xref: /linux-6.15/include/linux/netdevice.h (revision 1ce84604)
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  * void (*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, u32 handle,
972  *		       __be16 protocol, struct tc_to_netdev *tc);
973  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
974  *	This is always called from the stack with the rtnl lock held and netif
975  *	tx queues stopped. This allows the netdevice to perform queue
976  *	management safely.
977  *
978  *	Fiber Channel over Ethernet (FCoE) offload functions.
979  * int (*ndo_fcoe_enable)(struct net_device *dev);
980  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
981  *	so the underlying device can perform whatever needed configuration or
982  *	initialization to support acceleration of FCoE traffic.
983  *
984  * int (*ndo_fcoe_disable)(struct net_device *dev);
985  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
986  *	so the underlying device can perform whatever needed clean-ups to
987  *	stop supporting acceleration of FCoE traffic.
988  *
989  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
990  *			     struct scatterlist *sgl, unsigned int sgc);
991  *	Called when the FCoE Initiator wants to initialize an I/O that
992  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
993  *	perform necessary setup and returns 1 to indicate the device is set up
994  *	successfully to perform DDP on this I/O, otherwise this returns 0.
995  *
996  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
997  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
998  *	indicated by the FC exchange id 'xid', so the underlying device can
999  *	clean up and reuse resources for later DDP requests.
1000  *
1001  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1002  *			      struct scatterlist *sgl, unsigned int sgc);
1003  *	Called when the FCoE Target wants to initialize an I/O that
1004  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1005  *	perform necessary setup and returns 1 to indicate the device is set up
1006  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1007  *
1008  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1009  *			       struct netdev_fcoe_hbainfo *hbainfo);
1010  *	Called when the FCoE Protocol stack wants information on the underlying
1011  *	device. This information is utilized by the FCoE protocol stack to
1012  *	register attributes with Fiber Channel management service as per the
1013  *	FC-GS Fabric Device Management Information(FDMI) specification.
1014  *
1015  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1016  *	Called when the underlying device wants to override default World Wide
1017  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1018  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1019  *	protocol stack to use.
1020  *
1021  *	RFS acceleration.
1022  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1023  *			    u16 rxq_index, u32 flow_id);
1024  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1025  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1026  *	Return the filter ID on success, or a negative error code.
1027  *
1028  *	Slave management functions (for bridge, bonding, etc).
1029  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1030  *	Called to make another netdev an underling.
1031  *
1032  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1033  *	Called to release previously enslaved netdev.
1034  *
1035  *      Feature/offload setting functions.
1036  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1037  *		netdev_features_t features);
1038  *	Adjusts the requested feature flags according to device-specific
1039  *	constraints, and returns the resulting flags. Must not modify
1040  *	the device state.
1041  *
1042  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1043  *	Called to update device configuration to new features. Passed
1044  *	feature set might be less than what was returned by ndo_fix_features()).
1045  *	Must return >0 or -errno if it changed dev->features itself.
1046  *
1047  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1048  *		      struct net_device *dev,
1049  *		      const unsigned char *addr, u16 vid, u16 flags)
1050  *	Adds an FDB entry to dev for addr.
1051  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1052  *		      struct net_device *dev,
1053  *		      const unsigned char *addr, u16 vid)
1054  *	Deletes the FDB entry from dev coresponding to addr.
1055  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1056  *		       struct net_device *dev, struct net_device *filter_dev,
1057  *		       int *idx)
1058  *	Used to add FDB entries to dump requests. Implementers should add
1059  *	entries to skb and update idx with the number of entries.
1060  *
1061  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1062  *			     u16 flags)
1063  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1064  *			     struct net_device *dev, u32 filter_mask,
1065  *			     int nlflags)
1066  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1067  *			     u16 flags);
1068  *
1069  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1070  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1071  *	which do not represent real hardware may define this to allow their
1072  *	userspace components to manage their virtual carrier state. Devices
1073  *	that determine carrier state from physical hardware properties (eg
1074  *	network cables) or protocol-dependent mechanisms (eg
1075  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1076  *
1077  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1078  *			       struct netdev_phys_item_id *ppid);
1079  *	Called to get ID of physical port of this device. If driver does
1080  *	not implement this, it is assumed that the hw is not able to have
1081  *	multiple net devices on single physical port.
1082  *
1083  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1084  *			      struct udp_tunnel_info *ti);
1085  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1086  *	address family that a UDP tunnel is listnening to. It is called only
1087  *	when a new port starts listening. The operation is protected by the
1088  *	RTNL.
1089  *
1090  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1091  *			      struct udp_tunnel_info *ti);
1092  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1093  *	address family that the UDP tunnel is not listening to anymore. The
1094  *	operation is protected by the RTNL.
1095  *
1096  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1097  *				 struct net_device *dev)
1098  *	Called by upper layer devices to accelerate switching or other
1099  *	station functionality into hardware. 'pdev is the lowerdev
1100  *	to use for the offload and 'dev' is the net device that will
1101  *	back the offload. Returns a pointer to the private structure
1102  *	the upper layer will maintain.
1103  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1104  *	Called by upper layer device to delete the station created
1105  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1106  *	the station and priv is the structure returned by the add
1107  *	operation.
1108  * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1109  *				      struct net_device *dev,
1110  *				      void *priv);
1111  *	Callback to use for xmit over the accelerated station. This
1112  *	is used in place of ndo_start_xmit on accelerated net
1113  *	devices.
1114  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1115  *			     int queue_index, u32 maxrate);
1116  *	Called when a user wants to set a max-rate limitation of specific
1117  *	TX queue.
1118  * int (*ndo_get_iflink)(const struct net_device *dev);
1119  *	Called to get the iflink value of this device.
1120  * void (*ndo_change_proto_down)(struct net_device *dev,
1121  *				 bool proto_down);
1122  *	This function is used to pass protocol port error state information
1123  *	to the switch driver. The switch driver can react to the proto_down
1124  *      by doing a phys down on the associated switch port.
1125  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1126  *	This function is used to get egress tunnel information for given skb.
1127  *	This is useful for retrieving outer tunnel header parameters while
1128  *	sampling packet.
1129  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1130  *	This function is used to specify the headroom that the skb must
1131  *	consider when allocation skb during packet reception. Setting
1132  *	appropriate rx headroom value allows avoiding skb head copy on
1133  *	forward. Setting a negative value resets the rx headroom to the
1134  *	default value.
1135  * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1136  *	This function is used to set or query state related to XDP on the
1137  *	netdevice. See definition of enum xdp_netdev_command for details.
1138  *
1139  */
1140 struct net_device_ops {
1141 	int			(*ndo_init)(struct net_device *dev);
1142 	void			(*ndo_uninit)(struct net_device *dev);
1143 	int			(*ndo_open)(struct net_device *dev);
1144 	int			(*ndo_stop)(struct net_device *dev);
1145 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1146 						  struct net_device *dev);
1147 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1148 						      struct net_device *dev,
1149 						      netdev_features_t features);
1150 	u16			(*ndo_select_queue)(struct net_device *dev,
1151 						    struct sk_buff *skb,
1152 						    void *accel_priv,
1153 						    select_queue_fallback_t fallback);
1154 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1155 						       int flags);
1156 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1157 	int			(*ndo_set_mac_address)(struct net_device *dev,
1158 						       void *addr);
1159 	int			(*ndo_validate_addr)(struct net_device *dev);
1160 	int			(*ndo_do_ioctl)(struct net_device *dev,
1161 					        struct ifreq *ifr, int cmd);
1162 	int			(*ndo_set_config)(struct net_device *dev,
1163 					          struct ifmap *map);
1164 	int			(*ndo_change_mtu)(struct net_device *dev,
1165 						  int new_mtu);
1166 	int			(*ndo_neigh_setup)(struct net_device *dev,
1167 						   struct neigh_parms *);
1168 	void			(*ndo_tx_timeout) (struct net_device *dev);
1169 
1170 	void			(*ndo_get_stats64)(struct net_device *dev,
1171 						   struct rtnl_link_stats64 *storage);
1172 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1173 	int			(*ndo_get_offload_stats)(int attr_id,
1174 							 const struct net_device *dev,
1175 							 void *attr_data);
1176 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1177 
1178 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1179 						       __be16 proto, u16 vid);
1180 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1181 						        __be16 proto, u16 vid);
1182 #ifdef CONFIG_NET_POLL_CONTROLLER
1183 	void                    (*ndo_poll_controller)(struct net_device *dev);
1184 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1185 						     struct netpoll_info *info);
1186 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1187 #endif
1188 #ifdef CONFIG_NET_RX_BUSY_POLL
1189 	int			(*ndo_busy_poll)(struct napi_struct *dev);
1190 #endif
1191 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1192 						  int queue, u8 *mac);
1193 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1194 						   int queue, u16 vlan,
1195 						   u8 qos, __be16 proto);
1196 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1197 						   int vf, int min_tx_rate,
1198 						   int max_tx_rate);
1199 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1200 						       int vf, bool setting);
1201 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1202 						    int vf, bool setting);
1203 	int			(*ndo_get_vf_config)(struct net_device *dev,
1204 						     int vf,
1205 						     struct ifla_vf_info *ivf);
1206 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1207 							 int vf, int link_state);
1208 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1209 						    int vf,
1210 						    struct ifla_vf_stats
1211 						    *vf_stats);
1212 	int			(*ndo_set_vf_port)(struct net_device *dev,
1213 						   int vf,
1214 						   struct nlattr *port[]);
1215 	int			(*ndo_get_vf_port)(struct net_device *dev,
1216 						   int vf, struct sk_buff *skb);
1217 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1218 						   int vf, u64 guid,
1219 						   int guid_type);
1220 	int			(*ndo_set_vf_rss_query_en)(
1221 						   struct net_device *dev,
1222 						   int vf, bool setting);
1223 	int			(*ndo_setup_tc)(struct net_device *dev,
1224 						u32 handle,
1225 						__be16 protocol,
1226 						struct tc_to_netdev *tc);
1227 #if IS_ENABLED(CONFIG_FCOE)
1228 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1229 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1230 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1231 						      u16 xid,
1232 						      struct scatterlist *sgl,
1233 						      unsigned int sgc);
1234 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1235 						     u16 xid);
1236 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1237 						       u16 xid,
1238 						       struct scatterlist *sgl,
1239 						       unsigned int sgc);
1240 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1241 							struct netdev_fcoe_hbainfo *hbainfo);
1242 #endif
1243 
1244 #if IS_ENABLED(CONFIG_LIBFCOE)
1245 #define NETDEV_FCOE_WWNN 0
1246 #define NETDEV_FCOE_WWPN 1
1247 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1248 						    u64 *wwn, int type);
1249 #endif
1250 
1251 #ifdef CONFIG_RFS_ACCEL
1252 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1253 						     const struct sk_buff *skb,
1254 						     u16 rxq_index,
1255 						     u32 flow_id);
1256 #endif
1257 	int			(*ndo_add_slave)(struct net_device *dev,
1258 						 struct net_device *slave_dev);
1259 	int			(*ndo_del_slave)(struct net_device *dev,
1260 						 struct net_device *slave_dev);
1261 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1262 						    netdev_features_t features);
1263 	int			(*ndo_set_features)(struct net_device *dev,
1264 						    netdev_features_t features);
1265 	int			(*ndo_neigh_construct)(struct net_device *dev,
1266 						       struct neighbour *n);
1267 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1268 						     struct neighbour *n);
1269 
1270 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1271 					       struct nlattr *tb[],
1272 					       struct net_device *dev,
1273 					       const unsigned char *addr,
1274 					       u16 vid,
1275 					       u16 flags);
1276 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1277 					       struct nlattr *tb[],
1278 					       struct net_device *dev,
1279 					       const unsigned char *addr,
1280 					       u16 vid);
1281 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1282 						struct netlink_callback *cb,
1283 						struct net_device *dev,
1284 						struct net_device *filter_dev,
1285 						int *idx);
1286 
1287 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1288 						      struct nlmsghdr *nlh,
1289 						      u16 flags);
1290 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1291 						      u32 pid, u32 seq,
1292 						      struct net_device *dev,
1293 						      u32 filter_mask,
1294 						      int nlflags);
1295 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1296 						      struct nlmsghdr *nlh,
1297 						      u16 flags);
1298 	int			(*ndo_change_carrier)(struct net_device *dev,
1299 						      bool new_carrier);
1300 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1301 							struct netdev_phys_item_id *ppid);
1302 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1303 							  char *name, size_t len);
1304 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1305 						      struct udp_tunnel_info *ti);
1306 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1307 						      struct udp_tunnel_info *ti);
1308 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1309 							struct net_device *dev);
1310 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1311 							void *priv);
1312 
1313 	netdev_tx_t		(*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1314 							struct net_device *dev,
1315 							void *priv);
1316 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1317 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1318 						      int queue_index,
1319 						      u32 maxrate);
1320 	int			(*ndo_get_iflink)(const struct net_device *dev);
1321 	int			(*ndo_change_proto_down)(struct net_device *dev,
1322 							 bool proto_down);
1323 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1324 						       struct sk_buff *skb);
1325 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1326 						       int needed_headroom);
1327 	int			(*ndo_xdp)(struct net_device *dev,
1328 					   struct netdev_xdp *xdp);
1329 };
1330 
1331 /**
1332  * enum net_device_priv_flags - &struct net_device priv_flags
1333  *
1334  * These are the &struct net_device, they are only set internally
1335  * by drivers and used in the kernel. These flags are invisible to
1336  * userspace; this means that the order of these flags can change
1337  * during any kernel release.
