xref: /openbmc/linux/include/linux/netdevice.h (revision 56896ef5b990a3822da45f92e548c4c8ac301bb4)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Definitions for the Interfaces handler.
7  *
8  * Version:	@(#)dev.h	1.0.10	08/12/93
9  *
10  * Authors:	Ross Biro
11  *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12  *		Corey Minyard <wf-rch!minyard@relay.EU.net>
13  *		Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14  *		Alan Cox, <alan@lxorguk.ukuu.org.uk>
15  *		Bjorn Ekwall. <bj0rn@blox.se>
16  *              Pekka Riikonen <priikone@poseidon.pspt.fi>
17  *
18  *		This program is free software; you can redistribute it and/or
19  *		modify it under the terms of the GNU General Public License
20  *		as published by the Free Software Foundation; either version
21  *		2 of the License, or (at your option) any later version.
22  *
23  *		Moved to /usr/include/linux for NET3
24  */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27 
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35 
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41 
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49 
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
55 
56 struct netpoll_info;
57 struct device;
58 struct phy_device;
59 /* 802.11 specific */
60 struct wireless_dev;
61 /* 802.15.4 specific */
62 struct wpan_dev;
63 struct mpls_dev;
64 
65 void netdev_set_default_ethtool_ops(struct net_device *dev,
66 				    const struct ethtool_ops *ops);
67 
68 /* Backlog congestion levels */
69 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
70 #define NET_RX_DROP		1	/* packet dropped */
71 
72 /*
73  * Transmit return codes: transmit return codes originate from three different
74  * namespaces:
75  *
76  * - qdisc return codes
77  * - driver transmit return codes
78  * - errno values
79  *
80  * Drivers are allowed to return any one of those in their hard_start_xmit()
81  * function. Real network devices commonly used with qdiscs should only return
82  * the driver transmit return codes though - when qdiscs are used, the actual
83  * transmission happens asynchronously, so the value is not propagated to
84  * higher layers. Virtual network devices transmit synchronously; in this case
85  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
86  * others are propagated to higher layers.
87  */
88 
89 /* qdisc ->enqueue() return codes. */
90 #define NET_XMIT_SUCCESS	0x00
91 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
92 #define NET_XMIT_CN		0x02	/* congestion notification	*/
93 #define NET_XMIT_POLICED	0x03	/* skb is shot by police	*/
94 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
95 
96 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
97  * indicates that the device will soon be dropping packets, or already drops
98  * some packets of the same priority; prompting us to send less aggressively. */
99 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
100 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101 
102 /* Driver transmit return codes */
103 #define NETDEV_TX_MASK		0xf0
104 
105 enum netdev_tx {
106 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
107 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
108 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
109 	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 	bool	(*validate)(const char *ll_header, unsigned int len);
272 };
273 
274 /* These flag bits are private to the generic network queueing
275  * layer; they may not be explicitly referenced by any other
276  * code.
277  */
278 
279 enum netdev_state_t {
280 	__LINK_STATE_START,
281 	__LINK_STATE_PRESENT,
282 	__LINK_STATE_NOCARRIER,
283 	__LINK_STATE_LINKWATCH_PENDING,
284 	__LINK_STATE_DORMANT,
285 };
286 
287 
288 /*
289  * This structure holds boot-time configured netdevice settings. They
290  * are then used in the device probing.
291  */
292 struct netdev_boot_setup {
293 	char name[IFNAMSIZ];
294 	struct ifmap map;
295 };
296 #define NETDEV_BOOT_SETUP_MAX 8
297 
298 int __init netdev_boot_setup(char *str);
299 
300 /*
301  * Structure for NAPI scheduling similar to tasklet but with weighting
302  */
303 struct napi_struct {
304 	/* The poll_list must only be managed by the entity which
305 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
306 	 * whoever atomically sets that bit can add this napi_struct
307 	 * to the per-CPU poll_list, and whoever clears that bit
308 	 * can remove from the list right before clearing the bit.
309 	 */
310 	struct list_head	poll_list;
311 
312 	unsigned long		state;
313 	int			weight;
314 	unsigned int		gro_count;
315 	int			(*poll)(struct napi_struct *, int);
316 #ifdef CONFIG_NETPOLL
317 	spinlock_t		poll_lock;
318 	int			poll_owner;
319 #endif
320 	struct net_device	*dev;
321 	struct sk_buff		*gro_list;
322 	struct sk_buff		*skb;
323 	struct hrtimer		timer;
324 	struct list_head	dev_list;
325 	struct hlist_node	napi_hash_node;
326 	unsigned int		napi_id;
327 };
328 
329 enum {
330 	NAPI_STATE_SCHED,	/* Poll is scheduled */
331 	NAPI_STATE_DISABLE,	/* Disable pending */
332 	NAPI_STATE_NPSVC,	/* Netpoll - don't dequeue from poll_list */
333 	NAPI_STATE_HASHED,	/* In NAPI hash (busy polling possible) */
334 	NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
335 };
336 
337 enum gro_result {
338 	GRO_MERGED,
339 	GRO_MERGED_FREE,
340 	GRO_HELD,
341 	GRO_NORMAL,
342 	GRO_DROP,
343 };
344 typedef enum gro_result gro_result_t;
345 
346 /*
347  * enum rx_handler_result - Possible return values for rx_handlers.
348  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
349  * further.
350  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
351  * case skb->dev was changed by rx_handler.
352  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
353  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
354  *
355  * rx_handlers are functions called from inside __netif_receive_skb(), to do
356  * special processing of the skb, prior to delivery to protocol handlers.
357  *
358  * Currently, a net_device can only have a single rx_handler registered. Trying
359  * to register a second rx_handler will return -EBUSY.
360  *
361  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
362  * To unregister a rx_handler on a net_device, use
363  * netdev_rx_handler_unregister().
364  *
365  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
366  * do with the skb.
367  *
368  * If the rx_handler consumed the skb in some way, it should return
369  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
370  * the skb to be delivered in some other way.
371  *
372  * If the rx_handler changed skb->dev, to divert the skb to another
373  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
374  * new device will be called if it exists.
375  *
376  * If the rx_handler decides the skb should be ignored, it should return
377  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
378  * are registered on exact device (ptype->dev == skb->dev).
379  *
380  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
381  * delivered, it should return RX_HANDLER_PASS.
382  *
383  * A device without a registered rx_handler will behave as if rx_handler
384  * returned RX_HANDLER_PASS.
385  */
386 
387 enum rx_handler_result {
388 	RX_HANDLER_CONSUMED,
389 	RX_HANDLER_ANOTHER,
390 	RX_HANDLER_EXACT,
391 	RX_HANDLER_PASS,
392 };
393 typedef enum rx_handler_result rx_handler_result_t;
394 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
395 
396 void __napi_schedule(struct napi_struct *n);
397 void __napi_schedule_irqoff(struct napi_struct *n);
398 
399 static inline bool napi_disable_pending(struct napi_struct *n)
400 {
401 	return test_bit(NAPI_STATE_DISABLE, &n->state);
402 }
403 
404 /**
405  *	napi_schedule_prep - check if NAPI can be scheduled
406  *	@n: NAPI context
407  *
408  * Test if NAPI routine is already running, and if not mark
409  * it as running.  This is used as a condition variable to
410  * insure only one NAPI poll instance runs.  We also make
411  * sure there is no pending NAPI disable.
412  */
413 static inline bool napi_schedule_prep(struct napi_struct *n)
414 {
415 	return !napi_disable_pending(n) &&
416 		!test_and_set_bit(NAPI_STATE_SCHED, &n->state);
417 }
418 
419 /**
420  *	napi_schedule - schedule NAPI poll
421  *	@n: NAPI context
422  *
423  * Schedule NAPI poll routine to be called if it is not already
424  * running.
425  */
426 static inline void napi_schedule(struct napi_struct *n)
427 {
428 	if (napi_schedule_prep(n))
429 		__napi_schedule(n);
430 }
431 
432 /**
433  *	napi_schedule_irqoff - schedule NAPI poll
434  *	@n: NAPI context
435  *
436  * Variant of napi_schedule(), assuming hard irqs are masked.
437  */
438 static inline void napi_schedule_irqoff(struct napi_struct *n)
439 {
440 	if (napi_schedule_prep(n))
441 		__napi_schedule_irqoff(n);
442 }
443 
444 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
445 static inline bool napi_reschedule(struct napi_struct *napi)
446 {
447 	if (napi_schedule_prep(napi)) {
448 		__napi_schedule(napi);
449 		return true;
450 	}
451 	return false;
452 }
453 
454 void __napi_complete(struct napi_struct *n);
455 void napi_complete_done(struct napi_struct *n, int work_done);
456 /**
457  *	napi_complete - NAPI processing complete
458  *	@n: NAPI context
459  *
460  * Mark NAPI processing as complete.
461  * Consider using napi_complete_done() instead.
462  */
463 static inline void napi_complete(struct napi_struct *n)
464 {
465 	return napi_complete_done(n, 0);
466 }
467 
468 /**
469  *	napi_hash_add - add a NAPI to global hashtable
470  *	@napi: NAPI context
471  *
472  * Generate a new napi_id and store a @napi under it in napi_hash.
473  * Used for busy polling (CONFIG_NET_RX_BUSY_POLL).
474  * Note: This is normally automatically done from netif_napi_add(),
475  * so might disappear in a future Linux version.
476  */
477 void napi_hash_add(struct napi_struct *napi);
478 
479 /**
480  *	napi_hash_del - remove a NAPI from global table
481  *	@napi: NAPI context
482  *
483  * Warning: caller must observe RCU grace period
484  * before freeing memory containing @napi, if
485  * this function returns true.
486  * Note: core networking stack automatically calls it
487  * from netif_napi_del().
488  * Drivers might want to call this helper to combine all
489  * the needed RCU grace periods into a single one.
490  */
491 bool napi_hash_del(struct napi_struct *napi);
492 
493 /**
494  *	napi_disable - prevent NAPI from scheduling
495  *	@n: NAPI context
496  *
497  * Stop NAPI from being scheduled on this context.
498  * Waits till any outstanding processing completes.
499  */
500 void napi_disable(struct napi_struct *n);
501 
502 /**
503  *	napi_enable - enable NAPI scheduling
504  *	@n: NAPI context
505  *
506  * Resume NAPI from being scheduled on this context.
507  * Must be paired with napi_disable.
508  */
509 static inline void napi_enable(struct napi_struct *n)
510 {
511 	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
512 	smp_mb__before_atomic();
513 	clear_bit(NAPI_STATE_SCHED, &n->state);
514 	clear_bit(NAPI_STATE_NPSVC, &n->state);
515 }
516 
517 /**
518  *	napi_synchronize - wait until NAPI is not running
519  *	@n: NAPI context
520  *
521  * Wait until NAPI is done being scheduled on this context.
522  * Waits till any outstanding processing completes but
523  * does not disable future activations.
524  */
525 static inline void napi_synchronize(const struct napi_struct *n)
526 {
527 	if (IS_ENABLED(CONFIG_SMP))
528 		while (test_bit(NAPI_STATE_SCHED, &n->state))
529 			msleep(1);
530 	else
531 		barrier();
532 }
533 
534 enum netdev_queue_state_t {
535 	__QUEUE_STATE_DRV_XOFF,
536 	__QUEUE_STATE_STACK_XOFF,
537 	__QUEUE_STATE_FROZEN,
538 };
539 
540 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
541 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
542 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
543 
544 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
545 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
546 					QUEUE_STATE_FROZEN)
547 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
548 					QUEUE_STATE_FROZEN)
549 
550 /*
551  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
552  * netif_tx_* functions below are used to manipulate this flag.  The
553  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
554  * queue independently.  The netif_xmit_*stopped functions below are called
555  * to check if the queue has been stopped by the driver or stack (either
556  * of the XOFF bits are set in the state).  Drivers should not need to call
557  * netif_xmit*stopped functions, they should only be using netif_tx_*.
558  */
559 
560 struct netdev_queue {
561 /*
562  * read-mostly part
563  */
564 	struct net_device	*dev;
565 	struct Qdisc __rcu	*qdisc;
566 	struct Qdisc		*qdisc_sleeping;
567 #ifdef CONFIG_SYSFS
568 	struct kobject		kobj;
569 #endif
570 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
571 	int			numa_node;
572 #endif
573 /*
574  * write-mostly part
575  */
576 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
577 	int			xmit_lock_owner;
578 	/*
579 	 * please use this field instead of dev->trans_start
580 	 */
581 	unsigned long		trans_start;
582 
583 	/*
584 	 * Number of TX timeouts for this queue
585 	 * (/sys/class/net/DEV/Q/trans_timeout)
586 	 */
587 	unsigned long		trans_timeout;
588 
589 	unsigned long		state;
590 
591 #ifdef CONFIG_BQL
592 	struct dql		dql;
593 #endif
594 	unsigned long		tx_maxrate;
595 } ____cacheline_aligned_in_smp;
596 
597 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
598 {
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 	return q->numa_node;
601 #else
602 	return NUMA_NO_NODE;
603 #endif
604 }
605 
606 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
607 {
608 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
609 	q->numa_node = node;
610 #endif
611 }
612 
613 #ifdef CONFIG_RPS
614 /*
615  * This structure holds an RPS map which can be of variable length.  The
616  * map is an array of CPUs.
617  */
618 struct rps_map {
619 	unsigned int len;
620 	struct rcu_head rcu;
621 	u16 cpus[0];
622 };
623 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
624 
625 /*
626  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627  * tail pointer for that CPU's input queue at the time of last enqueue, and
628  * a hardware filter index.
629  */
630 struct rps_dev_flow {
631 	u16 cpu;
632 	u16 filter;
633 	unsigned int last_qtail;
634 };
635 #define RPS_NO_FILTER 0xffff
636 
637 /*
638  * The rps_dev_flow_table structure contains a table of flow mappings.
639  */
640 struct rps_dev_flow_table {
641 	unsigned int mask;
642 	struct rcu_head rcu;
643 	struct rps_dev_flow flows[0];
644 };
645 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
646     ((_num) * sizeof(struct rps_dev_flow)))
647 
648 /*
649  * The rps_sock_flow_table contains mappings of flows to the last CPU
650  * on which they were processed by the application (set in recvmsg).
651  * Each entry is a 32bit value. Upper part is the high-order bits
652  * of flow hash, lower part is CPU number.
653  * rps_cpu_mask is used to partition the space, depending on number of
654  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
656  * meaning we use 32-6=26 bits for the hash.
657  */
658 struct rps_sock_flow_table {
659 	u32	mask;
660 
661 	u32	ents[0] ____cacheline_aligned_in_smp;
662 };
663 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
664 
665 #define RPS_NO_CPU 0xffff
666 
667 extern u32 rps_cpu_mask;
668 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
669 
670 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
671 					u32 hash)
672 {
673 	if (table && hash) {
674 		unsigned int index = hash & table->mask;
675 		u32 val = hash & ~rps_cpu_mask;
676 
677 		/* We only give a hint, preemption can change CPU under us */
678 		val |= raw_smp_processor_id();
679 
680 		if (table->ents[index] != val)
681 			table->ents[index] = val;
682 	}
683 }
684 
685 #ifdef CONFIG_RFS_ACCEL
686 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
687 			 u16 filter_id);
688 #endif
689 #endif /* CONFIG_RPS */
690 
691 /* This structure contains an instance of an RX queue. */
692 struct netdev_rx_queue {
693 #ifdef CONFIG_RPS
694 	struct rps_map __rcu		*rps_map;
695 	struct rps_dev_flow_table __rcu	*rps_flow_table;
696 #endif
697 	struct kobject			kobj;
698 	struct net_device		*dev;
699 } ____cacheline_aligned_in_smp;
700 
701 /*
702  * RX queue sysfs structures and functions.
703  */
704 struct rx_queue_attribute {
705 	struct attribute attr;
706 	ssize_t (*show)(struct netdev_rx_queue *queue,
707 	    struct rx_queue_attribute *attr, char *buf);
708 	ssize_t (*store)(struct netdev_rx_queue *queue,
709 	    struct rx_queue_attribute *attr, const char *buf, size_t len);
710 };
711 
712 #ifdef CONFIG_XPS
713 /*
714  * This structure holds an XPS map which can be of variable length.  The
715  * map is an array of queues.
