xref: /linux-6.15/include/linux/netdevice.h (revision 0e9f9fd2)
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, <[email protected]>
12  *		Corey Minyard <[email protected]>
13  *		Donald J. Becker, <[email protected]>
14  *		Alan Cox, <[email protected]>
15  *		Bjorn Ekwall. <[email protected]>
16  *              Pekka Riikonen <[email protected]>
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/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35 
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41 
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49 
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
55 #include <linux/hashtable.h>
56 
57 struct netpoll_info;
58 struct device;
59 struct phy_device;
60 /* 802.11 specific */
61 struct wireless_dev;
62 /* 802.15.4 specific */
63 struct wpan_dev;
64 struct mpls_dev;
65 /* UDP Tunnel offloads */
66 struct udp_tunnel_info;
67 struct bpf_prog;
68 
69 void netdev_set_default_ethtool_ops(struct net_device *dev,
70 				    const struct ethtool_ops *ops);
71 
72 /* Backlog congestion levels */
73 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
74 #define NET_RX_DROP		1	/* packet dropped */
75 
76 /*
77  * Transmit return codes: transmit return codes originate from three different
78  * namespaces:
79  *
80  * - qdisc return codes
81  * - driver transmit return codes
82  * - errno values
83  *
84  * Drivers are allowed to return any one of those in their hard_start_xmit()
85  * function. Real network devices commonly used with qdiscs should only return
86  * the driver transmit return codes though - when qdiscs are used, the actual
87  * transmission happens asynchronously, so the value is not propagated to
88  * higher layers. Virtual network devices transmit synchronously; in this case
89  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
90  * others are propagated to higher layers.
91  */
92 
93 /* qdisc ->enqueue() return codes. */
94 #define NET_XMIT_SUCCESS	0x00
95 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
96 #define NET_XMIT_CN		0x02	/* congestion notification	*/
97 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
98 
99 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
100  * indicates that the device will soon be dropping packets, or already drops
101  * some packets of the same priority; prompting us to send less aggressively. */
102 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
103 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
104 
105 /* Driver transmit return codes */
106 #define NETDEV_TX_MASK		0xf0
107 
108 enum netdev_tx {
109 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
110 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
111 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
112 };
113 typedef enum netdev_tx netdev_tx_t;
114 
115 /*
116  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
117  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
118  */
119 static inline bool dev_xmit_complete(int rc)
120 {
121 	/*
122 	 * Positive cases with an skb consumed by a driver:
123 	 * - successful transmission (rc == NETDEV_TX_OK)
124 	 * - error while transmitting (rc < 0)
125 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
126 	 */
127 	if (likely(rc < NET_XMIT_MASK))
128 		return true;
129 
130 	return false;
131 }
132 
133 /*
134  *	Compute the worst-case header length according to the protocols
135  *	used.
136  */
137 
138 #if defined(CONFIG_HYPERV_NET)
139 # define LL_MAX_HEADER 128
140 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
141 # if defined(CONFIG_MAC80211_MESH)
142 #  define LL_MAX_HEADER 128
143 # else
144 #  define LL_MAX_HEADER 96
145 # endif
146 #else
147 # define LL_MAX_HEADER 32
148 #endif
149 
150 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
151     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
152 #define MAX_HEADER LL_MAX_HEADER
153 #else
154 #define MAX_HEADER (LL_MAX_HEADER + 48)
155 #endif
156 
157 /*
158  *	Old network device statistics. Fields are native words
159  *	(unsigned long) so they can be read and written atomically.
160  */
161 
162 struct net_device_stats {
163 	unsigned long	rx_packets;
164 	unsigned long	tx_packets;
165 	unsigned long	rx_bytes;
166 	unsigned long	tx_bytes;
167 	unsigned long	rx_errors;
168 	unsigned long	tx_errors;
169 	unsigned long	rx_dropped;
170 	unsigned long	tx_dropped;
171 	unsigned long	multicast;
172 	unsigned long	collisions;
173 	unsigned long	rx_length_errors;
174 	unsigned long	rx_over_errors;
175 	unsigned long	rx_crc_errors;
176 	unsigned long	rx_frame_errors;
177 	unsigned long	rx_fifo_errors;
178 	unsigned long	rx_missed_errors;
179 	unsigned long	tx_aborted_errors;
180 	unsigned long	tx_carrier_errors;
181 	unsigned long	tx_fifo_errors;
182 	unsigned long	tx_heartbeat_errors;
183 	unsigned long	tx_window_errors;
184 	unsigned long	rx_compressed;
185 	unsigned long	tx_compressed;
186 };
187 
188 
189 #include <linux/cache.h>
190 #include <linux/skbuff.h>
191 
192 #ifdef CONFIG_RPS
193 #include <linux/static_key.h>
194 extern struct static_key rps_needed;
195 extern struct static_key rfs_needed;
196 #endif
197 
198 struct neighbour;
199 struct neigh_parms;
200 struct sk_buff;
201 
202 struct netdev_hw_addr {
203 	struct list_head	list;
204 	unsigned char		addr[MAX_ADDR_LEN];
205 	unsigned char		type;
206 #define NETDEV_HW_ADDR_T_LAN		1
207 #define NETDEV_HW_ADDR_T_SAN		2
208 #define NETDEV_HW_ADDR_T_SLAVE		3
209 #define NETDEV_HW_ADDR_T_UNICAST	4
210 #define NETDEV_HW_ADDR_T_MULTICAST	5
211 	bool			global_use;
212 	int			sync_cnt;
213 	int			refcount;
214 	int			synced;
215 	struct rcu_head		rcu_head;
216 };
217 
218 struct netdev_hw_addr_list {
219 	struct list_head	list;
220 	int			count;
221 };
222 
223 #define netdev_hw_addr_list_count(l) ((l)->count)
224 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
225 #define netdev_hw_addr_list_for_each(ha, l) \
226 	list_for_each_entry(ha, &(l)->list, list)
227 
228 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
229 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
230 #define netdev_for_each_uc_addr(ha, dev) \
231 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
232 
233 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
234 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
235 #define netdev_for_each_mc_addr(ha, dev) \
236 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
237 
238 struct hh_cache {
239 	u16		hh_len;
240 	u16		__pad;
241 	seqlock_t	hh_lock;
242 
243 	/* cached hardware header; allow for machine alignment needs.        */
244 #define HH_DATA_MOD	16
245 #define HH_DATA_OFF(__len) \
246 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
247 #define HH_DATA_ALIGN(__len) \
248 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
249 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
250 };
251 
252 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
253  * Alternative is:
254  *   dev->hard_header_len ? (dev->hard_header_len +
255  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
256  *
257  * We could use other alignment values, but we must maintain the
258  * relationship HH alignment <= LL alignment.
259  */
260 #define LL_RESERVED_SPACE(dev) \
261 	((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
263 	((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 
265 struct header_ops {
266 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
267 			   unsigned short type, const void *daddr,
268 			   const void *saddr, unsigned int len);
269 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
270 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
271 	void	(*cache_update)(struct hh_cache *hh,
272 				const struct net_device *dev,
273 				const unsigned char *haddr);
274 	bool	(*validate)(const char *ll_header, unsigned int len);
275 };
276 
277 /* These flag bits are private to the generic network queueing
278  * layer; they may not be explicitly referenced by any other
279  * code.
280  */
281 
282 enum netdev_state_t {
283 	__LINK_STATE_START,
284 	__LINK_STATE_PRESENT,
285 	__LINK_STATE_NOCARRIER,
286 	__LINK_STATE_LINKWATCH_PENDING,
287 	__LINK_STATE_DORMANT,
288 };
289 
290 
291 /*
292  * This structure holds boot-time configured netdevice settings. They
293  * are then used in the device probing.
294  */
295 struct netdev_boot_setup {
296 	char name[IFNAMSIZ];
297 	struct ifmap map;
298 };
299 #define NETDEV_BOOT_SETUP_MAX 8
300 
301 int __init netdev_boot_setup(char *str);
302 
303 /*
304  * Structure for NAPI scheduling similar to tasklet but with weighting
305  */
306 struct napi_struct {
307 	/* The poll_list must only be managed by the entity which
308 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
309 	 * whoever atomically sets that bit can add this napi_struct
310 	 * to the per-CPU poll_list, and whoever clears that bit
311 	 * can remove from the list right before clearing the bit.
312 	 */
313 	struct list_head	poll_list;
314 
315 	unsigned long		state;
316 	int			weight;
317 	unsigned int		gro_count;
318 	int			(*poll)(struct napi_struct *, int);
319 #ifdef CONFIG_NETPOLL
320 	int			poll_owner;
321 #endif
322 	struct net_device	*dev;
323 	struct sk_buff		*gro_list;
324 	struct sk_buff		*skb;
325 	struct hrtimer		timer;
326 	struct list_head	dev_list;
327 	struct hlist_node	napi_hash_node;
328 	unsigned int		napi_id;
329 };
330 
331 enum {
332 	NAPI_STATE_SCHED,	/* Poll is scheduled */
333 	NAPI_STATE_DISABLE,	/* Disable pending */
334 	NAPI_STATE_NPSVC,	/* Netpoll - don't dequeue from poll_list */
335 	NAPI_STATE_HASHED,	/* In NAPI hash (busy polling possible) */
336 	NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
337 	NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
338 };
339 
340 enum {
341 	NAPIF_STATE_SCHED	 = (1UL << NAPI_STATE_SCHED),
342 	NAPIF_STATE_DISABLE	 = (1UL << NAPI_STATE_DISABLE),
343 	NAPIF_STATE_NPSVC	 = (1UL << NAPI_STATE_NPSVC),
344 	NAPIF_STATE_HASHED	 = (1UL << NAPI_STATE_HASHED),
345 	NAPIF_STATE_NO_BUSY_POLL = (1UL << NAPI_STATE_NO_BUSY_POLL),
346 	NAPIF_STATE_IN_BUSY_POLL = (1UL << NAPI_STATE_IN_BUSY_POLL),
347 };
348 
349 enum gro_result {
350 	GRO_MERGED,
351 	GRO_MERGED_FREE,
352 	GRO_HELD,
353 	GRO_NORMAL,
354 	GRO_DROP,
355 };
356 typedef enum gro_result gro_result_t;
357 
358 /*
359  * enum rx_handler_result - Possible return values for rx_handlers.
360  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
361  * further.
362  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
363  * case skb->dev was changed by rx_handler.
364  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
365  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
366  *
367  * rx_handlers are functions called from inside __netif_receive_skb(), to do
368  * special processing of the skb, prior to delivery to protocol handlers.
369  *
370  * Currently, a net_device can only have a single rx_handler registered. Trying
371  * to register a second rx_handler will return -EBUSY.
372  *
373  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
374  * To unregister a rx_handler on a net_device, use
375  * netdev_rx_handler_unregister().
376  *
377  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
378  * do with the skb.
379  *
380  * If the rx_handler consumed the skb in some way, it should return
381  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
382  * the skb to be delivered in some other way.
383  *
384  * If the rx_handler changed skb->dev, to divert the skb to another
385  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
386  * new device will be called if it exists.
387  *
388  * If the rx_handler decides the skb should be ignored, it should return
389  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
390  * are registered on exact device (ptype->dev == skb->dev).
391  *
392  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
393  * delivered, it should return RX_HANDLER_PASS.
394  *
395  * A device without a registered rx_handler will behave as if rx_handler
396  * returned RX_HANDLER_PASS.
397  */
398 
399 enum rx_handler_result {
400 	RX_HANDLER_CONSUMED,
401 	RX_HANDLER_ANOTHER,
402 	RX_HANDLER_EXACT,
403 	RX_HANDLER_PASS,
404 };
405 typedef enum rx_handler_result rx_handler_result_t;
406 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
407 
408 void __napi_schedule(struct napi_struct *n);
409 void __napi_schedule_irqoff(struct napi_struct *n);
410 
411 static inline bool napi_disable_pending(struct napi_struct *n)
412 {
413 	return test_bit(NAPI_STATE_DISABLE, &n->state);
414 }
415 
416 /**
417  *	napi_schedule_prep - check if NAPI can be scheduled
418  *	@n: NAPI context
419  *
420  * Test if NAPI routine is already running, and if not mark
421  * it as running.  This is used as a condition variable to
422  * insure only one NAPI poll instance runs.  We also make
423  * sure there is no pending NAPI disable.
424  */
425 static inline bool napi_schedule_prep(struct napi_struct *n)
426 {
427 	return !napi_disable_pending(n) &&
428 		!test_and_set_bit(NAPI_STATE_SCHED, &n->state);
429 }
430 
431 /**
432  *	napi_schedule - schedule NAPI poll
433  *	@n: NAPI context
434  *
435  * Schedule NAPI poll routine to be called if it is not already
436  * running.
437  */
438 static inline void napi_schedule(struct napi_struct *n)
439 {
440 	if (napi_schedule_prep(n))
441 		__napi_schedule(n);
442 }
443 
444 /**
445  *	napi_schedule_irqoff - schedule NAPI poll
446  *	@n: NAPI context
447  *
448  * Variant of napi_schedule(), assuming hard irqs are masked.
449  */
450 static inline void napi_schedule_irqoff(struct napi_struct *n)
451 {
452 	if (napi_schedule_prep(n))
453 		__napi_schedule_irqoff(n);
454 }
455 
456 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
457 static inline bool napi_reschedule(struct napi_struct *napi)
458 {
459 	if (napi_schedule_prep(napi)) {
460 		__napi_schedule(napi);
461 		return true;
462 	}
463 	return false;
464 }
465 
466 bool __napi_complete(struct napi_struct *n);
467 bool napi_complete_done(struct napi_struct *n, int work_done);
468 /**
469  *	napi_complete - NAPI processing complete
470  *	@n: NAPI context
471  *
472  * Mark NAPI processing as complete.
473  * Consider using napi_complete_done() instead.
474  * Return false if device should avoid rearming interrupts.
475  */
476 static inline bool napi_complete(struct napi_struct *n)
477 {
478 	return napi_complete_done(n, 0);
479 }
480 
481 /**
482  *	napi_hash_del - remove a NAPI from global table
483  *	@napi: NAPI context
484  *
485  * Warning: caller must observe RCU grace period
486  * before freeing memory containing @napi, if
487  * this function returns true.
488  * Note: core networking stack automatically calls it
489  * from netif_napi_del().
490  * Drivers might want to call this helper to combine all
491  * the needed RCU grace periods into a single one.
492  */
493 bool napi_hash_del(struct napi_struct *napi);
494 
495 /**
496  *	napi_disable - prevent NAPI from scheduling
497  *	@n: NAPI context
498  *
499  * Stop NAPI from being scheduled on this context.
500  * Waits till any outstanding processing completes.
501  */
502 void napi_disable(struct napi_struct *n);
503 
504 /**
505  *	napi_enable - enable NAPI scheduling
506  *	@n: NAPI context
507  *
508  * Resume NAPI from being scheduled on this context.
509  * Must be paired with napi_disable.
510  */
511 static inline void napi_enable(struct napi_struct *n)
512 {
513 	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
514 	smp_mb__before_atomic();
515 	clear_bit(NAPI_STATE_SCHED, &n->state);
516 	clear_bit(NAPI_STATE_NPSVC, &n->state);
517 }
518 
519 /**
520  *	napi_synchronize - wait until NAPI is not running
521  *	@n: NAPI context
522  *
523  * Wait until NAPI is done being scheduled on this context.
524  * Waits till any outstanding processing completes but
525  * does not disable future activations.
526  */
527 static inline void napi_synchronize(const struct napi_struct *n)
528 {
529 	if (IS_ENABLED(CONFIG_SMP))
530 		while (test_bit(NAPI_STATE_SCHED, &n->state))
531 			msleep(1);
532 	else
533 		barrier();
534 }
535 
536 enum netdev_queue_state_t {
537 	__QUEUE_STATE_DRV_XOFF,
538 	__QUEUE_STATE_STACK_XOFF,
539 	__QUEUE_STATE_FROZEN,
540 };
541 
542 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
543 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
545 
546 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
547 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
548 					QUEUE_STATE_FROZEN)
549 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
550 					QUEUE_STATE_FROZEN)
551 
552 /*
553  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
554  * netif_tx_* functions below are used to manipulate this flag.  The
555  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
556  * queue independently.  The netif_xmit_*stopped functions below are called
557  * to check if the queue has been stopped by the driver or stack (either
558  * of the XOFF bits are set in the state).  Drivers should not need to call
559  * netif_xmit*stopped functions, they should only be using netif_tx_*.
560  */
561 
562 struct netdev_queue {
563 /*
564  * read-mostly part
565  */
566 	struct net_device	*dev;
567 	struct Qdisc __rcu	*qdisc;
568 	struct Qdisc		*qdisc_sleeping;
569 #ifdef CONFIG_SYSFS
570 	struct kobject		kobj;
571 #endif
572 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
573 	int			numa_node;
574 #endif
575 	unsigned long		tx_maxrate;
576 	/*
577 	 * Number of TX timeouts for this queue
578 	 * (/sys/class/net/DEV/Q/trans_timeout)
579 	 */
580 	unsigned long		trans_timeout;
581 /*
582  * write-mostly part
583  */
584 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
585 	int			xmit_lock_owner;
586 	/*
587 	 * Time (in jiffies) of last Tx
588 	 */
589 	unsigned long		trans_start;
590 
591 	unsigned long		state;
592 
593 #ifdef CONFIG_BQL
594 	struct dql		dql;
595 #endif
596 } ____cacheline_aligned_in_smp;
597 
598 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
599 {
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 	return q->numa_node;
602 #else
603 	return NUMA_NO_NODE;
604 #endif
605 }
606 
607 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
608 {
609 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
610 	q->numa_node = node;
611 #endif
612 }
613 
614 #ifdef CONFIG_RPS
615 /*
616  * This structure holds an RPS map which can be of variable length.  The
617  * map is an array of CPUs.