1338  *
1339  * You should have a pretty good reason to be extending these flags.
1340  *
1341  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1342  * @IFF_EBRIDGE: Ethernet bridging device
1343  * @IFF_BONDING: bonding master or slave
1344  * @IFF_ISATAP: ISATAP interface (RFC4214)
1345  * @IFF_WAN_HDLC: WAN HDLC device
1346  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1347  *	release skb->dst
1348  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1349  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1350  * @IFF_MACVLAN_PORT: device used as macvlan port
1351  * @IFF_BRIDGE_PORT: device used as bridge port
1352  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1353  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1354  * @IFF_UNICAST_FLT: Supports unicast filtering
1355  * @IFF_TEAM_PORT: device used as team port
1356  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1357  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1358  *	change when it's running
1359  * @IFF_MACVLAN: Macvlan device
1360  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1361  *	underlying stacked devices
1362  * @IFF_IPVLAN_MASTER: IPvlan master device
1363  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1364  * @IFF_L3MDEV_MASTER: device is an L3 master device
1365  * @IFF_NO_QUEUE: device can run without qdisc attached
1366  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1367  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1368  * @IFF_TEAM: device is a team device
1369  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1370  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1371  *	entity (i.e. the master device for bridged veth)
1372  * @IFF_MACSEC: device is a MACsec device
1373  */
1374 enum netdev_priv_flags {
1375 	IFF_802_1Q_VLAN			= 1<<0,
1376 	IFF_EBRIDGE			= 1<<1,
1377 	IFF_BONDING			= 1<<2,
1378 	IFF_ISATAP			= 1<<3,
1379 	IFF_WAN_HDLC			= 1<<4,
1380 	IFF_XMIT_DST_RELEASE		= 1<<5,
1381 	IFF_DONT_BRIDGE			= 1<<6,
1382 	IFF_DISABLE_NETPOLL		= 1<<7,
1383 	IFF_MACVLAN_PORT		= 1<<8,
1384 	IFF_BRIDGE_PORT			= 1<<9,
1385 	IFF_OVS_DATAPATH		= 1<<10,
1386 	IFF_TX_SKB_SHARING		= 1<<11,
1387 	IFF_UNICAST_FLT			= 1<<12,
1388 	IFF_TEAM_PORT			= 1<<13,
1389 	IFF_SUPP_NOFCS			= 1<<14,
1390 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1391 	IFF_MACVLAN			= 1<<16,
1392 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1393 	IFF_IPVLAN_MASTER		= 1<<18,
1394 	IFF_IPVLAN_SLAVE		= 1<<19,
1395 	IFF_L3MDEV_MASTER		= 1<<20,
1396 	IFF_NO_QUEUE			= 1<<21,
1397 	IFF_OPENVSWITCH			= 1<<22,
1398 	IFF_L3MDEV_SLAVE		= 1<<23,
1399 	IFF_TEAM			= 1<<24,
1400 	IFF_RXFH_CONFIGURED		= 1<<25,
1401 	IFF_PHONY_HEADROOM		= 1<<26,
1402 	IFF_MACSEC			= 1<<27,
1403 };
1404 
1405 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1406 #define IFF_EBRIDGE			IFF_EBRIDGE
1407 #define IFF_BONDING			IFF_BONDING
1408 #define IFF_ISATAP			IFF_ISATAP
1409 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1410 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1411 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1412 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1413 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1414 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1415 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1416 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1417 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1418 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1419 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1420 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1421 #define IFF_MACVLAN			IFF_MACVLAN
1422 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1423 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1424 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1425 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1426 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1427 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1428 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1429 #define IFF_TEAM			IFF_TEAM
1430 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1431 #define IFF_MACSEC			IFF_MACSEC
1432 
1433 /**
1434  *	struct net_device - The DEVICE structure.
1435  *		Actually, this whole structure is a big mistake.  It mixes I/O
1436  *		data with strictly "high-level" data, and it has to know about
1437  *		almost every data structure used in the INET module.
1438  *
1439  *	@name:	This is the first field of the "visible" part of this structure
1440  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1441  *	 	of the interface.
1442  *
1443  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1444  *	@ifalias:	SNMP alias
1445  *	@mem_end:	Shared memory end
1446  *	@mem_start:	Shared memory start
1447  *	@base_addr:	Device I/O address
1448  *	@irq:		Device IRQ number
1449  *
1450  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1451  *
1452  *	@state:		Generic network queuing layer state, see netdev_state_t
1453  *	@dev_list:	The global list of network devices
1454  *	@napi_list:	List entry used for polling NAPI devices
1455  *	@unreg_list:	List entry  when we are unregistering the
1456  *			device; see the function unregister_netdev
1457  *	@close_list:	List entry used when we are closing the device
1458  *	@ptype_all:     Device-specific packet handlers for all protocols
1459  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1460  *
1461  *	@adj_list:	Directly linked devices, like slaves for bonding
1462  *	@features:	Currently active device features
1463  *	@hw_features:	User-changeable features
1464  *
1465  *	@wanted_features:	User-requested features
1466  *	@vlan_features:		Mask of features inheritable by VLAN devices
1467  *
1468  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1469  *				This field indicates what encapsulation
1470  *				offloads the hardware is capable of doing,
1471  *				and drivers will need to set them appropriately.
1472  *
1473  *	@mpls_features:	Mask of features inheritable by MPLS
1474  *
1475  *	@ifindex:	interface index
1476  *	@group:		The group the device belongs to
1477  *
1478  *	@stats:		Statistics struct, which was left as a legacy, use
1479  *			rtnl_link_stats64 instead
1480  *
1481  *	@rx_dropped:	Dropped packets by core network,
1482  *			do not use this in drivers
1483  *	@tx_dropped:	Dropped packets by core network,
1484  *			do not use this in drivers
1485  *	@rx_nohandler:	nohandler dropped packets by core network on
1486  *			inactive devices, do not use this in drivers
1487  *
1488  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1489  *				instead of ioctl,
1490  *				see <net/iw_handler.h> for details.
1491  *	@wireless_data:	Instance data managed by the core of wireless extensions
1492  *
1493  *	@netdev_ops:	Includes several pointers to callbacks,
1494  *			if one wants to override the ndo_*() functions
1495  *	@ethtool_ops:	Management operations
1496  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1497  *			discovery handling. Necessary for e.g. 6LoWPAN.
1498  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1499  *			of Layer 2 headers.
1500  *
1501  *	@flags:		Interface flags (a la BSD)
1502  *	@priv_flags:	Like 'flags' but invisible to userspace,
1503  *			see if.h for the definitions
1504  *	@gflags:	Global flags ( kept as legacy )
1505  *	@padded:	How much padding added by alloc_netdev()
1506  *	@operstate:	RFC2863 operstate
1507  *	@link_mode:	Mapping policy to operstate
1508  *	@if_port:	Selectable AUI, TP, ...
1509  *	@dma:		DMA channel
1510  *	@mtu:		Interface MTU value
1511  *	@min_mtu:	Interface Minimum MTU value
1512  *	@max_mtu:	Interface Maximum MTU value
1513  *	@type:		Interface hardware type
1514  *	@hard_header_len: Maximum hardware header length.
1515  *
1516  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1517  *			  cases can this be guaranteed
1518  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1519  *			  cases can this be guaranteed. Some cases also use
1520  *			  LL_MAX_HEADER instead to allocate the skb
1521  *
1522  *	interface address info:
1523  *
1524  * 	@perm_addr:		Permanent hw address
1525  * 	@addr_assign_type:	Hw address assignment type
1526  * 	@addr_len:		Hardware address length
1527  *	@neigh_priv_len:	Used in neigh_alloc()
1528  * 	@dev_id:		Used to differentiate devices that share
1529  * 				the same link layer address
1530  * 	@dev_port:		Used to differentiate devices that share
1531  * 				the same function
1532  *	@addr_list_lock:	XXX: need comments on this one
1533  *	@uc_promisc:		Counter that indicates promiscuous mode
1534  *				has been enabled due to the need to listen to
1535  *				additional unicast addresses in a device that
1536  *				does not implement ndo_set_rx_mode()
1537  *	@uc:			unicast mac addresses
1538  *	@mc:			multicast mac addresses
1539  *	@dev_addrs:		list of device hw addresses
1540  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1541  *	@promiscuity:		Number of times the NIC is told to work in
1542  *				promiscuous mode; if it becomes 0 the NIC will
1543  *				exit promiscuous mode
1544  *	@allmulti:		Counter, enables or disables allmulticast mode
1545  *
1546  *	@vlan_info:	VLAN info
1547  *	@dsa_ptr:	dsa specific data
1548  *	@tipc_ptr:	TIPC specific data
1549  *	@atalk_ptr:	AppleTalk link
1550  *	@ip_ptr:	IPv4 specific data
1551  *	@dn_ptr:	DECnet specific data
1552  *	@ip6_ptr:	IPv6 specific data
1553  *	@ax25_ptr:	AX.25 specific data
1554  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1555  *
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 	/* Interface address info used in eth_type_trans() */
1782 	unsigned char		*dev_addr;
1783 
1784 #ifdef CONFIG_SYSFS
1785 	struct netdev_rx_queue	*_rx;
1786 
1787 	unsigned int		num_rx_queues;
1788 	unsigned int		real_num_rx_queues;
1789 #endif
1790 
1791 	unsigned long		gro_flush_timeout;
1792 	rx_handler_func_t __rcu	*rx_handler;
1793 	void __rcu		*rx_handler_data;
1794 
1795 #ifdef CONFIG_NET_CLS_ACT
1796 	struct tcf_proto __rcu  *ingress_cl_list;
1797 #endif
1798 	struct netdev_queue __rcu *ingress_queue;
1799 #ifdef CONFIG_NETFILTER_INGRESS
1800 	struct nf_hook_entry __rcu *nf_hooks_ingress;
1801 #endif
1802 
1803 	unsigned char		broadcast[MAX_ADDR_LEN];
1804 #ifdef CONFIG_RFS_ACCEL
1805 	struct cpu_rmap		*rx_cpu_rmap;
1806 #endif
1807 	struct hlist_node	index_hlist;
1808 
1809 /*
1810  * Cache lines mostly used on transmit path
1811  */
1812 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1813 	unsigned int		num_tx_queues;
1814 	unsigned int		real_num_tx_queues;
1815 	struct Qdisc		*qdisc;
1816 #ifdef CONFIG_NET_SCHED
1817 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1818 #endif
1819 	unsigned long		tx_queue_len;
1820 	spinlock_t		tx_global_lock;
1821 	int			watchdog_timeo;
1822 
1823 #ifdef CONFIG_XPS
1824 	struct xps_dev_maps __rcu *xps_maps;
1825 #endif
1826 #ifdef CONFIG_NET_CLS_ACT
1827 	struct tcf_proto __rcu  *egress_cl_list;
1828 #endif
1829 
1830 	/* These may be needed for future network-power-down code. */
1831 	struct timer_list	watchdog_timer;
1832 
1833 	int __percpu		*pcpu_refcnt;
1834 	struct list_head	todo_list;
1835 
1836 	struct list_head	link_watch_list;
1837 
1838 	enum { NETREG_UNINITIALIZED=0,
1839 	       NETREG_REGISTERED,	/* completed register_netdevice */
1840 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1841 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1842 	       NETREG_RELEASED,		/* called free_netdev */
1843 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1844 	} reg_state:8;
1845 
1846 	bool dismantle;
1847 
1848 	enum {
1849 		RTNL_LINK_INITIALIZED,
1850 		RTNL_LINK_INITIALIZING,
1851 	} rtnl_link_state:16;
1852 
1853 	void (*destructor)(struct net_device *dev);
1854 
1855 #ifdef CONFIG_NETPOLL
1856 	struct netpoll_info __rcu	*npinfo;
1857 #endif
1858 
1859 	possible_net_t			nd_net;
1860 
1861 	/* mid-layer private */
1862 	union {
1863 		void					*ml_priv;
1864 		struct pcpu_lstats __percpu		*lstats;
1865 		struct pcpu_sw_netstats __percpu	*tstats;
1866 		struct pcpu_dstats __percpu		*dstats;
1867 		struct pcpu_vstats __percpu		*vstats;
