xref: /linux-6.15/include/linux/netdevice.h (revision 0d240e78)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Definitions for the Interfaces handler.
7  *
8  * Version:	@(#)dev.h	1.0.10	08/12/93
9  *
10  * Authors:	Ross Biro
11  *		Fred N. van Kempen, <[email protected]>
12  *		Corey Minyard <[email protected]>
13  *		Donald J. Becker, <[email protected]>
14  *		Alan Cox, <[email protected]>
15  *		Bjorn Ekwall. <[email protected]>
16  *              Pekka Riikonen <[email protected]>
17  *
18  *		This program is free software; you can redistribute it and/or
19  *		modify it under the terms of the GNU General Public License
20  *		as published by the Free Software Foundation; either version
21  *		2 of the License, or (at your option) any later version.
22  *
23  *		Moved to /usr/include/linux for NET3
24  */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27 
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35 
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/dmaengine.h>
39 #include <linux/workqueue.h>
40 #include <linux/dynamic_queue_limits.h>
41 
42 #include <linux/ethtool.h>
43 #include <net/net_namespace.h>
44 #include <net/dsa.h>
45 #ifdef CONFIG_DCB
46 #include <net/dcbnl.h>
47 #endif
48 #include <net/netprio_cgroup.h>
49 
50 #include <linux/netdev_features.h>
51 #include <linux/neighbour.h>
52 #include <uapi/linux/netdevice.h>
53 #include <uapi/linux/if_bonding.h>
54 #include <uapi/linux/pkt_cls.h>
55 
56 struct netpoll_info;
57 struct device;
58 struct phy_device;
59 /* 802.11 specific */
60 struct wireless_dev;
61 /* 802.15.4 specific */
62 struct wpan_dev;
63 struct mpls_dev;
64 /* UDP Tunnel offloads */
65 struct udp_tunnel_info;
66 
67 void netdev_set_default_ethtool_ops(struct net_device *dev,
68 				    const struct ethtool_ops *ops);
69 
70 /* Backlog congestion levels */
71 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
72 #define NET_RX_DROP		1	/* packet dropped */
73 
74 /*
75  * Transmit return codes: transmit return codes originate from three different
76  * namespaces:
77  *
78  * - qdisc return codes
79  * - driver transmit return codes
80  * - errno values
81  *
82  * Drivers are allowed to return any one of those in their hard_start_xmit()
83  * function. Real network devices commonly used with qdiscs should only return
84  * the driver transmit return codes though - when qdiscs are used, the actual
85  * transmission happens asynchronously, so the value is not propagated to
86  * higher layers. Virtual network devices transmit synchronously; in this case
87  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
88  * others are propagated to higher layers.
89  */
90 
91 /* qdisc ->enqueue() return codes. */
92 #define NET_XMIT_SUCCESS	0x00
93 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
94 #define NET_XMIT_CN		0x02	/* congestion notification	*/
95 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
96 
97 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
98  * indicates that the device will soon be dropping packets, or already drops
99  * some packets of the same priority; prompting us to send less aggressively. */
100 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
101 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
102 
103 /* Driver transmit return codes */
104 #define NETDEV_TX_MASK		0xf0
105 
106 enum netdev_tx {
107 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
108 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
109 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
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 	unsigned long		tx_maxrate;
574 	/*
575 	 * Number of TX timeouts for this queue
576 	 * (/sys/class/net/DEV/Q/trans_timeout)
577 	 */
578 	unsigned long		trans_timeout;
579 /*
580  * write-mostly part
581  */
582 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
583 	int			xmit_lock_owner;
584 	/*
585 	 * Time (in jiffies) of last Tx
586 	 */
587 	unsigned long		trans_start;
588 
589 	unsigned long		state;
590 
591 #ifdef CONFIG_BQL
592 	struct dql		dql;
593 #endif
594 } ____cacheline_aligned_in_smp;
595 
596 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
597 {
598 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
599 	return q->numa_node;
600 #else
601 	return NUMA_NO_NODE;
602 #endif
603 }
604 
605 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
606 {
607 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
608 	q->numa_node = node;
609 #endif
610 }
611 
612 #ifdef CONFIG_RPS
613 /*
614  * This structure holds an RPS map which can be of variable length.  The
615  * map is an array of CPUs.
616  */
617 struct rps_map {
618 	unsigned int len;
619 	struct rcu_head rcu;
620 	u16 cpus[0];
621 };
622 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
623 
624 /*
625  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
626  * tail pointer for that CPU's input queue at the time of last enqueue, and
627  * a hardware filter index.
628  */
629 struct rps_dev_flow {
630 	u16 cpu;
631 	u16 filter;
632 	unsigned int last_qtail;
633 };
634 #define RPS_NO_FILTER 0xffff
635 
636 /*
637  * The rps_dev_flow_table structure contains a table of flow mappings.
638  */
639 struct rps_dev_flow_table {
640 	unsigned int mask;
641 	struct rcu_head rcu;
642 	struct rps_dev_flow flows[0];
643 };
644 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
645     ((_num) * sizeof(struct rps_dev_flow)))
646 
647 /*
648  * The rps_sock_flow_table contains mappings of flows to the last CPU
649  * on which they were processed by the application (set in recvmsg).
650  * Each entry is a 32bit value. Upper part is the high-order bits
651  * of flow hash, lower part is CPU number.
652  * rps_cpu_mask is used to partition the space, depending on number of
653  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
654  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
655  * meaning we use 32-6=26 bits for the hash.
656  */
657 struct rps_sock_flow_table {
658 	u32	mask;
659 
660 	u32	ents[0] ____cacheline_aligned_in_smp;
661 };
662 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
663 
664 #define RPS_NO_CPU 0xffff
665 
666 extern u32 rps_cpu_mask;
667 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
668 
669 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
670 					u32 hash)
671 {
672 	if (table && hash) {
673 		unsigned int index = hash & table->mask;
674 		u32 val = hash & ~rps_cpu_mask;
675 
676 		/* We only give a hint, preemption can change CPU under us */
677 		val |= raw_smp_processor_id();
678 
679 		if (table->ents[index] != val)
680 			table->ents[index] = val;
681 	}
682 }
683 
684 #ifdef CONFIG_RFS_ACCEL
685 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
686 			 u16 filter_id);
687 #endif
688 #endif /* CONFIG_RPS */
689 
690 /* This structure contains an instance of an RX queue. */
691 struct netdev_rx_queue {
692 #ifdef CONFIG_RPS
693 	struct rps_map __rcu		*rps_map;
694 	struct rps_dev_flow_table __rcu	*rps_flow_table;
695 #endif
696 	struct kobject			kobj;
697 	struct net_device		*dev;
698 } ____cacheline_aligned_in_smp;
699 
700 /*
701  * RX queue sysfs structures and functions.
702  */
703 struct rx_queue_attribute {
704 	struct attribute attr;
705 	ssize_t (*show)(struct netdev_rx_queue *queue,
706 	    struct rx_queue_attribute *attr, char *buf);
707 	ssize_t (*store)(struct netdev_rx_queue *queue,
708 	    struct rx_queue_attribute *attr, const char *buf, size_t len);
709 };
710 
711 #ifdef CONFIG_XPS
712 /*
713  * This structure holds an XPS map which can be of variable length.  The
714  * map is an array of queues.
715  */
716 struct xps_map {
717 	unsigned int len;
718 	unsigned int alloc_len;
719 	struct rcu_head rcu;
720 	u16 queues[0];
721 };
722 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
723 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
724        - sizeof(struct xps_map)) / sizeof(u16))
725 
726 /*
727  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
728  */
729 struct xps_dev_maps {
730 	struct rcu_head rcu;
731 	struct xps_map __rcu *cpu_map[0];
732 };
733 #define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) +		\
734     (nr_cpu_ids * sizeof(struct xps_map *)))
735 #endif /* CONFIG_XPS */
736 
737 #define TC_MAX_QUEUE	16
738 #define TC_BITMASK	15
739 /* HW offloaded queuing disciplines txq count and offset maps */
740 struct netdev_tc_txq {
741 	u16 count;
742 	u16 offset;
743 };
744 
745 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
746 /*
747  * This structure is to hold information about the device
748  * configured to run FCoE protocol stack.
749  */
750 struct netdev_fcoe_hbainfo {
751 	char	manufacturer[64];
752 	char	serial_number[64];
753 	char	hardware_version[64];
754 	char	driver_version[64];
755 	char	optionrom_version[64];
756 	char	firmware_version[64];
757 	char	model[256];
758 	char	model_description[256];
759 };
760 #endif
761 
762 #define MAX_PHYS_ITEM_ID_LEN 32
763 
764 /* This structure holds a unique identifier to identify some
765  * physical item (port for example) used by a netdevice.
766  */
767 struct netdev_phys_item_id {
768 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
769 	unsigned char id_len;
770 };
771 
772 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
773 					    struct netdev_phys_item_id *b)
774 {
775 	return a->id_len == b->id_len &&
776 	       memcmp(a->id, b->id, a->id_len) == 0;
777 }
778 
779 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
780 				       struct sk_buff *skb);
781 
782 /* These structures hold the attributes of qdisc and classifiers
783  * that are being passed to the netdevice through the setup_tc op.
784  */
785 enum {
786 	TC_SETUP_MQPRIO,
787 	TC_SETUP_CLSU32,
788 	TC_SETUP_CLSFLOWER,
789 };
790 
791 struct tc_cls_u32_offload;
792 
793 struct tc_to_netdev {
794 	unsigned int type;
795 	union {
796 		u8 tc;
797 		struct tc_cls_u32_offload *cls_u32;
798 		struct tc_cls_flower_offload *cls_flower;
799 	};
800 };
801 
802 
803 /*
804  * This structure defines the management hooks for network devices.
805  * The following hooks can be defined; unless noted otherwise, they are
806  * optional and can be filled with a null pointer.
807  *
808  * int (*ndo_init)(struct net_device *dev);
809  *     This function is called once when a network device is registered.
810  *     The network device can use this for any late stage initialization
811  *     or semantic validation. It can fail with an error code which will
812  *     be propagated back to register_netdev.
813  *
814  * void (*ndo_uninit)(struct net_device *dev);
815  *     This function is called when device is unregistered or when registration
816  *     fails. It is not called if init fails.
817  *
818  * int (*ndo_open)(struct net_device *dev);
819  *     This function is called when a network device transitions to the up
820  *     state.
821  *
822  * int (*ndo_stop)(struct net_device *dev);
823  *     This function is called when a network device transitions to the down
824  *     state.
825  *
826  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
827  *                               struct net_device *dev);
828  *	Called when a packet needs to be transmitted.
829  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
830  *	the queue before that can happen; it's for obsolete devices and weird
831  *	corner cases, but the stack really does a non-trivial amount
832  *	of useless work if you return NETDEV_TX_BUSY.
833  *	Required; cannot be NULL.
834  *
835  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
836  *		netdev_features_t features);
837  *	Adjusts the requested feature flags according to device-specific
838  *	constraints, and returns the resulting flags. Must not modify
839  *	the device state.
840  *
841  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
842  *                         void *accel_priv, select_queue_fallback_t fallback);
843  *	Called to decide which queue to use when device supports multiple
844  *	transmit queues.
845  *
846  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
847  *	This function is called to allow device receiver to make
848  *	changes to configuration when multicast or promiscuous is enabled.
849  *
850  * void (*ndo_set_rx_mode)(struct net_device *dev);
851  *	This function is called device changes address list filtering.
852  *	If driver handles unicast address filtering, it should set
853  *	IFF_UNICAST_FLT in its priv_flags.
854  *
855  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
856  *	This function  is called when the Media Access Control address
857  *	needs to be changed. If this interface is not defined, the
858  *	MAC address can not be changed.
859  *
860  * int (*ndo_validate_addr)(struct net_device *dev);
861  *	Test if Media Access Control address is valid for the device.
862  *
863  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
864  *	Called when a user requests an ioctl which can't be handled by
865  *	the generic interface code. If not defined ioctls return
866  *	not supported error code.
867  *
868  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
869  *	Used to set network devices bus interface parameters. This interface
870  *	is retained for legacy reasons; new devices should use the bus
871  *	interface (PCI) for low level management.
872  *
873  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
874  *	Called when a user wants to change the Maximum Transfer Unit
875  *	of a device. If not defined, any request to change MTU will
876  *	will return an error.
877  *
878  * void (*ndo_tx_timeout)(struct net_device *dev);
879  *	Callback used when the transmitter has not made any progress
880  *	for dev->watchdog ticks.
881  *
882  * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
883  *                      struct rtnl_link_stats64 *storage);
884  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
885  *	Called when a user wants to get the network device usage
886  *	statistics. Drivers must do one of the following:
887  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
888  *	   rtnl_link_stats64 structure passed by the caller.
889  *	2. Define @ndo_get_stats to update a net_device_stats structure
890  *	   (which should normally be dev->stats) and return a pointer to
891  *	   it. The structure may be changed asynchronously only if each
892  *	   field is written atomically.
893  *	3. Update dev->stats asynchronously and atomically, and define
894  *	   neither operation.
895  *
896  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
897  *	If device supports VLAN filtering this function is called when a
898  *	VLAN id is registered.
899  *
900  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
901  *	If device supports VLAN filtering this function is called when a
902  *	VLAN id is unregistered.
903  *
904  * void (*ndo_poll_controller)(struct net_device *dev);
905  *
906  *	SR-IOV management functions.
907  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
908  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
909  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
910  *			  int max_tx_rate);
911  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
912  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
913  * int (*ndo_get_vf_config)(struct net_device *dev,
914  *			    int vf, struct ifla_vf_info *ivf);
915  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
916  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
917  *			  struct nlattr *port[]);
918  *
919  *      Enable or disable the VF ability to query its RSS Redirection Table and
920  *      Hash Key. This is needed since on some devices VF share this information
921  *      with PF and querying it may introduce a theoretical security risk.
922  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
923  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
924  * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
925  * 	Called to setup 'tc' number of traffic classes in the net device. This
926  * 	is always called from the stack with the rtnl lock held and netif tx
927  * 	queues stopped. This allows the netdevice to perform queue management
928  * 	safely.
929  *
930  *	Fiber Channel over Ethernet (FCoE) offload functions.
