xref: /linux-6.15/include/linux/netdevice.h (revision 166cc713)
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(), 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 	NETDEV_TX_LOCKED = 0x20,	/* driver tx lock was already taken */
110 };
111 typedef enum netdev_tx netdev_tx_t;
112 
113 /*
114  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
115  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116  */
117 static inline bool dev_xmit_complete(int rc)
118 {
119 	/*
120 	 * Positive cases with an skb consumed by a driver:
121 	 * - successful transmission (rc == NETDEV_TX_OK)
122 	 * - error while transmitting (rc < 0)
123 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 	 */
125 	if (likely(rc < NET_XMIT_MASK))
126 		return true;
127 
128 	return false;
129 }
130 
131 /*
132  *	Compute the worst case header length according to the protocols
133  *	used.
134  */
135 
136 #if defined(CONFIG_HYPERV_NET)
137 # define LL_MAX_HEADER 128
138 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
139 # if defined(CONFIG_MAC80211_MESH)
140 #  define LL_MAX_HEADER 128
141 # else
142 #  define LL_MAX_HEADER 96
143 # endif
144 #else
145 # define LL_MAX_HEADER 32
146 #endif
147 
148 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
149     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
150 #define MAX_HEADER LL_MAX_HEADER
151 #else
152 #define MAX_HEADER (LL_MAX_HEADER + 48)
153 #endif
154 
155 /*
156  *	Old network device statistics. Fields are native words
157  *	(unsigned long) so they can be read and written atomically.
158  */
159 
160 struct net_device_stats {
161 	unsigned long	rx_packets;
162 	unsigned long	tx_packets;
163 	unsigned long	rx_bytes;
164 	unsigned long	tx_bytes;
165 	unsigned long	rx_errors;
166 	unsigned long	tx_errors;
167 	unsigned long	rx_dropped;
168 	unsigned long	tx_dropped;
169 	unsigned long	multicast;
170 	unsigned long	collisions;
171 	unsigned long	rx_length_errors;
172 	unsigned long	rx_over_errors;
173 	unsigned long	rx_crc_errors;
174 	unsigned long	rx_frame_errors;
175 	unsigned long	rx_fifo_errors;
176 	unsigned long	rx_missed_errors;
177 	unsigned long	tx_aborted_errors;
178 	unsigned long	tx_carrier_errors;
179 	unsigned long	tx_fifo_errors;
180 	unsigned long	tx_heartbeat_errors;
181 	unsigned long	tx_window_errors;
182 	unsigned long	rx_compressed;
183 	unsigned long	tx_compressed;
184 };
185 
186 
187 #include <linux/cache.h>
188 #include <linux/skbuff.h>
189 
190 #ifdef CONFIG_RPS
191 #include <linux/static_key.h>
192 extern struct static_key rps_needed;
193 #endif
194 
195 struct neighbour;
196 struct neigh_parms;
197 struct sk_buff;
198 
199 struct netdev_hw_addr {
200 	struct list_head	list;
201 	unsigned char		addr[MAX_ADDR_LEN];
202 	unsigned char		type;
203 #define NETDEV_HW_ADDR_T_LAN		1
204 #define NETDEV_HW_ADDR_T_SAN		2
205 #define NETDEV_HW_ADDR_T_SLAVE		3
206 #define NETDEV_HW_ADDR_T_UNICAST	4
207 #define NETDEV_HW_ADDR_T_MULTICAST	5
208 	bool			global_use;
209 	int			sync_cnt;
210 	int			refcount;
211 	int			synced;
212 	struct rcu_head		rcu_head;
213 };
214 
215 struct netdev_hw_addr_list {
216 	struct list_head	list;
217 	int			count;
218 };
219 
220 #define netdev_hw_addr_list_count(l) ((l)->count)
221 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
222 #define netdev_hw_addr_list_for_each(ha, l) \
223 	list_for_each_entry(ha, &(l)->list, list)
224 
225 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
226 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
227 #define netdev_for_each_uc_addr(ha, dev) \
228 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
229 
230 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
231 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
232 #define netdev_for_each_mc_addr(ha, dev) \
233 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
234 
235 struct hh_cache {
236 	u16		hh_len;
237 	u16		__pad;
238 	seqlock_t	hh_lock;
239 
240 	/* cached hardware header; allow for machine alignment needs.        */
241 #define HH_DATA_MOD	16
242 #define HH_DATA_OFF(__len) \
243 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
244 #define HH_DATA_ALIGN(__len) \
245 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
246 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
247 };
248 
249 /* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
250  * Alternative is:
251  *   dev->hard_header_len ? (dev->hard_header_len +
252  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
253  *
254  * We could use other alignment values, but we must maintain the
255  * relationship HH alignment <= LL alignment.
256  */
257 #define LL_RESERVED_SPACE(dev) \
258 	((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
259 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
260 	((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
261 
262 struct header_ops {
263 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
264 			   unsigned short type, const void *daddr,
265 			   const void *saddr, unsigned int len);
266 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
267 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
268 	void	(*cache_update)(struct hh_cache *hh,
269 				const struct net_device *dev,
270 				const unsigned char *haddr);
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 at 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, passe 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 to 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 ways.
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 consider 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 changed skb->dev, but want 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
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 /*
573  * write mostly part
574  */
575 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
576 	int			xmit_lock_owner;
577 	/*
578 	 * please use this field instead of dev->trans_start
579 	 */
580 	unsigned long		trans_start;
581 
582 	/*
583 	 * Number of TX timeouts for this queue
584 	 * (/sys/class/net/DEV/Q/trans_timeout)
585 	 */
586 	unsigned long		trans_timeout;
587 
588 	unsigned long		state;
589 
590 #ifdef CONFIG_BQL
591 	struct dql		dql;
592 #endif
593 	unsigned long		tx_maxrate;
594 } ____cacheline_aligned_in_smp;
595 
596 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
597 {
598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 	return q->numa_node;
600 #else
601 	return NUMA_NO_NODE;
602 #endif
603 }
604 
605 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
606 {
607 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 	q->numa_node = node;
609 #endif
610 }
611 
612 #ifdef CONFIG_RPS
613 /*
614  * This structure holds an RPS map which can be of variable length.  The
615  * map is an array of CPUs.
616  */
617 struct rps_map {
618 	unsigned int len;
619 	struct rcu_head rcu;
620 	u16 cpus[0];
621 };
622 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
623 
624 /*
625  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
626  * tail pointer for that CPU's input queue at the time of last enqueue, and
627  * a hardware filter index.
628  */
629 struct rps_dev_flow {
630 	u16 cpu;
631 	u16 filter;
632 	unsigned int last_qtail;
633 };
634 #define RPS_NO_FILTER 0xffff
635 
636 /*
637  * The rps_dev_flow_table structure contains a table of flow mappings.
638  */
639 struct rps_dev_flow_table {
640 	unsigned int mask;
641 	struct rcu_head rcu;
642 	struct rps_dev_flow flows[0];
643 };
644 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
645     ((_num) * sizeof(struct rps_dev_flow)))
646 
647 /*
648  * The rps_sock_flow_table contains mappings of flows to the last CPU
649  * on which they were processed by the application (set in recvmsg).
650  * Each entry is a 32bit value. Upper part is the high order bits
651  * of flow hash, lower part is cpu number.
652  * rps_cpu_mask is used to partition the space, depending on number of
653  * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
654  * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
655  * meaning we use 32-6=26 bits for the hash.
656  */
657 struct rps_sock_flow_table {
658 	u32	mask;
659 
660 	u32	ents[0] ____cacheline_aligned_in_smp;
661 };
662 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
663 
664 #define RPS_NO_CPU 0xffff
665 
666 extern u32 rps_cpu_mask;
667 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
668 
669 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
670 					u32 hash)
671 {
672 	if (table && hash) {
673 		unsigned int index = hash & table->mask;
674 		u32 val = hash & ~rps_cpu_mask;
675 
676 		/* We only give a hint, preemption can change cpu under us */
677 		val |= raw_smp_processor_id();
678 
679 		if (table->ents[index] != val)
680 			table->ents[index] = val;
681 	}
682 }
683 
684 #ifdef CONFIG_RFS_ACCEL
685 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
686 			 u16 filter_id);
687 #endif
688 #endif /* CONFIG_RPS */
689 
690 /* This structure contains an instance of an RX queue. */
691 struct netdev_rx_queue {
692 #ifdef CONFIG_RPS
693 	struct rps_map __rcu		*rps_map;
694 	struct rps_dev_flow_table __rcu	*rps_flow_table;
695 #endif
696 	struct kobject			kobj;
697 	struct net_device		*dev;
698 } ____cacheline_aligned_in_smp;
699 
700 /*
701  * RX queue sysfs structures and functions.
702  */
703 struct rx_queue_attribute {
704 	struct attribute attr;
705 	ssize_t (*show)(struct netdev_rx_queue *queue,
706 	    struct rx_queue_attribute *attr, char *buf);
707 	ssize_t (*store)(struct netdev_rx_queue *queue,
708 	    struct rx_queue_attribute *attr, const char *buf, size_t len);
709 };
710 
711 #ifdef CONFIG_XPS
712 /*
713  * This structure holds an XPS map which can be of variable length.  The
714  * map is an array of queues.
715  */
716 struct xps_map {
717 	unsigned int len;
718 	unsigned int alloc_len;
719 	struct rcu_head rcu;
720 	u16 queues[0];
721 };
722 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
723 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
724        - sizeof(struct xps_map)) / sizeof(u16))
725 
726 /*
727  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
728  */
729 struct xps_dev_maps {
730 	struct rcu_head rcu;
731 	struct xps_map __rcu *cpu_map[0];
732 };
733 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) +		\
734     (nr_cpu_ids * sizeof(struct xps_map *)))
735 #endif /* CONFIG_XPS */
736 
737 #define TC_MAX_QUEUE	16
738 #define TC_BITMASK	15
739 /* HW offloaded queuing disciplines txq count and offset maps */
740 struct netdev_tc_txq {
741 	u16 count;
742 	u16 offset;
743 };
744 
745 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
746 /*
747  * This structure is to hold information about the device
748  * configured to run FCoE protocol stack.
749  */
750 struct netdev_fcoe_hbainfo {
751 	char	manufacturer[64];
752 	char	serial_number[64];
753 	char	hardware_version[64];
754 	char	driver_version[64];
755 	char	optionrom_version[64];
756 	char	firmware_version[64];
757 	char	model[256];
758 	char	model_description[256];
759 };
760 #endif
761 
762 #define MAX_PHYS_ITEM_ID_LEN 32
763 
764 /* This structure holds a unique identifier to identify some
765  * physical item (port for example) used by a netdevice.
766  */
767 struct netdev_phys_item_id {
768 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
769 	unsigned char id_len;
770 };
771 
772 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
773 					    struct netdev_phys_item_id *b)
774 {
775 	return a->id_len == b->id_len &&
776 	       memcmp(a->id, b->id, a->id_len) == 0;
777 }
778 
779 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
780 				       struct sk_buff *skb);
781 
782 /* These structures hold the attributes of qdisc and classifiers
783  * that are being passed to the netdevice through the setup_tc op.
784  */
785 enum {
786 	TC_SETUP_MQPRIO,
787 	TC_SETUP_CLSU32,
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 	};
798 };
799 
800 
801 /*
802  * This structure defines the management hooks for network devices.
803  * The following hooks can be defined; unless noted otherwise, they are
804  * optional and can be filled with a null pointer.
805  *
806  * int (*ndo_init)(struct net_device *dev);
807  *     This function is called once when network device is registered.
808  *     The network device can use this to any late stage initializaton
809  *     or semantic validattion. It can fail with an error code which will
810  *     be propogated back to register_netdev
811  *
812  * void (*ndo_uninit)(struct net_device *dev);
813  *     This function is called when device is unregistered or when registration
814  *     fails. It is not called if init fails.
815  *
816  * int (*ndo_open)(struct net_device *dev);
817  *     This function is called when network device transistions to the up
818  *     state.
819  *
820  * int (*ndo_stop)(struct net_device *dev);
821  *     This function is called when network device transistions to the down
822  *     state.
823  *
824  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
825  *                               struct net_device *dev);
826  *	Called when a packet needs to be transmitted.
827  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
828  *	the queue before that can happen; it's for obsolete devices and weird
829  *	corner cases, but the stack really does a non-trivial amount
830  *	of useless work if you return NETDEV_TX_BUSY.