618  */
619 struct rps_map {
620 	unsigned int len;
621 	struct rcu_head rcu;
622 	u16 cpus[0];
623 };
624 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
625 
626 /*
627  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
628  * tail pointer for that CPU's input queue at the time of last enqueue, and
629  * a hardware filter index.
630  */
631 struct rps_dev_flow {
632 	u16 cpu;
633 	u16 filter;
634 	unsigned int last_qtail;
635 };
636 #define RPS_NO_FILTER 0xffff
637 
638 /*
639  * The rps_dev_flow_table structure contains a table of flow mappings.
640  */
641 struct rps_dev_flow_table {
642 	unsigned int mask;
643 	struct rcu_head rcu;
644 	struct rps_dev_flow flows[0];
645 };
646 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
647     ((_num) * sizeof(struct rps_dev_flow)))
648 
649 /*
650  * The rps_sock_flow_table contains mappings of flows to the last CPU
651  * on which they were processed by the application (set in recvmsg).
652  * Each entry is a 32bit value. Upper part is the high-order bits
653  * of flow hash, lower part is CPU number.
654  * rps_cpu_mask is used to partition the space, depending on number of
655  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
656  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
657  * meaning we use 32-6=26 bits for the hash.
658  */
659 struct rps_sock_flow_table {
660 	u32	mask;
661 
662 	u32	ents[0] ____cacheline_aligned_in_smp;
663 };
664 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
665 
666 #define RPS_NO_CPU 0xffff
667 
668 extern u32 rps_cpu_mask;
669 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
670 
671 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
672 					u32 hash)
673 {
674 	if (table && hash) {
675 		unsigned int index = hash & table->mask;
676 		u32 val = hash & ~rps_cpu_mask;
677 
678 		/* We only give a hint, preemption can change CPU under us */
679 		val |= raw_smp_processor_id();
680 
681 		if (table->ents[index] != val)
682 			table->ents[index] = val;
683 	}
684 }
685 
686 #ifdef CONFIG_RFS_ACCEL
687 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
688 			 u16 filter_id);
689 #endif
690 #endif /* CONFIG_RPS */
691 
692 /* This structure contains an instance of an RX queue. */
693 struct netdev_rx_queue {
694 #ifdef CONFIG_RPS
695 	struct rps_map __rcu		*rps_map;
696 	struct rps_dev_flow_table __rcu	*rps_flow_table;
697 #endif
698 	struct kobject			kobj;
699 	struct net_device		*dev;
700 } ____cacheline_aligned_in_smp;
701 
702 /*
703  * RX queue sysfs structures and functions.
704  */
705 struct rx_queue_attribute {
706 	struct attribute attr;
707 	ssize_t (*show)(struct netdev_rx_queue *queue,
708 	    struct rx_queue_attribute *attr, char *buf);
709 	ssize_t (*store)(struct netdev_rx_queue *queue,
710 	    struct rx_queue_attribute *attr, const char *buf, size_t len);
711 };
712 
713 #ifdef CONFIG_XPS
714 /*
715  * This structure holds an XPS map which can be of variable length.  The
716  * map is an array of queues.
717  */
718 struct xps_map {
719 	unsigned int len;
720 	unsigned int alloc_len;
721 	struct rcu_head rcu;
722 	u16 queues[0];
723 };
724 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
725 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
726        - sizeof(struct xps_map)) / sizeof(u16))
727 
728 /*
729  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
730  */
731 struct xps_dev_maps {
732 	struct rcu_head rcu;
733 	struct xps_map __rcu *cpu_map[0];
734 };
735 #define XPS_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +		\
736 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
737 #endif /* CONFIG_XPS */
738 
739 #define TC_MAX_QUEUE	16
740 #define TC_BITMASK	15
741 /* HW offloaded queuing disciplines txq count and offset maps */
742 struct netdev_tc_txq {
743 	u16 count;
744 	u16 offset;
745 };
746 
747 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
748 /*
749  * This structure is to hold information about the device
750  * configured to run FCoE protocol stack.
751  */
752 struct netdev_fcoe_hbainfo {
753 	char	manufacturer[64];
754 	char	serial_number[64];
755 	char	hardware_version[64];
756 	char	driver_version[64];
757 	char	optionrom_version[64];
758 	char	firmware_version[64];
759 	char	model[256];
760 	char	model_description[256];
761 };
762 #endif
763 
764 #define MAX_PHYS_ITEM_ID_LEN 32
765 
766 /* This structure holds a unique identifier to identify some
767  * physical item (port for example) used by a netdevice.
768  */
769 struct netdev_phys_item_id {
770 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
771 	unsigned char id_len;
772 };
773 
774 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
775 					    struct netdev_phys_item_id *b)
776 {
777 	return a->id_len == b->id_len &&
778 	       memcmp(a->id, b->id, a->id_len) == 0;
779 }
780 
781 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
782 				       struct sk_buff *skb);
783 
784 /* These structures hold the attributes of qdisc and classifiers
785  * that are being passed to the netdevice through the setup_tc op.
786  */
787 enum {
788 	TC_SETUP_MQPRIO,
789 	TC_SETUP_CLSU32,
790 	TC_SETUP_CLSFLOWER,
791 	TC_SETUP_MATCHALL,
792 	TC_SETUP_CLSBPF,
793 };
794 
795 struct tc_cls_u32_offload;
796 
797 struct tc_to_netdev {
798 	unsigned int type;
799 	union {
800 		u8 tc;
801 		struct tc_cls_u32_offload *cls_u32;
802 		struct tc_cls_flower_offload *cls_flower;
803 		struct tc_cls_matchall_offload *cls_mall;
804 		struct tc_cls_bpf_offload *cls_bpf;
805 	};
806 	bool egress_dev;
807 };
808 
809 /* These structures hold the attributes of xdp state that are being passed
810  * to the netdevice through the xdp op.
811  */
812 enum xdp_netdev_command {
813 	/* Set or clear a bpf program used in the earliest stages of packet
814 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
815 	 * is responsible for calling bpf_prog_put on any old progs that are
816 	 * stored. In case of error, the callee need not release the new prog
817 	 * reference, but on success it takes ownership and must bpf_prog_put
818 	 * when it is no longer used.
819 	 */
820 	XDP_SETUP_PROG,
821 	/* Check if a bpf program is set on the device.  The callee should
822 	 * return true if a program is currently attached and running.
823 	 */
824 	XDP_QUERY_PROG,
825 };
826 
827 struct netdev_xdp {
828 	enum xdp_netdev_command command;
829 	union {
830 		/* XDP_SETUP_PROG */
831 		struct bpf_prog *prog;
832 		/* XDP_QUERY_PROG */
833 		bool prog_attached;
834 	};
835 };
836 
837 /*
838  * This structure defines the management hooks for network devices.
839  * The following hooks can be defined; unless noted otherwise, they are
840  * optional and can be filled with a null pointer.
841  *
842  * int (*ndo_init)(struct net_device *dev);
843  *     This function is called once when a network device is registered.
844  *     The network device can use this for any late stage initialization
845  *     or semantic validation. It can fail with an error code which will
846  *     be propagated back to register_netdev.
847  *
848  * void (*ndo_uninit)(struct net_device *dev);
849  *     This function is called when device is unregistered or when registration
850  *     fails. It is not called if init fails.
851  *
852  * int (*ndo_open)(struct net_device *dev);
853  *     This function is called when a network device transitions to the up
854  *     state.
855  *
856  * int (*ndo_stop)(struct net_device *dev);
857  *     This function is called when a network device transitions to the down
858  *     state.
859  *
860  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
861  *                               struct net_device *dev);
862  *	Called when a packet needs to be transmitted.
863  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
864  *	the queue before that can happen; it's for obsolete devices and weird
865  *	corner cases, but the stack really does a non-trivial amount
866  *	of useless work if you return NETDEV_TX_BUSY.
867  *	Required; cannot be NULL.
868  *
869  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
870  *					   struct net_device *dev
871  *					   netdev_features_t features);
872  *	Called by core transmit path to determine if device is capable of
873  *	performing offload operations on a given packet. This is to give
874  *	the device an opportunity to implement any restrictions that cannot
875  *	be otherwise expressed by feature flags. The check is called with
876  *	the set of features that the stack has calculated and it returns
877  *	those the driver believes to be appropriate.
878  *
879  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
880  *                         void *accel_priv, select_queue_fallback_t fallback);
881  *	Called to decide which queue to use when device supports multiple
882  *	transmit queues.
883  *
884  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
885  *	This function is called to allow device receiver to make
886  *	changes to configuration when multicast or promiscuous is enabled.
887  *
888  * void (*ndo_set_rx_mode)(struct net_device *dev);
889  *	This function is called device changes address list filtering.
890  *	If driver handles unicast address filtering, it should set
891  *	IFF_UNICAST_FLT in its priv_flags.
892  *
893  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
894  *	This function  is called when the Media Access Control address
895  *	needs to be changed. If this interface is not defined, the
896  *	MAC address can not be changed.
897  *
898  * int (*ndo_validate_addr)(struct net_device *dev);
899  *	Test if Media Access Control address is valid for the device.
900  *
901  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
902  *	Called when a user requests an ioctl which can't be handled by
903  *	the generic interface code. If not defined ioctls return
904  *	not supported error code.
905  *
906  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
907  *	Used to set network devices bus interface parameters. This interface
908  *	is retained for legacy reasons; new devices should use the bus
909  *	interface (PCI) for low level management.
910  *
911  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
912  *	Called when a user wants to change the Maximum Transfer Unit
913  *	of a device. If not defined, any request to change MTU will
914  *	will return an error.
915  *
916  * void (*ndo_tx_timeout)(struct net_device *dev);
917  *	Callback used when the transmitter has not made any progress
918  *	for dev->watchdog ticks.
919  *
920  * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
921  *                      struct rtnl_link_stats64 *storage);
922  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
923  *	Called when a user wants to get the network device usage
924  *	statistics. Drivers must do one of the following:
925  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
926  *	   rtnl_link_stats64 structure passed by the caller.
927  *	2. Define @ndo_get_stats to update a net_device_stats structure
928  *	   (which should normally be dev->stats) and return a pointer to
929  *	   it. The structure may be changed asynchronously only if each
930  *	   field is written atomically.
931  *	3. Update dev->stats asynchronously and atomically, and define
932  *	   neither operation.
933  *
934  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
935  *	Return true if this device supports offload stats of this attr_id.
936  *
937  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
938  *	void *attr_data)
939  *	Get statistics for offload operations by attr_id. Write it into the
940  *	attr_data pointer.
941  *
942  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
943  *	If device supports VLAN filtering this function is called when a
944  *	VLAN id is registered.
945  *
946  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
947  *	If device supports VLAN filtering this function is called when a
948  *	VLAN id is unregistered.
949  *
950  * void (*ndo_poll_controller)(struct net_device *dev);
951  *
952  *	SR-IOV management functions.
953  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
954  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
955  *			  u8 qos, __be16 proto);
956  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
957  *			  int max_tx_rate);
958  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
959  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
960  * int (*ndo_get_vf_config)(struct net_device *dev,
961  *			    int vf, struct ifla_vf_info *ivf);
962  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
963  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
964  *			  struct nlattr *port[]);
965  *
966  *      Enable or disable the VF ability to query its RSS Redirection Table and
967  *      Hash Key. This is needed since on some devices VF share this information
968  *      with PF and querying it may introduce a theoretical security risk.
969  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
970  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
971  * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
972  * 	Called to setup 'tc' number of traffic classes in the net device. This
973  * 	is always called from the stack with the rtnl lock held and netif tx
974  * 	queues stopped. This allows the netdevice to perform queue management
975  * 	safely.
976  *
977  *	Fiber Channel over Ethernet (FCoE) offload functions.
978  * int (*ndo_fcoe_enable)(struct net_device *dev);
979  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
980  *	so the underlying device can perform whatever needed configuration or
981  *	initialization to support acceleration of FCoE traffic.
982  *
983  * int (*ndo_fcoe_disable)(struct net_device *dev);
984  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
985  *	so the underlying device can perform whatever needed clean-ups to
986  *	stop supporting acceleration of FCoE traffic.
987  *
988  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
989  *			     struct scatterlist *sgl, unsigned int sgc);
990  *	Called when the FCoE Initiator wants to initialize an I/O that
991  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
992  *	perform necessary setup and returns 1 to indicate the device is set up
993  *	successfully to perform DDP on this I/O, otherwise this returns 0.
994  *
995  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
996  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
997  *	indicated by the FC exchange id 'xid', so the underlying device can
998  *	clean up and reuse resources for later DDP requests.
999  *
1000  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1001  *			      struct scatterlist *sgl, unsigned int sgc);
1002  *	Called when the FCoE Target wants to initialize an I/O that
1003  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1004  *	perform necessary setup and returns 1 to indicate the device is set up
1005  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1006  *
1007  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1008  *			       struct netdev_fcoe_hbainfo *hbainfo);
1009  *	Called when the FCoE Protocol stack wants information on the underlying
1010  *	device. This information is utilized by the FCoE protocol stack to
1011  *	register attributes with Fiber Channel management service as per the
1012  *	FC-GS Fabric Device Management Information(FDMI) specification.
1013  *
1014  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1015  *	Called when the underlying device wants to override default World Wide
1016  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1017  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1018  *	protocol stack to use.
1019  *
1020  *	RFS acceleration.
1021  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1022  *			    u16 rxq_index, u32 flow_id);
1023  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1024  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1025  *	Return the filter ID on success, or a negative error code.
1026  *
1027  *	Slave management functions (for bridge, bonding, etc).
1028  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1029  *	Called to make another netdev an underling.
1030  *
1031  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1032  *	Called to release previously enslaved netdev.
1033  *
1034  *      Feature/offload setting functions.
1035  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1036  *		netdev_features_t features);
1037  *	Adjusts the requested feature flags according to device-specific
1038  *	constraints, and returns the resulting flags. Must not modify
1039  *	the device state.
1040  *
1041  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1042  *	Called to update device configuration to new features. Passed
1043  *	feature set might be less than what was returned by ndo_fix_features()).
1044  *	Must return >0 or -errno if it changed dev->features itself.
1045  *
1046  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1047  *		      struct net_device *dev,
1048  *		      const unsigned char *addr, u16 vid, u16 flags)
1049  *	Adds an FDB entry to dev for addr.
1050  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1051  *		      struct net_device *dev,
1052  *		      const unsigned char *addr, u16 vid)
1053  *	Deletes the FDB entry from dev coresponding to addr.
1054  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1055  *		       struct net_device *dev, struct net_device *filter_dev,
1056  *		       int *idx)
1057  *	Used to add FDB entries to dump requests. Implementers should add
1058  *	entries to skb and update idx with the number of entries.
1059  *
1060  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1061  *			     u16 flags)
1062  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1063  *			     struct net_device *dev, u32 filter_mask,
1064  *			     int nlflags)
1065  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1066  *			     u16 flags);
1067  *
1068  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1069  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1070  *	which do not represent real hardware may define this to allow their
1071  *	userspace components to manage their virtual carrier state. Devices
1072  *	that determine carrier state from physical hardware properties (eg
1073  *	network cables) or protocol-dependent mechanisms (eg
1074  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1075  *
1076  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1077  *			       struct netdev_phys_item_id *ppid);
1078  *	Called to get ID of physical port of this device. If driver does
1079  *	not implement this, it is assumed that the hw is not able to have
1080  *	multiple net devices on single physical port.
1081  *
1082  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1083  *			      struct udp_tunnel_info *ti);
1084  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1085  *	address family that a UDP tunnel is listnening to. It is called only
1086  *	when a new port starts listening. The operation is protected by the
1087  *	RTNL.
1088  *
1089  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1090  *			      struct udp_tunnel_info *ti);
1091  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1092  *	address family that the UDP tunnel is not listening to anymore. The
1093  *	operation is protected by the RTNL.
1094  *
1095  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1096  *				 struct net_device *dev)
1097  *	Called by upper layer devices to accelerate switching or other
1098  *	station functionality into hardware. 'pdev is the lowerdev
1099  *	to use for the offload and 'dev' is the net device that will
1100  *	back the offload. Returns a pointer to the private structure
1101  *	the upper layer will maintain.
1102  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1103  *	Called by upper layer device to delete the station created
1104  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1105  *	the station and priv is the structure returned by the add
1106  *	operation.
1107  * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1108  *				      struct net_device *dev,
1109  *				      void *priv);
1110  *	Callback to use for xmit over the accelerated station. This
1111  *	is used in place of ndo_start_xmit on accelerated net
1112  *	devices.
