xref: /linux-6.15/include/linux/netdevice.h (revision 96c63fa7)
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 
56 struct netpoll_info;
57 struct device;
58 struct phy_device;
59 /* 802.11 specific */
60 struct wireless_dev;
61 /* 802.15.4 specific */
62 struct wpan_dev;
63 struct mpls_dev;
64 
65 void netdev_set_default_ethtool_ops(struct net_device *dev,
66 				    const struct ethtool_ops *ops);
67 
68 /* Backlog congestion levels */
69 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
70 #define NET_RX_DROP		1	/* packet dropped */
71 
72 /*
73  * Transmit return codes: transmit return codes originate from three different
74  * namespaces:
75  *
76  * - qdisc return codes
77  * - driver transmit return codes
78  * - errno values
79  *
80  * Drivers are allowed to return any one of those in their hard_start_xmit()
81  * function. Real network devices commonly used with qdiscs should only return
82  * the driver transmit return codes though - when qdiscs are used, the actual
83  * transmission happens asynchronously, so the value is not propagated to
84  * higher layers. Virtual network devices transmit synchronously; in this case
85  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
86  * others are propagated to higher layers.
87  */
88 
89 /* qdisc ->enqueue() return codes. */
90 #define NET_XMIT_SUCCESS	0x00
91 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
92 #define NET_XMIT_CN		0x02	/* congestion notification	*/
93 #define NET_XMIT_POLICED	0x03	/* skb is shot by police	*/
94 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
95 
96 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
97  * indicates that the device will soon be dropping packets, or already drops
98  * some packets of the same priority; prompting us to send less aggressively. */
99 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
100 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101 
102 /* Driver transmit return codes */
103 #define NETDEV_TX_MASK		0xf0
104 
105 enum netdev_tx {
106 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
107 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
108 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
109 };
110 typedef enum netdev_tx netdev_tx_t;
111 
112 /*
113  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
114  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
115  */
116 static inline bool dev_xmit_complete(int rc)
117 {
118 	/*
119 	 * Positive cases with an skb consumed by a driver:
120 	 * - successful transmission (rc == NETDEV_TX_OK)
121 	 * - error while transmitting (rc < 0)
122 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
123 	 */
124 	if (likely(rc < NET_XMIT_MASK))
125 		return true;
126 
127 	return false;
128 }
129 
130 /*
131  *	Compute the worst-case header length according to the protocols
132  *	used.
133  */
134 
135 #if defined(CONFIG_HYPERV_NET)
136 # define LL_MAX_HEADER 128
137 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
138 # if defined(CONFIG_MAC80211_MESH)
139 #  define LL_MAX_HEADER 128
140 # else
141 #  define LL_MAX_HEADER 96
142 # endif
143 #else
144 # define LL_MAX_HEADER 32
145 #endif
146 
147 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
148     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
149 #define MAX_HEADER LL_MAX_HEADER
150 #else
151 #define MAX_HEADER (LL_MAX_HEADER + 48)
152 #endif
153 
154 /*
155  *	Old network device statistics. Fields are native words
156  *	(unsigned long) so they can be read and written atomically.
157  */
158 
159 struct net_device_stats {
160 	unsigned long	rx_packets;
161 	unsigned long	tx_packets;
162 	unsigned long	rx_bytes;
163 	unsigned long	tx_bytes;
164 	unsigned long	rx_errors;
165 	unsigned long	tx_errors;
166 	unsigned long	rx_dropped;
167 	unsigned long	tx_dropped;
168 	unsigned long	multicast;
169 	unsigned long	collisions;
170 	unsigned long	rx_length_errors;
171 	unsigned long	rx_over_errors;
172 	unsigned long	rx_crc_errors;
173 	unsigned long	rx_frame_errors;
174 	unsigned long	rx_fifo_errors;
175 	unsigned long	rx_missed_errors;
176 	unsigned long	tx_aborted_errors;
177 	unsigned long	tx_carrier_errors;
178 	unsigned long	tx_fifo_errors;
179 	unsigned long	tx_heartbeat_errors;
180 	unsigned long	tx_window_errors;
181 	unsigned long	rx_compressed;
182 	unsigned long	tx_compressed;
183 };
184 
185 
186 #include <linux/cache.h>
187 #include <linux/skbuff.h>
188 
189 #ifdef CONFIG_RPS
190 #include <linux/static_key.h>
191 extern struct static_key rps_needed;
192 #endif
193 
194 struct neighbour;
195 struct neigh_parms;
196 struct sk_buff;
197 
198 struct netdev_hw_addr {
199 	struct list_head	list;
200 	unsigned char		addr[MAX_ADDR_LEN];
201 	unsigned char		type;
202 #define NETDEV_HW_ADDR_T_LAN		1
203 #define NETDEV_HW_ADDR_T_SAN		2
204 #define NETDEV_HW_ADDR_T_SLAVE		3
205 #define NETDEV_HW_ADDR_T_UNICAST	4
206 #define NETDEV_HW_ADDR_T_MULTICAST	5
207 	bool			global_use;
208 	int			sync_cnt;
209 	int			refcount;
210 	int			synced;
211 	struct rcu_head		rcu_head;
212 };
213 
214 struct netdev_hw_addr_list {
215 	struct list_head	list;
216 	int			count;
217 };
218 
219 #define netdev_hw_addr_list_count(l) ((l)->count)
220 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
221 #define netdev_hw_addr_list_for_each(ha, l) \
222 	list_for_each_entry(ha, &(l)->list, list)
223 
224 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
225 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
226 #define netdev_for_each_uc_addr(ha, dev) \
227 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
228 
229 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
230 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
231 #define netdev_for_each_mc_addr(ha, dev) \
232 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
233 
234 struct hh_cache {
235 	u16		hh_len;
236 	u16		__pad;
237 	seqlock_t	hh_lock;
238 
239 	/* cached hardware header; allow for machine alignment needs.        */
240 #define HH_DATA_MOD	16
241 #define HH_DATA_OFF(__len) \
242 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
243 #define HH_DATA_ALIGN(__len) \
244 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
245 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
246 };
247 
248 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
249  * Alternative is:
250  *   dev->hard_header_len ? (dev->hard_header_len +
251  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
252  *
253  * We could use other alignment values, but we must maintain the
254  * relationship HH alignment <= LL alignment.
255  */
256 #define LL_RESERVED_SPACE(dev) \
257 	((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
258 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
259 	((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 
261 struct header_ops {
262 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
263 			   unsigned short type, const void *daddr,
264 			   const void *saddr, unsigned int len);
265 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
266 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
267 	void	(*cache_update)(struct hh_cache *hh,
268 				const struct net_device *dev,
269 				const unsigned char *haddr);
270 	bool	(*validate)(const char *ll_header, unsigned int len);
271 };
272 
273 /* These flag bits are private to the generic network queueing
274  * layer; they may not be explicitly referenced by any other
275  * code.
276  */
277 
278 enum netdev_state_t {
279 	__LINK_STATE_START,
280 	__LINK_STATE_PRESENT,
281 	__LINK_STATE_NOCARRIER,
282 	__LINK_STATE_LINKWATCH_PENDING,
283 	__LINK_STATE_DORMANT,
284 };
285 
286 
287 /*
288  * This structure holds boot-time configured netdevice settings. They
289  * are then used in the device probing.
290  */
291 struct netdev_boot_setup {
292 	char name[IFNAMSIZ];
293 	struct ifmap map;
294 };
295 #define NETDEV_BOOT_SETUP_MAX 8
296 
297 int __init netdev_boot_setup(char *str);
298 
299 /*
300  * Structure for NAPI scheduling similar to tasklet but with weighting
301  */
302 struct napi_struct {
303 	/* The poll_list must only be managed by the entity which
304 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
305 	 * whoever atomically sets that bit can add this napi_struct
306 	 * to the per-CPU poll_list, and whoever clears that bit
307 	 * can remove from the list right before clearing the bit.
308 	 */
309 	struct list_head	poll_list;
310 
311 	unsigned long		state;
312 	int			weight;
313 	unsigned int		gro_count;
314 	int			(*poll)(struct napi_struct *, int);
315 #ifdef CONFIG_NETPOLL
316 	spinlock_t		poll_lock;
317 	int			poll_owner;
318 #endif
319 	struct net_device	*dev;
320 	struct sk_buff		*gro_list;
321 	struct sk_buff		*skb;
322 	struct hrtimer		timer;
323 	struct list_head	dev_list;
324 	struct hlist_node	napi_hash_node;
325 	unsigned int		napi_id;
326 };
327 
328 enum {
329 	NAPI_STATE_SCHED,	/* Poll is scheduled */
330 	NAPI_STATE_DISABLE,	/* Disable pending */
331 	NAPI_STATE_NPSVC,	/* Netpoll - don't dequeue from poll_list */
332 	NAPI_STATE_HASHED,	/* In NAPI hash (busy polling possible) */
333 	NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
334 };
335 
336 enum gro_result {
337 	GRO_MERGED,
338 	GRO_MERGED_FREE,
339 	GRO_HELD,
340 	GRO_NORMAL,
341 	GRO_DROP,
342 };
343 typedef enum gro_result gro_result_t;
344 
345 /*
346  * enum rx_handler_result - Possible return values for rx_handlers.
347  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
348  * further.
349  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
350  * case skb->dev was changed by rx_handler.
351  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
352  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
353  *
354  * rx_handlers are functions called from inside __netif_receive_skb(), to do
355  * special processing of the skb, prior to delivery to protocol handlers.
356  *
357  * Currently, a net_device can only have a single rx_handler registered. Trying
358  * to register a second rx_handler will return -EBUSY.
359  *
360  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
361  * To unregister a rx_handler on a net_device, use
362  * netdev_rx_handler_unregister().
363  *
364  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
365  * do with the skb.
366  *
367  * If the rx_handler consumed the skb in some way, it should return
368  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
369  * the skb to be delivered in some other way.
370  *
371  * If the rx_handler changed skb->dev, to divert the skb to another
372  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
373  * new device will be called if it exists.
374  *
375  * If the rx_handler decides the skb should be ignored, it should return
376  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
377  * are registered on exact device (ptype->dev == skb->dev).
378  *
379  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
380  * delivered, it should return RX_HANDLER_PASS.
381  *
382  * A device without a registered rx_handler will behave as if rx_handler
383  * returned RX_HANDLER_PASS.
384  */
385 
386 enum rx_handler_result {
387 	RX_HANDLER_CONSUMED,
388 	RX_HANDLER_ANOTHER,
389 	RX_HANDLER_EXACT,
390 	RX_HANDLER_PASS,
391 };
392 typedef enum rx_handler_result rx_handler_result_t;
393 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
394 
395 void __napi_schedule(struct napi_struct *n);
396 void __napi_schedule_irqoff(struct napi_struct *n);
397 
398 static inline bool napi_disable_pending(struct napi_struct *n)
399 {
400 	return test_bit(NAPI_STATE_DISABLE, &n->state);
401 }
402 
403 /**
404  *	napi_schedule_prep - check if NAPI can be scheduled
405  *	@n: NAPI context
406  *
407  * Test if NAPI routine is already running, and if not mark
408  * it as running.  This is used as a condition variable to
409  * insure only one NAPI poll instance runs.  We also make
410  * sure there is no pending NAPI disable.
411  */
412 static inline bool napi_schedule_prep(struct napi_struct *n)
413 {
414 	return !napi_disable_pending(n) &&
415 		!test_and_set_bit(NAPI_STATE_SCHED, &n->state);
416 }
417 
418 /**
419  *	napi_schedule - schedule NAPI poll
420  *	@n: NAPI context
421  *
422  * Schedule NAPI poll routine to be called if it is not already
423  * running.
424  */
425 static inline void napi_schedule(struct napi_struct *n)
426 {
427 	if (napi_schedule_prep(n))
428 		__napi_schedule(n);
429 }
430 
431 /**
432  *	napi_schedule_irqoff - schedule NAPI poll
433  *	@n: NAPI context
434  *
435  * Variant of napi_schedule(), assuming hard irqs are masked.
436  */
437 static inline void napi_schedule_irqoff(struct napi_struct *n)
438 {
439 	if (napi_schedule_prep(n))
440 		__napi_schedule_irqoff(n);
441 }
442 
443 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
444 static inline bool napi_reschedule(struct napi_struct *napi)
445 {
446 	if (napi_schedule_prep(napi)) {
447 		__napi_schedule(napi);
448 		return true;
449 	}
450 	return false;
451 }
452 
453 void __napi_complete(struct napi_struct *n);
454 void napi_complete_done(struct napi_struct *n, int work_done);
455 /**
456  *	napi_complete - NAPI processing complete
457  *	@n: NAPI context
458  *
459  * Mark NAPI processing as complete.
460  * Consider using napi_complete_done() instead.
461  */
462 static inline void napi_complete(struct napi_struct *n)
463 {
464 	return napi_complete_done(n, 0);
465 }
466 
467 /**
468  *	napi_hash_add - add a NAPI to global hashtable
469  *	@napi: NAPI context
470  *
471  * Generate a new napi_id and store a @napi under it in napi_hash.
472  * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
473  * Note: This is normally automatically done from netif_napi_add(),
474  * so might disappear in a future Linux version.
475  */
476 void napi_hash_add(struct napi_struct *napi);
477 
478 /**
479  *	napi_hash_del - remove a NAPI from global table
480  *	@napi: NAPI context
481  *
482  * Warning: caller must observe RCU grace period
483  * before freeing memory containing @napi, if
484  * this function returns true.
485  * Note: core networking stack automatically calls it
486  * from netif_napi_del().
487  * Drivers might want to call this helper to combine all
488  * the needed RCU grace periods into a single one.
489  */
490 bool napi_hash_del(struct napi_struct *napi);
491 
492 /**
493  *	napi_disable - prevent NAPI from scheduling
494  *	@n: NAPI context
495  *
496  * Stop NAPI from being scheduled on this context.
497  * Waits till any outstanding processing completes.
498  */
499 void napi_disable(struct napi_struct *n);
500 
501 /**
502  *	napi_enable - enable NAPI scheduling
503  *	@n: NAPI context
504  *
505  * Resume NAPI from being scheduled on this context.
506  * Must be paired with napi_disable.
507  */
508 static inline void napi_enable(struct napi_struct *n)
509 {
510 	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
511 	smp_mb__before_atomic();
512 	clear_bit(NAPI_STATE_SCHED, &n->state);
513 	clear_bit(NAPI_STATE_NPSVC, &n->state);
514 }
515 
516 /**
517  *	napi_synchronize - wait until NAPI is not running
518  *	@n: NAPI context
519  *
520  * Wait until NAPI is done being scheduled on this context.
521  * Waits till any outstanding processing completes but
522  * does not disable future activations.
523  */
524 static inline void napi_synchronize(const struct napi_struct *n)
525 {
526 	if (IS_ENABLED(CONFIG_SMP))
527 		while (test_bit(NAPI_STATE_SCHED, &n->state))
528 			msleep(1);
529 	else
530 		barrier();
531 }
532 
533 enum netdev_queue_state_t {
534 	__QUEUE_STATE_DRV_XOFF,
535 	__QUEUE_STATE_STACK_XOFF,
536 	__QUEUE_STATE_FROZEN,
537 };
538 
539 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
540 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
541 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
542 
543 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
544 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
545 					QUEUE_STATE_FROZEN)
546 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
547 					QUEUE_STATE_FROZEN)
548 
549 /*
550  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
551  * netif_tx_* functions below are used to manipulate this flag.  The
552  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
553  * queue independently.  The netif_xmit_*stopped functions below are called
554  * to check if the queue has been stopped by the driver or stack (either
555  * of the XOFF bits are set in the state).  Drivers should not need to call
556  * netif_xmit*stopped functions, they should only be using netif_tx_*.
557  */
558 
559 struct netdev_queue {
560 /*
561  * read-mostly part
562  */
563 	struct net_device	*dev;
564 	struct Qdisc __rcu	*qdisc;
565 	struct Qdisc		*qdisc_sleeping;
566 #ifdef CONFIG_SYSFS
567 	struct kobject		kobj;
568 #endif
569 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
570 	int			numa_node;
571 #endif
572 	unsigned long		tx_maxrate;
573 	/*
574 	 * Number of TX timeouts for this queue
575 	 * (/sys/class/net/DEV/Q/trans_timeout)
576 	 */
577 	unsigned long		trans_timeout;
578 /*
579  * write-mostly part
580  */
581 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
582 	int			xmit_lock_owner;
583 	/*
584 	 * Time (in jiffies) of last Tx
585 	 */
586 	unsigned long		trans_start;
587 
588 	unsigned long		state;
589 
590 #ifdef CONFIG_BQL
591 	struct dql		dql;
592 #endif
593 } ____cacheline_aligned_in_smp;
594 
595 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
596 {
597 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
598 	return q->numa_node;
599 #else
600 	return NUMA_NO_NODE;
601 #endif
602 }
603 
604 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
605 {
606 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
607 	q->numa_node = node;
608 #endif
609 }
610 
611 #ifdef CONFIG_RPS
612 /*
613  * This structure holds an RPS map which can be of variable length.  The
614  * map is an array of CPUs.
