xref: /linux-6.15/include/linux/netdevice.h (revision a5a7daa5)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		Definitions for the Interfaces handler.
8  *
9  * Version:	@(#)dev.h	1.0.10	08/12/93
10  *
11  * Authors:	Ross Biro
12  *		Fred N. van Kempen, <[email protected]>
13  *		Corey Minyard <[email protected]>
14  *		Donald J. Becker, <[email protected]>
15  *		Alan Cox, <[email protected]>
16  *		Bjorn Ekwall. <[email protected]>
17  *              Pekka Riikonen <[email protected]>
18  *
19  *		Moved to /usr/include/linux for NET3
20  */
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
23 
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
31 
32 #include <linux/percpu.h>
33 #include <linux/rculist.h>
34 #include <linux/workqueue.h>
35 #include <linux/dynamic_queue_limits.h>
36 
37 #include <linux/ethtool.h>
38 #include <net/net_namespace.h>
39 #ifdef CONFIG_DCB
40 #include <net/dcbnl.h>
41 #endif
42 #include <net/netprio_cgroup.h>
43 #include <net/xdp.h>
44 
45 #include <linux/netdev_features.h>
46 #include <linux/neighbour.h>
47 #include <uapi/linux/netdevice.h>
48 #include <uapi/linux/if_bonding.h>
49 #include <uapi/linux/pkt_cls.h>
50 #include <linux/hashtable.h>
51 
52 struct netpoll_info;
53 struct device;
54 struct phy_device;
55 struct dsa_port;
56 
57 struct sfp_bus;
58 /* 802.11 specific */
59 struct wireless_dev;
60 /* 802.15.4 specific */
61 struct wpan_dev;
62 struct mpls_dev;
63 /* UDP Tunnel offloads */
64 struct udp_tunnel_info;
65 struct bpf_prog;
66 struct xdp_buff;
67 
68 void netdev_set_default_ethtool_ops(struct net_device *dev,
69 				    const struct ethtool_ops *ops);
70 
71 /* Backlog congestion levels */
72 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
73 #define NET_RX_DROP		1	/* packet dropped */
74 
75 /*
76  * Transmit return codes: transmit return codes originate from three different
77  * namespaces:
78  *
79  * - qdisc return codes
80  * - driver transmit return codes
81  * - errno values
82  *
83  * Drivers are allowed to return any one of those in their hard_start_xmit()
84  * function. Real network devices commonly used with qdiscs should only return
85  * the driver transmit return codes though - when qdiscs are used, the actual
86  * transmission happens asynchronously, so the value is not propagated to
87  * higher layers. Virtual network devices transmit synchronously; in this case
88  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
89  * others are propagated to higher layers.
90  */
91 
92 /* qdisc ->enqueue() return codes. */
93 #define NET_XMIT_SUCCESS	0x00
94 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
95 #define NET_XMIT_CN		0x02	/* congestion notification	*/
96 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
97 
98 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
99  * indicates that the device will soon be dropping packets, or already drops
100  * some packets of the same priority; prompting us to send less aggressively. */
101 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
102 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
103 
104 /* Driver transmit return codes */
105 #define NETDEV_TX_MASK		0xf0
106 
107 enum netdev_tx {
108 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
109 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
110 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
111 };
112 typedef enum netdev_tx netdev_tx_t;
113 
114 /*
115  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
116  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
117  */
118 static inline bool dev_xmit_complete(int rc)
119 {
120 	/*
121 	 * Positive cases with an skb consumed by a driver:
122 	 * - successful transmission (rc == NETDEV_TX_OK)
123 	 * - error while transmitting (rc < 0)
124 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
125 	 */
126 	if (likely(rc < NET_XMIT_MASK))
127 		return true;
128 
129 	return false;
130 }
131 
132 /*
133  *	Compute the worst-case header length according to the protocols
134  *	used.
135  */
136 
137 #if defined(CONFIG_HYPERV_NET)
138 # define LL_MAX_HEADER 128
139 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
140 # if defined(CONFIG_MAC80211_MESH)
141 #  define LL_MAX_HEADER 128
142 # else
143 #  define LL_MAX_HEADER 96
144 # endif
145 #else
146 # define LL_MAX_HEADER 32
147 #endif
148 
149 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
150     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
151 #define MAX_HEADER LL_MAX_HEADER
152 #else
153 #define MAX_HEADER (LL_MAX_HEADER + 48)
154 #endif
155 
156 /*
157  *	Old network device statistics. Fields are native words
158  *	(unsigned long) so they can be read and written atomically.
159  */
160 
161 struct net_device_stats {
162 	unsigned long	rx_packets;
163 	unsigned long	tx_packets;
164 	unsigned long	rx_bytes;
165 	unsigned long	tx_bytes;
166 	unsigned long	rx_errors;
167 	unsigned long	tx_errors;
168 	unsigned long	rx_dropped;
169 	unsigned long	tx_dropped;
170 	unsigned long	multicast;
171 	unsigned long	collisions;
172 	unsigned long	rx_length_errors;
173 	unsigned long	rx_over_errors;
174 	unsigned long	rx_crc_errors;
175 	unsigned long	rx_frame_errors;
176 	unsigned long	rx_fifo_errors;
177 	unsigned long	rx_missed_errors;
178 	unsigned long	tx_aborted_errors;
179 	unsigned long	tx_carrier_errors;
180 	unsigned long	tx_fifo_errors;
181 	unsigned long	tx_heartbeat_errors;
182 	unsigned long	tx_window_errors;
183 	unsigned long	rx_compressed;
184 	unsigned long	tx_compressed;
185 };
186 
187 
188 #include <linux/cache.h>
189 #include <linux/skbuff.h>
190 
191 #ifdef CONFIG_RPS
192 #include <linux/static_key.h>
193 extern struct static_key_false rps_needed;
194 extern struct static_key_false rfs_needed;
195 #endif
196 
197 struct neighbour;
198 struct neigh_parms;
199 struct sk_buff;
200 
201 struct netdev_hw_addr {
202 	struct list_head	list;
203 	unsigned char		addr[MAX_ADDR_LEN];
204 	unsigned char		type;
205 #define NETDEV_HW_ADDR_T_LAN		1
206 #define NETDEV_HW_ADDR_T_SAN		2
207 #define NETDEV_HW_ADDR_T_SLAVE		3
208 #define NETDEV_HW_ADDR_T_UNICAST	4
209 #define NETDEV_HW_ADDR_T_MULTICAST	5
210 	bool			global_use;
211 	int			sync_cnt;
212 	int			refcount;
213 	int			synced;
214 	struct rcu_head		rcu_head;
215 };
216 
217 struct netdev_hw_addr_list {
218 	struct list_head	list;
219 	int			count;
220 };
221 
222 #define netdev_hw_addr_list_count(l) ((l)->count)
223 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
224 #define netdev_hw_addr_list_for_each(ha, l) \
225 	list_for_each_entry(ha, &(l)->list, list)
226 
227 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
228 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
229 #define netdev_for_each_uc_addr(ha, dev) \
230 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
231 
232 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
233 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
234 #define netdev_for_each_mc_addr(ha, dev) \
235 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
236 
237 struct hh_cache {
238 	unsigned int	hh_len;
239 	seqlock_t	hh_lock;
240 
241 	/* cached hardware header; allow for machine alignment needs.        */
242 #define HH_DATA_MOD	16
243 #define HH_DATA_OFF(__len) \
244 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
245 #define HH_DATA_ALIGN(__len) \
246 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
247 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
248 };
249 
250 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
251  * Alternative is:
252  *   dev->hard_header_len ? (dev->hard_header_len +
253  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
254  *
255  * We could use other alignment values, but we must maintain the
256  * relationship HH alignment <= LL alignment.
257  */
258 #define LL_RESERVED_SPACE(dev) \
259 	((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
260 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
261 	((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
262 
263 struct header_ops {
264 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
265 			   unsigned short type, const void *daddr,
266 			   const void *saddr, unsigned int len);
267 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
268 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
269 	void	(*cache_update)(struct hh_cache *hh,
270 				const struct net_device *dev,
271 				const unsigned char *haddr);
272 	bool	(*validate)(const char *ll_header, unsigned int len);
273 	__be16	(*parse_protocol)(const struct sk_buff *skb);
274 };
275 
276 /* These flag bits are private to the generic network queueing
277  * layer; they may not be explicitly referenced by any other
278  * code.
279  */
280 
281 enum netdev_state_t {
282 	__LINK_STATE_START,
283 	__LINK_STATE_PRESENT,
284 	__LINK_STATE_NOCARRIER,
285 	__LINK_STATE_LINKWATCH_PENDING,
286 	__LINK_STATE_DORMANT,
287 };
288 
289 
290 /*
291  * This structure holds boot-time configured netdevice settings. They
292  * are then used in the device probing.
293  */
294 struct netdev_boot_setup {
295 	char name[IFNAMSIZ];
296 	struct ifmap map;
297 };
298 #define NETDEV_BOOT_SETUP_MAX 8
299 
300 int __init netdev_boot_setup(char *str);
301 
302 struct gro_list {
303 	struct list_head	list;
304 	int			count;
305 };
306 
307 /*
308  * size of gro hash buckets, must less than bit number of
309  * napi_struct::gro_bitmask
310  */
311 #define GRO_HASH_BUCKETS	8
312 
313 /*
314  * Structure for NAPI scheduling similar to tasklet but with weighting
315  */
316 struct napi_struct {
317 	/* The poll_list must only be managed by the entity which
318 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
319 	 * whoever atomically sets that bit can add this napi_struct
320 	 * to the per-CPU poll_list, and whoever clears that bit
321 	 * can remove from the list right before clearing the bit.