1113  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1114  *			     int queue_index, u32 maxrate);
1115  *	Called when a user wants to set a max-rate limitation of specific
1116  *	TX queue.
1117  * int (*ndo_get_iflink)(const struct net_device *dev);
1118  *	Called to get the iflink value of this device.
1119  * void (*ndo_change_proto_down)(struct net_device *dev,
1120  *				 bool proto_down);
1121  *	This function is used to pass protocol port error state information
1122  *	to the switch driver. The switch driver can react to the proto_down
1123  *      by doing a phys down on the associated switch port.
1124  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1125  *	This function is used to get egress tunnel information for given skb.
1126  *	This is useful for retrieving outer tunnel header parameters while
1127  *	sampling packet.
1128  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1129  *	This function is used to specify the headroom that the skb must
1130  *	consider when allocation skb during packet reception. Setting
1131  *	appropriate rx headroom value allows avoiding skb head copy on
1132  *	forward. Setting a negative value resets the rx headroom to the
1133  *	default value.
1134  * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1135  *	This function is used to set or query state related to XDP on the
1136  *	netdevice. See definition of enum xdp_netdev_command for details.
1137  *
1138  */
1139 struct net_device_ops {
1140 	int			(*ndo_init)(struct net_device *dev);
1141 	void			(*ndo_uninit)(struct net_device *dev);
1142 	int			(*ndo_open)(struct net_device *dev);
1143 	int			(*ndo_stop)(struct net_device *dev);
1144 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1145 						  struct net_device *dev);
1146 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1147 						      struct net_device *dev,
1148 						      netdev_features_t features);
1149 	u16			(*ndo_select_queue)(struct net_device *dev,
1150 						    struct sk_buff *skb,
1151 						    void *accel_priv,
1152 						    select_queue_fallback_t fallback);
1153 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1154 						       int flags);
1155 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1156 	int			(*ndo_set_mac_address)(struct net_device *dev,
1157 						       void *addr);
1158 	int			(*ndo_validate_addr)(struct net_device *dev);
1159 	int			(*ndo_do_ioctl)(struct net_device *dev,
1160 					        struct ifreq *ifr, int cmd);
1161 	int			(*ndo_set_config)(struct net_device *dev,
1162 					          struct ifmap *map);
1163 	int			(*ndo_change_mtu)(struct net_device *dev,
1164 						  int new_mtu);
1165 	int			(*ndo_neigh_setup)(struct net_device *dev,
1166 						   struct neigh_parms *);
1167 	void			(*ndo_tx_timeout) (struct net_device *dev);
1168 
1169 	struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1170 						     struct rtnl_link_stats64 *storage);
1171 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1172 	int			(*ndo_get_offload_stats)(int attr_id,
1173 							 const struct net_device *dev,
1174 							 void *attr_data);
1175 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1176 
1177 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1178 						       __be16 proto, u16 vid);
1179 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1180 						        __be16 proto, u16 vid);
1181 #ifdef CONFIG_NET_POLL_CONTROLLER
1182 	void                    (*ndo_poll_controller)(struct net_device *dev);
1183 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1184 						     struct netpoll_info *info);
1185 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1186 #endif
1187 #ifdef CONFIG_NET_RX_BUSY_POLL
1188 	int			(*ndo_busy_poll)(struct napi_struct *dev);
1189 #endif
1190 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1191 						  int queue, u8 *mac);
1192 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1193 						   int queue, u16 vlan,
1194 						   u8 qos, __be16 proto);
1195 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1196 						   int vf, int min_tx_rate,
1197 						   int max_tx_rate);
1198 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1199 						       int vf, bool setting);
1200 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1201 						    int vf, bool setting);
1202 	int			(*ndo_get_vf_config)(struct net_device *dev,
1203 						     int vf,
1204 						     struct ifla_vf_info *ivf);
1205 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1206 							 int vf, int link_state);
1207 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1208 						    int vf,
1209 						    struct ifla_vf_stats
1210 						    *vf_stats);
1211 	int			(*ndo_set_vf_port)(struct net_device *dev,
1212 						   int vf,
1213 						   struct nlattr *port[]);
1214 	int			(*ndo_get_vf_port)(struct net_device *dev,
1215 						   int vf, struct sk_buff *skb);
1216 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1217 						   int vf, u64 guid,
1218 						   int guid_type);
1219 	int			(*ndo_set_vf_rss_query_en)(
1220 						   struct net_device *dev,
1221 						   int vf, bool setting);
1222 	int			(*ndo_setup_tc)(struct net_device *dev,
1223 						u32 handle,
1224 						__be16 protocol,
1225 						struct tc_to_netdev *tc);
1226 #if IS_ENABLED(CONFIG_FCOE)
1227 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1228 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1229 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1230 						      u16 xid,
1231 						      struct scatterlist *sgl,
1232 						      unsigned int sgc);
1233 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1234 						     u16 xid);
1235 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1236 						       u16 xid,
1237 						       struct scatterlist *sgl,
1238 						       unsigned int sgc);
1239 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1240 							struct netdev_fcoe_hbainfo *hbainfo);
1241 #endif
1242 
1243 #if IS_ENABLED(CONFIG_LIBFCOE)
1244 #define NETDEV_FCOE_WWNN 0
1245 #define NETDEV_FCOE_WWPN 1
1246 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1247 						    u64 *wwn, int type);
1248 #endif
1249 
1250 #ifdef CONFIG_RFS_ACCEL
1251 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1252 						     const struct sk_buff *skb,
1253 						     u16 rxq_index,
1254 						     u32 flow_id);
1255 #endif
1256 	int			(*ndo_add_slave)(struct net_device *dev,
1257 						 struct net_device *slave_dev);
1258 	int			(*ndo_del_slave)(struct net_device *dev,
1259 						 struct net_device *slave_dev);
1260 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1261 						    netdev_features_t features);
1262 	int			(*ndo_set_features)(struct net_device *dev,
1263 						    netdev_features_t features);
1264 	int			(*ndo_neigh_construct)(struct net_device *dev,
1265 						       struct neighbour *n);
1266 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1267 						     struct neighbour *n);
1268 
1269 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1270 					       struct nlattr *tb[],
1271 					       struct net_device *dev,
1272 					       const unsigned char *addr,
1273 					       u16 vid,
1274 					       u16 flags);
1275 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1276 					       struct nlattr *tb[],
1277 					       struct net_device *dev,
1278 					       const unsigned char *addr,
1279 					       u16 vid);
1280 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1281 						struct netlink_callback *cb,
1282 						struct net_device *dev,
1283 						struct net_device *filter_dev,
1284 						int *idx);
1285 
1286 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1287 						      struct nlmsghdr *nlh,
1288 						      u16 flags);
1289 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1290 						      u32 pid, u32 seq,
1291 						      struct net_device *dev,
1292 						      u32 filter_mask,
1293 						      int nlflags);
1294 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1295 						      struct nlmsghdr *nlh,
1296 						      u16 flags);
1297 	int			(*ndo_change_carrier)(struct net_device *dev,
1298 						      bool new_carrier);
1299 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1300 							struct netdev_phys_item_id *ppid);
1301 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1302 							  char *name, size_t len);
1303 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1304 						      struct udp_tunnel_info *ti);
1305 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1306 						      struct udp_tunnel_info *ti);
1307 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1308 							struct net_device *dev);
1309 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1310 							void *priv);
1311 
1312 	netdev_tx_t		(*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1313 							struct net_device *dev,
1314 							void *priv);
1315 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1316 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1317 						      int queue_index,
1318 						      u32 maxrate);
1319 	int			(*ndo_get_iflink)(const struct net_device *dev);
1320 	int			(*ndo_change_proto_down)(struct net_device *dev,
1321 							 bool proto_down);
1322 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1323 						       struct sk_buff *skb);
1324 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1325 						       int needed_headroom);
1326 	int			(*ndo_xdp)(struct net_device *dev,
1327 					   struct netdev_xdp *xdp);
1328 };
1329 
1330 /**
1331  * enum net_device_priv_flags - &struct net_device priv_flags
1332  *
1333  * These are the &struct net_device, they are only set internally
1334  * by drivers and used in the kernel. These flags are invisible to
1335  * userspace; this means that the order of these flags can change
1336  * during any kernel release.
1337  *
1338  * You should have a pretty good reason to be extending these flags.
1339  *
1340  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1341  * @IFF_EBRIDGE: Ethernet bridging device
1342  * @IFF_BONDING: bonding master or slave
1343  * @IFF_ISATAP: ISATAP interface (RFC4214)
1344  * @IFF_WAN_HDLC: WAN HDLC device
1345  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1346  *	release skb->dst
1347  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1348  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1349  * @IFF_MACVLAN_PORT: device used as macvlan port
1350  * @IFF_BRIDGE_PORT: device used as bridge port
1351  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1352  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1353  * @IFF_UNICAST_FLT: Supports unicast filtering
1354  * @IFF_TEAM_PORT: device used as team port
1355  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1356  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1357  *	change when it's running
1358  * @IFF_MACVLAN: Macvlan device
1359  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1360  *	underlying stacked devices
1361  * @IFF_IPVLAN_MASTER: IPvlan master device
1362  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1363  * @IFF_L3MDEV_MASTER: device is an L3 master device
1364  * @IFF_NO_QUEUE: device can run without qdisc attached
1365  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1366  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1367  * @IFF_TEAM: device is a team device
1368  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1369  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1370  *	entity (i.e. the master device for bridged veth)
1371  * @IFF_MACSEC: device is a MACsec device
1372  */
1373 enum netdev_priv_flags {
1374 	IFF_802_1Q_VLAN			= 1<<0,
1375 	IFF_EBRIDGE			= 1<<1,
1376 	IFF_BONDING			= 1<<2,
1377 	IFF_ISATAP			= 1<<3,
1378 	IFF_WAN_HDLC			= 1<<4,
1379 	IFF_XMIT_DST_RELEASE		= 1<<5,
1380 	IFF_DONT_BRIDGE			= 1<<6,
1381 	IFF_DISABLE_NETPOLL		= 1<<7,
1382 	IFF_MACVLAN_PORT		= 1<<8,
1383 	IFF_BRIDGE_PORT			= 1<<9,
1384 	IFF_OVS_DATAPATH		= 1<<10,
1385 	IFF_TX_SKB_SHARING		= 1<<11,
1386 	IFF_UNICAST_FLT			= 1<<12,
1387 	IFF_TEAM_PORT			= 1<<13,
1388 	IFF_SUPP_NOFCS			= 1<<14,
1389 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1390 	IFF_MACVLAN			= 1<<16,
1391 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1392 	IFF_IPVLAN_MASTER		= 1<<18,
1393 	IFF_IPVLAN_SLAVE		= 1<<19,
1394 	IFF_L3MDEV_MASTER		= 1<<20,
1395 	IFF_NO_QUEUE			= 1<<21,
1396 	IFF_OPENVSWITCH			= 1<<22,
1397 	IFF_L3MDEV_SLAVE		= 1<<23,
1398 	IFF_TEAM			= 1<<24,
1399 	IFF_RXFH_CONFIGURED		= 1<<25,
1400 	IFF_PHONY_HEADROOM		= 1<<26,
1401 	IFF_MACSEC			= 1<<27,
1402 };
1403 
1404 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1405 #define IFF_EBRIDGE			IFF_EBRIDGE
1406 #define IFF_BONDING			IFF_BONDING
1407 #define IFF_ISATAP			IFF_ISATAP
1408 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1409 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1410 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1411 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1412 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1413 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1414 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1415 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1416 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1417 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1418 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1419 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1420 #define IFF_MACVLAN			IFF_MACVLAN
1421 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1422 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1423 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1424 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1425 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1426 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1427 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1428 #define IFF_TEAM			IFF_TEAM
1429 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1430 #define IFF_MACSEC			IFF_MACSEC
1431 
1432 /**
1433  *	struct net_device - The DEVICE structure.
1434  *		Actually, this whole structure is a big mistake.  It mixes I/O
1435  *		data with strictly "high-level" data, and it has to know about
1436  *		almost every data structure used in the INET module.
1437  *
1438  *	@name:	This is the first field of the "visible" part of this structure
1439  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1440  *	 	of the interface.
1441  *
1442  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1443  *	@ifalias:	SNMP alias
1444  *	@mem_end:	Shared memory end
1445  *	@mem_start:	Shared memory start
1446  *	@base_addr:	Device I/O address
1447  *	@irq:		Device IRQ number
1448  *
1449  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1450  *
1451  *	@state:		Generic network queuing layer state, see netdev_state_t
1452  *	@dev_list:	The global list of network devices
1453  *	@napi_list:	List entry used for polling NAPI devices
1454  *	@unreg_list:	List entry  when we are unregistering the
1455  *			device; see the function unregister_netdev
1456  *	@close_list:	List entry used when we are closing the device
1457  *	@ptype_all:     Device-specific packet handlers for all protocols
1458  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1459  *
1460  *	@adj_list:	Directly linked devices, like slaves for bonding
1461  *	@features:	Currently active device features
1462  *	@hw_features:	User-changeable features
1463  *
1464  *	@wanted_features:	User-requested features
1465  *	@vlan_features:		Mask of features inheritable by VLAN devices
1466  *
1467  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1468  *				This field indicates what encapsulation
1469  *				offloads the hardware is capable of doing,
1470  *				and drivers will need to set them appropriately.
1471  *
1472  *	@mpls_features:	Mask of features inheritable by MPLS
1473  *
1474  *	@ifindex:	interface index
1475  *	@group:		The group the device belongs to
1476  *
1477  *	@stats:		Statistics struct, which was left as a legacy, use
1478  *			rtnl_link_stats64 instead
1479  *
1480  *	@rx_dropped:	Dropped packets by core network,
1481  *			do not use this in drivers
1482  *	@tx_dropped:	Dropped packets by core network,
1483  *			do not use this in drivers
1484  *	@rx_nohandler:	nohandler dropped packets by core network on
1485  *			inactive devices, do not use this in drivers
1486  *
1487  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1488  *				instead of ioctl,
1489  *				see <net/iw_handler.h> for details.
1490  *	@wireless_data:	Instance data managed by the core of wireless extensions
1491  *
1492  *	@netdev_ops:	Includes several pointers to callbacks,
1493  *			if one wants to override the ndo_*() functions
1494  *	@ethtool_ops:	Management operations
1495  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1496  *			discovery handling. Necessary for e.g. 6LoWPAN.
1497  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1498  *			of Layer 2 headers.
1499  *
1500  *	@flags:		Interface flags (a la BSD)
1501  *	@priv_flags:	Like 'flags' but invisible to userspace,
1502  *			see if.h for the definitions
1503  *	@gflags:	Global flags ( kept as legacy )
1504  *	@padded:	How much padding added by alloc_netdev()
1505  *	@operstate:	RFC2863 operstate
1506  *	@link_mode:	Mapping policy to operstate
1507  *	@if_port:	Selectable AUI, TP, ...
1508  *	@dma:		DMA channel
1509  *	@mtu:		Interface MTU value
1510  *	@min_mtu:	Interface Minimum MTU value
1511  *	@max_mtu:	Interface Maximum MTU value
1512  *	@type:		Interface hardware type
1513  *	@hard_header_len: Maximum hardware header length.
1514  *
1515  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1516  *			  cases can this be guaranteed
1517  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1518  *			  cases can this be guaranteed. Some cases also use
1519  *			  LL_MAX_HEADER instead to allocate the skb
1520  *
1521  *	interface address info:
1522  *
1523  * 	@perm_addr:		Permanent hw address
1524  * 	@addr_assign_type:	Hw address assignment type
1525  * 	@addr_len:		Hardware address length
1526  *	@neigh_priv_len:	Used in neigh_alloc()
1527  * 	@dev_id:		Used to differentiate devices that share
1528  * 				the same link layer address
1529  * 	@dev_port:		Used to differentiate devices that share
1530  * 				the same function
1531  *	@addr_list_lock:	XXX: need comments on this one
1532  *	@uc_promisc:		Counter that indicates promiscuous mode
1533  *				has been enabled due to the need to listen to
1534  *				additional unicast addresses in a device that
1535  *				does not implement ndo_set_rx_mode()
1536  *	@uc:			unicast mac addresses
1537  *	@mc:			multicast mac addresses
1538  *	@dev_addrs:		list of device hw addresses
1539  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1540  *	@promiscuity:		Number of times the NIC is told to work in
1541  *				promiscuous mode; if it becomes 0 the NIC will
1542  *				exit promiscuous mode
1543  *	@allmulti:		Counter, enables or disables allmulticast mode
1544  *
1545  *	@vlan_info:	VLAN info
1546  *	@dsa_ptr:	dsa specific data
1547  *	@tipc_ptr:	TIPC specific data
1548  *	@atalk_ptr:	AppleTalk link
1549  *	@ip_ptr:	IPv4 specific data
1550  *	@dn_ptr:	DECnet specific data
1551  *	@ip6_ptr:	IPv6 specific data
1552  *	@ax25_ptr:	AX.25 specific data
1553  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1554  *
1555  *	@last_rx:	Time of last Rx
1556  *	@dev_addr:	Hw address (before bcast,
1557  *			because most packets are unicast)
1558  *
1559  *	@_rx:			Array of RX queues
1560  *	@num_rx_queues:		Number of RX queues
1561  *				allocated at register_netdev() time
1562  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1563  *
1564  *	@rx_handler:		handler for received packets
1565  *	@rx_handler_data: 	XXX: need comments on this one
1566  *	@ingress_queue:		XXX: need comments on this one
1567  *	@broadcast:		hw bcast address
1568  *
1569  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1570  *			indexed by RX queue number. Assigned by driver.
