst,
1214 int dst_len,
1215 struct fib_info *fi,
1216 u8 tos, u8 type,
1217 u32 tb_id);
1218 #endif
1219 };
1220
1221 /**
1222 * enum net_device_priv_flags - &struct net_device priv_flags
1223 *
1224 * These are the &struct net_device, they are only set internally
1225 * by drivers and used in the kernel. These flags are invisible to
1226 * userspace, this means that the order of these flags can change
1227 * during any kernel release.
1228 *
1229 * You should have a pretty good reason to be extending these flags.
1230 *
1231 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1232 * @IFF_EBRIDGE: Ethernet bridging device
1233 * @IFF_SLAVE_INACTIVE: bonding slave not the curr. active
1234 * @IFF_MASTER_8023AD: bonding master, 802.3ad
1235 * @IFF_MASTER_ALB: bonding master, balance-alb
1236 * @IFF_BONDING: bonding master or slave
1237 * @IFF_SLAVE_NEEDARP: need ARPs for validation
1238 * @IFF_ISATAP: ISATAP interface (RFC4214)
1239 * @IFF_MASTER_ARPMON: bonding master, ARP mon in use
1240 * @IFF_WAN_HDLC: WAN HDLC device
1241 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1242 * release skb->dst
1243 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1244 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1245 * @IFF_MACVLAN_PORT: device used as macvlan port
1246 * @IFF_BRIDGE_PORT: device used as bridge port
1247 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1248 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1249 * @IFF_UNICAST_FLT: Supports unicast filtering
1250 * @IFF_TEAM_PORT: device used as team port
1251 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1252 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1253 * change when it's running
1254 * @IFF_MACVLAN: Macvlan device
1255 */
1256 enum netdev_priv_flags {
1257 IFF_802_1Q_VLAN = 1<<0,
1258 IFF_EBRIDGE = 1<<1,
1259 IFF_SLAVE_INACTIVE = 1<<2,
1260 IFF_MASTER_8023AD = 1<<3,
1261 IFF_MASTER_ALB = 1<<4,
1262 IFF_BONDING = 1<<5,
1263 IFF_SLAVE_NEEDARP = 1<<6,
1264 IFF_ISATAP = 1<<7,
1265 IFF_MASTER_ARPMON = 1<<8,
1266 IFF_WAN_HDLC = 1<<9,
1267 IFF_XMIT_DST_RELEASE = 1<<10,
1268 IFF_DONT_BRIDGE = 1<<11,
1269 IFF_DISABLE_NETPOLL = 1<<12,
1270 IFF_MACVLAN_PORT = 1<<13,
1271 IFF_BRIDGE_PORT = 1<<14,
1272 IFF_OVS_DATAPATH = 1<<15,
1273 IFF_TX_SKB_SHARING = 1<<16,
1274 IFF_UNICAST_FLT = 1<<17,
1275 IFF_TEAM_PORT = 1<<18,
1276 IFF_SUPP_NOFCS = 1<<19,
1277 IFF_LIVE_ADDR_CHANGE = 1<<20,
1278 IFF_MACVLAN = 1<<21,
1279 IFF_XMIT_DST_RELEASE_PERM = 1<<22,
1280 IFF_IPVLAN_MASTER = 1<<23,
1281 IFF_IPVLAN_SLAVE = 1<<24,
1282 };
1283
1284 #define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1285 #define IFF_EBRIDGE IFF_EBRIDGE
1286 #define IFF_SLAVE_INACTIVE IFF_SLAVE_INACTIVE
1287 #define IFF_MASTER_8023AD IFF_MASTER_8023AD
1288 #define IFF_MASTER_ALB IFF_MASTER_ALB
1289 #define IFF_BONDING IFF_BONDING
1290 #define IFF_SLAVE_NEEDARP IFF_SLAVE_NEEDARP
1291 #define IFF_ISATAP IFF_ISATAP
1292 #define IFF_MASTER_ARPMON IFF_MASTER_ARPMON
1293 #define IFF_WAN_HDLC IFF_WAN_HDLC
1294 #define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1295 #define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1296 #define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1297 #define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1298 #define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1299 #define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1300 #define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1301 #define IFF_UNICAST_FLT IFF_UNICAST_FLT
1302 #define IFF_TEAM_PORT IFF_TEAM_PORT
1303 #define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1304 #define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1305 #define IFF_MACVLAN IFF_MACVLAN
1306 #define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
1307 #define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1308 #define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
1309
1310 /**
1311 * struct net_device - The DEVICE structure.
1312 * Actually, this whole structure is a big mistake. It mixes I/O
1313 * data with strictly "high-level" data, and it has to know about
1314 * almost every data structure used in the INET module.
1315 *
1316 * @name: This is the first field of the "visible" part of this structure
1317 * (i.e. as seen by users in the "Space.c" file). It is the name
1318 * of the interface.
1319 *
1320 * @name_hlist: Device name hash chain, please keep it close to name[]
1321 * @ifalias: SNMP alias
1322 * @mem_end: Shared memory end
1323 * @mem_start: Shared memory start
1324 * @base_addr: Device I/O address
1325 * @irq: Device IRQ number
1326 *
1327 * @state: Generic network queuing layer state, see netdev_state_t
1328 * @dev_list: The global list of network devices
1329 * @napi_list: List entry, that is used for polling napi devices
1330 * @unreg_list: List entry, that is used, when we are unregistering the
1331 * device, see the function unregister_netdev
1332 * @close_list: List entry, that is used, when we are closing the device
1333 *
1334 * @adj_list: Directly linked devices, like slaves for bonding
1335 * @all_adj_list: All linked devices, *including* neighbours
1336 * @features: Currently active device features
1337 * @hw_features: User-changeable features
1338 *
1339 * @wanted_features: User-requested features
1340 * @vlan_features: Mask of features inheritable by VLAN devices
1341 *
1342 * @hw_enc_features: Mask of features inherited by encapsulating devices
1343 * This field indicates what encapsulation
1344 * offloads the hardware is capable of doing,
1345 * and drivers will need to set them appropriately.
1346 *
1347 * @mpls_features: Mask of features inheritable by MPLS
1348 *
1349 * @ifindex: interface index
1350 * @iflink: unique device identifier
1351 *
1352 * @stats: Statistics struct, which was left as a legacy, use
1353 * rtnl_link_stats64 instead
1354 *
1355 * @rx_dropped: Dropped packets by core network,
1356 * do not use this in drivers
1357 * @tx_dropped: Dropped packets by core network,
1358 * do not use this in drivers
1359 *
1360 * @carrier_changes: Stats to monitor carrier on<->off transitions
1361 *
1362 * @wireless_handlers: List of functions to handle Wireless Extensions,
1363 * instead of ioctl,
1364 * see <net/iw_handler.h> for details.
1365 * @wireless_data: Instance data managed by the core of wireless extensions
1366 *
1367 * @netdev_ops: Includes several pointers to callbacks,
1368 * if one wants to override the ndo_*() functions
1369 * @ethtool_ops: Management operations
1370 * @fwd_ops: Management operations
1371 * @header_ops: Includes callbacks for creating,parsing,caching,etc
1372 * of Layer 2 headers.
1373 *
1374 * @flags: Interface flags (a la BSD)
1375 * @priv_flags: Like 'flags' but invisible to userspace,
1376 * see if.h for the definitions
1377 * @gflags: Global flags ( kept as legacy )
1378 * @padded: How much padding added by alloc_netdev()
1379 * @operstate: RFC2863 operstate
1380 * @link_mode: Mapping policy to operstate
1381 * @if_port: Selectable AUI, TP, ...
1382 * @dma: DMA channel
1383 * @mtu: Interface MTU value
1384 * @type: Interface hardware type
1385 * @hard_header_len: Hardware header length
1386 *
1387 * @needed_headroom: Extra headroom the hardware may need, but not in all
1388 * cases can this be guaranteed
1389 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1390 * cases can this be guaranteed. Some cases also use
1391 * LL_MAX_HEADER instead to allocate the skb
1392 *
1393 * interface address info:
1394 *
1395 * @perm_addr: Permanent hw address
1396 * @addr_assign_type: Hw address assignment type
1397 * @addr_len: Hardware address length
1398 * @neigh_priv_len; Used in neigh_alloc(),
1399 * initialized only in atm/clip.c
1400 * @dev_id: Used to differentiate devices that share
1401 * the same link layer address
1402 * @dev_port: Used to differentiate devices that share
1403 * the same function
1404 * @addr_list_lock: XXX: need comments on this one
1405 * @uc: unicast mac addresses
1406 * @mc: multicast mac addresses
1407 * @dev_addrs: list of device hw addresses
1408 * @queues_kset: Group of all Kobjects in the Tx and RX queues
1409 * @uc_promisc: Counter, that indicates, that promiscuous mode
1410 * has been enabled due to the need to listen to
1411 * additional unicast addresses in a device that
1412 * does not implement ndo_set_rx_mode()
1413 * @promiscuity: Number of times, the NIC is told to work in
1414 * Promiscuous mode, if it becomes 0 the NIC will
1415 * exit from working in Promiscuous mode
1416 * @allmulti: Counter, enables or disables allmulticast mode
1417 *
1418 * @vlan_info: VLAN info
1419 * @dsa_ptr: dsa specific data
1420 * @tipc_ptr: TIPC specific data
1421 * @atalk_ptr: AppleTalk link
1422 * @ip_ptr: IPv4 specific data
1423 * @dn_ptr: DECnet specific data
1424 * @ip6_ptr: IPv6 specific data
1425 * @ax25_ptr: AX.25 specific data
1426 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1427 *
1428 * @last_rx: Time of last Rx
1429 * @dev_addr: Hw address (before bcast,
1430 * because most packets are unicast)
1431 *
1432 * @_rx: Array of RX queues
1433 * @num_rx_queues: Number of RX queues
1434 * allocated at register_netdev() time
1435 * @real_num_rx_queues: Number of RX queues currently active in device
1436 *
1437 * @rx_handler: handler for received packets
1438 * @rx_handler_data: XXX: need comments on this one
1439 * @ingress_queue: XXX: need comments on this one
1440 * @broadcast: hw bcast address
1441 *
1442 * @_tx: Array of TX queues
1443 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1444 * @real_num_tx_queues: Number of TX queues currently active in device
1445 * @qdisc: Root qdisc from userspace point of view
1446 * @tx_queue_len: Max frames per queue allowed
1447 * @tx_global_lock: XXX: need comments on this one
1448 *
1449 * @xps_maps: XXX: need comments on this one
1450 *
1451 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1452 * indexed by RX queue number. Assigned by driver.