322 	 */
323 	struct list_head	poll_list;
324 
325 	unsigned long		state;
326 	int			weight;
327 	unsigned long		gro_bitmask;
328 	int			(*poll)(struct napi_struct *, int);
329 #ifdef CONFIG_NETPOLL
330 	int			poll_owner;
331 #endif
332 	struct net_device	*dev;
333 	struct gro_list		gro_hash[GRO_HASH_BUCKETS];
334 	struct sk_buff		*skb;
335 	struct list_head	rx_list; /* Pending GRO_NORMAL skbs */
336 	int			rx_count; /* length of rx_list */
337 	struct hrtimer		timer;
338 	struct list_head	dev_list;
339 	struct hlist_node	napi_hash_node;
340 	unsigned int		napi_id;
341 };
342 
343 enum {
344 	NAPI_STATE_SCHED,	/* Poll is scheduled */
345 	NAPI_STATE_MISSED,	/* reschedule a napi */
346 	NAPI_STATE_DISABLE,	/* Disable pending */
347 	NAPI_STATE_NPSVC,	/* Netpoll - don't dequeue from poll_list */
348 	NAPI_STATE_HASHED,	/* In NAPI hash (busy polling possible) */
349 	NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
350 	NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
351 };
352 
353 enum {
354 	NAPIF_STATE_SCHED	 = BIT(NAPI_STATE_SCHED),
355 	NAPIF_STATE_MISSED	 = BIT(NAPI_STATE_MISSED),
356 	NAPIF_STATE_DISABLE	 = BIT(NAPI_STATE_DISABLE),
357 	NAPIF_STATE_NPSVC	 = BIT(NAPI_STATE_NPSVC),
358 	NAPIF_STATE_HASHED	 = BIT(NAPI_STATE_HASHED),
359 	NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
360 	NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
361 };
362 
363 enum gro_result {
364 	GRO_MERGED,
365 	GRO_MERGED_FREE,
366 	GRO_HELD,
367 	GRO_NORMAL,
368 	GRO_DROP,
369 	GRO_CONSUMED,
370 };
371 typedef enum gro_result gro_result_t;
372 
373 /*
374  * enum rx_handler_result - Possible return values for rx_handlers.
375  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
376  * further.
377  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
378  * case skb->dev was changed by rx_handler.
379  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
380  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
381  *
382  * rx_handlers are functions called from inside __netif_receive_skb(), to do
383  * special processing of the skb, prior to delivery to protocol handlers.
384  *
385  * Currently, a net_device can only have a single rx_handler registered. Trying
386  * to register a second rx_handler will return -EBUSY.
387  *
388  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
389  * To unregister a rx_handler on a net_device, use
390  * netdev_rx_handler_unregister().
391  *
392  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
393  * do with the skb.
394  *
395  * If the rx_handler consumed the skb in some way, it should return
396  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
397  * the skb to be delivered in some other way.
398  *
399  * If the rx_handler changed skb->dev, to divert the skb to another
400  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
401  * new device will be called if it exists.
402  *
403  * If the rx_handler decides the skb should be ignored, it should return
404  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
405  * are registered on exact device (ptype->dev == skb->dev).
406  *
407  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
408  * delivered, it should return RX_HANDLER_PASS.
409  *
410  * A device without a registered rx_handler will behave as if rx_handler
411  * returned RX_HANDLER_PASS.
412  */
413 
414 enum rx_handler_result {
415 	RX_HANDLER_CONSUMED,
416 	RX_HANDLER_ANOTHER,
417 	RX_HANDLER_EXACT,
418 	RX_HANDLER_PASS,
419 };
420 typedef enum rx_handler_result rx_handler_result_t;
421 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
422 
423 void __napi_schedule(struct napi_struct *n);
424 void __napi_schedule_irqoff(struct napi_struct *n);
425 
426 static inline bool napi_disable_pending(struct napi_struct *n)
427 {
428 	return test_bit(NAPI_STATE_DISABLE, &n->state);
429 }
430 
431 bool napi_schedule_prep(struct napi_struct *n);
432 
433 /**
434  *	napi_schedule - schedule NAPI poll
435  *	@n: NAPI context
436  *
437  * Schedule NAPI poll routine to be called if it is not already
438  * running.
439  */
440 static inline void napi_schedule(struct napi_struct *n)
441 {
442 	if (napi_schedule_prep(n))
443 		__napi_schedule(n);
444 }
445 
446 /**
447  *	napi_schedule_irqoff - schedule NAPI poll
448  *	@n: NAPI context
449  *
450  * Variant of napi_schedule(), assuming hard irqs are masked.
451  */
452 static inline void napi_schedule_irqoff(struct napi_struct *n)
453 {
454 	if (napi_schedule_prep(n))
455 		__napi_schedule_irqoff(n);
456 }
457 
458 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
459 static inline bool napi_reschedule(struct napi_struct *napi)
460 {
461 	if (napi_schedule_prep(napi)) {
462 		__napi_schedule(napi);
463 		return true;
464 	}
465 	return false;
466 }
467 
468 bool napi_complete_done(struct napi_struct *n, int work_done);
469 /**
470  *	napi_complete - NAPI processing complete
471  *	@n: NAPI context
472  *
473  * Mark NAPI processing as complete.
474  * Consider using napi_complete_done() instead.
475  * Return false if device should avoid rearming interrupts.
476  */
477 static inline bool napi_complete(struct napi_struct *n)
478 {
479 	return napi_complete_done(n, 0);
480 }
481 
482 /**
483  *	napi_hash_del - remove a NAPI from global table
484  *	@napi: NAPI context
485  *
486  * Warning: caller must observe RCU grace period
487  * before freeing memory containing @napi, if
488  * this function returns true.
489  * Note: core networking stack automatically calls it
490  * from netif_napi_del().
491  * Drivers might want to call this helper to combine all
492  * the needed RCU grace periods into a single one.
493  */
494 bool napi_hash_del(struct napi_struct *napi);
495 
496 /**
497  *	napi_disable - prevent NAPI from scheduling
498  *	@n: NAPI context
499  *
500  * Stop NAPI from being scheduled on this context.
501  * Waits till any outstanding processing completes.
502  */
503 void napi_disable(struct napi_struct *n);
504 
505 /**
506  *	napi_enable - enable NAPI scheduling
507  *	@n: NAPI context
508  *
509  * Resume NAPI from being scheduled on this context.
510  * Must be paired with napi_disable.
511  */
512 static inline void napi_enable(struct napi_struct *n)
513 {
514 	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
515 	smp_mb__before_atomic();
516 	clear_bit(NAPI_STATE_SCHED, &n->state);
517 	clear_bit(NAPI_STATE_NPSVC, &n->state);
518 }
519 
520 /**
521  *	napi_synchronize - wait until NAPI is not running
522  *	@n: NAPI context
523  *
524  * Wait until NAPI is done being scheduled on this context.
525  * Waits till any outstanding processing completes but
526  * does not disable future activations.
527  */
528 static inline void napi_synchronize(const struct napi_struct *n)
529 {
530 	if (IS_ENABLED(CONFIG_SMP))
531 		while (test_bit(NAPI_STATE_SCHED, &n->state))
532 			msleep(1);
533 	else
534 		barrier();
535 }
536 
537 /**
538  *	napi_if_scheduled_mark_missed - if napi is running, set the
539  *	NAPIF_STATE_MISSED
540  *	@n: NAPI context
541  *
542  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
543  * NAPI is scheduled.
544  **/
545 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
546 {
547 	unsigned long val, new;
548 
549 	do {
550 		val = READ_ONCE(n->state);
551 		if (val & NAPIF_STATE_DISABLE)
552 			return true;
553 
554 		if (!(val & NAPIF_STATE_SCHED))
555 			return false;
556 
557 		new = val | NAPIF_STATE_MISSED;
558 	} while (cmpxchg(&n->state, val, new) != val);
559 
560 	return true;
561 }
562 
563 enum netdev_queue_state_t {
564 	__QUEUE_STATE_DRV_XOFF,
565 	__QUEUE_STATE_STACK_XOFF,
566 	__QUEUE_STATE_FROZEN,
567 };
568 
569 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
570 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
571 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
572 
573 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
574 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
575 					QUEUE_STATE_FROZEN)
576 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
577 					QUEUE_STATE_FROZEN)
578 
579 /*
580  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
581  * netif_tx_* functions below are used to manipulate this flag.  The
582  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
583  * queue independently.  The netif_xmit_*stopped functions below are called
584  * to check if the queue has been stopped by the driver or stack (either
585  * of the XOFF bits are set in the state).  Drivers should not need to call
586  * netif_xmit*stopped functions, they should only be using netif_tx_*.
587  */
588 
589 struct netdev_queue {
590 /*
591  * read-mostly part
592  */
593 	struct net_device	*dev;
594 	struct Qdisc __rcu	*qdisc;
595 	struct Qdisc		*qdisc_sleeping;
596 #ifdef CONFIG_SYSFS
597 	struct kobject		kobj;
598 #endif
599 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 	int			numa_node;
601 #endif
602 	unsigned long		tx_maxrate;
603 	/*
604 	 * Number of TX timeouts for this queue
605 	 * (/sys/class/net/DEV/Q/trans_timeout)
606 	 */
607 	unsigned long		trans_timeout;
608 
609 	/* Subordinate device that the queue has been assigned to */
610 	struct net_device	*sb_dev;
611 #ifdef CONFIG_XDP_SOCKETS
612 	struct xdp_umem         *umem;
613 #endif
614 /*
615  * write-mostly part
616  */
617 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
618 	int			xmit_lock_owner;
619 	/*
620 	 * Time (in jiffies) of last Tx
621 	 */
622 	unsigned long		trans_start;
623 
624 	unsigned long		state;
625 
626 #ifdef CONFIG_BQL
627 	struct dql		dql;
628 #endif
629 } ____cacheline_aligned_in_smp;
630 
631 extern int sysctl_fb_tunnels_only_for_init_net;
632 extern int sysctl_devconf_inherit_init_net;
633 
634 static inline bool net_has_fallback_tunnels(const struct net *net)
635 {
636 	return net == &init_net ||
637 	       !IS_ENABLED(CONFIG_SYSCTL) ||
638 	       !sysctl_fb_tunnels_only_for_init_net;
639 }
640 
641 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
642 {
643 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
644 	return q->numa_node;
645 #else
646 	return NUMA_NO_NODE;
647 #endif
648 }
649 
650 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
651 {
652 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
653 	q->numa_node = node;
654 #endif
655 }
656 
657 #ifdef CONFIG_RPS
658 /*
659  * This structure holds an RPS map which can be of variable length.  The
660  * map is an array of CPUs.
661  */
662 struct rps_map {
663 	unsigned int len;
664 	struct rcu_head rcu;
665 	u16 cpus[0];
666 };
667 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
668 
669 /*
670  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
671  * tail pointer for that CPU's input queue at the time of last enqueue, and
672  * a hardware filter index.
673  */
674 struct rps_dev_flow {
675 	u16 cpu;
676 	u16 filter;
677 	unsigned int last_qtail;
678 };
679 #define RPS_NO_FILTER 0xffff
680 
681 /*
682  * The rps_dev_flow_table structure contains a table of flow mappings.