1571  *			This must only be set if the ndo_rx_flow_steer
1572  *			operation is defined
1573  *	@index_hlist:		Device index hash chain
1574  *
1575  *	@_tx:			Array of TX queues
1576  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1577  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1578  *	@qdisc:			Root qdisc from userspace point of view
1579  *	@tx_queue_len:		Max frames per queue allowed
1580  *	@tx_global_lock: 	XXX: need comments on this one
1581  *
1582  *	@xps_maps:	XXX: need comments on this one
1583  *
1584  *	@watchdog_timeo:	Represents the timeout that is used by
1585  *				the watchdog (see dev_watchdog())
1586  *	@watchdog_timer:	List of timers
1587  *
1588  *	@pcpu_refcnt:		Number of references to this device
1589  *	@todo_list:		Delayed register/unregister
1590  *	@link_watch_list:	XXX: need comments on this one
1591  *
1592  *	@reg_state:		Register/unregister state machine
1593  *	@dismantle:		Device is going to be freed
1594  *	@rtnl_link_state:	This enum represents the phases of creating
1595  *				a new link
1596  *
1597  *	@destructor:		Called from unregister,
1598  *				can be used to call free_netdev
1599  *	@npinfo:		XXX: need comments on this one
1600  * 	@nd_net:		Network namespace this network device is inside
1601  *
1602  * 	@ml_priv:	Mid-layer private
1603  * 	@lstats:	Loopback statistics
1604  * 	@tstats:	Tunnel statistics
1605  * 	@dstats:	Dummy statistics
1606  * 	@vstats:	Virtual ethernet statistics
1607  *
1608  *	@garp_port:	GARP
1609  *	@mrp_port:	MRP
1610  *
1611  *	@dev:		Class/net/name entry
1612  *	@sysfs_groups:	Space for optional device, statistics and wireless
1613  *			sysfs groups
1614  *
1615  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1616  *	@rtnl_link_ops:	Rtnl_link_ops
1617  *
1618  *	@gso_max_size:	Maximum size of generic segmentation offload
1619  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1620  *			NIC for GSO
1621  *
1622  *	@dcbnl_ops:	Data Center Bridging netlink ops
1623  *	@num_tc:	Number of traffic classes in the net device
1624  *	@tc_to_txq:	XXX: need comments on this one
1625  *	@prio_tc_map:	XXX: need comments on this one
1626  *
1627  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1628  *
1629  *	@priomap:	XXX: need comments on this one
1630  *	@phydev:	Physical device may attach itself
1631  *			for hardware timestamping
1632  *
1633  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1634  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1635  *
1636  *	@proto_down:	protocol port state information can be sent to the
1637  *			switch driver and used to set the phys state of the
1638  *			switch port.
1639  *
1640  *	FIXME: cleanup struct net_device such that network protocol info
1641  *	moves out.
1642  */
1643 
1644 struct net_device {
1645 	char			name[IFNAMSIZ];
1646 	struct hlist_node	name_hlist;
1647 	char 			*ifalias;
1648 	/*
1649 	 *	I/O specific fields
1650 	 *	FIXME: Merge these and struct ifmap into one
1651 	 */
1652 	unsigned long		mem_end;
1653 	unsigned long		mem_start;
1654 	unsigned long		base_addr;
1655 	int			irq;
1656 
1657 	atomic_t		carrier_changes;
1658 
1659 	/*
1660 	 *	Some hardware also needs these fields (state,dev_list,
1661 	 *	napi_list,unreg_list,close_list) but they are not
1662 	 *	part of the usual set specified in Space.c.
1663 	 */
1664 
1665 	unsigned long		state;
1666 
1667 	struct list_head	dev_list;
1668 	struct list_head	napi_list;
1669 	struct list_head	unreg_list;
1670 	struct list_head	close_list;
1671 	struct list_head	ptype_all;
1672 	struct list_head	ptype_specific;
1673 
1674 	struct {
1675 		struct list_head upper;
1676 		struct list_head lower;
1677 	} adj_list;
1678 
1679 	netdev_features_t	features;
1680 	netdev_features_t	hw_features;
1681 	netdev_features_t	wanted_features;
1682 	netdev_features_t	vlan_features;
1683 	netdev_features_t	hw_enc_features;
1684 	netdev_features_t	mpls_features;
1685 	netdev_features_t	gso_partial_features;
1686 
1687 	int			ifindex;
1688 	int			group;
1689 
1690 	struct net_device_stats	stats;
1691 
1692 	atomic_long_t		rx_dropped;
1693 	atomic_long_t		tx_dropped;
1694 	atomic_long_t		rx_nohandler;
1695 
1696 #ifdef CONFIG_WIRELESS_EXT
1697 	const struct iw_handler_def *wireless_handlers;
1698 	struct iw_public_data	*wireless_data;
1699 #endif
1700 	const struct net_device_ops *netdev_ops;
1701 	const struct ethtool_ops *ethtool_ops;
1702 #ifdef CONFIG_NET_SWITCHDEV
1703 	const struct switchdev_ops *switchdev_ops;
1704 #endif
1705 #ifdef CONFIG_NET_L3_MASTER_DEV
1706 	const struct l3mdev_ops	*l3mdev_ops;
1707 #endif
1708 #if IS_ENABLED(CONFIG_IPV6)
1709 	const struct ndisc_ops *ndisc_ops;
1710 #endif
1711 
1712 	const struct header_ops *header_ops;
1713 
1714 	unsigned int		flags;
1715 	unsigned int		priv_flags;
1716 
1717 	unsigned short		gflags;
1718 	unsigned short		padded;
1719 
1720 	unsigned char		operstate;
1721 	unsigned char		link_mode;
1722 
1723 	unsigned char		if_port;
1724 	unsigned char		dma;
1725 
1726 	unsigned int		mtu;
1727 	unsigned int		min_mtu;
1728 	unsigned int		max_mtu;
1729 	unsigned short		type;
1730 	unsigned short		hard_header_len;
1731 
1732 	unsigned short		needed_headroom;
1733 	unsigned short		needed_tailroom;
1734 
1735 	/* Interface address info. */
1736 	unsigned char		perm_addr[MAX_ADDR_LEN];
1737 	unsigned char		addr_assign_type;
1738 	unsigned char		addr_len;
1739 	unsigned short		neigh_priv_len;
1740 	unsigned short          dev_id;
1741 	unsigned short          dev_port;
1742 	spinlock_t		addr_list_lock;
1743 	unsigned char		name_assign_type;
1744 	bool			uc_promisc;
1745 	struct netdev_hw_addr_list	uc;
1746 	struct netdev_hw_addr_list	mc;
1747 	struct netdev_hw_addr_list	dev_addrs;
1748 
1749 #ifdef CONFIG_SYSFS
1750 	struct kset		*queues_kset;
1751 #endif
1752 	unsigned int		promiscuity;
1753 	unsigned int		allmulti;
1754 
1755 
1756 	/* Protocol-specific pointers */
1757 
1758 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1759 	struct vlan_info __rcu	*vlan_info;
1760 #endif
1761 #if IS_ENABLED(CONFIG_NET_DSA)
1762 	struct dsa_switch_tree	*dsa_ptr;
1763 #endif
1764 #if IS_ENABLED(CONFIG_TIPC)
1765 	struct tipc_bearer __rcu *tipc_ptr;
1766 #endif
1767 	void 			*atalk_ptr;
1768 	struct in_device __rcu	*ip_ptr;
1769 	struct dn_dev __rcu     *dn_ptr;
1770 	struct inet6_dev __rcu	*ip6_ptr;
1771 	void			*ax25_ptr;
1772 	struct wireless_dev	*ieee80211_ptr;
1773 	struct wpan_dev		*ieee802154_ptr;
1774 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1775 	struct mpls_dev __rcu	*mpls_ptr;
1776 #endif
1777 
1778 /*
1779  * Cache lines mostly used on receive path (including eth_type_trans())
1780  */
1781 	unsigned long		last_rx;
1782 
1783 	/* Interface address info used in eth_type_trans() */
1784 	unsigned char		*dev_addr;
1785 
1786 #ifdef CONFIG_SYSFS
1787 	struct netdev_rx_queue	*_rx;
1788 
1789 	unsigned int		num_rx_queues;
1790 	unsigned int		real_num_rx_queues;
1791 #endif
1792 
1793 	unsigned long		gro_flush_timeout;
1794 	rx_handler_func_t __rcu	*rx_handler;
1795 	void __rcu		*rx_handler_data;
1796 
1797 #ifdef CONFIG_NET_CLS_ACT
1798 	struct tcf_proto __rcu  *ingress_cl_list;
1799 #endif
1800 	struct netdev_queue __rcu *ingress_queue;
1801 #ifdef CONFIG_NETFILTER_INGRESS
1802 	struct nf_hook_entry __rcu *nf_hooks_ingress;
1803 #endif
1804 
1805 	unsigned char		broadcast[MAX_ADDR_LEN];
1806 #ifdef CONFIG_RFS_ACCEL
1807 	struct cpu_rmap		*rx_cpu_rmap;
1808 #endif
1809 	struct hlist_node	index_hlist;
1810 
1811 /*
1812  * Cache lines mostly used on transmit path
1813  */
1814 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1815 	unsigned int		num_tx_queues;
1816 	unsigned int		real_num_tx_queues;
1817 	struct Qdisc		*qdisc;
1818 #ifdef CONFIG_NET_SCHED
1819 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1820 #endif
1821 	unsigned long		tx_queue_len;
1822 	spinlock_t		tx_global_lock;
1823 	int			watchdog_timeo;
1824 
1825 #ifdef CONFIG_XPS
1826 	struct xps_dev_maps __rcu *xps_maps;
1827 #endif
1828 #ifdef CONFIG_NET_CLS_ACT
1829 	struct tcf_proto __rcu  *egress_cl_list;
1830 #endif
1831 
1832 	/* These may be needed for future network-power-down code. */
1833 	struct timer_list	watchdog_timer;
1834 
1835 	int __percpu		*pcpu_refcnt;
1836 	struct list_head	todo_list;
1837 
1838 	struct list_head	link_watch_list;
1839 
1840 	enum { NETREG_UNINITIALIZED=0,
1841 	       NETREG_REGISTERED,	/* completed register_netdevice */
1842 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1843 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1844 	       NETREG_RELEASED,		/* called free_netdev */
1845 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1846 	} reg_state:8;
1847 
1848 	bool dismantle;
1849 
1850 	enum {
1851 		RTNL_LINK_INITIALIZED,
1852 		RTNL_LINK_INITIALIZING,
1853 	} rtnl_link_state:16;
1854 
1855 	void (*destructor)(struct net_device *dev);
1856 
1857 #ifdef CONFIG_NETPOLL
1858 	struct netpoll_info __rcu	*npinfo;
1859 #endif
1860 
1861 	possible_net_t			nd_net;
1862 
1863 	/* mid-layer private */
1864 	union {
1865 		void					*ml_priv;
1866 		struct pcpu_lstats __percpu		*lstats;
1867 		struct pcpu_sw_netstats __percpu	*tstats;
1868 		struct pcpu_dstats __percpu		*dstats;
1869 		struct pcpu_vstats __percpu		*vstats;
1870 	};
1871 
1872 	struct garp_port __rcu	*garp_port;
1873 	struct mrp_port __rcu	*mrp_port;
1874 
1875 	struct device		dev;
1876 	const struct attribute_group *sysfs_groups[4];
1877 	const struct attribute_group *sysfs_rx_queue_group;
1878 
1879 	const struct rtnl_link_ops *rtnl_link_ops;
1880 
1881 	/* for setting kernel sock attribute on TCP connection setup */
1882 #define GSO_MAX_SIZE		65536
1883 	unsigned int		gso_max_size;
1884 #define GSO_MAX_SEGS		65535
1885 	u16			gso_max_segs;
1886 
1887 #ifdef CONFIG_DCB
1888 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1889 #endif
1890 	u8			num_tc;
1891 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1892 	u8			prio_tc_map[TC_BITMASK + 1];
1893 
1894 #if IS_ENABLED(CONFIG_FCOE)
1895 	unsigned int		fcoe_ddp_xid;
1896 #endif
1897 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1898 	struct netprio_map __rcu *priomap;
1899 #endif
1900 	struct phy_device	*phydev;
1901 	struct lock_class_key	*qdisc_tx_busylock;
1902 	struct lock_class_key	*qdisc_running_key;
1903 	bool			proto_down;
1904 };
1905 #define to_net_dev(d) container_of(d, struct net_device, dev)
1906 
1907 #define	NETDEV_ALIGN		32
1908 
1909 static inline
1910 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1911 {
1912 	return dev->prio_tc_map[prio & TC_BITMASK];
1913 }
1914 
1915 static inline
1916 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1917 {
1918 	if (tc >= dev->num_tc)
1919 		return -EINVAL;
1920 
1921 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1922 	return 0;
1923 }
1924 
1925 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1926 void netdev_reset_tc(struct net_device *dev);
1927 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1928 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1929 
1930 static inline
1931 int netdev_get_num_tc(struct net_device *dev)
1932 {
1933 	return dev->num_tc;
1934 }
1935 
1936 static inline
1937 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1938 					 unsigned int index)
1939 {
1940 	return &dev->_tx[index];
1941 }
1942 
1943 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1944 						    const struct sk_buff *skb)
1945 {
1946 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1947 }
1948 
1949 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1950 					    void (*f)(struct net_device *,
1951 						      struct netdev_queue *,
1952 						      void *),
1953 					    void *arg)
1954 {
1955 	unsigned int i;
1956 
1957 	for (i = 0; i < dev->num_tx_queues; i++)
1958 		f(dev, &dev->_tx[i], arg);
1959 }
1960 
1961 #define netdev_lockdep_set_classes(dev)				\
1962 {								\
1963 	static struct lock_class_key qdisc_tx_busylock_key;	\
1964 	static struct lock_class_key qdisc_running_key;		\
1965 	static struct lock_class_key qdisc_xmit_lock_key;	\
1966 	static struct lock_class_key dev_addr_list_lock_key;	\
1967 	unsigned int i;						\
1968 								\
1969 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
1970 	(dev)->qdisc_running_key = &qdisc_running_key;		\
1971 	lockdep_set_class(&(dev)->addr_list_lock,		\
1972 			  &dev_addr_list_lock_key); 		\
1973 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
1974 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
1975 				  &qdisc_xmit_lock_key);	\
1976 }
1977 
1978 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1979 				    struct sk_buff *skb,
1980 				    void *accel_priv);
1981 
1982 /* returns the headroom that the master device needs to take in account
1983  * when forwarding to this dev
1984  */
1985 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1986 {
1987 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1988 }
1989 
1990 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1991 {
1992 	if (dev->netdev_ops->ndo_set_rx_headroom)
1993 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1994 }
1995 
1996 /* set the device rx headroom to the dev's default */
1997 static inline void netdev_reset_rx_headroom(struct net_device *dev)
1998 {
1999 	netdev_set_rx_headroom(dev, -1);
2000 }
2001 
2002 /*
2003  * Net namespace inlines
2004  */
2005 static inline
2006 struct net *dev_net(const struct net_device *dev)
2007 {
2008 	return read_pnet(&dev->nd_net);
2009 }
2010 
2011 static inline
2012 void dev_net_set(struct net_device *dev, struct net *net)
2013 {
2014 	write_pnet(&dev->nd_net, net);
2015 }
2016 
2017 static inline bool netdev_uses_dsa(struct net_device *dev)
2018 {
2019 #if IS_ENABLED(CONFIG_NET_DSA)
2020 	if (dev->dsa_ptr != NULL)
2021 		return dsa_uses_tagged_protocol(dev->dsa_ptr);
2022 #endif
2023 	return false;
2024 }
2025 
2026 /**
2027  *	netdev_priv - access network device private data
2028  *	@dev: network device
2029  *
2030  * Get network device private data
2031  */
2032 static inline void *netdev_priv(const struct net_device *dev)
2033 {
2034 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2035 }
2036 
2037 /* Set the sysfs physical device reference for the network logical device
2038  * if set prior to registration will cause a symlink during initialization.
2039  */
2040 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2041 
2042 /* Set the sysfs device type for the network logical device to allow
2043  * fine-grained identification of different network device types. For
2044  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2045  */
2046 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2047 
2048 /* Default NAPI poll() weight
2049  * Device drivers are strongly advised to not use bigger value
2050  */
2051 #define NAPI_POLL_WEIGHT 64
2052 
2053 /**
2054  *	netif_napi_add - initialize a NAPI context
2055  *	@dev:  network device
2056  *	@napi: NAPI context
2057  *	@poll: polling function
2058  *	@weight: default weight
2059  *
2060  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2061  * *any* of the other NAPI-related functions.