831  *        (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
832  *	Required can not 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 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 promiscious 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 to 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 request an ioctl which can't be handled by
864  *	the generic interface code. If not defined ioctl's 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 reason, 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 uses 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 support 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 support 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 adduce 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 notiy a driver about the UDP port and socket
1031  *	address family that vxlan is listnening 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 reset 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_rss_query_en)(
1177 						   struct net_device *dev,
1178 						   int vf, bool setting);
1179 	int			(*ndo_setup_tc)(struct net_device *dev,
1180 						u32 handle,
1181 						__be16 protocol,
1182 						struct tc_to_netdev *tc);
1183 #if IS_ENABLED(CONFIG_FCOE)
1184 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1185 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1186 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1187 						      u16 xid,
1188 						      struct scatterlist *sgl,
1189 						      unsigned int sgc);
1190 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1191 						     u16 xid);
1192 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1193 						       u16 xid,
1194 						       struct scatterlist *sgl,
1195 						       unsigned int sgc);
1196 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1197 							struct netdev_fcoe_hbainfo *hbainfo);
1198 #endif
1199 
1200 #if IS_ENABLED(CONFIG_LIBFCOE)
1201 #define NETDEV_FCOE_WWNN 0
1202 #define NETDEV_FCOE_WWPN 1
1203 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1204 						    u64 *wwn, int type);
1205 #endif
1206 
1207 #ifdef CONFIG_RFS_ACCEL
1208 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1209 						     const struct sk_buff *skb,
1210 						     u16 rxq_index,
1211 						     u32 flow_id);
1212 #endif
1213 	int			(*ndo_add_slave)(struct net_device *dev,
1214 						 struct net_device *slave_dev);
1215 	int			(*ndo_del_slave)(struct net_device *dev,
1216 						 struct net_device *slave_dev);
1217 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1218 						    netdev_features_t features);
1219 	int			(*ndo_set_features)(struct net_device *dev,
1220 						    netdev_features_t features);
1221 	int			(*ndo_neigh_construct)(struct neighbour *n);
1222 	void			(*ndo_neigh_destroy)(struct neighbour *n);
1223 
1224 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1225 					       struct nlattr *tb[],
1226 					       struct net_device *dev,
1227 					       const unsigned char *addr,
1228 					       u16 vid,
1229 					       u16 flags);
1230 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1231 					       struct nlattr *tb[],
1232 					       struct net_device *dev,
1233 					       const unsigned char *addr,
1234 					       u16 vid);
1235 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1236 						struct netlink_callback *cb,
1237 						struct net_device *dev,
1238 						struct net_device *filter_dev,
1239 						int idx);
1240 
1241 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1242 						      struct nlmsghdr *nlh,
1243 						      u16 flags);
1244 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1245 						      u32 pid, u32 seq,
1246 						      struct net_device *dev,
1247 						      u32 filter_mask,
1248 						      int nlflags);
1249 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1250 						      struct nlmsghdr *nlh,
1251 						      u16 flags);
1252 	int			(*ndo_change_carrier)(struct net_device *dev,
1253 						      bool new_carrier);
1254 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1255 							struct netdev_phys_item_id *ppid);
1256 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1257 							  char *name, size_t len);
1258 	void			(*ndo_add_vxlan_port)(struct  net_device *dev,
1259 						      sa_family_t sa_family,
1260 						      __be16 port);
1261 	void			(*ndo_del_vxlan_port)(struct  net_device *dev,
1262 						      sa_family_t sa_family,
1263 						      __be16 port);
1264 	void			(*ndo_add_geneve_port)(struct  net_device *dev,
1265 						       sa_family_t sa_family,
1266 						       __be16 port);
1267 	void			(*ndo_del_geneve_port)(struct  net_device *dev,
1268 						       sa_family_t sa_family,
1269 						       __be16 port);
1270 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1271 							struct net_device *dev);
1272 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1273 							void *priv);
1274 
1275 	netdev_tx_t		(*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1276 							struct net_device *dev,
1277 							void *priv);
1278 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1279 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1280 						      int queue_index,
1281 						      u32 maxrate);
1282 	int			(*ndo_get_iflink)(const struct net_device *dev);
1283 	int			(*ndo_change_proto_down)(struct net_device *dev,
1284 							 bool proto_down);
1285 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1286 						       struct sk_buff *skb);
1287 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1288 						       int needed_headroom);
1289 };
1290 
1291 /**
1292  * enum net_device_priv_flags - &struct net_device priv_flags
1293  *
1294  * These are the &struct net_device, they are only set internally
1295  * by drivers and used in the kernel. These flags are invisible to
1296  * userspace, this means that the order of these flags can change
1297  * during any kernel release.
1298  *
1299  * You should have a pretty good reason to be extending these flags.
1300  *
1301  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1302  * @IFF_EBRIDGE: Ethernet bridging device
1303  * @IFF_BONDING: bonding master or slave
1304  * @IFF_ISATAP: ISATAP interface (RFC4214)
1305  * @IFF_WAN_HDLC: WAN HDLC device
1306  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1307  *	release skb->dst
1308  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1309  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1310  * @IFF_MACVLAN_PORT: device used as macvlan port
1311  * @IFF_BRIDGE_PORT: device used as bridge port
1312  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1313  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1314  * @IFF_UNICAST_FLT: Supports unicast filtering
1315  * @IFF_TEAM_PORT: device used as team port
1316  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1317  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1318  *	change when it's running
1319  * @IFF_MACVLAN: Macvlan device
1320  * @IFF_L3MDEV_MASTER: device is an L3 master device
1321  * @IFF_NO_QUEUE: device can run without qdisc attached
1322  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1323  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1324  * @IFF_TEAM: device is a team device
1325  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1326  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1327  *	entity (i.e. the master device for bridged veth)
1328  */
1329 enum netdev_priv_flags {
1330 	IFF_802_1Q_VLAN			= 1<<0,
1331 	IFF_EBRIDGE			= 1<<1,
1332 	IFF_BONDING			= 1<<2,
1333 	IFF_ISATAP			= 1<<3,
1334 	IFF_WAN_HDLC			= 1<<4,
1335 	IFF_XMIT_DST_RELEASE		= 1<<5,
1336 	IFF_DONT_BRIDGE			= 1<<6,
1337 	IFF_DISABLE_NETPOLL		= 1<<7,
1338 	IFF_MACVLAN_PORT		= 1<<8,
1339 	IFF_BRIDGE_PORT			= 1<<9,
1340 	IFF_OVS_DATAPATH		= 1<<10,
1341 	IFF_TX_SKB_SHARING		= 1<<11,
1342 	IFF_UNICAST_FLT			= 1<<12,
1343 	IFF_TEAM_PORT			= 1<<13,
1344 	IFF_SUPP_NOFCS			= 1<<14,
1345 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1346 	IFF_MACVLAN			= 1<<16,
1347 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1348 	IFF_IPVLAN_MASTER		= 1<<18,
1349 	IFF_IPVLAN_SLAVE		= 1<<19,
1350 	IFF_L3MDEV_MASTER		= 1<<20,
1351 	IFF_NO_QUEUE			= 1<<21,
1352 	IFF_OPENVSWITCH			= 1<<22,
1353 	IFF_L3MDEV_SLAVE		= 1<<23,
1354 	IFF_TEAM			= 1<<24,
1355 	IFF_RXFH_CONFIGURED		= 1<<25,
1356 	IFF_PHONY_HEADROOM		= 1<<26,
1357 };
1358 
1359 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1360 #define IFF_EBRIDGE			IFF_EBRIDGE
1361 #define IFF_BONDING			IFF_BONDING
1362 #define IFF_ISATAP			IFF_ISATAP
1363 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1364 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1365 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1366 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1367 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1368 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1369 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1370 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1371 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1372 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1373 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1374 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1375 #define IFF_MACVLAN			IFF_MACVLAN
1376 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1377 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1378 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1379 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1380 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1381 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1382 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1383 #define IFF_TEAM			IFF_TEAM
1384 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1385 
1386 /**
1387  *	struct net_device - The DEVICE structure.
1388  *		Actually, this whole structure is a big mistake.  It mixes I/O
1389  *		data with strictly "high-level" data, and it has to know about
1390  *		almost every data structure used in the INET module.
1391  *
1392  *	@name:	This is the first field of the "visible" part of this structure
1393  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1394  *	 	of the interface.
1395  *
1396  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1397  *	@ifalias:	SNMP alias
1398  *	@mem_end:	Shared memory end
1399  *	@mem_start:	Shared memory start
1400  *	@base_addr:	Device I/O address
1401  *	@irq:		Device IRQ number
1402  *
1403  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1404  *
1405  *	@state:		Generic network queuing layer state, see netdev_state_t
1406  *	@dev_list:	The global list of network devices
1407  *	@napi_list:	List entry, that is used for polling napi devices
1408  *	@unreg_list:	List entry, that is used, when we are unregistering the
1409  *			device, see the function unregister_netdev
1410  *	@close_list:	List entry, that is used, when we are closing the device
1411  *
1412  *	@adj_list:	Directly linked devices, like slaves for bonding
1413  *	@all_adj_list:	All linked devices, *including* neighbours
1414  *	@features:	Currently active device features
1415  *	@hw_features:	User-changeable features
1416  *
1417  *	@wanted_features:	User-requested features
1418  *	@vlan_features:		Mask of features inheritable by VLAN devices
1419  *
1420  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1421  *				This field indicates what encapsulation
1422  *				offloads the hardware is capable of doing,
1423  *				and drivers will need to set them appropriately.
1424  *
1425  *	@mpls_features:	Mask of features inheritable by MPLS
1426  *
1427  *	@ifindex:	interface index
1428  *	@group:		The group, that the device belongs to
1429  *
1430  *	@stats:		Statistics struct, which was left as a legacy, use
1431  *			rtnl_link_stats64 instead
1432  *
1433  *	@rx_dropped:	Dropped packets by core network,
1434  *			do not use this in drivers
1435  *	@tx_dropped:	Dropped packets by core network,
1436  *			do not use this in drivers
1437  *	@rx_nohandler:	nohandler dropped packets by core network on
1438  *			inactive devices, do not use this in drivers
1439  *
1440  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1441  *				instead of ioctl,
1442  *				see <net/iw_handler.h> for details.
1443  *	@wireless_data:	Instance data managed by the core of wireless extensions
1444  *
1445  *	@netdev_ops:	Includes several pointers to callbacks,
1446  *			if one wants to override the ndo_*() functions
1447  *	@ethtool_ops:	Management operations
1448  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1449  *			of Layer 2 headers.
1450  *
1451  *	@flags:		Interface flags (a la BSD)
1452  *	@priv_flags:	Like 'flags' but invisible to userspace,
1453  *			see if.h for the definitions
1454  *	@gflags:	Global flags ( kept as legacy )
1455  *	@padded:	How much padding added by alloc_netdev()
1456  *	@operstate:	RFC2863 operstate
1457  *	@link_mode:	Mapping policy to operstate
1458  *	@if_port:	Selectable AUI, TP, ...
1459  *	@dma:		DMA channel
1460  *	@mtu:		Interface MTU value
1461  *	@type:		Interface hardware type
1462  *	@hard_header_len: Hardware header length, which means that this is the
1463  *			  minimum size of a packet.
1464  *
1465  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1466  *			  cases can this be guaranteed
1467  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1468  *			  cases can this be guaranteed. Some cases also use
1469  *			  LL_MAX_HEADER instead to allocate the skb
1470  *
1471  *	interface address info:
1472  *
1473  * 	@perm_addr:		Permanent hw address
1474  * 	@addr_assign_type:	Hw address assignment type
1475  * 	@addr_len:		Hardware address length
1476  * 	@neigh_priv_len;	Used in neigh_alloc(),
1477  * 				initialized only in atm/clip.c
1478  * 	@dev_id:		Used to differentiate devices that share
1479  * 				the same link layer address
1480  * 	@dev_port:		Used to differentiate devices that share
1481  * 				the same function
1482  *	@addr_list_lock:	XXX: need comments on this one
1483  *	@uc_promisc:		Counter, that indicates, that promiscuous mode
1484  *				has been enabled due to the need to listen to
1485  *				additional unicast addresses in a device that
1486  *				does not implement ndo_set_rx_mode()
1487  *	@uc:			unicast mac addresses
1488  *	@mc:			multicast mac addresses
1489  *	@dev_addrs:		list of device hw addresses
1490  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1491  *	@promiscuity:		Number of times, the NIC is told to work in
1492  *				Promiscuous mode, if it becomes 0 the NIC will
1493  *				exit from working in Promiscuous mode
1494  *	@allmulti:		Counter, enables or disables allmulticast mode
1495  *
1496  *	@vlan_info:	VLAN info
1497  *	@dsa_ptr:	dsa specific data
1498  *	@tipc_ptr:	TIPC specific data
1499  *	@atalk_ptr:	AppleTalk link
1500  *	@ip_ptr:	IPv4 specific data
1501  *	@dn_ptr:	DECnet specific data
1502  *	@ip6_ptr:	IPv6 specific data
1503  *	@ax25_ptr:	AX.25 specific data
1504  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1505  *
1506  *	@last_rx:	Time of last Rx
1507  *	@dev_addr:	Hw address (before bcast,
1508  *			because most packets are unicast)
1509  *
1510  *	@_rx:			Array of RX queues
1511  *	@num_rx_queues:		Number of RX queues
1512  *				allocated at register_netdev() time
1513  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1514  *
1515  *	@rx_handler:		handler for received packets
1516  *	@rx_handler_data: 	XXX: need comments on this one
1517  *	@ingress_queue:		XXX: need comments on this one
1518  *	@broadcast:		hw bcast address
1519  *
1520  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1521  *			indexed by RX queue number. Assigned by driver.
1522  *			This must only be set if the ndo_rx_flow_steer
1523  *			operation is defined
1524  *	@index_hlist:		Device index hash chain
1525  *
1526  *	@_tx:			Array of TX queues
1527  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1528  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1529  *	@qdisc:			Root qdisc from userspace point of view
1530  *	@tx_queue_len:		Max frames per queue allowed
1531  *	@tx_global_lock: 	XXX: need comments on this one
1532  *
1533  *	@xps_maps:	XXX: need comments on this one
1534  *
1535  *	@offload_fwd_mark:	Offload device fwding mark
1536  *
1537  *	@trans_start:		Time (in jiffies) of last Tx
1538  *	@watchdog_timeo:	Represents the timeout that is used by
1539  *				the watchdog ( see dev_watchdog() )
1540  *	@watchdog_timer:	List of timers
1541  *
1542  *	@pcpu_refcnt:		Number of references to this device
1543  *	@todo_list:		Delayed register/unregister
1544  *	@link_watch_list:	XXX: need comments on this one
1545  *
1546  *	@reg_state:		Register/unregister state machine
1547  *	@dismantle:		Device is going to be freed
1548  *	@rtnl_link_state:	This enum represents the phases of creating
1549  *				a new link
1550  *
1551  *	@destructor:		Called from unregister,
1552  *				can be used to call free_netdev
1553  *	@npinfo:		XXX: need comments on this one
1554  * 	@nd_net:		Network namespace this network device is inside
1555  *
1556  * 	@ml_priv:	Mid-layer private
1557  * 	@lstats:	Loopback statistics
1558  * 	@tstats:	Tunnel statistics
1559  * 	@dstats:	Dummy statistics
1560  * 	@vstats:	Virtual ethernet statistics
1561  *
1562  *	@garp_port:	GARP
1563  *	@mrp_port:	MRP
1564  *
1565  *	@dev:		Class/net/name entry
1566  *	@sysfs_groups:	Space for optional device, statistics and wireless
1567  *			sysfs groups
1568  *
1569  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1570  *	@rtnl_link_ops:	Rtnl_link_ops
1571  *
1572  *	@gso_max_size:	Maximum size of generic segmentation offload
1573  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1574  *			NIC for GSO
1575  *	@gso_min_segs:	Minimum number of segments that can be passed to the
1576  *			NIC for GSO
1577  *
1578  *	@dcbnl_ops:	Data Center Bridging netlink ops
1579  *	@num_tc:	Number of traffic classes in the net device
1580  *	@tc_to_txq:	XXX: need comments on this one
1581  *	@prio_tc_map	XXX: need comments on this one
1582  *
1583  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1584  *
1585  *	@priomap:	XXX: need comments on this one
1586  *	@phydev:	Physical device may attach itself
1587  *			for hardware timestamping
1588  *
1589  *	@qdisc_tx_busylock:	XXX: need comments on this one
1590  *
1591  *	@proto_down:	protocol port state information can be sent to the
1592  *			switch driver and used to set the phys state of the
1593  *			switch port.