683  */
684 struct rps_dev_flow_table {
685 	unsigned int mask;
686 	struct rcu_head rcu;
687 	struct rps_dev_flow flows[0];
688 };
689 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
690     ((_num) * sizeof(struct rps_dev_flow)))
691 
692 /*
693  * The rps_sock_flow_table contains mappings of flows to the last CPU
694  * on which they were processed by the application (set in recvmsg).
695  * Each entry is a 32bit value. Upper part is the high-order bits
696  * of flow hash, lower part is CPU number.
697  * rps_cpu_mask is used to partition the space, depending on number of
698  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
699  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
700  * meaning we use 32-6=26 bits for the hash.
701  */
702 struct rps_sock_flow_table {
703 	u32	mask;
704 
705 	u32	ents[0] ____cacheline_aligned_in_smp;
706 };
707 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
708 
709 #define RPS_NO_CPU 0xffff
710 
711 extern u32 rps_cpu_mask;
712 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
713 
714 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
715 					u32 hash)
716 {
717 	if (table && hash) {
718 		unsigned int index = hash & table->mask;
719 		u32 val = hash & ~rps_cpu_mask;
720 
721 		/* We only give a hint, preemption can change CPU under us */
722 		val |= raw_smp_processor_id();
723 
724 		if (table->ents[index] != val)
725 			table->ents[index] = val;
726 	}
727 }
728 
729 #ifdef CONFIG_RFS_ACCEL
730 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
731 			 u16 filter_id);
732 #endif
733 #endif /* CONFIG_RPS */
734 
735 /* This structure contains an instance of an RX queue. */
736 struct netdev_rx_queue {
737 #ifdef CONFIG_RPS
738 	struct rps_map __rcu		*rps_map;
739 	struct rps_dev_flow_table __rcu	*rps_flow_table;
740 #endif
741 	struct kobject			kobj;
742 	struct net_device		*dev;
743 	struct xdp_rxq_info		xdp_rxq;
744 #ifdef CONFIG_XDP_SOCKETS
745 	struct xdp_umem                 *umem;
746 #endif
747 } ____cacheline_aligned_in_smp;
748 
749 /*
750  * RX queue sysfs structures and functions.
751  */
752 struct rx_queue_attribute {
753 	struct attribute attr;
754 	ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
755 	ssize_t (*store)(struct netdev_rx_queue *queue,
756 			 const char *buf, size_t len);
757 };
758 
759 #ifdef CONFIG_XPS
760 /*
761  * This structure holds an XPS map which can be of variable length.  The
762  * map is an array of queues.
763  */
764 struct xps_map {
765 	unsigned int len;
766 	unsigned int alloc_len;
767 	struct rcu_head rcu;
768 	u16 queues[0];
769 };
770 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
771 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
772        - sizeof(struct xps_map)) / sizeof(u16))
773 
774 /*
775  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
776  */
777 struct xps_dev_maps {
778 	struct rcu_head rcu;
779 	struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
780 };
781 
782 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +	\
783 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
784 
785 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
786 	(_rxqs * (_tcs) * sizeof(struct xps_map *)))
787 
788 #endif /* CONFIG_XPS */
789 
790 #define TC_MAX_QUEUE	16
791 #define TC_BITMASK	15
792 /* HW offloaded queuing disciplines txq count and offset maps */
793 struct netdev_tc_txq {
794 	u16 count;
795 	u16 offset;
796 };
797 
798 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
799 /*
800  * This structure is to hold information about the device
801  * configured to run FCoE protocol stack.
802  */
803 struct netdev_fcoe_hbainfo {
804 	char	manufacturer[64];
805 	char	serial_number[64];
806 	char	hardware_version[64];
807 	char	driver_version[64];
808 	char	optionrom_version[64];
809 	char	firmware_version[64];
810 	char	model[256];
811 	char	model_description[256];
812 };
813 #endif
814 
815 #define MAX_PHYS_ITEM_ID_LEN 32
816 
817 /* This structure holds a unique identifier to identify some
818  * physical item (port for example) used by a netdevice.
819  */
820 struct netdev_phys_item_id {
821 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
822 	unsigned char id_len;
823 };
824 
825 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
826 					    struct netdev_phys_item_id *b)
827 {
828 	return a->id_len == b->id_len &&
829 	       memcmp(a->id, b->id, a->id_len) == 0;
830 }
831 
832 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
833 				       struct sk_buff *skb,
834 				       struct net_device *sb_dev);
835 
836 enum tc_setup_type {
837 	TC_SETUP_QDISC_MQPRIO,
838 	TC_SETUP_CLSU32,
839 	TC_SETUP_CLSFLOWER,
840 	TC_SETUP_CLSMATCHALL,
841 	TC_SETUP_CLSBPF,
842 	TC_SETUP_BLOCK,
843 	TC_SETUP_QDISC_CBS,
844 	TC_SETUP_QDISC_RED,
845 	TC_SETUP_QDISC_PRIO,
846 	TC_SETUP_QDISC_MQ,
847 	TC_SETUP_QDISC_ETF,
848 	TC_SETUP_ROOT_QDISC,
849 	TC_SETUP_QDISC_GRED,
850 	TC_SETUP_QDISC_TAPRIO,
851 };
852 
853 /* These structures hold the attributes of bpf state that are being passed
854  * to the netdevice through the bpf op.
855  */
856 enum bpf_netdev_command {
857 	/* Set or clear a bpf program used in the earliest stages of packet
858 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
859 	 * is responsible for calling bpf_prog_put on any old progs that are
860 	 * stored. In case of error, the callee need not release the new prog
861 	 * reference, but on success it takes ownership and must bpf_prog_put
862 	 * when it is no longer used.
863 	 */
864 	XDP_SETUP_PROG,
865 	XDP_SETUP_PROG_HW,
866 	XDP_QUERY_PROG,
867 	XDP_QUERY_PROG_HW,
868 	/* BPF program for offload callbacks, invoked at program load time. */
869 	BPF_OFFLOAD_MAP_ALLOC,
870 	BPF_OFFLOAD_MAP_FREE,
871 	XDP_SETUP_XSK_UMEM,
872 };
873 
874 struct bpf_prog_offload_ops;
875 struct netlink_ext_ack;
876 struct xdp_umem;
877 
878 struct netdev_bpf {
879 	enum bpf_netdev_command command;
880 	union {
881 		/* XDP_SETUP_PROG */
882 		struct {
883 			u32 flags;
884 			struct bpf_prog *prog;
885 			struct netlink_ext_ack *extack;
886 		};
887 		/* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
888 		struct {
889 			u32 prog_id;
890 			/* flags with which program was installed */
891 			u32 prog_flags;
892 		};
893 		/* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
894 		struct {
895 			struct bpf_offloaded_map *offmap;
896 		};
897 		/* XDP_SETUP_XSK_UMEM */
898 		struct {
899 			struct xdp_umem *umem;
900 			u16 queue_id;
901 		} xsk;
902 	};
903 };
904 
905 /* Flags for ndo_xsk_wakeup. */
906 #define XDP_WAKEUP_RX (1 << 0)
907 #define XDP_WAKEUP_TX (1 << 1)
908 
909 #ifdef CONFIG_XFRM_OFFLOAD
910 struct xfrmdev_ops {
911 	int	(*xdo_dev_state_add) (struct xfrm_state *x);
912 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
913 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
914 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
915 				       struct xfrm_state *x);
916 	void	(*xdo_dev_state_advance_esn) (struct xfrm_state *x);
917 };
918 #endif
919 
920 struct dev_ifalias {
921 	struct rcu_head rcuhead;
922 	char ifalias[];
923 };
924 
925 struct devlink;
926 struct tlsdev_ops;
927 
928 struct netdev_name_node {
929 	struct hlist_node hlist;
930 	struct list_head list;
931 	struct net_device *dev;
932 	const char *name;
933 };
934 
935 int netdev_name_node_alt_create(struct net_device *dev, const char *name);
936 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name);
937 
938 /*
939  * This structure defines the management hooks for network devices.
940  * The following hooks can be defined; unless noted otherwise, they are
941  * optional and can be filled with a null pointer.
942  *
943  * int (*ndo_init)(struct net_device *dev);
944  *     This function is called once when a network device is registered.
945  *     The network device can use this for any late stage initialization
946  *     or semantic validation. It can fail with an error code which will
947  *     be propagated back to register_netdev.
948  *
949  * void (*ndo_uninit)(struct net_device *dev);
950  *     This function is called when device is unregistered or when registration
951  *     fails. It is not called if init fails.
952  *
953  * int (*ndo_open)(struct net_device *dev);
954  *     This function is called when a network device transitions to the up
955  *     state.
956  *
957  * int (*ndo_stop)(struct net_device *dev);
958  *     This function is called when a network device transitions to the down
959  *     state.
960  *
961  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
962  *                               struct net_device *dev);
963  *	Called when a packet needs to be transmitted.
964  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
965  *	the queue before that can happen; it's for obsolete devices and weird
966  *	corner cases, but the stack really does a non-trivial amount
967  *	of useless work if you return NETDEV_TX_BUSY.
968  *	Required; cannot be NULL.
969  *
970  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
971  *					   struct net_device *dev
972  *					   netdev_features_t features);
973  *	Called by core transmit path to determine if device is capable of
974  *	performing offload operations on a given packet. This is to give
975  *	the device an opportunity to implement any restrictions that cannot
976  *	be otherwise expressed by feature flags. The check is called with
977  *	the set of features that the stack has calculated and it returns
978  *	those the driver believes to be appropriate.
979  *
980  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
981  *                         struct net_device *sb_dev);
982  *	Called to decide which queue to use when device supports multiple
983  *	transmit queues.
984  *
985  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
986  *	This function is called to allow device receiver to make
987  *	changes to configuration when multicast or promiscuous is enabled.
988  *
989  * void (*ndo_set_rx_mode)(struct net_device *dev);
990  *	This function is called device changes address list filtering.
991  *	If driver handles unicast address filtering, it should set
992  *	IFF_UNICAST_FLT in its priv_flags.
993  *
994  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
995  *	This function  is called when the Media Access Control address
996  *	needs to be changed. If this interface is not defined, the
997  *	MAC address can not be changed.
998  *
999  * int (*ndo_validate_addr)(struct net_device *dev);
1000  *	Test if Media Access Control address is valid for the device.
1001  *
1002  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1003  *	Called when a user requests an ioctl which can't be handled by
1004  *	the generic interface code. If not defined ioctls return
1005  *	not supported error code.
1006  *
1007  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1008  *	Used to set network devices bus interface parameters. This interface
1009  *	is retained for legacy reasons; new devices should use the bus
1010  *	interface (PCI) for low level management.