1453 * This must only be set if the ndo_rx_flow_steer
1454 * operation is defined
1455 *
1456 * @trans_start: Time (in jiffies) of last Tx
1457 * @watchdog_timeo: Represents the timeout that is used by
1458 * the watchdog ( see dev_watchdog() )
1459 * @watchdog_timer: List of timers
1460 *
1461 * @pcpu_refcnt: Number of references to this device
1462 * @todo_list: Delayed register/unregister
1463 * @index_hlist: Device index hash chain
1464 * @link_watch_list: XXX: need comments on this one
1465 *
1466 * @reg_state: Register/unregister state machine
1467 * @dismantle: Device is going to be freed
1468 * @rtnl_link_state: This enum represents the phases of creating
1469 * a new link
1470 *
1471 * @destructor: Called from unregister,
1472 * can be used to call free_netdev
1473 * @npinfo: XXX: need comments on this one
1474 * @nd_net: Network namespace this network device is inside
1475 *
1476 * @ml_priv: Mid-layer private
1477 * @lstats: Loopback statistics
1478 * @tstats: Tunnel statistics
1479 * @dstats: Dummy statistics
1480 * @vstats: Virtual ethernet statistics
1481 *
1482 * @garp_port: GARP
1483 * @mrp_port: MRP
1484 *
1485 * @dev: Class/net/name entry
1486 * @sysfs_groups: Space for optional device, statistics and wireless
1487 * sysfs groups
1488 *
1489 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1490 * @rtnl_link_ops: Rtnl_link_ops
1491 *
1492 * @gso_max_size: Maximum size of generic segmentation offload
1493 * @gso_max_segs: Maximum number of segments that can be passed to the
1494 * NIC for GSO
1495 * @gso_min_segs: Minimum number of segments that can be passed to the
1496 * NIC for GSO
1497 *
1498 * @dcbnl_ops: Data Center Bridging netlink ops
1499 * @num_tc: Number of traffic classes in the net device
1500 * @tc_to_txq: XXX: need comments on this one
1501 * @prio_tc_map XXX: need comments on this one
1502 *
1503 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1504 *
1505 * @priomap: XXX: need comments on this one
1506 * @phydev: Physical device may attach itself
1507 * for hardware timestamping
1508 *
1509 * @qdisc_tx_busylock: XXX: need comments on this one
1510 *
1511 * @group: The group, that the device belongs to
1512 * @pm_qos_req: Power Management QoS object
1513 *
1514 * FIXME: cleanup struct net_device such that network protocol info
1515 * moves out.
1516 */
1517
1518 struct net_device {
1519 char name[IFNAMSIZ];
1520 struct hlist_node name_hlist;
1521 char *ifalias;
1522 /*
1523 * I/O specific fields
1524 * FIXME: Merge these and struct ifmap into one
1525 */
1526 unsigned long mem_end;
1527 unsigned long mem_start;
1528 unsigned long base_addr;
1529 int irq;
1530
1531 /*
1532 * Some hardware also needs these fields (state,dev_list,
1533 * napi_list,unreg_list,close_list) but they are not
1534 * part of the usual set specified in Space.c.
1535 */
1536
1537 unsigned long state;
1538
1539 struct list_head dev_list;
1540 struct list_head napi_list;
1541 struct list_head unreg_list;
1542 struct list_head close_list;
1543 struct list_head ptype_all;
1544 struct list_head ptype_specific;
1545
1546 struct {
1547 struct list_head upper;
1548 struct list_head lower;
1549 } adj_list;
1550
1551 struct {
1552 struct list_head upper;
1553 struct list_head lower;
1554 } all_adj_list;
1555
1556 netdev_features_t features;
1557 netdev_features_t hw_features;
1558 netdev_features_t wanted_features;
1559 netdev_features_t vlan_features;
1560 netdev_features_t hw_enc_features;
1561 netdev_features_t mpls_features;
1562
1563 int ifindex;
1564 int iflink;
1565
1566 struct net_device_stats stats;
1567
1568 atomic_long_t rx_dropped;
1569 atomic_long_t tx_dropped;
1570
1571 atomic_t carrier_changes;
1572
1573 #ifdef CONFIG_WIRELESS_EXT
1574 const struct iw_handler_def * wireless_handlers;
1575 struct iw_public_data * wireless_data;
1576 #endif
1577 const struct net_device_ops *netdev_ops;
1578 const struct ethtool_ops *ethtool_ops;
1579 const struct forwarding_accel_ops *fwd_ops;
1580
1581 const struct header_ops *header_ops;
1582
1583 unsigned int flags;
1584 unsigned int priv_flags;
1585
1586 unsigned short gflags;
1587 unsigned short padded;
1588
1589 unsigned char operstate;
1590 unsigned char link_mode;
1591
1592 unsigned char if_port;
1593 unsigned char dma;
1594
1595 unsigned int mtu;
1596 unsigned short type;
1597 unsigned short hard_header_len;
1598
1599 unsigned short needed_headroom;
1600 unsigned short needed_tailroom;
1601
1602 /* Interface address info. */
1603 unsigned char perm_addr[MAX_ADDR_LEN];
1604 unsigned char addr_assign_type;
1605 unsigned char addr_len;
1606 unsigned short neigh_priv_len;
1607 unsigned short dev_id;
1608 unsigned short dev_port;
1609 spinlock_t addr_list_lock;
1610 struct netdev_hw_addr_list uc;
1611 struct netdev_hw_addr_list mc;
1612 struct netdev_hw_addr_list dev_addrs;
1613
1614 #ifdef CONFIG_SYSFS
1615 struct kset *queues_kset;
1616 #endif
1617
1618 unsigned char name_assign_type;
1619
1620 bool uc_promisc;
1621 unsigned int promiscuity;
1622 unsigned int allmulti;
1623
1624
1625 /* Protocol specific pointers */
1626
1627 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1628 struct vlan_info __rcu *vlan_info;
1629 #endif
1630 #if IS_ENABLED(CONFIG_NET_DSA)
1631 struct dsa_switch_tree *dsa_ptr;
1632 #endif
1633 #if IS_ENABLED(CONFIG_TIPC)
1634 struct tipc_bearer __rcu *tipc_ptr;
1635 #endif
1636 void *atalk_ptr;
1637 struct in_device __rcu *ip_ptr;
1638 struct dn_dev __rcu *dn_ptr;
1639 struct inet6_dev __rcu *ip6_ptr;
1640 void *ax25_ptr;
1641 struct wireless_dev *ieee80211_ptr;
1642 struct wpan_dev *ieee802154_ptr;
1643
1644 /*
1645 * Cache lines mostly used on receive path (including eth_type_trans())
1646 */
1647 unsigned long last_rx;
1648
1649 /* Interface address info used in eth_type_trans() */
1650 unsigned char *dev_addr;
1651
1652
1653 #ifdef CONFIG_SYSFS
1654 struct netdev_rx_queue *_rx;
1655
1656 unsigned int num_rx_queues;
1657 unsigned int real_num_rx_queues;
1658
1659 #endif
1660
1661 unsigned long gro_flush_timeout;
1662 rx_handler_func_t __rcu *rx_handler;
1663 void __rcu *rx_handler_data;
1664
1665 struct netdev_queue __rcu *ingress_queue;
1666 unsigned char broadcast[MAX_ADDR_LEN];
1667
1668
1669 /*
1670 * Cache lines mostly used on transmit path
1671 */
1672 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1673 unsigned int num_tx_queues;
1674 unsigned int real_num_tx_queues;
1675 struct Qdisc *qdisc;
1676 unsigned long tx_queue_len;
1677 spinlock_t tx_global_lock;
1678
1679 #ifdef CONFIG_XPS
1680 struct xps_dev_maps __rcu *xps_maps;
1681 #endif
1682 #ifdef CONFIG_RFS_ACCEL
1683 struct cpu_rmap *rx_cpu_rmap;
1684 #endif
1685
1686 /* These may be needed for future network-power-down code. */
1687
1688 /*
1689 * trans_start here is expensive for high speed devices on SMP,
1690 * please use netdev_queue->trans_start instead.
1691 */
1692 unsigned long trans_start;
1693
1694 int watchdog_timeo;
1695 struct timer_list watchdog_timer;
1696
1697 int __percpu *pcpu_refcnt;
1698 struct list_head todo_list;
1699
1700 struct hlist_node index_hlist;
1701 struct list_head link_watch_list;
1702
1703 enum { NETREG_UNINITIALIZED=0,
1704 NETREG_REGISTERED, /* completed register_netdevice */
1705 NETREG_UNREGISTERING, /* called unregister_netdevice */
1706 NETREG_UNREGISTERED, /* completed unregister todo */
1707 NETREG_RELEASED, /* called free_netdev */
1708 NETREG_DUMMY, /* dummy device for NAPI poll */
1709 } reg_state:8;
1710
1711 bool dismantle;
1712
1713 enum {
1714 RTNL_LINK_INITIALIZED,
1715 RTNL_LINK_INITIALIZING,
1716 } rtnl_link_state:16;
1717
1718 void (*destructor)(struct net_device *dev);
1719
1720 #ifdef CONFIG_NETPOLL
1721 struct netpoll_info __rcu *npinfo;
1722 #endif
1723
1724 possible_net_t nd_net;
1725
1726 /* mid-layer private */
1727 union {
1728 void *ml_priv;
1729 struct pcpu_lstats __percpu *lstats;
1730 struct pcpu_sw_netstats __percpu *tstats;
1731 struct pcpu_dstats __percpu *dstats;
1732 struct pcpu_vstats __percpu *vstats;
1733 };
1734
1735 struct garp_port __rcu *garp_port;
1736 struct mrp_port __rcu *mrp_port;
1737
1738 struct device dev;
1739 const struct attribute_group *sysfs_groups[4];
1740 const struct attribute_group *sysfs_rx_queue_group;
1741
1742 const struct rtnl_link_ops *rtnl_link_ops;
1743
1744 /* for setting kernel sock attribute on TCP connection setup */
1745 #define GSO_MAX_SIZE 65536
1746 unsigned int gso_max_size;
1747 #define GSO_MAX_SEGS 65535
1748 u16 gso_max_segs;
1749 u16 gso_min_segs;
1750 #ifdef CONFIG_DCB
1751 const struct dcbnl_rtnl_ops *dcbnl_ops;
1752 #endif
1753 u8 num_tc;
1754 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1755 u8 prio_tc_map[TC_BITMASK + 1];
1756
1757 #if IS_ENABLED(CONFIG_FCOE)
1758 unsigned int fcoe_ddp_xid;
1759 #endif
1760 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1761 struct netprio_map __rcu *priomap;
1762 #endif
1763 struct phy_device *phydev;
1764 struct lock_class_key *qdisc_tx_busylock;
1765 int group;
1766 struct pm_qos_request pm_qos_req;
1767 };
1768 #define to_net_dev(d) container_of(d, struct net_device, dev)
1769
1770 #define NETDEV_ALIGN 32
1771
1772 static inline
1773 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1774 {
1775 return dev->prio_tc_map[prio & TC_BITMASK];
1776 }
1777
1778 static inline
1779 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1780 {
1781 if (tc >= dev->num_tc)
1782 return -EINVAL;
1783
1784 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1785 return 0;
1786 }
1787
1788 static inline
1789 void netdev_reset_tc(struct net_device *dev)
1790 {
1791 dev->num_tc = 0;
1792 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1793 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1794 }
1795
1796 static inline
1797 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1798 {
1799 if (tc >= dev->num_tc)
1800 return -EINVAL;
1801
1802 dev->tc_to_txq[tc].count = count;
1803 dev->tc_to_txq[tc].offset = offset;
1804 return 0;
1805 }
1806
1807 static inline
1808 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1809 {
1810 if (num_tc > TC_MAX_QUEUE)
1811 return -EINVAL;
1812
1813 dev->num_tc = num_tc;
1814 return 0;
1815 }
1816
1817 static inline
1818 int netdev_get_num_tc(struct net_device *dev)
1819 {
1820 return dev->num_tc;
1821 }
1822
1823 static inline
1824 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1825 unsigned int index)
1826 {
1827 return &dev->_tx[index];
1828 }
1829
1830 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1831 const struct sk_buff *skb)
1832 {
1833 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1834 }
1835
1836 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1837 void (*f)(struct net_device *,
1838 struct netdev_queue *,
1839 void *),
1840 void *arg)
1841 {
1842 unsigned int i;
1843
1844 for (i = 0; i < dev->num_tx_queues; i++)
1845 f(dev, &dev->_tx[i], arg);
1846 }
1847
1848 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1849 struct sk_buff *skb,
1850 void *accel_priv);
1851
1852 /*
1853 * Net namespace inlines
1854 */
1855 static inline
1856 struct net *dev_net(const struct net_device *dev)
1857 {
1858 return read_pnet(&dev->nd_net);
1859 }
1860
1861 static inline
1862 void dev_net_set(struct net_device *dev, struct net *net)
1863 {
1864 write_pnet(&dev->nd_net, net);
1865 }
1866
1867 static inline bool netdev_uses_dsa(struct net_device *dev)
1868 {
1869 #if IS_ENABLED(CONFIG_NET_DSA)
1870 if (dev->dsa_ptr != NULL)
1871 return dsa_uses_tagged_protocol(dev->dsa_ptr);
1872 #endif
1873 return false;
1874 }
1875
1876 /**
1877 * netdev_priv - access network device private data
1878 * @dev: network device
1879 *
1880 * Get network device private data
1881 */
1882 static inline void *netdev_priv(const struct net_device *dev)
1883 {
1884 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
1885 }
1886
1887 /* Set the sysfs physical device reference for the network logical device
1888 * if set prior to registration will cause a symlink during initialization.