716  */
717 struct xps_map {
718 	unsigned int len;
719 	unsigned int alloc_len;
720 	struct rcu_head rcu;
721 	u16 queues[0];
722 };
723 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
724 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725        - sizeof(struct xps_map)) / sizeof(u16))
726 
727 /*
728  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
729  */
730 struct xps_dev_maps {
731 	struct rcu_head rcu;
732 	struct xps_map __rcu *cpu_map[0];
733 };
734 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) +		\
735     (nr_cpu_ids * sizeof(struct xps_map *)))
736 #endif /* CONFIG_XPS */
737 
738 #define TC_MAX_QUEUE	16
739 #define TC_BITMASK	15
740 /* HW offloaded queuing disciplines txq count and offset maps */
741 struct netdev_tc_txq {
742 	u16 count;
743 	u16 offset;
744 };
745 
746 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
747 /*
748  * This structure is to hold information about the device
749  * configured to run FCoE protocol stack.
750  */
751 struct netdev_fcoe_hbainfo {
752 	char	manufacturer[64];
753 	char	serial_number[64];
754 	char	hardware_version[64];
755 	char	driver_version[64];
756 	char	optionrom_version[64];
757 	char	firmware_version[64];
758 	char	model[256];
759 	char	model_description[256];
760 };
761 #endif
762 
763 #define MAX_PHYS_ITEM_ID_LEN 32
764 
765 /* This structure holds a unique identifier to identify some
766  * physical item (port for example) used by a netdevice.
767  */
768 struct netdev_phys_item_id {
769 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
770 	unsigned char id_len;
771 };
772 
773 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 					    struct netdev_phys_item_id *b)
775 {
776 	return a->id_len == b->id_len &&
777 	       memcmp(a->id, b->id, a->id_len) == 0;
778 }
779 
780 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 				       struct sk_buff *skb);
782 
783 /* These structures hold the attributes of qdisc and classifiers
784  * that are being passed to the netdevice through the setup_tc op.
785  */
786 enum {
787 	TC_SETUP_MQPRIO,
788 	TC_SETUP_CLSU32,
789 	TC_SETUP_CLSFLOWER,
790 };
791 
792 struct tc_cls_u32_offload;
793 
794 struct tc_to_netdev {
795 	unsigned int type;
796 	union {
797 		u8 tc;
798 		struct tc_cls_u32_offload *cls_u32;
799 		struct tc_cls_flower_offload *cls_flower;
800 	};
801 };
802 
803 
804 /*
805  * This structure defines the management hooks for network devices.
806  * The following hooks can be defined; unless noted otherwise, they are
807  * optional and can be filled with a null pointer.
808  *
809  * int (*ndo_init)(struct net_device *dev);
810  *     This function is called once when a network device is registered.
811  *     The network device can use this for any late stage initialization
812  *     or semantic validation. It can fail with an error code which will
813  *     be propagated back to register_netdev.
814  *
815  * void (*ndo_uninit)(struct net_device *dev);
816  *     This function is called when device is unregistered or when registration
817  *     fails. It is not called if init fails.
818  *
819  * int (*ndo_open)(struct net_device *dev);
820  *     This function is called when a network device transitions to the up
821  *     state.
822  *
823  * int (*ndo_stop)(struct net_device *dev);
824  *     This function is called when a network device transitions to the down
825  *     state.
826  *
827  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
828  *                               struct net_device *dev);
829  *	Called when a packet needs to be transmitted.
830  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
831  *	the queue before that can happen; it's for obsolete devices and weird
832  *	corner cases, but the stack really does a non-trivial amount
833  *	of useless work if you return NETDEV_TX_BUSY.
834  *        (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
835  *	Required; cannot be NULL.
836  *
837  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
838  *		netdev_features_t features);
839  *	Adjusts the requested feature flags according to device-specific
840  *	constraints, and returns the resulting flags. Must not modify
841  *	the device state.
842  *
843  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
844  *                         void *accel_priv, select_queue_fallback_t fallback);
845  *	Called to decide which queue to use when device supports multiple
846  *	transmit queues.
847  *
848  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
849  *	This function is called to allow device receiver to make
850  *	changes to configuration when multicast or promiscuous is enabled.
851  *
852  * void (*ndo_set_rx_mode)(struct net_device *dev);
853  *	This function is called device changes address list filtering.
854  *	If driver handles unicast address filtering, it should set
855  *	IFF_UNICAST_FLT in its priv_flags.
856  *
857  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
858  *	This function  is called when the Media Access Control address
859  *	needs to be changed. If this interface is not defined, the
860  *	MAC address can not be changed.
861  *
862  * int (*ndo_validate_addr)(struct net_device *dev);
863  *	Test if Media Access Control address is valid for the device.
864  *
865  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
866  *	Called when a user requests an ioctl which can't be handled by
867  *	the generic interface code. If not defined ioctls return
868  *	not supported error code.
869  *
870  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
871  *	Used to set network devices bus interface parameters. This interface
872  *	is retained for legacy reasons; new devices should use the bus
873  *	interface (PCI) for low level management.
874  *
875  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
876  *	Called when a user wants to change the Maximum Transfer Unit
877  *	of a device. If not defined, any request to change MTU will
878  *	will return an error.
879  *
880  * void (*ndo_tx_timeout)(struct net_device *dev);
881  *	Callback used when the transmitter has not made any progress
882  *	for dev->watchdog ticks.
883  *
884  * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
885  *                      struct rtnl_link_stats64 *storage);
886  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
887  *	Called when a user wants to get the network device usage
888  *	statistics. Drivers must do one of the following:
889  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
890  *	   rtnl_link_stats64 structure passed by the caller.
891  *	2. Define @ndo_get_stats to update a net_device_stats structure
892  *	   (which should normally be dev->stats) and return a pointer to
893  *	   it. The structure may be changed asynchronously only if each
894  *	   field is written atomically.
895  *	3. Update dev->stats asynchronously and atomically, and define
896  *	   neither operation.
897  *
898  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
899  *	If device supports VLAN filtering this function is called when a
900  *	VLAN id is registered.
901  *
902  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
903  *	If device supports VLAN filtering this function is called when a
904  *	VLAN id is unregistered.
905  *
906  * void (*ndo_poll_controller)(struct net_device *dev);
907  *
908  *	SR-IOV management functions.
909  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
910  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
911  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
912  *			  int max_tx_rate);
913  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
914  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
915  * int (*ndo_get_vf_config)(struct net_device *dev,
916  *			    int vf, struct ifla_vf_info *ivf);
917  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
918  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
919  *			  struct nlattr *port[]);
920  *
921  *      Enable or disable the VF ability to query its RSS Redirection Table and
922  *      Hash Key. This is needed since on some devices VF share this information
923  *      with PF and querying it may introduce a theoretical security risk.
924  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
925  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
926  * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
927  * 	Called to setup 'tc' number of traffic classes in the net device. This
928  * 	is always called from the stack with the rtnl lock held and netif tx
929  * 	queues stopped. This allows the netdevice to perform queue management
930  * 	safely.
931  *
932  *	Fiber Channel over Ethernet (FCoE) offload functions.
933  * int (*ndo_fcoe_enable)(struct net_device *dev);
934  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
935  *	so the underlying device can perform whatever needed configuration or
936  *	initialization to support acceleration of FCoE traffic.
937  *
938  * int (*ndo_fcoe_disable)(struct net_device *dev);
939  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
940  *	so the underlying device can perform whatever needed clean-ups to
941  *	stop supporting acceleration of FCoE traffic.
942  *
943  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
944  *			     struct scatterlist *sgl, unsigned int sgc);
945  *	Called when the FCoE Initiator wants to initialize an I/O that
946  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
947  *	perform necessary setup and returns 1 to indicate the device is set up
948  *	successfully to perform DDP on this I/O, otherwise this returns 0.
949  *
950  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
951  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
952  *	indicated by the FC exchange id 'xid', so the underlying device can
953  *	clean up and reuse resources for later DDP requests.
954  *
955  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
956  *			      struct scatterlist *sgl, unsigned int sgc);
957  *	Called when the FCoE Target wants to initialize an I/O that
958  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
959  *	perform necessary setup and returns 1 to indicate the device is set up
960  *	successfully to perform DDP on this I/O, otherwise this returns 0.
961  *
962  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
963  *			       struct netdev_fcoe_hbainfo *hbainfo);
964  *	Called when the FCoE Protocol stack wants information on the underlying
965  *	device. This information is utilized by the FCoE protocol stack to
966  *	register attributes with Fiber Channel management service as per the
967  *	FC-GS Fabric Device Management Information(FDMI) specification.
968  *
969  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
970  *	Called when the underlying device wants to override default World Wide
971  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
972  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
973  *	protocol stack to use.
974  *
975  *	RFS acceleration.
976  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
977  *			    u16 rxq_index, u32 flow_id);
978  *	Set hardware filter for RFS.  rxq_index is the target queue index;
979  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
980  *	Return the filter ID on success, or a negative error code.
981  *
982  *	Slave management functions (for bridge, bonding, etc).
983  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
984  *	Called to make another netdev an underling.
985  *
986  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
987  *	Called to release previously enslaved netdev.
988  *
989  *      Feature/offload setting functions.
990  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
991  *	Called to update device configuration to new features. Passed
992  *	feature set might be less than what was returned by ndo_fix_features()).
993  *	Must return >0 or -errno if it changed dev->features itself.
994  *
995  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
996  *		      struct net_device *dev,
997  *		      const unsigned char *addr, u16 vid, u16 flags)
998  *	Adds an FDB entry to dev for addr.
999  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1000  *		      struct net_device *dev,
1001  *		      const unsigned char *addr, u16 vid)
1002  *	Deletes the FDB entry from dev coresponding to addr.
1003  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1004  *		       struct net_device *dev, struct net_device *filter_dev,
1005  *		       int idx)
1006  *	Used to add FDB entries to dump requests. Implementers should add
1007  *	entries to skb and update idx with the number of entries.
1008  *
1009  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1010  *			     u16 flags)
1011  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1012  *			     struct net_device *dev, u32 filter_mask,
1013  *			     int nlflags)
1014  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1015  *			     u16 flags);
1016  *
1017  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1018  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1019  *	which do not represent real hardware may define this to allow their
1020  *	userspace components to manage their virtual carrier state. Devices
1021  *	that determine carrier state from physical hardware properties (eg
1022  *	network cables) or protocol-dependent mechanisms (eg
1023  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1024  *
1025  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1026  *			       struct netdev_phys_item_id *ppid);
1027  *	Called to get ID of physical port of this device. If driver does
1028  *	not implement this, it is assumed that the hw is not able to have
1029  *	multiple net devices on single physical port.
1030  *
1031  * void (*ndo_add_vxlan_port)(struct  net_device *dev,
1032  *			      sa_family_t sa_family, __be16 port);
1033  *	Called by vxlan to notify a driver about the UDP port and socket
1034  *	address family that vxlan is listening to. It is called only when
1035  *	a new port starts listening. The operation is protected by the
1036  *	vxlan_net->sock_lock.
1037  *
1038  * void (*ndo_add_geneve_port)(struct net_device *dev,
1039  *			       sa_family_t sa_family, __be16 port);
1040  *	Called by geneve to notify a driver about the UDP port and socket
1041  *	address family that geneve is listnening to. It is called only when
1042  *	a new port starts listening. The operation is protected by the
1043  *	geneve_net->sock_lock.
1044  *
1045  * void (*ndo_del_geneve_port)(struct net_device *dev,
1046  *			       sa_family_t sa_family, __be16 port);
1047  *	Called by geneve to notify the driver about a UDP port and socket
1048  *	address family that geneve is not listening to anymore. The operation
1049  *	is protected by the geneve_net->sock_lock.
1050  *
1051  * void (*ndo_del_vxlan_port)(struct  net_device *dev,
1052  *			      sa_family_t sa_family, __be16 port);
1053  *	Called by vxlan to notify the driver about a UDP port and socket
1054  *	address family that vxlan is not listening to anymore. The operation
1055  *	is protected by the vxlan_net->sock_lock.
1056  *
1057  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1058  *				 struct net_device *dev)
1059  *	Called by upper layer devices to accelerate switching or other
1060  *	station functionality into hardware. 'pdev is the lowerdev
1061  *	to use for the offload and 'dev' is the net device that will
1062  *	back the offload. Returns a pointer to the private structure
1063  *	the upper layer will maintain.
1064  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1065  *	Called by upper layer device to delete the station created
1066  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1067  *	the station and priv is the structure returned by the add
1068  *	operation.
1069  * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1070  *				      struct net_device *dev,
1071  *				      void *priv);
1072  *	Callback to use for xmit over the accelerated station. This
1073  *	is used in place of ndo_start_xmit on accelerated net
1074  *	devices.
1075  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1076  *					   struct net_device *dev
1077  *					   netdev_features_t features);
1078  *	Called by core transmit path to determine if device is capable of
1079  *	performing offload operations on a given packet. This is to give
1080  *	the device an opportunity to implement any restrictions that cannot
1081  *	be otherwise expressed by feature flags. The check is called with
1082  *	the set of features that the stack has calculated and it returns
1083  *	those the driver believes to be appropriate.
1084  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1085  *			     int queue_index, u32 maxrate);
1086  *	Called when a user wants to set a max-rate limitation of specific
1087  *	TX queue.
1088  * int (*ndo_get_iflink)(const struct net_device *dev);
1089  *	Called to get the iflink value of this device.
1090  * void (*ndo_change_proto_down)(struct net_device *dev,
1091  *				 bool proto_down);
1092  *	This function is used to pass protocol port error state information
1093  *	to the switch driver. The switch driver can react to the proto_down
1094  *      by doing a phys down on the associated switch port.
1095  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1096  *	This function is used to get egress tunnel information for given skb.
1097  *	This is useful for retrieving outer tunnel header parameters while
1098  *	sampling packet.
1099  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1100  *	This function is used to specify the headroom that the skb must
1101  *	consider when allocation skb during packet reception. Setting
1102  *	appropriate rx headroom value allows avoiding skb head copy on
1103  *	forward. Setting a negative value resets the rx headroom to the
1104  *	default value.