1011  *
1012  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1013  *	Called when a user wants to change the Maximum Transfer Unit
1014  *	of a device.
1015  *
1016  * void (*ndo_tx_timeout)(struct net_device *dev);
1017  *	Callback used when the transmitter has not made any progress
1018  *	for dev->watchdog ticks.
1019  *
1020  * void (*ndo_get_stats64)(struct net_device *dev,
1021  *                         struct rtnl_link_stats64 *storage);
1022  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1023  *	Called when a user wants to get the network device usage
1024  *	statistics. Drivers must do one of the following:
1025  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
1026  *	   rtnl_link_stats64 structure passed by the caller.
1027  *	2. Define @ndo_get_stats to update a net_device_stats structure
1028  *	   (which should normally be dev->stats) and return a pointer to
1029  *	   it. The structure may be changed asynchronously only if each
1030  *	   field is written atomically.
1031  *	3. Update dev->stats asynchronously and atomically, and define
1032  *	   neither operation.
1033  *
1034  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1035  *	Return true if this device supports offload stats of this attr_id.
1036  *
1037  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1038  *	void *attr_data)
1039  *	Get statistics for offload operations by attr_id. Write it into the
1040  *	attr_data pointer.
1041  *
1042  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1043  *	If device supports VLAN filtering this function is called when a
1044  *	VLAN id is registered.
1045  *
1046  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1047  *	If device supports VLAN filtering this function is called when a
1048  *	VLAN id is unregistered.
1049  *
1050  * void (*ndo_poll_controller)(struct net_device *dev);
1051  *
1052  *	SR-IOV management functions.
1053  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1054  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1055  *			  u8 qos, __be16 proto);
1056  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1057  *			  int max_tx_rate);
1058  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1059  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1060  * int (*ndo_get_vf_config)(struct net_device *dev,
1061  *			    int vf, struct ifla_vf_info *ivf);
1062  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1063  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1064  *			  struct nlattr *port[]);
1065  *
1066  *      Enable or disable the VF ability to query its RSS Redirection Table and
1067  *      Hash Key. This is needed since on some devices VF share this information
1068  *      with PF and querying it may introduce a theoretical security risk.
1069  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1070  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1071  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1072  *		       void *type_data);
1073  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
1074  *	This is always called from the stack with the rtnl lock held and netif
1075  *	tx queues stopped. This allows the netdevice to perform queue
1076  *	management safely.
1077  *
1078  *	Fiber Channel over Ethernet (FCoE) offload functions.
1079  * int (*ndo_fcoe_enable)(struct net_device *dev);
1080  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
1081  *	so the underlying device can perform whatever needed configuration or
1082  *	initialization to support acceleration of FCoE traffic.
1083  *
1084  * int (*ndo_fcoe_disable)(struct net_device *dev);
1085  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
1086  *	so the underlying device can perform whatever needed clean-ups to
1087  *	stop supporting acceleration of FCoE traffic.
1088  *
1089  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1090  *			     struct scatterlist *sgl, unsigned int sgc);
1091  *	Called when the FCoE Initiator wants to initialize an I/O that
1092  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1093  *	perform necessary setup and returns 1 to indicate the device is set up
1094  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1095  *
1096  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1097  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1098  *	indicated by the FC exchange id 'xid', so the underlying device can
1099  *	clean up and reuse resources for later DDP requests.
1100  *
1101  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1102  *			      struct scatterlist *sgl, unsigned int sgc);
1103  *	Called when the FCoE Target wants to initialize an I/O that
1104  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1105  *	perform necessary setup and returns 1 to indicate the device is set up
1106  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1107  *
1108  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1109  *			       struct netdev_fcoe_hbainfo *hbainfo);
1110  *	Called when the FCoE Protocol stack wants information on the underlying
1111  *	device. This information is utilized by the FCoE protocol stack to
1112  *	register attributes with Fiber Channel management service as per the
1113  *	FC-GS Fabric Device Management Information(FDMI) specification.
1114  *
1115  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1116  *	Called when the underlying device wants to override default World Wide
1117  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1118  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1119  *	protocol stack to use.
1120  *
1121  *	RFS acceleration.
1122  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1123  *			    u16 rxq_index, u32 flow_id);
1124  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1125  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1126  *	Return the filter ID on success, or a negative error code.
1127  *
1128  *	Slave management functions (for bridge, bonding, etc).
1129  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1130  *	Called to make another netdev an underling.
1131  *
1132  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1133  *	Called to release previously enslaved netdev.
1134  *
1135  *      Feature/offload setting functions.
1136  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1137  *		netdev_features_t features);
1138  *	Adjusts the requested feature flags according to device-specific
1139  *	constraints, and returns the resulting flags. Must not modify
1140  *	the device state.
1141  *
1142  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1143  *	Called to update device configuration to new features. Passed
1144  *	feature set might be less than what was returned by ndo_fix_features()).
1145  *	Must return >0 or -errno if it changed dev->features itself.
1146  *
1147  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1148  *		      struct net_device *dev,
1149  *		      const unsigned char *addr, u16 vid, u16 flags,
1150  *		      struct netlink_ext_ack *extack);
1151  *	Adds an FDB entry to dev for addr.
1152  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1153  *		      struct net_device *dev,
1154  *		      const unsigned char *addr, u16 vid)
1155  *	Deletes the FDB entry from dev coresponding to addr.
1156  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1157  *		       struct net_device *dev, struct net_device *filter_dev,
1158  *		       int *idx)
1159  *	Used to add FDB entries to dump requests. Implementers should add
1160  *	entries to skb and update idx with the number of entries.
1161  *
1162  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1163  *			     u16 flags, struct netlink_ext_ack *extack)
1164  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1165  *			     struct net_device *dev, u32 filter_mask,
1166  *			     int nlflags)
1167  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1168  *			     u16 flags);
1169  *
1170  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1171  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1172  *	which do not represent real hardware may define this to allow their
1173  *	userspace components to manage their virtual carrier state. Devices
1174  *	that determine carrier state from physical hardware properties (eg
1175  *	network cables) or protocol-dependent mechanisms (eg
1176  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1177  *
1178  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1179  *			       struct netdev_phys_item_id *ppid);
1180  *	Called to get ID of physical port of this device. If driver does
1181  *	not implement this, it is assumed that the hw is not able to have
1182  *	multiple net devices on single physical port.
1183  *
1184  * int (*ndo_get_port_parent_id)(struct net_device *dev,
1185  *				 struct netdev_phys_item_id *ppid)
1186  *	Called to get the parent ID of the physical port of this device.
1187  *
1188  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1189  *			      struct udp_tunnel_info *ti);
1190  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1191  *	address family that a UDP tunnel is listnening to. It is called only
1192  *	when a new port starts listening. The operation is protected by the
1193  *	RTNL.
1194  *
1195  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1196  *			      struct udp_tunnel_info *ti);
1197  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1198  *	address family that the UDP tunnel is not listening to anymore. The
1199  *	operation is protected by the RTNL.
1200  *
1201  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1202  *				 struct net_device *dev)
1203  *	Called by upper layer devices to accelerate switching or other
1204  *	station functionality into hardware. 'pdev is the lowerdev
1205  *	to use for the offload and 'dev' is the net device that will
1206  *	back the offload. Returns a pointer to the private structure
1207  *	the upper layer will maintain.
1208  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1209  *	Called by upper layer device to delete the station created
1210  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1211  *	the station and priv is the structure returned by the add
1212  *	operation.
1213  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1214  *			     int queue_index, u32 maxrate);
1215  *	Called when a user wants to set a max-rate limitation of specific
1216  *	TX queue.
1217  * int (*ndo_get_iflink)(const struct net_device *dev);
1218  *	Called to get the iflink value of this device.
1219  * void (*ndo_change_proto_down)(struct net_device *dev,
1220  *				 bool proto_down);
1221  *	This function is used to pass protocol port error state information
1222  *	to the switch driver. The switch driver can react to the proto_down
1223  *      by doing a phys down on the associated switch port.
1224  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1225  *	This function is used to get egress tunnel information for given skb.
1226  *	This is useful for retrieving outer tunnel header parameters while
1227  *	sampling packet.
1228  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1229  *	This function is used to specify the headroom that the skb must
1230  *	consider when allocation skb during packet reception. Setting
1231  *	appropriate rx headroom value allows avoiding skb head copy on
1232  *	forward. Setting a negative value resets the rx headroom to the
1233  *	default value.
1234  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1235  *	This function is used to set or query state related to XDP on the
1236  *	netdevice and manage BPF offload. See definition of
1237  *	enum bpf_netdev_command for details.
1238  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1239  *			u32 flags);
1240  *	This function is used to submit @n XDP packets for transmit on a
1241  *	netdevice. Returns number of frames successfully transmitted, frames
1242  *	that got dropped are freed/returned via xdp_return_frame().
1243  *	Returns negative number, means general error invoking ndo, meaning
1244  *	no frames were xmit'ed and core-caller will free all frames.
1245  * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1246  *      This function is used to wake up the softirq, ksoftirqd or kthread
1247  *	responsible for sending and/or receiving packets on a specific
1248  *	queue id bound to an AF_XDP socket. The flags field specifies if
1249  *	only RX, only Tx, or both should be woken up using the flags
1250  *	XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1251  * struct devlink_port *(*ndo_get_devlink_port)(struct net_device *dev);
1252  *	Get devlink port instance associated with a given netdev.
1253  *	Called with a reference on the netdevice and devlink locks only,
1254  *	rtnl_lock is not held.