1889 */
1890 #define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
1891
1892 /* Set the sysfs device type for the network logical device to allow
1893 * fine-grained identification of different network device types. For
1894 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
1895 */
1896 #define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
1897
1898 /* Default NAPI poll() weight
1899 * Device drivers are strongly advised to not use bigger value
1900 */
1901 #define NAPI_POLL_WEIGHT 64
1902
1903 /**
1904 * netif_napi_add - initialize a napi context
1905 * @dev: network device
1906 * @napi: napi context
1907 * @poll: polling function
1908 * @weight: default weight
1909 *
1910 * netif_napi_add() must be used to initialize a napi context prior to calling
1911 * *any* of the other napi related functions.
1912 */
1913 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
1914 int (*poll)(struct napi_struct *, int), int weight);
1915
1916 /**
1917 * netif_napi_del - remove a napi context
1918 * @napi: napi context
1919 *
1920 * netif_napi_del() removes a napi context from the network device napi list
1921 */
1922 void netif_napi_del(struct napi_struct *napi);
1923
1924 struct napi_gro_cb {
1925 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
1926 void *frag0;
1927
1928 /* Length of frag0. */
1929 unsigned int frag0_len;
1930
1931 /* This indicates where we are processing relative to skb->data. */
1932 int data_offset;
1933
1934 /* This is non-zero if the packet cannot be merged with the new skb. */
1935 u16 flush;
1936
1937 /* Save the IP ID here and check when we get to the transport layer */
1938 u16 flush_id;
1939
1940 /* Number of segments aggregated. */
1941 u16 count;
1942
1943 /* Start offset for remote checksum offload */
1944 u16 gro_remcsum_start;
1945
1946 /* jiffies when first packet was created/queued */
1947 unsigned long age;
1948
1949 /* Used in ipv6_gro_receive() and foo-over-udp */
1950 u16 proto;
1951
1952 /* This is non-zero if the packet may be of the same flow. */
1953 u8 same_flow:1;
1954
1955 /* Used in udp_gro_receive */
1956 u8 udp_mark:1;
1957
1958 /* GRO checksum is valid */
1959 u8 csum_valid:1;
1960
1961 /* Number of checksums via CHECKSUM_UNNECESSARY */
1962 u8 csum_cnt:3;
1963
1964 /* Free the skb? */
1965 u8 free:2;
1966 #define NAPI_GRO_FREE 1
1967 #define NAPI_GRO_FREE_STOLEN_HEAD 2
1968
1969 /* Used in foo-over-udp, set in udp[46]_gro_receive */
1970 u8 is_ipv6:1;
1971
1972 /* 7 bit hole */
1973
1974 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
1975 __wsum csum;
1976
1977 /* used in skb_gro_receive() slow path */
1978 struct sk_buff *last;
1979 };
1980
1981 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
1982
1983 struct packet_type {
1984 __be16 type; /* This is really htons(ether_type). */
1985 struct net_device *dev; /* NULL is wildcarded here */
1986 int (*func) (struct sk_buff *,
1987 struct net_device *,
1988 struct packet_type *,
1989 struct net_device *);
1990 bool (*id_match)(struct packet_type *ptype,
1991 struct sock *sk);
1992 void *af_packet_priv;
1993 struct list_head list;
1994 };
1995
1996 struct offload_callbacks {
1997 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
1998 netdev_features_t features);
1999 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2000 struct sk_buff *skb);
2001 int (*gro_complete)(struct sk_buff *skb, int nhoff);
2002 };
2003
2004 struct packet_offload {
2005 __be16 type; /* This is really htons(ether_type). */
2006 struct offload_callbacks callbacks;
2007 struct list_head list;
2008 };
2009
2010 struct udp_offload;
2011
2012 struct udp_offload_callbacks {
2013 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2014 struct sk_buff *skb,
2015 struct udp_offload *uoff);
2016 int (*gro_complete)(struct sk_buff *skb,
2017 int nhoff,
2018 struct udp_offload *uoff);
2019 };
2020
2021 struct udp_offload {
2022 __be16 port;
2023 u8 ipproto;
2024 struct udp_offload_callbacks callbacks;
2025 };
2026
2027 /* often modified stats are per cpu, other are shared (netdev->stats) */
2028 struct pcpu_sw_netstats {
2029 u64 rx_packets;
2030 u64 rx_bytes;
2031 u64 tx_packets;
2032 u64 tx_bytes;
2033 struct u64_stats_sync syncp;
2034 };
2035
2036 #define netdev_alloc_pcpu_stats(type) \
2037 ({ \
2038 typeof(type) __percpu *pcpu_stats = alloc_percpu(type); \
2039 if (pcpu_stats) { \
2040 int i; \
2041 for_each_possible_cpu(i) { \
2042 typeof(type) *stat; \
2043 stat = per_cpu_ptr(pcpu_stats, i); \
2044 u64_stats_init(&stat->syncp); \
2045 } \
2046 } \
2047 pcpu_stats; \
2048 })
2049
2050 #include <linux/notifier.h>
2051
2052 /* netdevice notifier chain. Please remember to update the rtnetlink
2053 * notification exclusion list in rtnetlink_event() when adding new
2054 * types.