615  */
616 struct rps_map {
617 	unsigned int len;
618 	struct rcu_head rcu;
619 	u16 cpus[0];
620 };
621 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
622 
623 /*
624  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
625  * tail pointer for that CPU's input queue at the time of last enqueue, and
626  * a hardware filter index.
627  */
628 struct rps_dev_flow {
629 	u16 cpu;
630 	u16 filter;
631 	unsigned int last_qtail;
632 };
633 #define RPS_NO_FILTER 0xffff
634 
635 /*
636  * The rps_dev_flow_table structure contains a table of flow mappings.
637  */
638 struct rps_dev_flow_table {
639 	unsigned int mask;
640 	struct rcu_head rcu;
641 	struct rps_dev_flow flows[0];
642 };
643 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
644     ((_num) * sizeof(struct rps_dev_flow)))
645 
646 /*
647  * The rps_sock_flow_table contains mappings of flows to the last CPU
648  * on which they were processed by the application (set in recvmsg).
649  * Each entry is a 32bit value. Upper part is the high-order bits
650  * of flow hash, lower part is CPU number.
651  * rps_cpu_mask is used to partition the space, depending on number of
652  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
653  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
654  * meaning we use 32-6=26 bits for the hash.
655  */
656 struct rps_sock_flow_table {
657 	u32	mask;
658 
659 	u32	ents[0] ____cacheline_aligned_in_smp;
660 };
661 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
662 
663 #define RPS_NO_CPU 0xffff
664 
665 extern u32 rps_cpu_mask;
666 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
667 
668 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
669 					u32 hash)
670 {
671 	if (table && hash) {
672 		unsigned int index = hash & table->mask;
673 		u32 val = hash & ~rps_cpu_mask;
674 
675 		/* We only give a hint, preemption can change CPU under us */
676 		val |= raw_smp_processor_id();
677 
678 		if (table->ents[index] != val)
679 			table->ents[index] = val;
680 	}
681 }
682 
683 #ifdef CONFIG_RFS_ACCEL
684 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
685 			 u16 filter_id);
686 #endif
687 #endif /* CONFIG_RPS */
688 
689 /* This structure contains an instance of an RX queue. */
690 struct netdev_rx_queue {
691 #ifdef CONFIG_RPS
692 	struct rps_map __rcu		*rps_map;
693 	struct rps_dev_flow_table __rcu	*rps_flow_table;
694 #endif
695 	struct kobject			kobj;
696 	struct net_device		*dev;
697 } ____cacheline_aligned_in_smp;
698 
699 /*
700  * RX queue sysfs structures and functions.
701  */
702 struct rx_queue_attribute {
703 	struct attribute attr;
704 	ssize_t (*show)(struct netdev_rx_queue *queue,
705 	    struct rx_queue_attribute *attr, char *buf);
706 	ssize_t (*store)(struct netdev_rx_queue *queue,
707 	    struct rx_queue_attribute *attr, const char *buf, size_t len);
708 };
709 
710 #ifdef CONFIG_XPS
711 /*
712  * This structure holds an XPS map which can be of variable length.  The
713  * map is an array of queues.
714  */
715 struct xps_map {
716 	unsigned int len;
717 	unsigned int alloc_len;
718 	struct rcu_head rcu;
719 	u16 queues[0];
720 };
721 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
722 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
723        - sizeof(struct xps_map)) / sizeof(u16))
724 
725 /*
726  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
727  */
728 struct xps_dev_maps {
729 	struct rcu_head rcu;
730 	struct xps_map __rcu *cpu_map[0];
731 };
732 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) +		\
733     (nr_cpu_ids * sizeof(struct xps_map *)))
734 #endif /* CONFIG_XPS */
735 
736 #define TC_MAX_QUEUE	16
737 #define TC_BITMASK	15
738 /* HW offloaded queuing disciplines txq count and offset maps */
739 struct netdev_tc_txq {
740 	u16 count;
741 	u16 offset;
742 };
743 
744 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
745 /*
746  * This structure is to hold information about the device
747  * configured to run FCoE protocol stack.
748  */
749 struct netdev_fcoe_hbainfo {
750 	char	manufacturer[64];
751 	char	serial_number[64];
752 	char	hardware_version[64];
753 	char	driver_version[64];
754 	char	optionrom_version[64];
755 	char	firmware_version[64];
756 	char	model[256];
757 	char	model_description[256];
758 };
759 #endif
760 
761 #define MAX_PHYS_ITEM_ID_LEN 32
762 
763 /* This structure holds a unique identifier to identify some
764  * physical item (port for example) used by a netdevice.
765  */
766 struct netdev_phys_item_id {
767 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
768 	unsigned char id_len;
769 };
770 
771 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
772 					    struct netdev_phys_item_id *b)
773 {
774 	return a->id_len == b->id_len &&
775 	       memcmp(a->id, b->id, a->id_len) == 0;
776 }
777 
778 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
779 				       struct sk_buff *skb);
780 
781 /* These structures hold the attributes of qdisc and classifiers
782  * that are being passed to the netdevice through the setup_tc op.
783  */
784 enum {
785 	TC_SETUP_MQPRIO,
786 	TC_SETUP_CLSU32,
787 	TC_SETUP_CLSFLOWER,
788 };
789 
790 struct tc_cls_u32_offload;
791 
792 struct tc_to_netdev {
793 	unsigned int type;
794 	union {
795 		u8 tc;
796 		struct tc_cls_u32_offload *cls_u32;
797 		struct tc_cls_flower_offload *cls_flower;
798 	};
799 };
800 
801 
802 /*
803  * This structure defines the management hooks for network devices.
804  * The following hooks can be defined; unless noted otherwise, they are
805  * optional and can be filled with a null pointer.
806  *
807  * int (*ndo_init)(struct net_device *dev);
808  *     This function is called once when a network device is registered.
809  *     The network device can use this for any late stage initialization
810  *     or semantic validation. It can fail with an error code which will
811  *     be propagated back to register_netdev.
812  *
813  * void (*ndo_uninit)(struct net_device *dev);
814  *     This function is called when device is unregistered or when registration
815  *     fails. It is not called if init fails.
816  *
817  * int (*ndo_open)(struct net_device *dev);
818  *     This function is called when a network device transitions to the up
819  *     state.
820  *
821  * int (*ndo_stop)(struct net_device *dev);
822  *     This function is called when a network device transitions to the down
823  *     state.
824  *
825  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
826  *                               struct net_device *dev);
827  *	Called when a packet needs to be transmitted.
828  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
829  *	the queue before that can happen; it's for obsolete devices and weird
830  *	corner cases, but the stack really does a non-trivial amount
831  *	of useless work if you return NETDEV_TX_BUSY.
832  *	Required; cannot be NULL.
833  *
834  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
835  *		netdev_features_t features);
836  *	Adjusts the requested feature flags according to device-specific
837  *	constraints, and returns the resulting flags. Must not modify
838  *	the device state.
839  *
840  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
841  *                         void *accel_priv, select_queue_fallback_t fallback);
842  *	Called to decide which queue to use when device supports multiple
843  *	transmit queues.
844  *
845  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
846  *	This function is called to allow device receiver to make
847  *	changes to configuration when multicast or promiscuous is enabled.
848  *
849  * void (*ndo_set_rx_mode)(struct net_device *dev);
850  *	This function is called device changes address list filtering.
851  *	If driver handles unicast address filtering, it should set
852  *	IFF_UNICAST_FLT in its priv_flags.
853  *
854  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
855  *	This function  is called when the Media Access Control address
856  *	needs to be changed. If this interface is not defined, the
857  *	MAC address can not be changed.
858  *
859  * int (*ndo_validate_addr)(struct net_device *dev);
860  *	Test if Media Access Control address is valid for the device.
861  *
862  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
863  *	Called when a user requests an ioctl which can't be handled by
864  *	the generic interface code. If not defined ioctls return
865  *	not supported error code.
866  *
867  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
868  *	Used to set network devices bus interface parameters. This interface
869  *	is retained for legacy reasons; new devices should use the bus
870  *	interface (PCI) for low level management.
871  *
872  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
873  *	Called when a user wants to change the Maximum Transfer Unit
874  *	of a device. If not defined, any request to change MTU will
875  *	will return an error.
876  *
877  * void (*ndo_tx_timeout)(struct net_device *dev);
878  *	Callback used when the transmitter has not made any progress
879  *	for dev->watchdog ticks.
880  *
881  * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
882  *                      struct rtnl_link_stats64 *storage);
883  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
884  *	Called when a user wants to get the network device usage
885  *	statistics. Drivers must do one of the following:
886  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
887  *	   rtnl_link_stats64 structure passed by the caller.
888  *	2. Define @ndo_get_stats to update a net_device_stats structure
889  *	   (which should normally be dev->stats) and return a pointer to
890  *	   it. The structure may be changed asynchronously only if each
891  *	   field is written atomically.
892  *	3. Update dev->stats asynchronously and atomically, and define
893  *	   neither operation.
894  *
895  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
896  *	If device supports VLAN filtering this function is called when a
897  *	VLAN id is registered.
898  *
899  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
900  *	If device supports VLAN filtering this function is called when a
901  *	VLAN id is unregistered.
902  *
903  * void (*ndo_poll_controller)(struct net_device *dev);
904  *
905  *	SR-IOV management functions.
906  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
907  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
908  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
909  *			  int max_tx_rate);
910  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
911  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
912  * int (*ndo_get_vf_config)(struct net_device *dev,
913  *			    int vf, struct ifla_vf_info *ivf);
914  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
915  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
916  *			  struct nlattr *port[]);
917  *
918  *      Enable or disable the VF ability to query its RSS Redirection Table and
919  *      Hash Key. This is needed since on some devices VF share this information
920  *      with PF and querying it may introduce a theoretical security risk.
921  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
922  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
923  * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
924  * 	Called to setup 'tc' number of traffic classes in the net device. This
925  * 	is always called from the stack with the rtnl lock held and netif tx
926  * 	queues stopped. This allows the netdevice to perform queue management
927  * 	safely.
928  *
929  *	Fiber Channel over Ethernet (FCoE) offload functions.
930  * int (*ndo_fcoe_enable)(struct net_device *dev);
931  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
932  *	so the underlying device can perform whatever needed configuration or
933  *	initialization to support acceleration of FCoE traffic.
934  *
935  * int (*ndo_fcoe_disable)(struct net_device *dev);
936  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
937  *	so the underlying device can perform whatever needed clean-ups to
938  *	stop supporting acceleration of FCoE traffic.
939  *
940  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
941  *			     struct scatterlist *sgl, unsigned int sgc);
942  *	Called when the FCoE Initiator wants to initialize an I/O that
943  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
944  *	perform necessary setup and returns 1 to indicate the device is set up
945  *	successfully to perform DDP on this I/O, otherwise this returns 0.
946  *
947  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
948  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
949  *	indicated by the FC exchange id 'xid', so the underlying device can
950  *	clean up and reuse resources for later DDP requests.
951  *
952  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
953  *			      struct scatterlist *sgl, unsigned int sgc);
954  *	Called when the FCoE Target wants to initialize an I/O that
955  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
956  *	perform necessary setup and returns 1 to indicate the device is set up
957  *	successfully to perform DDP on this I/O, otherwise this returns 0.
958  *
959  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
960  *			       struct netdev_fcoe_hbainfo *hbainfo);
961  *	Called when the FCoE Protocol stack wants information on the underlying
962  *	device. This information is utilized by the FCoE protocol stack to
963  *	register attributes with Fiber Channel management service as per the
964  *	FC-GS Fabric Device Management Information(FDMI) specification.
965  *
966  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
967  *	Called when the underlying device wants to override default World Wide
968  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
969  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
970  *	protocol stack to use.
971  *
972  *	RFS acceleration.
973  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
974  *			    u16 rxq_index, u32 flow_id);
975  *	Set hardware filter for RFS.  rxq_index is the target queue index;
976  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
977  *	Return the filter ID on success, or a negative error code.
978  *
979  *	Slave management functions (for bridge, bonding, etc).
980  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
981  *	Called to make another netdev an underling.
982  *
983  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
984  *	Called to release previously enslaved netdev.
985  *
986  *      Feature/offload setting functions.
987  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
988  *	Called to update device configuration to new features. Passed
989  *	feature set might be less than what was returned by ndo_fix_features()).
990  *	Must return >0 or -errno if it changed dev->features itself.
991  *
992  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
993  *		      struct net_device *dev,
994  *		      const unsigned char *addr, u16 vid, u16 flags)
995  *	Adds an FDB entry to dev for addr.
996  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
997  *		      struct net_device *dev,
998  *		      const unsigned char *addr, u16 vid)
999  *	Deletes the FDB entry from dev coresponding to addr.
1000  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1001  *		       struct net_device *dev, struct net_device *filter_dev,
1002  *		       int idx)
1003  *	Used to add FDB entries to dump requests. Implementers should add
1004  *	entries to skb and update idx with the number of entries.
1005  *
1006  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1007  *			     u16 flags)
1008  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1009  *			     struct net_device *dev, u32 filter_mask,
1010  *			     int nlflags)
1011  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1012  *			     u16 flags);
1013  *
1014  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1015  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1016  *	which do not represent real hardware may define this to allow their
1017  *	userspace components to manage their virtual carrier state. Devices
1018  *	that determine carrier state from physical hardware properties (eg
1019  *	network cables) or protocol-dependent mechanisms (eg
1020  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1021  *
1022  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1023  *			       struct netdev_phys_item_id *ppid);
1024  *	Called to get ID of physical port of this device. If driver does
1025  *	not implement this, it is assumed that the hw is not able to have
1026  *	multiple net devices on single physical port.
1027  *
1028  * void (*ndo_add_vxlan_port)(struct  net_device *dev,
1029  *			      sa_family_t sa_family, __be16 port);
1030  *	Called by vxlan to notify a driver about the UDP port and socket
1031  *	address family that vxlan is listening to. It is called only when
1032  *	a new port starts listening. The operation is protected by the
1033  *	vxlan_net->sock_lock.
1034  *
1035  * void (*ndo_add_geneve_port)(struct net_device *dev,
1036  *			       sa_family_t sa_family, __be16 port);
1037  *	Called by geneve to notify a driver about the UDP port and socket
1038  *	address family that geneve is listnening to. It is called only when
1039  *	a new port starts listening. The operation is protected by the
1040  *	geneve_net->sock_lock.
1041  *
1042  * void (*ndo_del_geneve_port)(struct net_device *dev,
1043  *			       sa_family_t sa_family, __be16 port);
1044  *	Called by geneve to notify the driver about a UDP port and socket
1045  *	address family that geneve is not listening to anymore. The operation
1046  *	is protected by the geneve_net->sock_lock.
1047  *
1048  * void (*ndo_del_vxlan_port)(struct  net_device *dev,
1049  *			      sa_family_t sa_family, __be16 port);
1050  *	Called by vxlan to notify the driver about a UDP port and socket
1051  *	address family that vxlan is not listening to anymore. The operation
1052  *	is protected by the vxlan_net->sock_lock.
1053  *
1054  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1055  *				 struct net_device *dev)
1056  *	Called by upper layer devices to accelerate switching or other
1057  *	station functionality into hardware. 'pdev is the lowerdev
1058  *	to use for the offload and 'dev' is the net device that will
1059  *	back the offload. Returns a pointer to the private structure
1060  *	the upper layer will maintain.
1061  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1062  *	Called by upper layer device to delete the station created
1063  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1064  *	the station and priv is the structure returned by the add
1065  *	operation.
1066  * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1067  *				      struct net_device *dev,
1068  *				      void *priv);
1069  *	Callback to use for xmit over the accelerated station. This
1070  *	is used in place of ndo_start_xmit on accelerated net
1071  *	devices.
1072  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1073  *					   struct net_device *dev
1074  *					   netdev_features_t features);
1075  *	Called by core transmit path to determine if device is capable of
1076  *	performing offload operations on a given packet. This is to give
1077  *	the device an opportunity to implement any restrictions that cannot
1078  *	be otherwise expressed by feature flags. The check is called with
1079  *	the set of features that the stack has calculated and it returns
1080  *	those the driver believes to be appropriate.
1081  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1082  *			     int queue_index, u32 maxrate);
1083  *	Called when a user wants to set a max-rate limitation of specific
1084  *	TX queue.
1085  * int (*ndo_get_iflink)(const struct net_device *dev);
1086  *	Called to get the iflink value of this device.
1087  * void (*ndo_change_proto_down)(struct net_device *dev,
1088  *				 bool proto_down);
1089  *	This function is used to pass protocol port error state information
1090  *	to the switch driver. The switch driver can react to the proto_down
1091  *      by doing a phys down on the associated switch port.
1092  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1093  *	This function is used to get egress tunnel information for given skb.
1094  *	This is useful for retrieving outer tunnel header parameters while
1095  *	sampling packet.