1255  */
1256 struct net_device_ops {
1257 	int			(*ndo_init)(struct net_device *dev);
1258 	void			(*ndo_uninit)(struct net_device *dev);
1259 	int			(*ndo_open)(struct net_device *dev);
1260 	int			(*ndo_stop)(struct net_device *dev);
1261 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1262 						  struct net_device *dev);
1263 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1264 						      struct net_device *dev,
1265 						      netdev_features_t features);
1266 	u16			(*ndo_select_queue)(struct net_device *dev,
1267 						    struct sk_buff *skb,
1268 						    struct net_device *sb_dev);
1269 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1270 						       int flags);
1271 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1272 	int			(*ndo_set_mac_address)(struct net_device *dev,
1273 						       void *addr);
1274 	int			(*ndo_validate_addr)(struct net_device *dev);
1275 	int			(*ndo_do_ioctl)(struct net_device *dev,
1276 					        struct ifreq *ifr, int cmd);
1277 	int			(*ndo_set_config)(struct net_device *dev,
1278 					          struct ifmap *map);
1279 	int			(*ndo_change_mtu)(struct net_device *dev,
1280 						  int new_mtu);
1281 	int			(*ndo_neigh_setup)(struct net_device *dev,
1282 						   struct neigh_parms *);
1283 	void			(*ndo_tx_timeout) (struct net_device *dev);
1284 
1285 	void			(*ndo_get_stats64)(struct net_device *dev,
1286 						   struct rtnl_link_stats64 *storage);
1287 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1288 	int			(*ndo_get_offload_stats)(int attr_id,
1289 							 const struct net_device *dev,
1290 							 void *attr_data);
1291 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1292 
1293 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1294 						       __be16 proto, u16 vid);
1295 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1296 						        __be16 proto, u16 vid);
1297 #ifdef CONFIG_NET_POLL_CONTROLLER
1298 	void                    (*ndo_poll_controller)(struct net_device *dev);
1299 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1300 						     struct netpoll_info *info);
1301 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1302 #endif
1303 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1304 						  int queue, u8 *mac);
1305 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1306 						   int queue, u16 vlan,
1307 						   u8 qos, __be16 proto);
1308 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1309 						   int vf, int min_tx_rate,
1310 						   int max_tx_rate);
1311 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1312 						       int vf, bool setting);
1313 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1314 						    int vf, bool setting);
1315 	int			(*ndo_get_vf_config)(struct net_device *dev,
1316 						     int vf,
1317 						     struct ifla_vf_info *ivf);
1318 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1319 							 int vf, int link_state);
1320 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1321 						    int vf,
1322 						    struct ifla_vf_stats
1323 						    *vf_stats);
1324 	int			(*ndo_set_vf_port)(struct net_device *dev,
1325 						   int vf,
1326 						   struct nlattr *port[]);
1327 	int			(*ndo_get_vf_port)(struct net_device *dev,
1328 						   int vf, struct sk_buff *skb);
1329 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1330 						   int vf, u64 guid,
1331 						   int guid_type);
1332 	int			(*ndo_set_vf_rss_query_en)(
1333 						   struct net_device *dev,
1334 						   int vf, bool setting);
1335 	int			(*ndo_setup_tc)(struct net_device *dev,
1336 						enum tc_setup_type type,
1337 						void *type_data);
1338 #if IS_ENABLED(CONFIG_FCOE)
1339 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1340 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1341 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1342 						      u16 xid,
1343 						      struct scatterlist *sgl,
1344 						      unsigned int sgc);
1345 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1346 						     u16 xid);
1347 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1348 						       u16 xid,
1349 						       struct scatterlist *sgl,
1350 						       unsigned int sgc);
1351 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1352 							struct netdev_fcoe_hbainfo *hbainfo);
1353 #endif
1354 
1355 #if IS_ENABLED(CONFIG_LIBFCOE)
1356 #define NETDEV_FCOE_WWNN 0
1357 #define NETDEV_FCOE_WWPN 1
1358 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1359 						    u64 *wwn, int type);
1360 #endif
1361 
1362 #ifdef CONFIG_RFS_ACCEL
1363 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1364 						     const struct sk_buff *skb,
1365 						     u16 rxq_index,
1366 						     u32 flow_id);
1367 #endif
1368 	int			(*ndo_add_slave)(struct net_device *dev,
1369 						 struct net_device *slave_dev,
1370 						 struct netlink_ext_ack *extack);
1371 	int			(*ndo_del_slave)(struct net_device *dev,
1372 						 struct net_device *slave_dev);
1373 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1374 						    netdev_features_t features);
1375 	int			(*ndo_set_features)(struct net_device *dev,
1376 						    netdev_features_t features);
1377 	int			(*ndo_neigh_construct)(struct net_device *dev,
1378 						       struct neighbour *n);
1379 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1380 						     struct neighbour *n);
1381 
1382 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1383 					       struct nlattr *tb[],
1384 					       struct net_device *dev,
1385 					       const unsigned char *addr,
1386 					       u16 vid,
1387 					       u16 flags,
1388 					       struct netlink_ext_ack *extack);
1389 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1390 					       struct nlattr *tb[],
1391 					       struct net_device *dev,
1392 					       const unsigned char *addr,
1393 					       u16 vid);
1394 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1395 						struct netlink_callback *cb,
1396 						struct net_device *dev,
1397 						struct net_device *filter_dev,
1398 						int *idx);
1399 	int			(*ndo_fdb_get)(struct sk_buff *skb,
1400 					       struct nlattr *tb[],
1401 					       struct net_device *dev,
1402 					       const unsigned char *addr,
1403 					       u16 vid, u32 portid, u32 seq,
1404 					       struct netlink_ext_ack *extack);
1405 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1406 						      struct nlmsghdr *nlh,
1407 						      u16 flags,
1408 						      struct netlink_ext_ack *extack);
1409 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1410 						      u32 pid, u32 seq,
1411 						      struct net_device *dev,
1412 						      u32 filter_mask,
1413 						      int nlflags);
1414 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1415 						      struct nlmsghdr *nlh,
1416 						      u16 flags);
1417 	int			(*ndo_change_carrier)(struct net_device *dev,
1418 						      bool new_carrier);
1419 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1420 							struct netdev_phys_item_id *ppid);
1421 	int			(*ndo_get_port_parent_id)(struct net_device *dev,
1422 							  struct netdev_phys_item_id *ppid);
1423 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1424 							  char *name, size_t len);
1425 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1426 						      struct udp_tunnel_info *ti);
1427 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1428 						      struct udp_tunnel_info *ti);
1429 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1430 							struct net_device *dev);
1431 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1432 							void *priv);
1433 
1434 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1435 						      int queue_index,
1436 						      u32 maxrate);
1437 	int			(*ndo_get_iflink)(const struct net_device *dev);
1438 	int			(*ndo_change_proto_down)(struct net_device *dev,
1439 							 bool proto_down);
1440 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1441 						       struct sk_buff *skb);
1442 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1443 						       int needed_headroom);
1444 	int			(*ndo_bpf)(struct net_device *dev,
1445 					   struct netdev_bpf *bpf);
1446 	int			(*ndo_xdp_xmit)(struct net_device *dev, int n,
1447 						struct xdp_frame **xdp,
1448 						u32 flags);
1449 	int			(*ndo_xsk_wakeup)(struct net_device *dev,
1450 						  u32 queue_id, u32 flags);
1451 	struct devlink_port *	(*ndo_get_devlink_port)(struct net_device *dev);
1452 };
1453 
1454 /**
1455  * enum net_device_priv_flags - &struct net_device priv_flags
1456  *
1457  * These are the &struct net_device, they are only set internally
1458  * by drivers and used in the kernel. These flags are invisible to
1459  * userspace; this means that the order of these flags can change
1460  * during any kernel release.
1461  *
1462  * You should have a pretty good reason to be extending these flags.
1463  *
1464  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1465  * @IFF_EBRIDGE: Ethernet bridging device
1466  * @IFF_BONDING: bonding master or slave
1467  * @IFF_ISATAP: ISATAP interface (RFC4214)
1468  * @IFF_WAN_HDLC: WAN HDLC device
1469  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1470  *	release skb->dst
1471  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1472  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1473  * @IFF_MACVLAN_PORT: device used as macvlan port
1474  * @IFF_BRIDGE_PORT: device used as bridge port
1475  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1476  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1477  * @IFF_UNICAST_FLT: Supports unicast filtering
1478  * @IFF_TEAM_PORT: device used as team port
1479  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1480  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1481  *	change when it's running
1482  * @IFF_MACVLAN: Macvlan device
1483  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1484  *	underlying stacked devices
1485  * @IFF_L3MDEV_MASTER: device is an L3 master device
1486  * @IFF_NO_QUEUE: device can run without qdisc attached
1487  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1488  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1489  * @IFF_TEAM: device is a team device
1490  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1491  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1492  *	entity (i.e. the master device for bridged veth)
1493  * @IFF_MACSEC: device is a MACsec device
1494  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1495  * @IFF_FAILOVER: device is a failover master device
1496  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1497  * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1498  * @IFF_LIVE_RENAME_OK: rename is allowed while device is up and running
1499  */
1500 enum netdev_priv_flags {
1501 	IFF_802_1Q_VLAN			= 1<<0,
1502 	IFF_EBRIDGE			= 1<<1,
1503 	IFF_BONDING			= 1<<2,
1504 	IFF_ISATAP			= 1<<3,
1505 	IFF_WAN_HDLC			= 1<<4,
1506 	IFF_XMIT_DST_RELEASE		= 1<<5,
1507 	IFF_DONT_BRIDGE			= 1<<6,
1508 	IFF_DISABLE_NETPOLL		= 1<<7,
1509 	IFF_MACVLAN_PORT		= 1<<8,
1510 	IFF_BRIDGE_PORT			= 1<<9,
1511 	IFF_OVS_DATAPATH		= 1<<10,
1512 	IFF_TX_SKB_SHARING		= 1<<11,
1513 	IFF_UNICAST_FLT			= 1<<12,
1514 	IFF_TEAM_PORT			= 1<<13,
1515 	IFF_SUPP_NOFCS			= 1<<14,
1516 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1517 	IFF_MACVLAN			= 1<<16,
1518 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1519 	IFF_L3MDEV_MASTER		= 1<<18,
1520 	IFF_NO_QUEUE			= 1<<19,
1521 	IFF_OPENVSWITCH			= 1<<20,
1522 	IFF_L3MDEV_SLAVE		= 1<<21,
1523 	IFF_TEAM			= 1<<22,
1524 	IFF_RXFH_CONFIGURED		= 1<<23,
1525 	IFF_PHONY_HEADROOM		= 1<<24,
1526 	IFF_MACSEC			= 1<<25,
1527 	IFF_NO_RX_HANDLER		= 1<<26,
1528 	IFF_FAILOVER			= 1<<27,
1529 	IFF_FAILOVER_SLAVE		= 1<<28,
1530 	IFF_L3MDEV_RX_HANDLER		= 1<<29,
1531 	IFF_LIVE_RENAME_OK		= 1<<30,
1532 };
1533 
1534 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1535 #define IFF_EBRIDGE			IFF_EBRIDGE
1536 #define IFF_BONDING			IFF_BONDING
1537 #define IFF_ISATAP			IFF_ISATAP
1538 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1539 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1540 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1541 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1542 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1543 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1544 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1545 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1546 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1547 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1548 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1549 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1550 #define IFF_MACVLAN			IFF_MACVLAN
1551 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1552 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1553 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1554 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1555 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1556 #define IFF_TEAM			IFF_TEAM
1557 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1558 #define IFF_MACSEC			IFF_MACSEC
1559 #define IFF_NO_RX_HANDLER		IFF_NO_RX_HANDLER
1560 #define IFF_FAILOVER			IFF_FAILOVER
1561 #define IFF_FAILOVER_SLAVE		IFF_FAILOVER_SLAVE
1562 #define IFF_L3MDEV_RX_HANDLER		IFF_L3MDEV_RX_HANDLER
1563 #define IFF_LIVE_RENAME_OK		IFF_LIVE_RENAME_OK
1564 
1565 /**
1566  *	struct net_device - The DEVICE structure.