2055 */
2056 #define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2057 #define NETDEV_DOWN 0x0002
2058 #define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2059 detected a hardware crash and restarted
2060 - we can use this eg to kick tcp sessions
2061 once done */
2062 #define NETDEV_CHANGE 0x0004 /* Notify device state change */
2063 #define NETDEV_REGISTER 0x0005
2064 #define NETDEV_UNREGISTER 0x0006
2065 #define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
2066 #define NETDEV_CHANGEADDR 0x0008
2067 #define NETDEV_GOING_DOWN 0x0009
2068 #define NETDEV_CHANGENAME 0x000A
2069 #define NETDEV_FEAT_CHANGE 0x000B
2070 #define NETDEV_BONDING_FAILOVER 0x000C
2071 #define NETDEV_PRE_UP 0x000D
2072 #define NETDEV_PRE_TYPE_CHANGE 0x000E
2073 #define NETDEV_POST_TYPE_CHANGE 0x000F
2074 #define NETDEV_POST_INIT 0x0010
2075 #define NETDEV_UNREGISTER_FINAL 0x0011
2076 #define NETDEV_RELEASE 0x0012
2077 #define NETDEV_NOTIFY_PEERS 0x0013
2078 #define NETDEV_JOIN 0x0014
2079 #define NETDEV_CHANGEUPPER 0x0015
2080 #define NETDEV_RESEND_IGMP 0x0016
2081 #define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
2082 #define NETDEV_CHANGEINFODATA 0x0018
2083 #define NETDEV_BONDING_INFO 0x0019
2084
2085 int register_netdevice_notifier(struct notifier_block *nb);
2086 int unregister_netdevice_notifier(struct notifier_block *nb);
2087
2088 struct netdev_notifier_info {
2089 struct net_device *dev;
2090 };
2091
2092 struct netdev_notifier_change_info {
2093 struct netdev_notifier_info info; /* must be first */
2094 unsigned int flags_changed;
2095 };
2096
2097 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2098 struct net_device *dev)
2099 {
2100 info->dev = dev;
2101 }
2102
2103 static inline struct net_device *
2104 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2105 {
2106 return info->dev;
2107 }
2108
2109 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2110
2111
2112 extern rwlock_t dev_base_lock; /* Device list lock */
2113
2114 #define for_each_netdev(net, d) \
2115 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2116 #define for_each_netdev_reverse(net, d) \
2117 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2118 #define for_each_netdev_rcu(net, d) \
2119 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2120 #define for_each_netdev_safe(net, d, n) \
2121 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2122 #define for_each_netdev_continue(net, d) \
2123 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2124 #define for_each_netdev_continue_rcu(net, d) \
2125 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2126 #define for_each_netdev_in_bond_rcu(bond, slave) \
2127 for_each_netdev_rcu(&init_net, slave) \
2128 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2129 #define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2130
2131 static inline struct net_device *next_net_device(struct net_device *dev)
2132 {
2133 struct list_head *lh;
2134 struct net *net;
2135
2136 net = dev_net(dev);
2137 lh = dev->dev_list.next;
2138 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2139 }
2140
2141 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2142 {
2143 struct list_head *lh;
2144 struct net *net;
2145
2146 net = dev_net(dev);
2147 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2148 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2149 }
2150
2151 static inline struct net_device *first_net_device(struct net *net)
2152 {
2153 return list_empty(&net->dev_base_head) ? NULL :
2154 net_device_entry(net->dev_base_head.next);
2155 }
2156
2157 static inline struct net_device *first_net_device_rcu(struct net *net)
2158 {
2159 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2160
2161 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2162 }
2163
2164 int netdev_boot_setup_check(struct net_device *dev);
2165 unsigned long netdev_boot_base(const char *prefix, int unit);
2166 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2167 const char *hwaddr);
2168 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2169 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2170 void dev_add_pack(struct packet_type *pt);
2171 void dev_remove_pack(struct packet_type *pt);
2172 void __dev_remove_pack(struct packet_type *pt);
2173 void dev_add_offload(struct packet_offload *po);
2174 void dev_remove_offload(struct packet_offload *po);
2175
2176 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2177 unsigned short mask);
2178 struct net_device *dev_get_by_name(struct net *net, const char *name);
2179 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2180 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2181 int dev_alloc_name(struct net_device *dev, const char *name);
2182 int dev_open(struct net_device *dev);
2183 int dev_close(struct net_device *dev);
2184 void dev_disable_lro(struct net_device *dev);
2185 int dev_loopback_xmit(struct sk_buff *newskb);
2186 int dev_queue_xmit(struct sk_buff *skb);
2187 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2188 int register_netdevice(struct net_device *dev);
2189 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2190 void unregister_netdevice_many(struct list_head *head);
2191 static inline void unregister_netdevice(struct net_device *dev)
2192 {
2193 unregister_netdevice_queue(dev, NULL);
2194 }
2195
2196 int netdev_refcnt_read(const struct net_device *dev);
2197 void free_netdev(struct net_device *dev);
2198 void netdev_freemem(struct net_device *dev);
2199 void synchronize_net(void);
2200 int init_dummy_netdev(struct net_device *dev);
2201
2202 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2203 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2204 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2205 int netdev_get_name(struct net *net, char *name, int ifindex);
2206 int dev_restart(struct net_device *dev);
2207 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2208
2209 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2210 {
2211 return NAPI_GRO_CB(skb)->data_offset;
2212 }
2213
2214 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2215 {
2216 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2217 }
2218
2219 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2220 {
2221 NAPI_GRO_CB(skb)->data_offset += len;
2222 }
2223
2224 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2225 unsigned int offset)
2226 {
2227 return NAPI_GRO_CB(skb)->frag0 + offset;
2228 }
2229
2230 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2231 {
2232 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2233 }
2234
2235 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2236 unsigned int offset)
2237 {
2238 if (!pskb_may_pull(skb, hlen))
2239 return NULL;
2240
2241 NAPI_GRO_CB(skb)->frag0 = NULL;
2242 NAPI_GRO_CB(skb)->frag0_len = 0;
2243 return skb->data + offset;
2244 }
2245
2246 static inline void *skb_gro_network_header(struct sk_buff *skb)
2247 {
2248 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2249 skb_network_offset(skb);
2250 }
2251
2252 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2253 const void *start, unsigned int len)
2254 {
2255 if (NAPI_GRO_CB(skb)->csum_valid)
2256 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2257 csum_partial(start, len, 0));
2258 }
2259
2260 /* GRO checksum functions. These are logical equivalents of the normal
2261 * checksum functions (in skbuff.h) except that they operate on the GRO
2262 * offsets and fields in sk_buff.
2263 */
2264
2265 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2266
2267 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2268 {
2269 return (NAPI_GRO_CB(skb)->gro_remcsum_start - skb_headroom(skb) ==
2270 skb_gro_offset(skb));
2271 }
2272
2273 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2274 bool zero_okay,
2275 __sum16 check)
2276 {
2277 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2278 skb_checksum_start_offset(skb) <
2279 skb_gro_offset(skb)) &&
2280 !skb_at_gro_remcsum_start(skb) &&
2281 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2282 (!zero_okay || check));
2283 }
2284
2285 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2286 __wsum psum)
2287 {
2288 if (NAPI_GRO_CB(skb)->csum_valid &&
2289 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2290 return 0;
2291
2292 NAPI_GRO_CB(skb)->csum = psum;
2293
2294 return __skb_gro_checksum_complete(skb);
2295 }
2296
2297 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2298 {
2299 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2300 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2301 NAPI_GRO_CB(skb)->csum_cnt--;
2302 } else {
2303 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2304 * verified a new top level checksum or an encapsulated one
2305 * during GRO. This saves work if we fallback to normal path.
2306 */
2307 __skb_incr_checksum_unnecessary(skb);
2308 }
2309 }
2310
2311 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2312 compute_pseudo) \
2313 ({ \
2314 __sum16 __ret = 0; \
2315 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2316 __ret = __skb_gro_checksum_validate_complete(skb, \
2317 compute_pseudo(skb, proto)); \
2318 if (__ret) \
2319 __skb_mark_checksum_bad(skb); \
2320 else \
2321 skb_gro_incr_csum_unnecessary(skb); \
2322 __ret; \
2323 })
2324
2325 #define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2326 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2327
2328 #define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2329 compute_pseudo) \
2330 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2331
2332 #define skb_gro_checksum_simple_validate(skb) \
2333 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2334
2335 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2336 {
2337 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2338 !NAPI_GRO_CB(skb)->csum_valid);
2339 }
2340
2341 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2342 __sum16 check, __wsum pseudo)
2343 {
2344 NAPI_GRO_CB(skb)->csum = ~pseudo;
2345 NAPI_GRO_CB(skb)->csum_valid = 1;
2346 }
2347
2348 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2349 do { \
2350 if (__skb_gro_checksum_convert_check(skb)) \
2351 __skb_gro_checksum_convert(skb, check, \
2352 compute_pseudo(skb, proto)); \
2353 } while (0)
2354
2355 struct gro_remcsum {
2356 int offset;
2357 __wsum delta;
2358 };
2359
2360 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2361 {
2362 grc->offset = 0;
2363 grc->delta = 0;
2364 }
2365
2366 static inline void skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2367 int start, int offset,
2368 struct gro_remcsum *grc,
2369 bool nopartial)
2370 {
2371 __wsum delta;
2372
2373 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2374
2375 if (!nopartial) {
2376 NAPI_GRO_CB(skb)->gro_remcsum_start =
2377 ((unsigned char *)ptr + start) - skb->head;
2378 return;
2379 }
2380
2381 delta = remcsum_adjust(ptr, NAPI_GRO_CB(skb)->csum, start, offset);
2382
2383 /* Adjust skb->csum since we changed the packet */
2384 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2385
2386 grc->offset = (ptr + offset) - (void *)skb->head;
2387 grc->delta = delta;
2388 }
2389
2390 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2391 struct gro_remcsum *grc)
2392 {
2393 if (!grc->delta)
2394 return;
2395
2396 remcsum_unadjust((__sum16 *)(skb->head + grc->offset), grc->delta);
2397 }
2398
2399 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2400 unsigned short type,
2401 const void *daddr, const void *saddr,
2402 unsigned int len)
2403 {
2404 if (!dev->header_ops || !dev->header_ops->create)
2405 return 0;
2406
2407 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2408 }
2409
2410 static inline int dev_parse_header(const struct sk_buff *skb,
2411 unsigned char *haddr)
2412 {
2413 const struct net_device *dev = skb->dev;
2414
2415 if (!dev->header_ops || !dev->header_ops->parse)
2416 return 0;
2417 return dev->header_ops->parse(skb, haddr);
2418 }
2419
2420 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2421 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2422 static inline int unregister_gifconf(unsigned int family)
2423 {
2424 return register_gifconf(family, NULL);
2425 }
2426
2427 #ifdef CONFIG_NET_FLOW_LIMIT
2428 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
2429 struct sd_flow_limit {
2430 u64 count;
2431 unsigned int num_buckets;
2432 unsigned int history_head;
2433 u16 history[FLOW_LIMIT_HISTORY];
2434 u8 buckets[];
2435 };
2436
2437 extern int netdev_flow_limit_table_len;
2438 #endif /* CONFIG_NET_FLOW_LIMIT */
2439
2440 /*
2441 * Incoming packets are placed on per-cpu queues
2442 */
2443 struct softnet_data {
2444 struct list_head poll_list;
2445 struct sk_buff_head process_queue;
2446
2447 /* stats */
2448 unsigned int processed;
2449 unsigned int time_squeeze;
2450 unsigned int cpu_collision;
2451 unsigned int received_rps;
2452 #ifdef CONFIG_RPS
2453 struct softnet_data *rps_ipi_list;
2454 #endif
2455 #ifdef CONFIG_NET_FLOW_LIMIT
2456 struct sd_flow_limit __rcu *flow_limit;
2457 #endif
2458 struct Qdisc *output_queue;
2459 struct Qdisc **output_queue_tailp;
2460 struct sk_buff *completion_queue;
2461
2462 #ifdef CONFIG_RPS
2463 /* Elements below can be accessed between CPUs for RPS */
2464 struct call_single_data csd ____cacheline_aligned_in_smp;
2465 struct softnet_data *rps_ipi_next;
2466 unsigned int cpu;
2467 unsigned int input_queue_head;
2468 unsigned int input_queue_tail;
2469 #endif
2470 unsigned int dropped;
2471 struct sk_buff_head input_pkt_queue;
2472 struct napi_struct backlog;
2473
2474 };
2475
2476 static inline void input_queue_head_incr(struct softnet_data *sd)
2477 {
2478 #ifdef CONFIG_RPS
2479 sd->input_queue_head++;
2480 #endif
2481 }
2482
2483 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2484 unsigned int *qtail)
2485 {
2486 #ifdef CONFIG_RPS
2487 *qtail = ++sd->input_queue_tail;
2488 #endif
2489 }
2490
2491 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2492
2493 void __netif_schedule(struct Qdisc *q);
2494 void netif_schedule_queue(struct netdev_queue *txq);
2495
2496 static inline void netif_tx_schedule_all(struct net_device *dev)
2497 {
2498 unsigned int i;
2499
2500 for (i = 0; i < dev->num_tx_queues; i++)
2501 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2502 }
2503
2504 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2505 {
2506 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2507 }
2508
2509 /**
2510 * netif_start_queue - allow transmit
2511 * @dev: network device
2512 *
2513 * Allow upper layers to call the device hard_start_xmit routine.
2514 */
2515 static inline void netif_start_queue(struct net_device *dev)
2516 {
2517 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2518 }
2519
2520 static inline void netif_tx_start_all_queues(struct net_device *dev)
2521 {
2522 unsigned int i;
2523
2524 for (i = 0; i < dev->num_tx_queues; i++) {
2525 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2526 netif_tx_start_queue(txq);
2527 }
2528 }
2529
2530 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2531
2532 /**
2533 * netif_wake_queue - restart transmit
2534 * @dev: network device
2535 *
2536 * Allow upper layers to call the device hard_start_xmit routine.