1096  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1097  *	This function is used to specify the headroom that the skb must
1098  *	consider when allocation skb during packet reception. Setting
1099  *	appropriate rx headroom value allows avoiding skb head copy on
1100  *	forward. Setting a negative value resets the rx headroom to the
1101  *	default value.
1102  *
1103  */
1104 struct net_device_ops {
1105 	int			(*ndo_init)(struct net_device *dev);
1106 	void			(*ndo_uninit)(struct net_device *dev);
1107 	int			(*ndo_open)(struct net_device *dev);
1108 	int			(*ndo_stop)(struct net_device *dev);
1109 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1110 						  struct net_device *dev);
1111 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1112 						      struct net_device *dev,
1113 						      netdev_features_t features);
1114 	u16			(*ndo_select_queue)(struct net_device *dev,
1115 						    struct sk_buff *skb,
1116 						    void *accel_priv,
1117 						    select_queue_fallback_t fallback);
1118 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1119 						       int flags);
1120 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1121 	int			(*ndo_set_mac_address)(struct net_device *dev,
1122 						       void *addr);
1123 	int			(*ndo_validate_addr)(struct net_device *dev);
1124 	int			(*ndo_do_ioctl)(struct net_device *dev,
1125 					        struct ifreq *ifr, int cmd);
1126 	int			(*ndo_set_config)(struct net_device *dev,
1127 					          struct ifmap *map);
1128 	int			(*ndo_change_mtu)(struct net_device *dev,
1129 						  int new_mtu);
1130 	int			(*ndo_neigh_setup)(struct net_device *dev,
1131 						   struct neigh_parms *);
1132 	void			(*ndo_tx_timeout) (struct net_device *dev);
1133 
1134 	struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1135 						     struct rtnl_link_stats64 *storage);
1136 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1137 
1138 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1139 						       __be16 proto, u16 vid);
1140 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1141 						        __be16 proto, u16 vid);
1142 #ifdef CONFIG_NET_POLL_CONTROLLER
1143 	void                    (*ndo_poll_controller)(struct net_device *dev);
1144 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1145 						     struct netpoll_info *info);
1146 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1147 #endif
1148 #ifdef CONFIG_NET_RX_BUSY_POLL
1149 	int			(*ndo_busy_poll)(struct napi_struct *dev);
1150 #endif
1151 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1152 						  int queue, u8 *mac);
1153 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1154 						   int queue, u16 vlan, u8 qos);
1155 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1156 						   int vf, int min_tx_rate,
1157 						   int max_tx_rate);
1158 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1159 						       int vf, bool setting);
1160 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1161 						    int vf, bool setting);
1162 	int			(*ndo_get_vf_config)(struct net_device *dev,
1163 						     int vf,
1164 						     struct ifla_vf_info *ivf);
1165 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1166 							 int vf, int link_state);
1167 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1168 						    int vf,
1169 						    struct ifla_vf_stats
1170 						    *vf_stats);
1171 	int			(*ndo_set_vf_port)(struct net_device *dev,
1172 						   int vf,
1173 						   struct nlattr *port[]);
1174 	int			(*ndo_get_vf_port)(struct net_device *dev,
1175 						   int vf, struct sk_buff *skb);
1176 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1177 						   int vf, u64 guid,
1178 						   int guid_type);
1179 	int			(*ndo_set_vf_rss_query_en)(
1180 						   struct net_device *dev,
1181 						   int vf, bool setting);
1182 	int			(*ndo_setup_tc)(struct net_device *dev,
1183 						u32 handle,
1184 						__be16 protocol,
1185 						struct tc_to_netdev *tc);
1186 #if IS_ENABLED(CONFIG_FCOE)
1187 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1188 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1189 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1190 						      u16 xid,
1191 						      struct scatterlist *sgl,
1192 						      unsigned int sgc);
1193 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1194 						     u16 xid);
1195 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1196 						       u16 xid,
1197 						       struct scatterlist *sgl,
1198 						       unsigned int sgc);
1199 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1200 							struct netdev_fcoe_hbainfo *hbainfo);
1201 #endif
1202 
1203 #if IS_ENABLED(CONFIG_LIBFCOE)
1204 #define NETDEV_FCOE_WWNN 0
1205 #define NETDEV_FCOE_WWPN 1
1206 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1207 						    u64 *wwn, int type);
1208 #endif
1209 
1210 #ifdef CONFIG_RFS_ACCEL
1211 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1212 						     const struct sk_buff *skb,
1213 						     u16 rxq_index,
1214 						     u32 flow_id);
1215 #endif
1216 	int			(*ndo_add_slave)(struct net_device *dev,
1217 						 struct net_device *slave_dev);
1218 	int			(*ndo_del_slave)(struct net_device *dev,
1219 						 struct net_device *slave_dev);
1220 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1221 						    netdev_features_t features);
1222 	int			(*ndo_set_features)(struct net_device *dev,
1223 						    netdev_features_t features);
1224 	int			(*ndo_neigh_construct)(struct neighbour *n);
1225 	void			(*ndo_neigh_destroy)(struct neighbour *n);
1226 
1227 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1228 					       struct nlattr *tb[],
1229 					       struct net_device *dev,
1230 					       const unsigned char *addr,
1231 					       u16 vid,
1232 					       u16 flags);
1233 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1234 					       struct nlattr *tb[],
1235 					       struct net_device *dev,
1236 					       const unsigned char *addr,
1237 					       u16 vid);
1238 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1239 						struct netlink_callback *cb,
1240 						struct net_device *dev,
1241 						struct net_device *filter_dev,
1242 						int idx);
1243 
1244 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1245 						      struct nlmsghdr *nlh,
1246 						      u16 flags);
1247 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1248 						      u32 pid, u32 seq,
1249 						      struct net_device *dev,
1250 						      u32 filter_mask,
1251 						      int nlflags);
1252 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1253 						      struct nlmsghdr *nlh,
1254 						      u16 flags);
1255 	int			(*ndo_change_carrier)(struct net_device *dev,
1256 						      bool new_carrier);
1257 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1258 							struct netdev_phys_item_id *ppid);
1259 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1260 							  char *name, size_t len);
1261 	void			(*ndo_add_vxlan_port)(struct  net_device *dev,
1262 						      sa_family_t sa_family,
1263 						      __be16 port);
1264 	void			(*ndo_del_vxlan_port)(struct  net_device *dev,
1265 						      sa_family_t sa_family,
1266 						      __be16 port);
1267 	void			(*ndo_add_geneve_port)(struct  net_device *dev,
1268 						       sa_family_t sa_family,
1269 						       __be16 port);
1270 	void			(*ndo_del_geneve_port)(struct  net_device *dev,
1271 						       sa_family_t sa_family,
1272 						       __be16 port);
1273 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1274 							struct net_device *dev);
1275 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1276 							void *priv);
1277 
1278 	netdev_tx_t		(*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1279 							struct net_device *dev,
1280 							void *priv);
1281 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1282 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1283 						      int queue_index,
1284 						      u32 maxrate);
1285 	int			(*ndo_get_iflink)(const struct net_device *dev);
1286 	int			(*ndo_change_proto_down)(struct net_device *dev,
1287 							 bool proto_down);
1288 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1289 						       struct sk_buff *skb);
1290 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1291 						       int needed_headroom);
1292 };
1293 
1294 /**
1295  * enum net_device_priv_flags - &struct net_device priv_flags
1296  *
1297  * These are the &struct net_device, they are only set internally
1298  * by drivers and used in the kernel. These flags are invisible to
1299  * userspace; this means that the order of these flags can change
1300  * during any kernel release.
1301  *
1302  * You should have a pretty good reason to be extending these flags.
1303  *
1304  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1305  * @IFF_EBRIDGE: Ethernet bridging device
1306  * @IFF_BONDING: bonding master or slave
1307  * @IFF_ISATAP: ISATAP interface (RFC4214)
1308  * @IFF_WAN_HDLC: WAN HDLC device
1309  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1310  *	release skb->dst
1311  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1312  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1313  * @IFF_MACVLAN_PORT: device used as macvlan port
1314  * @IFF_BRIDGE_PORT: device used as bridge port
1315  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1316  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1317  * @IFF_UNICAST_FLT: Supports unicast filtering
1318  * @IFF_TEAM_PORT: device used as team port
1319  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1320  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1321  *	change when it's running
1322  * @IFF_MACVLAN: Macvlan device
1323  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1324  *	underlying stacked devices
1325  * @IFF_IPVLAN_MASTER: IPvlan master device
1326  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1327  * @IFF_L3MDEV_MASTER: device is an L3 master device
1328  * @IFF_NO_QUEUE: device can run without qdisc attached
1329  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1330  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1331  * @IFF_TEAM: device is a team device
1332  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1333  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1334  *	entity (i.e. the master device for bridged veth)
1335  * @IFF_MACSEC: device is a MACsec device
1336  */
1337 enum netdev_priv_flags {
1338 	IFF_802_1Q_VLAN			= 1<<0,
1339 	IFF_EBRIDGE			= 1<<1,
1340 	IFF_BONDING			= 1<<2,
1341 	IFF_ISATAP			= 1<<3,
1342 	IFF_WAN_HDLC			= 1<<4,
1343 	IFF_XMIT_DST_RELEASE		= 1<<5,
1344 	IFF_DONT_BRIDGE			= 1<<6,
1345 	IFF_DISABLE_NETPOLL		= 1<<7,
1346 	IFF_MACVLAN_PORT		= 1<<8,
1347 	IFF_BRIDGE_PORT			= 1<<9,
1348 	IFF_OVS_DATAPATH		= 1<<10,
1349 	IFF_TX_SKB_SHARING		= 1<<11,
1350 	IFF_UNICAST_FLT			= 1<<12,
1351 	IFF_TEAM_PORT			= 1<<13,
1352 	IFF_SUPP_NOFCS			= 1<<14,
1353 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1354 	IFF_MACVLAN			= 1<<16,
1355 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1356 	IFF_IPVLAN_MASTER		= 1<<18,
1357 	IFF_IPVLAN_SLAVE		= 1<<19,
1358 	IFF_L3MDEV_MASTER		= 1<<20,
1359 	IFF_NO_QUEUE			= 1<<21,
1360 	IFF_OPENVSWITCH			= 1<<22,
1361 	IFF_L3MDEV_SLAVE		= 1<<23,
1362 	IFF_TEAM			= 1<<24,
1363 	IFF_RXFH_CONFIGURED		= 1<<25,
1364 	IFF_PHONY_HEADROOM		= 1<<26,
1365 	IFF_MACSEC			= 1<<27,
1366 };
1367 
1368 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1369 #define IFF_EBRIDGE			IFF_EBRIDGE
1370 #define IFF_BONDING			IFF_BONDING
1371 #define IFF_ISATAP			IFF_ISATAP
1372 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1373 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1374 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1375 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1376 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1377 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1378 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1379 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1380 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1381 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1382 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1383 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1384 #define IFF_MACVLAN			IFF_MACVLAN
1385 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1386 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1387 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1388 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1389 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1390 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1391 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1392 #define IFF_TEAM			IFF_TEAM
1393 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1394 #define IFF_MACSEC			IFF_MACSEC
1395 
1396 /**
1397  *	struct net_device - The DEVICE structure.
1398  *		Actually, this whole structure is a big mistake.  It mixes I/O
1399  *		data with strictly "high-level" data, and it has to know about
1400  *		almost every data structure used in the INET module.
1401  *
1402  *	@name:	This is the first field of the "visible" part of this structure
1403  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1404  *	 	of the interface.
1405  *
1406  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1407  *	@ifalias:	SNMP alias
1408  *	@mem_end:	Shared memory end
1409  *	@mem_start:	Shared memory start
1410  *	@base_addr:	Device I/O address
1411  *	@irq:		Device IRQ number
1412  *
1413  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1414  *
1415  *	@state:		Generic network queuing layer state, see netdev_state_t
1416  *	@dev_list:	The global list of network devices
1417  *	@napi_list:	List entry used for polling NAPI devices
1418  *	@unreg_list:	List entry  when we are unregistering the
1419  *			device; see the function unregister_netdev
1420  *	@close_list:	List entry used when we are closing the device
1421  *	@ptype_all:     Device-specific packet handlers for all protocols
1422  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1423  *
1424  *	@adj_list:	Directly linked devices, like slaves for bonding
1425  *	@all_adj_list:	All linked devices, *including* neighbours
1426  *	@features:	Currently active device features
1427  *	@hw_features:	User-changeable features
1428  *
1429  *	@wanted_features:	User-requested features
1430  *	@vlan_features:		Mask of features inheritable by VLAN devices
1431  *
1432  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1433  *				This field indicates what encapsulation
1434  *				offloads the hardware is capable of doing,
1435  *				and drivers will need to set them appropriately.
1436  *
1437  *	@mpls_features:	Mask of features inheritable by MPLS
1438  *
1439  *	@ifindex:	interface index
1440  *	@group:		The group the device belongs to
1441  *
1442  *	@stats:		Statistics struct, which was left as a legacy, use
1443  *			rtnl_link_stats64 instead
1444  *
1445  *	@rx_dropped:	Dropped packets by core network,
1446  *			do not use this in drivers
1447  *	@tx_dropped:	Dropped packets by core network,
1448  *			do not use this in drivers
1449  *	@rx_nohandler:	nohandler dropped packets by core network on
1450  *			inactive devices, do not use this in drivers
1451  *
1452  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1453  *				instead of ioctl,
1454  *				see <net/iw_handler.h> for details.
1455  *	@wireless_data:	Instance data managed by the core of wireless extensions
1456  *
1457  *	@netdev_ops:	Includes several pointers to callbacks,
1458  *			if one wants to override the ndo_*() functions
1459  *	@ethtool_ops:	Management operations
1460  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1461  *			of Layer 2 headers.
1462  *
1463  *	@flags:		Interface flags (a la BSD)
1464  *	@priv_flags:	Like 'flags' but invisible to userspace,
1465  *			see if.h for the definitions
1466  *	@gflags:	Global flags ( kept as legacy )
1467  *	@padded:	How much padding added by alloc_netdev()
1468  *	@operstate:	RFC2863 operstate
1469  *	@link_mode:	Mapping policy to operstate
1470  *	@if_port:	Selectable AUI, TP, ...
1471  *	@dma:		DMA channel
1472  *	@mtu:		Interface MTU value
1473  *	@type:		Interface hardware type
1474  *	@hard_header_len: Maximum hardware header length.
1475  *
1476  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1477  *			  cases can this be guaranteed
1478  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1479  *			  cases can this be guaranteed. Some cases also use
1480  *			  LL_MAX_HEADER instead to allocate the skb
1481  *
1482  *	interface address info:
1483  *
1484  * 	@perm_addr:		Permanent hw address
1485  * 	@addr_assign_type:	Hw address assignment type
1486  * 	@addr_len:		Hardware address length
1487  * 	@neigh_priv_len;	Used in neigh_alloc(),
1488  * 				initialized only in atm/clip.c
1489  * 	@dev_id:		Used to differentiate devices that share
1490  * 				the same link layer address
1491  * 	@dev_port:		Used to differentiate devices that share
1492  * 				the same function
1493  *	@addr_list_lock:	XXX: need comments on this one
1494  *	@uc_promisc:		Counter that indicates promiscuous mode
1495  *				has been enabled due to the need to listen to
1496  *				additional unicast addresses in a device that
1497  *				does not implement ndo_set_rx_mode()
1498  *	@uc:			unicast mac addresses
1499  *	@mc:			multicast mac addresses
1500  *	@dev_addrs:		list of device hw addresses
1501  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1502  *	@promiscuity:		Number of times the NIC is told to work in
1503  *				promiscuous mode; if it becomes 0 the NIC will
1504  *				exit promiscuous mode
1505  *	@allmulti:		Counter, enables or disables allmulticast mode
1506  *
1507  *	@vlan_info:	VLAN info
1508  *	@dsa_ptr:	dsa specific data
1509  *	@tipc_ptr:	TIPC specific data
1510  *	@atalk_ptr:	AppleTalk link
1511  *	@ip_ptr:	IPv4 specific data
1512  *	@dn_ptr:	DECnet specific data
1513  *	@ip6_ptr:	IPv6 specific data
1514  *	@ax25_ptr:	AX.25 specific data
1515  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1516  *
1517  *	@last_rx:	Time of last Rx
1518  *	@dev_addr:	Hw address (before bcast,
1519  *			because most packets are unicast)
1520  *
1521  *	@_rx:			Array of RX queues
1522  *	@num_rx_queues:		Number of RX queues
1523  *				allocated at register_netdev() time
1524  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1525  *
1526  *	@rx_handler:		handler for received packets
1527  *	@rx_handler_data: 	XXX: need comments on this one
1528  *	@ingress_queue:		XXX: need comments on this one
1529  *	@broadcast:		hw bcast address
1530  *
1531  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1532  *			indexed by RX queue number. Assigned by driver.