1567  *
1568  *	Actually, this whole structure is a big mistake.  It mixes I/O
1569  *	data with strictly "high-level" data, and it has to know about
1570  *	almost every data structure used in the INET module.
1571  *
1572  *	@name:	This is the first field of the "visible" part of this structure
1573  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1574  *		of the interface.
1575  *
1576  *	@name_node:	Name hashlist node
1577  *	@ifalias:	SNMP alias
1578  *	@mem_end:	Shared memory end
1579  *	@mem_start:	Shared memory start
1580  *	@base_addr:	Device I/O address
1581  *	@irq:		Device IRQ number
1582  *
1583  *	@state:		Generic network queuing layer state, see netdev_state_t
1584  *	@dev_list:	The global list of network devices
1585  *	@napi_list:	List entry used for polling NAPI devices
1586  *	@unreg_list:	List entry  when we are unregistering the
1587  *			device; see the function unregister_netdev
1588  *	@close_list:	List entry used when we are closing the device
1589  *	@ptype_all:     Device-specific packet handlers for all protocols
1590  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1591  *
1592  *	@adj_list:	Directly linked devices, like slaves for bonding
1593  *	@features:	Currently active device features
1594  *	@hw_features:	User-changeable features
1595  *
1596  *	@wanted_features:	User-requested features
1597  *	@vlan_features:		Mask of features inheritable by VLAN devices
1598  *
1599  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1600  *				This field indicates what encapsulation
1601  *				offloads the hardware is capable of doing,
1602  *				and drivers will need to set them appropriately.
1603  *
1604  *	@mpls_features:	Mask of features inheritable by MPLS
1605  *
1606  *	@ifindex:	interface index
1607  *	@group:		The group the device belongs to
1608  *
1609  *	@stats:		Statistics struct, which was left as a legacy, use
1610  *			rtnl_link_stats64 instead
1611  *
1612  *	@rx_dropped:	Dropped packets by core network,
1613  *			do not use this in drivers
1614  *	@tx_dropped:	Dropped packets by core network,
1615  *			do not use this in drivers
1616  *	@rx_nohandler:	nohandler dropped packets by core network on
1617  *			inactive devices, do not use this in drivers
1618  *	@carrier_up_count:	Number of times the carrier has been up
1619  *	@carrier_down_count:	Number of times the carrier has been down
1620  *
1621  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1622  *				instead of ioctl,
1623  *				see <net/iw_handler.h> for details.
1624  *	@wireless_data:	Instance data managed by the core of wireless extensions
1625  *
1626  *	@netdev_ops:	Includes several pointers to callbacks,
1627  *			if one wants to override the ndo_*() functions
1628  *	@ethtool_ops:	Management operations
1629  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1630  *			discovery handling. Necessary for e.g. 6LoWPAN.
1631  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1632  *			of Layer 2 headers.
1633  *
1634  *	@flags:		Interface flags (a la BSD)
1635  *	@priv_flags:	Like 'flags' but invisible to userspace,
1636  *			see if.h for the definitions
1637  *	@gflags:	Global flags ( kept as legacy )
1638  *	@padded:	How much padding added by alloc_netdev()
1639  *	@operstate:	RFC2863 operstate
1640  *	@link_mode:	Mapping policy to operstate
1641  *	@if_port:	Selectable AUI, TP, ...
1642  *	@dma:		DMA channel
1643  *	@mtu:		Interface MTU value
1644  *	@min_mtu:	Interface Minimum MTU value
1645  *	@max_mtu:	Interface Maximum MTU value
1646  *	@type:		Interface hardware type
1647  *	@hard_header_len: Maximum hardware header length.
1648  *	@min_header_len:  Minimum hardware header length
1649  *
1650  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1651  *			  cases can this be guaranteed
1652  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1653  *			  cases can this be guaranteed. Some cases also use
1654  *			  LL_MAX_HEADER instead to allocate the skb
1655  *
1656  *	interface address info:
1657  *
1658  * 	@perm_addr:		Permanent hw address
1659  * 	@addr_assign_type:	Hw address assignment type
1660  * 	@addr_len:		Hardware address length
1661  *	@upper_level:		Maximum depth level of upper devices.
1662  *	@lower_level:		Maximum depth level of lower devices.
1663  *	@neigh_priv_len:	Used in neigh_alloc()
1664  * 	@dev_id:		Used to differentiate devices that share
1665  * 				the same link layer address
1666  * 	@dev_port:		Used to differentiate devices that share
1667  * 				the same function
1668  *	@addr_list_lock:	XXX: need comments on this one
1669  *	@uc_promisc:		Counter that indicates promiscuous mode
1670  *				has been enabled due to the need to listen to
1671  *				additional unicast addresses in a device that
1672  *				does not implement ndo_set_rx_mode()
1673  *	@uc:			unicast mac addresses
1674  *	@mc:			multicast mac addresses
1675  *	@dev_addrs:		list of device hw addresses
1676  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1677  *	@promiscuity:		Number of times the NIC is told to work in
1678  *				promiscuous mode; if it becomes 0 the NIC will
1679  *				exit promiscuous mode
1680  *	@allmulti:		Counter, enables or disables allmulticast mode
1681  *
1682  *	@vlan_info:	VLAN info
1683  *	@dsa_ptr:	dsa specific data
1684  *	@tipc_ptr:	TIPC specific data
1685  *	@atalk_ptr:	AppleTalk link
1686  *	@ip_ptr:	IPv4 specific data
1687  *	@dn_ptr:	DECnet specific data
1688  *	@ip6_ptr:	IPv6 specific data
1689  *	@ax25_ptr:	AX.25 specific data
1690  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1691  *
1692  *	@dev_addr:	Hw address (before bcast,
1693  *			because most packets are unicast)
1694  *
1695  *	@_rx:			Array of RX queues
1696  *	@num_rx_queues:		Number of RX queues
1697  *				allocated at register_netdev() time
1698  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1699  *
1700  *	@rx_handler:		handler for received packets
1701  *	@rx_handler_data: 	XXX: need comments on this one
1702  *	@miniq_ingress:		ingress/clsact qdisc specific data for
1703  *				ingress processing
1704  *	@ingress_queue:		XXX: need comments on this one
1705  *	@broadcast:		hw bcast address
1706  *
1707  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1708  *			indexed by RX queue number. Assigned by driver.
1709  *			This must only be set if the ndo_rx_flow_steer
1710  *			operation is defined
1711  *	@index_hlist:		Device index hash chain
1712  *
1713  *	@_tx:			Array of TX queues
1714  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1715  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1716  *	@qdisc:			Root qdisc from userspace point of view
1717  *	@tx_queue_len:		Max frames per queue allowed
1718  *	@tx_global_lock: 	XXX: need comments on this one
1719  *
1720  *	@xps_maps:	XXX: need comments on this one
1721  *	@miniq_egress:		clsact qdisc specific data for
1722  *				egress processing
1723  *	@watchdog_timeo:	Represents the timeout that is used by
1724  *				the watchdog (see dev_watchdog())
1725  *	@watchdog_timer:	List of timers
1726  *
1727  *	@pcpu_refcnt:		Number of references to this device
1728  *	@todo_list:		Delayed register/unregister
1729  *	@link_watch_list:	XXX: need comments on this one
1730  *
1731  *	@reg_state:		Register/unregister state machine
1732  *	@dismantle:		Device is going to be freed
1733  *	@rtnl_link_state:	This enum represents the phases of creating
1734  *				a new link
1735  *
1736  *	@needs_free_netdev:	Should unregister perform free_netdev?
1737  *	@priv_destructor:	Called from unregister
1738  *	@npinfo:		XXX: need comments on this one
1739  * 	@nd_net:		Network namespace this network device is inside
1740  *
1741  * 	@ml_priv:	Mid-layer private
1742  * 	@lstats:	Loopback statistics
1743  * 	@tstats:	Tunnel statistics
1744  * 	@dstats:	Dummy statistics
1745  * 	@vstats:	Virtual ethernet statistics
1746  *
1747  *	@garp_port:	GARP
1748  *	@mrp_port:	MRP
1749  *
1750  *	@dev:		Class/net/name entry
1751  *	@sysfs_groups:	Space for optional device, statistics and wireless
1752  *			sysfs groups
1753  *
1754  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1755  *	@rtnl_link_ops:	Rtnl_link_ops
1756  *
1757  *	@gso_max_size:	Maximum size of generic segmentation offload
1758  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1759  *			NIC for GSO
1760  *
1761  *	@dcbnl_ops:	Data Center Bridging netlink ops
1762  *	@num_tc:	Number of traffic classes in the net device
1763  *	@tc_to_txq:	XXX: need comments on this one
1764  *	@prio_tc_map:	XXX: need comments on this one
1765  *
1766  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1767  *
1768  *	@priomap:	XXX: need comments on this one
1769  *	@phydev:	Physical device may attach itself
1770  *			for hardware timestamping
1771  *	@sfp_bus:	attached &struct sfp_bus structure.
1772  *	@qdisc_tx_busylock_key: lockdep class annotating Qdisc->busylock
1773 				spinlock
1774  *	@qdisc_running_key:	lockdep class annotating Qdisc->running seqcount
1775  *	@qdisc_xmit_lock_key:	lockdep class annotating
1776  *				netdev_queue->_xmit_lock spinlock
1777  *	@addr_list_lock_key:	lockdep class annotating
1778  *				net_device->addr_list_lock spinlock
1779  *
1780  *	@proto_down:	protocol port state information can be sent to the
1781  *			switch driver and used to set the phys state of the
1782  *			switch port.
1783  *
1784  *	@wol_enabled:	Wake-on-LAN is enabled
1785  *
1786  *	FIXME: cleanup struct net_device such that network protocol info
1787  *	moves out.