1105  *
1106  */
1107 struct net_device_ops {
1108 	int			(*ndo_init)(struct net_device *dev);
1109 	void			(*ndo_uninit)(struct net_device *dev);
1110 	int			(*ndo_open)(struct net_device *dev);
1111 	int			(*ndo_stop)(struct net_device *dev);
1112 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1113 						  struct net_device *dev);
1114 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1115 						      struct net_device *dev,
1116 						      netdev_features_t features);
1117 	u16			(*ndo_select_queue)(struct net_device *dev,
1118 						    struct sk_buff *skb,
1119 						    void *accel_priv,
1120 						    select_queue_fallback_t fallback);
1121 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1122 						       int flags);
1123 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1124 	int			(*ndo_set_mac_address)(struct net_device *dev,
1125 						       void *addr);
1126 	int			(*ndo_validate_addr)(struct net_device *dev);
1127 	int			(*ndo_do_ioctl)(struct net_device *dev,
1128 					        struct ifreq *ifr, int cmd);
1129 	int			(*ndo_set_config)(struct net_device *dev,
1130 					          struct ifmap *map);
1131 	int			(*ndo_change_mtu)(struct net_device *dev,
1132 						  int new_mtu);
1133 	int			(*ndo_neigh_setup)(struct net_device *dev,
1134 						   struct neigh_parms *);
1135 	void			(*ndo_tx_timeout) (struct net_device *dev);
1136 
1137 	struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1138 						     struct rtnl_link_stats64 *storage);
1139 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1140 
1141 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1142 						       __be16 proto, u16 vid);
1143 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1144 						        __be16 proto, u16 vid);
1145 #ifdef CONFIG_NET_POLL_CONTROLLER
1146 	void                    (*ndo_poll_controller)(struct net_device *dev);
1147 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1148 						     struct netpoll_info *info);
1149 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1150 #endif
1151 #ifdef CONFIG_NET_RX_BUSY_POLL
1152 	int			(*ndo_busy_poll)(struct napi_struct *dev);
1153 #endif
1154 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1155 						  int queue, u8 *mac);
1156 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1157 						   int queue, u16 vlan, u8 qos);
1158 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1159 						   int vf, int min_tx_rate,
1160 						   int max_tx_rate);
1161 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1162 						       int vf, bool setting);
1163 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1164 						    int vf, bool setting);
1165 	int			(*ndo_get_vf_config)(struct net_device *dev,
1166 						     int vf,
1167 						     struct ifla_vf_info *ivf);
1168 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1169 							 int vf, int link_state);
1170 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1171 						    int vf,
1172 						    struct ifla_vf_stats
1173 						    *vf_stats);
1174 	int			(*ndo_set_vf_port)(struct net_device *dev,
1175 						   int vf,
1176 						   struct nlattr *port[]);
1177 	int			(*ndo_get_vf_port)(struct net_device *dev,
1178 						   int vf, struct sk_buff *skb);
1179 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1180 						   int vf, u64 guid,
1181 						   int guid_type);
1182 	int			(*ndo_set_vf_rss_query_en)(
1183 						   struct net_device *dev,
1184 						   int vf, bool setting);
1185 	int			(*ndo_setup_tc)(struct net_device *dev,
1186 						u32 handle,
1187 						__be16 protocol,
1188 						struct tc_to_netdev *tc);
1189 #if IS_ENABLED(CONFIG_FCOE)
1190 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1191 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1192 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1193 						      u16 xid,
1194 						      struct scatterlist *sgl,
1195 						      unsigned int sgc);
1196 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1197 						     u16 xid);
1198 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1199 						       u16 xid,
1200 						       struct scatterlist *sgl,
1201 						       unsigned int sgc);
1202 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1203 							struct netdev_fcoe_hbainfo *hbainfo);
1204 #endif
1205 
1206 #if IS_ENABLED(CONFIG_LIBFCOE)
1207 #define NETDEV_FCOE_WWNN 0
1208 #define NETDEV_FCOE_WWPN 1
1209 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1210 						    u64 *wwn, int type);
1211 #endif
1212 
1213 #ifdef CONFIG_RFS_ACCEL
1214 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1215 						     const struct sk_buff *skb,
1216 						     u16 rxq_index,
1217 						     u32 flow_id);
1218 #endif
1219 	int			(*ndo_add_slave)(struct net_device *dev,
1220 						 struct net_device *slave_dev);
1221 	int			(*ndo_del_slave)(struct net_device *dev,
1222 						 struct net_device *slave_dev);
1223 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1224 						    netdev_features_t features);
1225 	int			(*ndo_set_features)(struct net_device *dev,
1226 						    netdev_features_t features);
1227 	int			(*ndo_neigh_construct)(struct neighbour *n);
1228 	void			(*ndo_neigh_destroy)(struct neighbour *n);
1229 
1230 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1231 					       struct nlattr *tb[],
1232 					       struct net_device *dev,
1233 					       const unsigned char *addr,
1234 					       u16 vid,
1235 					       u16 flags);
1236 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1237 					       struct nlattr *tb[],
1238 					       struct net_device *dev,
1239 					       const unsigned char *addr,
1240 					       u16 vid);
1241 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1242 						struct netlink_callback *cb,
1243 						struct net_device *dev,
1244 						struct net_device *filter_dev,
1245 						int idx);
1246 
1247 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1248 						      struct nlmsghdr *nlh,
1249 						      u16 flags);
1250 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1251 						      u32 pid, u32 seq,
1252 						      struct net_device *dev,
1253 						      u32 filter_mask,
1254 						      int nlflags);
1255 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1256 						      struct nlmsghdr *nlh,
1257 						      u16 flags);
1258 	int			(*ndo_change_carrier)(struct net_device *dev,
1259 						      bool new_carrier);
1260 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1261 							struct netdev_phys_item_id *ppid);
1262 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1263 							  char *name, size_t len);
1264 	void			(*ndo_add_vxlan_port)(struct  net_device *dev,
1265 						      sa_family_t sa_family,
1266 						      __be16 port);
1267 	void			(*ndo_del_vxlan_port)(struct  net_device *dev,
1268 						      sa_family_t sa_family,
1269 						      __be16 port);
1270 	void			(*ndo_add_geneve_port)(struct  net_device *dev,
1271 						       sa_family_t sa_family,
1272 						       __be16 port);
1273 	void			(*ndo_del_geneve_port)(struct  net_device *dev,
1274 						       sa_family_t sa_family,
1275 						       __be16 port);
1276 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1277 							struct net_device *dev);
1278 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1279 							void *priv);
1280 
1281 	netdev_tx_t		(*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1282 							struct net_device *dev,
1283 							void *priv);
1284 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1285 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1286 						      int queue_index,
1287 						      u32 maxrate);
1288 	int			(*ndo_get_iflink)(const struct net_device *dev);
1289 	int			(*ndo_change_proto_down)(struct net_device *dev,
1290 							 bool proto_down);
1291 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1292 						       struct sk_buff *skb);
1293 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1294 						       int needed_headroom);
1295 };
1296 
1297 /**
1298  * enum net_device_priv_flags - &struct net_device priv_flags
1299  *
1300  * These are the &struct net_device, they are only set internally
1301  * by drivers and used in the kernel. These flags are invisible to
1302  * userspace; this means that the order of these flags can change
1303  * during any kernel release.
1304  *
1305  * You should have a pretty good reason to be extending these flags.
1306  *
1307  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1308  * @IFF_EBRIDGE: Ethernet bridging device
1309  * @IFF_BONDING: bonding master or slave
1310  * @IFF_ISATAP: ISATAP interface (RFC4214)
1311  * @IFF_WAN_HDLC: WAN HDLC device
1312  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1313  *	release skb->dst
1314  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1315  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1316  * @IFF_MACVLAN_PORT: device used as macvlan port
1317  * @IFF_BRIDGE_PORT: device used as bridge port
1318  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1319  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1320  * @IFF_UNICAST_FLT: Supports unicast filtering
1321  * @IFF_TEAM_PORT: device used as team port
1322  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1323  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1324  *	change when it's running
1325  * @IFF_MACVLAN: Macvlan device
1326  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1327  *	underlying stacked devices
1328  * @IFF_IPVLAN_MASTER: IPvlan master device
1329  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1330  * @IFF_L3MDEV_MASTER: device is an L3 master device
1331  * @IFF_NO_QUEUE: device can run without qdisc attached
1332  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1333  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1334  * @IFF_TEAM: device is a team device
1335  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1336  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1337  *	entity (i.e. the master device for bridged veth)
1338  * @IFF_MACSEC: device is a MACsec device
1339  */
1340 enum netdev_priv_flags {
1341 	IFF_802_1Q_VLAN			= 1<<0,
1342 	IFF_EBRIDGE			= 1<<1,
1343 	IFF_BONDING			= 1<<2,
1344 	IFF_ISATAP			= 1<<3,
1345 	IFF_WAN_HDLC			= 1<<4,
1346 	IFF_XMIT_DST_RELEASE		= 1<<5,
1347 	IFF_DONT_BRIDGE			= 1<<6,
1348 	IFF_DISABLE_NETPOLL		= 1<<7,
1349 	IFF_MACVLAN_PORT		= 1<<8,
1350 	IFF_BRIDGE_PORT			= 1<<9,
1351 	IFF_OVS_DATAPATH		= 1<<10,
1352 	IFF_TX_SKB_SHARING		= 1<<11,
1353 	IFF_UNICAST_FLT			= 1<<12,
1354 	IFF_TEAM_PORT			= 1<<13,
1355 	IFF_SUPP_NOFCS			= 1<<14,
1356 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1357 	IFF_MACVLAN			= 1<<16,
1358 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1359 	IFF_IPVLAN_MASTER		= 1<<18,
1360 	IFF_IPVLAN_SLAVE		= 1<<19,
1361 	IFF_L3MDEV_MASTER		= 1<<20,
1362 	IFF_NO_QUEUE			= 1<<21,
1363 	IFF_OPENVSWITCH			= 1<<22,
1364 	IFF_L3MDEV_SLAVE		= 1<<23,
1365 	IFF_TEAM			= 1<<24,
1366 	IFF_RXFH_CONFIGURED		= 1<<25,
1367 	IFF_PHONY_HEADROOM		= 1<<26,
1368 	IFF_MACSEC			= 1<<27,
1369 };
1370 
1371 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1372 #define IFF_EBRIDGE			IFF_EBRIDGE
1373 #define IFF_BONDING			IFF_BONDING
1374 #define IFF_ISATAP			IFF_ISATAP
1375 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1376 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1377 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1378 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1379 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1380 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1381 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1382 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1383 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1384 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1385 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1386 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1387 #define IFF_MACVLAN			IFF_MACVLAN
1388 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1389 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1390 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1391 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1392 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1393 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1394 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1395 #define IFF_TEAM			IFF_TEAM
1396 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1397 #define IFF_MACSEC			IFF_MACSEC
1398 
1399 /**
1400  *	struct net_device - The DEVICE structure.
1401  *		Actually, this whole structure is a big mistake.  It mixes I/O
1402  *		data with strictly "high-level" data, and it has to know about
1403  *		almost every data structure used in the INET module.
1404  *
1405  *	@name:	This is the first field of the "visible" part of this structure
1406  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1407  *	 	of the interface.
1408  *
1409  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1410  *	@ifalias:	SNMP alias
1411  *	@mem_end:	Shared memory end
1412  *	@mem_start:	Shared memory start
1413  *	@base_addr:	Device I/O address
1414  *	@irq:		Device IRQ number
1415  *
1416  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1417  *
1418  *	@state:		Generic network queuing layer state, see netdev_state_t
1419  *	@dev_list:	The global list of network devices
1420  *	@napi_list:	List entry used for polling NAPI devices
1421  *	@unreg_list:	List entry  when we are unregistering the
1422  *			device; see the function unregister_netdev
1423  *	@close_list:	List entry used when we are closing the device
1424  *	@ptype_all:     Device-specific packet handlers for all protocols
1425  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1426  *
1427  *	@adj_list:	Directly linked devices, like slaves for bonding
1428  *	@all_adj_list:	All linked devices, *including* neighbours
1429  *	@features:	Currently active device features
1430  *	@hw_features:	User-changeable features
1431  *
1432  *	@wanted_features:	User-requested features
1433  *	@vlan_features:		Mask of features inheritable by VLAN devices
1434  *
1435  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1436  *				This field indicates what encapsulation
1437  *				offloads the hardware is capable of doing,
1438  *				and drivers will need to set them appropriately.
1439  *
1440  *	@mpls_features:	Mask of features inheritable by MPLS
1441  *
1442  *	@ifindex:	interface index
1443  *	@group:		The group the device belongs to
1444  *
1445  *	@stats:		Statistics struct, which was left as a legacy, use
1446  *			rtnl_link_stats64 instead
1447  *
1448  *	@rx_dropped:	Dropped packets by core network,
1449  *			do not use this in drivers
1450  *	@tx_dropped:	Dropped packets by core network,
1451  *			do not use this in drivers
1452  *	@rx_nohandler:	nohandler dropped packets by core network on
1453  *			inactive devices, do not use this in drivers
1454  *
1455  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1456  *				instead of ioctl,
1457  *				see <net/iw_handler.h> for details.
1458  *	@wireless_data:	Instance data managed by the core of wireless extensions
1459  *
1460  *	@netdev_ops:	Includes several pointers to callbacks,
1461  *			if one wants to override the ndo_*() functions
1462  *	@ethtool_ops:	Management operations
1463  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1464  *			of Layer 2 headers.
1465  *
1466  *	@flags:		Interface flags (a la BSD)
1467  *	@priv_flags:	Like 'flags' but invisible to userspace,
1468  *			see if.h for the definitions
1469  *	@gflags:	Global flags ( kept as legacy )
1470  *	@padded:	How much padding added by alloc_netdev()
1471  *	@operstate:	RFC2863 operstate
1472  *	@link_mode:	Mapping policy to operstate
1473  *	@if_port:	Selectable AUI, TP, ...
1474  *	@dma:		DMA channel
1475  *	@mtu:		Interface MTU value
1476  *	@type:		Interface hardware type
1477  *	@hard_header_len: Maximum hardware header length.
1478  *
1479  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1480  *			  cases can this be guaranteed
1481  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1482  *			  cases can this be guaranteed. Some cases also use
1483  *			  LL_MAX_HEADER instead to allocate the skb
1484  *
1485  *	interface address info:
1486  *
1487  * 	@perm_addr:		Permanent hw address
1488  * 	@addr_assign_type:	Hw address assignment type
1489  * 	@addr_len:		Hardware address length
1490  * 	@neigh_priv_len;	Used in neigh_alloc(),
1491  * 				initialized only in atm/clip.c
1492  * 	@dev_id:		Used to differentiate devices that share
1493  * 				the same link layer address
1494  * 	@dev_port:		Used to differentiate devices that share
1495  * 				the same function
1496  *	@addr_list_lock:	XXX: need comments on this one
1497  *	@uc_promisc:		Counter that indicates promiscuous mode
1498  *				has been enabled due to the need to listen to
1499  *				additional unicast addresses in a device that
1500  *				does not implement ndo_set_rx_mode()
1501  *	@uc:			unicast mac addresses
1502  *	@mc:			multicast mac addresses
1503  *	@dev_addrs:		list of device hw addresses
1504  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1505  *	@promiscuity:		Number of times the NIC is told to work in
1506  *				promiscuous mode; if it becomes 0 the NIC will
1507  *				exit promiscuous mode
1508  *	@allmulti:		Counter, enables or disables allmulticast mode
1509  *
1510  *	@vlan_info:	VLAN info
1511  *	@dsa_ptr:	dsa specific data
1512  *	@tipc_ptr:	TIPC specific data
1513  *	@atalk_ptr:	AppleTalk link
1514  *	@ip_ptr:	IPv4 specific data
1515  *	@dn_ptr:	DECnet specific data
1516  *	@ip6_ptr:	IPv6 specific data
1517  *	@ax25_ptr:	AX.25 specific data
1518  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1519  *
1520  *	@last_rx:	Time of last Rx
1521  *	@dev_addr:	Hw address (before bcast,
1522  *			because most packets are unicast)
1523  *
1524  *	@_rx:			Array of RX queues
1525  *	@num_rx_queues:		Number of RX queues
1526  *				allocated at register_netdev() time
1527  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1528  *
1529  *	@rx_handler:		handler for received packets
1530  *	@rx_handler_data: 	XXX: need comments on this one
1531  *	@ingress_queue:		XXX: need comments on this one
1532  *	@broadcast:		hw bcast address
1533  *
1534  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1535  *			indexed by RX queue number. Assigned by driver.
1536  *			This must only be set if the ndo_rx_flow_steer
1537  *			operation is defined
1538  *	@index_hlist:		Device index hash chain
1539  *
1540  *	@_tx:			Array of TX queues
1541  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1542  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1543  *	@qdisc:			Root qdisc from userspace point of view
1544  *	@tx_queue_len:		Max frames per queue allowed
1545  *	@tx_global_lock: 	XXX: need comments on this one
1546  *
1547  *	@xps_maps:	XXX: need comments on this one
1548  *
1549  *	@offload_fwd_mark:	Offload device fwding mark
1550  *
1551  *	@trans_start:		Time (in jiffies) of last Tx
1552  *	@watchdog_timeo:	Represents the timeout that is used by
1553  *				the watchdog (see dev_watchdog())
1554  *	@watchdog_timer:	List of timers
1555  *
1556  *	@pcpu_refcnt:		Number of references to this device
1557  *	@todo_list:		Delayed register/unregister
1558  *	@link_watch_list:	XXX: need comments on this one
1559  *
1560  *	@reg_state:		Register/unregister state machine
1561  *	@dismantle:		Device is going to be freed
1562  *	@rtnl_link_state:	This enum represents the phases of creating
1563  *				a new link
1564  *
1565  *	@destructor:		Called from unregister,
1566  *				can be used to call free_netdev
1567  *	@npinfo:		XXX: need comments on this one
1568  * 	@nd_net:		Network namespace this network device is inside
1569  *
1570  * 	@ml_priv:	Mid-layer private
1571  * 	@lstats:	Loopback statistics
1572  * 	@tstats:	Tunnel statistics
1573  * 	@dstats:	Dummy statistics
1574  * 	@vstats:	Virtual ethernet statistics
1575  *
1576  *	@garp_port:	GARP
1577  *	@mrp_port:	MRP
1578  *
1579  *	@dev:		Class/net/name entry
1580  *	@sysfs_groups:	Space for optional device, statistics and wireless
1581  *			sysfs groups
1582  *
1583  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1584  *	@rtnl_link_ops:	Rtnl_link_ops
1585  *
1586  *	@gso_max_size:	Maximum size of generic segmentation offload
1587  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1588  *			NIC for GSO
1589  *	@gso_min_segs:	Minimum number of segments that can be passed to the
1590  *			NIC for GSO
1591  *
1592  *	@dcbnl_ops:	Data Center Bridging netlink ops
1593  *	@num_tc:	Number of traffic classes in the net device
1594  *	@tc_to_txq:	XXX: need comments on this one
1595  *	@prio_tc_map	XXX: need comments on this one
1596  *
1597  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1598  *
1599  *	@priomap:	XXX: need comments on this one
1600  *	@phydev:	Physical device may attach itself
1601  *			for hardware timestamping
1602  *
1603  *	@qdisc_tx_busylock:	XXX: need comments on this one
1604  *
1605  *	@proto_down:	protocol port state information can be sent to the
1606  *			switch driver and used to set the phys state of the
1607  *			switch port.