1533  *			This must only be set if the ndo_rx_flow_steer
1534  *			operation is defined
1535  *	@index_hlist:		Device index hash chain
1536  *
1537  *	@_tx:			Array of TX queues
1538  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1539  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1540  *	@qdisc:			Root qdisc from userspace point of view
1541  *	@tx_queue_len:		Max frames per queue allowed
1542  *	@tx_global_lock: 	XXX: need comments on this one
1543  *
1544  *	@xps_maps:	XXX: need comments on this one
1545  *
1546  *	@offload_fwd_mark:	Offload device fwding mark
1547  *
1548  *	@watchdog_timeo:	Represents the timeout that is used by
1549  *				the watchdog (see dev_watchdog())
1550  *	@watchdog_timer:	List of timers
1551  *
1552  *	@pcpu_refcnt:		Number of references to this device
1553  *	@todo_list:		Delayed register/unregister
1554  *	@link_watch_list:	XXX: need comments on this one
1555  *
1556  *	@reg_state:		Register/unregister state machine
1557  *	@dismantle:		Device is going to be freed
1558  *	@rtnl_link_state:	This enum represents the phases of creating
1559  *				a new link
1560  *
1561  *	@destructor:		Called from unregister,
1562  *				can be used to call free_netdev
1563  *	@npinfo:		XXX: need comments on this one
1564  * 	@nd_net:		Network namespace this network device is inside
1565  *
1566  * 	@ml_priv:	Mid-layer private
1567  * 	@lstats:	Loopback statistics
1568  * 	@tstats:	Tunnel statistics
1569  * 	@dstats:	Dummy statistics
1570  * 	@vstats:	Virtual ethernet statistics
1571  *
1572  *	@garp_port:	GARP
1573  *	@mrp_port:	MRP
1574  *
1575  *	@dev:		Class/net/name entry
1576  *	@sysfs_groups:	Space for optional device, statistics and wireless
1577  *			sysfs groups
1578  *
1579  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1580  *	@rtnl_link_ops:	Rtnl_link_ops
1581  *
1582  *	@gso_max_size:	Maximum size of generic segmentation offload
1583  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1584  *			NIC for GSO
1585  *
1586  *	@dcbnl_ops:	Data Center Bridging netlink ops
1587  *	@num_tc:	Number of traffic classes in the net device
1588  *	@tc_to_txq:	XXX: need comments on this one
1589  *	@prio_tc_map	XXX: need comments on this one
1590  *
1591  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1592  *
1593  *	@priomap:	XXX: need comments on this one
1594  *	@phydev:	Physical device may attach itself
1595  *			for hardware timestamping
1596  *
1597  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1598  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1599  *
1600  *	@proto_down:	protocol port state information can be sent to the
1601  *			switch driver and used to set the phys state of the
1602  *			switch port.
1603  *
1604  *	FIXME: cleanup struct net_device such that network protocol info
1605  *	moves out.
1606  */
1607 
1608 struct net_device {
1609 	char			name[IFNAMSIZ];
1610 	struct hlist_node	name_hlist;
1611 	char 			*ifalias;
1612 	/*
1613 	 *	I/O specific fields
1614 	 *	FIXME: Merge these and struct ifmap into one
1615 	 */
1616 	unsigned long		mem_end;
1617 	unsigned long		mem_start;
1618 	unsigned long		base_addr;
1619 	int			irq;
1620 
1621 	atomic_t		carrier_changes;
1622 
1623 	/*
1624 	 *	Some hardware also needs these fields (state,dev_list,
1625 	 *	napi_list,unreg_list,close_list) but they are not
1626 	 *	part of the usual set specified in Space.c.
1627 	 */
1628 
1629 	unsigned long		state;
1630 
1631 	struct list_head	dev_list;
1632 	struct list_head	napi_list;
1633 	struct list_head	unreg_list;
1634 	struct list_head	close_list;
1635 	struct list_head	ptype_all;
1636 	struct list_head	ptype_specific;
1637 
1638 	struct {
1639 		struct list_head upper;
1640 		struct list_head lower;
1641 	} adj_list;
1642 
1643 	struct {
1644 		struct list_head upper;
1645 		struct list_head lower;
1646 	} all_adj_list;
1647 
1648 	netdev_features_t	features;
1649 	netdev_features_t	hw_features;
1650 	netdev_features_t	wanted_features;
1651 	netdev_features_t	vlan_features;
1652 	netdev_features_t	hw_enc_features;
1653 	netdev_features_t	mpls_features;
1654 	netdev_features_t	gso_partial_features;
1655 
1656 	int			ifindex;
1657 	int			group;
1658 
1659 	struct net_device_stats	stats;
1660 
1661 	atomic_long_t		rx_dropped;
1662 	atomic_long_t		tx_dropped;
1663 	atomic_long_t		rx_nohandler;
1664 
1665 #ifdef CONFIG_WIRELESS_EXT
1666 	const struct iw_handler_def *wireless_handlers;
1667 	struct iw_public_data	*wireless_data;
1668 #endif
1669 	const struct net_device_ops *netdev_ops;
1670 	const struct ethtool_ops *ethtool_ops;
1671 #ifdef CONFIG_NET_SWITCHDEV
1672 	const struct switchdev_ops *switchdev_ops;
1673 #endif
1674 #ifdef CONFIG_NET_L3_MASTER_DEV
1675 	const struct l3mdev_ops	*l3mdev_ops;
1676 #endif
1677 
1678 	const struct header_ops *header_ops;
1679 
1680 	unsigned int		flags;
1681 	unsigned int		priv_flags;
1682 
1683 	unsigned short		gflags;
1684 	unsigned short		padded;
1685 
1686 	unsigned char		operstate;
1687 	unsigned char		link_mode;
1688 
1689 	unsigned char		if_port;
1690 	unsigned char		dma;
1691 
1692 	unsigned int		mtu;
1693 	unsigned short		type;
1694 	unsigned short		hard_header_len;
1695 
1696 	unsigned short		needed_headroom;
1697 	unsigned short		needed_tailroom;
1698 
1699 	/* Interface address info. */
1700 	unsigned char		perm_addr[MAX_ADDR_LEN];
1701 	unsigned char		addr_assign_type;
1702 	unsigned char		addr_len;
1703 	unsigned short		neigh_priv_len;
1704 	unsigned short          dev_id;
1705 	unsigned short          dev_port;
1706 	spinlock_t		addr_list_lock;
1707 	unsigned char		name_assign_type;
1708 	bool			uc_promisc;
1709 	struct netdev_hw_addr_list	uc;
1710 	struct netdev_hw_addr_list	mc;
1711 	struct netdev_hw_addr_list	dev_addrs;
1712 
1713 #ifdef CONFIG_SYSFS
1714 	struct kset		*queues_kset;
1715 #endif
1716 	unsigned int		promiscuity;
1717 	unsigned int		allmulti;
1718 
1719 
1720 	/* Protocol-specific pointers */
1721 
1722 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1723 	struct vlan_info __rcu	*vlan_info;
1724 #endif
1725 #if IS_ENABLED(CONFIG_NET_DSA)
1726 	struct dsa_switch_tree	*dsa_ptr;
1727 #endif
1728 #if IS_ENABLED(CONFIG_TIPC)
1729 	struct tipc_bearer __rcu *tipc_ptr;
1730 #endif
1731 	void 			*atalk_ptr;
1732 	struct in_device __rcu	*ip_ptr;
1733 	struct dn_dev __rcu     *dn_ptr;
1734 	struct inet6_dev __rcu	*ip6_ptr;
1735 	void			*ax25_ptr;
1736 	struct wireless_dev	*ieee80211_ptr;
1737 	struct wpan_dev		*ieee802154_ptr;
1738 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1739 	struct mpls_dev __rcu	*mpls_ptr;
1740 #endif
1741 
1742 /*
1743  * Cache lines mostly used on receive path (including eth_type_trans())
1744  */
1745 	unsigned long		last_rx;
1746 
1747 	/* Interface address info used in eth_type_trans() */
1748 	unsigned char		*dev_addr;
1749 
1750 #ifdef CONFIG_SYSFS
1751 	struct netdev_rx_queue	*_rx;
1752 
1753 	unsigned int		num_rx_queues;
1754 	unsigned int		real_num_rx_queues;
1755 #endif
1756 
1757 	unsigned long		gro_flush_timeout;
1758 	rx_handler_func_t __rcu	*rx_handler;
1759 	void __rcu		*rx_handler_data;
1760 
1761 #ifdef CONFIG_NET_CLS_ACT
1762 	struct tcf_proto __rcu  *ingress_cl_list;
1763 #endif
1764 	struct netdev_queue __rcu *ingress_queue;
1765 #ifdef CONFIG_NETFILTER_INGRESS
1766 	struct list_head	nf_hooks_ingress;
1767 #endif
1768 
1769 	unsigned char		broadcast[MAX_ADDR_LEN];
1770 #ifdef CONFIG_RFS_ACCEL
1771 	struct cpu_rmap		*rx_cpu_rmap;
1772 #endif
1773 	struct hlist_node	index_hlist;
1774 
1775 /*
1776  * Cache lines mostly used on transmit path
1777  */
1778 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1779 	unsigned int		num_tx_queues;
1780 	unsigned int		real_num_tx_queues;
1781 	struct Qdisc		*qdisc;
1782 	unsigned long		tx_queue_len;
1783 	spinlock_t		tx_global_lock;
1784 	int			watchdog_timeo;
1785 
1786 #ifdef CONFIG_XPS
1787 	struct xps_dev_maps __rcu *xps_maps;
1788 #endif
1789 #ifdef CONFIG_NET_CLS_ACT
1790 	struct tcf_proto __rcu  *egress_cl_list;
1791 #endif
1792 #ifdef CONFIG_NET_SWITCHDEV
1793 	u32			offload_fwd_mark;
1794 #endif
1795 
1796 	/* These may be needed for future network-power-down code. */
1797 	struct timer_list	watchdog_timer;
1798 
1799 	int __percpu		*pcpu_refcnt;
1800 	struct list_head	todo_list;
1801 
1802 	struct list_head	link_watch_list;
1803 
1804 	enum { NETREG_UNINITIALIZED=0,
1805 	       NETREG_REGISTERED,	/* completed register_netdevice */
1806 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1807 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1808 	       NETREG_RELEASED,		/* called free_netdev */
1809 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1810 	} reg_state:8;
1811 
1812 	bool dismantle;
1813 
1814 	enum {
1815 		RTNL_LINK_INITIALIZED,
1816 		RTNL_LINK_INITIALIZING,
1817 	} rtnl_link_state:16;
1818 
1819 	void (*destructor)(struct net_device *dev);
1820 
1821 #ifdef CONFIG_NETPOLL
1822 	struct netpoll_info __rcu	*npinfo;
1823 #endif
1824 
1825 	possible_net_t			nd_net;
1826 
1827 	/* mid-layer private */
1828 	union {
1829 		void					*ml_priv;
1830 		struct pcpu_lstats __percpu		*lstats;
1831 		struct pcpu_sw_netstats __percpu	*tstats;
1832 		struct pcpu_dstats __percpu		*dstats;
1833 		struct pcpu_vstats __percpu		*vstats;
1834 	};
1835 
1836 	struct garp_port __rcu	*garp_port;
1837 	struct mrp_port __rcu	*mrp_port;
1838 
1839 	struct device		dev;
1840 	const struct attribute_group *sysfs_groups[4];
1841 	const struct attribute_group *sysfs_rx_queue_group;
1842 
1843 	const struct rtnl_link_ops *rtnl_link_ops;
1844 
1845 	/* for setting kernel sock attribute on TCP connection setup */
1846 #define GSO_MAX_SIZE		65536
1847 	unsigned int		gso_max_size;
1848 #define GSO_MAX_SEGS		65535
1849 	u16			gso_max_segs;
1850 
1851 #ifdef CONFIG_DCB
1852 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1853 #endif
1854 	u8			num_tc;
1855 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1856 	u8			prio_tc_map[TC_BITMASK + 1];
1857 
1858 #if IS_ENABLED(CONFIG_FCOE)
1859 	unsigned int		fcoe_ddp_xid;
1860 #endif
1861 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1862 	struct netprio_map __rcu *priomap;
1863 #endif
1864 	struct phy_device	*phydev;
1865 	struct lock_class_key	*qdisc_tx_busylock;
1866 	struct lock_class_key	*qdisc_running_key;
1867 	bool			proto_down;
1868 };
1869 #define to_net_dev(d) container_of(d, struct net_device, dev)
1870 
1871 #define	NETDEV_ALIGN		32
1872 
1873 static inline
1874 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1875 {
1876 	return dev->prio_tc_map[prio & TC_BITMASK];
1877 }
1878 
1879 static inline
1880 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1881 {
1882 	if (tc >= dev->num_tc)
1883 		return -EINVAL;
1884 
1885 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1886 	return 0;
1887 }
1888 
1889 static inline
1890 void netdev_reset_tc(struct net_device *dev)
1891 {
1892 	dev->num_tc = 0;
1893 	memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1894 	memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1895 }
1896 
1897 static inline
1898 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1899 {
1900 	if (tc >= dev->num_tc)
1901 		return -EINVAL;
1902 
1903 	dev->tc_to_txq[tc].count = count;
1904 	dev->tc_to_txq[tc].offset = offset;
1905 	return 0;
1906 }
1907 
1908 static inline
1909 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1910 {
1911 	if (num_tc > TC_MAX_QUEUE)
1912 		return -EINVAL;
1913 
1914 	dev->num_tc = num_tc;
1915 	return 0;
1916 }
1917 
1918 static inline
1919 int netdev_get_num_tc(struct net_device *dev)
1920 {
1921 	return dev->num_tc;
1922 }
1923 
1924 static inline
1925 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1926 					 unsigned int index)
1927 {
1928 	return &dev->_tx[index];
1929 }
1930 
1931 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1932 						    const struct sk_buff *skb)
1933 {
1934 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1935 }
1936 
1937 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1938 					    void (*f)(struct net_device *,
1939 						      struct netdev_queue *,
1940 						      void *),
1941 					    void *arg)
1942 {
1943 	unsigned int i;
1944 
1945 	for (i = 0; i < dev->num_tx_queues; i++)
1946 		f(dev, &dev->_tx[i], arg);
1947 }
1948 
1949 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1950 				    struct sk_buff *skb,
1951 				    void *accel_priv);
1952 
1953 /* returns the headroom that the master device needs to take in account
1954  * when forwarding to this dev
1955  */
1956 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1957 {
1958 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1959 }
1960 
1961 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1962 {
1963 	if (dev->netdev_ops->ndo_set_rx_headroom)
1964 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1965 }
1966 
1967 /* set the device rx headroom to the dev's default */
1968 static inline void netdev_reset_rx_headroom(struct net_device *dev)
1969 {
1970 	netdev_set_rx_headroom(dev, -1);
1971 }
1972 
1973 /*
1974  * Net namespace inlines
1975  */
1976 static inline
1977 struct net *dev_net(const struct net_device *dev)
1978 {
1979 	return read_pnet(&dev->nd_net);
1980 }
1981 
1982 static inline
1983 void dev_net_set(struct net_device *dev, struct net *net)
1984 {
1985 	write_pnet(&dev->nd_net, net);
1986 }
1987 
1988 static inline bool netdev_uses_dsa(struct net_device *dev)
1989 {
1990 #if IS_ENABLED(CONFIG_NET_DSA)
1991 	if (dev->dsa_ptr != NULL)
1992 		return dsa_uses_tagged_protocol(dev->dsa_ptr);
1993 #endif
1994 	return false;
1995 }
1996 
1997 /**
1998  *	netdev_priv - access network device private data
1999  *	@dev: network device
2000  *
2001  * Get network device private data
2002  */
2003 static inline void *netdev_priv(const struct net_device *dev)
2004 {
2005 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2006 }
2007 
2008 /* Set the sysfs physical device reference for the network logical device
2009  * if set prior to registration will cause a symlink during initialization.
2010  */
2011 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2012 
2013 /* Set the sysfs device type for the network logical device to allow
2014  * fine-grained identification of different network device types. For
2015  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2016  */
2017 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2018 
2019 /* Default NAPI poll() weight
2020  * Device drivers are strongly advised to not use bigger value
2021  */
2022 #define NAPI_POLL_WEIGHT 64
2023 
2024 /**
2025  *	netif_napi_add - initialize a NAPI context
2026  *	@dev:  network device
2027  *	@napi: NAPI context
2028  *	@poll: polling function
2029  *	@weight: default weight
2030  *
2031  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2032  * *any* of the other NAPI-related functions.
2033  */
2034 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2035 		    int (*poll)(struct napi_struct *, int), int weight);
2036 
2037 /**
2038  *	netif_tx_napi_add - initialize a NAPI context
2039  *	@dev:  network device
2040  *	@napi: NAPI context
2041  *	@poll: polling function
2042  *	@weight: default weight
2043  *
2044  * This variant of netif_napi_add() should be used from drivers using NAPI
2045  * to exclusively poll a TX queue.