1788  */
1789 
1790 struct net_device {
1791 	char			name[IFNAMSIZ];
1792 	struct netdev_name_node	*name_node;
1793 	struct dev_ifalias	__rcu *ifalias;
1794 	/*
1795 	 *	I/O specific fields
1796 	 *	FIXME: Merge these and struct ifmap into one
1797 	 */
1798 	unsigned long		mem_end;
1799 	unsigned long		mem_start;
1800 	unsigned long		base_addr;
1801 	int			irq;
1802 
1803 	/*
1804 	 *	Some hardware also needs these fields (state,dev_list,
1805 	 *	napi_list,unreg_list,close_list) but they are not
1806 	 *	part of the usual set specified in Space.c.
1807 	 */
1808 
1809 	unsigned long		state;
1810 
1811 	struct list_head	dev_list;
1812 	struct list_head	napi_list;
1813 	struct list_head	unreg_list;
1814 	struct list_head	close_list;
1815 	struct list_head	ptype_all;
1816 	struct list_head	ptype_specific;
1817 
1818 	struct {
1819 		struct list_head upper;
1820 		struct list_head lower;
1821 	} adj_list;
1822 
1823 	netdev_features_t	features;
1824 	netdev_features_t	hw_features;
1825 	netdev_features_t	wanted_features;
1826 	netdev_features_t	vlan_features;
1827 	netdev_features_t	hw_enc_features;
1828 	netdev_features_t	mpls_features;
1829 	netdev_features_t	gso_partial_features;
1830 
1831 	int			ifindex;
1832 	int			group;
1833 
1834 	struct net_device_stats	stats;
1835 
1836 	atomic_long_t		rx_dropped;
1837 	atomic_long_t		tx_dropped;
1838 	atomic_long_t		rx_nohandler;
1839 
1840 	/* Stats to monitor link on/off, flapping */
1841 	atomic_t		carrier_up_count;
1842 	atomic_t		carrier_down_count;
1843 
1844 #ifdef CONFIG_WIRELESS_EXT
1845 	const struct iw_handler_def *wireless_handlers;
1846 	struct iw_public_data	*wireless_data;
1847 #endif
1848 	const struct net_device_ops *netdev_ops;
1849 	const struct ethtool_ops *ethtool_ops;
1850 #ifdef CONFIG_NET_L3_MASTER_DEV
1851 	const struct l3mdev_ops	*l3mdev_ops;
1852 #endif
1853 #if IS_ENABLED(CONFIG_IPV6)
1854 	const struct ndisc_ops *ndisc_ops;
1855 #endif
1856 
1857 #ifdef CONFIG_XFRM_OFFLOAD
1858 	const struct xfrmdev_ops *xfrmdev_ops;
1859 #endif
1860 
1861 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1862 	const struct tlsdev_ops *tlsdev_ops;
1863 #endif
1864 
1865 	const struct header_ops *header_ops;
1866 
1867 	unsigned int		flags;
1868 	unsigned int		priv_flags;
1869 
1870 	unsigned short		gflags;
1871 	unsigned short		padded;
1872 
1873 	unsigned char		operstate;
1874 	unsigned char		link_mode;
1875 
1876 	unsigned char		if_port;
1877 	unsigned char		dma;
1878 
1879 	unsigned int		mtu;
1880 	unsigned int		min_mtu;
1881 	unsigned int		max_mtu;
1882 	unsigned short		type;
1883 	unsigned short		hard_header_len;
1884 	unsigned char		min_header_len;
1885 
1886 	unsigned short		needed_headroom;
1887 	unsigned short		needed_tailroom;
1888 
1889 	/* Interface address info. */
1890 	unsigned char		perm_addr[MAX_ADDR_LEN];
1891 	unsigned char		addr_assign_type;
1892 	unsigned char		addr_len;
1893 	unsigned char		upper_level;
1894 	unsigned char		lower_level;
1895 	unsigned short		neigh_priv_len;
1896 	unsigned short          dev_id;
1897 	unsigned short          dev_port;
1898 	spinlock_t		addr_list_lock;
1899 	unsigned char		name_assign_type;
1900 	bool			uc_promisc;
1901 	struct netdev_hw_addr_list	uc;
1902 	struct netdev_hw_addr_list	mc;
1903 	struct netdev_hw_addr_list	dev_addrs;
1904 
1905 #ifdef CONFIG_SYSFS
1906 	struct kset		*queues_kset;
1907 #endif
1908 	unsigned int		promiscuity;
1909 	unsigned int		allmulti;
1910 
1911 
1912 	/* Protocol-specific pointers */
1913 
1914 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1915 	struct vlan_info __rcu	*vlan_info;
1916 #endif
1917 #if IS_ENABLED(CONFIG_NET_DSA)
1918 	struct dsa_port		*dsa_ptr;
1919 #endif
1920 #if IS_ENABLED(CONFIG_TIPC)
1921 	struct tipc_bearer __rcu *tipc_ptr;
1922 #endif
1923 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1924 	void 			*atalk_ptr;
1925 #endif
1926 	struct in_device __rcu	*ip_ptr;
1927 #if IS_ENABLED(CONFIG_DECNET)
1928 	struct dn_dev __rcu     *dn_ptr;
1929 #endif
1930 	struct inet6_dev __rcu	*ip6_ptr;
1931 #if IS_ENABLED(CONFIG_AX25)
1932 	void			*ax25_ptr;
1933 #endif
1934 	struct wireless_dev	*ieee80211_ptr;
1935 	struct wpan_dev		*ieee802154_ptr;
1936 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1937 	struct mpls_dev __rcu	*mpls_ptr;
1938 #endif
1939 
1940 /*
1941  * Cache lines mostly used on receive path (including eth_type_trans())
1942  */
1943 	/* Interface address info used in eth_type_trans() */
1944 	unsigned char		*dev_addr;
1945 
1946 	struct netdev_rx_queue	*_rx;
1947 	unsigned int		num_rx_queues;
1948 	unsigned int		real_num_rx_queues;
1949 
1950 	struct bpf_prog __rcu	*xdp_prog;
1951 	unsigned long		gro_flush_timeout;
1952 	rx_handler_func_t __rcu	*rx_handler;
1953 	void __rcu		*rx_handler_data;
1954 
1955 #ifdef CONFIG_NET_CLS_ACT
1956 	struct mini_Qdisc __rcu	*miniq_ingress;
1957 #endif
1958 	struct netdev_queue __rcu *ingress_queue;
1959 #ifdef CONFIG_NETFILTER_INGRESS
1960 	struct nf_hook_entries __rcu *nf_hooks_ingress;
1961 #endif
1962 
1963 	unsigned char		broadcast[MAX_ADDR_LEN];
1964 #ifdef CONFIG_RFS_ACCEL
1965 	struct cpu_rmap		*rx_cpu_rmap;
1966 #endif
1967 	struct hlist_node	index_hlist;
1968 
1969 /*
1970  * Cache lines mostly used on transmit path
1971  */
1972 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1973 	unsigned int		num_tx_queues;
1974 	unsigned int		real_num_tx_queues;
1975 	struct Qdisc		*qdisc;
1976 #ifdef CONFIG_NET_SCHED
1977 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1978 #endif
1979 	unsigned int		tx_queue_len;
1980 	spinlock_t		tx_global_lock;
1981 	int			watchdog_timeo;
1982 
1983 #ifdef CONFIG_XPS
1984 	struct xps_dev_maps __rcu *xps_cpus_map;
1985 	struct xps_dev_maps __rcu *xps_rxqs_map;
1986 #endif
1987 #ifdef CONFIG_NET_CLS_ACT
1988 	struct mini_Qdisc __rcu	*miniq_egress;
1989 #endif
1990 
1991 	/* These may be needed for future network-power-down code. */
1992 	struct timer_list	watchdog_timer;
1993 
1994 	int __percpu		*pcpu_refcnt;
1995 	struct list_head	todo_list;
1996 
1997 	struct list_head	link_watch_list;
1998 
1999 	enum { NETREG_UNINITIALIZED=0,
2000 	       NETREG_REGISTERED,	/* completed register_netdevice */
2001 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
2002 	       NETREG_UNREGISTERED,	/* completed unregister todo */
2003 	       NETREG_RELEASED,		/* called free_netdev */
2004 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
2005 	} reg_state:8;
2006 
2007 	bool dismantle;
2008 
2009 	enum {
2010 		RTNL_LINK_INITIALIZED,
2011 		RTNL_LINK_INITIALIZING,
2012 	} rtnl_link_state:16;
2013 
2014 	bool needs_free_netdev;
2015 	void (*priv_destructor)(struct net_device *dev);
2016 
2017 #ifdef CONFIG_NETPOLL
2018 	struct netpoll_info __rcu	*npinfo;
2019 #endif
2020 
2021 	possible_net_t			nd_net;
2022 
2023 	/* mid-layer private */
2024 	union {
2025 		void					*ml_priv;
2026 		struct pcpu_lstats __percpu		*lstats;
2027 		struct pcpu_sw_netstats __percpu	*tstats;
2028 		struct pcpu_dstats __percpu		*dstats;
2029 	};
2030 
2031 #if IS_ENABLED(CONFIG_GARP)
2032 	struct garp_port __rcu	*garp_port;
2033 #endif
2034 #if IS_ENABLED(CONFIG_MRP)
2035 	struct mrp_port __rcu	*mrp_port;
2036 #endif
2037 
2038 	struct device		dev;
2039 	const struct attribute_group *sysfs_groups[4];
2040 	const struct attribute_group *sysfs_rx_queue_group;
2041 
2042 	const struct rtnl_link_ops *rtnl_link_ops;
2043 
2044 	/* for setting kernel sock attribute on TCP connection setup */
2045 #define GSO_MAX_SIZE		65536
2046 	unsigned int		gso_max_size;
2047 #define GSO_MAX_SEGS		65535
2048 	u16			gso_max_segs;
2049 
2050 #ifdef CONFIG_DCB
2051 	const struct dcbnl_rtnl_ops *dcbnl_ops;
2052 #endif
2053 	s16			num_tc;
2054 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
2055 	u8			prio_tc_map[TC_BITMASK + 1];
2056 
2057 #if IS_ENABLED(CONFIG_FCOE)
2058 	unsigned int		fcoe_ddp_xid;
2059 #endif
2060 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2061 	struct netprio_map __rcu *priomap;
2062 #endif
2063 	struct phy_device	*phydev;
2064 	struct sfp_bus		*sfp_bus;
2065 	struct lock_class_key	qdisc_tx_busylock_key;
2066 	struct lock_class_key	qdisc_running_key;
2067 	struct lock_class_key	qdisc_xmit_lock_key;
2068 	struct lock_class_key	addr_list_lock_key;
2069 	bool			proto_down;
2070 	unsigned		wol_enabled:1;
2071 };
2072 #define to_net_dev(d) container_of(d, struct net_device, dev)
2073 
2074 static inline bool netif_elide_gro(const struct net_device *dev)
2075 {
2076 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2077 		return true;
2078 	return false;
2079 }
2080 
2081 #define	NETDEV_ALIGN		32
2082 
2083 static inline
2084 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2085 {
2086 	return dev->prio_tc_map[prio & TC_BITMASK];
2087 }
2088 
2089 static inline
2090 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2091 {
2092 	if (tc >= dev->num_tc)
2093 		return -EINVAL;
2094 
2095 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2096 	return 0;
2097 }
2098 
2099 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2100 void netdev_reset_tc(struct net_device *dev);
2101 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2102 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2103 
2104 static inline
2105 int netdev_get_num_tc(struct net_device *dev)
2106 {
2107 	return dev->num_tc;
2108 }
2109 
2110 void netdev_unbind_sb_channel(struct net_device *dev,
2111 			      struct net_device *sb_dev);
2112 int netdev_bind_sb_channel_queue(struct net_device *dev,
2113 				 struct net_device *sb_dev,
2114 				 u8 tc, u16 count, u16 offset);