931  * int (*ndo_fcoe_enable)(struct net_device *dev);
932  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
933  *	so the underlying device can perform whatever needed configuration or
934  *	initialization to support acceleration of FCoE traffic.
935  *
936  * int (*ndo_fcoe_disable)(struct net_device *dev);
937  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
938  *	so the underlying device can perform whatever needed clean-ups to
939  *	stop supporting acceleration of FCoE traffic.
940  *
941  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
942  *			     struct scatterlist *sgl, unsigned int sgc);
943  *	Called when the FCoE Initiator wants to initialize an I/O that
944  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
945  *	perform necessary setup and returns 1 to indicate the device is set up
946  *	successfully to perform DDP on this I/O, otherwise this returns 0.
947  *
948  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
949  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
950  *	indicated by the FC exchange id 'xid', so the underlying device can
951  *	clean up and reuse resources for later DDP requests.
952  *
953  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
954  *			      struct scatterlist *sgl, unsigned int sgc);
955  *	Called when the FCoE Target wants to initialize an I/O that
956  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
957  *	perform necessary setup and returns 1 to indicate the device is set up
958  *	successfully to perform DDP on this I/O, otherwise this returns 0.
959  *
960  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
961  *			       struct netdev_fcoe_hbainfo *hbainfo);
962  *	Called when the FCoE Protocol stack wants information on the underlying
963  *	device. This information is utilized by the FCoE protocol stack to
964  *	register attributes with Fiber Channel management service as per the
965  *	FC-GS Fabric Device Management Information(FDMI) specification.
966  *
967  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
968  *	Called when the underlying device wants to override default World Wide
969  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
970  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
971  *	protocol stack to use.
972  *
973  *	RFS acceleration.
974  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
975  *			    u16 rxq_index, u32 flow_id);
976  *	Set hardware filter for RFS.  rxq_index is the target queue index;
977  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
978  *	Return the filter ID on success, or a negative error code.
979  *
980  *	Slave management functions (for bridge, bonding, etc).
981  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
982  *	Called to make another netdev an underling.
983  *
984  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
985  *	Called to release previously enslaved netdev.
986  *
987  *      Feature/offload setting functions.
988  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
989  *	Called to update device configuration to new features. Passed
990  *	feature set might be less than what was returned by ndo_fix_features()).
991  *	Must return >0 or -errno if it changed dev->features itself.
992  *
993  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
994  *		      struct net_device *dev,
995  *		      const unsigned char *addr, u16 vid, u16 flags)
996  *	Adds an FDB entry to dev for addr.
997  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
998  *		      struct net_device *dev,
999  *		      const unsigned char *addr, u16 vid)
1000  *	Deletes the FDB entry from dev coresponding to addr.
1001  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1002  *		       struct net_device *dev, struct net_device *filter_dev,
1003  *		       int idx)
1004  *	Used to add FDB entries to dump requests. Implementers should add
1005  *	entries to skb and update idx with the number of entries.
1006  *
1007  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1008  *			     u16 flags)
1009  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1010  *			     struct net_device *dev, u32 filter_mask,
1011  *			     int nlflags)
1012  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1013  *			     u16 flags);
1014  *
1015  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1016  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1017  *	which do not represent real hardware may define this to allow their
1018  *	userspace components to manage their virtual carrier state. Devices
1019  *	that determine carrier state from physical hardware properties (eg
1020  *	network cables) or protocol-dependent mechanisms (eg
1021  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1022  *
1023  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1024  *			       struct netdev_phys_item_id *ppid);
1025  *	Called to get ID of physical port of this device. If driver does
1026  *	not implement this, it is assumed that the hw is not able to have
1027  *	multiple net devices on single physical port.
1028  *
1029  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1030  *			      struct udp_tunnel_info *ti);
1031  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1032  *	address family that a UDP tunnel is listnening to. It is called only
1033  *	when a new port starts listening. The operation is protected by the
1034  *	RTNL.
1035  *
1036  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1037  *			      struct udp_tunnel_info *ti);
1038  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1039  *	address family that the UDP tunnel is not listening to anymore. The
1040  *	operation is protected by the RTNL.
1041  *
1042  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1043  *				 struct net_device *dev)
1044  *	Called by upper layer devices to accelerate switching or other
1045  *	station functionality into hardware. 'pdev is the lowerdev
1046  *	to use for the offload and 'dev' is the net device that will
1047  *	back the offload. Returns a pointer to the private structure
1048  *	the upper layer will maintain.
1049  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1050  *	Called by upper layer device to delete the station created
1051  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1052  *	the station and priv is the structure returned by the add
1053  *	operation.
1054  * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1055  *				      struct net_device *dev,
1056  *				      void *priv);
1057  *	Callback to use for xmit over the accelerated station. This
1058  *	is used in place of ndo_start_xmit on accelerated net
1059  *	devices.
1060  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1061  *					   struct net_device *dev
1062  *					   netdev_features_t features);
1063  *	Called by core transmit path to determine if device is capable of
1064  *	performing offload operations on a given packet. This is to give
1065  *	the device an opportunity to implement any restrictions that cannot
1066  *	be otherwise expressed by feature flags. The check is called with
1067  *	the set of features that the stack has calculated and it returns
1068  *	those the driver believes to be appropriate.
1069  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1070  *			     int queue_index, u32 maxrate);
1071  *	Called when a user wants to set a max-rate limitation of specific
1072  *	TX queue.
1073  * int (*ndo_get_iflink)(const struct net_device *dev);
1074  *	Called to get the iflink value of this device.
1075  * void (*ndo_change_proto_down)(struct net_device *dev,
1076  *				 bool proto_down);
1077  *	This function is used to pass protocol port error state information
1078  *	to the switch driver. The switch driver can react to the proto_down
1079  *      by doing a phys down on the associated switch port.
1080  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1081  *	This function is used to get egress tunnel information for given skb.
1082  *	This is useful for retrieving outer tunnel header parameters while
1083  *	sampling packet.
1084  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1085  *	This function is used to specify the headroom that the skb must
1086  *	consider when allocation skb during packet reception. Setting
1087  *	appropriate rx headroom value allows avoiding skb head copy on
1088  *	forward. Setting a negative value resets the rx headroom to the
1089  *	default value.
1090  *
1091  */
1092 struct net_device_ops {
1093 	int			(*ndo_init)(struct net_device *dev);
1094 	void			(*ndo_uninit)(struct net_device *dev);
1095 	int			(*ndo_open)(struct net_device *dev);
1096 	int			(*ndo_stop)(struct net_device *dev);
1097 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1098 						  struct net_device *dev);
1099 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1100 						      struct net_device *dev,
1101 						      netdev_features_t features);
1102 	u16			(*ndo_select_queue)(struct net_device *dev,
1103 						    struct sk_buff *skb,
1104 						    void *accel_priv,
1105 						    select_queue_fallback_t fallback);
1106 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1107 						       int flags);
1108 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1109 	int			(*ndo_set_mac_address)(struct net_device *dev,
1110 						       void *addr);
1111 	int			(*ndo_validate_addr)(struct net_device *dev);
1112 	int			(*ndo_do_ioctl)(struct net_device *dev,
1113 					        struct ifreq *ifr, int cmd);
1114 	int			(*ndo_set_config)(struct net_device *dev,
1115 					          struct ifmap *map);
1116 	int			(*ndo_change_mtu)(struct net_device *dev,
1117 						  int new_mtu);
1118 	int			(*ndo_neigh_setup)(struct net_device *dev,
1119 						   struct neigh_parms *);
1120 	void			(*ndo_tx_timeout) (struct net_device *dev);
1121 
1122 	struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1123 						     struct rtnl_link_stats64 *storage);
1124 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1125 
1126 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1127 						       __be16 proto, u16 vid);
1128 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1129 						        __be16 proto, u16 vid);
1130 #ifdef CONFIG_NET_POLL_CONTROLLER
1131 	void                    (*ndo_poll_controller)(struct net_device *dev);
1132 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1133 						     struct netpoll_info *info);
1134 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1135 #endif
1136 #ifdef CONFIG_NET_RX_BUSY_POLL
1137 	int			(*ndo_busy_poll)(struct napi_struct *dev);
1138 #endif
1139 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1140 						  int queue, u8 *mac);
1141 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1142 						   int queue, u16 vlan, u8 qos);
1143 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1144 						   int vf, int min_tx_rate,
1145 						   int max_tx_rate);
1146 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1147 						       int vf, bool setting);
1148 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1149 						    int vf, bool setting);
1150 	int			(*ndo_get_vf_config)(struct net_device *dev,
1151 						     int vf,
1152 						     struct ifla_vf_info *ivf);
1153 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1154 							 int vf, int link_state);
1155 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1156 						    int vf,
1157 						    struct ifla_vf_stats
1158 						    *vf_stats);
1159 	int			(*ndo_set_vf_port)(struct net_device *dev,
1160 						   int vf,
1161 						   struct nlattr *port[]);
1162 	int			(*ndo_get_vf_port)(struct net_device *dev,
1163 						   int vf, struct sk_buff *skb);
1164 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1165 						   int vf, u64 guid,
1166 						   int guid_type);
1167 	int			(*ndo_set_vf_rss_query_en)(
1168 						   struct net_device *dev,
1169 						   int vf, bool setting);
1170 	int			(*ndo_setup_tc)(struct net_device *dev,
1171 						u32 handle,
1172 						__be16 protocol,
1173 						struct tc_to_netdev *tc);
1174 #if IS_ENABLED(CONFIG_FCOE)
1175 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1176 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1177 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1178 						      u16 xid,
1179 						      struct scatterlist *sgl,
1180 						      unsigned int sgc);
1181 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1182 						     u16 xid);
1183 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1184 						       u16 xid,
1185 						       struct scatterlist *sgl,
1186 						       unsigned int sgc);
1187 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1188 							struct netdev_fcoe_hbainfo *hbainfo);
1189 #endif
1190 
1191 #if IS_ENABLED(CONFIG_LIBFCOE)
1192 #define NETDEV_FCOE_WWNN 0
1193 #define NETDEV_FCOE_WWPN 1
1194 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1195 						    u64 *wwn, int type);
1196 #endif
1197 
1198 #ifdef CONFIG_RFS_ACCEL
1199 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1200 						     const struct sk_buff *skb,
1201 						     u16 rxq_index,
1202 						     u32 flow_id);
1203 #endif
1204 	int			(*ndo_add_slave)(struct net_device *dev,
1205 						 struct net_device *slave_dev);
1206 	int			(*ndo_del_slave)(struct net_device *dev,
1207 						 struct net_device *slave_dev);
1208 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1209 						    netdev_features_t features);
1210 	int			(*ndo_set_features)(struct net_device *dev,
1211 						    netdev_features_t features);
1212 	int			(*ndo_neigh_construct)(struct neighbour *n);
1213 	void			(*ndo_neigh_destroy)(struct neighbour *n);
1214 
1215 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1216 					       struct nlattr *tb[],
1217 					       struct net_device *dev,
1218 					       const unsigned char *addr,
1219 					       u16 vid,
1220 					       u16 flags);
1221 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1222 					       struct nlattr *tb[],
1223 					       struct net_device *dev,
1224 					       const unsigned char *addr,
1225 					       u16 vid);
1226 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1227 						struct netlink_callback *cb,
1228 						struct net_device *dev,
1229 						struct net_device *filter_dev,
1230 						int idx);
1231 
1232 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1233 						      struct nlmsghdr *nlh,
1234 						      u16 flags);
1235 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1236 						      u32 pid, u32 seq,
1237 						      struct net_device *dev,
1238 						      u32 filter_mask,
1239 						      int nlflags);
1240 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1241 						      struct nlmsghdr *nlh,
1242 						      u16 flags);
1243 	int			(*ndo_change_carrier)(struct net_device *dev,
1244 						      bool new_carrier);
1245 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1246 							struct netdev_phys_item_id *ppid);
1247 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1248 							  char *name, size_t len);
1249 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1250 						      struct udp_tunnel_info *ti);
1251 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1252 						      struct udp_tunnel_info *ti);
1253 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1254 							struct net_device *dev);
1255 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1256 							void *priv);
1257 
1258 	netdev_tx_t		(*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1259 							struct net_device *dev,
1260 							void *priv);
1261 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1262 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1263 						      int queue_index,
1264 						      u32 maxrate);
1265 	int			(*ndo_get_iflink)(const struct net_device *dev);
1266 	int			(*ndo_change_proto_down)(struct net_device *dev,
1267 							 bool proto_down);
1268 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1269 						       struct sk_buff *skb);
1270 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1271 						       int needed_headroom);
1272 };
1273 
1274 /**
1275  * enum net_device_priv_flags - &struct net_device priv_flags
1276  *
1277  * These are the &struct net_device, they are only set internally
1278  * by drivers and used in the kernel. These flags are invisible to
1279  * userspace; this means that the order of these flags can change
1280  * during any kernel release.
1281  *
1282  * You should have a pretty good reason to be extending these flags.