1594  *
1595  *	FIXME: cleanup struct net_device such that network protocol info
1596  *	moves out.
1597  */
1598 
1599 struct net_device {
1600 	char			name[IFNAMSIZ];
1601 	struct hlist_node	name_hlist;
1602 	char 			*ifalias;
1603 	/*
1604 	 *	I/O specific fields
1605 	 *	FIXME: Merge these and struct ifmap into one
1606 	 */
1607 	unsigned long		mem_end;
1608 	unsigned long		mem_start;
1609 	unsigned long		base_addr;
1610 	int			irq;
1611 
1612 	atomic_t		carrier_changes;
1613 
1614 	/*
1615 	 *	Some hardware also needs these fields (state,dev_list,
1616 	 *	napi_list,unreg_list,close_list) but they are not
1617 	 *	part of the usual set specified in Space.c.
1618 	 */
1619 
1620 	unsigned long		state;
1621 
1622 	struct list_head	dev_list;
1623 	struct list_head	napi_list;
1624 	struct list_head	unreg_list;
1625 	struct list_head	close_list;
1626 	struct list_head	ptype_all;
1627 	struct list_head	ptype_specific;
1628 
1629 	struct {
1630 		struct list_head upper;
1631 		struct list_head lower;
1632 	} adj_list;
1633 
1634 	struct {
1635 		struct list_head upper;
1636 		struct list_head lower;
1637 	} all_adj_list;
1638 
1639 	netdev_features_t	features;
1640 	netdev_features_t	hw_features;
1641 	netdev_features_t	wanted_features;
1642 	netdev_features_t	vlan_features;
1643 	netdev_features_t	hw_enc_features;
1644 	netdev_features_t	mpls_features;
1645 
1646 	int			ifindex;
1647 	int			group;
1648 
1649 	struct net_device_stats	stats;
1650 
1651 	atomic_long_t		rx_dropped;
1652 	atomic_long_t		tx_dropped;
1653 	atomic_long_t		rx_nohandler;
1654 
1655 #ifdef CONFIG_WIRELESS_EXT
1656 	const struct iw_handler_def *	wireless_handlers;
1657 	struct iw_public_data *	wireless_data;
1658 #endif
1659 	const struct net_device_ops *netdev_ops;
1660 	const struct ethtool_ops *ethtool_ops;
1661 #ifdef CONFIG_NET_SWITCHDEV
1662 	const struct switchdev_ops *switchdev_ops;
1663 #endif
1664 #ifdef CONFIG_NET_L3_MASTER_DEV
1665 	const struct l3mdev_ops	*l3mdev_ops;
1666 #endif
1667 
1668 	const struct header_ops *header_ops;
1669 
1670 	unsigned int		flags;
1671 	unsigned int		priv_flags;
1672 
1673 	unsigned short		gflags;
1674 	unsigned short		padded;
1675 
1676 	unsigned char		operstate;
1677 	unsigned char		link_mode;
1678 
1679 	unsigned char		if_port;
1680 	unsigned char		dma;
1681 
1682 	unsigned int		mtu;
1683 	unsigned short		type;
1684 	unsigned short		hard_header_len;
1685 
1686 	unsigned short		needed_headroom;
1687 	unsigned short		needed_tailroom;
1688 
1689 	/* Interface address info. */
1690 	unsigned char		perm_addr[MAX_ADDR_LEN];
1691 	unsigned char		addr_assign_type;
1692 	unsigned char		addr_len;
1693 	unsigned short		neigh_priv_len;
1694 	unsigned short          dev_id;
1695 	unsigned short          dev_port;
1696 	spinlock_t		addr_list_lock;
1697 	unsigned char		name_assign_type;
1698 	bool			uc_promisc;
1699 	struct netdev_hw_addr_list	uc;
1700 	struct netdev_hw_addr_list	mc;
1701 	struct netdev_hw_addr_list	dev_addrs;
1702 
1703 #ifdef CONFIG_SYSFS
1704 	struct kset		*queues_kset;
1705 #endif
1706 	unsigned int		promiscuity;
1707 	unsigned int		allmulti;
1708 
1709 
1710 	/* Protocol specific pointers */
1711 
1712 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1713 	struct vlan_info __rcu	*vlan_info;
1714 #endif
1715 #if IS_ENABLED(CONFIG_NET_DSA)
1716 	struct dsa_switch_tree	*dsa_ptr;
1717 #endif
1718 #if IS_ENABLED(CONFIG_TIPC)
1719 	struct tipc_bearer __rcu *tipc_ptr;
1720 #endif
1721 	void 			*atalk_ptr;
1722 	struct in_device __rcu	*ip_ptr;
1723 	struct dn_dev __rcu     *dn_ptr;
1724 	struct inet6_dev __rcu	*ip6_ptr;
1725 	void			*ax25_ptr;
1726 	struct wireless_dev	*ieee80211_ptr;
1727 	struct wpan_dev		*ieee802154_ptr;
1728 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1729 	struct mpls_dev __rcu	*mpls_ptr;
1730 #endif
1731 
1732 /*
1733  * Cache lines mostly used on receive path (including eth_type_trans())
1734  */
1735 	unsigned long		last_rx;
1736 
1737 	/* Interface address info used in eth_type_trans() */
1738 	unsigned char		*dev_addr;
1739 
1740 
1741 #ifdef CONFIG_SYSFS
1742 	struct netdev_rx_queue	*_rx;
1743 
1744 	unsigned int		num_rx_queues;
1745 	unsigned int		real_num_rx_queues;
1746 
1747 #endif
1748 
1749 	unsigned long		gro_flush_timeout;
1750 	rx_handler_func_t __rcu	*rx_handler;
1751 	void __rcu		*rx_handler_data;
1752 
1753 #ifdef CONFIG_NET_CLS_ACT
1754 	struct tcf_proto __rcu  *ingress_cl_list;
1755 #endif
1756 	struct netdev_queue __rcu *ingress_queue;
1757 #ifdef CONFIG_NETFILTER_INGRESS
1758 	struct list_head	nf_hooks_ingress;
1759 #endif
1760 
1761 	unsigned char		broadcast[MAX_ADDR_LEN];
1762 #ifdef CONFIG_RFS_ACCEL
1763 	struct cpu_rmap		*rx_cpu_rmap;
1764 #endif
1765 	struct hlist_node	index_hlist;
1766 
1767 /*
1768  * Cache lines mostly used on transmit path
1769  */
1770 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1771 	unsigned int		num_tx_queues;
1772 	unsigned int		real_num_tx_queues;
1773 	struct Qdisc		*qdisc;
1774 	unsigned long		tx_queue_len;
1775 	spinlock_t		tx_global_lock;
1776 	int			watchdog_timeo;
1777 
1778 #ifdef CONFIG_XPS
1779 	struct xps_dev_maps __rcu *xps_maps;
1780 #endif
1781 #ifdef CONFIG_NET_CLS_ACT
1782 	struct tcf_proto __rcu  *egress_cl_list;
1783 #endif
1784 #ifdef CONFIG_NET_SWITCHDEV
1785 	u32			offload_fwd_mark;
1786 #endif
1787 
1788 	/* These may be needed for future network-power-down code. */
1789 
1790 	/*
1791 	 * trans_start here is expensive for high speed devices on SMP,
1792 	 * please use netdev_queue->trans_start instead.
1793 	 */
1794 	unsigned long		trans_start;
1795 
1796 	struct timer_list	watchdog_timer;
1797 
1798 	int __percpu		*pcpu_refcnt;
1799 	struct list_head	todo_list;
1800 
1801 	struct list_head	link_watch_list;
1802 
1803 	enum { NETREG_UNINITIALIZED=0,
1804 	       NETREG_REGISTERED,	/* completed register_netdevice */
1805 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1806 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1807 	       NETREG_RELEASED,		/* called free_netdev */
1808 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1809 	} reg_state:8;
1810 
1811 	bool dismantle;
1812 
1813 	enum {
1814 		RTNL_LINK_INITIALIZED,
1815 		RTNL_LINK_INITIALIZING,
1816 	} rtnl_link_state:16;
1817 
1818 	void (*destructor)(struct net_device *dev);
1819 
1820 #ifdef CONFIG_NETPOLL
1821 	struct netpoll_info __rcu	*npinfo;
1822 #endif
1823 
1824 	possible_net_t			nd_net;
1825 
1826 	/* mid-layer private */
1827 	union {
1828 		void					*ml_priv;
1829 		struct pcpu_lstats __percpu		*lstats;
1830 		struct pcpu_sw_netstats __percpu	*tstats;
1831 		struct pcpu_dstats __percpu		*dstats;
1832 		struct pcpu_vstats __percpu		*vstats;
1833 	};
1834 
1835 	struct garp_port __rcu	*garp_port;
1836 	struct mrp_port __rcu	*mrp_port;
1837 
1838 	struct device	dev;
1839 	const struct attribute_group *sysfs_groups[4];
1840 	const struct attribute_group *sysfs_rx_queue_group;
1841 
1842 	const struct rtnl_link_ops *rtnl_link_ops;
1843 
1844 	/* for setting kernel sock attribute on TCP connection setup */
1845 #define GSO_MAX_SIZE		65536
1846 	unsigned int		gso_max_size;
1847 #define GSO_MAX_SEGS		65535
1848 	u16			gso_max_segs;
1849 	u16			gso_min_segs;
1850 #ifdef CONFIG_DCB
1851 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1852 #endif
1853 	u8 num_tc;
1854 	struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1855 	u8 prio_tc_map[TC_BITMASK + 1];
1856 
1857 #if IS_ENABLED(CONFIG_FCOE)
1858 	unsigned int		fcoe_ddp_xid;
1859 #endif
1860 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1861 	struct netprio_map __rcu *priomap;
1862 #endif
1863 	struct phy_device *phydev;
1864 	struct lock_class_key *qdisc_tx_busylock;
1865 	bool proto_down;
1866 };
1867 #define to_net_dev(d) container_of(d, struct net_device, dev)
1868 
1869 #define	NETDEV_ALIGN		32
1870 
1871 static inline
1872 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1873 {
1874 	return dev->prio_tc_map[prio & TC_BITMASK];
1875 }
1876 
1877 static inline
1878 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1879 {
1880 	if (tc >= dev->num_tc)
1881 		return -EINVAL;
1882 
1883 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1884 	return 0;
1885 }
1886 
1887 static inline
1888 void netdev_reset_tc(struct net_device *dev)
1889 {
1890 	dev->num_tc = 0;
1891 	memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1892 	memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1893 }
1894 
1895 static inline
1896 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1897 {
1898 	if (tc >= dev->num_tc)
1899 		return -EINVAL;
1900 
1901 	dev->tc_to_txq[tc].count = count;
1902 	dev->tc_to_txq[tc].offset = offset;
1903 	return 0;
1904 }
1905 
1906 static inline
1907 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1908 {
1909 	if (num_tc > TC_MAX_QUEUE)
1910 		return -EINVAL;
1911 
1912 	dev->num_tc = num_tc;
1913 	return 0;
1914 }
1915 
1916 static inline
1917 int netdev_get_num_tc(struct net_device *dev)
1918 {
1919 	return dev->num_tc;
1920 }
1921 
1922 static inline
1923 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1924 					 unsigned int index)
1925 {
1926 	return &dev->_tx[index];
1927 }
1928 
1929 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1930 						    const struct sk_buff *skb)
1931 {
1932 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1933 }
1934 
1935 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1936 					    void (*f)(struct net_device *,
1937 						      struct netdev_queue *,
1938 						      void *),
1939 					    void *arg)
1940 {
1941 	unsigned int i;
1942 
1943 	for (i = 0; i < dev->num_tx_queues; i++)
1944 		f(dev, &dev->_tx[i], arg);
1945 }
1946 
1947 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1948 				    struct sk_buff *skb,
1949 				    void *accel_priv);
1950 
1951 /* returns the headroom that the master device needs to take in account
1952  * when forwarding to this dev
1953  */
1954 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1955 {
1956 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1957 }
1958 
1959 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1960 {
1961 	if (dev->netdev_ops->ndo_set_rx_headroom)
1962 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1963 }
1964 
1965 /* set the device rx headroom to the dev's default */
1966 static inline void netdev_reset_rx_headroom(struct net_device *dev)
1967 {
1968 	netdev_set_rx_headroom(dev, -1);
1969 }
1970 
1971 /*
1972  * Net namespace inlines
1973  */
1974 static inline
1975 struct net *dev_net(const struct net_device *dev)
1976 {
1977 	return read_pnet(&dev->nd_net);
1978 }
1979 
1980 static inline
1981 void dev_net_set(struct net_device *dev, struct net *net)
1982 {
1983 	write_pnet(&dev->nd_net, net);
1984 }
1985 
1986 static inline bool netdev_uses_dsa(struct net_device *dev)
1987 {
1988 #if IS_ENABLED(CONFIG_NET_DSA)
1989 	if (dev->dsa_ptr != NULL)
1990 		return dsa_uses_tagged_protocol(dev->dsa_ptr);
1991 #endif
1992 	return false;
1993 }
1994 
1995 /**
1996  *	netdev_priv - access network device private data
1997  *	@dev: network device
1998  *
1999  * Get network device private data
2000  */
2001 static inline void *netdev_priv(const struct net_device *dev)
2002 {
2003 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2004 }
2005 
2006 /* Set the sysfs physical device reference for the network logical device
2007  * if set prior to registration will cause a symlink during initialization.