1868 	};
1869 
1870 	struct garp_port __rcu	*garp_port;
1871 	struct mrp_port __rcu	*mrp_port;
1872 
1873 	struct device		dev;
1874 	const struct attribute_group *sysfs_groups[4];
1875 	const struct attribute_group *sysfs_rx_queue_group;
1876 
1877 	const struct rtnl_link_ops *rtnl_link_ops;
1878 
1879 	/* for setting kernel sock attribute on TCP connection setup */
1880 #define GSO_MAX_SIZE		65536
1881 	unsigned int		gso_max_size;
1882 #define GSO_MAX_SEGS		65535
1883 	u16			gso_max_segs;
1884 
1885 #ifdef CONFIG_DCB
1886 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1887 #endif
1888 	u8			num_tc;
1889 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1890 	u8			prio_tc_map[TC_BITMASK + 1];
1891 
1892 #if IS_ENABLED(CONFIG_FCOE)
1893 	unsigned int		fcoe_ddp_xid;
1894 #endif
1895 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1896 	struct netprio_map __rcu *priomap;
1897 #endif
1898 	struct phy_device	*phydev;
1899 	struct lock_class_key	*qdisc_tx_busylock;
1900 	struct lock_class_key	*qdisc_running_key;
1901 	bool			proto_down;
1902 };
1903 #define to_net_dev(d) container_of(d, struct net_device, dev)
1904 
1905 #define	NETDEV_ALIGN		32
1906 
1907 static inline
1908 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1909 {
1910 	return dev->prio_tc_map[prio & TC_BITMASK];
1911 }
1912 
1913 static inline
1914 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1915 {
1916 	if (tc >= dev->num_tc)
1917 		return -EINVAL;
1918 
1919 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1920 	return 0;
1921 }
1922 
1923 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
1924 void netdev_reset_tc(struct net_device *dev);
1925 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1926 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
1927 
1928 static inline
1929 int netdev_get_num_tc(struct net_device *dev)
1930 {
1931 	return dev->num_tc;
1932 }
1933 
1934 static inline
1935 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1936 					 unsigned int index)
1937 {
1938 	return &dev->_tx[index];
1939 }
1940 
1941 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1942 						    const struct sk_buff *skb)
1943 {
1944 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1945 }
1946 
1947 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1948 					    void (*f)(struct net_device *,
1949 						      struct netdev_queue *,
1950 						      void *),
1951 					    void *arg)
1952 {
1953 	unsigned int i;
1954 
1955 	for (i = 0; i < dev->num_tx_queues; i++)
1956 		f(dev, &dev->_tx[i], arg);
1957 }
1958 
1959 #define netdev_lockdep_set_classes(dev)				\
1960 {								\
1961 	static struct lock_class_key qdisc_tx_busylock_key;	\
1962 	static struct lock_class_key qdisc_running_key;		\
1963 	static struct lock_class_key qdisc_xmit_lock_key;	\
1964 	static struct lock_class_key dev_addr_list_lock_key;	\
1965 	unsigned int i;						\
1966 								\
1967 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
1968 	(dev)->qdisc_running_key = &qdisc_running_key;		\
1969 	lockdep_set_class(&(dev)->addr_list_lock,		\
1970 			  &dev_addr_list_lock_key); 		\
1971 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
1972 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
1973 				  &qdisc_xmit_lock_key);	\
1974 }
1975 
1976 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1977 				    struct sk_buff *skb,
1978 				    void *accel_priv);
1979 
1980 /* returns the headroom that the master device needs to take in account
1981  * when forwarding to this dev
1982  */
1983 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1984 {
1985 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1986 }
1987 
1988 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1989 {
1990 	if (dev->netdev_ops->ndo_set_rx_headroom)
1991 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1992 }
1993 
1994 /* set the device rx headroom to the dev's default */
1995 static inline void netdev_reset_rx_headroom(struct net_device *dev)
1996 {
1997 	netdev_set_rx_headroom(dev, -1);
1998 }
1999 
2000 /*
2001  * Net namespace inlines
2002  */
2003 static inline
2004 struct net *dev_net(const struct net_device *dev)
2005 {
2006 	return read_pnet(&dev->nd_net);
2007 }
2008 
2009 static inline
2010 void dev_net_set(struct net_device *dev, struct net *net)
2011 {
2012 	write_pnet(&dev->nd_net, net);
2013 }
2014 
2015 static inline bool netdev_uses_dsa(struct net_device *dev)
2016 {
2017 #if IS_ENABLED(CONFIG_NET_DSA)
2018 	if (dev->dsa_ptr != NULL)
2019 		return dsa_uses_tagged_protocol(dev->dsa_ptr);
2020 #endif
2021 	return false;
2022 }
2023 
2024 /**
2025  *	netdev_priv - access network device private data
2026  *	@dev: network device
2027  *
2028  * Get network device private data
2029  */
2030 static inline void *netdev_priv(const struct net_device *dev)
2031 {
2032 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2033 }
2034 
2035 /* Set the sysfs physical device reference for the network logical device
2036  * if set prior to registration will cause a symlink during initialization.
2037  */
2038 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2039 
2040 /* Set the sysfs device type for the network logical device to allow
2041  * fine-grained identification of different network device types. For
2042  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2043  */
2044 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2045 
2046 /* Default NAPI poll() weight
2047  * Device drivers are strongly advised to not use bigger value
2048  */
2049 #define NAPI_POLL_WEIGHT 64
2050 
2051 /**
2052  *	netif_napi_add - initialize a NAPI context
2053  *	@dev:  network device
2054  *	@napi: NAPI context
2055  *	@poll: polling function
2056  *	@weight: default weight
2057  *
2058  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2059  * *any* of the other NAPI-related functions.
2060  */
2061 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2062 		    int (*poll)(struct napi_struct *, int), int weight);
2063 
2064 /**
2065  *	netif_tx_napi_add - initialize a NAPI context
2066  *	@dev:  network device
2067  *	@napi: NAPI context
2068  *	@poll: polling function
2069  *	@weight: default weight
2070  *
2071  * This variant of netif_napi_add() should be used from drivers using NAPI
2072  * to exclusively poll a TX queue.
2073  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2074  */
2075 static inline void netif_tx_napi_add(struct net_device *dev,
2076 				     struct napi_struct *napi,
2077 				     int (*poll)(struct napi_struct *, int),
2078 				     int weight)
2079 {
2080 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2081 	netif_napi_add(dev, napi, poll, weight);
2082 }
2083 
2084 /**
2085  *  netif_napi_del - remove a NAPI context
2086  *  @napi: NAPI context
2087  *
2088  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2089  */
2090 void netif_napi_del(struct napi_struct *napi);
2091 
2092 struct napi_gro_cb {
2093 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2094 	void	*frag0;
2095 
2096 	/* Length of frag0. */
2097 	unsigned int frag0_len;
2098 
2099 	/* This indicates where we are processing relative to skb->data. */
2100 	int	data_offset;
2101 
2102 	/* This is non-zero if the packet cannot be merged with the new skb. */
2103 	u16	flush;
2104 
2105 	/* Save the IP ID here and check when we get to the transport layer */
2106 	u16	flush_id;
2107 
2108 	/* Number of segments aggregated. */
2109 	u16	count;
2110 
2111 	/* Start offset for remote checksum offload */
2112 	u16	gro_remcsum_start;
2113 
2114 	/* jiffies when first packet was created/queued */
2115 	unsigned long age;
2116 
2117 	/* Used in ipv6_gro_receive() and foo-over-udp */
2118 	u16	proto;
2119 
2120 	/* This is non-zero if the packet may be of the same flow. */
2121 	u8	same_flow:1;
2122 
2123 	/* Used in tunnel GRO receive */
2124 	u8	encap_mark:1;
2125 
2126 	/* GRO checksum is valid */
2127 	u8	csum_valid:1;
2128 
2129 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2130 	u8	csum_cnt:3;
2131 
2132 	/* Free the skb? */
2133 	u8	free:2;
2134 #define NAPI_GRO_FREE		  1
2135 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2136 
2137 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2138 	u8	is_ipv6:1;
2139 
2140 	/* Used in GRE, set in fou/gue_gro_receive */
2141 	u8	is_fou:1;
2142 
2143 	/* Used to determine if flush_id can be ignored */
2144 	u8	is_atomic:1;
2145 
2146 	/* Number of gro_receive callbacks this packet already went through */
2147 	u8 recursion_counter:4;
2148 
2149 	/* 1 bit hole */
2150 
2151 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2152 	__wsum	csum;
2153 
2154 	/* used in skb_gro_receive() slow path */
2155 	struct sk_buff *last;
2156 };
2157 
2158 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2159 
2160 #define GRO_RECURSION_LIMIT 15
2161 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2162 {
2163 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2164 }
2165 
2166 typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2167 static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
2168 						struct sk_buff **head,
2169 						struct sk_buff *skb)
2170 {
2171 	if (unlikely(gro_recursion_inc_test(skb))) {
2172 		NAPI_GRO_CB(skb)->flush |= 1;
2173 		return NULL;
2174 	}
2175 
2176 	return cb(head, skb);
2177 }
2178 
2179 typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2180 					     struct sk_buff *);
2181 static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2182 						   struct sock *sk,
2183 						   struct sk_buff **head,
2184 						   struct sk_buff *skb)
2185 {
2186 	if (unlikely(gro_recursion_inc_test(skb))) {
2187 		NAPI_GRO_CB(skb)->flush |= 1;
2188 		return NULL;
2189 	}
2190 
2191 	return cb(sk, head, skb);
2192 }
2193 
2194 struct packet_type {
2195 	__be16			type;	/* This is really htons(ether_type). */
2196 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2197 	int			(*func) (struct sk_buff *,
2198 					 struct net_device *,
2199 					 struct packet_type *,
2200 					 struct net_device *);
2201 	bool			(*id_match)(struct packet_type *ptype,
2202 					    struct sock *sk);
2203 	void			*af_packet_priv;
2204 	struct list_head	list;
2205 };
2206 
2207 struct offload_callbacks {
2208 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2209 						netdev_features_t features);
2210 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2211 						 struct sk_buff *skb);
2212 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2213 };
2214 
2215 struct packet_offload {
2216 	__be16			 type;	/* This is really htons(ether_type). */
2217 	u16			 priority;
2218 	struct offload_callbacks callbacks;
2219 	struct list_head	 list;
2220 };
2221 
2222 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2223 struct pcpu_sw_netstats {
2224 	u64     rx_packets;
2225 	u64     rx_bytes;
2226 	u64     tx_packets;
2227 	u64     tx_bytes;
2228 	struct u64_stats_sync   syncp;
2229 };
2230 
2231 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2232 ({									\
2233 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2234 	if (pcpu_stats)	{						\
2235 		int __cpu;						\
2236 		for_each_possible_cpu(__cpu) {				\
2237 			typeof(type) *stat;				\
2238 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2239 			u64_stats_init(&stat->syncp);			\
2240 		}							\
2241 	}								\
2242 	pcpu_stats;							\
2243 })
2244 
2245 #define netdev_alloc_pcpu_stats(type)					\
2246 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2247 
2248 enum netdev_lag_tx_type {
2249 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2250 	NETDEV_LAG_TX_TYPE_RANDOM,
2251 	NETDEV_LAG_TX_TYPE_BROADCAST,
2252 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2253 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2254 	NETDEV_LAG_TX_TYPE_HASH,
2255 };
2256 
2257 struct netdev_lag_upper_info {
2258 	enum netdev_lag_tx_type tx_type;
2259 };
2260 
2261 struct netdev_lag_lower_state_info {
2262 	u8 link_up : 1,
2263 	   tx_enabled : 1;
2264 };
2265 
2266 #include <linux/notifier.h>
2267 
2268 /* netdevice notifier chain. Please remember to update the rtnetlink
2269  * notification exclusion list in rtnetlink_event() when adding new
2270  * types.