2537 * Used for flow control when transmit resources are available.
2538 */
2539 static inline void netif_wake_queue(struct net_device *dev)
2540 {
2541 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2542 }
2543
2544 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2545 {
2546 unsigned int i;
2547
2548 for (i = 0; i < dev->num_tx_queues; i++) {
2549 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2550 netif_tx_wake_queue(txq);
2551 }
2552 }
2553
2554 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2555 {
2556 if (WARN_ON(!dev_queue)) {
2557 pr_info("netif_stop_queue() cannot be called before register_netdev()\n");
2558 return;
2559 }
2560 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2561 }
2562
2563 /**
2564 * netif_stop_queue - stop transmitted packets
2565 * @dev: network device
2566 *
2567 * Stop upper layers calling the device hard_start_xmit routine.
2568 * Used for flow control when transmit resources are unavailable.
2569 */
2570 static inline void netif_stop_queue(struct net_device *dev)
2571 {
2572 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2573 }
2574
2575 static inline void netif_tx_stop_all_queues(struct net_device *dev)
2576 {
2577 unsigned int i;
2578
2579 for (i = 0; i < dev->num_tx_queues; i++) {
2580 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2581 netif_tx_stop_queue(txq);
2582 }
2583 }
2584
2585 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2586 {
2587 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2588 }
2589
2590 /**
2591 * netif_queue_stopped - test if transmit queue is flowblocked
2592 * @dev: network device
2593 *
2594 * Test if transmit queue on device is currently unable to send.
2595 */
2596 static inline bool netif_queue_stopped(const struct net_device *dev)
2597 {
2598 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2599 }
2600
2601 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2602 {
2603 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2604 }
2605
2606 static inline bool
2607 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2608 {
2609 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2610 }
2611
2612 static inline bool
2613 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2614 {
2615 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2616 }
2617
2618 /**
2619 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2620 * @dev_queue: pointer to transmit queue
2621 *
2622 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2623 * to give appropriate hint to the cpu.
2624 */
2625 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2626 {
2627 #ifdef CONFIG_BQL
2628 prefetchw(&dev_queue->dql.num_queued);
2629 #endif
2630 }
2631
2632 /**
2633 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2634 * @dev_queue: pointer to transmit queue
2635 *
2636 * BQL enabled drivers might use this helper in their TX completion path,
2637 * to give appropriate hint to the cpu.
2638 */
2639 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2640 {
2641 #ifdef CONFIG_BQL
2642 prefetchw(&dev_queue->dql.limit);
2643 #endif
2644 }
2645
2646 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2647 unsigned int bytes)
2648 {
2649 #ifdef CONFIG_BQL
2650 dql_queued(&dev_queue->dql, bytes);
2651
2652 if (likely(dql_avail(&dev_queue->dql) >= 0))
2653 return;
2654
2655 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2656
2657 /*
2658 * The XOFF flag must be set before checking the dql_avail below,
2659 * because in netdev_tx_completed_queue we update the dql_completed
2660 * before checking the XOFF flag.
2661 */
2662 smp_mb();
2663
2664 /* check again in case another CPU has just made room avail */
2665 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2666 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2667 #endif
2668 }
2669
2670 /**
2671 * netdev_sent_queue - report the number of bytes queued to hardware
2672 * @dev: network device
2673 * @bytes: number of bytes queued to the hardware device queue
2674 *
2675 * Report the number of bytes queued for sending/completion to the network
2676 * device hardware queue. @bytes should be a good approximation and should
2677 * exactly match netdev_completed_queue() @bytes
2678 */
2679 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2680 {
2681 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2682 }
2683
2684 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2685 unsigned int pkts, unsigned int bytes)
2686 {
2687 #ifdef CONFIG_BQL
2688 if (unlikely(!bytes))
2689 return;
2690
2691 dql_completed(&dev_queue->dql, bytes);
2692
2693 /*
2694 * Without the memory barrier there is a small possiblity that
2695 * netdev_tx_sent_queue will miss the update and cause the queue to
2696 * be stopped forever
2697 */
2698 smp_mb();
2699
2700 if (dql_avail(&dev_queue->dql) < 0)
2701 return;
2702
2703 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2704 netif_schedule_queue(dev_queue);
2705 #endif
2706 }
2707
2708 /**
2709 * netdev_completed_queue - report bytes and packets completed by device
2710 * @dev: network device
2711 * @pkts: actual number of packets sent over the medium
2712 * @bytes: actual number of bytes sent over the medium
2713 *
2714 * Report the number of bytes and packets transmitted by the network device
2715 * hardware queue over the physical medium, @bytes must exactly match the
2716 * @bytes amount passed to netdev_sent_queue()
2717 */
2718 static inline void netdev_completed_queue(struct net_device *dev,
2719 unsigned int pkts, unsigned int bytes)
2720 {
2721 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2722 }
2723
2724 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2725 {
2726 #ifdef CONFIG_BQL
2727 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2728 dql_reset(&q->dql);
2729 #endif
2730 }
2731
2732 /**
2733 * netdev_reset_queue - reset the packets and bytes count of a network device
2734 * @dev_queue: network device
2735 *
2736 * Reset the bytes and packet count of a network device and clear the
2737 * software flow control OFF bit for this network device
2738 */
2739 static inline void netdev_reset_queue(struct net_device *dev_queue)
2740 {
2741 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
2742 }
2743
2744 /**
2745 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
2746 * @dev: network device
2747 * @queue_index: given tx queue index
2748 *
2749 * Returns 0 if given tx queue index >= number of device tx queues,
2750 * otherwise returns the originally passed tx queue index.
2751 */
2752 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
2753 {
2754 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
2755 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
2756 dev->name, queue_index,
2757 dev->real_num_tx_queues);
2758 return 0;
2759 }
2760
2761 return queue_index;
2762 }
2763
2764 /**
2765 * netif_running - test if up
2766 * @dev: network device
2767 *
2768 * Test if the device has been brought up.
2769 */
2770 static inline bool netif_running(const struct net_device *dev)
2771 {
2772 return test_bit(__LINK_STATE_START, &dev->state);
2773 }
2774
2775 /*
2776 * Routines to manage the subqueues on a device. We only need start
2777 * stop, and a check if it's stopped. All other device management is
2778 * done at the overall netdevice level.
2779 * Also test the device if we're multiqueue.
2780 */
2781
2782 /**
2783 * netif_start_subqueue - allow sending packets on subqueue
2784 * @dev: network device
2785 * @queue_index: sub queue index
2786 *
2787 * Start individual transmit queue of a device with multiple transmit queues.
2788 */
2789 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
2790 {
2791 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2792
2793 netif_tx_start_queue(txq);
2794 }
2795
2796 /**
2797 * netif_stop_subqueue - stop sending packets on subqueue
2798 * @dev: network device
2799 * @queue_index: sub queue index
2800 *
2801 * Stop individual transmit queue of a device with multiple transmit queues.
2802 */
2803 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
2804 {
2805 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2806 netif_tx_stop_queue(txq);
2807 }
2808
2809 /**
2810 * netif_subqueue_stopped - test status of subqueue
2811 * @dev: network device
2812 * @queue_index: sub queue index
2813 *
2814 * Check individual transmit queue of a device with multiple transmit queues.
2815 */
2816 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
2817 u16 queue_index)
2818 {
2819 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
2820
2821 return netif_tx_queue_stopped(txq);
2822 }
2823
2824 static inline bool netif_subqueue_stopped(const struct net_device *dev,
2825 struct sk_buff *skb)
2826 {
2827 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
2828 }
2829
2830 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
2831
2832 #ifdef CONFIG_XPS
2833 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
2834 u16 index);
2835 #else
2836 static inline int netif_set_xps_queue(struct net_device *dev,
2837 const struct cpumask *mask,
2838 u16 index)
2839 {
2840 return 0;
2841 }
2842 #endif
2843
2844 /*
2845 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
2846 * as a distribution range limit for the returned value.
2847 */
2848 static inline u16 skb_tx_hash(const struct net_device *dev,
2849 struct sk_buff *skb)
2850 {
2851 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
2852 }
2853
2854 /**
2855 * netif_is_multiqueue - test if device has multiple transmit queues
2856 * @dev: network device
2857 *
2858 * Check if device has multiple transmit queues
2859 */
2860 static inline bool netif_is_multiqueue(const struct net_device *dev)
2861 {
2862 return dev->num_tx_queues > 1;
2863 }
2864
2865 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
2866
2867 #ifdef CONFIG_SYSFS
2868 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
2869 #else
2870 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
2871 unsigned int rxq)
2872 {
2873 return 0;
2874 }
2875 #endif
2876
2877 #ifdef CONFIG_SYSFS
2878 static inline unsigned int get_netdev_rx_queue_index(
2879 struct netdev_rx_queue *queue)
2880 {
2881 struct net_device *dev = queue->dev;
2882 int index = queue - dev->_rx;
2883
2884 BUG_ON(index >= dev->num_rx_queues);
2885 return index;
2886 }
2887 #endif
2888
2889 #define DEFAULT_MAX_NUM_RSS_QUEUES (8)
2890 int netif_get_num_default_rss_queues(void);
2891
2892 enum skb_free_reason {
2893 SKB_REASON_CONSUMED,
2894 SKB_REASON_DROPPED,
2895 };
2896
2897 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
2898 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
2899
2900 /*
2901 * It is not allowed to call kfree_skb() or consume_skb() from hardware
2902 * interrupt context or with hardware interrupts being disabled.