1608  *
1609  *	FIXME: cleanup struct net_device such that network protocol info
1610  *	moves out.
1611  */
1612 
1613 struct net_device {
1614 	char			name[IFNAMSIZ];
1615 	struct hlist_node	name_hlist;
1616 	char 			*ifalias;
1617 	/*
1618 	 *	I/O specific fields
1619 	 *	FIXME: Merge these and struct ifmap into one
1620 	 */
1621 	unsigned long		mem_end;
1622 	unsigned long		mem_start;
1623 	unsigned long		base_addr;
1624 	int			irq;
1625 
1626 	atomic_t		carrier_changes;
1627 
1628 	/*
1629 	 *	Some hardware also needs these fields (state,dev_list,
1630 	 *	napi_list,unreg_list,close_list) but they are not
1631 	 *	part of the usual set specified in Space.c.
1632 	 */
1633 
1634 	unsigned long		state;
1635 
1636 	struct list_head	dev_list;
1637 	struct list_head	napi_list;
1638 	struct list_head	unreg_list;
1639 	struct list_head	close_list;
1640 	struct list_head	ptype_all;
1641 	struct list_head	ptype_specific;
1642 
1643 	struct {
1644 		struct list_head upper;
1645 		struct list_head lower;
1646 	} adj_list;
1647 
1648 	struct {
1649 		struct list_head upper;
1650 		struct list_head lower;
1651 	} all_adj_list;
1652 
1653 	netdev_features_t	features;
1654 	netdev_features_t	hw_features;
1655 	netdev_features_t	wanted_features;
1656 	netdev_features_t	vlan_features;
1657 	netdev_features_t	hw_enc_features;
1658 	netdev_features_t	mpls_features;
1659 
1660 	int			ifindex;
1661 	int			group;
1662 
1663 	struct net_device_stats	stats;
1664 
1665 	atomic_long_t		rx_dropped;
1666 	atomic_long_t		tx_dropped;
1667 	atomic_long_t		rx_nohandler;
1668 
1669 #ifdef CONFIG_WIRELESS_EXT
1670 	const struct iw_handler_def *wireless_handlers;
1671 	struct iw_public_data	*wireless_data;
1672 #endif
1673 	const struct net_device_ops *netdev_ops;
1674 	const struct ethtool_ops *ethtool_ops;
1675 #ifdef CONFIG_NET_SWITCHDEV
1676 	const struct switchdev_ops *switchdev_ops;
1677 #endif
1678 #ifdef CONFIG_NET_L3_MASTER_DEV
1679 	const struct l3mdev_ops	*l3mdev_ops;
1680 #endif
1681 
1682 	const struct header_ops *header_ops;
1683 
1684 	unsigned int		flags;
1685 	unsigned int		priv_flags;
1686 
1687 	unsigned short		gflags;
1688 	unsigned short		padded;
1689 
1690 	unsigned char		operstate;
1691 	unsigned char		link_mode;
1692 
1693 	unsigned char		if_port;
1694 	unsigned char		dma;
1695 
1696 	unsigned int		mtu;
1697 	unsigned short		type;
1698 	unsigned short		hard_header_len;
1699 
1700 	unsigned short		needed_headroom;
1701 	unsigned short		needed_tailroom;
1702 
1703 	/* Interface address info. */
1704 	unsigned char		perm_addr[MAX_ADDR_LEN];
1705 	unsigned char		addr_assign_type;
1706 	unsigned char		addr_len;
1707 	unsigned short		neigh_priv_len;
1708 	unsigned short          dev_id;
1709 	unsigned short          dev_port;
1710 	spinlock_t		addr_list_lock;
1711 	unsigned char		name_assign_type;
1712 	bool			uc_promisc;
1713 	struct netdev_hw_addr_list	uc;
1714 	struct netdev_hw_addr_list	mc;
1715 	struct netdev_hw_addr_list	dev_addrs;
1716 
1717 #ifdef CONFIG_SYSFS
1718 	struct kset		*queues_kset;
1719 #endif
1720 	unsigned int		promiscuity;
1721 	unsigned int		allmulti;
1722 
1723 
1724 	/* Protocol-specific pointers */
1725 
1726 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1727 	struct vlan_info __rcu	*vlan_info;
1728 #endif
1729 #if IS_ENABLED(CONFIG_NET_DSA)
1730 	struct dsa_switch_tree	*dsa_ptr;
1731 #endif
1732 #if IS_ENABLED(CONFIG_TIPC)
1733 	struct tipc_bearer __rcu *tipc_ptr;
1734 #endif
1735 	void 			*atalk_ptr;
1736 	struct in_device __rcu	*ip_ptr;
1737 	struct dn_dev __rcu     *dn_ptr;
1738 	struct inet6_dev __rcu	*ip6_ptr;
1739 	void			*ax25_ptr;
1740 	struct wireless_dev	*ieee80211_ptr;
1741 	struct wpan_dev		*ieee802154_ptr;
1742 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1743 	struct mpls_dev __rcu	*mpls_ptr;
1744 #endif
1745 
1746 /*
1747  * Cache lines mostly used on receive path (including eth_type_trans())
1748  */
1749 	unsigned long		last_rx;
1750 
1751 	/* Interface address info used in eth_type_trans() */
1752 	unsigned char		*dev_addr;
1753 
1754 #ifdef CONFIG_SYSFS
1755 	struct netdev_rx_queue	*_rx;
1756 
1757 	unsigned int		num_rx_queues;
1758 	unsigned int		real_num_rx_queues;
1759 #endif
1760 
1761 	unsigned long		gro_flush_timeout;
1762 	rx_handler_func_t __rcu	*rx_handler;
1763 	void __rcu		*rx_handler_data;
1764 
1765 #ifdef CONFIG_NET_CLS_ACT
1766 	struct tcf_proto __rcu  *ingress_cl_list;
1767 #endif
1768 	struct netdev_queue __rcu *ingress_queue;
1769 #ifdef CONFIG_NETFILTER_INGRESS
1770 	struct list_head	nf_hooks_ingress;
1771 #endif
1772 
1773 	unsigned char		broadcast[MAX_ADDR_LEN];
1774 #ifdef CONFIG_RFS_ACCEL
1775 	struct cpu_rmap		*rx_cpu_rmap;
1776 #endif
1777 	struct hlist_node	index_hlist;
1778 
1779 /*
1780  * Cache lines mostly used on transmit path
1781  */
1782 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1783 	unsigned int		num_tx_queues;
1784 	unsigned int		real_num_tx_queues;
1785 	struct Qdisc		*qdisc;
1786 	unsigned long		tx_queue_len;
1787 	spinlock_t		tx_global_lock;
1788 	int			watchdog_timeo;
1789 
1790 #ifdef CONFIG_XPS
1791 	struct xps_dev_maps __rcu *xps_maps;
1792 #endif
1793 #ifdef CONFIG_NET_CLS_ACT
1794 	struct tcf_proto __rcu  *egress_cl_list;
1795 #endif
1796 #ifdef CONFIG_NET_SWITCHDEV
1797 	u32			offload_fwd_mark;
1798 #endif
1799 
1800 	/* These may be needed for future network-power-down code. */
1801 
1802 	/*
1803 	 * trans_start here is expensive for high speed devices on SMP,
1804 	 * please use netdev_queue->trans_start instead.
1805 	 */
1806 	unsigned long		trans_start;
1807 
1808 	struct timer_list	watchdog_timer;
1809 
1810 	int __percpu		*pcpu_refcnt;
1811 	struct list_head	todo_list;
1812 
1813 	struct list_head	link_watch_list;
1814 
1815 	enum { NETREG_UNINITIALIZED=0,
1816 	       NETREG_REGISTERED,	/* completed register_netdevice */
1817 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1818 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1819 	       NETREG_RELEASED,		/* called free_netdev */
1820 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1821 	} reg_state:8;
1822 
1823 	bool dismantle;
1824 
1825 	enum {
1826 		RTNL_LINK_INITIALIZED,
1827 		RTNL_LINK_INITIALIZING,
1828 	} rtnl_link_state:16;
1829 
1830 	void (*destructor)(struct net_device *dev);
1831 
1832 #ifdef CONFIG_NETPOLL
1833 	struct netpoll_info __rcu	*npinfo;
1834 #endif
1835 
1836 	possible_net_t			nd_net;
1837 
1838 	/* mid-layer private */
1839 	union {
1840 		void					*ml_priv;
1841 		struct pcpu_lstats __percpu		*lstats;
1842 		struct pcpu_sw_netstats __percpu	*tstats;
1843 		struct pcpu_dstats __percpu		*dstats;
1844 		struct pcpu_vstats __percpu		*vstats;
1845 	};
1846 
1847 	struct garp_port __rcu	*garp_port;
1848 	struct mrp_port __rcu	*mrp_port;
1849 
1850 	struct device		dev;
1851 	const struct attribute_group *sysfs_groups[4];
1852 	const struct attribute_group *sysfs_rx_queue_group;
1853 
1854 	const struct rtnl_link_ops *rtnl_link_ops;
1855 
1856 	/* for setting kernel sock attribute on TCP connection setup */
1857 #define GSO_MAX_SIZE		65536
1858 	unsigned int		gso_max_size;
1859 #define GSO_MAX_SEGS		65535
1860 	u16			gso_max_segs;
1861 	u16			gso_min_segs;
1862 #ifdef CONFIG_DCB
1863 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1864 #endif
1865 	u8			num_tc;
1866 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1867 	u8			prio_tc_map[TC_BITMASK + 1];
1868 
1869 #if IS_ENABLED(CONFIG_FCOE)
1870 	unsigned int		fcoe_ddp_xid;
1871 #endif
1872 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1873 	struct netprio_map __rcu *priomap;
1874 #endif
1875 	struct phy_device	*phydev;
1876 	struct lock_class_key	*qdisc_tx_busylock;
1877 	bool			proto_down;
1878 };
1879 #define to_net_dev(d) container_of(d, struct net_device, dev)
1880 
1881 #define	NETDEV_ALIGN		32
1882 
1883 static inline
1884 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1885 {
1886 	return dev->prio_tc_map[prio & TC_BITMASK];
1887 }
1888 
1889 static inline
1890 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1891 {
1892 	if (tc >= dev->num_tc)
1893 		return -EINVAL;
1894 
1895 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1896 	return 0;
1897 }
1898 
1899 static inline
1900 void netdev_reset_tc(struct net_device *dev)
1901 {
1902 	dev->num_tc = 0;
1903 	memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1904 	memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1905 }
1906 
1907 static inline
1908 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1909 {
1910 	if (tc >= dev->num_tc)
1911 		return -EINVAL;
1912 
1913 	dev->tc_to_txq[tc].count = count;
1914 	dev->tc_to_txq[tc].offset = offset;
1915 	return 0;
1916 }
1917 
1918 static inline
1919 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1920 {
1921 	if (num_tc > TC_MAX_QUEUE)
1922 		return -EINVAL;
1923 
1924 	dev->num_tc = num_tc;
1925 	return 0;
1926 }
1927 
1928 static inline
1929 int netdev_get_num_tc(struct net_device *dev)
1930 {
1931 	return dev->num_tc;
1932 }
1933 
1934 static inline
1935 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1936 					 unsigned int index)
1937 {
1938 	return &dev->_tx[index];
1939 }
1940 
1941 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1942 						    const struct sk_buff *skb)
1943 {
1944 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1945 }
1946 
1947 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1948 					    void (*f)(struct net_device *,
1949 						      struct netdev_queue *,
1950 						      void *),
1951 					    void *arg)
1952 {
1953 	unsigned int i;
1954 
1955 	for (i = 0; i < dev->num_tx_queues; i++)
1956 		f(dev, &dev->_tx[i], arg);
1957 }
1958 
1959 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1960 				    struct sk_buff *skb,
1961 				    void *accel_priv);
1962 
1963 /* returns the headroom that the master device needs to take in account
1964  * when forwarding to this dev
1965  */
1966 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1967 {
1968 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1969 }
1970 
1971 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1972 {
1973 	if (dev->netdev_ops->ndo_set_rx_headroom)
1974 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1975 }
1976 
1977 /* set the device rx headroom to the dev's default */
1978 static inline void netdev_reset_rx_headroom(struct net_device *dev)
1979 {
1980 	netdev_set_rx_headroom(dev, -1);
1981 }
1982 
1983 /*
1984  * Net namespace inlines
1985  */
1986 static inline
1987 struct net *dev_net(const struct net_device *dev)
1988 {
1989 	return read_pnet(&dev->nd_net);
1990 }
1991 
1992 static inline
1993 void dev_net_set(struct net_device *dev, struct net *net)
1994 {
1995 	write_pnet(&dev->nd_net, net);
1996 }
1997 
1998 static inline bool netdev_uses_dsa(struct net_device *dev)
1999 {
2000 #if IS_ENABLED(CONFIG_NET_DSA)
2001 	if (dev->dsa_ptr != NULL)
2002 		return dsa_uses_tagged_protocol(dev->dsa_ptr);
2003 #endif
2004 	return false;
2005 }
2006 
2007 /**
2008  *	netdev_priv - access network device private data
2009  *	@dev: network device
2010  *
2011  * Get network device private data
2012  */
2013 static inline void *netdev_priv(const struct net_device *dev)
2014 {
2015 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2016 }
2017 
2018 /* Set the sysfs physical device reference for the network logical device
2019  * if set prior to registration will cause a symlink during initialization.
2020  */
2021 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2022 
2023 /* Set the sysfs device type for the network logical device to allow
2024  * fine-grained identification of different network device types. For
2025  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2026  */
2027 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2028 
2029 /* Default NAPI poll() weight
2030  * Device drivers are strongly advised to not use bigger value
2031  */
2032 #define NAPI_POLL_WEIGHT 64
2033 
2034 /**
2035  *	netif_napi_add - initialize a NAPI context
2036  *	@dev:  network device
2037  *	@napi: NAPI context
2038  *	@poll: polling function
2039  *	@weight: default weight
2040  *
2041  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2042  * *any* of the other NAPI-related functions.
2043  */
2044 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2045 		    int (*poll)(struct napi_struct *, int), int weight);
2046 
2047 /**
2048  *	netif_tx_napi_add - initialize a NAPI context
2049  *	@dev:  network device
2050  *	@napi: NAPI context
2051  *	@poll: polling function
2052  *	@weight: default weight
2053  *
2054  * This variant of netif_napi_add() should be used from drivers using NAPI
2055  * to exclusively poll a TX queue.