1283  *
1284  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1285  * @IFF_EBRIDGE: Ethernet bridging device
1286  * @IFF_BONDING: bonding master or slave
1287  * @IFF_ISATAP: ISATAP interface (RFC4214)
1288  * @IFF_WAN_HDLC: WAN HDLC device
1289  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1290  *	release skb->dst
1291  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1292  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1293  * @IFF_MACVLAN_PORT: device used as macvlan port
1294  * @IFF_BRIDGE_PORT: device used as bridge port
1295  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1296  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1297  * @IFF_UNICAST_FLT: Supports unicast filtering
1298  * @IFF_TEAM_PORT: device used as team port
1299  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1300  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1301  *	change when it's running
1302  * @IFF_MACVLAN: Macvlan device
1303  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1304  *	underlying stacked devices
1305  * @IFF_IPVLAN_MASTER: IPvlan master device
1306  * @IFF_IPVLAN_SLAVE: IPvlan slave device
1307  * @IFF_L3MDEV_MASTER: device is an L3 master device
1308  * @IFF_NO_QUEUE: device can run without qdisc attached
1309  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1310  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1311  * @IFF_TEAM: device is a team device
1312  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1313  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1314  *	entity (i.e. the master device for bridged veth)
1315  * @IFF_MACSEC: device is a MACsec device
1316  */
1317 enum netdev_priv_flags {
1318 	IFF_802_1Q_VLAN			= 1<<0,
1319 	IFF_EBRIDGE			= 1<<1,
1320 	IFF_BONDING			= 1<<2,
1321 	IFF_ISATAP			= 1<<3,
1322 	IFF_WAN_HDLC			= 1<<4,
1323 	IFF_XMIT_DST_RELEASE		= 1<<5,
1324 	IFF_DONT_BRIDGE			= 1<<6,
1325 	IFF_DISABLE_NETPOLL		= 1<<7,
1326 	IFF_MACVLAN_PORT		= 1<<8,
1327 	IFF_BRIDGE_PORT			= 1<<9,
1328 	IFF_OVS_DATAPATH		= 1<<10,
1329 	IFF_TX_SKB_SHARING		= 1<<11,
1330 	IFF_UNICAST_FLT			= 1<<12,
1331 	IFF_TEAM_PORT			= 1<<13,
1332 	IFF_SUPP_NOFCS			= 1<<14,
1333 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1334 	IFF_MACVLAN			= 1<<16,
1335 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1336 	IFF_IPVLAN_MASTER		= 1<<18,
1337 	IFF_IPVLAN_SLAVE		= 1<<19,
1338 	IFF_L3MDEV_MASTER		= 1<<20,
1339 	IFF_NO_QUEUE			= 1<<21,
1340 	IFF_OPENVSWITCH			= 1<<22,
1341 	IFF_L3MDEV_SLAVE		= 1<<23,
1342 	IFF_TEAM			= 1<<24,
1343 	IFF_RXFH_CONFIGURED		= 1<<25,
1344 	IFF_PHONY_HEADROOM		= 1<<26,
1345 	IFF_MACSEC			= 1<<27,
1346 };
1347 
1348 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1349 #define IFF_EBRIDGE			IFF_EBRIDGE
1350 #define IFF_BONDING			IFF_BONDING
1351 #define IFF_ISATAP			IFF_ISATAP
1352 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1353 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1354 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1355 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1356 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1357 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1358 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1359 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1360 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1361 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1362 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1363 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1364 #define IFF_MACVLAN			IFF_MACVLAN
1365 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1366 #define IFF_IPVLAN_MASTER		IFF_IPVLAN_MASTER
1367 #define IFF_IPVLAN_SLAVE		IFF_IPVLAN_SLAVE
1368 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1369 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1370 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1371 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1372 #define IFF_TEAM			IFF_TEAM
1373 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1374 #define IFF_MACSEC			IFF_MACSEC
1375 
1376 /**
1377  *	struct net_device - The DEVICE structure.
1378  *		Actually, this whole structure is a big mistake.  It mixes I/O
1379  *		data with strictly "high-level" data, and it has to know about
1380  *		almost every data structure used in the INET module.
1381  *
1382  *	@name:	This is the first field of the "visible" part of this structure
1383  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1384  *	 	of the interface.
1385  *
1386  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1387  *	@ifalias:	SNMP alias
1388  *	@mem_end:	Shared memory end
1389  *	@mem_start:	Shared memory start
1390  *	@base_addr:	Device I/O address
1391  *	@irq:		Device IRQ number
1392  *
1393  *	@carrier_changes:	Stats to monitor carrier on<->off transitions
1394  *
1395  *	@state:		Generic network queuing layer state, see netdev_state_t
1396  *	@dev_list:	The global list of network devices
1397  *	@napi_list:	List entry used for polling NAPI devices
1398  *	@unreg_list:	List entry  when we are unregistering the
1399  *			device; see the function unregister_netdev
1400  *	@close_list:	List entry used when we are closing the device
1401  *	@ptype_all:     Device-specific packet handlers for all protocols
1402  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1403  *
1404  *	@adj_list:	Directly linked devices, like slaves for bonding
1405  *	@all_adj_list:	All linked devices, *including* neighbours
1406  *	@features:	Currently active device features
1407  *	@hw_features:	User-changeable features
1408  *
1409  *	@wanted_features:	User-requested features
1410  *	@vlan_features:		Mask of features inheritable by VLAN devices
1411  *
1412  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1413  *				This field indicates what encapsulation
1414  *				offloads the hardware is capable of doing,
1415  *				and drivers will need to set them appropriately.
1416  *
1417  *	@mpls_features:	Mask of features inheritable by MPLS
1418  *
1419  *	@ifindex:	interface index
1420  *	@group:		The group the device belongs to
1421  *
1422  *	@stats:		Statistics struct, which was left as a legacy, use
1423  *			rtnl_link_stats64 instead
1424  *
1425  *	@rx_dropped:	Dropped packets by core network,
1426  *			do not use this in drivers
1427  *	@tx_dropped:	Dropped packets by core network,
1428  *			do not use this in drivers
1429  *	@rx_nohandler:	nohandler dropped packets by core network on
1430  *			inactive devices, do not use this in drivers
1431  *
1432  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1433  *				instead of ioctl,
1434  *				see <net/iw_handler.h> for details.
1435  *	@wireless_data:	Instance data managed by the core of wireless extensions
1436  *
1437  *	@netdev_ops:	Includes several pointers to callbacks,
1438  *			if one wants to override the ndo_*() functions
1439  *	@ethtool_ops:	Management operations
1440  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1441  *			discovery handling. Necessary for e.g. 6LoWPAN.
1442  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1443  *			of Layer 2 headers.
1444  *
1445  *	@flags:		Interface flags (a la BSD)
1446  *	@priv_flags:	Like 'flags' but invisible to userspace,
1447  *			see if.h for the definitions
1448  *	@gflags:	Global flags ( kept as legacy )
1449  *	@padded:	How much padding added by alloc_netdev()
1450  *	@operstate:	RFC2863 operstate
1451  *	@link_mode:	Mapping policy to operstate
1452  *	@if_port:	Selectable AUI, TP, ...
1453  *	@dma:		DMA channel
1454  *	@mtu:		Interface MTU value
1455  *	@type:		Interface hardware type
1456  *	@hard_header_len: Maximum hardware header length.
1457  *
1458  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1459  *			  cases can this be guaranteed
1460  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1461  *			  cases can this be guaranteed. Some cases also use
1462  *			  LL_MAX_HEADER instead to allocate the skb
1463  *
1464  *	interface address info:
1465  *
1466  * 	@perm_addr:		Permanent hw address
1467  * 	@addr_assign_type:	Hw address assignment type
1468  * 	@addr_len:		Hardware address length
1469  *	@neigh_priv_len:	Used in neigh_alloc()
1470  * 	@dev_id:		Used to differentiate devices that share
1471  * 				the same link layer address
1472  * 	@dev_port:		Used to differentiate devices that share
1473  * 				the same function
1474  *	@addr_list_lock:	XXX: need comments on this one
1475  *	@uc_promisc:		Counter that indicates promiscuous mode
1476  *				has been enabled due to the need to listen to
1477  *				additional unicast addresses in a device that
1478  *				does not implement ndo_set_rx_mode()
1479  *	@uc:			unicast mac addresses
1480  *	@mc:			multicast mac addresses
1481  *	@dev_addrs:		list of device hw addresses
1482  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1483  *	@promiscuity:		Number of times the NIC is told to work in
1484  *				promiscuous mode; if it becomes 0 the NIC will
1485  *				exit promiscuous mode
1486  *	@allmulti:		Counter, enables or disables allmulticast mode
1487  *
1488  *	@vlan_info:	VLAN info
1489  *	@dsa_ptr:	dsa specific data
1490  *	@tipc_ptr:	TIPC specific data
1491  *	@atalk_ptr:	AppleTalk link
1492  *	@ip_ptr:	IPv4 specific data
1493  *	@dn_ptr:	DECnet specific data
1494  *	@ip6_ptr:	IPv6 specific data
1495  *	@ax25_ptr:	AX.25 specific data
1496  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1497  *
1498  *	@last_rx:	Time of last Rx
1499  *	@dev_addr:	Hw address (before bcast,
1500  *			because most packets are unicast)
1501  *
1502  *	@_rx:			Array of RX queues
1503  *	@num_rx_queues:		Number of RX queues
1504  *				allocated at register_netdev() time
1505  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1506  *
1507  *	@rx_handler:		handler for received packets
1508  *	@rx_handler_data: 	XXX: need comments on this one
1509  *	@ingress_queue:		XXX: need comments on this one
1510  *	@broadcast:		hw bcast address
1511  *
1512  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1513  *			indexed by RX queue number. Assigned by driver.
1514  *			This must only be set if the ndo_rx_flow_steer
1515  *			operation is defined
1516  *	@index_hlist:		Device index hash chain
1517  *
1518  *	@_tx:			Array of TX queues
1519  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1520  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1521  *	@qdisc:			Root qdisc from userspace point of view
1522  *	@tx_queue_len:		Max frames per queue allowed
1523  *	@tx_global_lock: 	XXX: need comments on this one
1524  *
1525  *	@xps_maps:	XXX: need comments on this one
1526  *
1527  *	@offload_fwd_mark:	Offload device fwding mark
1528  *
1529  *	@watchdog_timeo:	Represents the timeout that is used by
1530  *				the watchdog (see dev_watchdog())
1531  *	@watchdog_timer:	List of timers
1532  *
1533  *	@pcpu_refcnt:		Number of references to this device
1534  *	@todo_list:		Delayed register/unregister
1535  *	@link_watch_list:	XXX: need comments on this one
1536  *
1537  *	@reg_state:		Register/unregister state machine
1538  *	@dismantle:		Device is going to be freed
1539  *	@rtnl_link_state:	This enum represents the phases of creating
1540  *				a new link
1541  *
1542  *	@destructor:		Called from unregister,
1543  *				can be used to call free_netdev
1544  *	@npinfo:		XXX: need comments on this one
1545  * 	@nd_net:		Network namespace this network device is inside
1546  *
1547  * 	@ml_priv:	Mid-layer private
1548  * 	@lstats:	Loopback statistics
1549  * 	@tstats:	Tunnel statistics
1550  * 	@dstats:	Dummy statistics
1551  * 	@vstats:	Virtual ethernet statistics
1552  *
1553  *	@garp_port:	GARP
1554  *	@mrp_port:	MRP
1555  *
1556  *	@dev:		Class/net/name entry
1557  *	@sysfs_groups:	Space for optional device, statistics and wireless
1558  *			sysfs groups
1559  *
1560  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1561  *	@rtnl_link_ops:	Rtnl_link_ops
1562  *
1563  *	@gso_max_size:	Maximum size of generic segmentation offload
1564  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1565  *			NIC for GSO
1566  *
1567  *	@dcbnl_ops:	Data Center Bridging netlink ops
1568  *	@num_tc:	Number of traffic classes in the net device
1569  *	@tc_to_txq:	XXX: need comments on this one
1570  *	@prio_tc_map	XXX: need comments on this one
1571  *
1572  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1573  *
1574  *	@priomap:	XXX: need comments on this one
1575  *	@phydev:	Physical device may attach itself
1576  *			for hardware timestamping
1577  *
1578  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1579  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1580  *
1581  *	@proto_down:	protocol port state information can be sent to the
1582  *			switch driver and used to set the phys state of the
1583  *			switch port.
1584  *
1585  *	FIXME: cleanup struct net_device such that network protocol info
1586  *	moves out.
1587  */
1588 
1589 struct net_device {
1590 	char			name[IFNAMSIZ];
1591 	struct hlist_node	name_hlist;
1592 	char 			*ifalias;
1593 	/*
1594 	 *	I/O specific fields
1595 	 *	FIXME: Merge these and struct ifmap into one
1596 	 */
1597 	unsigned long		mem_end;
1598 	unsigned long		mem_start;
1599 	unsigned long		base_addr;
1600 	int			irq;
1601 
1602 	atomic_t		carrier_changes;
1603 
1604 	/*
1605 	 *	Some hardware also needs these fields (state,dev_list,
1606 	 *	napi_list,unreg_list,close_list) but they are not
1607 	 *	part of the usual set specified in Space.c.