2046  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2047  */
2048 static inline void netif_tx_napi_add(struct net_device *dev,
2049 				     struct napi_struct *napi,
2050 				     int (*poll)(struct napi_struct *, int),
2051 				     int weight)
2052 {
2053 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2054 	netif_napi_add(dev, napi, poll, weight);
2055 }
2056 
2057 /**
2058  *  netif_napi_del - remove a NAPI context
2059  *  @napi: NAPI context
2060  *
2061  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2062  */
2063 void netif_napi_del(struct napi_struct *napi);
2064 
2065 struct napi_gro_cb {
2066 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2067 	void	*frag0;
2068 
2069 	/* Length of frag0. */
2070 	unsigned int frag0_len;
2071 
2072 	/* This indicates where we are processing relative to skb->data. */
2073 	int	data_offset;
2074 
2075 	/* This is non-zero if the packet cannot be merged with the new skb. */
2076 	u16	flush;
2077 
2078 	/* Save the IP ID here and check when we get to the transport layer */
2079 	u16	flush_id;
2080 
2081 	/* Number of segments aggregated. */
2082 	u16	count;
2083 
2084 	/* Start offset for remote checksum offload */
2085 	u16	gro_remcsum_start;
2086 
2087 	/* jiffies when first packet was created/queued */
2088 	unsigned long age;
2089 
2090 	/* Used in ipv6_gro_receive() and foo-over-udp */
2091 	u16	proto;
2092 
2093 	/* This is non-zero if the packet may be of the same flow. */
2094 	u8	same_flow:1;
2095 
2096 	/* Used in tunnel GRO receive */
2097 	u8	encap_mark:1;
2098 
2099 	/* GRO checksum is valid */
2100 	u8	csum_valid:1;
2101 
2102 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2103 	u8	csum_cnt:3;
2104 
2105 	/* Free the skb? */
2106 	u8	free:2;
2107 #define NAPI_GRO_FREE		  1
2108 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2109 
2110 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2111 	u8	is_ipv6:1;
2112 
2113 	/* Used in GRE, set in fou/gue_gro_receive */
2114 	u8	is_fou:1;
2115 
2116 	/* Used to determine if flush_id can be ignored */
2117 	u8	is_atomic:1;
2118 
2119 	/* 5 bit hole */
2120 
2121 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2122 	__wsum	csum;
2123 
2124 	/* used in skb_gro_receive() slow path */
2125 	struct sk_buff *last;
2126 };
2127 
2128 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2129 
2130 struct packet_type {
2131 	__be16			type;	/* This is really htons(ether_type). */
2132 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2133 	int			(*func) (struct sk_buff *,
2134 					 struct net_device *,
2135 					 struct packet_type *,
2136 					 struct net_device *);
2137 	bool			(*id_match)(struct packet_type *ptype,
2138 					    struct sock *sk);
2139 	void			*af_packet_priv;
2140 	struct list_head	list;
2141 };
2142 
2143 struct offload_callbacks {
2144 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2145 						netdev_features_t features);
2146 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2147 						 struct sk_buff *skb);
2148 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2149 };
2150 
2151 struct packet_offload {
2152 	__be16			 type;	/* This is really htons(ether_type). */
2153 	u16			 priority;
2154 	struct offload_callbacks callbacks;
2155 	struct list_head	 list;
2156 };
2157 
2158 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2159 struct pcpu_sw_netstats {
2160 	u64     rx_packets;
2161 	u64     rx_bytes;
2162 	u64     tx_packets;
2163 	u64     tx_bytes;
2164 	struct u64_stats_sync   syncp;
2165 };
2166 
2167 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2168 ({									\
2169 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2170 	if (pcpu_stats)	{						\
2171 		int __cpu;						\
2172 		for_each_possible_cpu(__cpu) {				\
2173 			typeof(type) *stat;				\
2174 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2175 			u64_stats_init(&stat->syncp);			\
2176 		}							\
2177 	}								\
2178 	pcpu_stats;							\
2179 })
2180 
2181 #define netdev_alloc_pcpu_stats(type)					\
2182 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2183 
2184 enum netdev_lag_tx_type {
2185 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2186 	NETDEV_LAG_TX_TYPE_RANDOM,
2187 	NETDEV_LAG_TX_TYPE_BROADCAST,
2188 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2189 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2190 	NETDEV_LAG_TX_TYPE_HASH,
2191 };
2192 
2193 struct netdev_lag_upper_info {
2194 	enum netdev_lag_tx_type tx_type;
2195 };
2196 
2197 struct netdev_lag_lower_state_info {
2198 	u8 link_up : 1,
2199 	   tx_enabled : 1;
2200 };
2201 
2202 #include <linux/notifier.h>
2203 
2204 /* netdevice notifier chain. Please remember to update the rtnetlink
2205  * notification exclusion list in rtnetlink_event() when adding new
2206  * types.
2207  */
2208 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2209 #define NETDEV_DOWN	0x0002
2210 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2211 				   detected a hardware crash and restarted
2212 				   - we can use this eg to kick tcp sessions
2213 				   once done */
2214 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2215 #define NETDEV_REGISTER 0x0005
2216 #define NETDEV_UNREGISTER	0x0006
2217 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2218 #define NETDEV_CHANGEADDR	0x0008
2219 #define NETDEV_GOING_DOWN	0x0009
2220 #define NETDEV_CHANGENAME	0x000A
2221 #define NETDEV_FEAT_CHANGE	0x000B
2222 #define NETDEV_BONDING_FAILOVER 0x000C
2223 #define NETDEV_PRE_UP		0x000D
2224 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2225 #define NETDEV_POST_TYPE_CHANGE	0x000F
2226 #define NETDEV_POST_INIT	0x0010
2227 #define NETDEV_UNREGISTER_FINAL 0x0011
2228 #define NETDEV_RELEASE		0x0012
2229 #define NETDEV_NOTIFY_PEERS	0x0013
2230 #define NETDEV_JOIN		0x0014
2231 #define NETDEV_CHANGEUPPER	0x0015
2232 #define NETDEV_RESEND_IGMP	0x0016
2233 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2234 #define NETDEV_CHANGEINFODATA	0x0018
2235 #define NETDEV_BONDING_INFO	0x0019
2236 #define NETDEV_PRECHANGEUPPER	0x001A
2237 #define NETDEV_CHANGELOWERSTATE	0x001B
2238 #define NETDEV_OFFLOAD_PUSH_VXLAN	0x001C
2239 #define NETDEV_OFFLOAD_PUSH_GENEVE	0x001D
2240 
2241 int register_netdevice_notifier(struct notifier_block *nb);
2242 int unregister_netdevice_notifier(struct notifier_block *nb);
2243 
2244 struct netdev_notifier_info {
2245 	struct net_device *dev;
2246 };
2247 
2248 struct netdev_notifier_change_info {
2249 	struct netdev_notifier_info info; /* must be first */
2250 	unsigned int flags_changed;
2251 };
2252 
2253 struct netdev_notifier_changeupper_info {
2254 	struct netdev_notifier_info info; /* must be first */
2255 	struct net_device *upper_dev; /* new upper dev */
2256 	bool master; /* is upper dev master */
2257 	bool linking; /* is the notification for link or unlink */
2258 	void *upper_info; /* upper dev info */
2259 };
2260 
2261 struct netdev_notifier_changelowerstate_info {
2262 	struct netdev_notifier_info info; /* must be first */
2263 	void *lower_state_info; /* is lower dev state */
2264 };
2265 
2266 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2267 					     struct net_device *dev)
2268 {
2269 	info->dev = dev;
2270 }
2271 
2272 static inline struct net_device *
2273 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2274 {
2275 	return info->dev;
2276 }
2277 
2278 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2279 
2280 
2281 extern rwlock_t				dev_base_lock;		/* Device list lock */
2282 
2283 #define for_each_netdev(net, d)		\
2284 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2285 #define for_each_netdev_reverse(net, d)	\
2286 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2287 #define for_each_netdev_rcu(net, d)		\
2288 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2289 #define for_each_netdev_safe(net, d, n)	\
2290 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2291 #define for_each_netdev_continue(net, d)		\
2292 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2293 #define for_each_netdev_continue_rcu(net, d)		\
2294 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2295 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2296 		for_each_netdev_rcu(&init_net, slave)	\
2297 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2298 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2299 
2300 static inline struct net_device *next_net_device(struct net_device *dev)
2301 {
2302 	struct list_head *lh;
2303 	struct net *net;
2304 
2305 	net = dev_net(dev);
2306 	lh = dev->dev_list.next;
2307 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2308 }
2309 
2310 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2311 {
2312 	struct list_head *lh;
2313 	struct net *net;
2314 
2315 	net = dev_net(dev);
2316 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2317 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2318 }
2319 
2320 static inline struct net_device *first_net_device(struct net *net)
2321 {
2322 	return list_empty(&net->dev_base_head) ? NULL :
2323 		net_device_entry(net->dev_base_head.next);
2324 }
2325 
2326 static inline struct net_device *first_net_device_rcu(struct net *net)
2327 {
2328 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2329 
2330 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2331 }
2332 
2333 int netdev_boot_setup_check(struct net_device *dev);
2334 unsigned long netdev_boot_base(const char *prefix, int unit);
2335 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2336 				       const char *hwaddr);
2337 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2338 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2339 void dev_add_pack(struct packet_type *pt);
2340 void dev_remove_pack(struct packet_type *pt);
2341 void __dev_remove_pack(struct packet_type *pt);
2342 void dev_add_offload(struct packet_offload *po);
2343 void dev_remove_offload(struct packet_offload *po);
2344 
2345 int dev_get_iflink(const struct net_device *dev);
2346 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2347 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2348 				      unsigned short mask);
2349 struct net_device *dev_get_by_name(struct net *net, const char *name);
2350 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2351 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2352 int dev_alloc_name(struct net_device *dev, const char *name);
2353 int dev_open(struct net_device *dev);
2354 int dev_close(struct net_device *dev);
2355 int dev_close_many(struct list_head *head, bool unlink);
2356 void dev_disable_lro(struct net_device *dev);
2357 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2358 int dev_queue_xmit(struct sk_buff *skb);
2359 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2360 int register_netdevice(struct net_device *dev);
2361 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2362 void unregister_netdevice_many(struct list_head *head);
2363 static inline void unregister_netdevice(struct net_device *dev)
2364 {
2365 	unregister_netdevice_queue(dev, NULL);
2366 }
2367 
2368 int netdev_refcnt_read(const struct net_device *dev);
2369 void free_netdev(struct net_device *dev);
2370 void netdev_freemem(struct net_device *dev);
2371 void synchronize_net(void);
2372 int init_dummy_netdev(struct net_device *dev);
2373 
2374 DECLARE_PER_CPU(int, xmit_recursion);
2375 static inline int dev_recursion_level(void)
2376 {
2377 	return this_cpu_read(xmit_recursion);
2378 }
2379 
2380 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2381 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2382 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2383 int netdev_get_name(struct net *net, char *name, int ifindex);
2384 int dev_restart(struct net_device *dev);
2385 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2386 
2387 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2388 {
2389 	return NAPI_GRO_CB(skb)->data_offset;
2390 }
2391 
2392 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2393 {
2394 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2395 }
2396 
2397 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2398 {
2399 	NAPI_GRO_CB(skb)->data_offset += len;
2400 }
2401 
2402 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2403 					unsigned int offset)
2404 {
2405 	return NAPI_GRO_CB(skb)->frag0 + offset;
2406 }
2407 
2408 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2409 {
2410 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2411 }
2412 
2413 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2414 					unsigned int offset)
2415 {
2416 	if (!pskb_may_pull(skb, hlen))
2417 		return NULL;
2418 
2419 	NAPI_GRO_CB(skb)->frag0 = NULL;
2420 	NAPI_GRO_CB(skb)->frag0_len = 0;
2421 	return skb->data + offset;
2422 }
2423 
2424 static inline void *skb_gro_network_header(struct sk_buff *skb)
2425 {
2426 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2427 	       skb_network_offset(skb);
2428 }
2429 
2430 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2431 					const void *start, unsigned int len)
2432 {
2433 	if (NAPI_GRO_CB(skb)->csum_valid)
2434 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2435 						  csum_partial(start, len, 0));
2436 }
2437 
2438 /* GRO checksum functions. These are logical equivalents of the normal
2439  * checksum functions (in skbuff.h) except that they operate on the GRO
2440  * offsets and fields in sk_buff.
2441  */
2442 
2443 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2444 
2445 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2446 {
2447 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2448 }
2449 
2450 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2451 						      bool zero_okay,
2452 						      __sum16 check)
2453 {
2454 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2455 		skb_checksum_start_offset(skb) <
2456 		 skb_gro_offset(skb)) &&
2457 		!skb_at_gro_remcsum_start(skb) &&
2458 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2459 		(!zero_okay || check));
2460 }
2461 
2462 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2463 							   __wsum psum)
2464 {
2465 	if (NAPI_GRO_CB(skb)->csum_valid &&
2466 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2467 		return 0;
2468 
2469 	NAPI_GRO_CB(skb)->csum = psum;
2470 
2471 	return __skb_gro_checksum_complete(skb);
2472 }
2473 
2474 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2475 {
2476 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2477 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2478 		NAPI_GRO_CB(skb)->csum_cnt--;
2479 	} else {
2480 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2481 		 * verified a new top level checksum or an encapsulated one
2482 		 * during GRO. This saves work if we fallback to normal path.