2008  */
2009 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2010 
2011 /* Set the sysfs device type for the network logical device to allow
2012  * fine-grained identification of different network device types. For
2013  * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
2014  */
2015 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2016 
2017 /* Default NAPI poll() weight
2018  * Device drivers are strongly advised to not use bigger value
2019  */
2020 #define NAPI_POLL_WEIGHT 64
2021 
2022 /**
2023  *	netif_napi_add - initialize a napi context
2024  *	@dev:  network device
2025  *	@napi: napi context
2026  *	@poll: polling function
2027  *	@weight: default weight
2028  *
2029  * netif_napi_add() must be used to initialize a napi context prior to calling
2030  * *any* of the other napi related functions.
2031  */
2032 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2033 		    int (*poll)(struct napi_struct *, int), int weight);
2034 
2035 /**
2036  *	netif_tx_napi_add - initialize a napi context
2037  *	@dev:  network device
2038  *	@napi: napi context
2039  *	@poll: polling function
2040  *	@weight: default weight
2041  *
2042  * This variant of netif_napi_add() should be used from drivers using NAPI
2043  * to exclusively poll a TX queue.
2044  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2045  */
2046 static inline void netif_tx_napi_add(struct net_device *dev,
2047 				     struct napi_struct *napi,
2048 				     int (*poll)(struct napi_struct *, int),
2049 				     int weight)
2050 {
2051 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2052 	netif_napi_add(dev, napi, poll, weight);
2053 }
2054 
2055 /**
2056  *  netif_napi_del - remove a napi context
2057  *  @napi: napi context
2058  *
2059  *  netif_napi_del() removes a napi context from the network device napi list
2060  */
2061 void netif_napi_del(struct napi_struct *napi);
2062 
2063 struct napi_gro_cb {
2064 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2065 	void *frag0;
2066 
2067 	/* Length of frag0. */
2068 	unsigned int frag0_len;
2069 
2070 	/* This indicates where we are processing relative to skb->data. */
2071 	int data_offset;
2072 
2073 	/* This is non-zero if the packet cannot be merged with the new skb. */
2074 	u16	flush;
2075 
2076 	/* Save the IP ID here and check when we get to the transport layer */
2077 	u16	flush_id;
2078 
2079 	/* Number of segments aggregated. */
2080 	u16	count;
2081 
2082 	/* Start offset for remote checksum offload */
2083 	u16	gro_remcsum_start;
2084 
2085 	/* jiffies when first packet was created/queued */
2086 	unsigned long age;
2087 
2088 	/* Used in ipv6_gro_receive() and foo-over-udp */
2089 	u16	proto;
2090 
2091 	/* This is non-zero if the packet may be of the same flow. */
2092 	u8	same_flow:1;
2093 
2094 	/* Used in udp_gro_receive */
2095 	u8	udp_mark:1;
2096 
2097 	/* GRO checksum is valid */
2098 	u8	csum_valid:1;
2099 
2100 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2101 	u8	csum_cnt:3;
2102 
2103 	/* Free the skb? */
2104 	u8	free:2;
2105 #define NAPI_GRO_FREE		  1
2106 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2107 
2108 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2109 	u8	is_ipv6:1;
2110 
2111 	/* 7 bit hole */
2112 
2113 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2114 	__wsum	csum;
2115 
2116 	/* used in skb_gro_receive() slow path */
2117 	struct sk_buff *last;
2118 };
2119 
2120 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2121 
2122 struct packet_type {
2123 	__be16			type;	/* This is really htons(ether_type). */
2124 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2125 	int			(*func) (struct sk_buff *,
2126 					 struct net_device *,
2127 					 struct packet_type *,
2128 					 struct net_device *);
2129 	bool			(*id_match)(struct packet_type *ptype,
2130 					    struct sock *sk);
2131 	void			*af_packet_priv;
2132 	struct list_head	list;
2133 };
2134 
2135 struct offload_callbacks {
2136 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2137 						netdev_features_t features);
2138 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2139 						 struct sk_buff *skb);
2140 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2141 };
2142 
2143 struct packet_offload {
2144 	__be16			 type;	/* This is really htons(ether_type). */
2145 	u16			 priority;
2146 	struct offload_callbacks callbacks;
2147 	struct list_head	 list;
2148 };
2149 
2150 struct udp_offload;
2151 
2152 struct udp_offload_callbacks {
2153 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2154 						 struct sk_buff *skb,
2155 						 struct udp_offload *uoff);
2156 	int			(*gro_complete)(struct sk_buff *skb,
2157 						int nhoff,
2158 						struct udp_offload *uoff);
2159 };
2160 
2161 struct udp_offload {
2162 	__be16			 port;
2163 	u8			 ipproto;
2164 	struct udp_offload_callbacks callbacks;
2165 };
2166 
2167 /* often modified stats are per cpu, other are shared (netdev->stats) */
2168 struct pcpu_sw_netstats {
2169 	u64     rx_packets;
2170 	u64     rx_bytes;
2171 	u64     tx_packets;
2172 	u64     tx_bytes;
2173 	struct u64_stats_sync   syncp;
2174 };
2175 
2176 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2177 ({									\
2178 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2179 	if (pcpu_stats)	{						\
2180 		int __cpu;						\
2181 		for_each_possible_cpu(__cpu) {				\
2182 			typeof(type) *stat;				\
2183 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2184 			u64_stats_init(&stat->syncp);			\
2185 		}							\
2186 	}								\
2187 	pcpu_stats;							\
2188 })
2189 
2190 #define netdev_alloc_pcpu_stats(type)					\
2191 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2192 
2193 enum netdev_lag_tx_type {
2194 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2195 	NETDEV_LAG_TX_TYPE_RANDOM,
2196 	NETDEV_LAG_TX_TYPE_BROADCAST,
2197 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2198 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2199 	NETDEV_LAG_TX_TYPE_HASH,
2200 };
2201 
2202 struct netdev_lag_upper_info {
2203 	enum netdev_lag_tx_type tx_type;
2204 };
2205 
2206 struct netdev_lag_lower_state_info {
2207 	u8 link_up : 1,
2208 	   tx_enabled : 1;
2209 };
2210 
2211 #include <linux/notifier.h>
2212 
2213 /* netdevice notifier chain. Please remember to update the rtnetlink
2214  * notification exclusion list in rtnetlink_event() when adding new
2215  * types.
2216  */
2217 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2218 #define NETDEV_DOWN	0x0002
2219 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2220 				   detected a hardware crash and restarted
2221 				   - we can use this eg to kick tcp sessions
2222 				   once done */
2223 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2224 #define NETDEV_REGISTER 0x0005
2225 #define NETDEV_UNREGISTER	0x0006
2226 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2227 #define NETDEV_CHANGEADDR	0x0008
2228 #define NETDEV_GOING_DOWN	0x0009
2229 #define NETDEV_CHANGENAME	0x000A
2230 #define NETDEV_FEAT_CHANGE	0x000B
2231 #define NETDEV_BONDING_FAILOVER 0x000C
2232 #define NETDEV_PRE_UP		0x000D
2233 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2234 #define NETDEV_POST_TYPE_CHANGE	0x000F
2235 #define NETDEV_POST_INIT	0x0010
2236 #define NETDEV_UNREGISTER_FINAL 0x0011
2237 #define NETDEV_RELEASE		0x0012
2238 #define NETDEV_NOTIFY_PEERS	0x0013
2239 #define NETDEV_JOIN		0x0014
2240 #define NETDEV_CHANGEUPPER	0x0015
2241 #define NETDEV_RESEND_IGMP	0x0016
2242 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2243 #define NETDEV_CHANGEINFODATA	0x0018
2244 #define NETDEV_BONDING_INFO	0x0019
2245 #define NETDEV_PRECHANGEUPPER	0x001A
2246 #define NETDEV_CHANGELOWERSTATE	0x001B
2247 
2248 int register_netdevice_notifier(struct notifier_block *nb);
2249 int unregister_netdevice_notifier(struct notifier_block *nb);
2250 
2251 struct netdev_notifier_info {
2252 	struct net_device *dev;
2253 };
2254 
2255 struct netdev_notifier_change_info {
2256 	struct netdev_notifier_info info; /* must be first */
2257 	unsigned int flags_changed;
2258 };
2259 
2260 struct netdev_notifier_changeupper_info {
2261 	struct netdev_notifier_info info; /* must be first */
2262 	struct net_device *upper_dev; /* new upper dev */
2263 	bool master; /* is upper dev master */
2264 	bool linking; /* is the nofication for link or unlink */
2265 	void *upper_info; /* upper dev info */
2266 };
2267 
2268 struct netdev_notifier_changelowerstate_info {
2269 	struct netdev_notifier_info info; /* must be first */
2270 	void *lower_state_info; /* is lower dev state */
2271 };
2272 
2273 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2274 					     struct net_device *dev)
2275 {
2276 	info->dev = dev;
2277 }
2278 
2279 static inline struct net_device *
2280 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2281 {
2282 	return info->dev;
2283 }
2284 
2285 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2286 
2287 
2288 extern rwlock_t				dev_base_lock;		/* Device list lock */
2289 
2290 #define for_each_netdev(net, d)		\
2291 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2292 #define for_each_netdev_reverse(net, d)	\
2293 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2294 #define for_each_netdev_rcu(net, d)		\
2295 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2296 #define for_each_netdev_safe(net, d, n)	\
2297 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2298 #define for_each_netdev_continue(net, d)		\
2299 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2300 #define for_each_netdev_continue_rcu(net, d)		\
2301 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2302 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2303 		for_each_netdev_rcu(&init_net, slave)	\
2304 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2305 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2306 
2307 static inline struct net_device *next_net_device(struct net_device *dev)
2308 {
2309 	struct list_head *lh;
2310 	struct net *net;
2311 
2312 	net = dev_net(dev);
2313 	lh = dev->dev_list.next;
2314 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2315 }
2316 
2317 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2318 {
2319 	struct list_head *lh;
2320 	struct net *net;
2321 
2322 	net = dev_net(dev);
2323 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2324 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2325 }
2326 
2327 static inline struct net_device *first_net_device(struct net *net)
2328 {
2329 	return list_empty(&net->dev_base_head) ? NULL :
2330 		net_device_entry(net->dev_base_head.next);
2331 }
2332 
2333 static inline struct net_device *first_net_device_rcu(struct net *net)
2334 {
2335 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2336 
2337 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2338 }
2339 
2340 int netdev_boot_setup_check(struct net_device *dev);
2341 unsigned long netdev_boot_base(const char *prefix, int unit);
2342 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2343 				       const char *hwaddr);
2344 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2345 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2346 void dev_add_pack(struct packet_type *pt);
2347 void dev_remove_pack(struct packet_type *pt);
2348 void __dev_remove_pack(struct packet_type *pt);
2349 void dev_add_offload(struct packet_offload *po);
2350 void dev_remove_offload(struct packet_offload *po);
2351 
2352 int dev_get_iflink(const struct net_device *dev);
2353 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2354 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2355 				      unsigned short mask);
2356 struct net_device *dev_get_by_name(struct net *net, const char *name);
2357 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2358 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2359 int dev_alloc_name(struct net_device *dev, const char *name);
2360 int dev_open(struct net_device *dev);
2361 int dev_close(struct net_device *dev);
2362 int dev_close_many(struct list_head *head, bool unlink);
2363 void dev_disable_lro(struct net_device *dev);
2364 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2365 int dev_queue_xmit(struct sk_buff *skb);
2366 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2367 int register_netdevice(struct net_device *dev);
2368 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2369 void unregister_netdevice_many(struct list_head *head);
2370 static inline void unregister_netdevice(struct net_device *dev)
2371 {
2372 	unregister_netdevice_queue(dev, NULL);
2373 }
2374 
2375 int netdev_refcnt_read(const struct net_device *dev);
2376 void free_netdev(struct net_device *dev);
2377 void netdev_freemem(struct net_device *dev);
2378 void synchronize_net(void);
2379 int init_dummy_netdev(struct net_device *dev);
2380 
2381 DECLARE_PER_CPU(int, xmit_recursion);
2382 static inline int dev_recursion_level(void)
2383 {
2384 	return this_cpu_read(xmit_recursion);
2385 }
2386 
2387 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2388 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2389 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2390 int netdev_get_name(struct net *net, char *name, int ifindex);
2391 int dev_restart(struct net_device *dev);
2392 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2393 
2394 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2395 {
2396 	return NAPI_GRO_CB(skb)->data_offset;
2397 }
2398 
2399 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2400 {
2401 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2402 }
2403 
2404 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2405 {
2406 	NAPI_GRO_CB(skb)->data_offset += len;
2407 }
2408 
2409 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2410 					unsigned int offset)
2411 {
2412 	return NAPI_GRO_CB(skb)->frag0 + offset;
2413 }
2414 
2415 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2416 {
2417 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2418 }
2419 
2420 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2421 					unsigned int offset)
2422 {
2423 	if (!pskb_may_pull(skb, hlen))
2424 		return NULL;
2425 
2426 	NAPI_GRO_CB(skb)->frag0 = NULL;
2427 	NAPI_GRO_CB(skb)->frag0_len = 0;
2428 	return skb->data + offset;
2429 }
2430 
2431 static inline void *skb_gro_network_header(struct sk_buff *skb)
2432 {
2433 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2434 	       skb_network_offset(skb);
2435 }
2436 
2437 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2438 					const void *start, unsigned int len)
2439 {
2440 	if (NAPI_GRO_CB(skb)->csum_valid)
2441 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2442 						  csum_partial(start, len, 0));
2443 }
2444 
2445 /* GRO checksum functions. These are logical equivalents of the normal
2446  * checksum functions (in skbuff.h) except that they operate on the GRO
2447  * offsets and fields in sk_buff.