1608 	 */
1609 
1610 	unsigned long		state;
1611 
1612 	struct list_head	dev_list;
1613 	struct list_head	napi_list;
1614 	struct list_head	unreg_list;
1615 	struct list_head	close_list;
1616 	struct list_head	ptype_all;
1617 	struct list_head	ptype_specific;
1618 
1619 	struct {
1620 		struct list_head upper;
1621 		struct list_head lower;
1622 	} adj_list;
1623 
1624 	struct {
1625 		struct list_head upper;
1626 		struct list_head lower;
1627 	} all_adj_list;
1628 
1629 	netdev_features_t	features;
1630 	netdev_features_t	hw_features;
1631 	netdev_features_t	wanted_features;
1632 	netdev_features_t	vlan_features;
1633 	netdev_features_t	hw_enc_features;
1634 	netdev_features_t	mpls_features;
1635 	netdev_features_t	gso_partial_features;
1636 
1637 	int			ifindex;
1638 	int			group;
1639 
1640 	struct net_device_stats	stats;
1641 
1642 	atomic_long_t		rx_dropped;
1643 	atomic_long_t		tx_dropped;
1644 	atomic_long_t		rx_nohandler;
1645 
1646 #ifdef CONFIG_WIRELESS_EXT
1647 	const struct iw_handler_def *wireless_handlers;
1648 	struct iw_public_data	*wireless_data;
1649 #endif
1650 	const struct net_device_ops *netdev_ops;
1651 	const struct ethtool_ops *ethtool_ops;
1652 #ifdef CONFIG_NET_SWITCHDEV
1653 	const struct switchdev_ops *switchdev_ops;
1654 #endif
1655 #ifdef CONFIG_NET_L3_MASTER_DEV
1656 	const struct l3mdev_ops	*l3mdev_ops;
1657 #endif
1658 #if IS_ENABLED(CONFIG_IPV6)
1659 	const struct ndisc_ops *ndisc_ops;
1660 #endif
1661 
1662 	const struct header_ops *header_ops;
1663 
1664 	unsigned int		flags;
1665 	unsigned int		priv_flags;
1666 
1667 	unsigned short		gflags;
1668 	unsigned short		padded;
1669 
1670 	unsigned char		operstate;
1671 	unsigned char		link_mode;
1672 
1673 	unsigned char		if_port;
1674 	unsigned char		dma;
1675 
1676 	unsigned int		mtu;
1677 	unsigned short		type;
1678 	unsigned short		hard_header_len;
1679 
1680 	unsigned short		needed_headroom;
1681 	unsigned short		needed_tailroom;
1682 
1683 	/* Interface address info. */
1684 	unsigned char		perm_addr[MAX_ADDR_LEN];
1685 	unsigned char		addr_assign_type;
1686 	unsigned char		addr_len;
1687 	unsigned short		neigh_priv_len;
1688 	unsigned short          dev_id;
1689 	unsigned short          dev_port;
1690 	spinlock_t		addr_list_lock;
1691 	unsigned char		name_assign_type;
1692 	bool			uc_promisc;
1693 	struct netdev_hw_addr_list	uc;
1694 	struct netdev_hw_addr_list	mc;
1695 	struct netdev_hw_addr_list	dev_addrs;
1696 
1697 #ifdef CONFIG_SYSFS
1698 	struct kset		*queues_kset;
1699 #endif
1700 	unsigned int		promiscuity;
1701 	unsigned int		allmulti;
1702 
1703 
1704 	/* Protocol-specific pointers */
1705 
1706 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1707 	struct vlan_info __rcu	*vlan_info;
1708 #endif
1709 #if IS_ENABLED(CONFIG_NET_DSA)
1710 	struct dsa_switch_tree	*dsa_ptr;
1711 #endif
1712 #if IS_ENABLED(CONFIG_TIPC)
1713 	struct tipc_bearer __rcu *tipc_ptr;
1714 #endif
1715 	void 			*atalk_ptr;
1716 	struct in_device __rcu	*ip_ptr;
1717 	struct dn_dev __rcu     *dn_ptr;
1718 	struct inet6_dev __rcu	*ip6_ptr;
1719 	void			*ax25_ptr;
1720 	struct wireless_dev	*ieee80211_ptr;
1721 	struct wpan_dev		*ieee802154_ptr;
1722 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1723 	struct mpls_dev __rcu	*mpls_ptr;
1724 #endif
1725 
1726 /*
1727  * Cache lines mostly used on receive path (including eth_type_trans())
1728  */
1729 	unsigned long		last_rx;
1730 
1731 	/* Interface address info used in eth_type_trans() */
1732 	unsigned char		*dev_addr;
1733 
1734 #ifdef CONFIG_SYSFS
1735 	struct netdev_rx_queue	*_rx;
1736 
1737 	unsigned int		num_rx_queues;
1738 	unsigned int		real_num_rx_queues;
1739 #endif
1740 
1741 	unsigned long		gro_flush_timeout;
1742 	rx_handler_func_t __rcu	*rx_handler;
1743 	void __rcu		*rx_handler_data;
1744 
1745 #ifdef CONFIG_NET_CLS_ACT
1746 	struct tcf_proto __rcu  *ingress_cl_list;
1747 #endif
1748 	struct netdev_queue __rcu *ingress_queue;
1749 #ifdef CONFIG_NETFILTER_INGRESS
1750 	struct list_head	nf_hooks_ingress;
1751 #endif
1752 
1753 	unsigned char		broadcast[MAX_ADDR_LEN];
1754 #ifdef CONFIG_RFS_ACCEL
1755 	struct cpu_rmap		*rx_cpu_rmap;
1756 #endif
1757 	struct hlist_node	index_hlist;
1758 
1759 /*
1760  * Cache lines mostly used on transmit path
1761  */
1762 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1763 	unsigned int		num_tx_queues;
1764 	unsigned int		real_num_tx_queues;
1765 	struct Qdisc		*qdisc;
1766 	unsigned long		tx_queue_len;
1767 	spinlock_t		tx_global_lock;
1768 	int			watchdog_timeo;
1769 
1770 #ifdef CONFIG_XPS
1771 	struct xps_dev_maps __rcu *xps_maps;
1772 #endif
1773 #ifdef CONFIG_NET_CLS_ACT
1774 	struct tcf_proto __rcu  *egress_cl_list;
1775 #endif
1776 #ifdef CONFIG_NET_SWITCHDEV
1777 	u32			offload_fwd_mark;
1778 #endif
1779 
1780 	/* These may be needed for future network-power-down code. */
1781 	struct timer_list	watchdog_timer;
1782 
1783 	int __percpu		*pcpu_refcnt;
1784 	struct list_head	todo_list;
1785 
1786 	struct list_head	link_watch_list;
1787 
1788 	enum { NETREG_UNINITIALIZED=0,
1789 	       NETREG_REGISTERED,	/* completed register_netdevice */
1790 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1791 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1792 	       NETREG_RELEASED,		/* called free_netdev */
1793 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1794 	} reg_state:8;
1795 
1796 	bool dismantle;
1797 
1798 	enum {
1799 		RTNL_LINK_INITIALIZED,
1800 		RTNL_LINK_INITIALIZING,
1801 	} rtnl_link_state:16;
1802 
1803 	void (*destructor)(struct net_device *dev);
1804 
1805 #ifdef CONFIG_NETPOLL
1806 	struct netpoll_info __rcu	*npinfo;
1807 #endif
1808 
1809 	possible_net_t			nd_net;
1810 
1811 	/* mid-layer private */
1812 	union {
1813 		void					*ml_priv;
1814 		struct pcpu_lstats __percpu		*lstats;
1815 		struct pcpu_sw_netstats __percpu	*tstats;
1816 		struct pcpu_dstats __percpu		*dstats;
1817 		struct pcpu_vstats __percpu		*vstats;
1818 	};
1819 
1820 	struct garp_port __rcu	*garp_port;
1821 	struct mrp_port __rcu	*mrp_port;
1822 
1823 	struct device		dev;
1824 	const struct attribute_group *sysfs_groups[4];
1825 	const struct attribute_group *sysfs_rx_queue_group;
1826 
1827 	const struct rtnl_link_ops *rtnl_link_ops;
1828 
1829 	/* for setting kernel sock attribute on TCP connection setup */
1830 #define GSO_MAX_SIZE		65536
1831 	unsigned int		gso_max_size;
1832 #define GSO_MAX_SEGS		65535
1833 	u16			gso_max_segs;
1834 
1835 #ifdef CONFIG_DCB
1836 	const struct dcbnl_rtnl_ops *dcbnl_ops;
1837 #endif
1838 	u8			num_tc;
1839 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
1840 	u8			prio_tc_map[TC_BITMASK + 1];
1841 
1842 #if IS_ENABLED(CONFIG_FCOE)
1843 	unsigned int		fcoe_ddp_xid;
1844 #endif
1845 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
1846 	struct netprio_map __rcu *priomap;
1847 #endif
1848 	struct phy_device	*phydev;
1849 	struct lock_class_key	*qdisc_tx_busylock;
1850 	struct lock_class_key	*qdisc_running_key;
1851 	bool			proto_down;
1852 };
1853 #define to_net_dev(d) container_of(d, struct net_device, dev)
1854 
1855 #define	NETDEV_ALIGN		32
1856 
1857 static inline
1858 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1859 {
1860 	return dev->prio_tc_map[prio & TC_BITMASK];
1861 }
1862 
1863 static inline
1864 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1865 {
1866 	if (tc >= dev->num_tc)
1867 		return -EINVAL;
1868 
1869 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1870 	return 0;
1871 }
1872 
1873 static inline
1874 void netdev_reset_tc(struct net_device *dev)
1875 {
1876 	dev->num_tc = 0;
1877 	memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1878 	memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1879 }
1880 
1881 static inline
1882 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1883 {
1884 	if (tc >= dev->num_tc)
1885 		return -EINVAL;
1886 
1887 	dev->tc_to_txq[tc].count = count;
1888 	dev->tc_to_txq[tc].offset = offset;
1889 	return 0;
1890 }
1891 
1892 static inline
1893 int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1894 {
1895 	if (num_tc > TC_MAX_QUEUE)
1896 		return -EINVAL;
1897 
1898 	dev->num_tc = num_tc;
1899 	return 0;
1900 }
1901 
1902 static inline
1903 int netdev_get_num_tc(struct net_device *dev)
1904 {
1905 	return dev->num_tc;
1906 }
1907 
1908 static inline
1909 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1910 					 unsigned int index)
1911 {
1912 	return &dev->_tx[index];
1913 }
1914 
1915 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1916 						    const struct sk_buff *skb)
1917 {
1918 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1919 }
1920 
1921 static inline void netdev_for_each_tx_queue(struct net_device *dev,
1922 					    void (*f)(struct net_device *,
1923 						      struct netdev_queue *,
1924 						      void *),
1925 					    void *arg)
1926 {
1927 	unsigned int i;
1928 
1929 	for (i = 0; i < dev->num_tx_queues; i++)
1930 		f(dev, &dev->_tx[i], arg);
1931 }
1932 
1933 #define netdev_lockdep_set_classes(dev)				\
1934 {								\
1935 	static struct lock_class_key qdisc_tx_busylock_key;	\
1936 	static struct lock_class_key qdisc_running_key;		\
1937 	static struct lock_class_key qdisc_xmit_lock_key;	\
1938 	static struct lock_class_key dev_addr_list_lock_key;	\
1939 	unsigned int i;						\
1940 								\
1941 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
1942 	(dev)->qdisc_running_key = &qdisc_running_key;		\
1943 	lockdep_set_class(&(dev)->addr_list_lock,		\
1944 			  &dev_addr_list_lock_key); 		\
1945 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
1946 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
1947 				  &qdisc_xmit_lock_key);	\
1948 }
1949 
1950 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
1951 				    struct sk_buff *skb,
1952 				    void *accel_priv);
1953 
1954 /* returns the headroom that the master device needs to take in account
1955  * when forwarding to this dev
1956  */
1957 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1958 {
1959 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1960 }
1961 
1962 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1963 {
1964 	if (dev->netdev_ops->ndo_set_rx_headroom)
1965 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1966 }
1967 
1968 /* set the device rx headroom to the dev's default */
1969 static inline void netdev_reset_rx_headroom(struct net_device *dev)
1970 {
1971 	netdev_set_rx_headroom(dev, -1);
1972 }
1973 
1974 /*
1975  * Net namespace inlines
1976  */
1977 static inline
1978 struct net *dev_net(const struct net_device *dev)
1979 {
1980 	return read_pnet(&dev->nd_net);
1981 }
1982 
1983 static inline
1984 void dev_net_set(struct net_device *dev, struct net *net)
1985 {
1986 	write_pnet(&dev->nd_net, net);
1987 }
1988 
1989 static inline bool netdev_uses_dsa(struct net_device *dev)
1990 {
1991 #if IS_ENABLED(CONFIG_NET_DSA)
1992 	if (dev->dsa_ptr != NULL)
1993 		return dsa_uses_tagged_protocol(dev->dsa_ptr);
1994 #endif
1995 	return false;
1996 }
1997 
1998 /**
1999  *	netdev_priv - access network device private data
2000  *	@dev: network device
2001  *
2002  * Get network device private data
2003  */
2004 static inline void *netdev_priv(const struct net_device *dev)
2005 {
2006 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2007 }
2008 
2009 /* Set the sysfs physical device reference for the network logical device
2010  * if set prior to registration will cause a symlink during initialization.
2011  */
2012 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2013 
2014 /* Set the sysfs device type for the network logical device to allow
2015  * fine-grained identification of different network device types. For
2016  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2017  */
2018 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2019 
2020 /* Default NAPI poll() weight
2021  * Device drivers are strongly advised to not use bigger value
2022  */
2023 #define NAPI_POLL_WEIGHT 64
2024 
2025 /**
2026  *	netif_napi_add - initialize a NAPI context
2027  *	@dev:  network device
2028  *	@napi: NAPI context
2029  *	@poll: polling function
2030  *	@weight: default weight
2031  *
2032  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2033  * *any* of the other NAPI-related functions.
2034  */
2035 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2036 		    int (*poll)(struct napi_struct *, int), int weight);
2037 
2038 /**
2039  *	netif_tx_napi_add - initialize a NAPI context
2040  *	@dev:  network device
2041  *	@napi: NAPI context
2042  *	@poll: polling function
2043  *	@weight: default weight
2044  *
2045  * This variant of netif_napi_add() should be used from drivers using NAPI
2046  * to exclusively poll a TX queue.