2062  */
2063 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2064 		    int (*poll)(struct napi_struct *, int), int weight);
2065 
2066 /**
2067  *	netif_tx_napi_add - initialize a NAPI context
2068  *	@dev:  network device
2069  *	@napi: NAPI context
2070  *	@poll: polling function
2071  *	@weight: default weight
2072  *
2073  * This variant of netif_napi_add() should be used from drivers using NAPI
2074  * to exclusively poll a TX queue.
2075  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2076  */
2077 static inline void netif_tx_napi_add(struct net_device *dev,
2078 				     struct napi_struct *napi,
2079 				     int (*poll)(struct napi_struct *, int),
2080 				     int weight)
2081 {
2082 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2083 	netif_napi_add(dev, napi, poll, weight);
2084 }
2085 
2086 /**
2087  *  netif_napi_del - remove a NAPI context
2088  *  @napi: NAPI context
2089  *
2090  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2091  */
2092 void netif_napi_del(struct napi_struct *napi);
2093 
2094 struct napi_gro_cb {
2095 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2096 	void	*frag0;
2097 
2098 	/* Length of frag0. */
2099 	unsigned int frag0_len;
2100 
2101 	/* This indicates where we are processing relative to skb->data. */
2102 	int	data_offset;
2103 
2104 	/* This is non-zero if the packet cannot be merged with the new skb. */
2105 	u16	flush;
2106 
2107 	/* Save the IP ID here and check when we get to the transport layer */
2108 	u16	flush_id;
2109 
2110 	/* Number of segments aggregated. */
2111 	u16	count;
2112 
2113 	/* Start offset for remote checksum offload */
2114 	u16	gro_remcsum_start;
2115 
2116 	/* jiffies when first packet was created/queued */
2117 	unsigned long age;
2118 
2119 	/* Used in ipv6_gro_receive() and foo-over-udp */
2120 	u16	proto;
2121 
2122 	/* This is non-zero if the packet may be of the same flow. */
2123 	u8	same_flow:1;
2124 
2125 	/* Used in tunnel GRO receive */
2126 	u8	encap_mark:1;
2127 
2128 	/* GRO checksum is valid */
2129 	u8	csum_valid:1;
2130 
2131 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2132 	u8	csum_cnt:3;
2133 
2134 	/* Free the skb? */
2135 	u8	free:2;
2136 #define NAPI_GRO_FREE		  1
2137 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2138 
2139 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2140 	u8	is_ipv6:1;
2141 
2142 	/* Used in GRE, set in fou/gue_gro_receive */
2143 	u8	is_fou:1;
2144 
2145 	/* Used to determine if flush_id can be ignored */
2146 	u8	is_atomic:1;
2147 
2148 	/* Number of gro_receive callbacks this packet already went through */
2149 	u8 recursion_counter:4;
2150 
2151 	/* 1 bit hole */
2152 
2153 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2154 	__wsum	csum;
2155 
2156 	/* used in skb_gro_receive() slow path */
2157 	struct sk_buff *last;
2158 };
2159 
2160 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2161 
2162 #define GRO_RECURSION_LIMIT 15
2163 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2164 {
2165 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2166 }
2167 
2168 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2169 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2170 						struct sk_buff **head,
2171 						struct sk_buff *skb)
2172 {
2173 	if (unlikely(gro_recursion_inc_test(skb))) {
2174 		NAPI_GRO_CB(skb)->flush |= 1;
2175 		return NULL;
2176 	}
2177 
2178 	return cb(head, skb);
2179 }
2180 
2181 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2182 					     struct sk_buff *);
2183 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2184 						   struct sock *sk,
2185 						   struct sk_buff **head,
2186 						   struct sk_buff *skb)
2187 {
2188 	if (unlikely(gro_recursion_inc_test(skb))) {
2189 		NAPI_GRO_CB(skb)->flush |= 1;
2190 		return NULL;
2191 	}
2192 
2193 	return cb(sk, head, skb);
2194 }
2195 
2196 struct packet_type {
2197 	__be16			type;	/* This is really htons(ether_type). */
2198 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2199 	int			(*func) (struct sk_buff *,
2200 					 struct net_device *,
2201 					 struct packet_type *,
2202 					 struct net_device *);
2203 	bool			(*id_match)(struct packet_type *ptype,
2204 					    struct sock *sk);
2205 	void			*af_packet_priv;
2206 	struct list_head	list;
2207 };
2208 
2209 struct offload_callbacks {
2210 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2211 						netdev_features_t features);
2212 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2213 						 struct sk_buff *skb);
2214 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2215 };
2216 
2217 struct packet_offload {
2218 	__be16			 type;	/* This is really htons(ether_type). */
2219 	u16			 priority;
2220 	struct offload_callbacks callbacks;
2221 	struct list_head	 list;
2222 };
2223 
2224 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2225 struct pcpu_sw_netstats {
2226 	u64     rx_packets;
2227 	u64     rx_bytes;
2228 	u64     tx_packets;
2229 	u64     tx_bytes;
2230 	struct u64_stats_sync   syncp;
2231 };
2232 
2233 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2234 ({									\
2235 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2236 	if (pcpu_stats)	{						\
2237 		int __cpu;						\
2238 		for_each_possible_cpu(__cpu) {				\
2239 			typeof(type) *stat;				\
2240 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2241 			u64_stats_init(&stat->syncp);			\
2242 		}							\
2243 	}								\
2244 	pcpu_stats;							\
2245 })
2246 
2247 #define netdev_alloc_pcpu_stats(type)					\
2248 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2249 
2250 enum netdev_lag_tx_type {
2251 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2252 	NETDEV_LAG_TX_TYPE_RANDOM,
2253 	NETDEV_LAG_TX_TYPE_BROADCAST,
2254 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2255 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2256 	NETDEV_LAG_TX_TYPE_HASH,
2257 };
2258 
2259 struct netdev_lag_upper_info {
2260 	enum netdev_lag_tx_type tx_type;
2261 };
2262 
2263 struct netdev_lag_lower_state_info {
2264 	u8 link_up : 1,
2265 	   tx_enabled : 1;
2266 };
2267 
2268 #include <linux/notifier.h>
2269 
2270 /* netdevice notifier chain. Please remember to update the rtnetlink
2271  * notification exclusion list in rtnetlink_event() when adding new
2272  * types.
2273  */
2274 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2275 #define NETDEV_DOWN	0x0002
2276 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2277 				   detected a hardware crash and restarted
2278 				   - we can use this eg to kick tcp sessions
2279 				   once done */
2280 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2281 #define NETDEV_REGISTER 0x0005
2282 #define NETDEV_UNREGISTER	0x0006
2283 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2284 #define NETDEV_CHANGEADDR	0x0008
2285 #define NETDEV_GOING_DOWN	0x0009
2286 #define NETDEV_CHANGENAME	0x000A
2287 #define NETDEV_FEAT_CHANGE	0x000B
2288 #define NETDEV_BONDING_FAILOVER 0x000C
2289 #define NETDEV_PRE_UP		0x000D
2290 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2291 #define NETDEV_POST_TYPE_CHANGE	0x000F
2292 #define NETDEV_POST_INIT	0x0010
2293 #define NETDEV_UNREGISTER_FINAL 0x0011
2294 #define NETDEV_RELEASE		0x0012
2295 #define NETDEV_NOTIFY_PEERS	0x0013
2296 #define NETDEV_JOIN		0x0014
2297 #define NETDEV_CHANGEUPPER	0x0015
2298 #define NETDEV_RESEND_IGMP	0x0016
2299 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2300 #define NETDEV_CHANGEINFODATA	0x0018
2301 #define NETDEV_BONDING_INFO	0x0019
2302 #define NETDEV_PRECHANGEUPPER	0x001A
2303 #define NETDEV_CHANGELOWERSTATE	0x001B
2304 #define NETDEV_UDP_TUNNEL_PUSH_INFO	0x001C
2305 #define NETDEV_CHANGE_TX_QUEUE_LEN	0x001E
2306 
2307 int register_netdevice_notifier(struct notifier_block *nb);
2308 int unregister_netdevice_notifier(struct notifier_block *nb);
2309 
2310 struct netdev_notifier_info {
2311 	struct net_device *dev;
2312 };
2313 
2314 struct netdev_notifier_change_info {
2315 	struct netdev_notifier_info info; /* must be first */
2316 	unsigned int flags_changed;
2317 };
2318 
2319 struct netdev_notifier_changeupper_info {
2320 	struct netdev_notifier_info info; /* must be first */
2321 	struct net_device *upper_dev; /* new upper dev */
2322 	bool master; /* is upper dev master */
2323 	bool linking; /* is the notification for link or unlink */
2324 	void *upper_info; /* upper dev info */
2325 };
2326 
2327 struct netdev_notifier_changelowerstate_info {
2328 	struct netdev_notifier_info info; /* must be first */
2329 	void *lower_state_info; /* is lower dev state */
2330 };
2331 
2332 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2333 					     struct net_device *dev)
2334 {
2335 	info->dev = dev;
2336 }
2337 
2338 static inline struct net_device *
2339 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2340 {
2341 	return info->dev;
2342 }
2343 
2344 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2345 
2346 
2347 extern rwlock_t				dev_base_lock;		/* Device list lock */
2348 
2349 #define for_each_netdev(net, d)		\
2350 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2351 #define for_each_netdev_reverse(net, d)	\
2352 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2353 #define for_each_netdev_rcu(net, d)		\
2354 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2355 #define for_each_netdev_safe(net, d, n)	\
2356 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2357 #define for_each_netdev_continue(net, d)		\
2358 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2359 #define for_each_netdev_continue_rcu(net, d)		\
2360 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2361 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2362 		for_each_netdev_rcu(&init_net, slave)	\
2363 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2364 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2365 
2366 static inline struct net_device *next_net_device(struct net_device *dev)
2367 {
2368 	struct list_head *lh;
2369 	struct net *net;
2370 
2371 	net = dev_net(dev);
2372 	lh = dev->dev_list.next;
2373 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2374 }
2375 
2376 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2377 {
2378 	struct list_head *lh;
2379 	struct net *net;
2380 
2381 	net = dev_net(dev);
2382 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2383 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2384 }
2385 
2386 static inline struct net_device *first_net_device(struct net *net)
2387 {
2388 	return list_empty(&net->dev_base_head) ? NULL :
2389 		net_device_entry(net->dev_base_head.next);
2390 }
2391 
2392 static inline struct net_device *first_net_device_rcu(struct net *net)
2393 {
2394 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2395 
2396 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2397 }
2398 
2399 int netdev_boot_setup_check(struct net_device *dev);
2400 unsigned long netdev_boot_base(const char *prefix, int unit);
2401 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2402 				       const char *hwaddr);
2403 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2404 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2405 void dev_add_pack(struct packet_type *pt);
2406 void dev_remove_pack(struct packet_type *pt);
2407 void __dev_remove_pack(struct packet_type *pt);
2408 void dev_add_offload(struct packet_offload *po);
2409 void dev_remove_offload(struct packet_offload *po);
2410 
2411 int dev_get_iflink(const struct net_device *dev);
2412 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2413 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2414 				      unsigned short mask);
2415 struct net_device *dev_get_by_name(struct net *net, const char *name);
2416 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2417 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2418 int dev_alloc_name(struct net_device *dev, const char *name);
2419 int dev_open(struct net_device *dev);
2420 int dev_close(struct net_device *dev);
2421 int dev_close_many(struct list_head *head, bool unlink);
2422 void dev_disable_lro(struct net_device *dev);
2423 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2424 int dev_queue_xmit(struct sk_buff *skb);
2425 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2426 int register_netdevice(struct net_device *dev);
2427 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2428 void unregister_netdevice_many(struct list_head *head);
2429 static inline void unregister_netdevice(struct net_device *dev)
2430 {
2431 	unregister_netdevice_queue(dev, NULL);
2432 }
2433 
2434 int netdev_refcnt_read(const struct net_device *dev);
2435 void free_netdev(struct net_device *dev);
2436 void netdev_freemem(struct net_device *dev);
2437 void synchronize_net(void);
2438 int init_dummy_netdev(struct net_device *dev);
2439 
2440 DECLARE_PER_CPU(int, xmit_recursion);
2441 #define XMIT_RECURSION_LIMIT	10
2442 
2443 static inline int dev_recursion_level(void)
2444 {
2445 	return this_cpu_read(xmit_recursion);
2446 }
2447 
2448 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2449 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2450 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2451 int netdev_get_name(struct net *net, char *name, int ifindex);
2452 int dev_restart(struct net_device *dev);
2453 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2454 
2455 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2456 {
2457 	return NAPI_GRO_CB(skb)->data_offset;
2458 }
2459 
2460 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2461 {
2462 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2463 }
2464 
2465 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2466 {
2467 	NAPI_GRO_CB(skb)->data_offset += len;
2468 }
2469 
2470 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2471 					unsigned int offset)
2472 {
2473 	return NAPI_GRO_CB(skb)->frag0 + offset;
2474 }
2475 
2476 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2477 {
2478 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2479 }
2480 
2481 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2482 {
2483 	NAPI_GRO_CB(skb)->frag0 = NULL;
2484 	NAPI_GRO_CB(skb)->frag0_len = 0;
2485 }
2486 
2487 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2488 					unsigned int offset)
2489 {
2490 	if (!pskb_may_pull(skb, hlen))
2491 		return NULL;
2492 
2493 	skb_gro_frag0_invalidate(skb);
2494 	return skb->data + offset;
2495 }
2496 
2497 static inline void *skb_gro_network_header(struct sk_buff *skb)
2498 {
2499 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2500 	       skb_network_offset(skb);
2501 }
2502 
2503 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2504 					const void *start, unsigned int len)
2505 {
2506 	if (NAPI_GRO_CB(skb)->csum_valid)
2507 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2508 						  csum_partial(start, len, 0));
2509 }
2510 
2511 /* GRO checksum functions. These are logical equivalents of the normal
2512  * checksum functions (in skbuff.h) except that they operate on the GRO
2513  * offsets and fields in sk_buff.
2514  */
2515 
2516 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2517 
2518 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2519 {
2520 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2521 }
2522 
2523 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2524 						      bool zero_okay,
2525 						      __sum16 check)
2526 {
2527 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2528 		skb_checksum_start_offset(skb) <
2529 		 skb_gro_offset(skb)) &&
2530 		!skb_at_gro_remcsum_start(skb) &&
2531 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2532 		(!zero_okay || check));
2533 }
2534 
2535 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2536 							   __wsum psum)
2537 {
2538 	if (NAPI_GRO_CB(skb)->csum_valid &&
2539 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2540 		return 0;
2541 
2542 	NAPI_GRO_CB(skb)->csum = psum;
2543 
2544 	return __skb_gro_checksum_complete(skb);
2545 }
2546 
2547 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2548 {
2549 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2550 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2551 		NAPI_GRO_CB(skb)->csum_cnt--;
2552 	} else {
2553 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2554 		 * verified a new top level checksum or an encapsulated one
2555 		 * during GRO. This saves work if we fallback to normal path.