2448  */
2449 
2450 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2451 
2452 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2453 {
2454 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2455 }
2456 
2457 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2458 						      bool zero_okay,
2459 						      __sum16 check)
2460 {
2461 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2462 		skb_checksum_start_offset(skb) <
2463 		 skb_gro_offset(skb)) &&
2464 		!skb_at_gro_remcsum_start(skb) &&
2465 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2466 		(!zero_okay || check));
2467 }
2468 
2469 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2470 							   __wsum psum)
2471 {
2472 	if (NAPI_GRO_CB(skb)->csum_valid &&
2473 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2474 		return 0;
2475 
2476 	NAPI_GRO_CB(skb)->csum = psum;
2477 
2478 	return __skb_gro_checksum_complete(skb);
2479 }
2480 
2481 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2482 {
2483 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2484 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2485 		NAPI_GRO_CB(skb)->csum_cnt--;
2486 	} else {
2487 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2488 		 * verified a new top level checksum or an encapsulated one
2489 		 * during GRO. This saves work if we fallback to normal path.
2490 		 */
2491 		__skb_incr_checksum_unnecessary(skb);
2492 	}
2493 }
2494 
2495 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2496 				    compute_pseudo)			\
2497 ({									\
2498 	__sum16 __ret = 0;						\
2499 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2500 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2501 				compute_pseudo(skb, proto));		\
2502 	if (__ret)							\
2503 		__skb_mark_checksum_bad(skb);				\
2504 	else								\
2505 		skb_gro_incr_csum_unnecessary(skb);			\
2506 	__ret;								\
2507 })
2508 
2509 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2510 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2511 
2512 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2513 					     compute_pseudo)		\
2514 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2515 
2516 #define skb_gro_checksum_simple_validate(skb)				\
2517 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2518 
2519 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2520 {
2521 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2522 		!NAPI_GRO_CB(skb)->csum_valid);
2523 }
2524 
2525 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2526 					      __sum16 check, __wsum pseudo)
2527 {
2528 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2529 	NAPI_GRO_CB(skb)->csum_valid = 1;
2530 }
2531 
2532 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2533 do {									\
2534 	if (__skb_gro_checksum_convert_check(skb))			\
2535 		__skb_gro_checksum_convert(skb, check,			\
2536 					   compute_pseudo(skb, proto));	\
2537 } while (0)
2538 
2539 struct gro_remcsum {
2540 	int offset;
2541 	__wsum delta;
2542 };
2543 
2544 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2545 {
2546 	grc->offset = 0;
2547 	grc->delta = 0;
2548 }
2549 
2550 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2551 					    unsigned int off, size_t hdrlen,
2552 					    int start, int offset,
2553 					    struct gro_remcsum *grc,
2554 					    bool nopartial)
2555 {
2556 	__wsum delta;
2557 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2558 
2559 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2560 
2561 	if (!nopartial) {
2562 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2563 		return ptr;
2564 	}
2565 
2566 	ptr = skb_gro_header_fast(skb, off);
2567 	if (skb_gro_header_hard(skb, off + plen)) {
2568 		ptr = skb_gro_header_slow(skb, off + plen, off);
2569 		if (!ptr)
2570 			return NULL;
2571 	}
2572 
2573 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2574 			       start, offset);
2575 
2576 	/* Adjust skb->csum since we changed the packet */
2577 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2578 
2579 	grc->offset = off + hdrlen + offset;
2580 	grc->delta = delta;
2581 
2582 	return ptr;
2583 }
2584 
2585 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2586 					   struct gro_remcsum *grc)
2587 {
2588 	void *ptr;
2589 	size_t plen = grc->offset + sizeof(u16);
2590 
2591 	if (!grc->delta)
2592 		return;
2593 
2594 	ptr = skb_gro_header_fast(skb, grc->offset);
2595 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2596 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2597 		if (!ptr)
2598 			return;
2599 	}
2600 
2601 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2602 }
2603 
2604 struct skb_csum_offl_spec {
2605 	__u16		ipv4_okay:1,
2606 			ipv6_okay:1,
2607 			encap_okay:1,
2608 			ip_options_okay:1,
2609 			ext_hdrs_okay:1,
2610 			tcp_okay:1,
2611 			udp_okay:1,
2612 			sctp_okay:1,
2613 			vlan_okay:1,
2614 			no_encapped_ipv6:1,
2615 			no_not_encapped:1;
2616 };
2617 
2618 bool __skb_csum_offload_chk(struct sk_buff *skb,
2619 			    const struct skb_csum_offl_spec *spec,
2620 			    bool *csum_encapped,
2621 			    bool csum_help);
2622 
2623 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2624 					const struct skb_csum_offl_spec *spec,
2625 					bool *csum_encapped,
2626 					bool csum_help)
2627 {
2628 	if (skb->ip_summed != CHECKSUM_PARTIAL)
2629 		return false;
2630 
2631 	return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2632 }
2633 
2634 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2635 					     const struct skb_csum_offl_spec *spec)
2636 {
2637 	bool csum_encapped;
2638 
2639 	return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2640 }
2641 
2642 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2643 {
2644 	static const struct skb_csum_offl_spec csum_offl_spec = {
2645 		.ipv4_okay = 1,
2646 		.ip_options_okay = 1,
2647 		.ipv6_okay = 1,
2648 		.vlan_okay = 1,
2649 		.tcp_okay = 1,
2650 		.udp_okay = 1,
2651 	};
2652 
2653 	return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2654 }
2655 
2656 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2657 {
2658 	static const struct skb_csum_offl_spec csum_offl_spec = {
2659 		.ipv4_okay = 1,
2660 		.ip_options_okay = 1,
2661 		.tcp_okay = 1,
2662 		.udp_okay = 1,
2663 		.vlan_okay = 1,
2664 	};
2665 
2666 	return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2667 }
2668 
2669 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2670 				  unsigned short type,
2671 				  const void *daddr, const void *saddr,
2672 				  unsigned int len)
2673 {
2674 	if (!dev->header_ops || !dev->header_ops->create)
2675 		return 0;
2676 
2677 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2678 }
2679 
2680 static inline int dev_parse_header(const struct sk_buff *skb,
2681 				   unsigned char *haddr)
2682 {
2683 	const struct net_device *dev = skb->dev;
2684 
2685 	if (!dev->header_ops || !dev->header_ops->parse)
2686 		return 0;
2687 	return dev->header_ops->parse(skb, haddr);
2688 }
2689 
2690 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2691 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2692 static inline int unregister_gifconf(unsigned int family)
2693 {
2694 	return register_gifconf(family, NULL);
2695 }
2696 
2697 #ifdef CONFIG_NET_FLOW_LIMIT
2698 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2699 struct sd_flow_limit {
2700 	u64			count;
2701 	unsigned int		num_buckets;
2702 	unsigned int		history_head;
2703 	u16			history[FLOW_LIMIT_HISTORY];
2704 	u8			buckets[];
2705 };
2706 
2707 extern int netdev_flow_limit_table_len;
2708 #endif /* CONFIG_NET_FLOW_LIMIT */
2709 
2710 /*
2711  * Incoming packets are placed on per-cpu queues
2712  */
2713 struct softnet_data {
2714 	struct list_head	poll_list;
2715 	struct sk_buff_head	process_queue;
2716 
2717 	/* stats */
2718 	unsigned int		processed;
2719 	unsigned int		time_squeeze;
2720 	unsigned int		cpu_collision;
2721 	unsigned int		received_rps;
2722 #ifdef CONFIG_RPS
2723 	struct softnet_data	*rps_ipi_list;
2724 #endif
2725 #ifdef CONFIG_NET_FLOW_LIMIT
2726 	struct sd_flow_limit __rcu *flow_limit;
2727 #endif
2728 	struct Qdisc		*output_queue;
2729 	struct Qdisc		**output_queue_tailp;
2730 	struct sk_buff		*completion_queue;
2731 
2732 #ifdef CONFIG_RPS
2733 	/* Elements below can be accessed between CPUs for RPS */
2734 	struct call_single_data	csd ____cacheline_aligned_in_smp;
2735 	struct softnet_data	*rps_ipi_next;
2736 	unsigned int		cpu;
2737 	unsigned int		input_queue_head;
2738 	unsigned int		input_queue_tail;
2739 #endif
2740 	unsigned int		dropped;
2741 	struct sk_buff_head	input_pkt_queue;
2742 	struct napi_struct	backlog;
2743 
2744 };
2745 
2746 static inline void input_queue_head_incr(struct softnet_data *sd)
2747 {
2748 #ifdef CONFIG_RPS
2749 	sd->input_queue_head++;
2750 #endif
2751 }
2752 
2753 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2754 					      unsigned int *qtail)
2755 {
2756 #ifdef CONFIG_RPS
2757 	*qtail = ++sd->input_queue_tail;
2758 #endif
2759 }
2760 
2761 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2762 
2763 void __netif_schedule(struct Qdisc *q);
2764 void netif_schedule_queue(struct netdev_queue *txq);
2765 
2766 static inline void netif_tx_schedule_all(struct net_device *dev)
2767 {
2768 	unsigned int i;
2769 
2770 	for (i = 0; i < dev->num_tx_queues; i++)
2771 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2772 }
2773 
2774 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2775 {
2776 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2777 }
2778 
2779 /**
2780  *	netif_start_queue - allow transmit
2781  *	@dev: network device
2782  *
2783  *	Allow upper layers to call the device hard_start_xmit routine.
2784  */
2785 static inline void netif_start_queue(struct net_device *dev)
2786 {
2787 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2788 }
2789 
2790 static inline void netif_tx_start_all_queues(struct net_device *dev)
2791 {
2792 	unsigned int i;
2793 
2794 	for (i = 0; i < dev->num_tx_queues; i++) {
2795 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2796 		netif_tx_start_queue(txq);
2797 	}
2798 }
2799 
2800 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2801 
2802 /**
2803  *	netif_wake_queue - restart transmit
2804  *	@dev: network device
2805  *
2806  *	Allow upper layers to call the device hard_start_xmit routine.
2807  *	Used for flow control when transmit resources are available.
2808  */
2809 static inline void netif_wake_queue(struct net_device *dev)
2810 {
2811 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2812 }
2813 
2814 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2815 {
2816 	unsigned int i;
2817 
2818 	for (i = 0; i < dev->num_tx_queues; i++) {
2819 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2820 		netif_tx_wake_queue(txq);
2821 	}
2822 }
2823 
2824 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2825 {
2826 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2827 }
2828 
2829 /**
2830  *	netif_stop_queue - stop transmitted packets
2831  *	@dev: network device
2832  *
2833  *	Stop upper layers calling the device hard_start_xmit routine.
2834  *	Used for flow control when transmit resources are unavailable.
2835  */
2836 static inline void netif_stop_queue(struct net_device *dev)
2837 {
2838 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2839 }
2840 
2841 void netif_tx_stop_all_queues(struct net_device *dev);
2842 
2843 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2844 {
2845 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2846 }
2847 
2848 /**
2849  *	netif_queue_stopped - test if transmit queue is flowblocked
2850  *	@dev: network device
2851  *
2852  *	Test if transmit queue on device is currently unable to send.
2853  */
2854 static inline bool netif_queue_stopped(const struct net_device *dev)
2855 {
2856 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2857 }
2858 
2859 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2860 {
2861 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2862 }
2863 
2864 static inline bool
2865 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2866 {
2867 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2868 }
2869 
2870 static inline bool
2871 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2872 {
2873 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2874 }
2875 
2876 /**
2877  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2878  *	@dev_queue: pointer to transmit queue
2879  *
2880  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2881  * to give appropriate hint to the cpu.
2882  */
2883 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2884 {
2885 #ifdef CONFIG_BQL
2886 	prefetchw(&dev_queue->dql.num_queued);
2887 #endif
2888 }
2889 
2890 /**
2891  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2892  *	@dev_queue: pointer to transmit queue
2893  *
2894  * BQL enabled drivers might use this helper in their TX completion path,
2895  * to give appropriate hint to the cpu.
2896  */
2897 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2898 {
2899 #ifdef CONFIG_BQL
2900 	prefetchw(&dev_queue->dql.limit);
2901 #endif
2902 }
2903 
2904 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2905 					unsigned int bytes)
2906 {
2907 #ifdef CONFIG_BQL
2908 	dql_queued(&dev_queue->dql, bytes);
2909 
2910 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2911 		return;
2912 
2913 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2914 
2915 	/*
2916 	 * The XOFF flag must be set before checking the dql_avail below,
2917 	 * because in netdev_tx_completed_queue we update the dql_completed
2918 	 * before checking the XOFF flag.