2047  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2048  */
2049 static inline void netif_tx_napi_add(struct net_device *dev,
2050 				     struct napi_struct *napi,
2051 				     int (*poll)(struct napi_struct *, int),
2052 				     int weight)
2053 {
2054 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2055 	netif_napi_add(dev, napi, poll, weight);
2056 }
2057 
2058 /**
2059  *  netif_napi_del - remove a NAPI context
2060  *  @napi: NAPI context
2061  *
2062  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2063  */
2064 void netif_napi_del(struct napi_struct *napi);
2065 
2066 struct napi_gro_cb {
2067 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2068 	void	*frag0;
2069 
2070 	/* Length of frag0. */
2071 	unsigned int frag0_len;
2072 
2073 	/* This indicates where we are processing relative to skb->data. */
2074 	int	data_offset;
2075 
2076 	/* This is non-zero if the packet cannot be merged with the new skb. */
2077 	u16	flush;
2078 
2079 	/* Save the IP ID here and check when we get to the transport layer */
2080 	u16	flush_id;
2081 
2082 	/* Number of segments aggregated. */
2083 	u16	count;
2084 
2085 	/* Start offset for remote checksum offload */
2086 	u16	gro_remcsum_start;
2087 
2088 	/* jiffies when first packet was created/queued */
2089 	unsigned long age;
2090 
2091 	/* Used in ipv6_gro_receive() and foo-over-udp */
2092 	u16	proto;
2093 
2094 	/* This is non-zero if the packet may be of the same flow. */
2095 	u8	same_flow:1;
2096 
2097 	/* Used in tunnel GRO receive */
2098 	u8	encap_mark:1;
2099 
2100 	/* GRO checksum is valid */
2101 	u8	csum_valid:1;
2102 
2103 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2104 	u8	csum_cnt:3;
2105 
2106 	/* Free the skb? */
2107 	u8	free:2;
2108 #define NAPI_GRO_FREE		  1
2109 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2110 
2111 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2112 	u8	is_ipv6:1;
2113 
2114 	/* Used in GRE, set in fou/gue_gro_receive */
2115 	u8	is_fou:1;
2116 
2117 	/* Used to determine if flush_id can be ignored */
2118 	u8	is_atomic:1;
2119 
2120 	/* 5 bit hole */
2121 
2122 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2123 	__wsum	csum;
2124 
2125 	/* used in skb_gro_receive() slow path */
2126 	struct sk_buff *last;
2127 };
2128 
2129 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2130 
2131 struct packet_type {
2132 	__be16			type;	/* This is really htons(ether_type). */
2133 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2134 	int			(*func) (struct sk_buff *,
2135 					 struct net_device *,
2136 					 struct packet_type *,
2137 					 struct net_device *);
2138 	bool			(*id_match)(struct packet_type *ptype,
2139 					    struct sock *sk);
2140 	void			*af_packet_priv;
2141 	struct list_head	list;
2142 };
2143 
2144 struct offload_callbacks {
2145 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2146 						netdev_features_t features);
2147 	struct sk_buff		**(*gro_receive)(struct sk_buff **head,
2148 						 struct sk_buff *skb);
2149 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2150 };
2151 
2152 struct packet_offload {
2153 	__be16			 type;	/* This is really htons(ether_type). */
2154 	u16			 priority;
2155 	struct offload_callbacks callbacks;
2156 	struct list_head	 list;
2157 };
2158 
2159 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2160 struct pcpu_sw_netstats {
2161 	u64     rx_packets;
2162 	u64     rx_bytes;
2163 	u64     tx_packets;
2164 	u64     tx_bytes;
2165 	struct u64_stats_sync   syncp;
2166 };
2167 
2168 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2169 ({									\
2170 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2171 	if (pcpu_stats)	{						\
2172 		int __cpu;						\
2173 		for_each_possible_cpu(__cpu) {				\
2174 			typeof(type) *stat;				\
2175 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2176 			u64_stats_init(&stat->syncp);			\
2177 		}							\
2178 	}								\
2179 	pcpu_stats;							\
2180 })
2181 
2182 #define netdev_alloc_pcpu_stats(type)					\
2183 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2184 
2185 enum netdev_lag_tx_type {
2186 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2187 	NETDEV_LAG_TX_TYPE_RANDOM,
2188 	NETDEV_LAG_TX_TYPE_BROADCAST,
2189 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2190 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2191 	NETDEV_LAG_TX_TYPE_HASH,
2192 };
2193 
2194 struct netdev_lag_upper_info {
2195 	enum netdev_lag_tx_type tx_type;
2196 };
2197 
2198 struct netdev_lag_lower_state_info {
2199 	u8 link_up : 1,
2200 	   tx_enabled : 1;
2201 };
2202 
2203 #include <linux/notifier.h>
2204 
2205 /* netdevice notifier chain. Please remember to update the rtnetlink
2206  * notification exclusion list in rtnetlink_event() when adding new
2207  * types.
2208  */
2209 #define NETDEV_UP	0x0001	/* For now you can't veto a device up/down */
2210 #define NETDEV_DOWN	0x0002
2211 #define NETDEV_REBOOT	0x0003	/* Tell a protocol stack a network interface
2212 				   detected a hardware crash and restarted
2213 				   - we can use this eg to kick tcp sessions
2214 				   once done */
2215 #define NETDEV_CHANGE	0x0004	/* Notify device state change */
2216 #define NETDEV_REGISTER 0x0005
2217 #define NETDEV_UNREGISTER	0x0006
2218 #define NETDEV_CHANGEMTU	0x0007 /* notify after mtu change happened */
2219 #define NETDEV_CHANGEADDR	0x0008
2220 #define NETDEV_GOING_DOWN	0x0009
2221 #define NETDEV_CHANGENAME	0x000A
2222 #define NETDEV_FEAT_CHANGE	0x000B
2223 #define NETDEV_BONDING_FAILOVER 0x000C
2224 #define NETDEV_PRE_UP		0x000D
2225 #define NETDEV_PRE_TYPE_CHANGE	0x000E
2226 #define NETDEV_POST_TYPE_CHANGE	0x000F
2227 #define NETDEV_POST_INIT	0x0010
2228 #define NETDEV_UNREGISTER_FINAL 0x0011
2229 #define NETDEV_RELEASE		0x0012
2230 #define NETDEV_NOTIFY_PEERS	0x0013
2231 #define NETDEV_JOIN		0x0014
2232 #define NETDEV_CHANGEUPPER	0x0015
2233 #define NETDEV_RESEND_IGMP	0x0016
2234 #define NETDEV_PRECHANGEMTU	0x0017 /* notify before mtu change happened */
2235 #define NETDEV_CHANGEINFODATA	0x0018
2236 #define NETDEV_BONDING_INFO	0x0019
2237 #define NETDEV_PRECHANGEUPPER	0x001A
2238 #define NETDEV_CHANGELOWERSTATE	0x001B
2239 #define NETDEV_UDP_TUNNEL_PUSH_INFO	0x001C
2240 
2241 int register_netdevice_notifier(struct notifier_block *nb);
2242 int unregister_netdevice_notifier(struct notifier_block *nb);
2243 
2244 struct netdev_notifier_info {
2245 	struct net_device *dev;
2246 };
2247 
2248 struct netdev_notifier_change_info {
2249 	struct netdev_notifier_info info; /* must be first */
2250 	unsigned int flags_changed;
2251 };
2252 
2253 struct netdev_notifier_changeupper_info {
2254 	struct netdev_notifier_info info; /* must be first */
2255 	struct net_device *upper_dev; /* new upper dev */
2256 	bool master; /* is upper dev master */
2257 	bool linking; /* is the notification for link or unlink */
2258 	void *upper_info; /* upper dev info */
2259 };
2260 
2261 struct netdev_notifier_changelowerstate_info {
2262 	struct netdev_notifier_info info; /* must be first */
2263 	void *lower_state_info; /* is lower dev state */
2264 };
2265 
2266 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2267 					     struct net_device *dev)
2268 {
2269 	info->dev = dev;
2270 }
2271 
2272 static inline struct net_device *
2273 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2274 {
2275 	return info->dev;
2276 }
2277 
2278 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2279 
2280 
2281 extern rwlock_t				dev_base_lock;		/* Device list lock */
2282 
2283 #define for_each_netdev(net, d)		\
2284 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2285 #define for_each_netdev_reverse(net, d)	\
2286 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2287 #define for_each_netdev_rcu(net, d)		\
2288 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2289 #define for_each_netdev_safe(net, d, n)	\
2290 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2291 #define for_each_netdev_continue(net, d)		\
2292 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2293 #define for_each_netdev_continue_rcu(net, d)		\
2294 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2295 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2296 		for_each_netdev_rcu(&init_net, slave)	\
2297 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2298 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2299 
2300 static inline struct net_device *next_net_device(struct net_device *dev)
2301 {
2302 	struct list_head *lh;
2303 	struct net *net;
2304 
2305 	net = dev_net(dev);
2306 	lh = dev->dev_list.next;
2307 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2308 }
2309 
2310 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2311 {
2312 	struct list_head *lh;
2313 	struct net *net;
2314 
2315 	net = dev_net(dev);
2316 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2317 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2318 }
2319 
2320 static inline struct net_device *first_net_device(struct net *net)
2321 {
2322 	return list_empty(&net->dev_base_head) ? NULL :
2323 		net_device_entry(net->dev_base_head.next);
2324 }
2325 
2326 static inline struct net_device *first_net_device_rcu(struct net *net)
2327 {
2328 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2329 
2330 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2331 }
2332 
2333 int netdev_boot_setup_check(struct net_device *dev);
2334 unsigned long netdev_boot_base(const char *prefix, int unit);
2335 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2336 				       const char *hwaddr);
2337 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2338 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2339 void dev_add_pack(struct packet_type *pt);
2340 void dev_remove_pack(struct packet_type *pt);
2341 void __dev_remove_pack(struct packet_type *pt);
2342 void dev_add_offload(struct packet_offload *po);
2343 void dev_remove_offload(struct packet_offload *po);
2344 
2345 int dev_get_iflink(const struct net_device *dev);
2346 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2347 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2348 				      unsigned short mask);
2349 struct net_device *dev_get_by_name(struct net *net, const char *name);
2350 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2351 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2352 int dev_alloc_name(struct net_device *dev, const char *name);
2353 int dev_open(struct net_device *dev);
2354 int dev_close(struct net_device *dev);
2355 int dev_close_many(struct list_head *head, bool unlink);
2356 void dev_disable_lro(struct net_device *dev);
2357 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2358 int dev_queue_xmit(struct sk_buff *skb);
2359 int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
2360 int register_netdevice(struct net_device *dev);
2361 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2362 void unregister_netdevice_many(struct list_head *head);
2363 static inline void unregister_netdevice(struct net_device *dev)
2364 {
2365 	unregister_netdevice_queue(dev, NULL);
2366 }
2367 
2368 int netdev_refcnt_read(const struct net_device *dev);
2369 void free_netdev(struct net_device *dev);
2370 void netdev_freemem(struct net_device *dev);
2371 void synchronize_net(void);
2372 int init_dummy_netdev(struct net_device *dev);
2373 
2374 DECLARE_PER_CPU(int, xmit_recursion);
2375 #define XMIT_RECURSION_LIMIT	10
2376 
2377 static inline int dev_recursion_level(void)
2378 {
2379 	return this_cpu_read(xmit_recursion);
2380 }
2381 
2382 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2383 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2384 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2385 int netdev_get_name(struct net *net, char *name, int ifindex);
2386 int dev_restart(struct net_device *dev);
2387 int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
2388 
2389 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2390 {
2391 	return NAPI_GRO_CB(skb)->data_offset;
2392 }
2393 
2394 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2395 {
2396 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2397 }
2398 
2399 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2400 {
2401 	NAPI_GRO_CB(skb)->data_offset += len;
2402 }
2403 
2404 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2405 					unsigned int offset)
2406 {
2407 	return NAPI_GRO_CB(skb)->frag0 + offset;
2408 }
2409 
2410 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2411 {
2412 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2413 }
2414 
2415 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2416 					unsigned int offset)
2417 {
2418 	if (!pskb_may_pull(skb, hlen))
2419 		return NULL;
2420 
2421 	NAPI_GRO_CB(skb)->frag0 = NULL;
2422 	NAPI_GRO_CB(skb)->frag0_len = 0;
2423 	return skb->data + offset;
2424 }
2425 
2426 static inline void *skb_gro_network_header(struct sk_buff *skb)
2427 {
2428 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2429 	       skb_network_offset(skb);
2430 }
2431 
2432 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2433 					const void *start, unsigned int len)
2434 {
2435 	if (NAPI_GRO_CB(skb)->csum_valid)
2436 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2437 						  csum_partial(start, len, 0));
2438 }
2439 
2440 /* GRO checksum functions. These are logical equivalents of the normal
2441  * checksum functions (in skbuff.h) except that they operate on the GRO
2442  * offsets and fields in sk_buff.
2443  */
2444 
2445 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2446 
2447 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2448 {
2449 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2450 }
2451 
2452 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2453 						      bool zero_okay,
2454 						      __sum16 check)
2455 {
2456 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2457 		skb_checksum_start_offset(skb) <
2458 		 skb_gro_offset(skb)) &&
2459 		!skb_at_gro_remcsum_start(skb) &&
2460 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2461 		(!zero_okay || check));
2462 }
2463 
2464 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2465 							   __wsum psum)
2466 {
2467 	if (NAPI_GRO_CB(skb)->csum_valid &&
2468 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2469 		return 0;
2470 
2471 	NAPI_GRO_CB(skb)->csum = psum;
2472 
2473 	return __skb_gro_checksum_complete(skb);
2474 }
2475 
2476 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2477 {
2478 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2479 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2480 		NAPI_GRO_CB(skb)->csum_cnt--;
2481 	} else {
2482 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2483 		 * verified a new top level checksum or an encapsulated one
2484 		 * during GRO. This saves work if we fallback to normal path.