2556 		 */
2557 		__skb_incr_checksum_unnecessary(skb);
2558 	}
2559 }
2560 
2561 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2562 				    compute_pseudo)			\
2563 ({									\
2564 	__sum16 __ret = 0;						\
2565 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2566 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2567 				compute_pseudo(skb, proto));		\
2568 	if (__ret)							\
2569 		__skb_mark_checksum_bad(skb);				\
2570 	else								\
2571 		skb_gro_incr_csum_unnecessary(skb);			\
2572 	__ret;								\
2573 })
2574 
2575 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2576 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2577 
2578 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2579 					     compute_pseudo)		\
2580 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2581 
2582 #define skb_gro_checksum_simple_validate(skb)				\
2583 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2584 
2585 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2586 {
2587 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2588 		!NAPI_GRO_CB(skb)->csum_valid);
2589 }
2590 
2591 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2592 					      __sum16 check, __wsum pseudo)
2593 {
2594 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2595 	NAPI_GRO_CB(skb)->csum_valid = 1;
2596 }
2597 
2598 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2599 do {									\
2600 	if (__skb_gro_checksum_convert_check(skb))			\
2601 		__skb_gro_checksum_convert(skb, check,			\
2602 					   compute_pseudo(skb, proto));	\
2603 } while (0)
2604 
2605 struct gro_remcsum {
2606 	int offset;
2607 	__wsum delta;
2608 };
2609 
2610 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2611 {
2612 	grc->offset = 0;
2613 	grc->delta = 0;
2614 }
2615 
2616 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2617 					    unsigned int off, size_t hdrlen,
2618 					    int start, int offset,
2619 					    struct gro_remcsum *grc,
2620 					    bool nopartial)
2621 {
2622 	__wsum delta;
2623 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2624 
2625 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2626 
2627 	if (!nopartial) {
2628 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2629 		return ptr;
2630 	}
2631 
2632 	ptr = skb_gro_header_fast(skb, off);
2633 	if (skb_gro_header_hard(skb, off + plen)) {
2634 		ptr = skb_gro_header_slow(skb, off + plen, off);
2635 		if (!ptr)
2636 			return NULL;
2637 	}
2638 
2639 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2640 			       start, offset);
2641 
2642 	/* Adjust skb->csum since we changed the packet */
2643 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2644 
2645 	grc->offset = off + hdrlen + offset;
2646 	grc->delta = delta;
2647 
2648 	return ptr;
2649 }
2650 
2651 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2652 					   struct gro_remcsum *grc)
2653 {
2654 	void *ptr;
2655 	size_t plen = grc->offset + sizeof(u16);
2656 
2657 	if (!grc->delta)
2658 		return;
2659 
2660 	ptr = skb_gro_header_fast(skb, grc->offset);
2661 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2662 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2663 		if (!ptr)
2664 			return;
2665 	}
2666 
2667 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2668 }
2669 
2670 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2671 				  unsigned short type,
2672 				  const void *daddr, const void *saddr,
2673 				  unsigned int len)
2674 {
2675 	if (!dev->header_ops || !dev->header_ops->create)
2676 		return 0;
2677 
2678 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2679 }
2680 
2681 static inline int dev_parse_header(const struct sk_buff *skb,
2682 				   unsigned char *haddr)
2683 {
2684 	const struct net_device *dev = skb->dev;
2685 
2686 	if (!dev->header_ops || !dev->header_ops->parse)
2687 		return 0;
2688 	return dev->header_ops->parse(skb, haddr);
2689 }
2690 
2691 /* ll_header must have at least hard_header_len allocated */
2692 static inline bool dev_validate_header(const struct net_device *dev,
2693 				       char *ll_header, int len)
2694 {
2695 	if (likely(len >= dev->hard_header_len))
2696 		return true;
2697 
2698 	if (capable(CAP_SYS_RAWIO)) {
2699 		memset(ll_header + len, 0, dev->hard_header_len - len);
2700 		return true;
2701 	}
2702 
2703 	if (dev->header_ops && dev->header_ops->validate)
2704 		return dev->header_ops->validate(ll_header, len);
2705 
2706 	return false;
2707 }
2708 
2709 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2710 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2711 static inline int unregister_gifconf(unsigned int family)
2712 {
2713 	return register_gifconf(family, NULL);
2714 }
2715 
2716 #ifdef CONFIG_NET_FLOW_LIMIT
2717 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2718 struct sd_flow_limit {
2719 	u64			count;
2720 	unsigned int		num_buckets;
2721 	unsigned int		history_head;
2722 	u16			history[FLOW_LIMIT_HISTORY];
2723 	u8			buckets[];
2724 };
2725 
2726 extern int netdev_flow_limit_table_len;
2727 #endif /* CONFIG_NET_FLOW_LIMIT */
2728 
2729 /*
2730  * Incoming packets are placed on per-CPU queues
2731  */
2732 struct softnet_data {
2733 	struct list_head	poll_list;
2734 	struct sk_buff_head	process_queue;
2735 
2736 	/* stats */
2737 	unsigned int		processed;
2738 	unsigned int		time_squeeze;
2739 	unsigned int		received_rps;
2740 #ifdef CONFIG_RPS
2741 	struct softnet_data	*rps_ipi_list;
2742 #endif
2743 #ifdef CONFIG_NET_FLOW_LIMIT
2744 	struct sd_flow_limit __rcu *flow_limit;
2745 #endif
2746 	struct Qdisc		*output_queue;
2747 	struct Qdisc		**output_queue_tailp;
2748 	struct sk_buff		*completion_queue;
2749 
2750 #ifdef CONFIG_RPS
2751 	/* input_queue_head should be written by cpu owning this struct,
2752 	 * and only read by other cpus. Worth using a cache line.
2753 	 */
2754 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2755 
2756 	/* Elements below can be accessed between CPUs for RPS/RFS */
2757 	struct call_single_data	csd ____cacheline_aligned_in_smp;
2758 	struct softnet_data	*rps_ipi_next;
2759 	unsigned int		cpu;
2760 	unsigned int		input_queue_tail;
2761 #endif
2762 	unsigned int		dropped;
2763 	struct sk_buff_head	input_pkt_queue;
2764 	struct napi_struct	backlog;
2765 
2766 };
2767 
2768 static inline void input_queue_head_incr(struct softnet_data *sd)
2769 {
2770 #ifdef CONFIG_RPS
2771 	sd->input_queue_head++;
2772 #endif
2773 }
2774 
2775 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2776 					      unsigned int *qtail)
2777 {
2778 #ifdef CONFIG_RPS
2779 	*qtail = ++sd->input_queue_tail;
2780 #endif
2781 }
2782 
2783 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2784 
2785 void __netif_schedule(struct Qdisc *q);
2786 void netif_schedule_queue(struct netdev_queue *txq);
2787 
2788 static inline void netif_tx_schedule_all(struct net_device *dev)
2789 {
2790 	unsigned int i;
2791 
2792 	for (i = 0; i < dev->num_tx_queues; i++)
2793 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2794 }
2795 
2796 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2797 {
2798 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2799 }
2800 
2801 /**
2802  *	netif_start_queue - allow transmit
2803  *	@dev: network device
2804  *
2805  *	Allow upper layers to call the device hard_start_xmit routine.
2806  */
2807 static inline void netif_start_queue(struct net_device *dev)
2808 {
2809 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2810 }
2811 
2812 static inline void netif_tx_start_all_queues(struct net_device *dev)
2813 {
2814 	unsigned int i;
2815 
2816 	for (i = 0; i < dev->num_tx_queues; i++) {
2817 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2818 		netif_tx_start_queue(txq);
2819 	}
2820 }
2821 
2822 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2823 
2824 /**
2825  *	netif_wake_queue - restart transmit
2826  *	@dev: network device
2827  *
2828  *	Allow upper layers to call the device hard_start_xmit routine.
2829  *	Used for flow control when transmit resources are available.
2830  */
2831 static inline void netif_wake_queue(struct net_device *dev)
2832 {
2833 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2834 }
2835 
2836 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2837 {
2838 	unsigned int i;
2839 
2840 	for (i = 0; i < dev->num_tx_queues; i++) {
2841 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2842 		netif_tx_wake_queue(txq);
2843 	}
2844 }
2845 
2846 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2847 {
2848 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2849 }
2850 
2851 /**
2852  *	netif_stop_queue - stop transmitted packets
2853  *	@dev: network device
2854  *
2855  *	Stop upper layers calling the device hard_start_xmit routine.
2856  *	Used for flow control when transmit resources are unavailable.
2857  */
2858 static inline void netif_stop_queue(struct net_device *dev)
2859 {
2860 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2861 }
2862 
2863 void netif_tx_stop_all_queues(struct net_device *dev);
2864 
2865 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2866 {
2867 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2868 }
2869 
2870 /**
2871  *	netif_queue_stopped - test if transmit queue is flowblocked
2872  *	@dev: network device
2873  *
2874  *	Test if transmit queue on device is currently unable to send.
2875  */
2876 static inline bool netif_queue_stopped(const struct net_device *dev)
2877 {
2878 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2879 }
2880 
2881 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2882 {
2883 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2884 }
2885 
2886 static inline bool
2887 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2888 {
2889 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2890 }
2891 
2892 static inline bool
2893 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2894 {
2895 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2896 }
2897 
2898 /**
2899  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2900  *	@dev_queue: pointer to transmit queue
2901  *
2902  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2903  * to give appropriate hint to the CPU.
2904  */
2905 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2906 {
2907 #ifdef CONFIG_BQL
2908 	prefetchw(&dev_queue->dql.num_queued);
2909 #endif
2910 }
2911 
2912 /**
2913  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2914  *	@dev_queue: pointer to transmit queue
2915  *
2916  * BQL enabled drivers might use this helper in their TX completion path,
2917  * to give appropriate hint to the CPU.
2918  */
2919 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2920 {
2921 #ifdef CONFIG_BQL
2922 	prefetchw(&dev_queue->dql.limit);
2923 #endif
2924 }
2925 
2926 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2927 					unsigned int bytes)
2928 {
2929 #ifdef CONFIG_BQL
2930 	dql_queued(&dev_queue->dql, bytes);
2931 
2932 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2933 		return;
2934 
2935 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2936 
2937 	/*
2938 	 * The XOFF flag must be set before checking the dql_avail below,
2939 	 * because in netdev_tx_completed_queue we update the dql_completed
2940 	 * before checking the XOFF flag.
2941 	 */
2942 	smp_mb();
2943 
2944 	/* check again in case another CPU has just made room avail */
2945 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2946 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2947 #endif
2948 }
2949 
2950 /**
2951  * 	netdev_sent_queue - report the number of bytes queued to hardware
2952  * 	@dev: network device
2953  * 	@bytes: number of bytes queued to the hardware device queue
2954  *
2955  * 	Report the number of bytes queued for sending/completion to the network
2956  * 	device hardware queue. @bytes should be a good approximation and should
2957  * 	exactly match netdev_completed_queue() @bytes
2958  */
2959 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2960 {
2961 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2962 }
2963 
2964 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2965 					     unsigned int pkts, unsigned int bytes)
2966 {
2967 #ifdef CONFIG_BQL
2968 	if (unlikely(!bytes))
2969 		return;
2970 
2971 	dql_completed(&dev_queue->dql, bytes);
2972 
2973 	/*
2974 	 * Without the memory barrier there is a small possiblity that
2975 	 * netdev_tx_sent_queue will miss the update and cause the queue to
2976 	 * be stopped forever
2977 	 */
2978 	smp_mb();
2979 
2980 	if (dql_avail(&dev_queue->dql) < 0)
2981 		return;
2982 
2983 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2984 		netif_schedule_queue(dev_queue);
2985 #endif
2986 }
2987 
2988 /**
2989  * 	netdev_completed_queue - report bytes and packets completed by device
2990  * 	@dev: network device
2991  * 	@pkts: actual number of packets sent over the medium
2992  * 	@bytes: actual number of bytes sent over the medium
2993  *
2994  * 	Report the number of bytes and packets transmitted by the network device
2995  * 	hardware queue over the physical medium, @bytes must exactly match the
2996  * 	@bytes amount passed to netdev_sent_queue()
2997  */
2998 static inline void netdev_completed_queue(struct net_device *dev,
2999 					  unsigned int pkts, unsigned int bytes)
3000 {
3001 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3002 }
3003 
3004 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3005 {
3006 #ifdef CONFIG_BQL
3007 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3008 	dql_reset(&q->dql);
3009 #endif
3010 }
3011 
3012 /**
3013  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3014  * 	@dev_queue: network device
3015  *
3016  * 	Reset the bytes and packet count of a network device and clear the
3017  * 	software flow control OFF bit for this network device
3018  */
3019 static inline void netdev_reset_queue(struct net_device *dev_queue)
3020 {
3021 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3022 }
3023 
3024 /**
3025  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3026  * 	@dev: network device
3027  * 	@queue_index: given tx queue index
3028  *
3029  * 	Returns 0 if given tx queue index >= number of device tx queues,
3030  * 	otherwise returns the originally passed tx queue index.
3031  */
3032 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3033 {
3034 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3035 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3036 				     dev->name, queue_index,
3037 				     dev->real_num_tx_queues);
3038 		return 0;
3039 	}
3040 
3041 	return queue_index;
3042 }
3043 
3044 /**
3045  *	netif_running - test if up
3046  *	@dev: network device
3047  *
3048  *	Test if the device has been brought up.
3049  */
3050 static inline bool netif_running(const struct net_device *dev)
3051 {
3052 	return test_bit(__LINK_STATE_START, &dev->state);
3053 }
3054 
3055 /*
3056  * Routines to manage the subqueues on a device.  We only need start,
3057  * stop, and a check if it's stopped.  All other device management is
3058  * done at the overall netdevice level.
3059  * Also test the device if we're multiqueue.
3060  */
3061 
3062 /**
3063  *	netif_start_subqueue - allow sending packets on subqueue
3064  *	@dev: network device
3065  *	@queue_index: sub queue index
3066  *
3067  * Start individual transmit queue of a device with multiple transmit queues.
3068  */
3069 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3070 {
3071 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3072 
3073 	netif_tx_start_queue(txq);
3074 }
3075 
3076 /**
3077  *	netif_stop_subqueue - stop sending packets on subqueue
3078  *	@dev: network device
3079  *	@queue_index: sub queue index
3080  *
3081  * Stop individual transmit queue of a device with multiple transmit queues.
3082  */
3083 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3084 {
3085 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3086 	netif_tx_stop_queue(txq);
3087 }
3088 
3089 /**
3090  *	netif_subqueue_stopped - test status of subqueue
3091  *	@dev: network device
3092  *	@queue_index: sub queue index
3093  *
3094  * Check individual transmit queue of a device with multiple transmit queues.
3095  */
3096 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3097 					    u16 queue_index)
3098 {
3099 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3100 
3101 	return netif_tx_queue_stopped(txq);
3102 }
3103 
3104 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3105 					  struct sk_buff *skb)
3106 {
3107 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3108 }
3109 
3110 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3111 
3112 #ifdef CONFIG_XPS
3113 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3114 			u16 index);
3115 #else
3116 static inline int netif_set_xps_queue(struct net_device *dev,
3117 				      const struct cpumask *mask,
3118 				      u16 index)
3119 {
3120 	return 0;
3121 }
3122 #endif
3123 
3124 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3125 		  unsigned int num_tx_queues);
3126 
3127 /*
3128  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3129  * as a distribution range limit for the returned value.
3130  */
3131 static inline u16 skb_tx_hash(const struct net_device *dev,
3132 			      struct sk_buff *skb)
3133 {
3134 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3135 }
3136 
3137 /**
3138  *	netif_is_multiqueue - test if device has multiple transmit queues
3139  *	@dev: network device
3140  *
3141  * Check if device has multiple transmit queues
3142  */
3143 static inline bool netif_is_multiqueue(const struct net_device *dev)
3144 {
3145 	return dev->num_tx_queues > 1;
3146 }
3147 
3148 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3149 
3150 #ifdef CONFIG_SYSFS
3151 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3152 #else
3153 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3154 						unsigned int rxq)
3155 {
3156 	return 0;
3157 }
3158 #endif
3159 
3160 #ifdef CONFIG_SYSFS
3161 static inline unsigned int get_netdev_rx_queue_index(
3162 		struct netdev_rx_queue *queue)
3163 {
3164 	struct net_device *dev = queue->dev;
3165 	int index = queue - dev->_rx;
3166 
3167 	BUG_ON(index >= dev->num_rx_queues);
3168 	return index;
3169 }
3170 #endif
3171 
3172 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3173 int netif_get_num_default_rss_queues(void);
3174 
3175 enum skb_free_reason {
3176 	SKB_REASON_CONSUMED,
3177 	SKB_REASON_DROPPED,
3178 };
3179 
3180 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3181 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3182 
3183 /*
3184  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3185  * interrupt context or with hardware interrupts being disabled.
3186  * (in_irq() || irqs_disabled())
3187  *
3188  * We provide four helpers that can be used in following contexts :
3189  *
3190  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3191  *  replacing kfree_skb(skb)
3192  *
3193  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3194  *  Typically used in place of consume_skb(skb) in TX completion path
3195  *
3196  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3197  *  replacing kfree_skb(skb)
3198  *
3199  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3200  *  and consumed a packet. Used in place of consume_skb(skb)
3201  */
3202 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3203 {
3204 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3205 }
3206 
3207 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3208 {
3209 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3210 }
3211 
3212 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3213 {
3214 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3215 }
3216 
3217 static inline void dev_consume_skb_any(struct sk_buff *skb)
3218 {
3219 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3220 }
3221 
3222 int netif_rx(struct sk_buff *skb);
3223 int netif_rx_ni(struct sk_buff *skb);
3224 int netif_receive_skb(struct sk_buff *skb);
3225 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3226 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3227 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3228 gro_result_t napi_gro_frags(struct napi_struct *napi);
3229 struct packet_offload *gro_find_receive_by_type(__be16 type);
3230 struct packet_offload *gro_find_complete_by_type(__be16 type);
3231 
3232 static inline void napi_free_frags(struct napi_struct *napi)
3233 {
3234 	kfree_skb(napi->skb);
3235 	napi->skb = NULL;
3236 }
3237 
3238 bool netdev_is_rx_handler_busy(struct net_device *dev);
3239 int netdev_rx_handler_register(struct net_device *dev,
3240 			       rx_handler_func_t *rx_handler,
3241 			       void *rx_handler_data);
3242 void netdev_rx_handler_unregister(struct net_device *dev);
3243 
3244 bool dev_valid_name(const char *name);
3245 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3246 int dev_ethtool(struct net *net, struct ifreq *);
3247 unsigned int dev_get_flags(const struct net_device *);
3248 int __dev_change_flags(struct net_device *, unsigned int flags);
3249 int dev_change_flags(struct net_device *, unsigned int);
3250 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3251 			unsigned int gchanges);
3252 int dev_change_name(struct net_device *, const char *);
3253 int dev_set_alias(struct net_device *, const char *, size_t);
3254 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3255 int dev_set_mtu(struct net_device *, int);
3256 void dev_set_group(struct net_device *, int);
3257 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3258 int dev_change_carrier(struct net_device *, bool new_carrier);
3259 int dev_get_phys_port_id(struct net_device *dev,
3260 			 struct netdev_phys_item_id *ppid);
3261 int dev_get_phys_port_name(struct net_device *dev,
3262 			   char *name, size_t len);
3263 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3264 int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
3265 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3266 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3267 				    struct netdev_queue *txq, int *ret);
3268 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3269 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3270 bool is_skb_forwardable(const struct net_device *dev,
3271 			const struct sk_buff *skb);
3272 
3273 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3274 					       struct sk_buff *skb)
3275 {
3276 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3277 	    unlikely(!is_skb_forwardable(dev, skb))) {
3278 		atomic_long_inc(&dev->rx_dropped);
3279 		kfree_skb(skb);
3280 		return NET_RX_DROP;
3281 	}
3282 
3283 	skb_scrub_packet(skb, true);
3284 	skb->priority = 0;
3285 	return 0;
3286 }
3287 
3288 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3289 
3290 extern int		netdev_budget;
3291 
3292 /* Called by rtnetlink.c:rtnl_unlock() */
3293 void netdev_run_todo(void);
3294 
3295 /**
3296  *	dev_put - release reference to device
3297  *	@dev: network device
3298  *
3299  * Release reference to device to allow it to be freed.