2919 	 */
2920 	smp_mb();
2921 
2922 	/* check again in case another CPU has just made room avail */
2923 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2924 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2925 #endif
2926 }
2927 
2928 /**
2929  * 	netdev_sent_queue - report the number of bytes queued to hardware
2930  * 	@dev: network device
2931  * 	@bytes: number of bytes queued to the hardware device queue
2932  *
2933  * 	Report the number of bytes queued for sending/completion to the network
2934  * 	device hardware queue. @bytes should be a good approximation and should
2935  * 	exactly match netdev_completed_queue() @bytes
2936  */
2937 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2938 {
2939 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2940 }
2941 
2942 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2943 					     unsigned int pkts, unsigned int bytes)
2944 {
2945 #ifdef CONFIG_BQL
2946 	if (unlikely(!bytes))
2947 		return;
2948 
2949 	dql_completed(&dev_queue->dql, bytes);
2950 
2951 	/*
2952 	 * Without the memory barrier there is a small possiblity that
2953 	 * netdev_tx_sent_queue will miss the update and cause the queue to
2954 	 * be stopped forever
2955 	 */
2956 	smp_mb();
2957 
2958 	if (dql_avail(&dev_queue->dql) < 0)
2959 		return;
2960 
2961 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2962 		netif_schedule_queue(dev_queue);
2963 #endif
2964 }
2965 
2966 /**
2967  * 	netdev_completed_queue - report bytes and packets completed by device
2968  * 	@dev: network device
2969  * 	@pkts: actual number of packets sent over the medium
2970  * 	@bytes: actual number of bytes sent over the medium
2971  *
2972  * 	Report the number of bytes and packets transmitted by the network device
2973  * 	hardware queue over the physical medium, @bytes must exactly match the
2974  * 	@bytes amount passed to netdev_sent_queue()
2975  */
2976 static inline void netdev_completed_queue(struct net_device *dev,
2977 					  unsigned int pkts, unsigned int bytes)
2978 {
2979 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2980 }
2981 
2982 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2983 {
2984 #ifdef CONFIG_BQL
2985 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
2986 	dql_reset(&q->dql);
2987 #endif
2988 }
2989 
2990 /**
2991  * 	netdev_reset_queue - reset the packets and bytes count of a network device
2992  * 	@dev_queue: network device
2993  *
2994  * 	Reset the bytes and packet count of a network device and clear the
2995  * 	software flow control OFF bit for this network device
2996  */
2997 static inline void netdev_reset_queue(struct net_device *dev_queue)
2998 {
2999 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3000 }
3001 
3002 /**
3003  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3004  * 	@dev: network device
3005  * 	@queue_index: given tx queue index
3006  *
3007  * 	Returns 0 if given tx queue index >= number of device tx queues,
3008  * 	otherwise returns the originally passed tx queue index.
3009  */
3010 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3011 {
3012 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3013 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3014 				     dev->name, queue_index,
3015 				     dev->real_num_tx_queues);
3016 		return 0;
3017 	}
3018 
3019 	return queue_index;
3020 }
3021 
3022 /**
3023  *	netif_running - test if up
3024  *	@dev: network device
3025  *
3026  *	Test if the device has been brought up.
3027  */
3028 static inline bool netif_running(const struct net_device *dev)
3029 {
3030 	return test_bit(__LINK_STATE_START, &dev->state);
3031 }
3032 
3033 /*
3034  * Routines to manage the subqueues on a device.  We only need start
3035  * stop, and a check if it's stopped.  All other device management is
3036  * done at the overall netdevice level.
3037  * Also test the device if we're multiqueue.
3038  */
3039 
3040 /**
3041  *	netif_start_subqueue - allow sending packets on subqueue
3042  *	@dev: network device
3043  *	@queue_index: sub queue index
3044  *
3045  * Start individual transmit queue of a device with multiple transmit queues.
3046  */
3047 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3048 {
3049 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3050 
3051 	netif_tx_start_queue(txq);
3052 }
3053 
3054 /**
3055  *	netif_stop_subqueue - stop sending packets on subqueue
3056  *	@dev: network device
3057  *	@queue_index: sub queue index
3058  *
3059  * Stop individual transmit queue of a device with multiple transmit queues.
3060  */
3061 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3062 {
3063 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3064 	netif_tx_stop_queue(txq);
3065 }
3066 
3067 /**
3068  *	netif_subqueue_stopped - test status of subqueue
3069  *	@dev: network device
3070  *	@queue_index: sub queue index
3071  *
3072  * Check individual transmit queue of a device with multiple transmit queues.
3073  */
3074 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3075 					    u16 queue_index)
3076 {
3077 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3078 
3079 	return netif_tx_queue_stopped(txq);
3080 }
3081 
3082 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3083 					  struct sk_buff *skb)
3084 {
3085 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3086 }
3087 
3088 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3089 
3090 #ifdef CONFIG_XPS
3091 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3092 			u16 index);
3093 #else
3094 static inline int netif_set_xps_queue(struct net_device *dev,
3095 				      const struct cpumask *mask,
3096 				      u16 index)
3097 {
3098 	return 0;
3099 }
3100 #endif
3101 
3102 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3103 		  unsigned int num_tx_queues);
3104 
3105 /*
3106  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3107  * as a distribution range limit for the returned value.
3108  */
3109 static inline u16 skb_tx_hash(const struct net_device *dev,
3110 			      struct sk_buff *skb)
3111 {
3112 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3113 }
3114 
3115 /**
3116  *	netif_is_multiqueue - test if device has multiple transmit queues
3117  *	@dev: network device
3118  *
3119  * Check if device has multiple transmit queues
3120  */
3121 static inline bool netif_is_multiqueue(const struct net_device *dev)
3122 {
3123 	return dev->num_tx_queues > 1;
3124 }
3125 
3126 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3127 
3128 #ifdef CONFIG_SYSFS
3129 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3130 #else
3131 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3132 						unsigned int rxq)
3133 {
3134 	return 0;
3135 }
3136 #endif
3137 
3138 #ifdef CONFIG_SYSFS
3139 static inline unsigned int get_netdev_rx_queue_index(
3140 		struct netdev_rx_queue *queue)
3141 {
3142 	struct net_device *dev = queue->dev;
3143 	int index = queue - dev->_rx;
3144 
3145 	BUG_ON(index >= dev->num_rx_queues);
3146 	return index;
3147 }
3148 #endif
3149 
3150 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3151 int netif_get_num_default_rss_queues(void);
3152 
3153 enum skb_free_reason {
3154 	SKB_REASON_CONSUMED,
3155 	SKB_REASON_DROPPED,
3156 };
3157 
3158 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3159 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3160 
3161 /*
3162  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3163  * interrupt context or with hardware interrupts being disabled.
3164  * (in_irq() || irqs_disabled())
3165  *
3166  * We provide four helpers that can be used in following contexts :
3167  *
3168  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3169  *  replacing kfree_skb(skb)
3170  *
3171  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3172  *  Typically used in place of consume_skb(skb) in TX completion path
3173  *
3174  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3175  *  replacing kfree_skb(skb)
3176  *
3177  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3178  *  and consumed a packet. Used in place of consume_skb(skb)
3179  */
3180 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3181 {
3182 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3183 }
3184 
3185 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3186 {
3187 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3188 }
3189 
3190 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3191 {
3192 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3193 }
3194 
3195 static inline void dev_consume_skb_any(struct sk_buff *skb)
3196 {
3197 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3198 }
3199 
3200 int netif_rx(struct sk_buff *skb);
3201 int netif_rx_ni(struct sk_buff *skb);
3202 int netif_receive_skb(struct sk_buff *skb);
3203 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3204 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3205 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3206 gro_result_t napi_gro_frags(struct napi_struct *napi);
3207 struct packet_offload *gro_find_receive_by_type(__be16 type);
3208 struct packet_offload *gro_find_complete_by_type(__be16 type);
3209 
3210 static inline void napi_free_frags(struct napi_struct *napi)
3211 {
3212 	kfree_skb(napi->skb);
3213 	napi->skb = NULL;
3214 }
3215 
3216 int netdev_rx_handler_register(struct net_device *dev,
3217 			       rx_handler_func_t *rx_handler,
3218 			       void *rx_handler_data);
3219 void netdev_rx_handler_unregister(struct net_device *dev);
3220 
3221 bool dev_valid_name(const char *name);
3222 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3223 int dev_ethtool(struct net *net, struct ifreq *);
3224 unsigned int dev_get_flags(const struct net_device *);
3225 int __dev_change_flags(struct net_device *, unsigned int flags);
3226 int dev_change_flags(struct net_device *, unsigned int);
3227 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3228 			unsigned int gchanges);
3229 int dev_change_name(struct net_device *, const char *);
3230 int dev_set_alias(struct net_device *, const char *, size_t);
3231 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3232 int dev_set_mtu(struct net_device *, int);
3233 void dev_set_group(struct net_device *, int);
3234 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3235 int dev_change_carrier(struct net_device *, bool new_carrier);
3236 int dev_get_phys_port_id(struct net_device *dev,
3237 			 struct netdev_phys_item_id *ppid);
3238 int dev_get_phys_port_name(struct net_device *dev,
3239 			   char *name, size_t len);
3240 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3241 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3242 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3243 				    struct netdev_queue *txq, int *ret);
3244 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3245 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3246 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3247 
3248 extern int		netdev_budget;
3249 
3250 /* Called by rtnetlink.c:rtnl_unlock() */
3251 void netdev_run_todo(void);
3252 
3253 /**
3254  *	dev_put - release reference to device
3255  *	@dev: network device
3256  *
3257  * Release reference to device to allow it to be freed.
3258  */
3259 static inline void dev_put(struct net_device *dev)
3260 {
3261 	this_cpu_dec(*dev->pcpu_refcnt);
3262 }
3263 
3264 /**
3265  *	dev_hold - get reference to device
3266  *	@dev: network device
3267  *
3268  * Hold reference to device to keep it from being freed.
3269  */
3270 static inline void dev_hold(struct net_device *dev)
3271 {
3272 	this_cpu_inc(*dev->pcpu_refcnt);
3273 }
3274 
3275 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3276  * and _off may be called from IRQ context, but it is caller
3277  * who is responsible for serialization of these calls.
3278  *
3279  * The name carrier is inappropriate, these functions should really be
3280  * called netif_lowerlayer_*() because they represent the state of any
3281  * kind of lower layer not just hardware media.
3282  */
3283 
3284 void linkwatch_init_dev(struct net_device *dev);
3285 void linkwatch_fire_event(struct net_device *dev);
3286 void linkwatch_forget_dev(struct net_device *dev);
3287 
3288 /**
3289  *	netif_carrier_ok - test if carrier present
3290  *	@dev: network device
3291  *
3292  * Check if carrier is present on device
3293  */
3294 static inline bool netif_carrier_ok(const struct net_device *dev)
3295 {
3296 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3297 }
3298 
3299 unsigned long dev_trans_start(struct net_device *dev);
3300 
3301 void __netdev_watchdog_up(struct net_device *dev);
3302 
3303 void netif_carrier_on(struct net_device *dev);
3304 
3305 void netif_carrier_off(struct net_device *dev);
3306 
3307 /**
3308  *	netif_dormant_on - mark device as dormant.
3309  *	@dev: network device
3310  *
3311  * Mark device as dormant (as per RFC2863).
3312  *
3313  * The dormant state indicates that the relevant interface is not
3314  * actually in a condition to pass packets (i.e., it is not 'up') but is
3315  * in a "pending" state, waiting for some external event.  For "on-
3316  * demand" interfaces, this new state identifies the situation where the
3317  * interface is waiting for events to place it in the up state.
3318  *
3319  */
3320 static inline void netif_dormant_on(struct net_device *dev)
3321 {
3322 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3323 		linkwatch_fire_event(dev);
3324 }
3325 
3326 /**
3327  *	netif_dormant_off - set device as not dormant.
3328  *	@dev: network device
3329  *
3330  * Device is not in dormant state.
3331  */
3332 static inline void netif_dormant_off(struct net_device *dev)
3333 {
3334 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3335 		linkwatch_fire_event(dev);
3336 }
3337 
3338 /**
3339  *	netif_dormant - test if carrier present
3340  *	@dev: network device
3341  *
3342  * Check if carrier is present on device
3343  */
3344 static inline bool netif_dormant(const struct net_device *dev)
3345 {
3346 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3347 }
3348 
3349 
3350 /**
3351  *	netif_oper_up - test if device is operational
3352  *	@dev: network device
3353  *
3354  * Check if carrier is operational
3355  */
3356 static inline bool netif_oper_up(const struct net_device *dev)
3357 {
3358 	return (dev->operstate == IF_OPER_UP ||
3359 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3360 }
3361 
3362 /**
3363  *	netif_device_present - is device available or removed
3364  *	@dev: network device
3365  *
3366  * Check if device has not been removed from system.