2903 * (in_irq() || irqs_disabled())
2904 *
2905 * We provide four helpers that can be used in following contexts :
2906 *
2907 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
2908 * replacing kfree_skb(skb)
2909 *
2910 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
2911 * Typically used in place of consume_skb(skb) in TX completion path
2912 *
2913 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
2914 * replacing kfree_skb(skb)
2915 *
2916 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
2917 * and consumed a packet. Used in place of consume_skb(skb)
2918 */
2919 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
2920 {
2921 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
2922 }
2923
2924 static inline void dev_consume_skb_irq(struct sk_buff *skb)
2925 {
2926 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
2927 }
2928
2929 static inline void dev_kfree_skb_any(struct sk_buff *skb)
2930 {
2931 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
2932 }
2933
2934 static inline void dev_consume_skb_any(struct sk_buff *skb)
2935 {
2936 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
2937 }
2938
2939 int netif_rx(struct sk_buff *skb);
2940 int netif_rx_ni(struct sk_buff *skb);
2941 int netif_receive_skb(struct sk_buff *skb);
2942 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
2943 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
2944 struct sk_buff *napi_get_frags(struct napi_struct *napi);
2945 gro_result_t napi_gro_frags(struct napi_struct *napi);
2946 struct packet_offload *gro_find_receive_by_type(__be16 type);
2947 struct packet_offload *gro_find_complete_by_type(__be16 type);
2948
2949 static inline void napi_free_frags(struct napi_struct *napi)
2950 {
2951 kfree_skb(napi->skb);
2952 napi->skb = NULL;
2953 }
2954
2955 int netdev_rx_handler_register(struct net_device *dev,
2956 rx_handler_func_t *rx_handler,
2957 void *rx_handler_data);
2958 void netdev_rx_handler_unregister(struct net_device *dev);
2959
2960 bool dev_valid_name(const char *name);
2961 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
2962 int dev_ethtool(struct net *net, struct ifreq *);
2963 unsigned int dev_get_flags(const struct net_device *);
2964 int __dev_change_flags(struct net_device *, unsigned int flags);
2965 int dev_change_flags(struct net_device *, unsigned int);
2966 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
2967 unsigned int gchanges);
2968 int dev_change_name(struct net_device *, const char *);
2969 int dev_set_alias(struct net_device *, const char *, size_t);
2970 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
2971 int dev_set_mtu(struct net_device *, int);
2972 void dev_set_group(struct net_device *, int);
2973 int dev_set_mac_address(struct net_device *, struct sockaddr *);
2974 int dev_change_carrier(struct net_device *, bool new_carrier);
2975 int dev_get_phys_port_id(struct net_device *dev,
2976 struct netdev_phys_item_id *ppid);
2977 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
2978 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
2979 struct netdev_queue *txq, int *ret);
2980 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2981 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
2982 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
2983
2984 extern int netdev_budget;
2985
2986 /* Called by rtnetlink.c:rtnl_unlock() */
2987 void netdev_run_todo(void);
2988
2989 /**
2990 * dev_put - release reference to device
2991 * @dev: network device
2992 *
2993 * Release reference to device to allow it to be freed.
2994 */
2995 static inline void dev_put(struct net_device *dev)
2996 {
2997 this_cpu_dec(*dev->pcpu_refcnt);
2998 }
2999
3000 /**
3001 * dev_hold - get reference to device
3002 * @dev: network device
3003 *
3004 * Hold reference to device to keep it from being freed.
3005 */
3006 static inline void dev_hold(struct net_device *dev)
3007 {
3008 this_cpu_inc(*dev->pcpu_refcnt);
3009 }
3010
3011 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3012 * and _off may be called from IRQ context, but it is caller
3013 * who is responsible for serialization of these calls.
3014 *
3015 * The name carrier is inappropriate, these functions should really be
3016 * called netif_lowerlayer_*() because they represent the state of any
3017 * kind of lower layer not just hardware media.
3018 */
3019
3020 void linkwatch_init_dev(struct net_device *dev);
3021 void linkwatch_fire_event(struct net_device *dev);
3022 void linkwatch_forget_dev(struct net_device *dev);
3023
3024 /**
3025 * netif_carrier_ok - test if carrier present
3026 * @dev: network device
3027 *
3028 * Check if carrier is present on device
3029 */
3030 static inline bool netif_carrier_ok(const struct net_device *dev)
3031 {
3032 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3033 }
3034
3035 unsigned long dev_trans_start(struct net_device *dev);
3036
3037 void __netdev_watchdog_up(struct net_device *dev);
3038
3039 void netif_carrier_on(struct net_device *dev);
3040
3041 void netif_carrier_off(struct net_device *dev);
3042
3043 /**
3044 * netif_dormant_on - mark device as dormant.
3045 * @dev: network device
3046 *
3047 * Mark device as dormant (as per RFC2863).
3048 *
3049 * The dormant state indicates that the relevant interface is not
3050 * actually in a condition to pass packets (i.e., it is not 'up') but is
3051 * in a "pending" state, waiting for some external event. For "on-
3052 * demand" interfaces, this new state identifies the situation where the
3053 * interface is waiting for events to place it in the up state.
3054 *
3055 */
3056 static inline void netif_dormant_on(struct net_device *dev)
3057 {
3058 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3059 linkwatch_fire_event(dev);
3060 }
3061
3062 /**
3063 * netif_dormant_off - set device as not dormant.
3064 * @dev: network device
3065 *
3066 * Device is not in dormant state.
3067 */
3068 static inline void netif_dormant_off(struct net_device *dev)
3069 {
3070 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3071 linkwatch_fire_event(dev);
3072 }
3073
3074 /**
3075 * netif_dormant - test if carrier present
3076 * @dev: network device
3077 *
3078 * Check if carrier is present on device
3079 */
3080 static inline bool netif_dormant(const struct net_device *dev)
3081 {
3082 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3083 }
3084
3085
3086 /**
3087 * netif_oper_up - test if device is operational
3088 * @dev: network device
3089 *
3090 * Check if carrier is operational
3091 */
3092 static inline bool netif_oper_up(const struct net_device *dev)
3093 {
3094 return (dev->operstate == IF_OPER_UP ||
3095 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3096 }
3097
3098 /**
3099 * netif_device_present - is device available or removed
3100 * @dev: network device
3101 *
3102 * Check if device has not been removed from system.
3103 */
3104 static inline bool netif_device_present(struct net_device *dev)
3105 {
3106 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3107 }
3108
3109 void netif_device_detach(struct net_device *dev);
3110
3111 void netif_device_attach(struct net_device *dev);
3112
3113 /*
3114 * Network interface message level settings
3115 */
3116
3117 enum {
3118 NETIF_MSG_DRV = 0x0001,
3119 NETIF_MSG_PROBE = 0x0002,
3120 NETIF_MSG_LINK = 0x0004,
3121 NETIF_MSG_TIMER = 0x0008,
3122 NETIF_MSG_IFDOWN = 0x0010,
3123 NETIF_MSG_IFUP = 0x0020,
3124 NETIF_MSG_RX_ERR = 0x0040,
3125 NETIF_MSG_TX_ERR = 0x0080,
3126 NETIF_MSG_TX_QUEUED = 0x0100,
3127 NETIF_MSG_INTR = 0x0200,
3128 NETIF_MSG_TX_DONE = 0x0400,
3129 NETIF_MSG_RX_STATUS = 0x0800,
3130 NETIF_MSG_PKTDATA = 0x1000,
3131 NETIF_MSG_HW = 0x2000,
3132 NETIF_MSG_WOL = 0x4000,
3133 };
3134
3135 #define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3136 #define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3137 #define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3138 #define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3139 #define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3140 #define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3141 #define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3142 #define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3143 #define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3144 #define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3145 #define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3146 #define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3147 #define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3148 #define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3149 #define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3150
3151 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3152 {
3153 /* use default */
3154 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3155 return default_msg_enable_bits;
3156 if (debug_value == 0) /* no output */
3157 return 0;
3158 /* set low N bits */
3159 return (1 << debug_value) - 1;
3160 }
3161
3162 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3163 {
3164 spin_lock(&txq->_xmit_lock);
3165 txq->xmit_lock_owner = cpu;
3166 }
3167
3168 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3169 {
3170 spin_lock_bh(&txq->_xmit_lock);
3171 txq->xmit_lock_owner = smp_processor_id();
3172 }
3173
3174 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3175 {
3176 bool ok = spin_trylock(&txq->_xmit_lock);
3177 if (likely(ok))
3178 txq->xmit_lock_owner = smp_processor_id();
3179 return ok;
3180 }
3181
3182 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3183 {
3184 txq->xmit_lock_owner = -1;
3185 spin_unlock(&txq->_xmit_lock);
3186 }
3187
3188 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3189 {
3190 txq->xmit_lock_owner = -1;
3191 spin_unlock_bh(&txq->_xmit_lock);
3192 }
3193
3194 static inline void txq_trans_update(struct netdev_queue *txq)
3195 {
3196 if (txq->xmit_lock_owner != -1)
3197 txq->trans_start = jiffies;
3198 }
3199
3200 /**
3201 * netif_tx_lock - grab network device transmit lock
3202 * @dev: network device
3203 *
3204 * Get network device transmit lock
3205 */
3206 static inline void netif_tx_lock(struct net_device *dev)
3207 {
3208 unsigned int i;
3209 int cpu;
3210
3211 spin_lock(&dev->tx_global_lock);
3212 cpu = smp_processor_id();
3213 for (i = 0; i < dev->num_tx_queues; i++) {
3214 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3215
3216 /* We are the only thread of execution doing a
3217 * freeze, but we have to grab the _xmit_lock in
3218 * order to synchronize with threads which are in
3219 * the ->hard_start_xmit() handler and already
3220 * checked the frozen bit.
3221 */
3222 __netif_tx_lock(txq, cpu);
3223 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3224 __netif_tx_unlock(txq);
3225 }
3226 }
3227
3228 static inline void netif_tx_lock_bh(struct net_device *dev)
3229 {
3230 local_bh_disable();
3231 netif_tx_lock(dev);
3232 }
3233
3234 static inline void netif_tx_unlock(struct net_device *dev)
3235 {
3236 unsigned int i;
3237
3238 for (i = 0; i < dev->num_tx_queues; i++) {
3239 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3240
3241 /* No need to grab the _xmit_lock here. If the
3242 * queue is not stopped for another reason, we
3243 * force a schedule.
3244 */
3245 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3246 netif_schedule_queue(txq);
3247 }
3248 spin_unlock(&dev->tx_global_lock);
3249 }
3250
3251 static inline void netif_tx_unlock_bh(struct net_device *dev)
3252 {
3253 netif_tx_unlock(dev);
3254 local_bh_enable();
3255 }
3256
3257 #define HARD_TX_LOCK(dev, txq, cpu) { \
3258 if ((dev->features & NETIF_F_LLTX) == 0) { \
3259 __netif_tx_lock(txq, cpu); \
3260 } \
3261 }
3262
3263 #define HARD_TX_TRYLOCK(dev, txq) \
3264 (((dev->features & NETIF_F_LLTX) == 0) ? \
3265 __netif_tx_trylock(txq) : \
3266 true )
3267
3268 #define HARD_TX_UNLOCK(dev, txq) { \
3269 if ((dev->features & NETIF_F_LLTX) == 0) { \
3270 __netif_tx_unlock(txq); \
3271 } \
3272 }
3273
3274 static inline void netif_tx_disable(struct net_device *dev)
3275 {
3276 unsigned int i;
3277 int cpu;
3278
3279 local_bh_disable();
3280 cpu = smp_processor_id();
3281 for (i = 0; i < dev->num_tx_queues; i++) {
3282 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3283
3284 __netif_tx_lock(txq, cpu);
3285 netif_tx_stop_queue(txq);
3286 __netif_tx_unlock(txq);
3287 }
3288 local_bh_enable();
3289 }
3290
3291 static inline void netif_addr_lock(struct net_device *dev)
3292 {
3293 spin_lock(&dev->addr_list_lock);
3294 }
3295
3296 static inline void netif_addr_lock_nested(struct net_device *dev)
3297 {
3298 int subclass = SINGLE_DEPTH_NESTING;
3299
3300 if (dev->netdev_ops->ndo_get_lock_subclass)
3301 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3302
3303 spin_lock_nested(&dev->addr_list_lock, subclass);
3304 }
3305
3306 static inline void netif_addr_lock_bh(struct net_device *dev)
3307 {
3308 spin_lock_bh(&dev->addr_list_lock);
3309 }
3310
3311 static inline void netif_addr_unlock(struct net_device *dev)
3312 {
3313 spin_unlock(&dev->addr_list_lock);
3314 }
3315
3316 static inline void netif_addr_unlock_bh(struct net_device *dev)
3317 {
3318 spin_unlock_bh(&dev->addr_list_lock);
3319 }
3320
3321 /*
3322 * dev_addrs walker. Should be used only for read access. Call with
3323 * rcu_read_lock held.