2271  */
2272 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2273 #define NETDEV_DOWN	0x0002
2274 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2275 				   detected a hardware crash and restarted
2276 				   - we can use this eg to kick tcp sessions
2277 				   once done */
2278 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2279 #define NETDEV_REGISTER 0x0005
2280 #define NETDEV_UNREGISTER	0x0006
2281 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2282 #define NETDEV_CHANGEADDR	0x0008
2283 #define NETDEV_GOING_DOWN	0x0009
2284 #define NETDEV_CHANGENAME	0x000A
2285 #define NETDEV_FEAT_CHANGE	0x000B
2286 #define NETDEV_BONDING_FAILOVER 0x000C
2287 #define NETDEV_PRE_UP		0x000D
2288 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2289 #define NETDEV_POST_TYPE_CHANGE	0x000F
2290 #define NETDEV_POST_INIT	0x0010
2291 #define NETDEV_UNREGISTER_FINAL 0x0011
2292 #define NETDEV_RELEASE		0x0012
2293 #define NETDEV_NOTIFY_PEERS	0x0013
2294 #define NETDEV_JOIN		0x0014
2295 #define NETDEV_CHANGEUPPER	0x0015
2296 #define NETDEV_RESEND_IGMP	0x0016
2297 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2298 #define NETDEV_CHANGEINFODATA	0x0018
2299 #define NETDEV_BONDING_INFO	0x0019
2300 #define NETDEV_PRECHANGEUPPER	0x001A
2301 #define NETDEV_CHANGELOWERSTATE	0x001B
2302 #define NETDEV_UDP_TUNNEL_PUSH_INFO	0x001C
2303 #define NETDEV_CHANGE_TX_QUEUE_LEN	0x001E
2304 
2305 int register_netdevice_notifier(struct notifier_block *nb);
2306 int unregister_netdevice_notifier(struct notifier_block *nb);
2307 
2308 struct netdev_notifier_info {
2309 	struct net_device *dev;
2310 };
2311 
2312 struct netdev_notifier_change_info {
2313 	struct netdev_notifier_info info; /* must be first */
2314 	unsigned int flags_changed;
2315 };
2316 
2317 struct netdev_notifier_changeupper_info {
2318 	struct netdev_notifier_info info; /* must be first */
2319 	struct net_device *upper_dev; /* new upper dev */
2320 	bool master; /* is upper dev master */
2321 	bool linking; /* is the notification for link or unlink */
2322 	void *upper_info; /* upper dev info */
2323 };
2324 
2325 struct netdev_notifier_changelowerstate_info {
2326 	struct netdev_notifier_info info; /* must be first */
2327 	void *lower_state_info; /* is lower dev state */
2328 };
2329 
2330 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2331 					     struct net_device *dev)
2332 {
2333 	info->dev = dev;
2334 }
2335 
2336 static inline struct net_device *
2337 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2338 {
2339 	return info->dev;
2340 }
2341 
2342 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2343 
2344 
2345 extern rwlock_t				dev_base_lock;		/* Device list lock */
2346 
2347 #define for_each_netdev(net, d)		\
2348 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2349 #define for_each_netdev_reverse(net, d)	\
2350 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2351 #define for_each_netdev_rcu(net, d)		\
2352 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2353 #define for_each_netdev_safe(net, d, n)	\
2354 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2355 #define for_each_netdev_continue(net, d)		\
2356 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2357 #define for_each_netdev_continue_rcu(net, d)		\
2358 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2359 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2360 		for_each_netdev_rcu(&init_net, slave)	\
2361 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2362 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2363 
2364 static inline struct net_device *next_net_device(struct net_device *dev)
2365 {
2366 	struct list_head *lh;
2367 	struct net *net;
2368 
2369 	net = dev_net(dev);
2370 	lh = dev->dev_list.next;
2371 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2372 }
2373 
2374 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2375 {
2376 	struct list_head *lh;
2377 	struct net *net;
2378 
2379 	net = dev_net(dev);
2380 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2381 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2382 }
2383 
2384 static inline struct net_device *first_net_device(struct net *net)
2385 {
2386 	return list_empty(&net->dev_base_head) ? NULL :
2387 		net_device_entry(net->dev_base_head.next);
2388 }
2389 
2390 static inline struct net_device *first_net_device_rcu(struct net *net)
2391 {
2392 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2393 
2394 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2395 }
2396 
2397 int netdev_boot_setup_check(struct net_device *dev);
2398 unsigned long netdev_boot_base(const char *prefix, int unit);
2399 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2400 				       const char *hwaddr);
2401 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2402 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2403 void dev_add_pack(struct packet_type *pt);
2404 void dev_remove_pack(struct packet_type *pt);
2405 void __dev_remove_pack(struct packet_type *pt);
2406 void dev_add_offload(struct packet_offload *po);
2407 void dev_remove_offload(struct packet_offload *po);
2408 
2409 int dev_get_iflink(const struct net_device *dev);
2410 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2411 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2412 				      unsigned short mask);
2413 struct net_device *dev_get_by_name(struct net *net, const char *name);
2414 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2415 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2416 int dev_alloc_name(struct net_device *dev, const char *name);
2417 int dev_open(struct net_device *dev);
2418 int dev_close(struct net_device *dev);
2419 int dev_close_many(struct list_head *head, bool unlink);
2420 void dev_disable_lro(struct net_device *dev);
2421 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2422 int dev_queue_xmit(struct sk_buff *skb);
2423 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2424 int register_netdevice(struct net_device *dev);
2425 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2426 void unregister_netdevice_many(struct list_head *head);
2427 static inline void unregister_netdevice(struct net_device *dev)
2428 {
2429 	unregister_netdevice_queue(dev, NULL);
2430 }
2431 
2432 int netdev_refcnt_read(const struct net_device *dev);
2433 void free_netdev(struct net_device *dev);
2434 void netdev_freemem(struct net_device *dev);
2435 void synchronize_net(void);
2436 int init_dummy_netdev(struct net_device *dev);
2437 
2438 DECLARE_PER_CPU(int, xmit_recursion);
2439 #define XMIT_RECURSION_LIMIT	10
2440 
2441 static inline int dev_recursion_level(void)
2442 {
2443 	return this_cpu_read(xmit_recursion);
2444 }
2445 
2446 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2447 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2448 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2449 int netdev_get_name(struct net *net, char *name, int ifindex);
2450 int dev_restart(struct net_device *dev);
2451 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2452 
2453 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2454 {
2455 	return NAPI_GRO_CB(skb)->data_offset;
2456 }
2457 
2458 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2459 {
2460 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2461 }
2462 
2463 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2464 {
2465 	NAPI_GRO_CB(skb)->data_offset += len;
2466 }
2467 
2468 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2469 					unsigned int offset)
2470 {
2471 	return NAPI_GRO_CB(skb)->frag0 + offset;
2472 }
2473 
2474 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2475 {
2476 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2477 }
2478 
2479 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2480 {
2481 	NAPI_GRO_CB(skb)->frag0 = NULL;
2482 	NAPI_GRO_CB(skb)->frag0_len = 0;
2483 }
2484 
2485 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2486 					unsigned int offset)
2487 {
2488 	if (!pskb_may_pull(skb, hlen))
2489 		return NULL;
2490 
2491 	skb_gro_frag0_invalidate(skb);
2492 	return skb->data + offset;
2493 }
2494 
2495 static inline void *skb_gro_network_header(struct sk_buff *skb)
2496 {
2497 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2498 	       skb_network_offset(skb);
2499 }
2500 
2501 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2502 					const void *start, unsigned int len)
2503 {
2504 	if (NAPI_GRO_CB(skb)->csum_valid)
2505 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2506 						  csum_partial(start, len, 0));
2507 }
2508 
2509 /* GRO checksum functions. These are logical equivalents of the normal
2510  * checksum functions (in skbuff.h) except that they operate on the GRO
2511  * offsets and fields in sk_buff.
2512  */
2513 
2514 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2515 
2516 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2517 {
2518 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2519 }
2520 
2521 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2522 						      bool zero_okay,
2523 						      __sum16 check)
2524 {
2525 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2526 		skb_checksum_start_offset(skb) <
2527 		 skb_gro_offset(skb)) &&
2528 		!skb_at_gro_remcsum_start(skb) &&
2529 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2530 		(!zero_okay || check));
2531 }
2532 
2533 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2534 							   __wsum psum)
2535 {
2536 	if (NAPI_GRO_CB(skb)->csum_valid &&
2537 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2538 		return 0;
2539 
2540 	NAPI_GRO_CB(skb)->csum = psum;
2541 
2542 	return __skb_gro_checksum_complete(skb);
2543 }
2544 
2545 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2546 {
2547 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2548 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2549 		NAPI_GRO_CB(skb)->csum_cnt--;
2550 	} else {
2551 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2552 		 * verified a new top level checksum or an encapsulated one
2553 		 * during GRO. This saves work if we fallback to normal path.
2554 		 */
2555 		__skb_incr_checksum_unnecessary(skb);
2556 	}
2557 }
2558 
2559 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2560 				    compute_pseudo)			\
2561 ({									\
2562 	__sum16 __ret = 0;						\
2563 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2564 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2565 				compute_pseudo(skb, proto));		\
2566 	if (__ret)							\
2567 		__skb_mark_checksum_bad(skb);				\
2568 	else								\
2569 		skb_gro_incr_csum_unnecessary(skb);			\
2570 	__ret;								\
2571 })
2572 
2573 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2574 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2575 
2576 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2577 					     compute_pseudo)		\
2578 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2579 
2580 #define skb_gro_checksum_simple_validate(skb)				\
2581 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2582 
2583 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2584 {
2585 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2586 		!NAPI_GRO_CB(skb)->csum_valid);
2587 }
2588 
2589 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2590 					      __sum16 check, __wsum pseudo)
2591 {
2592 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2593 	NAPI_GRO_CB(skb)->csum_valid = 1;
2594 }
2595 
2596 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2597 do {									\
2598 	if (__skb_gro_checksum_convert_check(skb))			\
2599 		__skb_gro_checksum_convert(skb, check,			\
2600 					   compute_pseudo(skb, proto));	\
2601 } while (0)
2602 
2603 struct gro_remcsum {
2604 	int offset;
2605 	__wsum delta;
2606 };
2607 
2608 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2609 {
2610 	grc->offset = 0;
2611 	grc->delta = 0;
2612 }
2613 
2614 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2615 					    unsigned int off, size_t hdrlen,
2616 					    int start, int offset,
2617 					    struct gro_remcsum *grc,
2618 					    bool nopartial)
2619 {
2620 	__wsum delta;
2621 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2622 
2623 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2624 
2625 	if (!nopartial) {
2626 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2627 		return ptr;
2628 	}
2629 
2630 	ptr = skb_gro_header_fast(skb, off);
2631 	if (skb_gro_header_hard(skb, off + plen)) {
2632 		ptr = skb_gro_header_slow(skb, off + plen, off);
2633 		if (!ptr)
2634 			return NULL;
2635 	}
2636 
2637 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2638 			       start, offset);
2639 
2640 	/* Adjust skb->csum since we changed the packet */
2641 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2642 
2643 	grc->offset = off + hdrlen + offset;
2644 	grc->delta = delta;
2645 
2646 	return ptr;
2647 }
2648 
2649 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2650 					   struct gro_remcsum *grc)
2651 {
2652 	void *ptr;
2653 	size_t plen = grc->offset + sizeof(u16);
2654 
2655 	if (!grc->delta)
2656 		return;
2657 
2658 	ptr = skb_gro_header_fast(skb, grc->offset);
2659 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2660 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2661 		if (!ptr)
2662 			return;
2663 	}
2664 
2665 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2666 }
2667 
2668 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2669 				  unsigned short type,
2670 				  const void *daddr, const void *saddr,
2671 				  unsigned int len)
2672 {
2673 	if (!dev->header_ops || !dev->header_ops->create)
2674 		return 0;
2675 
2676 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2677 }
2678 
2679 static inline int dev_parse_header(const struct sk_buff *skb,
2680 				   unsigned char *haddr)
2681 {
2682 	const struct net_device *dev = skb->dev;
2683 
2684 	if (!dev->header_ops || !dev->header_ops->parse)
2685 		return 0;
2686 	return dev->header_ops->parse(skb, haddr);
2687 }
2688 
2689 /* ll_header must have at least hard_header_len allocated */
2690 static inline bool dev_validate_header(const struct net_device *dev,
2691 				       char *ll_header, int len)
2692 {
2693 	if (likely(len >= dev->hard_header_len))
2694 		return true;
2695 
2696 	if (capable(CAP_SYS_RAWIO)) {
2697 		memset(ll_header + len, 0, dev->hard_header_len - len);
2698 		return true;
2699 	}
2700 
2701 	if (dev->header_ops && dev->header_ops->validate)
2702 		return dev->header_ops->validate(ll_header, len);
2703 
2704 	return false;
2705 }
2706 
2707 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2708 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2709 static inline int unregister_gifconf(unsigned int family)
2710 {
2711 	return register_gifconf(family, NULL);
2712 }
2713 
2714 #ifdef CONFIG_NET_FLOW_LIMIT
2715 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2716 struct sd_flow_limit {
2717 	u64			count;
2718 	unsigned int		num_buckets;
2719 	unsigned int		history_head;
2720 	u16			history[FLOW_LIMIT_HISTORY];
2721 	u8			buckets[];
2722 };
2723 
2724 extern int netdev_flow_limit_table_len;
2725 #endif /* CONFIG_NET_FLOW_LIMIT */
2726 
2727 /*
2728  * Incoming packets are placed on per-CPU queues
2729  */
2730 struct softnet_data {
2731 	struct list_head	poll_list;
2732 	struct sk_buff_head	process_queue;
2733 
2734 	/* stats */
2735 	unsigned int		processed;
2736 	unsigned int		time_squeeze;
2737 	unsigned int		received_rps;
2738 #ifdef CONFIG_RPS
2739 	struct softnet_data	*rps_ipi_list;
2740 #endif
2741 #ifdef CONFIG_NET_FLOW_LIMIT
2742 	struct sd_flow_limit __rcu *flow_limit;
2743 #endif
2744 	struct Qdisc		*output_queue;
2745 	struct Qdisc		**output_queue_tailp;
2746 	struct sk_buff		*completion_queue;
2747 
2748 #ifdef CONFIG_RPS
2749 	/* input_queue_head should be written by cpu owning this struct,
2750 	 * and only read by other cpus. Worth using a cache line.