3367  */
3368 static inline bool netif_device_present(struct net_device *dev)
3369 {
3370 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3371 }
3372 
3373 void netif_device_detach(struct net_device *dev);
3374 
3375 void netif_device_attach(struct net_device *dev);
3376 
3377 /*
3378  * Network interface message level settings
3379  */
3380 
3381 enum {
3382 	NETIF_MSG_DRV		= 0x0001,
3383 	NETIF_MSG_PROBE		= 0x0002,
3384 	NETIF_MSG_LINK		= 0x0004,
3385 	NETIF_MSG_TIMER		= 0x0008,
3386 	NETIF_MSG_IFDOWN	= 0x0010,
3387 	NETIF_MSG_IFUP		= 0x0020,
3388 	NETIF_MSG_RX_ERR	= 0x0040,
3389 	NETIF_MSG_TX_ERR	= 0x0080,
3390 	NETIF_MSG_TX_QUEUED	= 0x0100,
3391 	NETIF_MSG_INTR		= 0x0200,
3392 	NETIF_MSG_TX_DONE	= 0x0400,
3393 	NETIF_MSG_RX_STATUS	= 0x0800,
3394 	NETIF_MSG_PKTDATA	= 0x1000,
3395 	NETIF_MSG_HW		= 0x2000,
3396 	NETIF_MSG_WOL		= 0x4000,
3397 };
3398 
3399 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3400 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3401 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3402 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3403 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3404 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3405 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3406 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3407 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3408 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3409 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3410 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3411 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3412 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3413 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3414 
3415 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3416 {
3417 	/* use default */
3418 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3419 		return default_msg_enable_bits;
3420 	if (debug_value == 0)	/* no output */
3421 		return 0;
3422 	/* set low N bits */
3423 	return (1 << debug_value) - 1;
3424 }
3425 
3426 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3427 {
3428 	spin_lock(&txq->_xmit_lock);
3429 	txq->xmit_lock_owner = cpu;
3430 }
3431 
3432 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3433 {
3434 	spin_lock_bh(&txq->_xmit_lock);
3435 	txq->xmit_lock_owner = smp_processor_id();
3436 }
3437 
3438 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3439 {
3440 	bool ok = spin_trylock(&txq->_xmit_lock);
3441 	if (likely(ok))
3442 		txq->xmit_lock_owner = smp_processor_id();
3443 	return ok;
3444 }
3445 
3446 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3447 {
3448 	txq->xmit_lock_owner = -1;
3449 	spin_unlock(&txq->_xmit_lock);
3450 }
3451 
3452 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3453 {
3454 	txq->xmit_lock_owner = -1;
3455 	spin_unlock_bh(&txq->_xmit_lock);
3456 }
3457 
3458 static inline void txq_trans_update(struct netdev_queue *txq)
3459 {
3460 	if (txq->xmit_lock_owner != -1)
3461 		txq->trans_start = jiffies;
3462 }
3463 
3464 /**
3465  *	netif_tx_lock - grab network device transmit lock
3466  *	@dev: network device
3467  *
3468  * Get network device transmit lock
3469  */
3470 static inline void netif_tx_lock(struct net_device *dev)
3471 {
3472 	unsigned int i;
3473 	int cpu;
3474 
3475 	spin_lock(&dev->tx_global_lock);
3476 	cpu = smp_processor_id();
3477 	for (i = 0; i < dev->num_tx_queues; i++) {
3478 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3479 
3480 		/* We are the only thread of execution doing a
3481 		 * freeze, but we have to grab the _xmit_lock in
3482 		 * order to synchronize with threads which are in
3483 		 * the ->hard_start_xmit() handler and already
3484 		 * checked the frozen bit.
3485 		 */
3486 		__netif_tx_lock(txq, cpu);
3487 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3488 		__netif_tx_unlock(txq);
3489 	}
3490 }
3491 
3492 static inline void netif_tx_lock_bh(struct net_device *dev)
3493 {
3494 	local_bh_disable();
3495 	netif_tx_lock(dev);
3496 }
3497 
3498 static inline void netif_tx_unlock(struct net_device *dev)
3499 {
3500 	unsigned int i;
3501 
3502 	for (i = 0; i < dev->num_tx_queues; i++) {
3503 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3504 
3505 		/* No need to grab the _xmit_lock here.  If the
3506 		 * queue is not stopped for another reason, we
3507 		 * force a schedule.
3508 		 */
3509 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3510 		netif_schedule_queue(txq);
3511 	}
3512 	spin_unlock(&dev->tx_global_lock);
3513 }
3514 
3515 static inline void netif_tx_unlock_bh(struct net_device *dev)
3516 {
3517 	netif_tx_unlock(dev);
3518 	local_bh_enable();
3519 }
3520 
3521 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3522 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3523 		__netif_tx_lock(txq, cpu);		\
3524 	}						\
3525 }
3526 
3527 #define HARD_TX_TRYLOCK(dev, txq)			\
3528 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3529 		__netif_tx_trylock(txq) :		\
3530 		true )
3531 
3532 #define HARD_TX_UNLOCK(dev, txq) {			\
3533 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3534 		__netif_tx_unlock(txq);			\
3535 	}						\
3536 }
3537 
3538 static inline void netif_tx_disable(struct net_device *dev)
3539 {
3540 	unsigned int i;
3541 	int cpu;
3542 
3543 	local_bh_disable();
3544 	cpu = smp_processor_id();
3545 	for (i = 0; i < dev->num_tx_queues; i++) {
3546 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3547 
3548 		__netif_tx_lock(txq, cpu);
3549 		netif_tx_stop_queue(txq);
3550 		__netif_tx_unlock(txq);
3551 	}
3552 	local_bh_enable();
3553 }
3554 
3555 static inline void netif_addr_lock(struct net_device *dev)
3556 {
3557 	spin_lock(&dev->addr_list_lock);
3558 }
3559 
3560 static inline void netif_addr_lock_nested(struct net_device *dev)
3561 {
3562 	int subclass = SINGLE_DEPTH_NESTING;
3563 
3564 	if (dev->netdev_ops->ndo_get_lock_subclass)
3565 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3566 
3567 	spin_lock_nested(&dev->addr_list_lock, subclass);
3568 }
3569 
3570 static inline void netif_addr_lock_bh(struct net_device *dev)
3571 {
3572 	spin_lock_bh(&dev->addr_list_lock);
3573 }
3574 
3575 static inline void netif_addr_unlock(struct net_device *dev)
3576 {
3577 	spin_unlock(&dev->addr_list_lock);
3578 }
3579 
3580 static inline void netif_addr_unlock_bh(struct net_device *dev)
3581 {
3582 	spin_unlock_bh(&dev->addr_list_lock);
3583 }
3584 
3585 /*
3586  * dev_addrs walker. Should be used only for read access. Call with
3587  * rcu_read_lock held.
3588  */
3589 #define for_each_dev_addr(dev, ha) \
3590 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3591 
3592 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3593 
3594 void ether_setup(struct net_device *dev);
3595 
3596 /* Support for loadable net-drivers */
3597 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3598 				    unsigned char name_assign_type,
3599 				    void (*setup)(struct net_device *),
3600 				    unsigned int txqs, unsigned int rxqs);
3601 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3602 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3603 
3604 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3605 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3606 			 count)
3607 
3608 int register_netdev(struct net_device *dev);
3609 void unregister_netdev(struct net_device *dev);
3610 
3611 /* General hardware address lists handling functions */
3612 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3613 		   struct netdev_hw_addr_list *from_list, int addr_len);
3614 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3615 		      struct netdev_hw_addr_list *from_list, int addr_len);
3616 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3617 		       struct net_device *dev,
3618 		       int (*sync)(struct net_device *, const unsigned char *),
3619 		       int (*unsync)(struct net_device *,
3620 				     const unsigned char *));
3621 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3622 			  struct net_device *dev,
3623 			  int (*unsync)(struct net_device *,
3624 					const unsigned char *));
3625 void __hw_addr_init(struct netdev_hw_addr_list *list);
3626 
3627 /* Functions used for device addresses handling */
3628 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3629 		 unsigned char addr_type);
3630 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3631 		 unsigned char addr_type);
3632 void dev_addr_flush(struct net_device *dev);
3633 int dev_addr_init(struct net_device *dev);
3634 
3635 /* Functions used for unicast addresses handling */
3636 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3637 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3638 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3639 int dev_uc_sync(struct net_device *to, struct net_device *from);
3640 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3641 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3642 void dev_uc_flush(struct net_device *dev);
3643 void dev_uc_init(struct net_device *dev);
3644 
3645 /**
3646  *  __dev_uc_sync - Synchonize device's unicast list
3647  *  @dev:  device to sync
3648  *  @sync: function to call if address should be added
3649  *  @unsync: function to call if address should be removed
3650  *
3651  *  Add newly added addresses to the interface, and release
3652  *  addresses that have been deleted.
3653  **/
3654 static inline int __dev_uc_sync(struct net_device *dev,
3655 				int (*sync)(struct net_device *,
3656 					    const unsigned char *),
3657 				int (*unsync)(struct net_device *,
3658 					      const unsigned char *))
3659 {
3660 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3661 }
3662 
3663 /**
3664  *  __dev_uc_unsync - Remove synchronized addresses from device
3665  *  @dev:  device to sync
3666  *  @unsync: function to call if address should be removed
3667  *
3668  *  Remove all addresses that were added to the device by dev_uc_sync().
3669  **/
3670 static inline void __dev_uc_unsync(struct net_device *dev,
3671 				   int (*unsync)(struct net_device *,
3672 						 const unsigned char *))
3673 {
3674 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3675 }
3676 
3677 /* Functions used for multicast addresses handling */
3678 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3679 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3680 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3681 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3682 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3683 int dev_mc_sync(struct net_device *to, struct net_device *from);
3684 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3685 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3686 void dev_mc_flush(struct net_device *dev);
3687 void dev_mc_init(struct net_device *dev);
3688 
3689 /**
3690  *  __dev_mc_sync - Synchonize device's multicast list
3691  *  @dev:  device to sync
3692  *  @sync: function to call if address should be added
3693  *  @unsync: function to call if address should be removed
3694  *
3695  *  Add newly added addresses to the interface, and release
3696  *  addresses that have been deleted.
3697  **/
3698 static inline int __dev_mc_sync(struct net_device *dev,
3699 				int (*sync)(struct net_device *,
3700 					    const unsigned char *),
3701 				int (*unsync)(struct net_device *,
3702 					      const unsigned char *))
3703 {
3704 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3705 }
3706 
3707 /**
3708  *  __dev_mc_unsync - Remove synchronized addresses from device
3709  *  @dev:  device to sync
3710  *  @unsync: function to call if address should be removed
3711  *
3712  *  Remove all addresses that were added to the device by dev_mc_sync().
3713  **/
3714 static inline void __dev_mc_unsync(struct net_device *dev,
3715 				   int (*unsync)(struct net_device *,
3716 						 const unsigned char *))
3717 {
3718 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3719 }
3720 
3721 /* Functions used for secondary unicast and multicast support */
3722 void dev_set_rx_mode(struct net_device *dev);
3723 void __dev_set_rx_mode(struct net_device *dev);
3724 int dev_set_promiscuity(struct net_device *dev, int inc);
3725 int dev_set_allmulti(struct net_device *dev, int inc);
3726 void netdev_state_change(struct net_device *dev);
3727 void netdev_notify_peers(struct net_device *dev);
3728 void netdev_features_change(struct net_device *dev);
3729 /* Load a device via the kmod */
3730 void dev_load(struct net *net, const char *name);
3731 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3732 					struct rtnl_link_stats64 *storage);
3733 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3734 			     const struct net_device_stats *netdev_stats);
3735 
3736 extern int		netdev_max_backlog;
3737 extern int		netdev_tstamp_prequeue;
3738 extern int		weight_p;
3739 extern int		bpf_jit_enable;
3740 
3741 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3742 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3743 						     struct list_head **iter);
3744 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3745 						     struct list_head **iter);
3746 
3747 /* iterate through upper list, must be called under RCU read lock */
3748 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3749 	for (iter = &(dev)->adj_list.upper, \
3750 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3751 	     updev; \
3752 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3753 
3754 /* iterate through upper list, must be called under RCU read lock */
3755 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3756 	for (iter = &(dev)->all_adj_list.upper, \
3757 	     updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3758 	     updev; \
3759 	     updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3760 
3761 void *netdev_lower_get_next_private(struct net_device *dev,
3762 				    struct list_head **iter);
3763 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3764 					struct list_head **iter);
3765 
3766 #define netdev_for_each_lower_private(dev, priv, iter) \
3767 	for (iter = (dev)->adj_list.lower.next, \
3768 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3769 	     priv; \
3770 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3771 
3772 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3773 	for (iter = &(dev)->adj_list.lower, \
3774 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3775 	     priv; \
3776 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3777 
3778 void *netdev_lower_get_next(struct net_device *dev,
3779 				struct list_head **iter);
3780 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3781 	for (iter = (dev)->adj_list.lower.next, \
3782 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3783 	     ldev; \
3784 	     ldev = netdev_lower_get_next(dev, &(iter)))
3785 
3786 void *netdev_adjacent_get_private(struct list_head *adj_list);
3787 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3788 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3789 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3790 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3791 int netdev_master_upper_dev_link(struct net_device *dev,
3792 				 struct net_device *upper_dev,
3793 				 void *upper_priv, void *upper_info);
3794 void netdev_upper_dev_unlink(struct net_device *dev,
3795 			     struct net_device *upper_dev);
3796 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3797 void *netdev_lower_dev_get_private(struct net_device *dev,
3798 				   struct net_device *lower_dev);
3799 void netdev_lower_state_changed(struct net_device *lower_dev,
3800 				void *lower_state_info);
3801 
3802 /* RSS keys are 40 or 52 bytes long */
3803 #define NETDEV_RSS_KEY_LEN 52
3804 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3805 void netdev_rss_key_fill(void *buffer, size_t len);
3806 
3807 int dev_get_nest_level(struct net_device *dev,
3808 		       bool (*type_check)(const struct net_device *dev));
3809 int skb_checksum_help(struct sk_buff *skb);
3810 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3811 				  netdev_features_t features, bool tx_path);
3812 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3813 				    netdev_features_t features);
3814 
3815 struct netdev_bonding_info {
3816 	ifslave	slave;
3817 	ifbond	master;
3818 };
3819 
3820 struct netdev_notifier_bonding_info {
3821 	struct netdev_notifier_info info; /* must be first */
3822 	struct netdev_bonding_info  bonding_info;
3823 };
3824 
3825 void netdev_bonding_info_change(struct net_device *dev,
3826 				struct netdev_bonding_info *bonding_info);
3827 
3828 static inline
3829 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3830 {
3831 	return __skb_gso_segment(skb, features, true);
3832 }
3833 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3834 
3835 static inline bool can_checksum_protocol(netdev_features_t features,
3836 					 __be16 protocol)
3837 {
3838 	if (protocol == htons(ETH_P_FCOE))
3839 		return !!(features & NETIF_F_FCOE_CRC);
3840 
3841 	/* Assume this is an IP checksum (not SCTP CRC) */
3842 
3843 	if (features & NETIF_F_HW_CSUM) {
3844 		/* Can checksum everything */
3845 		return true;
3846 	}
3847 
3848 	switch (protocol) {
3849 	case htons(ETH_P_IP):
3850 		return !!(features & NETIF_F_IP_CSUM);
3851 	case htons(ETH_P_IPV6):
3852 		return !!(features & NETIF_F_IPV6_CSUM);
3853 	default:
3854 		return false;
3855 	}
3856 }
3857 
3858 /* Map an ethertype into IP protocol if possible */
3859 static inline int eproto_to_ipproto(int eproto)
3860 {
3861 	switch (eproto) {
3862 	case htons(ETH_P_IP):
3863 		return IPPROTO_IP;
3864 	case htons(ETH_P_IPV6):
3865 		return IPPROTO_IPV6;
3866 	default:
3867 		return -1;
3868 	}
3869 }
3870 
3871 #ifdef CONFIG_BUG
3872 void netdev_rx_csum_fault(struct net_device *dev);
3873 #else
3874 static inline void netdev_rx_csum_fault(struct net_device *dev)
3875 {
3876 }
3877 #endif
3878 /* rx skb timestamps */
3879 void net_enable_timestamp(void);
3880 void net_disable_timestamp(void);
3881 
3882 #ifdef CONFIG_PROC_FS
3883 int __init dev_proc_init(void);
3884 #else
3885 #define dev_proc_init() 0
3886 #endif
3887 
3888 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3889 					      struct sk_buff *skb, struct net_device *dev,
3890 					      bool more)
3891 {
3892 	skb->xmit_more = more ? 1 : 0;
3893 	return ops->ndo_start_xmit(skb, dev);
3894 }
3895 
3896 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3897 					    struct netdev_queue *txq, bool more)
3898 {
3899 	const struct net_device_ops *ops = dev->netdev_ops;
3900 	int rc;
3901 
3902 	rc = __netdev_start_xmit(ops, skb, dev, more);
3903 	if (rc == NETDEV_TX_OK)
3904 		txq_trans_update(txq);
3905 
3906 	return rc;
3907 }
3908 
3909 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3910 				const void *ns);
3911 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3912 				 const void *ns);
3913 
3914 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3915 {
3916 	return netdev_class_create_file_ns(class_attr, NULL);
3917 }
3918 
3919 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3920 {
3921 	netdev_class_remove_file_ns(class_attr, NULL);
3922 }
3923 
3924 extern struct kobj_ns_type_operations net_ns_type_operations;
3925 
3926 const char *netdev_drivername(const struct net_device *dev);
3927 
3928 void linkwatch_run_queue(void);
3929 
3930 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3931 							  netdev_features_t f2)
3932 {
3933 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3934 		if (f1 & NETIF_F_HW_CSUM)
3935 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3936 		else
3937 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3938 	}
3939 
3940 	return f1 & f2;
3941 }
3942 
3943 static inline netdev_features_t netdev_get_wanted_features(
3944 	struct net_device *dev)
3945 {
3946 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
3947 }
3948 netdev_features_t netdev_increment_features(netdev_features_t all,
3949 	netdev_features_t one, netdev_features_t mask);
3950 
3951 /* Allow TSO being used on stacked device :
3952  * Performing the GSO segmentation before last device
3953  * is a performance improvement.