2056  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2057  */
2058 static inline void netif_tx_napi_add(struct net_device *dev,
2059 				     struct napi_struct *napi,
2060 				     int (*poll)(struct napi_struct *, int),
2061 				     int weight)
2062 {
2063 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2064 	netif_napi_add(dev, napi, poll, weight);
2065 }
2066 
2067 /**
2068  *  netif_napi_del - remove a NAPI context
2069  *  @napi: NAPI context
2070  *
2071  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2072  */
2073 void netif_napi_del(struct napi_struct *napi);
2074 
2075 struct napi_gro_cb {
2076 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2077 	void	*frag0;
2078 
2079 	/* Length of frag0. */
2080 	unsigned int frag0_len;
2081 
2082 	/* This indicates where we are processing relative to skb->data. */
2083 	int	data_offset;
2084 
2085 	/* This is non-zero if the packet cannot be merged with the new skb. */
2086 	u16	flush;
2087 
2088 	/* Save the IP ID here and check when we get to the transport layer */
2089 	u16	flush_id;
2090 
2091 	/* Number of segments aggregated. */
2092 	u16	count;
2093 
2094 	/* Start offset for remote checksum offload */
2095 	u16	gro_remcsum_start;
2096 
2097 	/* jiffies when first packet was created/queued */
2098 	unsigned long age;
2099 
2100 	/* Used in ipv6_gro_receive() and foo-over-udp */
2101 	u16	proto;
2102 
2103 	/* This is non-zero if the packet may be of the same flow. */
2104 	u8	same_flow:1;
2105 
2106 	/* Used in tunnel GRO receive */
2107 	u8	encap_mark:1;
2108 
2109 	/* GRO checksum is valid */
2110 	u8	csum_valid:1;
2111 
2112 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2113 	u8	csum_cnt:3;
2114 
2115 	/* Free the skb? */
2116 	u8	free:2;
2117 #define NAPI_GRO_FREE		  1
2118 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2119 
2120 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2121 	u8	is_ipv6:1;
2122 
2123 	/* 7 bit hole */
2124 
2125 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2126 	__wsum	csum;
2127 
2128 	/* used in skb_gro_receive() slow path */
2129 	struct sk_buff *last;
2130 };
2131 
2132 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2133 
2134 struct packet_type {
2135 	__be16			type;	/* This is really htons(ether_type). */
2136 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2137 	int			(*func) (struct sk_buff *,
2138 					 struct net_device *,
2139 					 struct packet_type *,
2140 					 struct net_device *);
2141 	bool			(*id_match)(struct packet_type *ptype,
2142 					    struct sock *sk);
2143 	void			*af_packet_priv;
2144 	struct list_head	list;
2145 };
2146 
2147 struct offload_callbacks {
2148 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2149 						netdev_features_t features);
2150 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2151 						 struct sk_buff *skb);
2152 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2153 };
2154 
2155 struct packet_offload {
2156 	__be16			 type;	/* This is really htons(ether_type). */
2157 	u16			 priority;
2158 	struct offload_callbacks callbacks;
2159 	struct list_head	 list;
2160 };
2161 
2162 struct udp_offload;
2163 
2164 struct udp_offload_callbacks {
2165 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2166 						 struct sk_buff *skb,
2167 						 struct udp_offload *uoff);
2168 	int			(*gro_complete)(struct sk_buff *skb,
2169 						int nhoff,
2170 						struct udp_offload *uoff);
2171 };
2172 
2173 struct udp_offload {
2174 	__be16			 port;
2175 	u8			 ipproto;
2176 	struct udp_offload_callbacks callbacks;
2177 };
2178 
2179 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2180 struct pcpu_sw_netstats {
2181 	u64     rx_packets;
2182 	u64     rx_bytes;
2183 	u64     tx_packets;
2184 	u64     tx_bytes;
2185 	struct u64_stats_sync   syncp;
2186 };
2187 
2188 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2189 ({									\
2190 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2191 	if (pcpu_stats)	{						\
2192 		int __cpu;						\
2193 		for_each_possible_cpu(__cpu) {				\
2194 			typeof(type) *stat;				\
2195 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2196 			u64_stats_init(&stat->syncp);			\
2197 		}							\
2198 	}								\
2199 	pcpu_stats;							\
2200 })
2201 
2202 #define netdev_alloc_pcpu_stats(type)					\
2203 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2204 
2205 enum netdev_lag_tx_type {
2206 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2207 	NETDEV_LAG_TX_TYPE_RANDOM,
2208 	NETDEV_LAG_TX_TYPE_BROADCAST,
2209 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2210 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2211 	NETDEV_LAG_TX_TYPE_HASH,
2212 };
2213 
2214 struct netdev_lag_upper_info {
2215 	enum netdev_lag_tx_type tx_type;
2216 };
2217 
2218 struct netdev_lag_lower_state_info {
2219 	u8 link_up : 1,
2220 	   tx_enabled : 1;
2221 };
2222 
2223 #include <linux/notifier.h>
2224 
2225 /* netdevice notifier chain. Please remember to update the rtnetlink
2226  * notification exclusion list in rtnetlink_event() when adding new
2227  * types.
2228  */
2229 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2230 #define NETDEV_DOWN	0x0002
2231 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2232 				   detected a hardware crash and restarted
2233 				   - we can use this eg to kick tcp sessions
2234 				   once done */
2235 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2236 #define NETDEV_REGISTER 0x0005
2237 #define NETDEV_UNREGISTER	0x0006
2238 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2239 #define NETDEV_CHANGEADDR	0x0008
2240 #define NETDEV_GOING_DOWN	0x0009
2241 #define NETDEV_CHANGENAME	0x000A
2242 #define NETDEV_FEAT_CHANGE	0x000B
2243 #define NETDEV_BONDING_FAILOVER 0x000C
2244 #define NETDEV_PRE_UP		0x000D
2245 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2246 #define NETDEV_POST_TYPE_CHANGE	0x000F
2247 #define NETDEV_POST_INIT	0x0010
2248 #define NETDEV_UNREGISTER_FINAL 0x0011
2249 #define NETDEV_RELEASE		0x0012
2250 #define NETDEV_NOTIFY_PEERS	0x0013
2251 #define NETDEV_JOIN		0x0014
2252 #define NETDEV_CHANGEUPPER	0x0015
2253 #define NETDEV_RESEND_IGMP	0x0016
2254 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2255 #define NETDEV_CHANGEINFODATA	0x0018
2256 #define NETDEV_BONDING_INFO	0x0019
2257 #define NETDEV_PRECHANGEUPPER	0x001A
2258 #define NETDEV_CHANGELOWERSTATE	0x001B
2259 
2260 int register_netdevice_notifier(struct notifier_block *nb);
2261 int unregister_netdevice_notifier(struct notifier_block *nb);
2262 
2263 struct netdev_notifier_info {
2264 	struct net_device *dev;
2265 };
2266 
2267 struct netdev_notifier_change_info {
2268 	struct netdev_notifier_info info; /* must be first */
2269 	unsigned int flags_changed;
2270 };
2271 
2272 struct netdev_notifier_changeupper_info {
2273 	struct netdev_notifier_info info; /* must be first */
2274 	struct net_device *upper_dev; /* new upper dev */
2275 	bool master; /* is upper dev master */
2276 	bool linking; /* is the notification for link or unlink */
2277 	void *upper_info; /* upper dev info */
2278 };
2279 
2280 struct netdev_notifier_changelowerstate_info {
2281 	struct netdev_notifier_info info; /* must be first */
2282 	void *lower_state_info; /* is lower dev state */
2283 };
2284 
2285 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2286 					     struct net_device *dev)
2287 {
2288 	info->dev = dev;
2289 }
2290 
2291 static inline struct net_device *
2292 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2293 {
2294 	return info->dev;
2295 }
2296 
2297 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2298 
2299 
2300 extern rwlock_t				dev_base_lock;		/* Device list lock */
2301 
2302 #define for_each_netdev(net, d)		\
2303 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2304 #define for_each_netdev_reverse(net, d)	\
2305 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2306 #define for_each_netdev_rcu(net, d)		\
2307 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2308 #define for_each_netdev_safe(net, d, n)	\
2309 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2310 #define for_each_netdev_continue(net, d)		\
2311 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2312 #define for_each_netdev_continue_rcu(net, d)		\
2313 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2314 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2315 		for_each_netdev_rcu(&init_net, slave)	\
2316 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2317 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2318 
2319 static inline struct net_device *next_net_device(struct net_device *dev)
2320 {
2321 	struct list_head *lh;
2322 	struct net *net;
2323 
2324 	net = dev_net(dev);
2325 	lh = dev->dev_list.next;
2326 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2327 }
2328 
2329 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2330 {
2331 	struct list_head *lh;
2332 	struct net *net;
2333 
2334 	net = dev_net(dev);
2335 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2336 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2337 }
2338 
2339 static inline struct net_device *first_net_device(struct net *net)
2340 {
2341 	return list_empty(&net->dev_base_head) ? NULL :
2342 		net_device_entry(net->dev_base_head.next);
2343 }
2344 
2345 static inline struct net_device *first_net_device_rcu(struct net *net)
2346 {
2347 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2348 
2349 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2350 }
2351 
2352 int netdev_boot_setup_check(struct net_device *dev);
2353 unsigned long netdev_boot_base(const char *prefix, int unit);
2354 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2355 				       const char *hwaddr);
2356 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2357 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2358 void dev_add_pack(struct packet_type *pt);
2359 void dev_remove_pack(struct packet_type *pt);
2360 void __dev_remove_pack(struct packet_type *pt);
2361 void dev_add_offload(struct packet_offload *po);
2362 void dev_remove_offload(struct packet_offload *po);
2363 
2364 int dev_get_iflink(const struct net_device *dev);
2365 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2366 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2367 				      unsigned short mask);
2368 struct net_device *dev_get_by_name(struct net *net, const char *name);
2369 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2370 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2371 int dev_alloc_name(struct net_device *dev, const char *name);
2372 int dev_open(struct net_device *dev);
2373 int dev_close(struct net_device *dev);
2374 int dev_close_many(struct list_head *head, bool unlink);
2375 void dev_disable_lro(struct net_device *dev);
2376 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2377 int dev_queue_xmit(struct sk_buff *skb);
2378 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2379 int register_netdevice(struct net_device *dev);
2380 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2381 void unregister_netdevice_many(struct list_head *head);
2382 static inline void unregister_netdevice(struct net_device *dev)
2383 {
2384 	unregister_netdevice_queue(dev, NULL);
2385 }
2386 
2387 int netdev_refcnt_read(const struct net_device *dev);
2388 void free_netdev(struct net_device *dev);
2389 void netdev_freemem(struct net_device *dev);
2390 void synchronize_net(void);
2391 int init_dummy_netdev(struct net_device *dev);
2392 
2393 DECLARE_PER_CPU(int, xmit_recursion);
2394 static inline int dev_recursion_level(void)
2395 {
2396 	return this_cpu_read(xmit_recursion);
2397 }
2398 
2399 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2400 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2401 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2402 int netdev_get_name(struct net *net, char *name, int ifindex);
2403 int dev_restart(struct net_device *dev);
2404 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2405 
2406 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2407 {
2408 	return NAPI_GRO_CB(skb)->data_offset;
2409 }
2410 
2411 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2412 {
2413 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2414 }
2415 
2416 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2417 {
2418 	NAPI_GRO_CB(skb)->data_offset += len;
2419 }
2420 
2421 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2422 					unsigned int offset)
2423 {
2424 	return NAPI_GRO_CB(skb)->frag0 + offset;
2425 }
2426 
2427 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2428 {
2429 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2430 }
2431 
2432 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2433 					unsigned int offset)
2434 {
2435 	if (!pskb_may_pull(skb, hlen))
2436 		return NULL;
2437 
2438 	NAPI_GRO_CB(skb)->frag0 = NULL;
2439 	NAPI_GRO_CB(skb)->frag0_len = 0;
2440 	return skb->data + offset;
2441 }
2442 
2443 static inline void *skb_gro_network_header(struct sk_buff *skb)
2444 {
2445 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2446 	       skb_network_offset(skb);
2447 }
2448 
2449 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2450 					const void *start, unsigned int len)
2451 {
2452 	if (NAPI_GRO_CB(skb)->csum_valid)
2453 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2454 						  csum_partial(start, len, 0));
2455 }
2456 
2457 /* GRO checksum functions. These are logical equivalents of the normal
2458  * checksum functions (in skbuff.h) except that they operate on the GRO
2459  * offsets and fields in sk_buff.
2460  */
2461 
2462 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2463 
2464 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2465 {
2466 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2467 }
2468 
2469 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2470 						      bool zero_okay,
2471 						      __sum16 check)
2472 {
2473 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2474 		skb_checksum_start_offset(skb) <
2475 		 skb_gro_offset(skb)) &&
2476 		!skb_at_gro_remcsum_start(skb) &&
2477 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2478 		(!zero_okay || check));
2479 }
2480 
2481 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2482 							   __wsum psum)
2483 {
2484 	if (NAPI_GRO_CB(skb)->csum_valid &&
2485 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2486 		return 0;
2487 
2488 	NAPI_GRO_CB(skb)->csum = psum;
2489 
2490 	return __skb_gro_checksum_complete(skb);
2491 }
2492 
2493 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2494 {
2495 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2496 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2497 		NAPI_GRO_CB(skb)->csum_cnt--;
2498 	} else {
2499 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2500 		 * verified a new top level checksum or an encapsulated one
2501 		 * during GRO. This saves work if we fallback to normal path.
2502 		 */
2503 		__skb_incr_checksum_unnecessary(skb);
2504 	}
2505 }
2506 
2507 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2508 				    compute_pseudo)			\
2509 ({									\
2510 	__sum16 __ret = 0;						\
2511 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2512 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2513 				compute_pseudo(skb, proto));		\
2514 	if (__ret)							\
2515 		__skb_mark_checksum_bad(skb);				\
2516 	else								\
2517 		skb_gro_incr_csum_unnecessary(skb);			\
2518 	__ret;								\
2519 })
2520 
2521 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2522 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2523 
2524 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2525 					     compute_pseudo)		\
2526 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2527 
2528 #define skb_gro_checksum_simple_validate(skb)				\
2529 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2530 
2531 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2532 {
2533 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2534 		!NAPI_GRO_CB(skb)->csum_valid);
2535 }
2536 
2537 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2538 					      __sum16 check, __wsum pseudo)
2539 {
2540 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2541 	NAPI_GRO_CB(skb)->csum_valid = 1;
2542 }
2543 
2544 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2545 do {									\
2546 	if (__skb_gro_checksum_convert_check(skb))			\
2547 		__skb_gro_checksum_convert(skb, check,			\
2548 					   compute_pseudo(skb, proto));	\
2549 } while (0)
2550 
2551 struct gro_remcsum {
2552 	int offset;
2553 	__wsum delta;
2554 };
2555 
2556 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2557 {
2558 	grc->offset = 0;
2559 	grc->delta = 0;
2560 }
2561 
2562 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2563 					    unsigned int off, size_t hdrlen,
2564 					    int start, int offset,
2565 					    struct gro_remcsum *grc,
2566 					    bool nopartial)
2567 {
2568 	__wsum delta;
2569 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2570 
2571 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2572 
2573 	if (!nopartial) {
2574 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2575 		return ptr;
2576 	}
2577 
2578 	ptr = skb_gro_header_fast(skb, off);
2579 	if (skb_gro_header_hard(skb, off + plen)) {
2580 		ptr = skb_gro_header_slow(skb, off + plen, off);
2581 		if (!ptr)
2582 			return NULL;
2583 	}
2584 
2585 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2586 			       start, offset);
2587 
2588 	/* Adjust skb->csum since we changed the packet */
2589 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2590 
2591 	grc->offset = off + hdrlen + offset;
2592 	grc->delta = delta;
2593 
2594 	return ptr;
2595 }
2596 
2597 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2598 					   struct gro_remcsum *grc)
2599 {
2600 	void *ptr;
2601 	size_t plen = grc->offset + sizeof(u16);
2602 
2603 	if (!grc->delta)
2604 		return;
2605 
2606 	ptr = skb_gro_header_fast(skb, grc->offset);
2607 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2608 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2609 		if (!ptr)
2610 			return;
2611 	}
2612 
2613 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2614 }
2615 
2616 struct skb_csum_offl_spec {
2617 	__u16		ipv4_okay:1,
2618 			ipv6_okay:1,
2619 			encap_okay:1,
2620 			ip_options_okay:1,
2621 			ext_hdrs_okay:1,
2622 			tcp_okay:1,
2623 			udp_okay:1,
2624 			sctp_okay:1,
2625 			vlan_okay:1,
2626 			no_encapped_ipv6:1,
2627 			no_not_encapped:1;
2628 };
2629 
2630 bool __skb_csum_offload_chk(struct sk_buff *skb,
2631 			    const struct skb_csum_offl_spec *spec,
2632 			    bool *csum_encapped,
2633 			    bool csum_help);
2634 
2635 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2636 					const struct skb_csum_offl_spec *spec,
2637 					bool *csum_encapped,
2638 					bool csum_help)
2639 {
2640 	if (skb->ip_summed != CHECKSUM_PARTIAL)
2641 		return false;
2642 
2643 	return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2644 }
2645 
2646 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2647 					     const struct skb_csum_offl_spec *spec)
2648 {
2649 	bool csum_encapped;
2650 
2651 	return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2652 }
2653 
2654 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2655 {
2656 	static const struct skb_csum_offl_spec csum_offl_spec = {
2657 		.ipv4_okay = 1,
2658 		.ip_options_okay = 1,
2659 		.ipv6_okay = 1,
2660 		.vlan_okay = 1,
2661 		.tcp_okay = 1,
2662 		.udp_okay = 1,
2663 	};
2664 
2665 	return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2666 }
2667 
2668 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2669 {
2670 	static const struct skb_csum_offl_spec csum_offl_spec = {
2671 		.ipv4_okay = 1,
2672 		.ip_options_okay = 1,
2673 		.tcp_okay = 1,
2674 		.udp_okay = 1,
2675 		.vlan_okay = 1,
2676 	};
2677 
2678 	return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2679 }
2680 
2681 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2682 				  unsigned short type,
2683 				  const void *daddr, const void *saddr,
2684 				  unsigned int len)
2685 {
2686 	if (!dev->header_ops || !dev->header_ops->create)
2687 		return 0;
2688 
2689 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2690 }
2691 
2692 static inline int dev_parse_header(const struct sk_buff *skb,
2693 				   unsigned char *haddr)
2694 {
2695 	const struct net_device *dev = skb->dev;
2696 
2697 	if (!dev->header_ops || !dev->header_ops->parse)
2698 		return 0;
2699 	return dev->header_ops->parse(skb, haddr);
2700 }
2701 
2702 /* ll_header must have at least hard_header_len allocated */
2703 static inline bool dev_validate_header(const struct net_device *dev,
2704 				       char *ll_header, int len)
2705 {
2706 	if (likely(len >= dev->hard_header_len))
2707 		return true;
2708 
2709 	if (capable(CAP_SYS_RAWIO)) {
2710 		memset(ll_header + len, 0, dev->hard_header_len - len);
2711 		return true;
2712 	}
2713 
2714 	if (dev->header_ops && dev->header_ops->validate)
2715 		return dev->header_ops->validate(ll_header, len);
2716 
2717 	return false;
2718 }
2719 
2720 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2721 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2722 static inline int unregister_gifconf(unsigned int family)
2723 {
2724 	return register_gifconf(family, NULL);
2725 }
2726 
2727 #ifdef CONFIG_NET_FLOW_LIMIT
2728 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2729 struct sd_flow_limit {
2730 	u64			count;
2731 	unsigned int		num_buckets;
2732 	unsigned int		history_head;
2733 	u16			history[FLOW_LIMIT_HISTORY];
2734 	u8			buckets[];
2735 };
2736 
2737 extern int netdev_flow_limit_table_len;
2738 #endif /* CONFIG_NET_FLOW_LIMIT */
2739 
2740 /*
2741  * Incoming packets are placed on per-CPU queues
2742  */
2743 struct softnet_data {
2744 	struct list_head	poll_list;
2745 	struct sk_buff_head	process_queue;
2746 
2747 	/* stats */
2748 	unsigned int		processed;
2749 	unsigned int		time_squeeze;
2750 	unsigned int		cpu_collision;
2751 	unsigned int		received_rps;
2752 #ifdef CONFIG_RPS
2753 	struct softnet_data	*rps_ipi_list;
2754 #endif
2755 #ifdef CONFIG_NET_FLOW_LIMIT
2756 	struct sd_flow_limit __rcu *flow_limit;
2757 #endif
2758 	struct Qdisc		*output_queue;
2759 	struct Qdisc		**output_queue_tailp;
2760 	struct sk_buff		*completion_queue;
2761 
2762 #ifdef CONFIG_RPS
2763 	/* Elements below can be accessed between CPUs for RPS */
2764 	struct call_single_data	csd ____cacheline_aligned_in_smp;
2765 	struct softnet_data	*rps_ipi_next;
2766 	unsigned int		cpu;
2767 	unsigned int		input_queue_head;
2768 	unsigned int		input_queue_tail;
2769 #endif
2770 	unsigned int		dropped;
2771 	struct sk_buff_head	input_pkt_queue;
2772 	struct napi_struct	backlog;
2773 
2774 };
2775 
2776 static inline void input_queue_head_incr(struct softnet_data *sd)
2777 {
2778 #ifdef CONFIG_RPS
2779 	sd->input_queue_head++;
2780 #endif
2781 }
2782 
2783 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2784 					      unsigned int *qtail)
2785 {
2786 #ifdef CONFIG_RPS
2787 	*qtail = ++sd->input_queue_tail;
2788 #endif
2789 }
2790 
2791 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2792 
2793 void __netif_schedule(struct Qdisc *q);
2794 void netif_schedule_queue(struct netdev_queue *txq);
2795 
2796 static inline void netif_tx_schedule_all(struct net_device *dev)
2797 {
2798 	unsigned int i;
2799 
2800 	for (i = 0; i < dev->num_tx_queues; i++)
2801 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2802 }
2803 
2804 static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2805 {
2806 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2807 }
2808 
2809 /**
2810  *	netif_start_queue - allow transmit
2811  *	@dev: network device
2812  *
2813  *	Allow upper layers to call the device hard_start_xmit routine.