2485 		 */
2486 		__skb_incr_checksum_unnecessary(skb);
2487 	}
2488 }
2489 
2490 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2491 				    compute_pseudo)			\
2492 ({									\
2493 	__sum16 __ret = 0;						\
2494 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2495 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2496 				compute_pseudo(skb, proto));		\
2497 	if (__ret)							\
2498 		__skb_mark_checksum_bad(skb);				\
2499 	else								\
2500 		skb_gro_incr_csum_unnecessary(skb);			\
2501 	__ret;								\
2502 })
2503 
2504 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2505 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2506 
2507 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2508 					     compute_pseudo)		\
2509 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2510 
2511 #define skb_gro_checksum_simple_validate(skb)				\
2512 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2513 
2514 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2515 {
2516 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2517 		!NAPI_GRO_CB(skb)->csum_valid);
2518 }
2519 
2520 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2521 					      __sum16 check, __wsum pseudo)
2522 {
2523 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2524 	NAPI_GRO_CB(skb)->csum_valid = 1;
2525 }
2526 
2527 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2528 do {									\
2529 	if (__skb_gro_checksum_convert_check(skb))			\
2530 		__skb_gro_checksum_convert(skb, check,			\
2531 					   compute_pseudo(skb, proto));	\
2532 } while (0)
2533 
2534 struct gro_remcsum {
2535 	int offset;
2536 	__wsum delta;
2537 };
2538 
2539 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2540 {
2541 	grc->offset = 0;
2542 	grc->delta = 0;
2543 }
2544 
2545 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2546 					    unsigned int off, size_t hdrlen,
2547 					    int start, int offset,
2548 					    struct gro_remcsum *grc,
2549 					    bool nopartial)
2550 {
2551 	__wsum delta;
2552 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2553 
2554 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2555 
2556 	if (!nopartial) {
2557 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2558 		return ptr;
2559 	}
2560 
2561 	ptr = skb_gro_header_fast(skb, off);
2562 	if (skb_gro_header_hard(skb, off + plen)) {
2563 		ptr = skb_gro_header_slow(skb, off + plen, off);
2564 		if (!ptr)
2565 			return NULL;
2566 	}
2567 
2568 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2569 			       start, offset);
2570 
2571 	/* Adjust skb->csum since we changed the packet */
2572 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2573 
2574 	grc->offset = off + hdrlen + offset;
2575 	grc->delta = delta;
2576 
2577 	return ptr;
2578 }
2579 
2580 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2581 					   struct gro_remcsum *grc)
2582 {
2583 	void *ptr;
2584 	size_t plen = grc->offset + sizeof(u16);
2585 
2586 	if (!grc->delta)
2587 		return;
2588 
2589 	ptr = skb_gro_header_fast(skb, grc->offset);
2590 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2591 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2592 		if (!ptr)
2593 			return;
2594 	}
2595 
2596 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2597 }
2598 
2599 struct skb_csum_offl_spec {
2600 	__u16		ipv4_okay:1,
2601 			ipv6_okay:1,
2602 			encap_okay:1,
2603 			ip_options_okay:1,
2604 			ext_hdrs_okay:1,
2605 			tcp_okay:1,
2606 			udp_okay:1,
2607 			sctp_okay:1,
2608 			vlan_okay:1,
2609 			no_encapped_ipv6:1,
2610 			no_not_encapped:1;
2611 };
2612 
2613 bool __skb_csum_offload_chk(struct sk_buff *skb,
2614 			    const struct skb_csum_offl_spec *spec,
2615 			    bool *csum_encapped,
2616 			    bool csum_help);
2617 
2618 static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2619 					const struct skb_csum_offl_spec *spec,
2620 					bool *csum_encapped,
2621 					bool csum_help)
2622 {
2623 	if (skb->ip_summed != CHECKSUM_PARTIAL)
2624 		return false;
2625 
2626 	return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2627 }
2628 
2629 static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2630 					     const struct skb_csum_offl_spec *spec)
2631 {
2632 	bool csum_encapped;
2633 
2634 	return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2635 }
2636 
2637 static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2638 {
2639 	static const struct skb_csum_offl_spec csum_offl_spec = {
2640 		.ipv4_okay = 1,
2641 		.ip_options_okay = 1,
2642 		.ipv6_okay = 1,
2643 		.vlan_okay = 1,
2644 		.tcp_okay = 1,
2645 		.udp_okay = 1,
2646 	};
2647 
2648 	return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2649 }
2650 
2651 static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2652 {
2653 	static const struct skb_csum_offl_spec csum_offl_spec = {
2654 		.ipv4_okay = 1,
2655 		.ip_options_okay = 1,
2656 		.tcp_okay = 1,
2657 		.udp_okay = 1,
2658 		.vlan_okay = 1,
2659 	};
2660 
2661 	return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2662 }
2663 
2664 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2665 				  unsigned short type,
2666 				  const void *daddr, const void *saddr,
2667 				  unsigned int len)
2668 {
2669 	if (!dev->header_ops || !dev->header_ops->create)
2670 		return 0;
2671 
2672 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2673 }
2674 
2675 static inline int dev_parse_header(const struct sk_buff *skb,
2676 				   unsigned char *haddr)
2677 {
2678 	const struct net_device *dev = skb->dev;
2679 
2680 	if (!dev->header_ops || !dev->header_ops->parse)
2681 		return 0;
2682 	return dev->header_ops->parse(skb, haddr);
2683 }
2684 
2685 /* ll_header must have at least hard_header_len allocated */
2686 static inline bool dev_validate_header(const struct net_device *dev,
2687 				       char *ll_header, int len)
2688 {
2689 	if (likely(len >= dev->hard_header_len))
2690 		return true;
2691 
2692 	if (capable(CAP_SYS_RAWIO)) {
2693 		memset(ll_header + len, 0, dev->hard_header_len - len);
2694 		return true;
2695 	}
2696 
2697 	if (dev->header_ops && dev->header_ops->validate)
2698 		return dev->header_ops->validate(ll_header, len);
2699 
2700 	return false;
2701 }
2702 
2703 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
2704 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2705 static inline int unregister_gifconf(unsigned int family)
2706 {
2707 	return register_gifconf(family, NULL);
2708 }
2709 
2710 #ifdef CONFIG_NET_FLOW_LIMIT
2711 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2712 struct sd_flow_limit {
2713 	u64			count;
2714 	unsigned int		num_buckets;
2715 	unsigned int		history_head;
2716 	u16			history[FLOW_LIMIT_HISTORY];
2717 	u8			buckets[];
2718 };
2719 
2720 extern int netdev_flow_limit_table_len;
2721 #endif /* CONFIG_NET_FLOW_LIMIT */
2722 
2723 /*
2724  * Incoming packets are placed on per-CPU queues
2725  */
2726 struct softnet_data {
2727 	struct list_head	poll_list;
2728 	struct sk_buff_head	process_queue;
2729 
2730 	/* stats */
2731 	unsigned int		processed;
2732 	unsigned int		time_squeeze;
2733 	unsigned int		received_rps;
2734 #ifdef CONFIG_RPS
2735 	struct softnet_data	*rps_ipi_list;
2736 #endif
2737 #ifdef CONFIG_NET_FLOW_LIMIT
2738 	struct sd_flow_limit __rcu *flow_limit;
2739 #endif
2740 	struct Qdisc		*output_queue;
2741 	struct Qdisc		**output_queue_tailp;
2742 	struct sk_buff		*completion_queue;
2743 
2744 #ifdef CONFIG_RPS
2745 	/* input_queue_head should be written by cpu owning this struct,
2746 	 * and only read by other cpus. Worth using a cache line.
2747 	 */
2748 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2749 
2750 	/* Elements below can be accessed between CPUs for RPS/RFS */
2751 	struct call_single_data	csd ____cacheline_aligned_in_smp;
2752 	struct softnet_data	*rps_ipi_next;
2753 	unsigned int		cpu;
2754 	unsigned int		input_queue_tail;
2755 #endif
2756 	unsigned int		dropped;
2757 	struct sk_buff_head	input_pkt_queue;
2758 	struct napi_struct	backlog;
2759 
2760 };
2761 
2762 static inline void input_queue_head_incr(struct softnet_data *sd)
2763 {
2764 #ifdef CONFIG_RPS
2765 	sd->input_queue_head++;
2766 #endif
2767 }
2768 
2769 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2770 					      unsigned int *qtail)
2771 {
2772 #ifdef CONFIG_RPS
2773 	*qtail = ++sd->input_queue_tail;
2774 #endif
2775 }
2776 
2777 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
2778 
2779 void __netif_schedule(struct Qdisc *q);
2780 void netif_schedule_queue(struct netdev_queue *txq);
2781 
2782 static inline void netif_tx_schedule_all(struct net_device *dev)
2783 {
2784 	unsigned int i;
2785 
2786 	for (i = 0; i < dev->num_tx_queues; i++)
2787 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
2788 }
2789 
2790 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2791 {
2792 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2793 }
2794 
2795 /**
2796  *	netif_start_queue - allow transmit
2797  *	@dev: network device
2798  *
2799  *	Allow upper layers to call the device hard_start_xmit routine.
2800  */
2801 static inline void netif_start_queue(struct net_device *dev)
2802 {
2803 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
2804 }
2805 
2806 static inline void netif_tx_start_all_queues(struct net_device *dev)
2807 {
2808 	unsigned int i;
2809 
2810 	for (i = 0; i < dev->num_tx_queues; i++) {
2811 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2812 		netif_tx_start_queue(txq);
2813 	}
2814 }
2815 
2816 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
2817 
2818 /**
2819  *	netif_wake_queue - restart transmit
2820  *	@dev: network device
2821  *
2822  *	Allow upper layers to call the device hard_start_xmit routine.
2823  *	Used for flow control when transmit resources are available.
2824  */
2825 static inline void netif_wake_queue(struct net_device *dev)
2826 {
2827 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
2828 }
2829 
2830 static inline void netif_tx_wake_all_queues(struct net_device *dev)
2831 {
2832 	unsigned int i;
2833 
2834 	for (i = 0; i < dev->num_tx_queues; i++) {
2835 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2836 		netif_tx_wake_queue(txq);
2837 	}
2838 }
2839 
2840 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2841 {
2842 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2843 }
2844 
2845 /**
2846  *	netif_stop_queue - stop transmitted packets
2847  *	@dev: network device
2848  *
2849  *	Stop upper layers calling the device hard_start_xmit routine.
2850  *	Used for flow control when transmit resources are unavailable.
2851  */
2852 static inline void netif_stop_queue(struct net_device *dev)
2853 {
2854 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
2855 }
2856 
2857 void netif_tx_stop_all_queues(struct net_device *dev);
2858 
2859 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
2860 {
2861 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
2862 }
2863 
2864 /**
2865  *	netif_queue_stopped - test if transmit queue is flowblocked
2866  *	@dev: network device
2867  *
2868  *	Test if transmit queue on device is currently unable to send.
2869  */
2870 static inline bool netif_queue_stopped(const struct net_device *dev)
2871 {
2872 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
2873 }
2874 
2875 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
2876 {
2877 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2878 }
2879 
2880 static inline bool
2881 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
2882 {
2883 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2884 }
2885 
2886 static inline bool
2887 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2888 {
2889 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2890 }
2891 
2892 /**
2893  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2894  *	@dev_queue: pointer to transmit queue
2895  *
2896  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2897  * to give appropriate hint to the CPU.
2898  */
2899 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2900 {
2901 #ifdef CONFIG_BQL
2902 	prefetchw(&dev_queue->dql.num_queued);
2903 #endif
2904 }
2905 
2906 /**
2907  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2908  *	@dev_queue: pointer to transmit queue
2909  *
2910  * BQL enabled drivers might use this helper in their TX completion path,
2911  * to give appropriate hint to the CPU.
2912  */
2913 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2914 {
2915 #ifdef CONFIG_BQL
2916 	prefetchw(&dev_queue->dql.limit);
2917 #endif
2918 }
2919 
2920 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2921 					unsigned int bytes)
2922 {
2923 #ifdef CONFIG_BQL
2924 	dql_queued(&dev_queue->dql, bytes);
2925 
2926 	if (likely(dql_avail(&dev_queue->dql) >= 0))
2927 		return;
2928 
2929 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2930 
2931 	/*
2932 	 * The XOFF flag must be set before checking the dql_avail below,
2933 	 * because in netdev_tx_completed_queue we update the dql_completed
2934 	 * before checking the XOFF flag.
2935 	 */
2936 	smp_mb();
2937 
2938 	/* check again in case another CPU has just made room avail */
2939 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2940 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2941 #endif
2942 }
2943 
2944 /**
2945  * 	netdev_sent_queue - report the number of bytes queued to hardware
2946  * 	@dev: network device
2947  * 	@bytes: number of bytes queued to the hardware device queue
2948  *
2949  * 	Report the number of bytes queued for sending/completion to the network
2950  * 	device hardware queue. @bytes should be a good approximation and should
2951  * 	exactly match netdev_completed_queue() @bytes
2952  */
2953 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2954 {
2955 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2956 }
2957 
2958 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
2959 					     unsigned int pkts, unsigned int bytes)
2960 {
2961 #ifdef CONFIG_BQL
2962 	if (unlikely(!bytes))
2963 		return;
2964 
2965 	dql_completed(&dev_queue->dql, bytes);
2966 
2967 	/*
2968 	 * Without the memory barrier there is a small possiblity that
2969 	 * netdev_tx_sent_queue will miss the update and cause the queue to
2970 	 * be stopped forever
2971 	 */
2972 	smp_mb();
2973 
2974 	if (dql_avail(&dev_queue->dql) < 0)
2975 		return;
2976 
2977 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2978 		netif_schedule_queue(dev_queue);
2979 #endif
2980 }
2981 
2982 /**
2983  * 	netdev_completed_queue - report bytes and packets completed by device
2984  * 	@dev: network device
2985  * 	@pkts: actual number of packets sent over the medium
2986  * 	@bytes: actual number of bytes sent over the medium
2987  *
2988  * 	Report the number of bytes and packets transmitted by the network device
2989  * 	hardware queue over the physical medium, @bytes must exactly match the
2990  * 	@bytes amount passed to netdev_sent_queue()
2991  */
2992 static inline void netdev_completed_queue(struct net_device *dev,
2993 					  unsigned int pkts, unsigned int bytes)
2994 {
2995 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
2996 }
2997 
2998 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
2999 {
3000 #ifdef CONFIG_BQL
3001 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3002 	dql_reset(&q->dql);
3003 #endif
3004 }
3005 
3006 /**
3007  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3008  * 	@dev_queue: network device
3009  *
3010  * 	Reset the bytes and packet count of a network device and clear the
3011  * 	software flow control OFF bit for this network device
3012  */
3013 static inline void netdev_reset_queue(struct net_device *dev_queue)
3014 {
3015 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3016 }
3017 
3018 /**
3019  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3020  * 	@dev: network device
3021  * 	@queue_index: given tx queue index
3022  *
3023  * 	Returns 0 if given tx queue index >= number of device tx queues,
3024  * 	otherwise returns the originally passed tx queue index.
3025  */
3026 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3027 {
3028 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3029 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3030 				     dev->name, queue_index,
3031 				     dev->real_num_tx_queues);
3032 		return 0;
3033 	}
3034 
3035 	return queue_index;
3036 }
3037 
3038 /**
3039  *	netif_running - test if up
3040  *	@dev: network device
3041  *
3042  *	Test if the device has been brought up.
3043  */
3044 static inline bool netif_running(const struct net_device *dev)
3045 {
3046 	return test_bit(__LINK_STATE_START, &dev->state);
3047 }
3048 
3049 /*
3050  * Routines to manage the subqueues on a device.  We only need start,
3051  * stop, and a check if it's stopped.  All other device management is
3052  * done at the overall netdevice level.
3053  * Also test the device if we're multiqueue.
3054  */
3055 
3056 /**
3057  *	netif_start_subqueue - allow sending packets on subqueue
3058  *	@dev: network device
3059  *	@queue_index: sub queue index
3060  *
3061  * Start individual transmit queue of a device with multiple transmit queues.
3062  */
3063 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3064 {
3065 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3066 
3067 	netif_tx_start_queue(txq);
3068 }
3069 
3070 /**
3071  *	netif_stop_subqueue - stop sending packets on subqueue
3072  *	@dev: network device
3073  *	@queue_index: sub queue index
3074  *
3075  * Stop individual transmit queue of a device with multiple transmit queues.
3076  */
3077 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3078 {
3079 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3080 	netif_tx_stop_queue(txq);
3081 }
3082 
3083 /**
3084  *	netif_subqueue_stopped - test status of subqueue
3085  *	@dev: network device
3086  *	@queue_index: sub queue index
3087  *
3088  * Check individual transmit queue of a device with multiple transmit queues.