2751 	 */
2752 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2753 
2754 	/* Elements below can be accessed between CPUs for RPS/RFS */
2755 	struct call_single_data	csd ____cacheline_aligned_in_smp;
2756 	struct softnet_data	*rps_ipi_next;
2757 	unsigned int		cpu;
2758 	unsigned int		input_queue_tail;
2759 #endif
2760 	unsigned int		dropped;
2761 	struct sk_buff_head	input_pkt_queue;
2762 	struct napi_struct	backlog;
2763 
2764 };
2765 
2766 static inline void input_queue_head_incr(struct softnet_data *sd)
2767 {
2768 #ifdef CONFIG_RPS
2769 	sd->input_queue_head++;
2770 #endif
2771 }
2772 
2773 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2774 					      unsigned int *qtail)
2775 {
2776 #ifdef CONFIG_RPS
2777 	*qtail = ++sd->input_queue_tail;
2778 #endif
2779 }
2780 
2781 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2782 
2783 void __netif_schedule(struct Qdisc *q);
2784 void netif_schedule_queue(struct netdev_queue *txq);
2785 
2786 static inline void netif_tx_schedule_all(struct net_device *dev)
2787 {
2788 	unsigned int i;
2789 
2790 	for (i = 0; i < dev->num_tx_queues; i++)
2791 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2792 }
2793 
2794 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2795 {
2796 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2797 }
2798 
2799 /**
2800  *	netif_start_queue - allow transmit
2801  *	@dev: network device
2802  *
2803  *	Allow upper layers to call the device hard_start_xmit routine.
2804  */
2805 static inline void netif_start_queue(struct net_device *dev)
2806 {
2807 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2808 }
2809 
2810 static inline void netif_tx_start_all_queues(struct net_device *dev)
2811 {
2812 	unsigned int i;
2813 
2814 	for (i = 0; i < dev->num_tx_queues; i++) {
2815 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2816 		netif_tx_start_queue(txq);
2817 	}
2818 }
2819 
2820 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2821 
2822 /**
2823  *	netif_wake_queue - restart transmit
2824  *	@dev: network device
2825  *
2826  *	Allow upper layers to call the device hard_start_xmit routine.
2827  *	Used for flow control when transmit resources are available.
2828  */
2829 static inline void netif_wake_queue(struct net_device *dev)
2830 {
2831 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2832 }
2833 
2834 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2835 {
2836 	unsigned int i;
2837 
2838 	for (i = 0; i < dev->num_tx_queues; i++) {
2839 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2840 		netif_tx_wake_queue(txq);
2841 	}
2842 }
2843 
2844 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2845 {
2846 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2847 }
2848 
2849 /**
2850  *	netif_stop_queue - stop transmitted packets
2851  *	@dev: network device
2852  *
2853  *	Stop upper layers calling the device hard_start_xmit routine.
2854  *	Used for flow control when transmit resources are unavailable.
2855  */
2856 static inline void netif_stop_queue(struct net_device *dev)
2857 {
2858 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2859 }
2860 
2861 void netif_tx_stop_all_queues(struct net_device *dev);
2862 
2863 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2864 {
2865 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2866 }
2867 
2868 /**
2869  *	netif_queue_stopped - test if transmit queue is flowblocked
2870  *	@dev: network device
2871  *
2872  *	Test if transmit queue on device is currently unable to send.
2873  */
2874 static inline bool netif_queue_stopped(const struct net_device *dev)
2875 {
2876 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2877 }
2878 
2879 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2880 {
2881 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2882 }
2883 
2884 static inline bool
2885 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2886 {
2887 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2888 }
2889 
2890 static inline bool
2891 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2892 {
2893 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2894 }
2895 
2896 /**
2897  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2898  *	@dev_queue: pointer to transmit queue
2899  *
2900  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2901  * to give appropriate hint to the CPU.
2902  */
2903 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2904 {
2905 #ifdef CONFIG_BQL
2906 	prefetchw(&dev_queue->dql.num_queued);
2907 #endif
2908 }
2909 
2910 /**
2911  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2912  *	@dev_queue: pointer to transmit queue
2913  *
2914  * BQL enabled drivers might use this helper in their TX completion path,
2915  * to give appropriate hint to the CPU.
2916  */
2917 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2918 {
2919 #ifdef CONFIG_BQL
2920 	prefetchw(&dev_queue->dql.limit);
2921 #endif
2922 }
2923 
2924 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2925 					unsigned int bytes)
2926 {
2927 #ifdef CONFIG_BQL
2928 	dql_queued(&dev_queue->dql, bytes);
2929 
2930 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2931 		return;
2932 
2933 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2934 
2935 	/*
2936 	 * The XOFF flag must be set before checking the dql_avail below,
2937 	 * because in netdev_tx_completed_queue we update the dql_completed
2938 	 * before checking the XOFF flag.
2939 	 */
2940 	smp_mb();
2941 
2942 	/* check again in case another CPU has just made room avail */
2943 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2944 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2945 #endif
2946 }
2947 
2948 /**
2949  * 	netdev_sent_queue - report the number of bytes queued to hardware
2950  * 	@dev: network device
2951  * 	@bytes: number of bytes queued to the hardware device queue
2952  *
2953  * 	Report the number of bytes queued for sending/completion to the network
2954  * 	device hardware queue. @bytes should be a good approximation and should
2955  * 	exactly match netdev_completed_queue() @bytes
2956  */
2957 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2958 {
2959 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2960 }
2961 
2962 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2963 					     unsigned int pkts, unsigned int bytes)
2964 {
2965 #ifdef CONFIG_BQL
2966 	if (unlikely(!bytes))
2967 		return;
2968 
2969 	dql_completed(&dev_queue->dql, bytes);
2970 
2971 	/*
2972 	 * Without the memory barrier there is a small possiblity that
2973 	 * netdev_tx_sent_queue will miss the update and cause the queue to
2974 	 * be stopped forever
2975 	 */
2976 	smp_mb();
2977 
2978 	if (dql_avail(&dev_queue->dql) < 0)
2979 		return;
2980 
2981 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2982 		netif_schedule_queue(dev_queue);
2983 #endif
2984 }
2985 
2986 /**
2987  * 	netdev_completed_queue - report bytes and packets completed by device
2988  * 	@dev: network device
2989  * 	@pkts: actual number of packets sent over the medium
2990  * 	@bytes: actual number of bytes sent over the medium
2991  *
2992  * 	Report the number of bytes and packets transmitted by the network device
2993  * 	hardware queue over the physical medium, @bytes must exactly match the
2994  * 	@bytes amount passed to netdev_sent_queue()
2995  */
2996 static inline void netdev_completed_queue(struct net_device *dev,
2997 					  unsigned int pkts, unsigned int bytes)
2998 {
2999 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3000 }
3001 
3002 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3003 {
3004 #ifdef CONFIG_BQL
3005 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3006 	dql_reset(&q->dql);
3007 #endif
3008 }
3009 
3010 /**
3011  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3012  * 	@dev_queue: network device
3013  *
3014  * 	Reset the bytes and packet count of a network device and clear the
3015  * 	software flow control OFF bit for this network device
3016  */
3017 static inline void netdev_reset_queue(struct net_device *dev_queue)
3018 {
3019 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3020 }
3021 
3022 /**
3023  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3024  * 	@dev: network device
3025  * 	@queue_index: given tx queue index
3026  *
3027  * 	Returns 0 if given tx queue index >= number of device tx queues,
3028  * 	otherwise returns the originally passed tx queue index.
3029  */
3030 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3031 {
3032 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3033 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3034 				     dev->name, queue_index,
3035 				     dev->real_num_tx_queues);
3036 		return 0;
3037 	}
3038 
3039 	return queue_index;
3040 }
3041 
3042 /**
3043  *	netif_running - test if up
3044  *	@dev: network device
3045  *
3046  *	Test if the device has been brought up.
3047  */
3048 static inline bool netif_running(const struct net_device *dev)
3049 {
3050 	return test_bit(__LINK_STATE_START, &dev->state);
3051 }
3052 
3053 /*
3054  * Routines to manage the subqueues on a device.  We only need start,
3055  * stop, and a check if it's stopped.  All other device management is
3056  * done at the overall netdevice level.
3057  * Also test the device if we're multiqueue.
3058  */
3059 
3060 /**
3061  *	netif_start_subqueue - allow sending packets on subqueue
3062  *	@dev: network device
3063  *	@queue_index: sub queue index
3064  *
3065  * Start individual transmit queue of a device with multiple transmit queues.
3066  */
3067 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3068 {
3069 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3070 
3071 	netif_tx_start_queue(txq);
3072 }
3073 
3074 /**
3075  *	netif_stop_subqueue - stop sending packets on subqueue
3076  *	@dev: network device
3077  *	@queue_index: sub queue index
3078  *
3079  * Stop individual transmit queue of a device with multiple transmit queues.
3080  */
3081 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3082 {
3083 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3084 	netif_tx_stop_queue(txq);
3085 }
3086 
3087 /**
3088  *	netif_subqueue_stopped - test status of subqueue
3089  *	@dev: network device
3090  *	@queue_index: sub queue index
3091  *
3092  * Check individual transmit queue of a device with multiple transmit queues.
3093  */
3094 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3095 					    u16 queue_index)
3096 {
3097 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3098 
3099 	return netif_tx_queue_stopped(txq);
3100 }
3101 
3102 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3103 					  struct sk_buff *skb)
3104 {
3105 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3106 }
3107 
3108 /**
3109  *	netif_wake_subqueue - allow sending packets on subqueue
3110  *	@dev: network device
3111  *	@queue_index: sub queue index
3112  *
3113  * Resume individual transmit queue of a device with multiple transmit queues.
3114  */
3115 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3116 {
3117 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3118 
3119 	netif_tx_wake_queue(txq);
3120 }
3121 
3122 #ifdef CONFIG_XPS
3123 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3124 			u16 index);
3125 #else
3126 static inline int netif_set_xps_queue(struct net_device *dev,
3127 				      const struct cpumask *mask,
3128 				      u16 index)
3129 {
3130 	return 0;
3131 }
3132 #endif
3133 
3134 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3135 		  unsigned int num_tx_queues);
3136 
3137 /*
3138  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3139  * as a distribution range limit for the returned value.
3140  */
3141 static inline u16 skb_tx_hash(const struct net_device *dev,
3142 			      struct sk_buff *skb)
3143 {
3144 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3145 }
3146 
3147 /**
3148  *	netif_is_multiqueue - test if device has multiple transmit queues
3149  *	@dev: network device
3150  *
3151  * Check if device has multiple transmit queues
3152  */
3153 static inline bool netif_is_multiqueue(const struct net_device *dev)
3154 {
3155 	return dev->num_tx_queues > 1;
3156 }
3157 
3158 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3159 
3160 #ifdef CONFIG_SYSFS
3161 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3162 #else
3163 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3164 						unsigned int rxq)
3165 {
3166 	return 0;
3167 }
3168 #endif
3169 
3170 #ifdef CONFIG_SYSFS
3171 static inline unsigned int get_netdev_rx_queue_index(
3172 		struct netdev_rx_queue *queue)
3173 {
3174 	struct net_device *dev = queue->dev;
3175 	int index = queue - dev->_rx;
3176 
3177 	BUG_ON(index >= dev->num_rx_queues);
3178 	return index;
3179 }
3180 #endif
3181 
3182 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3183 int netif_get_num_default_rss_queues(void);
3184 
3185 enum skb_free_reason {
3186 	SKB_REASON_CONSUMED,
3187 	SKB_REASON_DROPPED,
3188 };
3189 
3190 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3191 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3192 
3193 /*
3194  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3195  * interrupt context or with hardware interrupts being disabled.