2814  */
2815 static inline void netif_start_queue(struct net_device *dev)
2816 {
2817 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2818 }
2819 
2820 static inline void netif_tx_start_all_queues(struct net_device *dev)
2821 {
2822 	unsigned int i;
2823 
2824 	for (i = 0; i < dev->num_tx_queues; i++) {
2825 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2826 		netif_tx_start_queue(txq);
2827 	}
2828 }
2829 
2830 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2831 
2832 /**
2833  *	netif_wake_queue - restart transmit
2834  *	@dev: network device
2835  *
2836  *	Allow upper layers to call the device hard_start_xmit routine.
2837  *	Used for flow control when transmit resources are available.
2838  */
2839 static inline void netif_wake_queue(struct net_device *dev)
2840 {
2841 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2842 }
2843 
2844 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2845 {
2846 	unsigned int i;
2847 
2848 	for (i = 0; i < dev->num_tx_queues; i++) {
2849 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2850 		netif_tx_wake_queue(txq);
2851 	}
2852 }
2853 
2854 static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2855 {
2856 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2857 }
2858 
2859 /**
2860  *	netif_stop_queue - stop transmitted packets
2861  *	@dev: network device
2862  *
2863  *	Stop upper layers calling the device hard_start_xmit routine.
2864  *	Used for flow control when transmit resources are unavailable.
2865  */
2866 static inline void netif_stop_queue(struct net_device *dev)
2867 {
2868 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2869 }
2870 
2871 void netif_tx_stop_all_queues(struct net_device *dev);
2872 
2873 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2874 {
2875 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2876 }
2877 
2878 /**
2879  *	netif_queue_stopped - test if transmit queue is flowblocked
2880  *	@dev: network device
2881  *
2882  *	Test if transmit queue on device is currently unable to send.
2883  */
2884 static inline bool netif_queue_stopped(const struct net_device *dev)
2885 {
2886 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2887 }
2888 
2889 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2890 {
2891 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2892 }
2893 
2894 static inline bool
2895 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2896 {
2897 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2898 }
2899 
2900 static inline bool
2901 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2902 {
2903 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2904 }
2905 
2906 /**
2907  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2908  *	@dev_queue: pointer to transmit queue
2909  *
2910  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2911  * to give appropriate hint to the CPU.
2912  */
2913 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2914 {
2915 #ifdef CONFIG_BQL
2916 	prefetchw(&dev_queue->dql.num_queued);
2917 #endif
2918 }
2919 
2920 /**
2921  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2922  *	@dev_queue: pointer to transmit queue
2923  *
2924  * BQL enabled drivers might use this helper in their TX completion path,
2925  * to give appropriate hint to the CPU.
2926  */
2927 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2928 {
2929 #ifdef CONFIG_BQL
2930 	prefetchw(&dev_queue->dql.limit);
2931 #endif
2932 }
2933 
2934 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2935 					unsigned int bytes)
2936 {
2937 #ifdef CONFIG_BQL
2938 	dql_queued(&dev_queue->dql, bytes);
2939 
2940 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2941 		return;
2942 
2943 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2944 
2945 	/*
2946 	 * The XOFF flag must be set before checking the dql_avail below,
2947 	 * because in netdev_tx_completed_queue we update the dql_completed
2948 	 * before checking the XOFF flag.
2949 	 */
2950 	smp_mb();
2951 
2952 	/* check again in case another CPU has just made room avail */
2953 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2954 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2955 #endif
2956 }
2957 
2958 /**
2959  * 	netdev_sent_queue - report the number of bytes queued to hardware
2960  * 	@dev: network device
2961  * 	@bytes: number of bytes queued to the hardware device queue
2962  *
2963  * 	Report the number of bytes queued for sending/completion to the network
2964  * 	device hardware queue. @bytes should be a good approximation and should
2965  * 	exactly match netdev_completed_queue() @bytes
2966  */
2967 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2968 {
2969 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2970 }
2971 
2972 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2973 					     unsigned int pkts, unsigned int bytes)
2974 {
2975 #ifdef CONFIG_BQL
2976 	if (unlikely(!bytes))
2977 		return;
2978 
2979 	dql_completed(&dev_queue->dql, bytes);
2980 
2981 	/*
2982 	 * Without the memory barrier there is a small possiblity that
2983 	 * netdev_tx_sent_queue will miss the update and cause the queue to
2984 	 * be stopped forever
2985 	 */
2986 	smp_mb();
2987 
2988 	if (dql_avail(&dev_queue->dql) < 0)
2989 		return;
2990 
2991 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2992 		netif_schedule_queue(dev_queue);
2993 #endif
2994 }
2995 
2996 /**
2997  * 	netdev_completed_queue - report bytes and packets completed by device
2998  * 	@dev: network device
2999  * 	@pkts: actual number of packets sent over the medium
3000  * 	@bytes: actual number of bytes sent over the medium
3001  *
3002  * 	Report the number of bytes and packets transmitted by the network device
3003  * 	hardware queue over the physical medium, @bytes must exactly match the
3004  * 	@bytes amount passed to netdev_sent_queue()
3005  */
3006 static inline void netdev_completed_queue(struct net_device *dev,
3007 					  unsigned int pkts, unsigned int bytes)
3008 {
3009 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3010 }
3011 
3012 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3013 {
3014 #ifdef CONFIG_BQL
3015 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3016 	dql_reset(&q->dql);
3017 #endif
3018 }
3019 
3020 /**
3021  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3022  * 	@dev_queue: network device
3023  *
3024  * 	Reset the bytes and packet count of a network device and clear the
3025  * 	software flow control OFF bit for this network device
3026  */
3027 static inline void netdev_reset_queue(struct net_device *dev_queue)
3028 {
3029 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3030 }
3031 
3032 /**
3033  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3034  * 	@dev: network device
3035  * 	@queue_index: given tx queue index
3036  *
3037  * 	Returns 0 if given tx queue index >= number of device tx queues,
3038  * 	otherwise returns the originally passed tx queue index.
3039  */
3040 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3041 {
3042 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3043 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3044 				     dev->name, queue_index,
3045 				     dev->real_num_tx_queues);
3046 		return 0;
3047 	}
3048 
3049 	return queue_index;
3050 }
3051 
3052 /**
3053  *	netif_running - test if up
3054  *	@dev: network device
3055  *
3056  *	Test if the device has been brought up.
3057  */
3058 static inline bool netif_running(const struct net_device *dev)
3059 {
3060 	return test_bit(__LINK_STATE_START, &dev->state);
3061 }
3062 
3063 /*
3064  * Routines to manage the subqueues on a device.  We only need start,
3065  * stop, and a check if it's stopped.  All other device management is
3066  * done at the overall netdevice level.
3067  * Also test the device if we're multiqueue.
3068  */
3069 
3070 /**
3071  *	netif_start_subqueue - allow sending packets on subqueue
3072  *	@dev: network device
3073  *	@queue_index: sub queue index
3074  *
3075  * Start individual transmit queue of a device with multiple transmit queues.
3076  */
3077 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3078 {
3079 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3080 
3081 	netif_tx_start_queue(txq);
3082 }
3083 
3084 /**
3085  *	netif_stop_subqueue - stop sending packets on subqueue
3086  *	@dev: network device
3087  *	@queue_index: sub queue index
3088  *
3089  * Stop individual transmit queue of a device with multiple transmit queues.
3090  */
3091 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3092 {
3093 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3094 	netif_tx_stop_queue(txq);
3095 }
3096 
3097 /**
3098  *	netif_subqueue_stopped - test status of subqueue
3099  *	@dev: network device
3100  *	@queue_index: sub queue index
3101  *
3102  * Check individual transmit queue of a device with multiple transmit queues.
3103  */
3104 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3105 					    u16 queue_index)
3106 {
3107 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3108 
3109 	return netif_tx_queue_stopped(txq);
3110 }
3111 
3112 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3113 					  struct sk_buff *skb)
3114 {
3115 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3116 }
3117 
3118 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3119 
3120 #ifdef CONFIG_XPS
3121 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3122 			u16 index);
3123 #else
3124 static inline int netif_set_xps_queue(struct net_device *dev,
3125 				      const struct cpumask *mask,
3126 				      u16 index)
3127 {
3128 	return 0;
3129 }
3130 #endif
3131 
3132 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3133 		  unsigned int num_tx_queues);
3134 
3135 /*
3136  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3137  * as a distribution range limit for the returned value.
3138  */
3139 static inline u16 skb_tx_hash(const struct net_device *dev,
3140 			      struct sk_buff *skb)
3141 {
3142 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3143 }
3144 
3145 /**
3146  *	netif_is_multiqueue - test if device has multiple transmit queues
3147  *	@dev: network device
3148  *
3149  * Check if device has multiple transmit queues
3150  */
3151 static inline bool netif_is_multiqueue(const struct net_device *dev)
3152 {
3153 	return dev->num_tx_queues > 1;
3154 }
3155 
3156 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3157 
3158 #ifdef CONFIG_SYSFS
3159 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3160 #else
3161 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3162 						unsigned int rxq)
3163 {
3164 	return 0;
3165 }
3166 #endif
3167 
3168 #ifdef CONFIG_SYSFS
3169 static inline unsigned int get_netdev_rx_queue_index(
3170 		struct netdev_rx_queue *queue)
3171 {
3172 	struct net_device *dev = queue->dev;
3173 	int index = queue - dev->_rx;
3174 
3175 	BUG_ON(index >= dev->num_rx_queues);
3176 	return index;
3177 }
3178 #endif
3179 
3180 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3181 int netif_get_num_default_rss_queues(void);
3182 
3183 enum skb_free_reason {
3184 	SKB_REASON_CONSUMED,
3185 	SKB_REASON_DROPPED,
3186 };
3187 
3188 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3189 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3190 
3191 /*
3192  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3193  * interrupt context or with hardware interrupts being disabled.
3194  * (in_irq() || irqs_disabled())
3195  *
3196  * We provide four helpers that can be used in following contexts :
3197  *
3198  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3199  *  replacing kfree_skb(skb)
3200  *
3201  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3202  *  Typically used in place of consume_skb(skb) in TX completion path
3203  *
3204  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3205  *  replacing kfree_skb(skb)
3206  *
3207  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3208  *  and consumed a packet. Used in place of consume_skb(skb)
3209  */
3210 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3211 {
3212 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3213 }
3214 
3215 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3216 {
3217 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3218 }
3219 
3220 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3221 {
3222 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3223 }
3224 
3225 static inline void dev_consume_skb_any(struct sk_buff *skb)
3226 {
3227 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3228 }
3229 
3230 int netif_rx(struct sk_buff *skb);
3231 int netif_rx_ni(struct sk_buff *skb);
3232 int netif_receive_skb(struct sk_buff *skb);
3233 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3234 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3235 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3236 gro_result_t napi_gro_frags(struct napi_struct *napi);
3237 struct packet_offload *gro_find_receive_by_type(__be16 type);
3238 struct packet_offload *gro_find_complete_by_type(__be16 type);
3239 
3240 static inline void napi_free_frags(struct napi_struct *napi)
3241 {
3242 	kfree_skb(napi->skb);
3243 	napi->skb = NULL;
3244 }
3245 
3246 int netdev_rx_handler_register(struct net_device *dev,
3247 			       rx_handler_func_t *rx_handler,
3248 			       void *rx_handler_data);
3249 void netdev_rx_handler_unregister(struct net_device *dev);
3250 
3251 bool dev_valid_name(const char *name);
3252 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3253 int dev_ethtool(struct net *net, struct ifreq *);
3254 unsigned int dev_get_flags(const struct net_device *);
3255 int __dev_change_flags(struct net_device *, unsigned int flags);
3256 int dev_change_flags(struct net_device *, unsigned int);
3257 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3258 			unsigned int gchanges);
3259 int dev_change_name(struct net_device *, const char *);
3260 int dev_set_alias(struct net_device *, const char *, size_t);
3261 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3262 int dev_set_mtu(struct net_device *, int);
3263 void dev_set_group(struct net_device *, int);
3264 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3265 int dev_change_carrier(struct net_device *, bool new_carrier);
3266 int dev_get_phys_port_id(struct net_device *dev,
3267 			 struct netdev_phys_item_id *ppid);
3268 int dev_get_phys_port_name(struct net_device *dev,
3269 			   char *name, size_t len);
3270 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3271 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3272 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3273 				    struct netdev_queue *txq, int *ret);
3274 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3275 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3276 bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
3277 
3278 extern int		netdev_budget;
3279 
3280 /* Called by rtnetlink.c:rtnl_unlock() */
3281 void netdev_run_todo(void);
3282 
3283 /**
3284  *	dev_put - release reference to device
3285  *	@dev: network device
3286  *
3287  * Release reference to device to allow it to be freed.