3089  */
3090 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3091 					    u16 queue_index)
3092 {
3093 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3094 
3095 	return netif_tx_queue_stopped(txq);
3096 }
3097 
3098 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3099 					  struct sk_buff *skb)
3100 {
3101 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3102 }
3103 
3104 void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
3105 
3106 #ifdef CONFIG_XPS
3107 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3108 			u16 index);
3109 #else
3110 static inline int netif_set_xps_queue(struct net_device *dev,
3111 				      const struct cpumask *mask,
3112 				      u16 index)
3113 {
3114 	return 0;
3115 }
3116 #endif
3117 
3118 u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3119 		  unsigned int num_tx_queues);
3120 
3121 /*
3122  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3123  * as a distribution range limit for the returned value.
3124  */
3125 static inline u16 skb_tx_hash(const struct net_device *dev,
3126 			      struct sk_buff *skb)
3127 {
3128 	return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3129 }
3130 
3131 /**
3132  *	netif_is_multiqueue - test if device has multiple transmit queues
3133  *	@dev: network device
3134  *
3135  * Check if device has multiple transmit queues
3136  */
3137 static inline bool netif_is_multiqueue(const struct net_device *dev)
3138 {
3139 	return dev->num_tx_queues > 1;
3140 }
3141 
3142 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3143 
3144 #ifdef CONFIG_SYSFS
3145 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3146 #else
3147 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3148 						unsigned int rxq)
3149 {
3150 	return 0;
3151 }
3152 #endif
3153 
3154 #ifdef CONFIG_SYSFS
3155 static inline unsigned int get_netdev_rx_queue_index(
3156 		struct netdev_rx_queue *queue)
3157 {
3158 	struct net_device *dev = queue->dev;
3159 	int index = queue - dev->_rx;
3160 
3161 	BUG_ON(index >= dev->num_rx_queues);
3162 	return index;
3163 }
3164 #endif
3165 
3166 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3167 int netif_get_num_default_rss_queues(void);
3168 
3169 enum skb_free_reason {
3170 	SKB_REASON_CONSUMED,
3171 	SKB_REASON_DROPPED,
3172 };
3173 
3174 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3175 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3176 
3177 /*
3178  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3179  * interrupt context or with hardware interrupts being disabled.
3180  * (in_irq() || irqs_disabled())
3181  *
3182  * We provide four helpers that can be used in following contexts :
3183  *
3184  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3185  *  replacing kfree_skb(skb)
3186  *
3187  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3188  *  Typically used in place of consume_skb(skb) in TX completion path
3189  *
3190  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3191  *  replacing kfree_skb(skb)
3192  *
3193  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3194  *  and consumed a packet. Used in place of consume_skb(skb)
3195  */
3196 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3197 {
3198 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3199 }
3200 
3201 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3202 {
3203 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3204 }
3205 
3206 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3207 {
3208 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3209 }
3210 
3211 static inline void dev_consume_skb_any(struct sk_buff *skb)
3212 {
3213 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3214 }
3215 
3216 int netif_rx(struct sk_buff *skb);
3217 int netif_rx_ni(struct sk_buff *skb);
3218 int netif_receive_skb(struct sk_buff *skb);
3219 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3220 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3221 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3222 gro_result_t napi_gro_frags(struct napi_struct *napi);
3223 struct packet_offload *gro_find_receive_by_type(__be16 type);
3224 struct packet_offload *gro_find_complete_by_type(__be16 type);
3225 
3226 static inline void napi_free_frags(struct napi_struct *napi)
3227 {
3228 	kfree_skb(napi->skb);
3229 	napi->skb = NULL;
3230 }
3231 
3232 int netdev_rx_handler_register(struct net_device *dev,
3233 			       rx_handler_func_t *rx_handler,
3234 			       void *rx_handler_data);
3235 void netdev_rx_handler_unregister(struct net_device *dev);
3236 
3237 bool dev_valid_name(const char *name);
3238 int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3239 int dev_ethtool(struct net *net, struct ifreq *);
3240 unsigned int dev_get_flags(const struct net_device *);
3241 int __dev_change_flags(struct net_device *, unsigned int flags);
3242 int dev_change_flags(struct net_device *, unsigned int);
3243 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3244 			unsigned int gchanges);
3245 int dev_change_name(struct net_device *, const char *);
3246 int dev_set_alias(struct net_device *, const char *, size_t);
3247 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3248 int dev_set_mtu(struct net_device *, int);
3249 void dev_set_group(struct net_device *, int);
3250 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3251 int dev_change_carrier(struct net_device *, bool new_carrier);
3252 int dev_get_phys_port_id(struct net_device *dev,
3253 			 struct netdev_phys_item_id *ppid);
3254 int dev_get_phys_port_name(struct net_device *dev,
3255 			   char *name, size_t len);
3256 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3257 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
3258 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3259 				    struct netdev_queue *txq, int *ret);
3260 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3261 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3262 bool is_skb_forwardable(const struct net_device *dev,
3263 			const struct sk_buff *skb);
3264 
3265 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3266 
3267 extern int		netdev_budget;
3268 
3269 /* Called by rtnetlink.c:rtnl_unlock() */
3270 void netdev_run_todo(void);
3271 
3272 /**
3273  *	dev_put - release reference to device
3274  *	@dev: network device
3275  *
3276  * Release reference to device to allow it to be freed.
3277  */
3278 static inline void dev_put(struct net_device *dev)
3279 {
3280 	this_cpu_dec(*dev->pcpu_refcnt);
3281 }
3282 
3283 /**
3284  *	dev_hold - get reference to device
3285  *	@dev: network device
3286  *
3287  * Hold reference to device to keep it from being freed.
3288  */
3289 static inline void dev_hold(struct net_device *dev)
3290 {
3291 	this_cpu_inc(*dev->pcpu_refcnt);
3292 }
3293 
3294 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3295  * and _off may be called from IRQ context, but it is caller
3296  * who is responsible for serialization of these calls.
3297  *
3298  * The name carrier is inappropriate, these functions should really be
3299  * called netif_lowerlayer_*() because they represent the state of any
3300  * kind of lower layer not just hardware media.
3301  */
3302 
3303 void linkwatch_init_dev(struct net_device *dev);
3304 void linkwatch_fire_event(struct net_device *dev);
3305 void linkwatch_forget_dev(struct net_device *dev);
3306 
3307 /**
3308  *	netif_carrier_ok - test if carrier present
3309  *	@dev: network device
3310  *
3311  * Check if carrier is present on device
3312  */
3313 static inline bool netif_carrier_ok(const struct net_device *dev)
3314 {
3315 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3316 }
3317 
3318 unsigned long dev_trans_start(struct net_device *dev);
3319 
3320 void __netdev_watchdog_up(struct net_device *dev);
3321 
3322 void netif_carrier_on(struct net_device *dev);
3323 
3324 void netif_carrier_off(struct net_device *dev);
3325 
3326 /**
3327  *	netif_dormant_on - mark device as dormant.
3328  *	@dev: network device
3329  *
3330  * Mark device as dormant (as per RFC2863).
3331  *
3332  * The dormant state indicates that the relevant interface is not
3333  * actually in a condition to pass packets (i.e., it is not 'up') but is
3334  * in a "pending" state, waiting for some external event.  For "on-
3335  * demand" interfaces, this new state identifies the situation where the
3336  * interface is waiting for events to place it in the up state.
3337  */
3338 static inline void netif_dormant_on(struct net_device *dev)
3339 {
3340 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3341 		linkwatch_fire_event(dev);
3342 }
3343 
3344 /**
3345  *	netif_dormant_off - set device as not dormant.
3346  *	@dev: network device
3347  *
3348  * Device is not in dormant state.
3349  */
3350 static inline void netif_dormant_off(struct net_device *dev)
3351 {
3352 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3353 		linkwatch_fire_event(dev);
3354 }
3355 
3356 /**
3357  *	netif_dormant - test if carrier present
3358  *	@dev: network device
3359  *
3360  * Check if carrier is present on device
3361  */
3362 static inline bool netif_dormant(const struct net_device *dev)
3363 {
3364 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3365 }
3366 
3367 
3368 /**
3369  *	netif_oper_up - test if device is operational
3370  *	@dev: network device
3371  *
3372  * Check if carrier is operational
3373  */
3374 static inline bool netif_oper_up(const struct net_device *dev)
3375 {
3376 	return (dev->operstate == IF_OPER_UP ||
3377 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3378 }
3379 
3380 /**
3381  *	netif_device_present - is device available or removed
3382  *	@dev: network device
3383  *
3384  * Check if device has not been removed from system.
3385  */
3386 static inline bool netif_device_present(struct net_device *dev)
3387 {
3388 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3389 }
3390 
3391 void netif_device_detach(struct net_device *dev);
3392 
3393 void netif_device_attach(struct net_device *dev);
3394 
3395 /*
3396  * Network interface message level settings
3397  */
3398 
3399 enum {
3400 	NETIF_MSG_DRV		= 0x0001,
3401 	NETIF_MSG_PROBE		= 0x0002,
3402 	NETIF_MSG_LINK		= 0x0004,
3403 	NETIF_MSG_TIMER		= 0x0008,
3404 	NETIF_MSG_IFDOWN	= 0x0010,
3405 	NETIF_MSG_IFUP		= 0x0020,
3406 	NETIF_MSG_RX_ERR	= 0x0040,
3407 	NETIF_MSG_TX_ERR	= 0x0080,
3408 	NETIF_MSG_TX_QUEUED	= 0x0100,
3409 	NETIF_MSG_INTR		= 0x0200,
3410 	NETIF_MSG_TX_DONE	= 0x0400,
3411 	NETIF_MSG_RX_STATUS	= 0x0800,
3412 	NETIF_MSG_PKTDATA	= 0x1000,
3413 	NETIF_MSG_HW		= 0x2000,
3414 	NETIF_MSG_WOL		= 0x4000,
3415 };
3416 
3417 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3418 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3419 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3420 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3421 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3422 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3423 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3424 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3425 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3426 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3427 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3428 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3429 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3430 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3431 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3432 
3433 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3434 {
3435 	/* use default */
3436 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3437 		return default_msg_enable_bits;
3438 	if (debug_value == 0)	/* no output */
3439 		return 0;
3440 	/* set low N bits */
3441 	return (1 << debug_value) - 1;
3442 }
3443 
3444 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3445 {
3446 	spin_lock(&txq->_xmit_lock);
3447 	txq->xmit_lock_owner = cpu;
3448 }
3449 
3450 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3451 {
3452 	spin_lock_bh(&txq->_xmit_lock);
3453 	txq->xmit_lock_owner = smp_processor_id();
3454 }
3455 
3456 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3457 {
3458 	bool ok = spin_trylock(&txq->_xmit_lock);
3459 	if (likely(ok))
3460 		txq->xmit_lock_owner = smp_processor_id();
3461 	return ok;
3462 }
3463 
3464 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3465 {
3466 	txq->xmit_lock_owner = -1;
3467 	spin_unlock(&txq->_xmit_lock);
3468 }
3469 
3470 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3471 {
3472 	txq->xmit_lock_owner = -1;
3473 	spin_unlock_bh(&txq->_xmit_lock);
3474 }
3475 
3476 static inline void txq_trans_update(struct netdev_queue *txq)
3477 {
3478 	if (txq->xmit_lock_owner != -1)
3479 		txq->trans_start = jiffies;
3480 }
3481 
3482 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3483 static inline void netif_trans_update(struct net_device *dev)
3484 {
3485 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3486 
3487 	if (txq->trans_start != jiffies)
3488 		txq->trans_start = jiffies;
3489 }
3490 
3491 /**
3492  *	netif_tx_lock - grab network device transmit lock
3493  *	@dev: network device
3494  *
3495  * Get network device transmit lock
3496  */
3497 static inline void netif_tx_lock(struct net_device *dev)
3498 {
3499 	unsigned int i;
3500 	int cpu;
3501 
3502 	spin_lock(&dev->tx_global_lock);
3503 	cpu = smp_processor_id();
3504 	for (i = 0; i < dev->num_tx_queues; i++) {
3505 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3506 
3507 		/* We are the only thread of execution doing a
3508 		 * freeze, but we have to grab the _xmit_lock in
3509 		 * order to synchronize with threads which are in
3510 		 * the ->hard_start_xmit() handler and already
3511 		 * checked the frozen bit.
3512 		 */
3513 		__netif_tx_lock(txq, cpu);
3514 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3515 		__netif_tx_unlock(txq);
3516 	}
3517 }
3518 
3519 static inline void netif_tx_lock_bh(struct net_device *dev)
3520 {
3521 	local_bh_disable();
3522 	netif_tx_lock(dev);
3523 }
3524 
3525 static inline void netif_tx_unlock(struct net_device *dev)
3526 {
3527 	unsigned int i;
3528 
3529 	for (i = 0; i < dev->num_tx_queues; i++) {
3530 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3531 
3532 		/* No need to grab the _xmit_lock here.  If the
3533 		 * queue is not stopped for another reason, we
3534 		 * force a schedule.