3954  */
3955 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3956 							netdev_features_t mask)
3957 {
3958 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3959 }
3960 
3961 int __netdev_update_features(struct net_device *dev);
3962 void netdev_update_features(struct net_device *dev);
3963 void netdev_change_features(struct net_device *dev);
3964 
3965 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3966 					struct net_device *dev);
3967 
3968 netdev_features_t passthru_features_check(struct sk_buff *skb,
3969 					  struct net_device *dev,
3970 					  netdev_features_t features);
3971 netdev_features_t netif_skb_features(struct sk_buff *skb);
3972 
3973 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
3974 {
3975 	netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
3976 
3977 	/* check flags correspondence */
3978 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
3979 	BUILD_BUG_ON(SKB_GSO_UDP     != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
3980 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
3981 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
3982 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
3983 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
3984 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
3985 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
3986 	BUILD_BUG_ON(SKB_GSO_IPIP    != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
3987 	BUILD_BUG_ON(SKB_GSO_SIT     != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
3988 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
3989 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
3990 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
3991 
3992 	return (features & feature) == feature;
3993 }
3994 
3995 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
3996 {
3997 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
3998 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
3999 }
4000 
4001 static inline bool netif_needs_gso(struct sk_buff *skb,
4002 				   netdev_features_t features)
4003 {
4004 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4005 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4006 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4007 }
4008 
4009 static inline void netif_set_gso_max_size(struct net_device *dev,
4010 					  unsigned int size)
4011 {
4012 	dev->gso_max_size = size;
4013 }
4014 
4015 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4016 					int pulled_hlen, u16 mac_offset,
4017 					int mac_len)
4018 {
4019 	skb->protocol = protocol;
4020 	skb->encapsulation = 1;
4021 	skb_push(skb, pulled_hlen);
4022 	skb_reset_transport_header(skb);
4023 	skb->mac_header = mac_offset;
4024 	skb->network_header = skb->mac_header + mac_len;
4025 	skb->mac_len = mac_len;
4026 }
4027 
4028 static inline bool netif_is_macvlan(const struct net_device *dev)
4029 {
4030 	return dev->priv_flags & IFF_MACVLAN;
4031 }
4032 
4033 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4034 {
4035 	return dev->priv_flags & IFF_MACVLAN_PORT;
4036 }
4037 
4038 static inline bool netif_is_ipvlan(const struct net_device *dev)
4039 {
4040 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4041 }
4042 
4043 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4044 {
4045 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4046 }
4047 
4048 static inline bool netif_is_bond_master(const struct net_device *dev)
4049 {
4050 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4051 }
4052 
4053 static inline bool netif_is_bond_slave(const struct net_device *dev)
4054 {
4055 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4056 }
4057 
4058 static inline bool netif_supports_nofcs(struct net_device *dev)
4059 {
4060 	return dev->priv_flags & IFF_SUPP_NOFCS;
4061 }
4062 
4063 static inline bool netif_is_l3_master(const struct net_device *dev)
4064 {
4065 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4066 }
4067 
4068 static inline bool netif_is_l3_slave(const struct net_device *dev)
4069 {
4070 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4071 }
4072 
4073 static inline bool netif_is_bridge_master(const struct net_device *dev)
4074 {
4075 	return dev->priv_flags & IFF_EBRIDGE;
4076 }
4077 
4078 static inline bool netif_is_bridge_port(const struct net_device *dev)
4079 {
4080 	return dev->priv_flags & IFF_BRIDGE_PORT;
4081 }
4082 
4083 static inline bool netif_is_ovs_master(const struct net_device *dev)
4084 {
4085 	return dev->priv_flags & IFF_OPENVSWITCH;
4086 }
4087 
4088 static inline bool netif_is_team_master(const struct net_device *dev)
4089 {
4090 	return dev->priv_flags & IFF_TEAM;
4091 }
4092 
4093 static inline bool netif_is_team_port(const struct net_device *dev)
4094 {
4095 	return dev->priv_flags & IFF_TEAM_PORT;
4096 }
4097 
4098 static inline bool netif_is_lag_master(const struct net_device *dev)
4099 {
4100 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4101 }
4102 
4103 static inline bool netif_is_lag_port(const struct net_device *dev)
4104 {
4105 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4106 }
4107 
4108 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4109 {
4110 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4111 }
4112 
4113 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4114 static inline void netif_keep_dst(struct net_device *dev)
4115 {
4116 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4117 }
4118 
4119 extern struct pernet_operations __net_initdata loopback_net_ops;
4120 
4121 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4122 
4123 /* netdev_printk helpers, similar to dev_printk */
4124 
4125 static inline const char *netdev_name(const struct net_device *dev)
4126 {
4127 	if (!dev->name[0] || strchr(dev->name, '%'))
4128 		return "(unnamed net_device)";
4129 	return dev->name;
4130 }
4131 
4132 static inline const char *netdev_reg_state(const struct net_device *dev)
4133 {
4134 	switch (dev->reg_state) {
4135 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4136 	case NETREG_REGISTERED: return "";
4137 	case NETREG_UNREGISTERING: return " (unregistering)";
4138 	case NETREG_UNREGISTERED: return " (unregistered)";
4139 	case NETREG_RELEASED: return " (released)";
4140 	case NETREG_DUMMY: return " (dummy)";
4141 	}
4142 
4143 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4144 	return " (unknown)";
4145 }
4146 
4147 __printf(3, 4)
4148 void netdev_printk(const char *level, const struct net_device *dev,
4149 		   const char *format, ...);
4150 __printf(2, 3)
4151 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4152 __printf(2, 3)
4153 void netdev_alert(const struct net_device *dev, const char *format, ...);
4154 __printf(2, 3)
4155 void netdev_crit(const struct net_device *dev, const char *format, ...);
4156 __printf(2, 3)
4157 void netdev_err(const struct net_device *dev, const char *format, ...);
4158 __printf(2, 3)
4159 void netdev_warn(const struct net_device *dev, const char *format, ...);
4160 __printf(2, 3)
4161 void netdev_notice(const struct net_device *dev, const char *format, ...);
4162 __printf(2, 3)
4163 void netdev_info(const struct net_device *dev, const char *format, ...);
4164 
4165 #define MODULE_ALIAS_NETDEV(device) \
4166 	MODULE_ALIAS("netdev-" device)
4167 
4168 #if defined(CONFIG_DYNAMIC_DEBUG)
4169 #define netdev_dbg(__dev, format, args...)			\
4170 do {								\
4171 	dynamic_netdev_dbg(__dev, format, ##args);		\
4172 } while (0)
4173 #elif defined(DEBUG)
4174 #define netdev_dbg(__dev, format, args...)			\
4175 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4176 #else
4177 #define netdev_dbg(__dev, format, args...)			\
4178 ({								\
4179 	if (0)							\
4180 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4181 })
4182 #endif
4183 
4184 #if defined(VERBOSE_DEBUG)
4185 #define netdev_vdbg	netdev_dbg
4186 #else
4187 
4188 #define netdev_vdbg(dev, format, args...)			\
4189 ({								\
4190 	if (0)							\
4191 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4192 	0;							\
4193 })
4194 #endif
4195 
4196 /*
4197  * netdev_WARN() acts like dev_printk(), but with the key difference
4198  * of using a WARN/WARN_ON to get the message out, including the
4199  * file/line information and a backtrace.
4200  */
4201 #define netdev_WARN(dev, format, args...)			\
4202 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4203 	     netdev_reg_state(dev), ##args)
4204 
4205 /* netif printk helpers, similar to netdev_printk */
4206 
4207 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4208 do {					  			\
4209 	if (netif_msg_##type(priv))				\
4210 		netdev_printk(level, (dev), fmt, ##args);	\
4211 } while (0)
4212 
4213 #define netif_level(level, priv, type, dev, fmt, args...)	\
4214 do {								\
4215 	if (netif_msg_##type(priv))				\
4216 		netdev_##level(dev, fmt, ##args);		\
4217 } while (0)
4218 
4219 #define netif_emerg(priv, type, dev, fmt, args...)		\
4220 	netif_level(emerg, priv, type, dev, fmt, ##args)
4221 #define netif_alert(priv, type, dev, fmt, args...)		\
4222 	netif_level(alert, priv, type, dev, fmt, ##args)
4223 #define netif_crit(priv, type, dev, fmt, args...)		\
4224 	netif_level(crit, priv, type, dev, fmt, ##args)
4225 #define netif_err(priv, type, dev, fmt, args...)		\
4226 	netif_level(err, priv, type, dev, fmt, ##args)
4227 #define netif_warn(priv, type, dev, fmt, args...)		\
4228 	netif_level(warn, priv, type, dev, fmt, ##args)
4229 #define netif_notice(priv, type, dev, fmt, args...)		\
4230 	netif_level(notice, priv, type, dev, fmt, ##args)
4231 #define netif_info(priv, type, dev, fmt, args...)		\
4232 	netif_level(info, priv, type, dev, fmt, ##args)
4233 
4234 #if defined(CONFIG_DYNAMIC_DEBUG)
4235 #define netif_dbg(priv, type, netdev, format, args...)		\
4236 do {								\
4237 	if (netif_msg_##type(priv))				\
4238 		dynamic_netdev_dbg(netdev, format, ##args);	\
4239 } while (0)
4240 #elif defined(DEBUG)
4241 #define netif_dbg(priv, type, dev, format, args...)		\
4242 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4243 #else
4244 #define netif_dbg(priv, type, dev, format, args...)			\
4245 ({									\
4246 	if (0)								\
4247 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4248 	0;								\
4249 })
4250 #endif
4251 
4252 #if defined(VERBOSE_DEBUG)
4253 #define netif_vdbg	netif_dbg
4254 #else
4255 #define netif_vdbg(priv, type, dev, format, args...)		\
4256 ({								\
4257 	if (0)							\
4258 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4259 	0;							\
4260 })
4261 #endif
4262 
4263 /*
4264  *	The list of packet types we will receive (as opposed to discard)
4265  *	and the routines to invoke.
4266  *
4267  *	Why 16. Because with 16 the only overlap we get on a hash of the
4268  *	low nibble of the protocol value is RARP/SNAP/X.25.
4269  *
4270  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4271  *             sure which should go first, but I bet it won't make much
4272  *             difference if we are running VLANs.  The good news is that
4273  *             this protocol won't be in the list unless compiled in, so
4274  *             the average user (w/out VLANs) will not be adversely affected.
4275  *             --BLG
4276  *
4277  *		0800	IP
4278  *		8100    802.1Q VLAN
4279  *		0001	802.3
4280  *		0002	AX.25
4281  *		0004	802.2
4282  *		8035	RARP
4283  *		0005	SNAP
4284  *		0805	X.25
4285  *		0806	ARP
4286  *		8137	IPX
4287  *		0009	Localtalk
4288  *		86DD	IPv6
4289  */
4290 #define PTYPE_HASH_SIZE	(16)
4291 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4292 
4293 #endif	/* _LINUX_NETDEVICE_H */
4294