3288  */
3289 static inline void dev_put(struct net_device *dev)
3290 {
3291 	this_cpu_dec(*dev->pcpu_refcnt);
3292 }
3293 
3294 /**
3295  *	dev_hold - get reference to device
3296  *	@dev: network device
3297  *
3298  * Hold reference to device to keep it from being freed.
3299  */
3300 static inline void dev_hold(struct net_device *dev)
3301 {
3302 	this_cpu_inc(*dev->pcpu_refcnt);
3303 }
3304 
3305 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3306  * and _off may be called from IRQ context, but it is caller
3307  * who is responsible for serialization of these calls.
3308  *
3309  * The name carrier is inappropriate, these functions should really be
3310  * called netif_lowerlayer_*() because they represent the state of any
3311  * kind of lower layer not just hardware media.
3312  */
3313 
3314 void linkwatch_init_dev(struct net_device *dev);
3315 void linkwatch_fire_event(struct net_device *dev);
3316 void linkwatch_forget_dev(struct net_device *dev);
3317 
3318 /**
3319  *	netif_carrier_ok - test if carrier present
3320  *	@dev: network device
3321  *
3322  * Check if carrier is present on device
3323  */
3324 static inline bool netif_carrier_ok(const struct net_device *dev)
3325 {
3326 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3327 }
3328 
3329 unsigned long dev_trans_start(struct net_device *dev);
3330 
3331 void __netdev_watchdog_up(struct net_device *dev);
3332 
3333 void netif_carrier_on(struct net_device *dev);
3334 
3335 void netif_carrier_off(struct net_device *dev);
3336 
3337 /**
3338  *	netif_dormant_on - mark device as dormant.
3339  *	@dev: network device
3340  *
3341  * Mark device as dormant (as per RFC2863).
3342  *
3343  * The dormant state indicates that the relevant interface is not
3344  * actually in a condition to pass packets (i.e., it is not 'up') but is
3345  * in a "pending" state, waiting for some external event.  For "on-
3346  * demand" interfaces, this new state identifies the situation where the
3347  * interface is waiting for events to place it in the up state.
3348  */
3349 static inline void netif_dormant_on(struct net_device *dev)
3350 {
3351 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3352 		linkwatch_fire_event(dev);
3353 }
3354 
3355 /**
3356  *	netif_dormant_off - set device as not dormant.
3357  *	@dev: network device
3358  *
3359  * Device is not in dormant state.
3360  */
3361 static inline void netif_dormant_off(struct net_device *dev)
3362 {
3363 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3364 		linkwatch_fire_event(dev);
3365 }
3366 
3367 /**
3368  *	netif_dormant - test if carrier present
3369  *	@dev: network device
3370  *
3371  * Check if carrier is present on device
3372  */
3373 static inline bool netif_dormant(const struct net_device *dev)
3374 {
3375 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3376 }
3377 
3378 
3379 /**
3380  *	netif_oper_up - test if device is operational
3381  *	@dev: network device
3382  *
3383  * Check if carrier is operational
3384  */
3385 static inline bool netif_oper_up(const struct net_device *dev)
3386 {
3387 	return (dev->operstate == IF_OPER_UP ||
3388 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3389 }
3390 
3391 /**
3392  *	netif_device_present - is device available or removed
3393  *	@dev: network device
3394  *
3395  * Check if device has not been removed from system.
3396  */
3397 static inline bool netif_device_present(struct net_device *dev)
3398 {
3399 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3400 }
3401 
3402 void netif_device_detach(struct net_device *dev);
3403 
3404 void netif_device_attach(struct net_device *dev);
3405 
3406 /*
3407  * Network interface message level settings
3408  */
3409 
3410 enum {
3411 	NETIF_MSG_DRV		= 0x0001,
3412 	NETIF_MSG_PROBE		= 0x0002,
3413 	NETIF_MSG_LINK		= 0x0004,
3414 	NETIF_MSG_TIMER		= 0x0008,
3415 	NETIF_MSG_IFDOWN	= 0x0010,
3416 	NETIF_MSG_IFUP		= 0x0020,
3417 	NETIF_MSG_RX_ERR	= 0x0040,
3418 	NETIF_MSG_TX_ERR	= 0x0080,
3419 	NETIF_MSG_TX_QUEUED	= 0x0100,
3420 	NETIF_MSG_INTR		= 0x0200,
3421 	NETIF_MSG_TX_DONE	= 0x0400,
3422 	NETIF_MSG_RX_STATUS	= 0x0800,
3423 	NETIF_MSG_PKTDATA	= 0x1000,
3424 	NETIF_MSG_HW		= 0x2000,
3425 	NETIF_MSG_WOL		= 0x4000,
3426 };
3427 
3428 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3429 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3430 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3431 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3432 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3433 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3434 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3435 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3436 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3437 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3438 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3439 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3440 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3441 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3442 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3443 
3444 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3445 {
3446 	/* use default */
3447 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3448 		return default_msg_enable_bits;
3449 	if (debug_value == 0)	/* no output */
3450 		return 0;
3451 	/* set low N bits */
3452 	return (1 << debug_value) - 1;
3453 }
3454 
3455 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3456 {
3457 	spin_lock(&txq->_xmit_lock);
3458 	txq->xmit_lock_owner = cpu;
3459 }
3460 
3461 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3462 {
3463 	spin_lock_bh(&txq->_xmit_lock);
3464 	txq->xmit_lock_owner = smp_processor_id();
3465 }
3466 
3467 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3468 {
3469 	bool ok = spin_trylock(&txq->_xmit_lock);
3470 	if (likely(ok))
3471 		txq->xmit_lock_owner = smp_processor_id();
3472 	return ok;
3473 }
3474 
3475 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3476 {
3477 	txq->xmit_lock_owner = -1;
3478 	spin_unlock(&txq->_xmit_lock);
3479 }
3480 
3481 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3482 {
3483 	txq->xmit_lock_owner = -1;
3484 	spin_unlock_bh(&txq->_xmit_lock);
3485 }
3486 
3487 static inline void txq_trans_update(struct netdev_queue *txq)
3488 {
3489 	if (txq->xmit_lock_owner != -1)
3490 		txq->trans_start = jiffies;
3491 }
3492 
3493 /**
3494  *	netif_tx_lock - grab network device transmit lock
3495  *	@dev: network device
3496  *
3497  * Get network device transmit lock
3498  */
3499 static inline void netif_tx_lock(struct net_device *dev)
3500 {
3501 	unsigned int i;
3502 	int cpu;
3503 
3504 	spin_lock(&dev->tx_global_lock);
3505 	cpu = smp_processor_id();
3506 	for (i = 0; i < dev->num_tx_queues; i++) {
3507 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3508 
3509 		/* We are the only thread of execution doing a
3510 		 * freeze, but we have to grab the _xmit_lock in
3511 		 * order to synchronize with threads which are in
3512 		 * the ->hard_start_xmit() handler and already
3513 		 * checked the frozen bit.
3514 		 */
3515 		__netif_tx_lock(txq, cpu);
3516 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3517 		__netif_tx_unlock(txq);
3518 	}
3519 }
3520 
3521 static inline void netif_tx_lock_bh(struct net_device *dev)
3522 {
3523 	local_bh_disable();
3524 	netif_tx_lock(dev);
3525 }
3526 
3527 static inline void netif_tx_unlock(struct net_device *dev)
3528 {
3529 	unsigned int i;
3530 
3531 	for (i = 0; i < dev->num_tx_queues; i++) {
3532 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3533 
3534 		/* No need to grab the _xmit_lock here.  If the
3535 		 * queue is not stopped for another reason, we
3536 		 * force a schedule.
3537 		 */
3538 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3539 		netif_schedule_queue(txq);
3540 	}
3541 	spin_unlock(&dev->tx_global_lock);
3542 }
3543 
3544 static inline void netif_tx_unlock_bh(struct net_device *dev)
3545 {
3546 	netif_tx_unlock(dev);
3547 	local_bh_enable();
3548 }
3549 
3550 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3551 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3552 		__netif_tx_lock(txq, cpu);		\
3553 	}						\
3554 }
3555 
3556 #define HARD_TX_TRYLOCK(dev, txq)			\
3557 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3558 		__netif_tx_trylock(txq) :		\
3559 		true )
3560 
3561 #define HARD_TX_UNLOCK(dev, txq) {			\
3562 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3563 		__netif_tx_unlock(txq);			\
3564 	}						\
3565 }
3566 
3567 static inline void netif_tx_disable(struct net_device *dev)
3568 {
3569 	unsigned int i;
3570 	int cpu;
3571 
3572 	local_bh_disable();
3573 	cpu = smp_processor_id();
3574 	for (i = 0; i < dev->num_tx_queues; i++) {
3575 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3576 
3577 		__netif_tx_lock(txq, cpu);
3578 		netif_tx_stop_queue(txq);
3579 		__netif_tx_unlock(txq);
3580 	}
3581 	local_bh_enable();
3582 }
3583 
3584 static inline void netif_addr_lock(struct net_device *dev)
3585 {
3586 	spin_lock(&dev->addr_list_lock);
3587 }
3588 
3589 static inline void netif_addr_lock_nested(struct net_device *dev)
3590 {
3591 	int subclass = SINGLE_DEPTH_NESTING;
3592 
3593 	if (dev->netdev_ops->ndo_get_lock_subclass)
3594 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3595 
3596 	spin_lock_nested(&dev->addr_list_lock, subclass);
3597 }
3598 
3599 static inline void netif_addr_lock_bh(struct net_device *dev)
3600 {
3601 	spin_lock_bh(&dev->addr_list_lock);
3602 }
3603 
3604 static inline void netif_addr_unlock(struct net_device *dev)
3605 {
3606 	spin_unlock(&dev->addr_list_lock);
3607 }
3608 
3609 static inline void netif_addr_unlock_bh(struct net_device *dev)
3610 {
3611 	spin_unlock_bh(&dev->addr_list_lock);
3612 }
3613 
3614 /*
3615  * dev_addrs walker. Should be used only for read access. Call with
3616  * rcu_read_lock held.
3617  */
3618 #define for_each_dev_addr(dev, ha) \
3619 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3620 
3621 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3622 
3623 void ether_setup(struct net_device *dev);
3624 
3625 /* Support for loadable net-drivers */
3626 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3627 				    unsigned char name_assign_type,
3628 				    void (*setup)(struct net_device *),
3629 				    unsigned int txqs, unsigned int rxqs);
3630 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3631 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3632 
3633 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3634 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3635 			 count)
3636 
3637 int register_netdev(struct net_device *dev);
3638 void unregister_netdev(struct net_device *dev);
3639 
3640 /* General hardware address lists handling functions */
3641 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3642 		   struct netdev_hw_addr_list *from_list, int addr_len);
3643 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3644 		      struct netdev_hw_addr_list *from_list, int addr_len);
3645 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3646 		       struct net_device *dev,
3647 		       int (*sync)(struct net_device *, const unsigned char *),
3648 		       int (*unsync)(struct net_device *,
3649 				     const unsigned char *));
3650 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3651 			  struct net_device *dev,
3652 			  int (*unsync)(struct net_device *,
3653 					const unsigned char *));
3654 void __hw_addr_init(struct netdev_hw_addr_list *list);
3655 
3656 /* Functions used for device addresses handling */
3657 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3658 		 unsigned char addr_type);
3659 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3660 		 unsigned char addr_type);
3661 void dev_addr_flush(struct net_device *dev);
3662 int dev_addr_init(struct net_device *dev);
3663 
3664 /* Functions used for unicast addresses handling */
3665 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3666 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3667 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3668 int dev_uc_sync(struct net_device *to, struct net_device *from);
3669 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3670 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3671 void dev_uc_flush(struct net_device *dev);
3672 void dev_uc_init(struct net_device *dev);
3673 
3674 /**
3675  *  __dev_uc_sync - Synchonize device's unicast list
3676  *  @dev:  device to sync
3677  *  @sync: function to call if address should be added
3678  *  @unsync: function to call if address should be removed
3679  *
3680  *  Add newly added addresses to the interface, and release
3681  *  addresses that have been deleted.
3682  */
3683 static inline int __dev_uc_sync(struct net_device *dev,
3684 				int (*sync)(struct net_device *,
3685 					    const unsigned char *),
3686 				int (*unsync)(struct net_device *,
3687 					      const unsigned char *))
3688 {
3689 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3690 }
3691 
3692 /**
3693  *  __dev_uc_unsync - Remove synchronized addresses from device
3694  *  @dev:  device to sync
3695  *  @unsync: function to call if address should be removed
3696  *
3697  *  Remove all addresses that were added to the device by dev_uc_sync().
3698  */
3699 static inline void __dev_uc_unsync(struct net_device *dev,
3700 				   int (*unsync)(struct net_device *,
3701 						 const unsigned char *))
3702 {
3703 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3704 }
3705 
3706 /* Functions used for multicast addresses handling */
3707 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3708 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3709 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3710 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3711 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3712 int dev_mc_sync(struct net_device *to, struct net_device *from);
3713 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3714 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3715 void dev_mc_flush(struct net_device *dev);
3716 void dev_mc_init(struct net_device *dev);
3717 
3718 /**
3719  *  __dev_mc_sync - Synchonize device's multicast list
3720  *  @dev:  device to sync
3721  *  @sync: function to call if address should be added
3722  *  @unsync: function to call if address should be removed
3723  *
3724  *  Add newly added addresses to the interface, and release
3725  *  addresses that have been deleted.
3726  */
3727 static inline int __dev_mc_sync(struct net_device *dev,
3728 				int (*sync)(struct net_device *,
3729 					    const unsigned char *),
3730 				int (*unsync)(struct net_device *,
3731 					      const unsigned char *))
3732 {
3733 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3734 }
3735 
3736 /**
3737  *  __dev_mc_unsync - Remove synchronized addresses from device
3738  *  @dev:  device to sync
3739  *  @unsync: function to call if address should be removed
3740  *
3741  *  Remove all addresses that were added to the device by dev_mc_sync().