3535 		 */
3536 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3537 		netif_schedule_queue(txq);
3538 	}
3539 	spin_unlock(&dev->tx_global_lock);
3540 }
3541 
3542 static inline void netif_tx_unlock_bh(struct net_device *dev)
3543 {
3544 	netif_tx_unlock(dev);
3545 	local_bh_enable();
3546 }
3547 
3548 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3549 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3550 		__netif_tx_lock(txq, cpu);		\
3551 	}						\
3552 }
3553 
3554 #define HARD_TX_TRYLOCK(dev, txq)			\
3555 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3556 		__netif_tx_trylock(txq) :		\
3557 		true )
3558 
3559 #define HARD_TX_UNLOCK(dev, txq) {			\
3560 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3561 		__netif_tx_unlock(txq);			\
3562 	}						\
3563 }
3564 
3565 static inline void netif_tx_disable(struct net_device *dev)
3566 {
3567 	unsigned int i;
3568 	int cpu;
3569 
3570 	local_bh_disable();
3571 	cpu = smp_processor_id();
3572 	for (i = 0; i < dev->num_tx_queues; i++) {
3573 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3574 
3575 		__netif_tx_lock(txq, cpu);
3576 		netif_tx_stop_queue(txq);
3577 		__netif_tx_unlock(txq);
3578 	}
3579 	local_bh_enable();
3580 }
3581 
3582 static inline void netif_addr_lock(struct net_device *dev)
3583 {
3584 	spin_lock(&dev->addr_list_lock);
3585 }
3586 
3587 static inline void netif_addr_lock_nested(struct net_device *dev)
3588 {
3589 	int subclass = SINGLE_DEPTH_NESTING;
3590 
3591 	if (dev->netdev_ops->ndo_get_lock_subclass)
3592 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3593 
3594 	spin_lock_nested(&dev->addr_list_lock, subclass);
3595 }
3596 
3597 static inline void netif_addr_lock_bh(struct net_device *dev)
3598 {
3599 	spin_lock_bh(&dev->addr_list_lock);
3600 }
3601 
3602 static inline void netif_addr_unlock(struct net_device *dev)
3603 {
3604 	spin_unlock(&dev->addr_list_lock);
3605 }
3606 
3607 static inline void netif_addr_unlock_bh(struct net_device *dev)
3608 {
3609 	spin_unlock_bh(&dev->addr_list_lock);
3610 }
3611 
3612 /*
3613  * dev_addrs walker. Should be used only for read access. Call with
3614  * rcu_read_lock held.
3615  */
3616 #define for_each_dev_addr(dev, ha) \
3617 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
3618 
3619 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
3620 
3621 void ether_setup(struct net_device *dev);
3622 
3623 /* Support for loadable net-drivers */
3624 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
3625 				    unsigned char name_assign_type,
3626 				    void (*setup)(struct net_device *),
3627 				    unsigned int txqs, unsigned int rxqs);
3628 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3629 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
3630 
3631 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3632 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3633 			 count)
3634 
3635 int register_netdev(struct net_device *dev);
3636 void unregister_netdev(struct net_device *dev);
3637 
3638 /* General hardware address lists handling functions */
3639 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3640 		   struct netdev_hw_addr_list *from_list, int addr_len);
3641 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3642 		      struct netdev_hw_addr_list *from_list, int addr_len);
3643 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3644 		       struct net_device *dev,
3645 		       int (*sync)(struct net_device *, const unsigned char *),
3646 		       int (*unsync)(struct net_device *,
3647 				     const unsigned char *));
3648 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3649 			  struct net_device *dev,
3650 			  int (*unsync)(struct net_device *,
3651 					const unsigned char *));
3652 void __hw_addr_init(struct netdev_hw_addr_list *list);
3653 
3654 /* Functions used for device addresses handling */
3655 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3656 		 unsigned char addr_type);
3657 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3658 		 unsigned char addr_type);
3659 void dev_addr_flush(struct net_device *dev);
3660 int dev_addr_init(struct net_device *dev);
3661 
3662 /* Functions used for unicast addresses handling */
3663 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3664 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3665 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3666 int dev_uc_sync(struct net_device *to, struct net_device *from);
3667 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3668 void dev_uc_unsync(struct net_device *to, struct net_device *from);
3669 void dev_uc_flush(struct net_device *dev);
3670 void dev_uc_init(struct net_device *dev);
3671 
3672 /**
3673  *  __dev_uc_sync - Synchonize device's unicast list
3674  *  @dev:  device to sync
3675  *  @sync: function to call if address should be added
3676  *  @unsync: function to call if address should be removed
3677  *
3678  *  Add newly added addresses to the interface, and release
3679  *  addresses that have been deleted.
3680  */
3681 static inline int __dev_uc_sync(struct net_device *dev,
3682 				int (*sync)(struct net_device *,
3683 					    const unsigned char *),
3684 				int (*unsync)(struct net_device *,
3685 					      const unsigned char *))
3686 {
3687 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3688 }
3689 
3690 /**
3691  *  __dev_uc_unsync - Remove synchronized addresses from device
3692  *  @dev:  device to sync
3693  *  @unsync: function to call if address should be removed
3694  *
3695  *  Remove all addresses that were added to the device by dev_uc_sync().
3696  */
3697 static inline void __dev_uc_unsync(struct net_device *dev,
3698 				   int (*unsync)(struct net_device *,
3699 						 const unsigned char *))
3700 {
3701 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
3702 }
3703 
3704 /* Functions used for multicast addresses handling */
3705 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3706 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3707 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3708 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3709 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3710 int dev_mc_sync(struct net_device *to, struct net_device *from);
3711 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3712 void dev_mc_unsync(struct net_device *to, struct net_device *from);
3713 void dev_mc_flush(struct net_device *dev);
3714 void dev_mc_init(struct net_device *dev);
3715 
3716 /**
3717  *  __dev_mc_sync - Synchonize device's multicast list
3718  *  @dev:  device to sync
3719  *  @sync: function to call if address should be added
3720  *  @unsync: function to call if address should be removed
3721  *
3722  *  Add newly added addresses to the interface, and release
3723  *  addresses that have been deleted.
3724  */
3725 static inline int __dev_mc_sync(struct net_device *dev,
3726 				int (*sync)(struct net_device *,
3727 					    const unsigned char *),
3728 				int (*unsync)(struct net_device *,
3729 					      const unsigned char *))
3730 {
3731 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3732 }
3733 
3734 /**
3735  *  __dev_mc_unsync - Remove synchronized addresses from device
3736  *  @dev:  device to sync
3737  *  @unsync: function to call if address should be removed
3738  *
3739  *  Remove all addresses that were added to the device by dev_mc_sync().
3740  */
3741 static inline void __dev_mc_unsync(struct net_device *dev,
3742 				   int (*unsync)(struct net_device *,
3743 						 const unsigned char *))
3744 {
3745 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
3746 }
3747 
3748 /* Functions used for secondary unicast and multicast support */
3749 void dev_set_rx_mode(struct net_device *dev);
3750 void __dev_set_rx_mode(struct net_device *dev);
3751 int dev_set_promiscuity(struct net_device *dev, int inc);
3752 int dev_set_allmulti(struct net_device *dev, int inc);
3753 void netdev_state_change(struct net_device *dev);
3754 void netdev_notify_peers(struct net_device *dev);
3755 void netdev_features_change(struct net_device *dev);
3756 /* Load a device via the kmod */
3757 void dev_load(struct net *net, const char *name);
3758 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3759 					struct rtnl_link_stats64 *storage);
3760 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3761 			     const struct net_device_stats *netdev_stats);
3762 
3763 extern int		netdev_max_backlog;
3764 extern int		netdev_tstamp_prequeue;
3765 extern int		weight_p;
3766 
3767 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
3768 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3769 						     struct list_head **iter);
3770 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3771 						     struct list_head **iter);
3772 
3773 /* iterate through upper list, must be called under RCU read lock */
3774 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3775 	for (iter = &(dev)->adj_list.upper, \
3776 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3777 	     updev; \
3778 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3779 
3780 /* iterate through upper list, must be called under RCU read lock */
3781 #define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3782 	for (iter = &(dev)->all_adj_list.upper, \
3783 	     updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3784 	     updev; \
3785 	     updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
3786 
3787 void *netdev_lower_get_next_private(struct net_device *dev,
3788 				    struct list_head **iter);
3789 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3790 					struct list_head **iter);
3791 
3792 #define netdev_for_each_lower_private(dev, priv, iter) \
3793 	for (iter = (dev)->adj_list.lower.next, \
3794 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
3795 	     priv; \
3796 	     priv = netdev_lower_get_next_private(dev, &(iter)))
3797 
3798 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3799 	for (iter = &(dev)->adj_list.lower, \
3800 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3801 	     priv; \
3802 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3803 
3804 void *netdev_lower_get_next(struct net_device *dev,
3805 				struct list_head **iter);
3806 #define netdev_for_each_lower_dev(dev, ldev, iter) \
3807 	for (iter = (dev)->adj_list.lower.next, \
3808 	     ldev = netdev_lower_get_next(dev, &(iter)); \
3809 	     ldev; \
3810 	     ldev = netdev_lower_get_next(dev, &(iter)))
3811 
3812 void *netdev_adjacent_get_private(struct list_head *adj_list);
3813 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
3814 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3815 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3816 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3817 int netdev_master_upper_dev_link(struct net_device *dev,
3818 				 struct net_device *upper_dev,
3819 				 void *upper_priv, void *upper_info);
3820 void netdev_upper_dev_unlink(struct net_device *dev,
3821 			     struct net_device *upper_dev);
3822 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
3823 void *netdev_lower_dev_get_private(struct net_device *dev,
3824 				   struct net_device *lower_dev);
3825 void netdev_lower_state_changed(struct net_device *lower_dev,
3826 				void *lower_state_info);
3827 
3828 /* RSS keys are 40 or 52 bytes long */
3829 #define NETDEV_RSS_KEY_LEN 52
3830 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
3831 void netdev_rss_key_fill(void *buffer, size_t len);
3832 
3833 int dev_get_nest_level(struct net_device *dev,
3834 		       bool (*type_check)(const struct net_device *dev));
3835 int skb_checksum_help(struct sk_buff *skb);
3836 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3837 				  netdev_features_t features, bool tx_path);
3838 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3839 				    netdev_features_t features);
3840 
3841 struct netdev_bonding_info {
3842 	ifslave	slave;
3843 	ifbond	master;
3844 };
3845 
3846 struct netdev_notifier_bonding_info {
3847 	struct netdev_notifier_info info; /* must be first */
3848 	struct netdev_bonding_info  bonding_info;
3849 };
3850 
3851 void netdev_bonding_info_change(struct net_device *dev,
3852 				struct netdev_bonding_info *bonding_info);
3853 
3854 static inline
3855 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3856 {
3857 	return __skb_gso_segment(skb, features, true);
3858 }
3859 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
3860 
3861 static inline bool can_checksum_protocol(netdev_features_t features,
3862 					 __be16 protocol)
3863 {
3864 	if (protocol == htons(ETH_P_FCOE))
3865 		return !!(features & NETIF_F_FCOE_CRC);
3866 
3867 	/* Assume this is an IP checksum (not SCTP CRC) */
3868 
3869 	if (features & NETIF_F_HW_CSUM) {
3870 		/* Can checksum everything */
3871 		return true;
3872 	}
3873 
3874 	switch (protocol) {
3875 	case htons(ETH_P_IP):
3876 		return !!(features & NETIF_F_IP_CSUM);
3877 	case htons(ETH_P_IPV6):
3878 		return !!(features & NETIF_F_IPV6_CSUM);
3879 	default:
3880 		return false;
3881 	}
3882 }
3883 
3884 /* Map an ethertype into IP protocol if possible */
3885 static inline int eproto_to_ipproto(int eproto)
3886 {
3887 	switch (eproto) {
3888 	case htons(ETH_P_IP):
3889 		return IPPROTO_IP;
3890 	case htons(ETH_P_IPV6):
3891 		return IPPROTO_IPV6;
3892 	default:
3893 		return -1;
3894 	}
3895 }
3896 
3897 #ifdef CONFIG_BUG
3898 void netdev_rx_csum_fault(struct net_device *dev);
3899 #else
3900 static inline void netdev_rx_csum_fault(struct net_device *dev)
3901 {
3902 }
3903 #endif
3904 /* rx skb timestamps */
3905 void net_enable_timestamp(void);
3906 void net_disable_timestamp(void);
3907 
3908 #ifdef CONFIG_PROC_FS
3909 int __init dev_proc_init(void);
3910 #else
3911 #define dev_proc_init() 0
3912 #endif
3913 
3914 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
3915 					      struct sk_buff *skb, struct net_device *dev,
3916 					      bool more)
3917 {
3918 	skb->xmit_more = more ? 1 : 0;
3919 	return ops->ndo_start_xmit(skb, dev);
3920 }
3921 
3922 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
3923 					    struct netdev_queue *txq, bool more)
3924 {
3925 	const struct net_device_ops *ops = dev->netdev_ops;
3926 	int rc;
3927 
3928 	rc = __netdev_start_xmit(ops, skb, dev, more);
3929 	if (rc == NETDEV_TX_OK)
3930 		txq_trans_update(txq);
3931 
3932 	return rc;
3933 }
3934 
3935 int netdev_class_create_file_ns(struct class_attribute *class_attr,
3936 				const void *ns);
3937 void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3938 				 const void *ns);
3939 
3940 static inline int netdev_class_create_file(struct class_attribute *class_attr)
3941 {
3942 	return netdev_class_create_file_ns(class_attr, NULL);
3943 }
3944 
3945 static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3946 {
3947 	netdev_class_remove_file_ns(class_attr, NULL);
3948 }
3949 
3950 extern struct kobj_ns_type_operations net_ns_type_operations;
3951 
3952 const char *netdev_drivername(const struct net_device *dev);
3953 
3954 void linkwatch_run_queue(void);
3955 
3956 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3957 							  netdev_features_t f2)
3958 {
3959 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3960 		if (f1 & NETIF_F_HW_CSUM)
3961 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3962 		else
3963 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
3964 	}
3965 
3966 	return f1 & f2;
3967 }
3968 
3969 static inline netdev_features_t netdev_get_wanted_features(
3970 	struct net_device *dev)
3971 {
3972 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
3973 }
3974 netdev_features_t netdev_increment_features(netdev_features_t all,
3975 	netdev_features_t one, netdev_features_t mask);
3976 
3977 /* Allow TSO being used on stacked device :
3978  * Performing the GSO segmentation before last device
3979  * is a performance improvement.
3980  */
3981 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3982 							netdev_features_t mask)
3983 {
3984 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3985 }
3986 
3987 int __netdev_update_features(struct net_device *dev);
3988 void netdev_update_features(struct net_device *dev);
3989 void netdev_change_features(struct net_device *dev);
3990 
3991 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3992 					struct net_device *dev);
3993 
3994 netdev_features_t passthru_features_check(struct sk_buff *skb,
3995 					  struct net_device *dev,
3996 					  netdev_features_t features);
3997 netdev_features_t netif_skb_features(struct sk_buff *skb);
3998 
3999 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4000 {
4001 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4002 
4003 	/* check flags correspondence */
4004 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4005 	BUILD_BUG_ON(SKB_GSO_UDP     != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4006 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4007 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4008 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4009 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4010 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4011 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4012 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4013 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4014 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4015 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4016 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4017 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4018 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4019 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> 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