3324 */
3325 #define for_each_dev_addr(dev, ha) \
3326 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3327
3328 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3329
3330 void ether_setup(struct net_device *dev);
3331
3332 /* Support for loadable net-drivers */
3333 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3334 unsigned char name_assign_type,
3335 void (*setup)(struct net_device *),
3336 unsigned int txqs, unsigned int rxqs);
3337 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3338 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3339
3340 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3341 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3342 count)
3343
3344 int register_netdev(struct net_device *dev);
3345 void unregister_netdev(struct net_device *dev);
3346
3347 /* General hardware address lists handling functions */
3348 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3349 struct netdev_hw_addr_list *from_list, int addr_len);
3350 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3351 struct netdev_hw_addr_list *from_list, int addr_len);
3352 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3353 struct net_device *dev,
3354 int (*sync)(struct net_device *, const unsigned char *),
3355 int (*unsync)(struct net_device *,
3356 const unsigned char *));
3357 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3358 struct net_device *dev,
3359 int (*unsync)(struct net_device *,
3360 const unsigned char *));
3361 void __hw_addr_init(struct netdev_hw_addr_list *list);
3362
3363 /* Functions used for device addresses handling */
3364 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3365 unsigned char addr_type);
3366 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3367 unsigned char addr_type);
3368 void dev_addr_flush(struct net_device *dev);
3369 int dev_addr_init(struct net_device *dev);
3370
3371 /* Functions used for unicast addresses handling */
3372 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3373 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3374 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3375 int dev_uc_sync(struct net_device *to, struct net_device *from);
3376 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3377 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3378 void dev_uc_flush(struct net_device *dev);
3379 void dev_uc_init(struct net_device *dev);
3380
3381 /**
3382 * __dev_uc_sync - Synchonize device's unicast list
3383 * @dev: device to sync
3384 * @sync: function to call if address should be added
3385 * @unsync: function to call if address should be removed
3386 *
3387 * Add newly added addresses to the interface, and release
3388 * addresses that have been deleted.
3389 **/
3390 static inline int __dev_uc_sync(struct net_device *dev,
3391 int (*sync)(struct net_device *,
3392 const unsigned char *),
3393 int (*unsync)(struct net_device *,
3394 const unsigned char *))
3395 {
3396 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3397 }
3398
3399 /**
3400 * __dev_uc_unsync - Remove synchronized addresses from device
3401 * @dev: device to sync
3402 * @unsync: function to call if address should be removed
3403 *
3404 * Remove all addresses that were added to the device by dev_uc_sync().
3405 **/
3406 static inline void __dev_uc_unsync(struct net_device *dev,
3407 int (*unsync)(struct net_device *,
3408 const unsigned char *))
3409 {
3410 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3411 }
3412
3413 /* Functions used for multicast addresses handling */
3414 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3415 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3416 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3417 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3418 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3419 int dev_mc_sync(struct net_device *to, struct net_device *from);
3420 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3421 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3422 void dev_mc_flush(struct net_device *dev);
3423 void dev_mc_init(struct net_device *dev);
3424
3425 /**
3426 * __dev_mc_sync - Synchonize device's multicast list
3427 * @dev: device to sync
3428 * @sync: function to call if address should be added
3429 * @unsync: function to call if address should be removed
3430 *
3431 * Add newly added addresses to the interface, and release
3432 * addresses that have been deleted.
3433 **/
3434 static inline int __dev_mc_sync(struct net_device *dev,
3435 int (*sync)(struct net_device *,
3436 const unsigned char *),
3437 int (*unsync)(struct net_device *,
3438 const unsigned char *))
3439 {
3440 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3441 }
3442
3443 /**
3444 * __dev_mc_unsync - Remove synchronized addresses from device
3445 * @dev: device to sync
3446 * @unsync: function to call if address should be removed
3447 *
3448 * Remove all addresses that were added to the device by dev_mc_sync().
3449 **/
3450 static inline void __dev_mc_unsync(struct net_device *dev,
3451 int (*unsync)(struct net_device *,
3452 const unsigned char *))
3453 {
3454 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3455 }
3456
3457 /* Functions used for secondary unicast and multicast support */
3458 void dev_set_rx_mode(struct net_device *dev);
3459 void __dev_set_rx_mode(struct net_device *dev);
3460 int dev_set_promiscuity(struct net_device *dev, int inc);
3461 int dev_set_allmulti(struct net_device *dev, int inc);
3462 void netdev_state_change(struct net_device *dev);
3463 void netdev_notify_peers(struct net_device *dev);
3464 void netdev_features_change(struct net_device *dev);
3465 /* Load a device via the kmod */
3466 void dev_load(struct net *net, const char *name);
3467 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3468 struct rtnl_link_stats64 *storage);
3469 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3470 const struct net_device_stats *netdev_stats);
3471
3472 extern int netdev_max_backlog;
3473 extern int netdev_tstamp_prequeue;
3474 extern int weight_p;
3475 extern int bpf_jit_enable;
3476
3477 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3478 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3479 struct list_head **iter);
3480 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3481 struct list_head **iter);
3482
3483 /* iterate through upper list, must be called under RCU read lock */
3484 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3485 for (iter = &(dev)->adj_list.upper, \
3486 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3487 updev; \
3488 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3489
3490 /* iterate through upper list, must be called under RCU read lock */
3491 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3492 for (iter = &(dev)->all_adj_list.upper, \
3493 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3494 updev; \
3495 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3496
3497 void *netdev_lower_get_next_private(struct net_device *dev,
3498 struct list_head **iter);
3499 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3500 struct list_head **iter);
3501
3502 #define netdev_for_each_lower_private(dev, priv, iter) \
3503 for (iter = (dev)->adj_list.lower.next, \
3504 priv = netdev_lower_get_next_private(dev, &(iter)); \
3505 priv; \
3506 priv = netdev_lower_get_next_private(dev, &(iter)))
3507
3508 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3509 for (iter = &(dev)->adj_list.lower, \
3510 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3511 priv; \
3512 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3513
3514 void *netdev_lower_get_next(struct net_device *dev,
3515 struct list_head **iter);
3516 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3517 for (iter = &(dev)->adj_list.lower, \
3518 ldev = netdev_lower_get_next(dev, &(iter)); \
3519 ldev; \
3520 ldev = netdev_lower_get_next(dev, &(iter)))
3521
3522 void *netdev_adjacent_get_private(struct list_head *adj_list);
3523 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3524 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3525 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3526 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3527 int netdev_master_upper_dev_link(struct net_device *dev,
3528 struct net_device *upper_dev);
3529 int netdev_master_upper_dev_link_private(struct net_device *dev,
3530 struct net_device *upper_dev,
3531 void *private);
3532 void netdev_upper_dev_unlink(struct net_device *dev,
3533 struct net_device *upper_dev);
3534 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3535 void *netdev_lower_dev_get_private(struct net_device *dev,
3536 struct net_device *lower_dev);
3537
3538 /* RSS keys are 40 or 52 bytes long */
3539 #define NETDEV_RSS_KEY_LEN 52
3540 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN];
3541 void netdev_rss_key_fill(void *buffer, size_t len);
3542
3543 int dev_get_nest_level(struct net_device *dev,
3544 bool (*type_check)(struct net_device *dev));
3545 int skb_checksum_help(struct sk_buff *skb);
3546 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3547 netdev_features_t features, bool tx_path);
3548 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3549 netdev_features_t features);
3550
3551 struct netdev_bonding_info {
3552 ifslave slave;
3553 ifbond master;
3554 };
3555
3556 struct netdev_notifier_bonding_info {
3557 struct netdev_notifier_info info; /* must be first */
3558 struct netdev_bonding_info bonding_info;
3559 };
3560
3561 void netdev_bonding_info_change(struct net_device *dev,
3562 struct netdev_bonding_info *bonding_info);
3563
3564 static inline
3565 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3566 {
3567 return __skb_gso_segment(skb, features, true);
3568 }
3569 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3570
3571 static inline bool can_checksum_protocol(netdev_features_t features,
3572 __be16 protocol)
3573 {
3574 return ((features & NETIF_F_GEN_CSUM) ||
3575 ((features & NETIF_F_V4_CSUM) &&
3576 protocol == htons(ETH_P_IP)) ||
3577 ((features & NETIF_F_V6_CSUM) &&
3578 protocol == htons(ETH_P_IPV6)) ||
3579 ((features & NETIF_F_FCOE_CRC) &&
3580 protocol == htons(ETH_P_FCOE)));
3581 }
3582
3583 #ifdef CONFIG_BUG
3584 void netdev_rx_csum_fault(struct net_device *dev);
3585 #else
3586 static inline void netdev_rx_csum_fault(struct net_device *dev)
3587 {
3588 }
3589 #endif
3590 /* rx skb timestamps */
3591 void net_enable_timestamp(void);
3592 void net_disable_timestamp(void);
3593
3594 #ifdef CONFIG_PROC_FS
3595 int __init dev_proc_init(void);
3596 #else
3597 #define dev_proc_init() 0
3598 #endif
3599
3600 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3601 struct sk_buff *skb, struct net_device *dev,
3602 bool more)
3603 {
3604 skb->xmit_more = more ? 1 : 0;
3605 return ops->ndo_start_xmit(skb, dev);
3606 }
3607
3608 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3609 struct netdev_queue *txq, bool more)
3610 {
3611 const struct net_device_ops *ops = dev->netdev_ops;
3612 int rc;
3613
3614 rc = __netdev_start_xmit(ops, skb, dev, more);
3615 if (rc == NETDEV_TX_OK)
3616 txq_trans_update(txq);
3617
3618 return rc;
3619 }
3620
3621 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3622 const void *ns);
3623 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3624 const void *ns);
3625
3626 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3627 {
3628 return netdev_class_create_file_ns(class_attr, NULL);
3629 }
3630
3631 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3632 {
3633 netdev_class_remove_file_ns(class_attr, NULL);
3634 }
3635
3636 extern struct kobj_ns_type_operations net_ns_type_operations;
3637
3638 const char *netdev_drivername(const struct net_device *dev);
3639
3640 void linkwatch_run_queue(void);
3641
3642 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3643 netdev_features_t f2)
3644 {
3645 if (f1 & NETIF_F_GEN_CSUM)
3646 f1 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3647 if (f2 & NETIF_F_GEN_CSUM)
3648 f2 |= (NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3649 f1 &= f2;
3650 if (f1 & NETIF_F_GEN_CSUM)
3651 f1 &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
3652
3653 return f1;
3654 }
3655
3656 static inline netdev_features_t netdev_get_wanted_features(
3657 struct net_device *dev)
3658 {
3659 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3660 }
3661 netdev_features_t netdev_increment_features(netdev_features_t all,
3662 netdev_features_t one, netdev_features_t mask);
3663
3664 /* Allow TSO being used on stacked device :
3665 * Performing the GSO segmentation before last device
3666 * is a performance improvement.