3742  */
3743 static inline void __dev_mc_unsync(struct net_device *dev,
3744 				   int (*unsync)(struct net_device *,
3745 						 const unsigned char *))
3746 {
3747 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3748 }
3749 
3750 /* Functions used for secondary unicast and multicast support */
3751 void dev_set_rx_mode(struct net_device *dev);
3752 void __dev_set_rx_mode(struct net_device *dev);
3753 int dev_set_promiscuity(struct net_device *dev, int inc);
3754 int dev_set_allmulti(struct net_device *dev, int inc);
3755 void netdev_state_change(struct net_device *dev);
3756 void netdev_notify_peers(struct net_device *dev);
3757 void netdev_features_change(struct net_device *dev);
3758 /* Load a device via the kmod */
3759 void dev_load(struct net *net, const char *name);
3760 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3761 					struct rtnl_link_stats64 *storage);
3762 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3763 			     const struct net_device_stats *netdev_stats);
3764 
3765 extern int		netdev_max_backlog;
3766 extern int		netdev_tstamp_prequeue;
3767 extern int		weight_p;
3768 extern int		bpf_jit_enable;
3769 
3770 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3771 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3772 						     struct list_head **iter);
3773 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3774 						     struct list_head **iter);
3775 
3776 /* iterate through upper list, must be called under RCU read lock */
3777 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3778 	for (iter = &(dev)->adj_list.upper, \
3779 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3780 	     updev; \
3781 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3782 
3783 /* iterate through upper list, must be called under RCU read lock */
3784 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3785 	for (iter = &(dev)->all_adj_list.upper, \
3786 	     updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3787 	     updev; \
3788 	     updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3789 
3790 void *netdev_lower_get_next_private(struct net_device *dev,
3791 				    struct list_head **iter);
3792 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3793 					struct list_head **iter);
3794 
3795 #define netdev_for_each_lower_private(dev, priv, iter) \
3796 	for (iter = (dev)->adj_list.lower.next, \
3797 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3798 	     priv; \
3799 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3800 
3801 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3802 	for (iter = &(dev)->adj_list.lower, \
3803 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3804 	     priv; \
3805 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3806 
3807 void *netdev_lower_get_next(struct net_device *dev,
3808 				struct list_head **iter);
3809 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3810 	for (iter = (dev)->adj_list.lower.next, \
3811 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3812 	     ldev; \
3813 	     ldev = netdev_lower_get_next(dev, &(iter)))
3814 
3815 void *netdev_adjacent_get_private(struct list_head *adj_list);
3816 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3817 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3818 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3819 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3820 int netdev_master_upper_dev_link(struct net_device *dev,
3821 				 struct net_device *upper_dev,
3822 				 void *upper_priv, void *upper_info);
3823 void netdev_upper_dev_unlink(struct net_device *dev,
3824 			     struct net_device *upper_dev);
3825 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3826 void *netdev_lower_dev_get_private(struct net_device *dev,
3827 				   struct net_device *lower_dev);
3828 void netdev_lower_state_changed(struct net_device *lower_dev,
3829 				void *lower_state_info);
3830 
3831 /* RSS keys are 40 or 52 bytes long */
3832 #define NETDEV_RSS_KEY_LEN 52
3833 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3834 void netdev_rss_key_fill(void *buffer, size_t len);
3835 
3836 int dev_get_nest_level(struct net_device *dev,
3837 		       bool (*type_check)(const struct net_device *dev));
3838 int skb_checksum_help(struct sk_buff *skb);
3839 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3840 				  netdev_features_t features, bool tx_path);
3841 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3842 				    netdev_features_t features);
3843 
3844 struct netdev_bonding_info {
3845 	ifslave	slave;
3846 	ifbond	master;
3847 };
3848 
3849 struct netdev_notifier_bonding_info {
3850 	struct netdev_notifier_info info; /* must be first */
3851 	struct netdev_bonding_info  bonding_info;
3852 };
3853 
3854 void netdev_bonding_info_change(struct net_device *dev,
3855 				struct netdev_bonding_info *bonding_info);
3856 
3857 static inline
3858 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3859 {
3860 	return __skb_gso_segment(skb, features, true);
3861 }
3862 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3863 
3864 static inline bool can_checksum_protocol(netdev_features_t features,
3865 					 __be16 protocol)
3866 {
3867 	if (protocol == htons(ETH_P_FCOE))
3868 		return !!(features & NETIF_F_FCOE_CRC);
3869 
3870 	/* Assume this is an IP checksum (not SCTP CRC) */
3871 
3872 	if (features & NETIF_F_HW_CSUM) {
3873 		/* Can checksum everything */
3874 		return true;
3875 	}
3876 
3877 	switch (protocol) {
3878 	case htons(ETH_P_IP):
3879 		return !!(features & NETIF_F_IP_CSUM);
3880 	case htons(ETH_P_IPV6):
3881 		return !!(features & NETIF_F_IPV6_CSUM);
3882 	default:
3883 		return false;
3884 	}
3885 }
3886 
3887 /* Map an ethertype into IP protocol if possible */
3888 static inline int eproto_to_ipproto(int eproto)
3889 {
3890 	switch (eproto) {
3891 	case htons(ETH_P_IP):
3892 		return IPPROTO_IP;
3893 	case htons(ETH_P_IPV6):
3894 		return IPPROTO_IPV6;
3895 	default:
3896 		return -1;
3897 	}
3898 }
3899 
3900 #ifdef CONFIG_BUG
3901 void netdev_rx_csum_fault(struct net_device *dev);
3902 #else
3903 static inline void netdev_rx_csum_fault(struct net_device *dev)
3904 {
3905 }
3906 #endif
3907 /* rx skb timestamps */
3908 void net_enable_timestamp(void);
3909 void net_disable_timestamp(void);
3910 
3911 #ifdef CONFIG_PROC_FS
3912 int __init dev_proc_init(void);
3913 #else
3914 #define dev_proc_init() 0
3915 #endif
3916 
3917 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3918 					      struct sk_buff *skb, struct net_device *dev,
3919 					      bool more)
3920 {
3921 	skb->xmit_more = more ? 1 : 0;
3922 	return ops->ndo_start_xmit(skb, dev);
3923 }
3924 
3925 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3926 					    struct netdev_queue *txq, bool more)
3927 {
3928 	const struct net_device_ops *ops = dev->netdev_ops;
3929 	int rc;
3930 
3931 	rc = __netdev_start_xmit(ops, skb, dev, more);
3932 	if (rc == NETDEV_TX_OK)
3933 		txq_trans_update(txq);
3934 
3935 	return rc;
3936 }
3937 
3938 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3939 				const void *ns);
3940 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3941 				 const void *ns);
3942 
3943 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3944 {
3945 	return netdev_class_create_file_ns(class_attr, NULL);
3946 }
3947 
3948 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3949 {
3950 	netdev_class_remove_file_ns(class_attr, NULL);
3951 }
3952 
3953 extern struct kobj_ns_type_operations net_ns_type_operations;
3954 
3955 const char *netdev_drivername(const struct net_device *dev);
3956 
3957 void linkwatch_run_queue(void);
3958 
3959 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3960 							  netdev_features_t f2)
3961 {
3962 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3963 		if (f1 & NETIF_F_HW_CSUM)
3964 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3965 		else
3966 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3967 	}
3968 
3969 	return f1 & f2;
3970 }
3971 
3972 static inline netdev_features_t netdev_get_wanted_features(
3973 	struct net_device *dev)
3974 {
3975 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
3976 }
3977 netdev_features_t netdev_increment_features(netdev_features_t all,
3978 	netdev_features_t one, netdev_features_t mask);
3979 
3980 /* Allow TSO being used on stacked device :
3981  * Performing the GSO segmentation before last device
3982  * is a performance improvement.
3983  */
3984 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3985 							netdev_features_t mask)
3986 {
3987 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3988 }
3989 
3990 int __netdev_update_features(struct net_device *dev);
3991 void netdev_update_features(struct net_device *dev);
3992 void netdev_change_features(struct net_device *dev);
3993 
3994 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3995 					struct net_device *dev);
3996 
3997 netdev_features_t passthru_features_check(struct sk_buff *skb,
3998 					  struct net_device *dev,
3999 					  netdev_features_t features);
4000 netdev_features_t netif_skb_features(struct sk_buff *skb);
4001 
4002 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4003 {
4004 	netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
4005 
4006 	/* check flags correspondence */
4007 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4008 	BUILD_BUG_ON(SKB_GSO_UDP     != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4009 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4010 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4011 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4012 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4013 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4014 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4015 	BUILD_BUG_ON(SKB_GSO_IPIP    != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
4016 	BUILD_BUG_ON(SKB_GSO_SIT     != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
4017 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4018 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4019 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4020 
4021 	return (features & feature) == feature;
4022 }
4023 
4024 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4025 {
4026 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4027 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4028 }
4029 
4030 static inline bool netif_needs_gso(struct sk_buff *skb,
4031 				   netdev_features_t features)
4032 {
4033 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4034 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4035 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4036 }
4037 
4038 static inline void netif_set_gso_max_size(struct net_device *dev,
4039 					  unsigned int size)
4040 {
4041 	dev->gso_max_size = size;
4042 }
4043 
4044 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4045 					int pulled_hlen, u16 mac_offset,
4046 					int mac_len)
4047 {
4048 	skb->protocol = protocol;
4049 	skb->encapsulation = 1;
4050 	skb_push(skb, pulled_hlen);
4051 	skb_reset_transport_header(skb);
4052 	skb->mac_header = mac_offset;
4053 	skb->network_header = skb->mac_header + mac_len;
4054 	skb->mac_len = mac_len;
4055 }
4056 
4057 static inline bool netif_is_macsec(const struct net_device *dev)
4058 {
4059 	return dev->priv_flags & IFF_MACSEC;
4060 }
4061 
4062 static inline bool netif_is_macvlan(const struct net_device *dev)
4063 {
4064 	return dev->priv_flags & IFF_MACVLAN;
4065 }
4066 
4067 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4068 {
4069 	return dev->priv_flags & IFF_MACVLAN_PORT;
4070 }
4071 
4072 static inline bool netif_is_ipvlan(const struct net_device *dev)
4073 {
4074 	return dev->priv_flags & IFF_IPVLAN_SLAVE;
4075 }
4076 
4077 static inline bool netif_is_ipvlan_port(const struct net_device *dev)
4078 {
4079 	return dev->priv_flags & IFF_IPVLAN_MASTER;
4080 }
4081 
4082 static inline bool netif_is_bond_master(const struct net_device *dev)
4083 {
4084 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4085 }
4086 
4087 static inline bool netif_is_bond_slave(const struct net_device *dev)
4088 {
4089 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4090 }
4091 
4092 static inline bool netif_supports_nofcs(struct net_device *dev)
4093 {
4094 	return dev->priv_flags & IFF_SUPP_NOFCS;
4095 }
4096 
4097 static inline bool netif_is_l3_master(const struct net_device *dev)
4098 {
4099 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4100 }
4101 
4102 static inline bool netif_is_l3_slave(const struct net_device *dev)
4103 {
4104 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4105 }
4106 
4107 static inline bool netif_is_bridge_master(const struct net_device *dev)
4108 {
4109 	return dev->priv_flags & IFF_EBRIDGE;
4110 }
4111 
4112 static inline bool netif_is_bridge_port(const struct net_device *dev)
4113 {
4114 	return dev->priv_flags & IFF_BRIDGE_PORT;
4115 }
4116 
4117 static inline bool netif_is_ovs_master(const struct net_device *dev)
4118 {
4119 	return dev->priv_flags & IFF_OPENVSWITCH;
4120 }
4121 
4122 static inline bool netif_is_team_master(const struct net_device *dev)
4123 {
4124 	return dev->priv_flags & IFF_TEAM;
4125 }
4126 
4127 static inline bool netif_is_team_port(const struct net_device *dev)
4128 {
4129 	return dev->priv_flags & IFF_TEAM_PORT;
4130 }
4131 
4132 static inline bool netif_is_lag_master(const struct net_device *dev)
4133 {
4134 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4135 }
4136 
4137 static inline bool netif_is_lag_port(const struct net_device *dev)
4138 {
4139 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4140 }
4141 
4142 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4143 {
4144 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4145 }
4146 
4147 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4148 static inline void netif_keep_dst(struct net_device *dev)
4149 {
4150 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4151 }
4152 
4153 extern struct pernet_operations __net_initdata loopback_net_ops;
4154 
4155 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4156 
4157 /* netdev_printk helpers, similar to dev_printk */
4158 
4159 static inline const char *netdev_name(const struct net_device *dev)
4160 {
4161 	if (!dev->name[0] || strchr(dev->name, '%'))
4162 		return "(unnamed net_device)";
4163 	return dev->name;
4164 }
4165 
4166 static inline const char *netdev_reg_state(const struct net_device *dev)
4167 {
4168 	switch (dev->reg_state) {
4169 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4170 	case NETREG_REGISTERED: return "";
4171 	case NETREG_UNREGISTERING: return " (unregistering)";
4172 	case NETREG_UNREGISTERED: return " (unregistered)";
4173 	case NETREG_RELEASED: return " (released)";
4174 	case NETREG_DUMMY: return " (dummy)";
4175 	}
4176 
4177 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4178 	return " (unknown)";
4179 }
4180 
4181 __printf(3, 4)
4182 void netdev_printk(const char *level, const struct net_device *dev,
4183 		   const char *format, ...);
4184 __printf(2, 3)
4185 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4186 __printf(2, 3)
4187 void netdev_alert(const struct net_device *dev, const char *format, ...);
4188 __printf(2, 3)
4189 void netdev_crit(const struct net_device *dev, const char *format, ...);
4190 __printf(2, 3)
4191 void netdev_err(const struct net_device *dev, const char *format, ...);
4192 __printf(2, 3)
4193 void netdev_warn(const struct net_device *dev, const char *format, ...);
4194 __printf(2, 3)
4195 void netdev_notice(const struct net_device *dev, const char *format, ...);
4196 __printf(2, 3)
4197 void netdev_info(const struct net_device *dev, const char *format, ...);
4198 
4199 #define MODULE_ALIAS_NETDEV(device) \
4200 	MODULE_ALIAS("netdev-" device)
4201 
4202 #if defined(CONFIG_DYNAMIC_DEBUG)
4203 #define netdev_dbg(__dev, format, args...)			\
4204 do {								\
4205 	dynamic_netdev_dbg(__dev, format, ##args);		\
4206 } while (0)
4207 #elif defined(DEBUG)
4208 #define netdev_dbg(__dev, format, args...)			\
4209 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4210 #else
4211 #define netdev_dbg(__dev, format, args...)			\
4212 ({								\
4213 	if (0)							\
4214 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4215 })
4216 #endif
4217 
4218 #if defined(VERBOSE_DEBUG)
4219 #define netdev_vdbg	netdev_dbg
4220 #else
4221 
4222 #define netdev_vdbg(dev, format, args...)			\
4223 ({								\
4224 	if (0)							\
4225 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4226 	0;							\
4227 })
4228 #endif
4229 
4230 /*
4231  * netdev_WARN() acts like dev_printk(), but with the key difference
4232  * of using a WARN/WARN_ON to get the message out, including the
4233  * file/line information and a backtrace.
4234  */
4235 #define netdev_WARN(dev, format, args...)			\
4236 	WARN(1, "netdevice: %s%s\n" format, netdev_name(dev),	\
4237 	     netdev_reg_state(dev), ##args)
4238 
4239 /* netif printk helpers, similar to netdev_printk */
4240 
4241 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4242 do {					  			\
4243 	if (netif_msg_##type(priv))				\
4244 		netdev_printk(level, (dev), fmt, ##args);	\
4245 } while (0)
4246 
4247 #define netif_level(level, priv, type, dev, fmt, args...)	\
4248 do {								\
4249 	if (netif_msg_##type(priv))				\
4250 		netdev_##level(dev, fmt, ##args);		\
4251 } while (0)
4252 
4253 #define netif_emerg(priv, type, dev, fmt, args...)		\
4254 	netif_level(emerg, priv, type, dev, fmt, ##args)
4255 #define netif_alert(priv, type, dev, fmt, args...)		\
4256 	netif_level(alert, priv, type, dev, fmt, ##args)
4257 #define netif_crit(priv, type, dev, fmt, args...)		\
4258 	netif_level(crit, priv, type, dev, fmt, ##args)
4259 #define netif_err(priv, type, dev, fmt, args...)		\
4260 	netif_level(err, priv, type, dev, fmt, ##args)
4261 #define netif_warn(priv, type, dev, fmt, args...)		\
4262 	netif_level(warn, priv, type, dev, fmt, ##args)
4263 #define netif_notice(priv, type, dev, fmt, args...)		\
4264 	netif_level(notice, priv, type, dev, fmt, ##args)
4265 #define netif_info(priv, type, dev, fmt, args...)		\
4266 	netif_level(info, priv, type, dev, fmt, ##args)
4267 
4268 #if defined(CONFIG_DYNAMIC_DEBUG)
4269 #define netif_dbg(priv, type, netdev, format, args...)		\
4270 do {								\
4271 	if (netif_msg_##type(priv))				\
4272 		dynamic_netdev_dbg(netdev, format, ##args);	\
4273 } while (0)
4274 #elif defined(DEBUG)
4275 #define netif_dbg(priv, type, dev, format, args...)		\
4276 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4277 #else
4278 #define netif_dbg(priv, type, dev, format, args...)			\
4279 ({									\
4280 	if (0)								\
4281 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4282 	0;								\
4283 })
4284 #endif
4285 
4286 #if defined(VERBOSE_DEBUG)
4287 #define netif_vdbg	netif_dbg
4288 #else
4289 #define netif_vdbg(priv, type, dev, format, args...)		\
4290 ({								\
4291 	if (0)							\
4292 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4293 	0;							\
4294 })
4295 #endif
4296 
4297 /*
4298  *	The list of packet types we will receive (as opposed to discard)
4299  *	and the routines to invoke.
4300  *
4301  *	Why 16. Because with 16 the only overlap we get on a hash of the
4302  *	low nibble of the protocol value is RARP/SNAP/X.25.
4303  *
4304  *      NOTE:  That is no longer true with the addition of VLAN tags.  Not
4305  *             sure which should go first, but I bet it won't make much
4306  *             difference if we are running VLANs.  The good news is that
4307  *             this protocol won't be in the list unless compiled in, so
4308  *             the average user (w/out VLANs) will not be adversely affected.
4309  *             --BLG
4310  *
4311  *		0800	IP
4312  *		8100    802.1Q VLAN
4313  *		0001	802.3
4314  *		0002	AX.25
4315  *		0004	802.2
4316  *		8035	RARP
4317  *		0005	SNAP
4318  *		0805	X.25
4319  *		0806	ARP
4320  *		8137	IPX
4321  *		0009	Localtalk
4322  *		86DD	IPv6
4323  */
4324 #define PTYPE_HASH_SIZE	(16)
4325 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4326 
4327 #endif	/* _LINUX_NETDEVICE_H */
4328