2483 		 */
2484 		__skb_incr_checksum_unnecessary(skb);
2485 	}
2486 }
2487 
2488 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2489 				    compute_pseudo)			\
2490 ({									\
2491 	__sum16 __ret = 0;						\
2492 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2493 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2494 				compute_pseudo(skb, proto));		\
2495 	if (__ret)							\
2496 		__skb_mark_checksum_bad(skb);				\
2497 	else								\
2498 		skb_gro_incr_csum_unnecessary(skb);			\
2499 	__ret;								\
2500 })
2501 
2502 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2503 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2504 
2505 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2506 					     compute_pseudo)		\
2507 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2508 
2509 #define skb_gro_checksum_simple_validate(skb)				\
2510 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2511 
2512 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2513 {
2514 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2515 		!NAPI_GRO_CB(skb)->csum_valid);
2516 }
2517 
2518 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2519 					      __sum16 check, __wsum pseudo)
2520 {
2521 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2522 	NAPI_GRO_CB(skb)->csum_valid = 1;
2523 }
2524 
2525 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2526 do {									\
2527 	if (__skb_gro_checksum_convert_check(skb))			\
2528 		__skb_gro_checksum_convert(skb, check,			\
2529 					   compute_pseudo(skb, proto));	\
2530 } while (0)
2531 
2532 struct gro_remcsum {
2533 	int offset;
2534 	__wsum delta;
2535 };
2536 
2537 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2538 {
2539 	grc->offset = 0;
2540 	grc->delta = 0;
2541 }
2542 
2543 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2544 					    unsigned int off, size_t hdrlen,
2545 					    int start, int offset,
2546 					    struct gro_remcsum *grc,
2547 					    bool nopartial)
2548 {
2549 	__wsum delta;
2550 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2551 
2552 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2553 
2554 	if (!nopartial) {
2555 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2556 		return ptr;
2557 	}
2558 
2559 	ptr = skb_gro_header_fast(skb, off);
2560 	if (skb_gro_header_hard(skb, off + plen)) {
2561 		ptr = skb_gro_header_slow(skb, off + plen, off);
2562 		if (!ptr)
2563 			return NULL;
2564 	}
2565 
2566 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2567 			       start, offset);
2568 
2569 	/* Adjust skb->csum since we changed the packet */
2570 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2571 
2572 	grc->offset = off + hdrlen + offset;
2573 	grc->delta = delta;
2574 
2575 	return ptr;
2576 }
2577 
2578 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2579 					   struct gro_remcsum *grc)
2580 {
2581 	void *ptr;
2582 	size_t plen = grc->offset + sizeof(u16);
2583 
2584 	if (!grc->delta)
2585 		return;
2586 
2587 	ptr = skb_gro_header_fast(skb, grc->offset);
2588 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2589 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2590 		if (!ptr)
2591 			return;
2592 	}
2593 
2594 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2595 }
2596 
2597 struct skb_csum_offl_spec {
2598 	__u16		ipv4_okay:1,
2599 			ipv6_okay:1,
2600 			encap_okay:1,
2601 			ip_options_okay:1,
2602 			ext_hdrs_okay:1,
2603 			tcp_okay:1,
2604 			udp_okay:1,
2605 			sctp_okay:1,
2606 			vlan_okay:1,
2607 			no_encapped_ipv6:1,
2608 			no_not_encapped:1;
2609 };
2610 
2611 bool __skb_csum_offload_chk(struct sk_buff *skb,
2612 			    const struct skb_csum_offl_spec *spec,
2613 			    bool *csum_encapped,
2614 			    bool csum_help);
2615 
2616 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2617 					const struct skb_csum_offl_spec *spec,
2618 					bool *csum_encapped,
2619 					bool csum_help)
2620 {
2621 	if (skb->ip_summed != CHECKSUM_PARTIAL)
2622 		return false;
2623 
2624 	return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2625 }
2626 
2627 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2628 					     const struct skb_csum_offl_spec *spec)
2629 {
2630 	bool csum_encapped;
2631 
2632 	return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2633 }
2634 
2635 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2636 {
2637 	static const struct skb_csum_offl_spec csum_offl_spec = {
2638 		.ipv4_okay = 1,
2639 		.ip_options_okay = 1,
2640 		.ipv6_okay = 1,
2641 		.vlan_okay = 1,
2642 		.tcp_okay = 1,
2643 		.udp_okay = 1,
2644 	};
2645 
2646 	return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2647 }
2648 
2649 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2650 {
2651 	static const struct skb_csum_offl_spec csum_offl_spec = {
2652 		.ipv4_okay = 1,
2653 		.ip_options_okay = 1,
2654 		.tcp_okay = 1,
2655 		.udp_okay = 1,
2656 		.vlan_okay = 1,
2657 	};
2658 
2659 	return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2660 }
2661 
2662 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2663 				  unsigned short type,
2664 				  const void *daddr, const void *saddr,
2665 				  unsigned int len)
2666 {
2667 	if (!dev->header_ops || !dev->header_ops->create)
2668 		return 0;
2669 
2670 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2671 }
2672 
2673 static inline int dev_parse_header(const struct sk_buff *skb,
2674 				   unsigned char *haddr)
2675 {
2676 	const struct net_device *dev = skb->dev;
2677 
2678 	if (!dev->header_ops || !dev->header_ops->parse)
2679 		return 0;
2680 	return dev->header_ops->parse(skb, haddr);
2681 }
2682 
2683 /* ll_header must have at least hard_header_len allocated */
2684 static inline bool dev_validate_header(const struct net_device *dev,
2685 				       char *ll_header, int len)
2686 {
2687 	if (likely(len >= dev->hard_header_len))
2688 		return true;
2689 
2690 	if (capable(CAP_SYS_RAWIO)) {
2691 		memset(ll_header + len, 0, dev->hard_header_len - len);
2692 		return true;
2693 	}
2694 
2695 	if (dev->header_ops && dev->header_ops->validate)
2696 		return dev->header_ops->validate(ll_header, len);
2697 
2698 	return false;
2699 }
2700 
2701 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2702 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2703 static inline int unregister_gifconf(unsigned int family)
2704 {
2705 	return register_gifconf(family, NULL);
2706 }
2707 
2708 #ifdef CONFIG_NET_FLOW_LIMIT
2709 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2710 struct sd_flow_limit {
2711 	u64			count;
2712 	unsigned int		num_buckets;
2713 	unsigned int		history_head;
2714 	u16			history[FLOW_LIMIT_HISTORY];
2715 	u8			buckets[];
2716 };
2717 
2718 extern int netdev_flow_limit_table_len;
2719 #endif /* CONFIG_NET_FLOW_LIMIT */
2720 
2721 /*
2722  * Incoming packets are placed on per-CPU queues
2723  */
2724 struct softnet_data {
2725 	struct list_head	poll_list;
2726 	struct sk_buff_head	process_queue;
2727 
2728 	/* stats */
2729 	unsigned int		processed;
2730 	unsigned int		time_squeeze;
2731 	unsigned int		received_rps;
2732 #ifdef CONFIG_RPS
2733 	struct softnet_data	*rps_ipi_list;
2734 #endif
2735 #ifdef CONFIG_NET_FLOW_LIMIT
2736 	struct sd_flow_limit __rcu *flow_limit;
2737 #endif
2738 	struct Qdisc		*output_queue;
2739 	struct Qdisc		**output_queue_tailp;
2740 	struct sk_buff		*completion_queue;
2741 
2742 #ifdef CONFIG_RPS
2743 	/* input_queue_head should be written by cpu owning this struct,
2744 	 * and only read by other cpus. Worth using a cache line.
2745 	 */
2746 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2747 
2748 	/* Elements below can be accessed between CPUs for RPS/RFS */
2749 	struct call_single_data	csd ____cacheline_aligned_in_smp;
2750 	struct softnet_data	*rps_ipi_next;
2751 	unsigned int		cpu;
2752 	unsigned int		input_queue_tail;
2753 #endif
2754 	unsigned int		dropped;
2755 	struct sk_buff_head	input_pkt_queue;
2756 	struct napi_struct	backlog;
2757 
2758 };
2759 
2760 static inline void input_queue_head_incr(struct softnet_data *sd)
2761 {
2762 #ifdef CONFIG_RPS
2763 	sd->input_queue_head++;
2764 #endif
2765 }
2766 
2767 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2768 					      unsigned int *qtail)
2769 {
2770 #ifdef CONFIG_RPS
2771 	*qtail = ++sd->input_queue_tail;
2772 #endif
2773 }
2774 
2775 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2776 
2777 void __netif_schedule(struct Qdisc *q);
2778 void netif_schedule_queue(struct netdev_queue *txq);
2779 
2780 static inline void netif_tx_schedule_all(struct net_device *dev)
2781 {
2782 	unsigned int i;
2783 
2784 	for (i = 0; i < dev->num_tx_queues; i++)
2785 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2786 }
2787 
2788 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2789 {
2790 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2791 }
2792 
2793 /**
2794  *	netif_start_queue - allow transmit
2795  *	@dev: network device
2796  *
2797  *	Allow upper layers to call the device hard_start_xmit routine.
2798  */
2799 static inline void netif_start_queue(struct net_device *dev)
2800 {
2801 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2802 }
2803 
2804 static inline void netif_tx_start_all_queues(struct net_device *dev)
2805 {
2806 	unsigned int i;
2807 
2808 	for (i = 0; i < dev->num_tx_queues; i++) {
2809 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2810 		netif_tx_start_queue(txq);
2811 	}
2812 }
2813 
2814 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2815 
2816 /**
2817  *	netif_wake_queue - restart transmit
2818  *	@dev: network device
2819  *
2820  *	Allow upper layers to call the device hard_start_xmit routine.
2821  *	Used for flow control when transmit resources are available.
2822  */
2823 static inline void netif_wake_queue(struct net_device *dev)
2824 {
2825 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2826 }
2827 
2828 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2829 {
2830 	unsigned int i;
2831 
2832 	for (i = 0; i < dev->num_tx_queues; i++) {
2833 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2834 		netif_tx_wake_queue(txq);
2835 	}
2836 }
2837 
2838 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2839 {
2840 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2841 }
2842 
2843 /**
2844  *	netif_stop_queue - stop transmitted packets
2845  *	@dev: network device
2846  *
2847  *	Stop upper layers calling the device hard_start_xmit routine.
2848  *	Used for flow control when transmit resources are unavailable.
2849  */
2850 static inline void netif_stop_queue(struct net_device *dev)
2851 {
2852 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2853 }
2854 
2855 void netif_tx_stop_all_queues(struct net_device *dev);
2856 
2857 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2858 {
2859 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2860 }
2861 
2862 /**
2863  *	netif_queue_stopped - test if transmit queue is flowblocked
2864  *	@dev: network device
2865  *
2866  *	Test if transmit queue on device is currently unable to send.
2867  */
2868 static inline bool netif_queue_stopped(const struct net_device *dev)
2869 {
2870 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2871 }
2872 
2873 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2874 {
2875 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2876 }
2877 
2878 static inline bool
2879 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2880 {
2881 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2882 }
2883 
2884 static inline bool
2885 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2886 {
2887 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2888 }
2889 
2890 /**
2891  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2892  *	@dev_queue: pointer to transmit queue
2893  *
2894  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2895  * to give appropriate hint to the CPU.
2896  */
2897 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2898 {
2899 #ifdef CONFIG_BQL
2900 	prefetchw(&dev_queue->dql.num_queued);
2901 #endif
2902 }
2903 
2904 /**
2905  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2906  *	@dev_queue: pointer to transmit queue
2907  *
2908  * BQL enabled drivers might use this helper in their TX completion path,
2909  * to give appropriate hint to the CPU.
2910  */
2911 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2912 {
2913 #ifdef CONFIG_BQL
2914 	prefetchw(&dev_queue->dql.limit);
2915 #endif
2916 }
2917 
2918 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2919 					unsigned int bytes)
2920 {
2921 #ifdef CONFIG_BQL
2922 	dql_queued(&dev_queue->dql, bytes);
2923 
2924 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2925 		return;
2926 
2927 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2928 
2929 	/*
2930 	 * The XOFF flag must be set before checking the dql_avail below,
2931 	 * because in netdev_tx_completed_queue we update the dql_completed
2932 	 * before checking the XOFF flag.
2933 	 */
2934 	smp_mb();
2935 
2936 	/* check again in case another CPU has just made room avail */
2937 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2938 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2939 #endif
2940 }
2941 
2942 /**
2943  * 	netdev_sent_queue - report the number of bytes queued to hardware
2944  * 	@dev: network device
2945  * 	@bytes: number of bytes queued to the hardware device queue
2946  *
2947  * 	Report the number of bytes queued for sending/completion to the network
2948  * 	device hardware queue. @bytes should be a good approximation and should
2949  * 	exactly match netdev_completed_queue() @bytes
2950  */
2951 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2952 {
2953 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2954 }
2955 
2956 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2957 					     unsigned int pkts, unsigned int bytes)
2958 {
2959 #ifdef CONFIG_BQL
2960 	if (unlikely(!bytes))
2961 		return;
2962 
2963 	dql_completed(&dev_queue->dql, bytes);
2964 
2965 	/*
2966 	 * Without the memory barrier there is a small possiblity that
2967 	 * netdev_tx_sent_queue will miss the update and cause the queue to
2968 	 * be stopped forever
2969 	 */
2970 	smp_mb();
2971 
2972 	if (dql_avail(&dev_queue->dql) < 0)
2973 		return;
2974 
2975 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2976 		netif_schedule_queue(dev_queue);
2977 #endif
2978 }
2979 
2980 /**
2981  * 	netdev_completed_queue - report bytes and packets completed by device
2982  * 	@dev: network device
2983  * 	@pkts: actual number of packets sent over the medium
2984  * 	@bytes: actual number of bytes sent over the medium
2985  *
2986  * 	Report the number of bytes and packets transmitted by the network device
2987  * 	hardware queue over the physical medium, @bytes must exactly match the
2988  * 	@bytes amount passed to netdev_sent_queue()
2989  */
2990 static inline void netdev_completed_queue(struct net_device *dev,
2991 					  unsigned int pkts, unsigned int bytes)
2992 {
2993 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2994 }
2995 
2996 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2997 {
2998 #ifdef CONFIG_BQL
2999 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3000 	dql_reset(&q->dql);
3001 #endif
3002 }
3003 
3004 /**
3005  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3006  * 	@dev_queue: network device
3007  *
3008  * 	Reset the bytes and packet count of a network device and clear the
3009  * 	software flow control OFF bit for this network device
3010  */
3011 static inline void netdev_reset_queue(struct net_device *dev_queue)
3012 {
3013 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3014 }
3015 
3016 /**
3017  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3018  * 	@dev: network device
3019  * 	@queue_index: given tx queue index
3020  *
3021  * 	Returns 0 if given tx queue index >= number of device tx queues,
3022  * 	otherwise returns the originally passed tx queue index.
3023  */
3024 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3025 {
3026 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3027 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3028 				     dev->name, queue_index,
3029 				     dev->real_num_tx_queues);
3030 		return 0;
3031 	}
3032 
3033 	return queue_index;
3034 }
3035 
3036 /**
3037  *	netif_running - test if up
3038  *	@dev: network device
3039  *
3040  *	Test if the device has been brought up.
3041  */
3042 static inline bool netif_running(const struct net_device *dev)
3043 {
3044 	return test_bit(__LINK_STATE_START, &dev->state);
3045 }
3046 
3047 /*
3048  * Routines to manage the subqueues on a device.  We only need start,
3049  * stop, and a check if it's stopped.  All other device management is
3050  * done at the overall netdevice level.
3051  * Also test the device if we're multiqueue.
3052  */
3053 
3054 /**
3055  *	netif_start_subqueue - allow sending packets on subqueue
3056  *	@dev: network device
3057  *	@queue_index: sub queue index
3058  *
3059  * Start individual transmit queue of a device with multiple transmit queues.
3060  */
3061 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3062 {
3063 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3064 
3065 	netif_tx_start_queue(txq);
3066 }
3067 
3068 /**
3069  *	netif_stop_subqueue - stop sending packets on subqueue
3070  *	@dev: network device
3071  *	@queue_index: sub queue index
3072  *
3073  * Stop individual transmit queue of a device with multiple transmit queues.
3074  */
3075 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3076 {
3077 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3078 	netif_tx_stop_queue(txq);
3079 }
3080 
3081 /**
3082  *	netif_subqueue_stopped - test status of subqueue
3083  *	@dev: network device
3084  *	@queue_index: sub queue index
3085  *
3086  * Check individual transmit queue of a device with multiple transmit queues.
3087  */
3088 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3089 					    u16 queue_index)
3090 {
3091 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3092 
3093 	return netif_tx_queue_stopped(txq);
3094 }
3095 
3096 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3097 					  struct sk_buff *skb)
3098 {
3099 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3100 }
3101 
3102 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3103 
3104 #ifdef CONFIG_XPS
3105 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3106 			u16 index);
3107 #else
3108 static inline int netif_set_xps_queue(struct net_device *dev,
3109 				      const struct cpumask *mask,
3110 				      u16 index)
3111 {
3112 	return 0;
3113 }
3114 #endif
3115 
3116 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3117 		  unsigned int num_tx_queues);
3118 
3119 /*
3120  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3121  * as a distribution range limit for the returned value.
3122  */
3123 static inline u16 skb_tx_hash(const struct net_device *dev,
3124 			      struct sk_buff *skb)
3125 {
3126 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3127 }
3128 
3129 /**
3130  *	netif_is_multiqueue - test if device has multiple transmit queues
3131  *	@dev: network device
3132  *
3133  * Check if device has multiple transmit queues
3134  */
3135 static inline bool netif_is_multiqueue(const struct net_device *dev)
3136 {
3137 	return dev->num_tx_queues > 1;
3138 }
3139 
3140 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3141 
3142 #ifdef CONFIG_SYSFS
3143 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3144 #else
3145 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3146 						unsigned int rxq)
3147 {
3148 	return 0;
3149 }
3150 #endif
3151 
3152 #ifdef CONFIG_SYSFS
3153 static inline unsigned int get_netdev_rx_queue_index(
3154 		struct netdev_rx_queue *queue)
3155 {
3156 	struct net_device *dev = queue->dev;
3157 	int index = queue - dev->_rx;
3158 
3159 	BUG_ON(index >= dev->num_rx_queues);
3160 	return index;
3161 }
3162 #endif
3163 
3164 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3165 int netif_get_num_default_rss_queues(void);
3166 
3167 enum skb_free_reason {
3168 	SKB_REASON_CONSUMED,
3169 	SKB_REASON_DROPPED,
3170 };
3171 
3172 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3173 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3174 
3175 /*
3176  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3177  * interrupt context or with hardware interrupts being disabled.
3178  * (in_irq() || irqs_disabled())
3179  *
3180  * We provide four helpers that can be used in following contexts :
3181  *
3182  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3183  *  replacing kfree_skb(skb)
3184  *
3185  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3186  *  Typically used in place of consume_skb(skb) in TX completion path
3187  *
3188  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3189  *  replacing kfree_skb(skb)
3190  *
3191  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3192  *  and consumed a packet. Used in place of consume_skb(skb)
3193  */
3194 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3195 {
3196 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3197 }
3198 
3199 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3200 {
3201 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3202 }
3203 
3204 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3205 {
3206 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3207 }
3208 
3209 static inline void dev_consume_skb_any(struct sk_buff *skb)
3210 {
3211 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3212 }
3213 
3214 int netif_rx(struct sk_buff *skb);
3215 int netif_rx_ni(struct sk_buff *skb);
3216 int netif_receive_skb(struct sk_buff *skb);
3217 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3218 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3219 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3220 gro_result_t napi_gro_frags(struct napi_struct *napi);
3221 struct packet_offload *gro_find_receive_by_type(__be16 type);
3222 struct packet_offload *gro_find_complete_by_type(__be16 type);
3223 
3224 static inline void napi_free_frags(struct napi_struct *napi)
3225 {
3226 	kfree_skb(napi->skb);
3227 	napi->skb = NULL;
3228 }
3229 
3230 int netdev_rx_handler_register(struct net_device *dev,
3231 			       rx_handler_func_t *rx_handler,
3232 			       void *rx_handler_data);
3233 void netdev_rx_handler_unregister(struct net_device *dev);
3234 
3235 bool dev_valid_name(const char *name);
3236 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3237 int dev_ethtool(struct net *net, struct ifreq *);
3238 unsigned int dev_get_flags(const struct net_device *);
3239 int __dev_change_flags(struct net_device *, unsigned int flags);
3240 int dev_change_flags(struct net_device *, unsigned int);
3241 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3242 			unsigned int gchanges);
3243 int dev_change_name(struct net_device *, const char *);
3244 int dev_set_alias(struct net_device *, const char *, size_t);
3245 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3246 int dev_set_mtu(struct net_device *, int);
3247 void dev_set_group(struct net_device *, int);
3248 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3249 int dev_change_carrier(struct net_device *, bool new_carrier);
3250 int dev_get_phys_port_id(struct net_device *dev,
3251 			 struct netdev_phys_item_id *ppid);
3252 int dev_get_phys_port_name(struct net_device *dev,
3253 			   char *name, size_t len);
3254 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3255 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3256 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3257 				    struct netdev_queue *txq, int *ret);
3258 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3259 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3260 bool is_skb_forwardable(const struct net_device *dev,
3261 			const struct sk_buff *skb);
3262 
3263 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3264 
3265 extern int		netdev_budget;
3266 
3267 /* Called by rtnetlink.c:rtnl_unlock() */
3268 void netdev_run_todo(void);
3269 
3270 /**
3271  *	dev_put - release reference to device
3272  *	@dev: network device
3273  *
3274  * Release reference to device to allow it to be freed.