3667 */
3668 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3669 netdev_features_t mask)
3670 {
3671 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3672 }
3673
3674 int __netdev_update_features(struct net_device *dev);
3675 void netdev_update_features(struct net_device *dev);
3676 void netdev_change_features(struct net_device *dev);
3677
3678 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3679 struct net_device *dev);
3680
3681 netdev_features_t netif_skb_features(struct sk_buff *skb);
3682
3683 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3684 {
3685 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3686
3687 /* check flags correspondence */
3688 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3689 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3690 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3691 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3692 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3693 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3694 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3695 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3696 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3697 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3698 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3699 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3700 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3701
3702 return (features & feature) == feature;
3703 }
3704
3705 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3706 {
3707 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3708 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3709 }
3710
3711 static inline bool netif_needs_gso(struct net_device *dev, struct sk_buff *skb,
3712 netdev_features_t features)
3713 {
3714 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
3715 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
3716 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
3717 }
3718
3719 static inline void netif_set_gso_max_size(struct net_device *dev,
3720 unsigned int size)
3721 {
3722 dev->gso_max_size = size;
3723 }
3724
3725 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
3726 int pulled_hlen, u16 mac_offset,
3727 int mac_len)
3728 {
3729 skb->protocol = protocol;
3730 skb->encapsulation = 1;
3731 skb_push(skb, pulled_hlen);
3732 skb_reset_transport_header(skb);
3733 skb->mac_header = mac_offset;
3734 skb->network_header = skb->mac_header + mac_len;
3735 skb->mac_len = mac_len;
3736 }
3737
3738 static inline bool netif_is_macvlan(struct net_device *dev)
3739 {
3740 return dev->priv_flags & IFF_MACVLAN;
3741 }
3742
3743 static inline bool netif_is_macvlan_port(struct net_device *dev)
3744 {
3745 return dev->priv_flags & IFF_MACVLAN_PORT;
3746 }
3747
3748 static inline bool netif_is_ipvlan(struct net_device *dev)
3749 {
3750 return dev->priv_flags & IFF_IPVLAN_SLAVE;
3751 }
3752
3753 static inline bool netif_is_ipvlan_port(struct net_device *dev)
3754 {
3755 return dev->priv_flags & IFF_IPVLAN_MASTER;
3756 }
3757
3758 static inline bool netif_is_bond_master(struct net_device *dev)
3759 {
3760 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
3761 }
3762
3763 static inline bool netif_is_bond_slave(struct net_device *dev)
3764 {
3765 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
3766 }
3767
3768 static inline bool netif_supports_nofcs(struct net_device *dev)
3769 {
3770 return dev->priv_flags & IFF_SUPP_NOFCS;
3771 }
3772
3773 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
3774 static inline void netif_keep_dst(struct net_device *dev)
3775 {
3776 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
3777 }
3778
3779 extern struct pernet_operations __net_initdata loopback_net_ops;
3780
3781 /* Logging, debugging and troubleshooting/diagnostic helpers. */
3782
3783 /* netdev_printk helpers, similar to dev_printk */
3784
3785 static inline const char *netdev_name(const struct net_device *dev)
3786 {
3787 if (!dev->name[0] || strchr(dev->name, '%'))
3788 return "(unnamed net_device)";
3789 return dev->name;
3790 }
3791
3792 static inline const char *netdev_reg_state(const struct net_device *dev)
3793 {
3794 switch (dev->reg_state) {
3795 case NETREG_UNINITIALIZED: return " (uninitialized)";
3796 case NETREG_REGISTERED: return "";
3797 case NETREG_UNREGISTERING: return " (unregistering)";
3798 case NETREG_UNREGISTERED: return " (unregistered)";
3799 case NETREG_RELEASED: return " (released)";
3800 case NETREG_DUMMY: return " (dummy)";
3801 }
3802
3803 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
3804 return " (unknown)";
3805 }
3806
3807 __printf(3, 4)
3808 void netdev_printk(const char *level, const struct net_device *dev,
3809 const char *format, ...);
3810 __printf(2, 3)
3811 void netdev_emerg(const struct net_device *dev, const char *format, ...);
3812 __printf(2, 3)
3813 void netdev_alert(const struct net_device *dev, const char *format, ...);
3814 __printf(2, 3)
3815 void netdev_crit(const struct net_device *dev, const char *format, ...);
3816 __printf(2, 3)
3817 void netdev_err(const struct net_device *dev, const char *format, ...);
3818 __printf(2, 3)
3819 void netdev_warn(const struct net_device *dev, const char *format, ...);
3820 __printf(2, 3)
3821 void netdev_notice(const struct net_device *dev, const char *format, ...);
3822 __printf(2, 3)
3823 void netdev_info(const struct net_device *dev, const char *format, ...);
3824
3825 #define MODULE_ALIAS_NETDEV(device) \
3826 MODULE_ALIAS("netdev-" device)
3827
3828 #if defined(CONFIG_DYNAMIC_DEBUG)
3829 #define netdev_dbg(__dev, format, args...) \
3830 do { \
3831 dynamic_netdev_dbg(__dev, format, ##args); \
3832 } while (0)
3833 #elif defined(DEBUG)
3834 #define netdev_dbg(__dev, format, args...) \
3835 netdev_printk(KERN_DEBUG, __dev, format, ##args)
3836 #else
3837 #define netdev_dbg(__dev, format, args...) \
3838 ({ \
3839 if (0) \
3840 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
3841 })
3842 #endif
3843
3844 #if defined(VERBOSE_DEBUG)
3845 #define netdev_vdbg netdev_dbg
3846 #else
3847
3848 #define netdev_vdbg(dev, format, args...) \
3849 ({ \
3850 if (0) \
3851 netdev_printk(KERN_DEBUG, dev, format, ##args); \
3852 0; \
3853 })
3854 #endif
3855
3856 /*
3857 * netdev_WARN() acts like dev_printk(), but with the key difference
3858 * of using a WARN/WARN_ON to get the message out, including the
3859 * file/line information and a backtrace.
3860 */
3861 #define netdev_WARN(dev, format, args...) \
3862 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
3863 netdev_reg_state(dev), ##args)
3864
3865 /* netif printk helpers, similar to netdev_printk */
3866
3867 #define netif_printk(priv, type, level, dev, fmt, args...) \
3868 do { \
3869 if (netif_msg_##type(priv)) \
3870 netdev_printk(level, (dev), fmt, ##args); \
3871 } while (0)
3872
3873 #define netif_level(level, priv, type, dev, fmt, args...) \
3874 do { \
3875 if (netif_msg_##type(priv)) \
3876 netdev_##level(dev, fmt, ##args); \
3877 } while (0)
3878
3879 #define netif_emerg(priv, type, dev, fmt, args...) \
3880 netif_level(emerg, priv, type, dev, fmt, ##args)
3881 #define netif_alert(priv, type, dev, fmt, args...) \
3882 netif_level(alert, priv, type, dev, fmt, ##args)
3883 #define netif_crit(priv, type, dev, fmt, args...) \
3884 netif_level(crit, priv, type, dev, fmt, ##args)
3885 #define netif_err(priv, type, dev, fmt, args...) \
3886 netif_level(err, priv, type, dev, fmt, ##args)
3887 #define netif_warn(priv, type, dev, fmt, args...) \
3888 netif_level(warn, priv, type, dev, fmt, ##args)
3889 #define netif_notice(priv, type, dev, fmt, args...) \
3890 netif_level(notice, priv, type, dev, fmt, ##args)
3891 #define netif_info(priv, type, dev, fmt, args...) \
3892 netif_level(info, priv, type, dev, fmt, ##args)
3893
3894 #if defined(CONFIG_DYNAMIC_DEBUG)
3895 #define netif_dbg(priv, type, netdev, format, args...) \
3896 do { \
3897 if (netif_msg_##type(priv)) \
3898 dynamic_netdev_dbg(netdev, format, ##args); \
3899 } while (0)
3900 #elif defined(DEBUG)
3901 #define netif_dbg(priv, type, dev, format, args...) \
3902 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
3903 #else
3904 #define netif_dbg(priv, type, dev, format, args...) \
3905 ({ \
3906 if (0) \
3907 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3908 0; \
3909 })
3910 #endif
3911
3912 #if defined(VERBOSE_DEBUG)
3913 #define netif_vdbg netif_dbg
3914 #else
3915 #define netif_vdbg(priv, type, dev, format, args...) \
3916 ({ \
3917 if (0) \
3918 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
3919 0; \
3920 })
3921 #endif
3922
3923 /*
3924 * The list of packet types we will receive (as opposed to discard)
3925 * and the routines to invoke.
3926 *
3927 * Why 16. Because with 16 the only overlap we get on a hash of the
3928 * low nibble of the protocol value is RARP/SNAP/X.25.
3929 *
3930 * NOTE: That is no longer true with the addition of VLAN tags. Not
3931 * sure which should go first, but I bet it won't make much
3932 * difference if we are running VLANs. The good news is that
3933 * this protocol won't be in the list unless compiled in, so
3934 * the average user (w/out VLANs) will not be adversely affected.
3935 * --BLG
3936 *
3937 * 0800 IP
3938 * 8100 802.1Q VLAN
3939 * 0001 802.3
3940 * 0002 AX.25
3941 * 0004 802.2
3942 * 8035 RARP
3943 * 0005 SNAP
3944 * 0805 X.25
3945 * 0806 ARP
3946 * 8137 IPX
3947 * 0009 Localtalk
3948 * 86DD IPv6
3949 */
3950 #define PTYPE_HASH_SIZE (16)
3951 #define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
3952
3953 #endif /* _LINUX_NETDEVICE_H */
This page took 0.111567 seconds and 5 git commands to generate.