3196  * (in_irq() || irqs_disabled())
3197  *
3198  * We provide four helpers that can be used in following contexts :
3199  *
3200  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3201  *  replacing kfree_skb(skb)
3202  *
3203  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3204  *  Typically used in place of consume_skb(skb) in TX completion path
3205  *
3206  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3207  *  replacing kfree_skb(skb)
3208  *
3209  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3210  *  and consumed a packet. Used in place of consume_skb(skb)
3211  */
3212 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3213 {
3214 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3215 }
3216 
3217 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3218 {
3219 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3220 }
3221 
3222 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3223 {
3224 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3225 }
3226 
3227 static inline void dev_consume_skb_any(struct sk_buff *skb)
3228 {
3229 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3230 }
3231 
3232 int netif_rx(struct sk_buff *skb);
3233 int netif_rx_ni(struct sk_buff *skb);
3234 int netif_receive_skb(struct sk_buff *skb);
3235 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3236 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3237 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3238 gro_result_t napi_gro_frags(struct napi_struct *napi);
3239 struct packet_offload *gro_find_receive_by_type(__be16 type);
3240 struct packet_offload *gro_find_complete_by_type(__be16 type);
3241 
3242 static inline void napi_free_frags(struct napi_struct *napi)
3243 {
3244 	kfree_skb(napi->skb);
3245 	napi->skb = NULL;
3246 }
3247 
3248 bool netdev_is_rx_handler_busy(struct net_device *dev);
3249 int netdev_rx_handler_register(struct net_device *dev,
3250 			       rx_handler_func_t *rx_handler,
3251 			       void *rx_handler_data);
3252 void netdev_rx_handler_unregister(struct net_device *dev);
3253 
3254 bool dev_valid_name(const char *name);
3255 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3256 int dev_ethtool(struct net *net, struct ifreq *);
3257 unsigned int dev_get_flags(const struct net_device *);
3258 int __dev_change_flags(struct net_device *, unsigned int flags);
3259 int dev_change_flags(struct net_device *, unsigned int);
3260 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3261 			unsigned int gchanges);
3262 int dev_change_name(struct net_device *, const char *);
3263 int dev_set_alias(struct net_device *, const char *, size_t);
3264 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3265 int dev_set_mtu(struct net_device *, int);
3266 void dev_set_group(struct net_device *, int);
3267 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3268 int dev_change_carrier(struct net_device *, bool new_carrier);
3269 int dev_get_phys_port_id(struct net_device *dev,
3270 			 struct netdev_phys_item_id *ppid);
3271 int dev_get_phys_port_name(struct net_device *dev,
3272 			   char *name, size_t len);
3273 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3274 int dev_change_xdp_fd(struct net_device *dev, int fd, u32 flags);
3275 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3276 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3277 				    struct netdev_queue *txq, int *ret);
3278 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3279 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3280 bool is_skb_forwardable(const struct net_device *dev,
3281 			const struct sk_buff *skb);
3282 
3283 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3284 					       struct sk_buff *skb)
3285 {
3286 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3287 	    unlikely(!is_skb_forwardable(dev, skb))) {
3288 		atomic_long_inc(&dev->rx_dropped);
3289 		kfree_skb(skb);
3290 		return NET_RX_DROP;
3291 	}
3292 
3293 	skb_scrub_packet(skb, true);
3294 	skb->priority = 0;
3295 	return 0;
3296 }
3297 
3298 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3299 
3300 extern int		netdev_budget;
3301 
3302 /* Called by rtnetlink.c:rtnl_unlock() */
3303 void netdev_run_todo(void);
3304 
3305 /**
3306  *	dev_put - release reference to device
3307  *	@dev: network device
3308  *
3309  * Release reference to device to allow it to be freed.
3310  */
3311 static inline void dev_put(struct net_device *dev)
3312 {
3313 	this_cpu_dec(*dev->pcpu_refcnt);
3314 }
3315 
3316 /**
3317  *	dev_hold - get reference to device
3318  *	@dev: network device
3319  *
3320  * Hold reference to device to keep it from being freed.
3321  */
3322 static inline void dev_hold(struct net_device *dev)
3323 {
3324 	this_cpu_inc(*dev->pcpu_refcnt);
3325 }
3326 
3327 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3328  * and _off may be called from IRQ context, but it is caller
3329  * who is responsible for serialization of these calls.
3330  *
3331  * The name carrier is inappropriate, these functions should really be
3332  * called netif_lowerlayer_*() because they represent the state of any
3333  * kind of lower layer not just hardware media.
3334  */
3335 
3336 void linkwatch_init_dev(struct net_device *dev);
3337 void linkwatch_fire_event(struct net_device *dev);
3338 void linkwatch_forget_dev(struct net_device *dev);
3339 
3340 /**
3341  *	netif_carrier_ok - test if carrier present
3342  *	@dev: network device
3343  *
3344  * Check if carrier is present on device
3345  */
3346 static inline bool netif_carrier_ok(const struct net_device *dev)
3347 {
3348 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3349 }
3350 
3351 unsigned long dev_trans_start(struct net_device *dev);
3352 
3353 void __netdev_watchdog_up(struct net_device *dev);
3354 
3355 void netif_carrier_on(struct net_device *dev);
3356 
3357 void netif_carrier_off(struct net_device *dev);
3358 
3359 /**
3360  *	netif_dormant_on - mark device as dormant.
3361  *	@dev: network device
3362  *
3363  * Mark device as dormant (as per RFC2863).
3364  *
3365  * The dormant state indicates that the relevant interface is not
3366  * actually in a condition to pass packets (i.e., it is not 'up') but is
3367  * in a "pending" state, waiting for some external event.  For "on-
3368  * demand" interfaces, this new state identifies the situation where the
3369  * interface is waiting for events to place it in the up state.
3370  */
3371 static inline void netif_dormant_on(struct net_device *dev)
3372 {
3373 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3374 		linkwatch_fire_event(dev);
3375 }
3376 
3377 /**
3378  *	netif_dormant_off - set device as not dormant.
3379  *	@dev: network device
3380  *
3381  * Device is not in dormant state.
3382  */
3383 static inline void netif_dormant_off(struct net_device *dev)
3384 {
3385 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3386 		linkwatch_fire_event(dev);
3387 }
3388 
3389 /**
3390  *	netif_dormant - test if carrier present
3391  *	@dev: network device
3392  *
3393  * Check if carrier is present on device
3394  */
3395 static inline bool netif_dormant(const struct net_device *dev)
3396 {
3397 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3398 }
3399 
3400 
3401 /**
3402  *	netif_oper_up - test if device is operational
3403  *	@dev: network device
3404  *
3405  * Check if carrier is operational
3406  */
3407 static inline bool netif_oper_up(const struct net_device *dev)
3408 {
3409 	return (dev->operstate == IF_OPER_UP ||
3410 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3411 }
3412 
3413 /**
3414  *	netif_device_present - is device available or removed
3415  *	@dev: network device
3416  *
3417  * Check if device has not been removed from system.
3418  */
3419 static inline bool netif_device_present(struct net_device *dev)
3420 {
3421 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3422 }
3423 
3424 void netif_device_detach(struct net_device *dev);
3425 
3426 void netif_device_attach(struct net_device *dev);
3427 
3428 /*
3429  * Network interface message level settings
3430  */
3431 
3432 enum {
3433 	NETIF_MSG_DRV		= 0x0001,
3434 	NETIF_MSG_PROBE		= 0x0002,
3435 	NETIF_MSG_LINK		= 0x0004,
3436 	NETIF_MSG_TIMER		= 0x0008,
3437 	NETIF_MSG_IFDOWN	= 0x0010,
3438 	NETIF_MSG_IFUP		= 0x0020,
3439 	NETIF_MSG_RX_ERR	= 0x0040,
3440 	NETIF_MSG_TX_ERR	= 0x0080,
3441 	NETIF_MSG_TX_QUEUED	= 0x0100,
3442 	NETIF_MSG_INTR		= 0x0200,
3443 	NETIF_MSG_TX_DONE	= 0x0400,
3444 	NETIF_MSG_RX_STATUS	= 0x0800,
3445 	NETIF_MSG_PKTDATA	= 0x1000,
3446 	NETIF_MSG_HW		= 0x2000,
3447 	NETIF_MSG_WOL		= 0x4000,
3448 };
3449 
3450 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3451 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3452 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3453 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3454 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3455 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3456 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3457 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3458 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3459 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3460 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3461 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3462 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3463 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3464 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3465 
3466 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3467 {
3468 	/* use default */
3469 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3470 		return default_msg_enable_bits;
3471 	if (debug_value == 0)	/* no output */
3472 		return 0;
3473 	/* set low N bits */
3474 	return (1 << debug_value) - 1;
3475 }
3476 
3477 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3478 {
3479 	spin_lock(&txq->_xmit_lock);
3480 	txq->xmit_lock_owner = cpu;
3481 }
3482 
3483 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3484 {
3485 	__acquire(&txq->_xmit_lock);
3486 	return true;
3487 }
3488 
3489 static inline void __netif_tx_release(struct netdev_queue *txq)
3490 {
3491 	__release(&txq->_xmit_lock);
3492 }
3493 
3494 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3495 {
3496 	spin_lock_bh(&txq->_xmit_lock);
3497 	txq->xmit_lock_owner = smp_processor_id();
3498 }
3499 
3500 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3501 {
3502 	bool ok = spin_trylock(&txq->_xmit_lock);
3503 	if (likely(ok))
3504 		txq->xmit_lock_owner = smp_processor_id();
3505 	return ok;
3506 }
3507 
3508 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3509 {
3510 	txq->xmit_lock_owner = -1;
3511 	spin_unlock(&txq->_xmit_lock);
3512 }
3513 
3514 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3515 {
3516 	txq->xmit_lock_owner = -1;
3517 	spin_unlock_bh(&txq->_xmit_lock);
3518 }
3519 
3520 static inline void txq_trans_update(struct netdev_queue *txq)
3521 {
3522 	if (txq->xmit_lock_owner != -1)
3523 		txq->trans_start = jiffies;
3524 }
3525 
3526 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3527 static inline void netif_trans_update(struct net_device *dev)
3528 {
3529 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3530 
3531 	if (txq->trans_start != jiffies)
3532 		txq->trans_start = jiffies;
3533 }
3534 
3535 /**
3536  *	netif_tx_lock - grab network device transmit lock
3537  *	@dev: network device
3538  *
3539  * Get network device transmit lock
3540  */
3541 static inline void netif_tx_lock(struct net_device *dev)
3542 {
3543 	unsigned int i;
3544 	int cpu;
3545 
3546 	spin_lock(&dev->tx_global_lock);
3547 	cpu = smp_processor_id();
3548 	for (i = 0; i < dev->num_tx_queues; i++) {
3549 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3550 
3551 		/* We are the only thread of execution doing a
3552 		 * freeze, but we have to grab the _xmit_lock in
3553 		 * order to synchronize with threads which are in
3554 		 * the ->hard_start_xmit() handler and already
3555 		 * checked the frozen bit.
3556 		 */
3557 		__netif_tx_lock(txq, cpu);
3558 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3559 		__netif_tx_unlock(txq);
3560 	}
3561 }
3562 
3563 static inline void netif_tx_lock_bh(struct net_device *dev)
3564 {
3565 	local_bh_disable();
3566 	netif_tx_lock(dev);
3567 }
3568 
3569 static inline void netif_tx_unlock(struct net_device *dev)
3570 {
3571 	unsigned int i;
3572 
3573 	for (i = 0; i < dev->num_tx_queues; i++) {
3574 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3575 
3576 		/* No need to grab the _xmit_lock here.  If the
3577 		 * queue is not stopped for another reason, we
3578 		 * force a schedule.