3275  */
3276 static inline void dev_put(struct net_device *dev)
3277 {
3278 	this_cpu_dec(*dev->pcpu_refcnt);
3279 }
3280 
3281 /**
3282  *	dev_hold - get reference to device
3283  *	@dev: network device
3284  *
3285  * Hold reference to device to keep it from being freed.
3286  */
3287 static inline void dev_hold(struct net_device *dev)
3288 {
3289 	this_cpu_inc(*dev->pcpu_refcnt);
3290 }
3291 
3292 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3293  * and _off may be called from IRQ context, but it is caller
3294  * who is responsible for serialization of these calls.
3295  *
3296  * The name carrier is inappropriate, these functions should really be
3297  * called netif_lowerlayer_*() because they represent the state of any
3298  * kind of lower layer not just hardware media.
3299  */
3300 
3301 void linkwatch_init_dev(struct net_device *dev);
3302 void linkwatch_fire_event(struct net_device *dev);
3303 void linkwatch_forget_dev(struct net_device *dev);
3304 
3305 /**
3306  *	netif_carrier_ok - test if carrier present
3307  *	@dev: network device
3308  *
3309  * Check if carrier is present on device
3310  */
3311 static inline bool netif_carrier_ok(const struct net_device *dev)
3312 {
3313 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3314 }
3315 
3316 unsigned long dev_trans_start(struct net_device *dev);
3317 
3318 void __netdev_watchdog_up(struct net_device *dev);
3319 
3320 void netif_carrier_on(struct net_device *dev);
3321 
3322 void netif_carrier_off(struct net_device *dev);
3323 
3324 /**
3325  *	netif_dormant_on - mark device as dormant.
3326  *	@dev: network device
3327  *
3328  * Mark device as dormant (as per RFC2863).
3329  *
3330  * The dormant state indicates that the relevant interface is not
3331  * actually in a condition to pass packets (i.e., it is not 'up') but is
3332  * in a "pending" state, waiting for some external event.  For "on-
3333  * demand" interfaces, this new state identifies the situation where the
3334  * interface is waiting for events to place it in the up state.
3335  */
3336 static inline void netif_dormant_on(struct net_device *dev)
3337 {
3338 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3339 		linkwatch_fire_event(dev);
3340 }
3341 
3342 /**
3343  *	netif_dormant_off - set device as not dormant.
3344  *	@dev: network device
3345  *
3346  * Device is not in dormant state.
3347  */
3348 static inline void netif_dormant_off(struct net_device *dev)
3349 {
3350 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3351 		linkwatch_fire_event(dev);
3352 }
3353 
3354 /**
3355  *	netif_dormant - test if carrier present
3356  *	@dev: network device
3357  *
3358  * Check if carrier is present on device
3359  */
3360 static inline bool netif_dormant(const struct net_device *dev)
3361 {
3362 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3363 }
3364 
3365 
3366 /**
3367  *	netif_oper_up - test if device is operational
3368  *	@dev: network device
3369  *
3370  * Check if carrier is operational
3371  */
3372 static inline bool netif_oper_up(const struct net_device *dev)
3373 {
3374 	return (dev->operstate == IF_OPER_UP ||
3375 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3376 }
3377 
3378 /**
3379  *	netif_device_present - is device available or removed
3380  *	@dev: network device
3381  *
3382  * Check if device has not been removed from system.
3383  */
3384 static inline bool netif_device_present(struct net_device *dev)
3385 {
3386 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3387 }
3388 
3389 void netif_device_detach(struct net_device *dev);
3390 
3391 void netif_device_attach(struct net_device *dev);
3392 
3393 /*
3394  * Network interface message level settings
3395  */
3396 
3397 enum {
3398 	NETIF_MSG_DRV		= 0x0001,
3399 	NETIF_MSG_PROBE		= 0x0002,
3400 	NETIF_MSG_LINK		= 0x0004,
3401 	NETIF_MSG_TIMER		= 0x0008,
3402 	NETIF_MSG_IFDOWN	= 0x0010,
3403 	NETIF_MSG_IFUP		= 0x0020,
3404 	NETIF_MSG_RX_ERR	= 0x0040,
3405 	NETIF_MSG_TX_ERR	= 0x0080,
3406 	NETIF_MSG_TX_QUEUED	= 0x0100,
3407 	NETIF_MSG_INTR		= 0x0200,
3408 	NETIF_MSG_TX_DONE	= 0x0400,
3409 	NETIF_MSG_RX_STATUS	= 0x0800,
3410 	NETIF_MSG_PKTDATA	= 0x1000,
3411 	NETIF_MSG_HW		= 0x2000,
3412 	NETIF_MSG_WOL		= 0x4000,
3413 };
3414 
3415 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3416 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3417 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3418 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3419 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3420 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3421 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3422 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3423 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3424 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3425 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3426 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3427 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3428 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3429 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3430 
3431 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3432 {
3433 	/* use default */
3434 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3435 		return default_msg_enable_bits;
3436 	if (debug_value == 0)	/* no output */
3437 		return 0;
3438 	/* set low N bits */
3439 	return (1 << debug_value) - 1;
3440 }
3441 
3442 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3443 {
3444 	spin_lock(&txq->_xmit_lock);
3445 	txq->xmit_lock_owner = cpu;
3446 }
3447 
3448 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3449 {
3450 	spin_lock_bh(&txq->_xmit_lock);
3451 	txq->xmit_lock_owner = smp_processor_id();
3452 }
3453 
3454 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3455 {
3456 	bool ok = spin_trylock(&txq->_xmit_lock);
3457 	if (likely(ok))
3458 		txq->xmit_lock_owner = smp_processor_id();
3459 	return ok;
3460 }
3461 
3462 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3463 {
3464 	txq->xmit_lock_owner = -1;
3465 	spin_unlock(&txq->_xmit_lock);
3466 }
3467 
3468 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3469 {
3470 	txq->xmit_lock_owner = -1;
3471 	spin_unlock_bh(&txq->_xmit_lock);
3472 }
3473 
3474 static inline void txq_trans_update(struct netdev_queue *txq)
3475 {
3476 	if (txq->xmit_lock_owner != -1)
3477 		txq->trans_start = jiffies;
3478 }
3479 
3480 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3481 static inline void netif_trans_update(struct net_device *dev)
3482 {
3483 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3484 
3485 	if (txq->trans_start != jiffies)
3486 		txq->trans_start = jiffies;
3487 }
3488 
3489 /**
3490  *	netif_tx_lock - grab network device transmit lock
3491  *	@dev: network device
3492  *
3493  * Get network device transmit lock
3494  */
3495 static inline void netif_tx_lock(struct net_device *dev)
3496 {
3497 	unsigned int i;
3498 	int cpu;
3499 
3500 	spin_lock(&dev->tx_global_lock);
3501 	cpu = smp_processor_id();
3502 	for (i = 0; i < dev->num_tx_queues; i++) {
3503 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3504 
3505 		/* We are the only thread of execution doing a
3506 		 * freeze, but we have to grab the _xmit_lock in
3507 		 * order to synchronize with threads which are in
3508 		 * the ->hard_start_xmit() handler and already
3509 		 * checked the frozen bit.
3510 		 */
3511 		__netif_tx_lock(txq, cpu);
3512 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3513 		__netif_tx_unlock(txq);
3514 	}
3515 }
3516 
3517 static inline void netif_tx_lock_bh(struct net_device *dev)
3518 {
3519 	local_bh_disable();
3520 	netif_tx_lock(dev);
3521 }
3522 
3523 static inline void netif_tx_unlock(struct net_device *dev)
3524 {
3525 	unsigned int i;
3526 
3527 	for (i = 0; i < dev->num_tx_queues; i++) {
3528 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3529 
3530 		/* No need to grab the _xmit_lock here.  If the
3531 		 * queue is not stopped for another reason, we
3532 		 * force a schedule.
3533 		 */
3534 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3535 		netif_schedule_queue(txq);
3536 	}
3537 	spin_unlock(&dev->tx_global_lock);
3538 }
3539 
3540 static inline void netif_tx_unlock_bh(struct net_device *dev)
3541 {
3542 	netif_tx_unlock(dev);
3543 	local_bh_enable();
3544 }
3545 
3546 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3547 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3548 		__netif_tx_lock(txq, cpu);		\
3549 	}						\
3550 }
3551 
3552 #define HARD_TX_TRYLOCK(dev, txq)			\
3553 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3554 		__netif_tx_trylock(txq) :		\
3555 		true )
3556 
3557 #define HARD_TX_UNLOCK(dev, txq) {			\
3558 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3559 		__netif_tx_unlock(txq);			\
3560 	}						\
3561 }
3562 
3563 static inline void netif_tx_disable(struct net_device *dev)
3564 {
3565 	unsigned int i;
3566 	int cpu;
3567 
3568 	local_bh_disable();
3569 	cpu = smp_processor_id();
3570 	for (i = 0; i < dev->num_tx_queues; i++) {
3571 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3572 
3573 		__netif_tx_lock(txq, cpu);
3574 		netif_tx_stop_queue(txq);
3575 		__netif_tx_unlock(txq);
3576 	}
3577 	local_bh_enable();
3578 }
3579 
3580 static inline void netif_addr_lock(struct net_device *dev)
3581 {
3582 	spin_lock(&dev->addr_list_lock);
3583 }
3584 
3585 static inline void netif_addr_lock_nested(struct net_device *dev)
3586 {
3587 	int subclass = SINGLE_DEPTH_NESTING;
3588 
3589 	if (dev->netdev_ops->ndo_get_lock_subclass)
3590 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3591 
3592 	spin_lock_nested(&dev->addr_list_lock, subclass);
3593 }
3594 
3595 static inline void netif_addr_lock_bh(struct net_device *dev)
3596 {
3597 	spin_lock_bh(&dev->addr_list_lock);
3598 }
3599 
3600 static inline void netif_addr_unlock(struct net_device *dev)
3601 {
3602 	spin_unlock(&dev->addr_list_lock);
3603 }
3604 
3605 static inline void netif_addr_unlock_bh(struct net_device *dev)
3606 {
3607 	spin_unlock_bh(&dev->addr_list_lock);
3608 }
3609 
3610 /*
3611  * dev_addrs walker. Should be used only for read access. Call with
3612  * rcu_read_lock held.
3613  */
3614 #define for_each_dev_addr(dev, ha) \
3615 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3616 
3617 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3618 
3619 void ether_setup(struct net_device *dev);
3620 
3621 /* Support for loadable net-drivers */
3622 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3623 				    unsigned char name_assign_type,
3624 				    void (*setup)(struct net_device *),
3625 				    unsigned int txqs, unsigned int rxqs);
3626 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3627 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3628 
3629 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3630 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3631 			 count)
3632 
3633 int register_netdev(struct net_device *dev);
3634 void unregister_netdev(struct net_device *dev);
3635 
3636 /* General hardware address lists handling functions */
3637 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3638 		   struct netdev_hw_addr_list *from_list, int addr_len);
3639 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3640 		      struct netdev_hw_addr_list *from_list, int addr_len);
3641 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3642 		       struct net_device *dev,
3643 		       int (*sync)(struct net_device *, const unsigned char *),
3644 		       int (*unsync)(struct net_device *,
3645 				     const unsigned char *));
3646 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3647 			  struct net_device *dev,
3648 			  int (*unsync)(struct net_device *,
3649 					const unsigned char *));
3650 void __hw_addr_init(struct netdev_hw_addr_list *list);
3651 
3652 /* Functions used for device addresses handling */
3653 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3654 		 unsigned char addr_type);
3655 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3656 		 unsigned char addr_type);
3657 void dev_addr_flush(struct net_device *dev);
3658 int dev_addr_init(struct net_device *dev);
3659 
3660 /* Functions used for unicast addresses handling */
3661 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3662 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3663 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3664 int dev_uc_sync(struct net_device *to, struct net_device *from);
3665 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3666 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3667 void dev_uc_flush(struct net_device *dev);
3668 void dev_uc_init(struct net_device *dev);
3669 
3670 /**
3671  *  __dev_uc_sync - Synchonize device's unicast list
3672  *  @dev:  device to sync
3673  *  @sync: function to call if address should be added
3674  *  @unsync: function to call if address should be removed
3675  *
3676  *  Add newly added addresses to the interface, and release
3677  *  addresses that have been deleted.
3678  */
3679 static inline int __dev_uc_sync(struct net_device *dev,
3680 				int (*sync)(struct net_device *,
3681 					    const unsigned char *),
3682 				int (*unsync)(struct net_device *,
3683 					      const unsigned char *))
3684 {
3685 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3686 }
3687 
3688 /**
3689  *  __dev_uc_unsync - Remove synchronized addresses from device
3690  *  @dev:  device to sync
3691  *  @unsync: function to call if address should be removed
3692  *
3693  *  Remove all addresses that were added to the device by dev_uc_sync().
3694  */
3695 static inline void __dev_uc_unsync(struct net_device *dev,
3696 				   int (*unsync)(struct net_device *,
3697 						 const unsigned char *))
3698 {
3699 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3700 }
3701 
3702 /* Functions used for multicast addresses handling */
3703 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3704 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3705 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3706 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3707 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3708 int dev_mc_sync(struct net_device *to, struct net_device *from);
3709 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3710 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3711 void dev_mc_flush(struct net_device *dev);
3712 void dev_mc_init(struct net_device *dev);
3713 
3714 /**
3715  *  __dev_mc_sync - Synchonize device's multicast list
3716  *  @dev:  device to sync
3717  *  @sync: function to call if address should be added
3718  *  @unsync: function to call if address should be removed
3719  *
3720  *  Add newly added addresses to the interface, and release
3721  *  addresses that have been deleted.
3722  */
3723 static inline int __dev_mc_sync(struct net_device *dev,
3724 				int (*sync)(struct net_device *,
3725 					    const unsigned char *),
3726 				int (*unsync)(struct net_device *,
3727 					      const unsigned char *))
3728 {
3729 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3730 }
3731 
3732 /**
3733  *  __dev_mc_unsync - Remove synchronized addresses from device
3734  *  @dev:  device to sync
3735  *  @unsync: function to call if address should be removed
3736  *
3737  *  Remove all addresses that were added to the device by dev_mc_sync().