3300  */
3301 static inline void dev_put(struct net_device *dev)
3302 {
3303 	this_cpu_dec(*dev->pcpu_refcnt);
3304 }
3305 
3306 /**
3307  *	dev_hold - get reference to device
3308  *	@dev: network device
3309  *
3310  * Hold reference to device to keep it from being freed.
3311  */
3312 static inline void dev_hold(struct net_device *dev)
3313 {
3314 	this_cpu_inc(*dev->pcpu_refcnt);
3315 }
3316 
3317 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3318  * and _off may be called from IRQ context, but it is caller
3319  * who is responsible for serialization of these calls.
3320  *
3321  * The name carrier is inappropriate, these functions should really be
3322  * called netif_lowerlayer_*() because they represent the state of any
3323  * kind of lower layer not just hardware media.
3324  */
3325 
3326 void linkwatch_init_dev(struct net_device *dev);
3327 void linkwatch_fire_event(struct net_device *dev);
3328 void linkwatch_forget_dev(struct net_device *dev);
3329 
3330 /**
3331  *	netif_carrier_ok - test if carrier present
3332  *	@dev: network device
3333  *
3334  * Check if carrier is present on device
3335  */
3336 static inline bool netif_carrier_ok(const struct net_device *dev)
3337 {
3338 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3339 }
3340 
3341 unsigned long dev_trans_start(struct net_device *dev);
3342 
3343 void __netdev_watchdog_up(struct net_device *dev);
3344 
3345 void netif_carrier_on(struct net_device *dev);
3346 
3347 void netif_carrier_off(struct net_device *dev);
3348 
3349 /**
3350  *	netif_dormant_on - mark device as dormant.
3351  *	@dev: network device
3352  *
3353  * Mark device as dormant (as per RFC2863).
3354  *
3355  * The dormant state indicates that the relevant interface is not
3356  * actually in a condition to pass packets (i.e., it is not 'up') but is
3357  * in a "pending" state, waiting for some external event.  For "on-
3358  * demand" interfaces, this new state identifies the situation where the
3359  * interface is waiting for events to place it in the up state.
3360  */
3361 static inline void netif_dormant_on(struct net_device *dev)
3362 {
3363 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3364 		linkwatch_fire_event(dev);
3365 }
3366 
3367 /**
3368  *	netif_dormant_off - set device as not dormant.
3369  *	@dev: network device
3370  *
3371  * Device is not in dormant state.
3372  */
3373 static inline void netif_dormant_off(struct net_device *dev)
3374 {
3375 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3376 		linkwatch_fire_event(dev);
3377 }
3378 
3379 /**
3380  *	netif_dormant - test if carrier present
3381  *	@dev: network device
3382  *
3383  * Check if carrier is present on device
3384  */
3385 static inline bool netif_dormant(const struct net_device *dev)
3386 {
3387 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3388 }
3389 
3390 
3391 /**
3392  *	netif_oper_up - test if device is operational
3393  *	@dev: network device
3394  *
3395  * Check if carrier is operational
3396  */
3397 static inline bool netif_oper_up(const struct net_device *dev)
3398 {
3399 	return (dev->operstate == IF_OPER_UP ||
3400 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3401 }
3402 
3403 /**
3404  *	netif_device_present - is device available or removed
3405  *	@dev: network device
3406  *
3407  * Check if device has not been removed from system.
3408  */
3409 static inline bool netif_device_present(struct net_device *dev)
3410 {
3411 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3412 }
3413 
3414 void netif_device_detach(struct net_device *dev);
3415 
3416 void netif_device_attach(struct net_device *dev);
3417 
3418 /*
3419  * Network interface message level settings
3420  */
3421 
3422 enum {
3423 	NETIF_MSG_DRV		= 0x0001,
3424 	NETIF_MSG_PROBE		= 0x0002,
3425 	NETIF_MSG_LINK		= 0x0004,
3426 	NETIF_MSG_TIMER		= 0x0008,
3427 	NETIF_MSG_IFDOWN	= 0x0010,
3428 	NETIF_MSG_IFUP		= 0x0020,
3429 	NETIF_MSG_RX_ERR	= 0x0040,
3430 	NETIF_MSG_TX_ERR	= 0x0080,
3431 	NETIF_MSG_TX_QUEUED	= 0x0100,
3432 	NETIF_MSG_INTR		= 0x0200,
3433 	NETIF_MSG_TX_DONE	= 0x0400,
3434 	NETIF_MSG_RX_STATUS	= 0x0800,
3435 	NETIF_MSG_PKTDATA	= 0x1000,
3436 	NETIF_MSG_HW		= 0x2000,
3437 	NETIF_MSG_WOL		= 0x4000,
3438 };
3439 
3440 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3441 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3442 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3443 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3444 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3445 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3446 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3447 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3448 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3449 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3450 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3451 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3452 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3453 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3454 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3455 
3456 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3457 {
3458 	/* use default */
3459 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3460 		return default_msg_enable_bits;
3461 	if (debug_value == 0)	/* no output */
3462 		return 0;
3463 	/* set low N bits */
3464 	return (1 << debug_value) - 1;
3465 }
3466 
3467 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3468 {
3469 	spin_lock(&txq->_xmit_lock);
3470 	txq->xmit_lock_owner = cpu;
3471 }
3472 
3473 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3474 {
3475 	__acquire(&txq->_xmit_lock);
3476 	return true;
3477 }
3478 
3479 static inline void __netif_tx_release(struct netdev_queue *txq)
3480 {
3481 	__release(&txq->_xmit_lock);
3482 }
3483 
3484 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3485 {
3486 	spin_lock_bh(&txq->_xmit_lock);
3487 	txq->xmit_lock_owner = smp_processor_id();
3488 }
3489 
3490 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3491 {
3492 	bool ok = spin_trylock(&txq->_xmit_lock);
3493 	if (likely(ok))
3494 		txq->xmit_lock_owner = smp_processor_id();
3495 	return ok;
3496 }
3497 
3498 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3499 {
3500 	txq->xmit_lock_owner = -1;
3501 	spin_unlock(&txq->_xmit_lock);
3502 }
3503 
3504 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3505 {
3506 	txq->xmit_lock_owner = -1;
3507 	spin_unlock_bh(&txq->_xmit_lock);
3508 }
3509 
3510 static inline void txq_trans_update(struct netdev_queue *txq)
3511 {
3512 	if (txq->xmit_lock_owner != -1)
3513 		txq->trans_start = jiffies;
3514 }
3515 
3516 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3517 static inline void netif_trans_update(struct net_device *dev)
3518 {
3519 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3520 
3521 	if (txq->trans_start != jiffies)
3522 		txq->trans_start = jiffies;
3523 }
3524 
3525 /**
3526  *	netif_tx_lock - grab network device transmit lock
3527  *	@dev: network device
3528  *
3529  * Get network device transmit lock
3530  */
3531 static inline void netif_tx_lock(struct net_device *dev)
3532 {
3533 	unsigned int i;
3534 	int cpu;
3535 
3536 	spin_lock(&dev->tx_global_lock);
3537 	cpu = smp_processor_id();
3538 	for (i = 0; i < dev->num_tx_queues; i++) {
3539 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3540 
3541 		/* We are the only thread of execution doing a
3542 		 * freeze, but we have to grab the _xmit_lock in
3543 		 * order to synchronize with threads which are in
3544 		 * the ->hard_start_xmit() handler and already
3545 		 * checked the frozen bit.
3546 		 */
3547 		__netif_tx_lock(txq, cpu);
3548 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3549 		__netif_tx_unlock(txq);
3550 	}
3551 }
3552 
3553 static inline void netif_tx_lock_bh(struct net_device *dev)
3554 {
3555 	local_bh_disable();
3556 	netif_tx_lock(dev);
3557 }
3558 
3559 static inline void netif_tx_unlock(struct net_device *dev)
3560 {
3561 	unsigned int i;
3562 
3563 	for (i = 0; i < dev->num_tx_queues; i++) {
3564 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3565 
3566 		/* No need to grab the _xmit_lock here.  If the
3567 		 * queue is not stopped for another reason, we
3568 		 * force a schedule.
3569 		 */
3570 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3571 		netif_schedule_queue(txq);
3572 	}
3573 	spin_unlock(&dev->tx_global_lock);
3574 }
3575 
3576 static inline void netif_tx_unlock_bh(struct net_device *dev)
3577 {
3578 	netif_tx_unlock(dev);
3579 	local_bh_enable();
3580 }
3581 
3582 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3583 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3584 		__netif_tx_lock(txq, cpu);		\
3585 	} else {					\
3586 		__netif_tx_acquire(txq);		\
3587 	}						\
3588 }
3589 
3590 #define HARD_TX_TRYLOCK(dev, txq)			\
3591 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3592 		__netif_tx_trylock(txq) :		\
3593 		__netif_tx_acquire(txq))
3594 
3595 #define HARD_TX_UNLOCK(dev, txq) {			\
3596 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3597 		__netif_tx_unlock(txq);			\
3598 	} else {					\
3599 		__netif_tx_release(txq);		\
3600 	}						\
3601 }
3602 
3603 static inline void netif_tx_disable(struct net_device *dev)
3604 {
3605 	unsigned int i;
3606 	int cpu;
3607 
3608 	local_bh_disable();
3609 	cpu = smp_processor_id();
3610 	for (i = 0; i < dev->num_tx_queues; i++) {
3611 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3612 
3613 		__netif_tx_lock(txq, cpu);
3614 		netif_tx_stop_queue(txq);
3615 		__netif_tx_unlock(txq);
3616 	}
3617 	local_bh_enable();
3618 }
3619 
3620 static inline void netif_addr_lock(struct net_device *dev)
3621 {
3622 	spin_lock(&dev->addr_list_lock);
3623 }
3624 
3625 static inline void netif_addr_lock_nested(struct net_device *dev)
3626 {
3627 	int subclass = SINGLE_DEPTH_NESTING;
3628 
3629 	if (dev->netdev_ops->ndo_get_lock_subclass)
3630 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3631 
3632 	spin_lock_nested(&dev->addr_list_lock, subclass);
3633 }
3634 
3635 static inline void netif_addr_lock_bh(struct net_device *dev)
3636 {
3637 	spin_lock_bh(&dev->addr_list_lock);
3638 }
3639 
3640 static inline void netif_addr_unlock(struct net_device *dev)
3641 {
3642 	spin_unlock(&dev->addr_list_lock);
3643 }
3644 
3645 static inline void netif_addr_unlock_bh(struct net_device *dev)
3646 {
3647 	spin_unlock_bh(&dev->addr_list_lock);
3648 }
3649 
3650 /*
3651  * dev_addrs walker. Should be used only for read access. Call with
3652  * rcu_read_lock held.
3653  */
3654 #define for_each_dev_addr(dev, ha) \
3655 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3656 
3657 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3658 
3659 void ether_setup(struct net_device *dev);
3660 
3661 /* Support for loadable net-drivers */
3662 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3663 				    unsigned char name_assign_type,
3664 				    void (*setup)(struct net_device *),
3665 				    unsigned int txqs, unsigned int rxqs);
3666 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3667 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3668 
3669 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3670 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3671 			 count)
3672 
3673 int register_netdev(struct net_device *dev);
3674 void unregister_netdev(struct net_device *dev);
3675 
3676 /* General hardware address lists handling functions */
3677 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3678 		   struct netdev_hw_addr_list *from_list, int addr_len);
3679 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3680 		      struct netdev_hw_addr_list *from_list, int addr_len);
3681 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3682 		       struct net_device *dev,
3683 		       int (*sync)(struct net_device *, const unsigned char *),
3684 		       int (*unsync)(struct net_device *,
3685 				     const unsigned char *));
3686 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3687 			  struct net_device *dev,
3688 			  int (*unsync)(struct net_device *,
3689 					const unsigned char *));
3690 void __hw_addr_init(struct netdev_hw_addr_list *list);
3691 
3692 /* Functions used for device addresses handling */
3693 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3694 		 unsigned char addr_type);
3695 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3696 		 unsigned char addr_type);
3697 void dev_addr_flush(struct net_device *dev);
3698 int dev_addr_init(struct net_device *dev);
3699 
3700 /* Functions used for unicast addresses handling */
3701 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3702 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3703 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3704 int dev_uc_sync(struct net_device *to, struct net_device *from);
3705 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3706 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3707 void dev_uc_flush(struct net_device *dev);
3708 void dev_uc_init(struct net_device *dev);
3709 
3710 /**
3711  *  __dev_uc_sync - Synchonize device's unicast list
3712  *  @dev:  device to sync
3713  *  @sync: function to call if address should be added
3714  *  @unsync: function to call if address should be removed
3715  *
3716  *  Add newly added addresses to the interface, and release
3717  *  addresses that have been deleted.
3718  */
3719 static inline int __dev_uc_sync(struct net_device *dev,
3720 				int (*sync)(struct net_device *,
3721 					    const unsigned char *),
3722 				int (*unsync)(struct net_device *,
3723 					      const unsigned char *))
3724 {
3725 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3726 }
3727 
3728 /**
3729  *  __dev_uc_unsync - Remove synchronized addresses from device
3730  *  @dev:  device to sync
3731  *  @unsync: function to call if address should be removed
3732  *
3733  *  Remove all addresses that were added to the device by dev_uc_sync().
3734  */
3735 static inline void __dev_uc_unsync(struct net_device *dev,
3736 				   int (*unsync)(struct net_device *,
3737 						 const unsigned char *))
3738 {
3739 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3740 }
3741 
3742 /* Functions used for multicast addresses handling */
3743 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3744 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3745 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3746 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3747 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3748 int dev_mc_sync(struct net_device *to, struct net_device *from);
3749 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3750 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3751 void dev_mc_flush(struct net_device *dev);
3752 void dev_mc_init(struct net_device *dev);
3753 
3754 /**
3755  *  __dev_mc_sync - Synchonize device's multicast list
3756  *  @dev:  device to sync
3757  *  @sync: function to call if address should be added
3758  *  @unsync: function to call if address should be removed
3759  *
3760  *  Add newly added addresses to the interface, and release
3761  *  addresses that have been deleted.
3762  */
3763 static inline int __dev_mc_sync(struct net_device *dev,
3764 				int (*sync)(struct net_device *,
3765 					    const unsigned char *),
3766 				int (*unsync)(struct net_device *,
3767 					      const unsigned char *))
3768 {
3769 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3770 }
3771 
3772 /**
3773  *  __dev_mc_unsync - Remove synchronized addresses from device
3774  *  @dev:  device to sync
3775  *  @unsync: function to call if address should be removed
3776  *
3777  *  Remove all addresses that were added to the device by dev_mc_sync().