3579 		 */
3580 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3581 		netif_schedule_queue(txq);
3582 	}
3583 	spin_unlock(&dev->tx_global_lock);
3584 }
3585 
3586 static inline void netif_tx_unlock_bh(struct net_device *dev)
3587 {
3588 	netif_tx_unlock(dev);
3589 	local_bh_enable();
3590 }
3591 
3592 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3593 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3594 		__netif_tx_lock(txq, cpu);		\
3595 	} else {					\
3596 		__netif_tx_acquire(txq);		\
3597 	}						\
3598 }
3599 
3600 #define HARD_TX_TRYLOCK(dev, txq)			\
3601 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3602 		__netif_tx_trylock(txq) :		\
3603 		__netif_tx_acquire(txq))
3604 
3605 #define HARD_TX_UNLOCK(dev, txq) {			\
3606 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3607 		__netif_tx_unlock(txq);			\
3608 	} else {					\
3609 		__netif_tx_release(txq);		\
3610 	}						\
3611 }
3612 
3613 static inline void netif_tx_disable(struct net_device *dev)
3614 {
3615 	unsigned int i;
3616 	int cpu;
3617 
3618 	local_bh_disable();
3619 	cpu = smp_processor_id();
3620 	for (i = 0; i < dev->num_tx_queues; i++) {
3621 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3622 
3623 		__netif_tx_lock(txq, cpu);
3624 		netif_tx_stop_queue(txq);
3625 		__netif_tx_unlock(txq);
3626 	}
3627 	local_bh_enable();
3628 }
3629 
3630 static inline void netif_addr_lock(struct net_device *dev)
3631 {
3632 	spin_lock(&dev->addr_list_lock);
3633 }
3634 
3635 static inline void netif_addr_lock_nested(struct net_device *dev)
3636 {
3637 	int subclass = SINGLE_DEPTH_NESTING;
3638 
3639 	if (dev->netdev_ops->ndo_get_lock_subclass)
3640 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3641 
3642 	spin_lock_nested(&dev->addr_list_lock, subclass);
3643 }
3644 
3645 static inline void netif_addr_lock_bh(struct net_device *dev)
3646 {
3647 	spin_lock_bh(&dev->addr_list_lock);
3648 }
3649 
3650 static inline void netif_addr_unlock(struct net_device *dev)
3651 {
3652 	spin_unlock(&dev->addr_list_lock);
3653 }
3654 
3655 static inline void netif_addr_unlock_bh(struct net_device *dev)
3656 {
3657 	spin_unlock_bh(&dev->addr_list_lock);
3658 }
3659 
3660 /*
3661  * dev_addrs walker. Should be used only for read access. Call with
3662  * rcu_read_lock held.
3663  */
3664 #define for_each_dev_addr(dev, ha) \
3665 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3666 
3667 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3668 
3669 void ether_setup(struct net_device *dev);
3670 
3671 /* Support for loadable net-drivers */
3672 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3673 				    unsigned char name_assign_type,
3674 				    void (*setup)(struct net_device *),
3675 				    unsigned int txqs, unsigned int rxqs);
3676 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3677 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3678 
3679 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3680 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3681 			 count)
3682 
3683 int register_netdev(struct net_device *dev);
3684 void unregister_netdev(struct net_device *dev);
3685 
3686 /* General hardware address lists handling functions */
3687 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3688 		   struct netdev_hw_addr_list *from_list, int addr_len);
3689 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3690 		      struct netdev_hw_addr_list *from_list, int addr_len);
3691 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3692 		       struct net_device *dev,
3693 		       int (*sync)(struct net_device *, const unsigned char *),
3694 		       int (*unsync)(struct net_device *,
3695 				     const unsigned char *));
3696 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3697 			  struct net_device *dev,
3698 			  int (*unsync)(struct net_device *,
3699 					const unsigned char *));
3700 void __hw_addr_init(struct netdev_hw_addr_list *list);
3701 
3702 /* Functions used for device addresses handling */
3703 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3704 		 unsigned char addr_type);
3705 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3706 		 unsigned char addr_type);
3707 void dev_addr_flush(struct net_device *dev);
3708 int dev_addr_init(struct net_device *dev);
3709 
3710 /* Functions used for unicast addresses handling */
3711 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3712 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3713 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3714 int dev_uc_sync(struct net_device *to, struct net_device *from);
3715 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3716 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3717 void dev_uc_flush(struct net_device *dev);
3718 void dev_uc_init(struct net_device *dev);
3719 
3720 /**
3721  *  __dev_uc_sync - Synchonize device's unicast list
3722  *  @dev:  device to sync
3723  *  @sync: function to call if address should be added
3724  *  @unsync: function to call if address should be removed
3725  *
3726  *  Add newly added addresses to the interface, and release
3727  *  addresses that have been deleted.
3728  */
3729 static inline int __dev_uc_sync(struct net_device *dev,
3730 				int (*sync)(struct net_device *,
3731 					    const unsigned char *),
3732 				int (*unsync)(struct net_device *,
3733 					      const unsigned char *))
3734 {
3735 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3736 }
3737 
3738 /**
3739  *  __dev_uc_unsync - Remove synchronized addresses from device
3740  *  @dev:  device to sync
3741  *  @unsync: function to call if address should be removed
3742  *
3743  *  Remove all addresses that were added to the device by dev_uc_sync().
3744  */
3745 static inline void __dev_uc_unsync(struct net_device *dev,
3746 				   int (*unsync)(struct net_device *,
3747 						 const unsigned char *))
3748 {
3749 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3750 }
3751 
3752 /* Functions used for multicast addresses handling */
3753 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3754 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3755 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3756 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3757 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3758 int dev_mc_sync(struct net_device *to, struct net_device *from);
3759 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3760 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3761 void dev_mc_flush(struct net_device *dev);
3762 void dev_mc_init(struct net_device *dev);
3763 
3764 /**
3765  *  __dev_mc_sync - Synchonize device's multicast list
3766  *  @dev:  device to sync
3767  *  @sync: function to call if address should be added
3768  *  @unsync: function to call if address should be removed
3769  *
3770  *  Add newly added addresses to the interface, and release
3771  *  addresses that have been deleted.
3772  */
3773 static inline int __dev_mc_sync(struct net_device *dev,
3774 				int (*sync)(struct net_device *,
3775 					    const unsigned char *),
3776 				int (*unsync)(struct net_device *,
3777 					      const unsigned char *))
3778 {
3779 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3780 }
3781 
3782 /**
3783  *  __dev_mc_unsync - Remove synchronized addresses from device
3784  *  @dev:  device to sync
3785  *  @unsync: function to call if address should be removed
3786  *
3787  *  Remove all addresses that were added to the device by dev_mc_sync().
3788  */
3789 static inline void __dev_mc_unsync(struct net_device *dev,
3790 				   int (*unsync)(struct net_device *,
3791 						 const unsigned char *))
3792 {
3793 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3794 }
3795 
3796 /* Functions used for secondary unicast and multicast support */
3797 void dev_set_rx_mode(struct net_device *dev);
3798 void __dev_set_rx_mode(struct net_device *dev);
3799 int dev_set_promiscuity(struct net_device *dev, int inc);
3800 int dev_set_allmulti(struct net_device *dev, int inc);
3801 void netdev_state_change(struct net_device *dev);
3802 void netdev_notify_peers(struct net_device *dev);
3803 void netdev_features_change(struct net_device *dev);
3804 /* Load a device via the kmod */
3805 void dev_load(struct net *net, const char *name);
3806 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3807 					struct rtnl_link_stats64 *storage);
3808 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3809 			     const struct net_device_stats *netdev_stats);
3810 
3811 extern int		netdev_max_backlog;
3812 extern int		netdev_tstamp_prequeue;
3813 extern int		weight_p;
3814 extern int		dev_weight_rx_bias;
3815 extern int		dev_weight_tx_bias;
3816 extern int		dev_rx_weight;
3817 extern int		dev_tx_weight;
3818 
3819 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3820 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3821 						     struct list_head **iter);
3822 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3823 						     struct list_head **iter);
3824 
3825 /* iterate through upper list, must be called under RCU read lock */
3826 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3827 	for (iter = &(dev)->adj_list.upper, \
3828 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3829 	     updev; \
3830 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3831 
3832 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3833 				  int (*fn)(struct net_device *upper_dev,
3834 					    void *data),
3835 				  void *data);
3836 
3837 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3838 				  struct net_device *upper_dev);
3839 
3840 void *netdev_lower_get_next_private(struct net_device *dev,
3841 				    struct list_head **iter);
3842 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3843 					struct list_head **iter);
3844 
3845 #define netdev_for_each_lower_private(dev, priv, iter) \
3846 	for (iter = (dev)->adj_list.lower.next, \
3847 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3848 	     priv; \
3849 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3850 
3851 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3852 	for (iter = &(dev)->adj_list.lower, \
3853 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3854 	     priv; \
3855 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3856 
3857 void *netdev_lower_get_next(struct net_device *dev,
3858 				struct list_head **iter);
3859 
3860 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3861 	for (iter = (dev)->adj_list.lower.next, \
3862 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3863 	     ldev; \
3864 	     ldev = netdev_lower_get_next(dev, &(iter)))
3865 
3866 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3867 					     struct list_head **iter);
3868 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3869 						 struct list_head **iter);
3870 
3871 int netdev_walk_all_lower_dev(struct net_device *dev,
3872 			      int (*fn)(struct net_device *lower_dev,
3873 					void *data),
3874 			      void *data);
3875 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3876 				  int (*fn)(struct net_device *lower_dev,
3877 					    void *data),
3878 				  void *data);
3879 
3880 void *netdev_adjacent_get_private(struct list_head *adj_list);
3881 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3882 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3883 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3884 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3885 int netdev_master_upper_dev_link(struct net_device *dev,
3886 				 struct net_device *upper_dev,
3887 				 void *upper_priv, void *upper_info);
3888 void netdev_upper_dev_unlink(struct net_device *dev,
3889 			     struct net_device *upper_dev);
3890 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3891 void *netdev_lower_dev_get_private(struct net_device *dev,
3892 				   struct net_device *lower_dev);
3893 void netdev_lower_state_changed(struct net_device *lower_dev,
3894 				void *lower_state_info);
3895 int netdev_default_l2upper_neigh_construct(struct net_device *dev,
3896 					   struct neighbour *n);
3897 void netdev_default_l2upper_neigh_destroy(struct net_device *dev,
3898 					  struct neighbour *n);
3899 
3900 /* RSS keys are 40 or 52 bytes long */
3901 #define NETDEV_RSS_KEY_LEN 52
3902 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3903 void netdev_rss_key_fill(void *buffer, size_t len);
3904 
3905 int dev_get_nest_level(struct net_device *dev);
3906 int skb_checksum_help(struct sk_buff *skb);
3907 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3908 				  netdev_features_t features, bool tx_path);
3909 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3910 				    netdev_features_t features);
3911 
3912 struct netdev_bonding_info {
3913 	ifslave	slave;
3914 	ifbond	master;
3915 };
3916 
3917 struct netdev_notifier_bonding_info {
3918 	struct netdev_notifier_info info; /* must be first */
3919 	struct netdev_bonding_info  bonding_info;
3920 };
3921 
3922 void netdev_bonding_info_change(struct net_device *dev,
3923 				struct netdev_bonding_info *bonding_info);
3924 
3925 static inline
3926 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3927 {
3928 	return __skb_gso_segment(skb, features, true);
3929 }
3930 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3931 
3932 static inline bool can_checksum_protocol(netdev_features_t features,
3933 					 __be16 protocol)
3934 {
3935 	if (protocol == htons(ETH_P_FCOE))
3936 		return !!(features & NETIF_F_FCOE_CRC);
3937 
3938 	/* Assume this is an IP checksum (not SCTP CRC) */
3939 
3940 	if (features & NETIF_F_HW_CSUM) {
3941 		/* Can checksum everything */
3942 		return true;
3943 	}
3944 
3945 	switch (protocol) {
3946 	case htons(ETH_P_IP):
3947 		return !!(features & NETIF_F_IP_CSUM);
3948 	case htons(ETH_P_IPV6):
3949 		return !!(features & NETIF_F_IPV6_CSUM);
3950 	default:
3951 		return false;
3952 	}
3953 }
3954 
3955 #ifdef CONFIG_BUG
3956 void netdev_rx_csum_fault(struct net_device *dev);
3957 #else
3958 static inline void netdev_rx_csum_fault(struct net_device *dev)
3959 {
3960 }
3961 #endif
3962 /* rx skb timestamps */
3963 void net_enable_timestamp(void);
3964 void net_disable_timestamp(void);
3965 
3966 #ifdef CONFIG_PROC_FS
3967 int __init dev_proc_init(void);
3968 #else
3969 #define dev_proc_init() 0
3970 #endif
3971 
3972 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3973 					      struct sk_buff *skb, struct net_device *dev,
3974 					      bool more)
3975 {
3976 	skb->xmit_more = more ? 1 : 0;
3977 	return ops->ndo_start_xmit(skb, dev);
3978 }
3979 
3980 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3981 					    struct netdev_queue *txq, bool more)
3982 {
3983 	const struct net_device_ops *ops = dev->netdev_ops;
3984 	int rc;
3985 
3986 	rc = __netdev_start_xmit(ops, skb, dev, more);
3987 	if (rc == NETDEV_TX_OK)
3988 		txq_trans_update(txq);
3989 
3990 	return rc;
3991 }
3992 
3993 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3994 				const void *ns);
3995 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3996 				 const void *ns);
3997 
3998 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3999 {
4000 	return netdev_class_create_file_ns(class_attr, NULL);
4001 }
4002 
4003 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4004 {
4005 	netdev_class_remove_file_ns(class_attr, NULL);
4006 }
4007 
4008 extern struct kobj_ns_type_operations net_ns_type_operations;
4009 
4010 const char *netdev_drivername(const struct net_device *dev);
4011 
4012 void linkwatch_run_queue(void);
4013 
4014 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4015 							  netdev_features_t f2)
4016 {
4017 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4018 		if (f1 & NETIF_F_HW_CSUM)
4019 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4020 		else
4021 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4022 	}
4023 
4024 	return f1 & f2;
4025 }
4026 
4027 static inline netdev_features_t netdev_get_wanted_features(
4028 	struct net_device *dev)
4029 {
4030 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4031 }
4032 netdev_features_t netdev_increment_features(netdev_features_t all,
4033 	netdev_features_t one, netdev_features_t mask);
4034 
4035 /* Allow TSO being used on stacked device :
4036  * Performing the GSO segmentation before last device
4037  * is a performance improvement.