3738  */
3739 static inline void __dev_mc_unsync(struct net_device *dev,
3740 				   int (*unsync)(struct net_device *,
3741 						 const unsigned char *))
3742 {
3743 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3744 }
3745 
3746 /* Functions used for secondary unicast and multicast support */
3747 void dev_set_rx_mode(struct net_device *dev);
3748 void __dev_set_rx_mode(struct net_device *dev);
3749 int dev_set_promiscuity(struct net_device *dev, int inc);
3750 int dev_set_allmulti(struct net_device *dev, int inc);
3751 void netdev_state_change(struct net_device *dev);
3752 void netdev_notify_peers(struct net_device *dev);
3753 void netdev_features_change(struct net_device *dev);
3754 /* Load a device via the kmod */
3755 void dev_load(struct net *net, const char *name);
3756 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3757 					struct rtnl_link_stats64 *storage);
3758 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3759 			     const struct net_device_stats *netdev_stats);
3760 
3761 extern int		netdev_max_backlog;
3762 extern int		netdev_tstamp_prequeue;
3763 extern int		weight_p;
3764 
3765 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3766 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3767 						     struct list_head **iter);
3768 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3769 						     struct list_head **iter);
3770 
3771 /* iterate through upper list, must be called under RCU read lock */
3772 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3773 	for (iter = &(dev)->adj_list.upper, \
3774 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3775 	     updev; \
3776 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3777 
3778 /* iterate through upper list, must be called under RCU read lock */
3779 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3780 	for (iter = &(dev)->all_adj_list.upper, \
3781 	     updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3782 	     updev; \
3783 	     updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3784 
3785 void *netdev_lower_get_next_private(struct net_device *dev,
3786 				    struct list_head **iter);
3787 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3788 					struct list_head **iter);
3789 
3790 #define netdev_for_each_lower_private(dev, priv, iter) \
3791 	for (iter = (dev)->adj_list.lower.next, \
3792 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3793 	     priv; \
3794 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3795 
3796 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3797 	for (iter = &(dev)->adj_list.lower, \
3798 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3799 	     priv; \
3800 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3801 
3802 void *netdev_lower_get_next(struct net_device *dev,
3803 				struct list_head **iter);
3804 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3805 	for (iter = (dev)->adj_list.lower.next, \
3806 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3807 	     ldev; \
3808 	     ldev = netdev_lower_get_next(dev, &(iter)))
3809 
3810 void *netdev_adjacent_get_private(struct list_head *adj_list);
3811 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3812 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3813 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3814 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3815 int netdev_master_upper_dev_link(struct net_device *dev,
3816 				 struct net_device *upper_dev,
3817 				 void *upper_priv, void *upper_info);
3818 void netdev_upper_dev_unlink(struct net_device *dev,
3819 			     struct net_device *upper_dev);
3820 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3821 void *netdev_lower_dev_get_private(struct net_device *dev,
3822 				   struct net_device *lower_dev);
3823 void netdev_lower_state_changed(struct net_device *lower_dev,
3824 				void *lower_state_info);
3825 
3826 /* RSS keys are 40 or 52 bytes long */
3827 #define NETDEV_RSS_KEY_LEN 52
3828 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3829 void netdev_rss_key_fill(void *buffer, size_t len);
3830 
3831 int dev_get_nest_level(struct net_device *dev,
3832 		       bool (*type_check)(const struct net_device *dev));
3833 int skb_checksum_help(struct sk_buff *skb);
3834 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3835 				  netdev_features_t features, bool tx_path);
3836 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3837 				    netdev_features_t features);
3838 
3839 struct netdev_bonding_info {
3840 	ifslave	slave;
3841 	ifbond	master;
3842 };
3843 
3844 struct netdev_notifier_bonding_info {
3845 	struct netdev_notifier_info info; /* must be first */
3846 	struct netdev_bonding_info  bonding_info;
3847 };
3848 
3849 void netdev_bonding_info_change(struct net_device *dev,
3850 				struct netdev_bonding_info *bonding_info);
3851 
3852 static inline
3853 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3854 {
3855 	return __skb_gso_segment(skb, features, true);
3856 }
3857 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3858 
3859 static inline bool can_checksum_protocol(netdev_features_t features,
3860 					 __be16 protocol)
3861 {
3862 	if (protocol == htons(ETH_P_FCOE))
3863 		return !!(features & NETIF_F_FCOE_CRC);
3864 
3865 	/* Assume this is an IP checksum (not SCTP CRC) */
3866 
3867 	if (features & NETIF_F_HW_CSUM) {
3868 		/* Can checksum everything */
3869 		return true;
3870 	}
3871 
3872 	switch (protocol) {
3873 	case htons(ETH_P_IP):
3874 		return !!(features & NETIF_F_IP_CSUM);
3875 	case htons(ETH_P_IPV6):
3876 		return !!(features & NETIF_F_IPV6_CSUM);
3877 	default:
3878 		return false;
3879 	}
3880 }
3881 
3882 /* Map an ethertype into IP protocol if possible */
3883 static inline int eproto_to_ipproto(int eproto)
3884 {
3885 	switch (eproto) {
3886 	case htons(ETH_P_IP):
3887 		return IPPROTO_IP;
3888 	case htons(ETH_P_IPV6):
3889 		return IPPROTO_IPV6;
3890 	default:
3891 		return -1;
3892 	}
3893 }
3894 
3895 #ifdef CONFIG_BUG
3896 void netdev_rx_csum_fault(struct net_device *dev);
3897 #else
3898 static inline void netdev_rx_csum_fault(struct net_device *dev)
3899 {
3900 }
3901 #endif
3902 /* rx skb timestamps */
3903 void net_enable_timestamp(void);
3904 void net_disable_timestamp(void);
3905 
3906 #ifdef CONFIG_PROC_FS
3907 int __init dev_proc_init(void);
3908 #else
3909 #define dev_proc_init() 0
3910 #endif
3911 
3912 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3913 					      struct sk_buff *skb, struct net_device *dev,
3914 					      bool more)
3915 {
3916 	skb->xmit_more = more ? 1 : 0;
3917 	return ops->ndo_start_xmit(skb, dev);
3918 }
3919 
3920 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3921 					    struct netdev_queue *txq, bool more)
3922 {
3923 	const struct net_device_ops *ops = dev->netdev_ops;
3924 	int rc;
3925 
3926 	rc = __netdev_start_xmit(ops, skb, dev, more);
3927 	if (rc == NETDEV_TX_OK)
3928 		txq_trans_update(txq);
3929 
3930 	return rc;
3931 }
3932 
3933 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3934 				const void *ns);
3935 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3936 				 const void *ns);
3937 
3938 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3939 {
3940 	return netdev_class_create_file_ns(class_attr, NULL);
3941 }
3942 
3943 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3944 {
3945 	netdev_class_remove_file_ns(class_attr, NULL);
3946 }
3947 
3948 extern struct kobj_ns_type_operations net_ns_type_operations;
3949 
3950 const char *netdev_drivername(const struct net_device *dev);
3951 
3952 void linkwatch_run_queue(void);
3953 
3954 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3955 							  netdev_features_t f2)
3956 {
3957 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3958 		if (f1 & NETIF_F_HW_CSUM)
3959 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3960 		else
3961 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3962 	}
3963 
3964 	return f1 & f2;
3965 }
3966 
3967 static inline netdev_features_t netdev_get_wanted_features(
3968 	struct net_device *dev)
3969 {
3970 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
3971 }
3972 netdev_features_t netdev_increment_features(netdev_features_t all,
3973 	netdev_features_t one, netdev_features_t mask);
3974 
3975 /* Allow TSO being used on stacked device :
3976  * Performing the GSO segmentation before last device
3977  * is a performance improvement.
3978  */
3979 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3980 							netdev_features_t mask)
3981 {
3982 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3983 }
3984 
3985 int __netdev_update_features(struct net_device *dev);
3986 void netdev_update_features(struct net_device *dev);
3987 void netdev_change_features(struct net_device *dev);
3988 
3989 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3990 					struct net_device *dev);
3991 
3992 netdev_features_t passthru_features_check(struct sk_buff *skb,
3993 					  struct net_device *dev,
3994 					  netdev_features_t features);
3995 netdev_features_t netif_skb_features(struct sk_buff *skb);
3996 
3997 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3998 {
3999 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4000 
4001 	/* check flags correspondence */
4002 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4003 	BUILD_BUG_ON(SKB_GSO_UDP     != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4004 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4005 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4006 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4007 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4008 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4009 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4010 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4011 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4012 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4013 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4014 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4015 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4016 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4017 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4018 
4019 	return (features & feature) == feature;
4020 }
4021 
4022 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4023 {
4024 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4025 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4026 }
4027 
4028 static inline bool netif_needs_gso(struct sk_buff *skb,
4029 				   netdev_features_t features)
4030 {
4031 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4032 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4033 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4034 }
4035 
4036 static inline void netif_set_gso_max_size(struct net_device *dev,
4037 					  unsigned int size)
4038 {
4039 	dev->gso_max_size = size;
4040 }
4041 
4042 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4043 					int pulled_hlen, u16 mac_offset,
4044 					int mac_len)
4045 {
4046 	skb->protocol = protocol;
4047 	skb->encapsulation = 1;
4048 	skb_push(skb, pulled_hlen);
4049 	skb_reset_transport_header(skb);
4050 	skb->mac_header = mac_offset;
4051 	skb->network_header = skb->mac_header + mac_len;
4052 	skb->mac_len = mac_len;
4053 }
4054 
4055 static inline bool netif_is_macsec(const struct net_device *dev)
4056 {
4057 	return dev->priv_flags & IFF_MACSEC;
4058 }
4059 
4060 static inline bool netif_is_macvlan(const struct net_device *dev)
4061 {
4062 	return dev->priv_flags & IFF_MACVLAN;
4063 }
4064 
4065 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4066 {
4067 	return dev->priv_flags & IFF_MACVLAN_PORT;
4068 }
4069 
4070 static inline bool netif_is_ipvlan(const struct net_device *dev)
4071 {
4072 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4073 }
4074 
4075 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4076 {
4077 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4078 }
4079 
4080 static inline bool netif_is_bond_master(const struct net_device *dev)
4081 {
4082 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4083 }
4084 
4085 static inline bool netif_is_bond_slave(const struct net_device *dev)
4086 {
4087 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4088 }
4089 
4090 static inline bool netif_supports_nofcs(struct net_device *dev)
4091 {
4092 	return dev->priv_flags & IFF_SUPP_NOFCS;
4093 }
4094 
4095 static inline bool netif_is_l3_master(const struct net_device *dev)
4096 {
4097 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4098 }
4099 
4100 static inline bool netif_is_l3_slave(const struct net_device *dev)
4101 {
4102 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4103 }
4104 
4105 static inline bool netif_is_bridge_master(const struct net_device *dev)
4106 {
4107 	return dev->priv_flags & IFF_EBRIDGE;
4108 }
4109 
4110 static inline bool netif_is_bridge_port(const struct net_device *dev)
4111 {
4112 	return dev->priv_flags & IFF_BRIDGE_PORT;
4113 }
4114 
4115 static inline bool netif_is_ovs_master(const struct net_device *dev)
4116 {
4117 	return dev->priv_flags & IFF_OPENVSWITCH;
4118 }
4119 
4120 static inline bool netif_is_team_master(const struct net_device *dev)
4121 {
4122 	return dev->priv_flags & IFF_TEAM;
4123 }
4124 
4125 static inline bool netif_is_team_port(const struct net_device *dev)
4126 {
4127 	return dev->priv_flags & IFF_TEAM_PORT;
4128 }
4129 
4130 static inline bool netif_is_lag_master(const struct net_device *dev)
4131 {
4132 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4133 }
4134 
4135 static inline bool netif_is_lag_port(const struct net_device *dev)
4136 {
4137 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4138 }
4139 
4140 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4141 {
4142 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4143 }
4144 
4145 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4146 static inline void netif_keep_dst(struct net_device *dev)
4147 {
4148 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4149 }
4150 
4151 extern struct pernet_operations __net_initdata loopback_net_ops;
4152 
4153 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4154 
4155 /* netdev_printk helpers, similar to dev_printk */
4156 
4157 static inline const char *netdev_name(const struct net_device *dev)
4158 {
4159 	if (!dev->name[0] || strchr(dev->name, '%'))
4160 		return "(unnamed net_device)";
4161 	return dev->name;
4162 }
4163 
4164 static inline const char *netdev_reg_state(const struct net_device *dev)
4165 {
4166 	switch (dev->reg_state) {
4167 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4168 	case NETREG_REGISTERED: return "";
4169 	case NETREG_UNREGISTERING: return " (unregistering)";
4170 	case NETREG_UNREGISTERED: return " (unregistered)";
4171 	case NETREG_RELEASED: return " (released)";
4172 	case NETREG_DUMMY: return " (dummy)";
4173 	}
4174 
4175 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4176 	return " (unknown)";
4177 }
4178 
4179 __printf(3, 4)
4180 void netdev_printk(const char *level, const struct net_device *dev,
4181 		   const char *format, ...);
4182 __printf(2, 3)
4183 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4184 __printf(2, 3)
4185 void netdev_alert(const struct net_device *dev, const char *format, ...);
4186 __printf(2, 3)
4187 void netdev_crit(const struct net_device *dev, const char *format, ...);
4188 __printf(2, 3)
4189 void netdev_err(const struct net_device *dev, const char *format, ...);
4190 __printf(2, 3)
4191 void netdev_warn(const struct net_device *dev, const char *format, ...);
4192 __printf(2, 3)
4193 void netdev_notice(const struct net_device *dev, const char *format, ...);
4194 __printf(2, 3)
4195 void netdev_info(const struct net_device *dev, const char *format, ...);
4196 
4197 #define MODULE_ALIAS_NETDEV(device) \
4198 	MODULE_ALIAS("netdev-" device)
4199 
4200 #if defined(CONFIG_DYNAMIC_DEBUG)
4201 #define netdev_dbg(__dev, format, args...)			\
4202 do {								\
4203 	dynamic_netdev_dbg(__dev, format, ##args);		\
4204 } while (0)
4205 #elif defined(DEBUG)
4206 #define netdev_dbg(__dev, format, args...)			\
4207 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4208 #else
4209 #define netdev_dbg(__dev, format, args...)			\
4210 ({								\
4211 	if (0)							\
4212 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4213 })
4214 #endif
4215 
4216 #if defined(VERBOSE_DEBUG)
4217 #define netdev_vdbg	netdev_dbg
4218 #else
4219 
4220 #define netdev_vdbg(dev, format, args...)			\
4221 ({								\
4222 	if (0)							\
4223 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4224 	0;							\
4225 })
4226 #endif
4227 
4228 /*
4229  * netdev_WARN() acts like dev_printk(), but with the key difference
4230  * of using a WARN/WARN_ON to get the message out, including the
4231  * file/line information and a backtrace.
4232  */
4233 #define netdev_WARN(dev, format, args...)			\
4234 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4235 	     netdev_reg_state(dev), ##args)
4236 
4237 /* netif printk helpers, similar to netdev_printk */
4238 
4239 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4240 do {					  			\
4241 	if (netif_msg_##type(priv))				\
4242 		netdev_printk(level, (dev), fmt, ##args);	\
4243 } while (0)
4244 
4245 #define netif_level(level, priv, type, dev, fmt, args...)	\
4246 do {								\
4247 	if (netif_msg_##type(priv))				\
4248 		netdev_##level(dev, fmt, ##args);		\
4249 } while (0)
4250 
4251 #define netif_emerg(priv, type, dev, fmt, args...)		\
4252 	netif_level(emerg, priv, type, dev, fmt, ##args)
4253 #define netif_alert(priv, type, dev, fmt, args...)		\
4254 	netif_level(alert, priv, type, dev, fmt, ##args)
4255 #define netif_crit(priv, type, dev, fmt, args...)		\
4256 	netif_level(crit, priv, type, dev, fmt, ##args)
4257 #define netif_err(priv, type, dev, fmt, args...)		\
4258 	netif_level(err, priv, type, dev, fmt, ##args)
4259 #define netif_warn(priv, type, dev, fmt, args...)		\
4260 	netif_level(warn, priv, type, dev, fmt, ##args)
4261 #define netif_notice(priv, type, dev, fmt, args...)		\
4262 	netif_level(notice, priv, type, dev, fmt, ##args)
4263 #define netif_info(priv, type, dev, fmt, args...)		\
4264 	netif_level(info, priv, type, dev, fmt, ##args)
4265 
4266 #if defined(CONFIG_DYNAMIC_DEBUG)
4267 #define netif_dbg(priv, type, netdev, format, args...)		\
4268 do {								\
4269 	if (netif_msg_##type(priv))				\
4270 		dynamic_netdev_dbg(netdev, format, ##args);	\
4271 } while (0)
4272 #elif defined(DEBUG)
4273 #define netif_dbg(priv, type, dev, format, args...)		\
4274 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4275 #else
4276 #define netif_dbg(priv, type, dev, format, args...)			\
4277 ({									\
4278 	if (0)								\
4279 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4280 	0;								\
4281 })
4282 #endif
4283 
4284 #if defined(VERBOSE_DEBUG)
4285 #define netif_vdbg	netif_dbg
4286 #else
4287 #define netif_vdbg(priv, type, dev, format, args...)		\
4288 ({								\
4289 	if (0)							\
4290 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4291 	0;							\
4292 })
4293 #endif
4294 
4295 /*
4296  *	The list of packet types we will receive (as opposed to discard)
4297  *	and the routines to invoke.
4298  *
4299  *	Why 16. Because with 16 the only overlap we get on a hash of the
4300  *	low nibble of the protocol value is RARP/SNAP/X.25.
4301  *
4302  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4303  *             sure which should go first, but I bet it won't make much
4304  *             difference if we are running VLANs.  The good news is that
4305  *             this protocol won't be in the list unless compiled in, so
4306  *             the average user (w/out VLANs) will not be adversely affected.
4307  *             --BLG
4308  *
4309  *		0800	IP
4310  *		8100    802.1Q VLAN
4311  *		0001	802.3
4312  *		0002	AX.25
4313  *		0004	802.2
4314  *		8035	RARP
4315  *		0005	SNAP
4316  *		0805	X.25
4317  *		0806	ARP
4318  *		8137	IPX
4319  *		0009	Localtalk
4320  *		86DD	IPv6
4321  */
4322 #define PTYPE_HASH_SIZE	(16)
4323 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4324 
4325 #endif	/* _LINUX_NETDEVICE_H */
4326