3778  */
3779 static inline void __dev_mc_unsync(struct net_device *dev,
3780 				   int (*unsync)(struct net_device *,
3781 						 const unsigned char *))
3782 {
3783 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3784 }
3785 
3786 /* Functions used for secondary unicast and multicast support */
3787 void dev_set_rx_mode(struct net_device *dev);
3788 void __dev_set_rx_mode(struct net_device *dev);
3789 int dev_set_promiscuity(struct net_device *dev, int inc);
3790 int dev_set_allmulti(struct net_device *dev, int inc);
3791 void netdev_state_change(struct net_device *dev);
3792 void netdev_notify_peers(struct net_device *dev);
3793 void netdev_features_change(struct net_device *dev);
3794 /* Load a device via the kmod */
3795 void dev_load(struct net *net, const char *name);
3796 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3797 					struct rtnl_link_stats64 *storage);
3798 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3799 			     const struct net_device_stats *netdev_stats);
3800 
3801 extern int		netdev_max_backlog;
3802 extern int		netdev_tstamp_prequeue;
3803 extern int		weight_p;
3804 
3805 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3806 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3807 						     struct list_head **iter);
3808 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3809 						     struct list_head **iter);
3810 
3811 /* iterate through upper list, must be called under RCU read lock */
3812 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3813 	for (iter = &(dev)->adj_list.upper, \
3814 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3815 	     updev; \
3816 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3817 
3818 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3819 				  int (*fn)(struct net_device *upper_dev,
3820 					    void *data),
3821 				  void *data);
3822 
3823 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3824 				  struct net_device *upper_dev);
3825 
3826 void *netdev_lower_get_next_private(struct net_device *dev,
3827 				    struct list_head **iter);
3828 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3829 					struct list_head **iter);
3830 
3831 #define netdev_for_each_lower_private(dev, priv, iter) \
3832 	for (iter = (dev)->adj_list.lower.next, \
3833 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3834 	     priv; \
3835 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3836 
3837 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3838 	for (iter = &(dev)->adj_list.lower, \
3839 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3840 	     priv; \
3841 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3842 
3843 void *netdev_lower_get_next(struct net_device *dev,
3844 				struct list_head **iter);
3845 
3846 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3847 	for (iter = (dev)->adj_list.lower.next, \
3848 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3849 	     ldev; \
3850 	     ldev = netdev_lower_get_next(dev, &(iter)))
3851 
3852 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3853 					     struct list_head **iter);
3854 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3855 						 struct list_head **iter);
3856 
3857 int netdev_walk_all_lower_dev(struct net_device *dev,
3858 			      int (*fn)(struct net_device *lower_dev,
3859 					void *data),
3860 			      void *data);
3861 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3862 				  int (*fn)(struct net_device *lower_dev,
3863 					    void *data),
3864 				  void *data);
3865 
3866 void *netdev_adjacent_get_private(struct list_head *adj_list);
3867 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3868 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3869 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3870 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3871 int netdev_master_upper_dev_link(struct net_device *dev,
3872 				 struct net_device *upper_dev,
3873 				 void *upper_priv, void *upper_info);
3874 void netdev_upper_dev_unlink(struct net_device *dev,
3875 			     struct net_device *upper_dev);
3876 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3877 void *netdev_lower_dev_get_private(struct net_device *dev,
3878 				   struct net_device *lower_dev);
3879 void netdev_lower_state_changed(struct net_device *lower_dev,
3880 				void *lower_state_info);
3881 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3882 					   struct neighbour *n);
3883 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3884 					  struct neighbour *n);
3885 
3886 /* RSS keys are 40 or 52 bytes long */
3887 #define NETDEV_RSS_KEY_LEN 52
3888 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3889 void netdev_rss_key_fill(void *buffer, size_t len);
3890 
3891 int dev_get_nest_level(struct net_device *dev);
3892 int skb_checksum_help(struct sk_buff *skb);
3893 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3894 				  netdev_features_t features, bool tx_path);
3895 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3896 				    netdev_features_t features);
3897 
3898 struct netdev_bonding_info {
3899 	ifslave	slave;
3900 	ifbond	master;
3901 };
3902 
3903 struct netdev_notifier_bonding_info {
3904 	struct netdev_notifier_info info; /* must be first */
3905 	struct netdev_bonding_info  bonding_info;
3906 };
3907 
3908 void netdev_bonding_info_change(struct net_device *dev,
3909 				struct netdev_bonding_info *bonding_info);
3910 
3911 static inline
3912 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3913 {
3914 	return __skb_gso_segment(skb, features, true);
3915 }
3916 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3917 
3918 static inline bool can_checksum_protocol(netdev_features_t features,
3919 					 __be16 protocol)
3920 {
3921 	if (protocol == htons(ETH_P_FCOE))
3922 		return !!(features & NETIF_F_FCOE_CRC);
3923 
3924 	/* Assume this is an IP checksum (not SCTP CRC) */
3925 
3926 	if (features & NETIF_F_HW_CSUM) {
3927 		/* Can checksum everything */
3928 		return true;
3929 	}
3930 
3931 	switch (protocol) {
3932 	case htons(ETH_P_IP):
3933 		return !!(features & NETIF_F_IP_CSUM);
3934 	case htons(ETH_P_IPV6):
3935 		return !!(features & NETIF_F_IPV6_CSUM);
3936 	default:
3937 		return false;
3938 	}
3939 }
3940 
3941 #ifdef CONFIG_BUG
3942 void netdev_rx_csum_fault(struct net_device *dev);
3943 #else
3944 static inline void netdev_rx_csum_fault(struct net_device *dev)
3945 {
3946 }
3947 #endif
3948 /* rx skb timestamps */
3949 void net_enable_timestamp(void);
3950 void net_disable_timestamp(void);
3951 
3952 #ifdef CONFIG_PROC_FS
3953 int __init dev_proc_init(void);
3954 #else
3955 #define dev_proc_init() 0
3956 #endif
3957 
3958 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3959 					      struct sk_buff *skb, struct net_device *dev,
3960 					      bool more)
3961 {
3962 	skb->xmit_more = more ? 1 : 0;
3963 	return ops->ndo_start_xmit(skb, dev);
3964 }
3965 
3966 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3967 					    struct netdev_queue *txq, bool more)
3968 {
3969 	const struct net_device_ops *ops = dev->netdev_ops;
3970 	int rc;
3971 
3972 	rc = __netdev_start_xmit(ops, skb, dev, more);
3973 	if (rc == NETDEV_TX_OK)
3974 		txq_trans_update(txq);
3975 
3976 	return rc;
3977 }
3978 
3979 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3980 				const void *ns);
3981 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3982 				 const void *ns);
3983 
3984 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3985 {
3986 	return netdev_class_create_file_ns(class_attr, NULL);
3987 }
3988 
3989 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3990 {
3991 	netdev_class_remove_file_ns(class_attr, NULL);
3992 }
3993 
3994 extern struct kobj_ns_type_operations net_ns_type_operations;
3995 
3996 const char *netdev_drivername(const struct net_device *dev);
3997 
3998 void linkwatch_run_queue(void);
3999 
4000 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4001 							  netdev_features_t f2)
4002 {
4003 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4004 		if (f1 & NETIF_F_HW_CSUM)
4005 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4006 		else
4007 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4008 	}
4009 
4010 	return f1 & f2;
4011 }
4012 
4013 static inline netdev_features_t netdev_get_wanted_features(
4014 	struct net_device *dev)
4015 {
4016 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4017 }
4018 netdev_features_t netdev_increment_features(netdev_features_t all,
4019 	netdev_features_t one, netdev_features_t mask);
4020 
4021 /* Allow TSO being used on stacked device :
4022  * Performing the GSO segmentation before last device
4023  * is a performance improvement.
4024  */
4025 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4026 							netdev_features_t mask)
4027 {
4028 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4029 }
4030 
4031 int __netdev_update_features(struct net_device *dev);
4032 void netdev_update_features(struct net_device *dev);
4033 void netdev_change_features(struct net_device *dev);
4034 
4035 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4036 					struct net_device *dev);
4037 
4038 netdev_features_t passthru_features_check(struct sk_buff *skb,
4039 					  struct net_device *dev,
4040 					  netdev_features_t features);
4041 netdev_features_t netif_skb_features(struct sk_buff *skb);
4042 
4043 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4044 {
4045 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4046 
4047 	/* check flags correspondence */
4048 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4049 	BUILD_BUG_ON(SKB_GSO_UDP     != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4050 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4051 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4052 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4053 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4054 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4055 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4056 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4057 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4058 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4059 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4060 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4061 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4062 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4063 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4064 
4065 	return (features & feature) == feature;
4066 }
4067 
4068 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4069 {
4070 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4071 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4072 }
4073 
4074 static inline bool netif_needs_gso(struct sk_buff *skb,
4075 				   netdev_features_t features)
4076 {
4077 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4078 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4079 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4080 }
4081 
4082 static inline void netif_set_gso_max_size(struct net_device *dev,
4083 					  unsigned int size)
4084 {
4085 	dev->gso_max_size = size;
4086 }
4087 
4088 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4089 					int pulled_hlen, u16 mac_offset,
4090 					int mac_len)
4091 {
4092 	skb->protocol = protocol;
4093 	skb->encapsulation = 1;
4094 	skb_push(skb, pulled_hlen);
4095 	skb_reset_transport_header(skb);
4096 	skb->mac_header = mac_offset;
4097 	skb->network_header = skb->mac_header + mac_len;
4098 	skb->mac_len = mac_len;
4099 }
4100 
4101 static inline bool netif_is_macsec(const struct net_device *dev)
4102 {
4103 	return dev->priv_flags & IFF_MACSEC;
4104 }
4105 
4106 static inline bool netif_is_macvlan(const struct net_device *dev)
4107 {
4108 	return dev->priv_flags & IFF_MACVLAN;
4109 }
4110 
4111 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4112 {
4113 	return dev->priv_flags & IFF_MACVLAN_PORT;
4114 }
4115 
4116 static inline bool netif_is_ipvlan(const struct net_device *dev)
4117 {
4118 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4119 }
4120 
4121 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4122 {
4123 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4124 }
4125 
4126 static inline bool netif_is_bond_master(const struct net_device *dev)
4127 {
4128 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4129 }
4130 
4131 static inline bool netif_is_bond_slave(const struct net_device *dev)
4132 {
4133 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4134 }
4135 
4136 static inline bool netif_supports_nofcs(struct net_device *dev)
4137 {
4138 	return dev->priv_flags & IFF_SUPP_NOFCS;
4139 }
4140 
4141 static inline bool netif_is_l3_master(const struct net_device *dev)
4142 {
4143 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4144 }
4145 
4146 static inline bool netif_is_l3_slave(const struct net_device *dev)
4147 {
4148 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4149 }
4150 
4151 static inline bool netif_is_bridge_master(const struct net_device *dev)
4152 {
4153 	return dev->priv_flags & IFF_EBRIDGE;
4154 }
4155 
4156 static inline bool netif_is_bridge_port(const struct net_device *dev)
4157 {
4158 	return dev->priv_flags & IFF_BRIDGE_PORT;
4159 }
4160 
4161 static inline bool netif_is_ovs_master(const struct net_device *dev)
4162 {
4163 	return dev->priv_flags & IFF_OPENVSWITCH;
4164 }
4165 
4166 static inline bool netif_is_team_master(const struct net_device *dev)
4167 {
4168 	return dev->priv_flags & IFF_TEAM;
4169 }
4170 
4171 static inline bool netif_is_team_port(const struct net_device *dev)
4172 {
4173 	return dev->priv_flags & IFF_TEAM_PORT;
4174 }
4175 
4176 static inline bool netif_is_lag_master(const struct net_device *dev)
4177 {
4178 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4179 }
4180 
4181 static inline bool netif_is_lag_port(const struct net_device *dev)
4182 {
4183 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4184 }
4185 
4186 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4187 {
4188 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4189 }
4190 
4191 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4192 static inline void netif_keep_dst(struct net_device *dev)
4193 {
4194 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4195 }
4196 
4197 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4198 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4199 {
4200 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4201 	return dev->priv_flags & IFF_MACSEC;
4202 }
4203 
4204 extern struct pernet_operations __net_initdata loopback_net_ops;
4205 
4206 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4207 
4208 /* netdev_printk helpers, similar to dev_printk */
4209 
4210 static inline const char *netdev_name(const struct net_device *dev)
4211 {
4212 	if (!dev->name[0] || strchr(dev->name, '%'))
4213 		return "(unnamed net_device)";
4214 	return dev->name;
4215 }
4216 
4217 static inline const char *netdev_reg_state(const struct net_device *dev)
4218 {
4219 	switch (dev->reg_state) {
4220 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4221 	case NETREG_REGISTERED: return "";
4222 	case NETREG_UNREGISTERING: return " (unregistering)";
4223 	case NETREG_UNREGISTERED: return " (unregistered)";
4224 	case NETREG_RELEASED: return " (released)";
4225 	case NETREG_DUMMY: return " (dummy)";
4226 	}
4227 
4228 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4229 	return " (unknown)";
4230 }
4231 
4232 __printf(3, 4)
4233 void netdev_printk(const char *level, const struct net_device *dev,
4234 		   const char *format, ...);
4235 __printf(2, 3)
4236 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4237 __printf(2, 3)
4238 void netdev_alert(const struct net_device *dev, const char *format, ...);
4239 __printf(2, 3)
4240 void netdev_crit(const struct net_device *dev, const char *format, ...);
4241 __printf(2, 3)
4242 void netdev_err(const struct net_device *dev, const char *format, ...);
4243 __printf(2, 3)
4244 void netdev_warn(const struct net_device *dev, const char *format, ...);
4245 __printf(2, 3)
4246 void netdev_notice(const struct net_device *dev, const char *format, ...);
4247 __printf(2, 3)
4248 void netdev_info(const struct net_device *dev, const char *format, ...);
4249 
4250 #define MODULE_ALIAS_NETDEV(device) \
4251 	MODULE_ALIAS("netdev-" device)
4252 
4253 #if defined(CONFIG_DYNAMIC_DEBUG)
4254 #define netdev_dbg(__dev, format, args...)			\
4255 do {								\
4256 	dynamic_netdev_dbg(__dev, format, ##args);		\
4257 } while (0)
4258 #elif defined(DEBUG)
4259 #define netdev_dbg(__dev, format, args...)			\
4260 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4261 #else
4262 #define netdev_dbg(__dev, format, args...)			\
4263 ({								\
4264 	if (0)							\
4265 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4266 })
4267 #endif
4268 
4269 #if defined(VERBOSE_DEBUG)
4270 #define netdev_vdbg	netdev_dbg
4271 #else
4272 
4273 #define netdev_vdbg(dev, format, args...)			\
4274 ({								\
4275 	if (0)							\
4276 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4277 	0;							\
4278 })
4279 #endif
4280 
4281 /*
4282  * netdev_WARN() acts like dev_printk(), but with the key difference
4283  * of using a WARN/WARN_ON to get the message out, including the
4284  * file/line information and a backtrace.
4285  */
4286 #define netdev_WARN(dev, format, args...)			\
4287 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4288 	     netdev_reg_state(dev), ##args)
4289 
4290 /* netif printk helpers, similar to netdev_printk */
4291 
4292 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4293 do {					  			\
4294 	if (netif_msg_##type(priv))				\
4295 		netdev_printk(level, (dev), fmt, ##args);	\
4296 } while (0)
4297 
4298 #define netif_level(level, priv, type, dev, fmt, args...)	\
4299 do {								\
4300 	if (netif_msg_##type(priv))				\
4301 		netdev_##level(dev, fmt, ##args);		\
4302 } while (0)
4303 
4304 #define netif_emerg(priv, type, dev, fmt, args...)		\
4305 	netif_level(emerg, priv, type, dev, fmt, ##args)
4306 #define netif_alert(priv, type, dev, fmt, args...)		\
4307 	netif_level(alert, priv, type, dev, fmt, ##args)
4308 #define netif_crit(priv, type, dev, fmt, args...)		\
4309 	netif_level(crit, priv, type, dev, fmt, ##args)
4310 #define netif_err(priv, type, dev, fmt, args...)		\
4311 	netif_level(err, priv, type, dev, fmt, ##args)
4312 #define netif_warn(priv, type, dev, fmt, args...)		\
4313 	netif_level(warn, priv, type, dev, fmt, ##args)
4314 #define netif_notice(priv, type, dev, fmt, args...)		\
4315 	netif_level(notice, priv, type, dev, fmt, ##args)
4316 #define netif_info(priv, type, dev, fmt, args...)		\
4317 	netif_level(info, priv, type, dev, fmt, ##args)
4318 
4319 #if defined(CONFIG_DYNAMIC_DEBUG)
4320 #define netif_dbg(priv, type, netdev, format, args...)		\
4321 do {								\
4322 	if (netif_msg_##type(priv))				\
4323 		dynamic_netdev_dbg(netdev, format, ##args);	\
4324 } while (0)
4325 #elif defined(DEBUG)
4326 #define netif_dbg(priv, type, dev, format, args...)		\
4327 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4328 #else
4329 #define netif_dbg(priv, type, dev, format, args...)			\
4330 ({									\
4331 	if (0)								\
4332 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4333 	0;								\
4334 })
4335 #endif
4336 
4337 #if defined(VERBOSE_DEBUG)
4338 #define netif_vdbg	netif_dbg
4339 #else
4340 #define netif_vdbg(priv, type, dev, format, args...)		\
4341 ({								\
4342 	if (0)							\
4343 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4344 	0;							\
4345 })
4346 #endif
4347 
4348 /*
4349  *	The list of packet types we will receive (as opposed to discard)
4350  *	and the routines to invoke.
4351  *
4352  *	Why 16. Because with 16 the only overlap we get on a hash of the
4353  *	low nibble of the protocol value is RARP/SNAP/X.25.
4354  *
4355  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4356  *             sure which should go first, but I bet it won't make much
4357  *             difference if we are running VLANs.  The good news is that
4358  *             this protocol won't be in the list unless compiled in, so
4359  *             the average user (w/out VLANs) will not be adversely affected.
4360  *             --BLG
4361  *
4362  *		0800	IP
4363  *		8100    802.1Q VLAN
4364  *		0001	802.3
4365  *		0002	AX.25
4366  *		0004	802.2
4367  *		8035	RARP
4368  *		0005	SNAP
4369  *		0805	X.25
4370  *		0806	ARP
4371  *		8137	IPX
4372  *		0009	Localtalk
4373  *		86DD	IPv6
4374  */
4375 #define PTYPE_HASH_SIZE	(16)
4376 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4377 
4378 #endif	/* _LINUX_NETDEVICE_H */
4379