4038  */
4039 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4040 							netdev_features_t mask)
4041 {
4042 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4043 }
4044 
4045 int __netdev_update_features(struct net_device *dev);
4046 void netdev_update_features(struct net_device *dev);
4047 void netdev_change_features(struct net_device *dev);
4048 
4049 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4050 					struct net_device *dev);
4051 
4052 netdev_features_t passthru_features_check(struct sk_buff *skb,
4053 					  struct net_device *dev,
4054 					  netdev_features_t features);
4055 netdev_features_t netif_skb_features(struct sk_buff *skb);
4056 
4057 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4058 {
4059 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4060 
4061 	/* check flags correspondence */
4062 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4063 	BUILD_BUG_ON(SKB_GSO_UDP     != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4064 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4065 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4066 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4067 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4068 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4069 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4070 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4071 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4072 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4073 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4074 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4075 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4076 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4077 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4078 
4079 	return (features & feature) == feature;
4080 }
4081 
4082 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4083 {
4084 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4085 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4086 }
4087 
4088 static inline bool netif_needs_gso(struct sk_buff *skb,
4089 				   netdev_features_t features)
4090 {
4091 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4092 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4093 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4094 }
4095 
4096 static inline void netif_set_gso_max_size(struct net_device *dev,
4097 					  unsigned int size)
4098 {
4099 	dev->gso_max_size = size;
4100 }
4101 
4102 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4103 					int pulled_hlen, u16 mac_offset,
4104 					int mac_len)
4105 {
4106 	skb->protocol = protocol;
4107 	skb->encapsulation = 1;
4108 	skb_push(skb, pulled_hlen);
4109 	skb_reset_transport_header(skb);
4110 	skb->mac_header = mac_offset;
4111 	skb->network_header = skb->mac_header + mac_len;
4112 	skb->mac_len = mac_len;
4113 }
4114 
4115 static inline bool netif_is_macsec(const struct net_device *dev)
4116 {
4117 	return dev->priv_flags & IFF_MACSEC;
4118 }
4119 
4120 static inline bool netif_is_macvlan(const struct net_device *dev)
4121 {
4122 	return dev->priv_flags & IFF_MACVLAN;
4123 }
4124 
4125 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4126 {
4127 	return dev->priv_flags & IFF_MACVLAN_PORT;
4128 }
4129 
4130 static inline bool netif_is_ipvlan(const struct net_device *dev)
4131 {
4132 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4133 }
4134 
4135 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4136 {
4137 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4138 }
4139 
4140 static inline bool netif_is_bond_master(const struct net_device *dev)
4141 {
4142 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4143 }
4144 
4145 static inline bool netif_is_bond_slave(const struct net_device *dev)
4146 {
4147 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4148 }
4149 
4150 static inline bool netif_supports_nofcs(struct net_device *dev)
4151 {
4152 	return dev->priv_flags & IFF_SUPP_NOFCS;
4153 }
4154 
4155 static inline bool netif_is_l3_master(const struct net_device *dev)
4156 {
4157 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4158 }
4159 
4160 static inline bool netif_is_l3_slave(const struct net_device *dev)
4161 {
4162 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4163 }
4164 
4165 static inline bool netif_is_bridge_master(const struct net_device *dev)
4166 {
4167 	return dev->priv_flags & IFF_EBRIDGE;
4168 }
4169 
4170 static inline bool netif_is_bridge_port(const struct net_device *dev)
4171 {
4172 	return dev->priv_flags & IFF_BRIDGE_PORT;
4173 }
4174 
4175 static inline bool netif_is_ovs_master(const struct net_device *dev)
4176 {
4177 	return dev->priv_flags & IFF_OPENVSWITCH;
4178 }
4179 
4180 static inline bool netif_is_team_master(const struct net_device *dev)
4181 {
4182 	return dev->priv_flags & IFF_TEAM;
4183 }
4184 
4185 static inline bool netif_is_team_port(const struct net_device *dev)
4186 {
4187 	return dev->priv_flags & IFF_TEAM_PORT;
4188 }
4189 
4190 static inline bool netif_is_lag_master(const struct net_device *dev)
4191 {
4192 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4193 }
4194 
4195 static inline bool netif_is_lag_port(const struct net_device *dev)
4196 {
4197 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4198 }
4199 
4200 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4201 {
4202 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4203 }
4204 
4205 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4206 static inline void netif_keep_dst(struct net_device *dev)
4207 {
4208 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4209 }
4210 
4211 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4212 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4213 {
4214 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4215 	return dev->priv_flags & IFF_MACSEC;
4216 }
4217 
4218 extern struct pernet_operations __net_initdata loopback_net_ops;
4219 
4220 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4221 
4222 /* netdev_printk helpers, similar to dev_printk */
4223 
4224 static inline const char *netdev_name(const struct net_device *dev)
4225 {
4226 	if (!dev->name[0] || strchr(dev->name, '%'))
4227 		return "(unnamed net_device)";
4228 	return dev->name;
4229 }
4230 
4231 static inline const char *netdev_reg_state(const struct net_device *dev)
4232 {
4233 	switch (dev->reg_state) {
4234 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4235 	case NETREG_REGISTERED: return "";
4236 	case NETREG_UNREGISTERING: return " (unregistering)";
4237 	case NETREG_UNREGISTERED: return " (unregistered)";
4238 	case NETREG_RELEASED: return " (released)";
4239 	case NETREG_DUMMY: return " (dummy)";
4240 	}
4241 
4242 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4243 	return " (unknown)";
4244 }
4245 
4246 __printf(3, 4)
4247 void netdev_printk(const char *level, const struct net_device *dev,
4248 		   const char *format, ...);
4249 __printf(2, 3)
4250 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4251 __printf(2, 3)
4252 void netdev_alert(const struct net_device *dev, const char *format, ...);
4253 __printf(2, 3)
4254 void netdev_crit(const struct net_device *dev, const char *format, ...);
4255 __printf(2, 3)
4256 void netdev_err(const struct net_device *dev, const char *format, ...);
4257 __printf(2, 3)
4258 void netdev_warn(const struct net_device *dev, const char *format, ...);
4259 __printf(2, 3)
4260 void netdev_notice(const struct net_device *dev, const char *format, ...);
4261 __printf(2, 3)
4262 void netdev_info(const struct net_device *dev, const char *format, ...);
4263 
4264 #define MODULE_ALIAS_NETDEV(device) \
4265 	MODULE_ALIAS("netdev-" device)
4266 
4267 #if defined(CONFIG_DYNAMIC_DEBUG)
4268 #define netdev_dbg(__dev, format, args...)			\
4269 do {								\
4270 	dynamic_netdev_dbg(__dev, format, ##args);		\
4271 } while (0)
4272 #elif defined(DEBUG)
4273 #define netdev_dbg(__dev, format, args...)			\
4274 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4275 #else
4276 #define netdev_dbg(__dev, format, args...)			\
4277 ({								\
4278 	if (0)							\
4279 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4280 })
4281 #endif
4282 
4283 #if defined(VERBOSE_DEBUG)
4284 #define netdev_vdbg	netdev_dbg
4285 #else
4286 
4287 #define netdev_vdbg(dev, format, args...)			\
4288 ({								\
4289 	if (0)							\
4290 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4291 	0;							\
4292 })
4293 #endif
4294 
4295 /*
4296  * netdev_WARN() acts like dev_printk(), but with the key difference
4297  * of using a WARN/WARN_ON to get the message out, including the
4298  * file/line information and a backtrace.
4299  */
4300 #define netdev_WARN(dev, format, args...)			\
4301 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4302 	     netdev_reg_state(dev), ##args)
4303 
4304 /* netif printk helpers, similar to netdev_printk */
4305 
4306 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4307 do {					  			\
4308 	if (netif_msg_##type(priv))				\
4309 		netdev_printk(level, (dev), fmt, ##args);	\
4310 } while (0)
4311 
4312 #define netif_level(level, priv, type, dev, fmt, args...)	\
4313 do {								\
4314 	if (netif_msg_##type(priv))				\
4315 		netdev_##level(dev, fmt, ##args);		\
4316 } while (0)
4317 
4318 #define netif_emerg(priv, type, dev, fmt, args...)		\
4319 	netif_level(emerg, priv, type, dev, fmt, ##args)
4320 #define netif_alert(priv, type, dev, fmt, args...)		\
4321 	netif_level(alert, priv, type, dev, fmt, ##args)
4322 #define netif_crit(priv, type, dev, fmt, args...)		\
4323 	netif_level(crit, priv, type, dev, fmt, ##args)
4324 #define netif_err(priv, type, dev, fmt, args...)		\
4325 	netif_level(err, priv, type, dev, fmt, ##args)
4326 #define netif_warn(priv, type, dev, fmt, args...)		\
4327 	netif_level(warn, priv, type, dev, fmt, ##args)
4328 #define netif_notice(priv, type, dev, fmt, args...)		\
4329 	netif_level(notice, priv, type, dev, fmt, ##args)
4330 #define netif_info(priv, type, dev, fmt, args...)		\
4331 	netif_level(info, priv, type, dev, fmt, ##args)
4332 
4333 #if defined(CONFIG_DYNAMIC_DEBUG)
4334 #define netif_dbg(priv, type, netdev, format, args...)		\
4335 do {								\
4336 	if (netif_msg_##type(priv))				\
4337 		dynamic_netdev_dbg(netdev, format, ##args);	\
4338 } while (0)
4339 #elif defined(DEBUG)
4340 #define netif_dbg(priv, type, dev, format, args...)		\
4341 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4342 #else
4343 #define netif_dbg(priv, type, dev, format, args...)			\
4344 ({									\
4345 	if (0)								\
4346 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4347 	0;								\
4348 })
4349 #endif
4350 
4351 /* if @cond then downgrade to debug, else print at @level */
4352 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4353 	do {                                                              \
4354 		if (cond)                                                 \
4355 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4356 		else                                                      \
4357 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4358 	} while (0)
4359 
4360 #if defined(VERBOSE_DEBUG)
4361 #define netif_vdbg	netif_dbg
4362 #else
4363 #define netif_vdbg(priv, type, dev, format, args...)		\
4364 ({								\
4365 	if (0)							\
4366 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4367 	0;							\
4368 })
4369 #endif
4370 
4371 /*
4372  *	The list of packet types we will receive (as opposed to discard)
4373  *	and the routines to invoke.
4374  *
4375  *	Why 16. Because with 16 the only overlap we get on a hash of the
4376  *	low nibble of the protocol value is RARP/SNAP/X.25.
4377  *
4378  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4379  *             sure which should go first, but I bet it won't make much
4380  *             difference if we are running VLANs.  The good news is that
4381  *             this protocol won't be in the list unless compiled in, so
4382  *             the average user (w/out VLANs) will not be adversely affected.
4383  *             --BLG
4384  *
4385  *		0800	IP
4386  *		8100    802.1Q VLAN
4387  *		0001	802.3
4388  *		0002	AX.25
4389  *		0004	802.2
4390  *		8035	RARP
4391  *		0005	SNAP
4392  *		0805	X.25
4393  *		0806	ARP
4394  *		8137	IPX
4395  *		0009	Localtalk
4396  *		86DD	IPv6
4397  */
4398 #define PTYPE_HASH_SIZE	(16)
4399 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4400 
4401 #endif	/* _LINUX_NETDEVICE_H */
4402