2115 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2116 static inline int netdev_get_sb_channel(struct net_device *dev)
2117 {
2118 	return max_t(int, -dev->num_tc, 0);
2119 }
2120 
2121 static inline
2122 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2123 					 unsigned int index)
2124 {
2125 	return &dev->_tx[index];
2126 }
2127 
2128 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2129 						    const struct sk_buff *skb)
2130 {
2131 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2132 }
2133 
2134 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2135 					    void (*f)(struct net_device *,
2136 						      struct netdev_queue *,
2137 						      void *),
2138 					    void *arg)
2139 {
2140 	unsigned int i;
2141 
2142 	for (i = 0; i < dev->num_tx_queues; i++)
2143 		f(dev, &dev->_tx[i], arg);
2144 }
2145 
2146 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2147 		     struct net_device *sb_dev);
2148 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2149 					 struct sk_buff *skb,
2150 					 struct net_device *sb_dev);
2151 
2152 /* returns the headroom that the master device needs to take in account
2153  * when forwarding to this dev
2154  */
2155 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2156 {
2157 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2158 }
2159 
2160 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2161 {
2162 	if (dev->netdev_ops->ndo_set_rx_headroom)
2163 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2164 }
2165 
2166 /* set the device rx headroom to the dev's default */
2167 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2168 {
2169 	netdev_set_rx_headroom(dev, -1);
2170 }
2171 
2172 /*
2173  * Net namespace inlines
2174  */
2175 static inline
2176 struct net *dev_net(const struct net_device *dev)
2177 {
2178 	return read_pnet(&dev->nd_net);
2179 }
2180 
2181 static inline
2182 void dev_net_set(struct net_device *dev, struct net *net)
2183 {
2184 	write_pnet(&dev->nd_net, net);
2185 }
2186 
2187 /**
2188  *	netdev_priv - access network device private data
2189  *	@dev: network device
2190  *
2191  * Get network device private data
2192  */
2193 static inline void *netdev_priv(const struct net_device *dev)
2194 {
2195 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2196 }
2197 
2198 /* Set the sysfs physical device reference for the network logical device
2199  * if set prior to registration will cause a symlink during initialization.
2200  */
2201 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2202 
2203 /* Set the sysfs device type for the network logical device to allow
2204  * fine-grained identification of different network device types. For
2205  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2206  */
2207 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2208 
2209 /* Default NAPI poll() weight
2210  * Device drivers are strongly advised to not use bigger value
2211  */
2212 #define NAPI_POLL_WEIGHT 64
2213 
2214 /**
2215  *	netif_napi_add - initialize a NAPI context
2216  *	@dev:  network device
2217  *	@napi: NAPI context
2218  *	@poll: polling function
2219  *	@weight: default weight
2220  *
2221  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2222  * *any* of the other NAPI-related functions.
2223  */
2224 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2225 		    int (*poll)(struct napi_struct *, int), int weight);
2226 
2227 /**
2228  *	netif_tx_napi_add - initialize a NAPI context
2229  *	@dev:  network device
2230  *	@napi: NAPI context
2231  *	@poll: polling function
2232  *	@weight: default weight
2233  *
2234  * This variant of netif_napi_add() should be used from drivers using NAPI
2235  * to exclusively poll a TX queue.
2236  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2237  */
2238 static inline void netif_tx_napi_add(struct net_device *dev,
2239 				     struct napi_struct *napi,
2240 				     int (*poll)(struct napi_struct *, int),
2241 				     int weight)
2242 {
2243 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2244 	netif_napi_add(dev, napi, poll, weight);
2245 }
2246 
2247 /**
2248  *  netif_napi_del - remove a NAPI context
2249  *  @napi: NAPI context
2250  *
2251  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2252  */
2253 void netif_napi_del(struct napi_struct *napi);
2254 
2255 struct napi_gro_cb {
2256 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2257 	void	*frag0;
2258 
2259 	/* Length of frag0. */
2260 	unsigned int frag0_len;
2261 
2262 	/* This indicates where we are processing relative to skb->data. */
2263 	int	data_offset;
2264 
2265 	/* This is non-zero if the packet cannot be merged with the new skb. */
2266 	u16	flush;
2267 
2268 	/* Save the IP ID here and check when we get to the transport layer */
2269 	u16	flush_id;
2270 
2271 	/* Number of segments aggregated. */
2272 	u16	count;
2273 
2274 	/* Start offset for remote checksum offload */
2275 	u16	gro_remcsum_start;
2276 
2277 	/* jiffies when first packet was created/queued */
2278 	unsigned long age;
2279 
2280 	/* Used in ipv6_gro_receive() and foo-over-udp */
2281 	u16	proto;
2282 
2283 	/* This is non-zero if the packet may be of the same flow. */
2284 	u8	same_flow:1;
2285 
2286 	/* Used in tunnel GRO receive */
2287 	u8	encap_mark:1;
2288 
2289 	/* GRO checksum is valid */
2290 	u8	csum_valid:1;
2291 
2292 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2293 	u8	csum_cnt:3;
2294 
2295 	/* Free the skb? */
2296 	u8	free:2;
2297 #define NAPI_GRO_FREE		  1
2298 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2299 
2300 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2301 	u8	is_ipv6:1;
2302 
2303 	/* Used in GRE, set in fou/gue_gro_receive */
2304 	u8	is_fou:1;
2305 
2306 	/* Used to determine if flush_id can be ignored */
2307 	u8	is_atomic:1;
2308 
2309 	/* Number of gro_receive callbacks this packet already went through */
2310 	u8 recursion_counter:4;
2311 
2312 	/* 1 bit hole */
2313 
2314 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2315 	__wsum	csum;
2316 
2317 	/* used in skb_gro_receive() slow path */
2318 	struct sk_buff *last;
2319 };
2320 
2321 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2322 
2323 #define GRO_RECURSION_LIMIT 15
2324 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2325 {
2326 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2327 }
2328 
2329 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2330 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2331 					       struct list_head *head,
2332 					       struct sk_buff *skb)
2333 {
2334 	if (unlikely(gro_recursion_inc_test(skb))) {
2335 		NAPI_GRO_CB(skb)->flush |= 1;
2336 		return NULL;
2337 	}
2338 
2339 	return cb(head, skb);
2340 }
2341 
2342 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2343 					    struct sk_buff *);
2344 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2345 						  struct sock *sk,
2346 						  struct list_head *head,
2347 						  struct sk_buff *skb)
2348 {
2349 	if (unlikely(gro_recursion_inc_test(skb))) {
2350 		NAPI_GRO_CB(skb)->flush |= 1;
2351 		return NULL;
2352 	}
2353 
2354 	return cb(sk, head, skb);
2355 }
2356 
2357 struct packet_type {
2358 	__be16			type;	/* This is really htons(ether_type). */
2359 	bool			ignore_outgoing;
2360 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2361 	int			(*func) (struct sk_buff *,
2362 					 struct net_device *,
2363 					 struct packet_type *,
2364 					 struct net_device *);
2365 	void			(*list_func) (struct list_head *,
2366 					      struct packet_type *,
2367 					      struct net_device *);
2368 	bool			(*id_match)(struct packet_type *ptype,
2369 					    struct sock *sk);
2370 	void			*af_packet_priv;
2371 	struct list_head	list;
2372 };
2373 
2374 struct offload_callbacks {
2375 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2376 						netdev_features_t features);
2377 	struct sk_buff		*(*gro_receive)(struct list_head *head,
2378 						struct sk_buff *skb);
2379 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2380 };
2381 
2382 struct packet_offload {
2383 	__be16			 type;	/* This is really htons(ether_type). */
2384 	u16			 priority;
2385 	struct offload_callbacks callbacks;
2386 	struct list_head	 list;
2387 };
2388 
2389 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2390 struct pcpu_sw_netstats {
2391 	u64     rx_packets;
2392 	u64     rx_bytes;
2393 	u64     tx_packets;
2394 	u64     tx_bytes;
2395 	struct u64_stats_sync   syncp;
2396 } __aligned(4 * sizeof(u64));
2397 
2398 struct pcpu_lstats {
2399 	u64 packets;
2400 	u64 bytes;
2401 	struct u64_stats_sync syncp;
2402 } __aligned(2 * sizeof(u64));
2403 
2404 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2405 ({									\
2406 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2407 	if (pcpu_stats)	{						\
2408 		int __cpu;						\
2409 		for_each_possible_cpu(__cpu) {				\
2410 			typeof(type) *stat;				\
2411 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2412 			u64_stats_init(&stat->syncp);			\
2413 		}							\
2414 	}								\
2415 	pcpu_stats;							\
2416 })
2417 
2418 #define netdev_alloc_pcpu_stats(type)					\
2419 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2420 
2421 enum netdev_lag_tx_type {
2422 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2423 	NETDEV_LAG_TX_TYPE_RANDOM,
2424 	NETDEV_LAG_TX_TYPE_BROADCAST,
2425 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2426 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2427 	NETDEV_LAG_TX_TYPE_HASH,
2428 };
2429 
2430 enum netdev_lag_hash {
2431 	NETDEV_LAG_HASH_NONE,
2432 	NETDEV_LAG_HASH_L2,
2433 	NETDEV_LAG_HASH_L34,
2434 	NETDEV_LAG_HASH_L23,
2435 	NETDEV_LAG_HASH_E23,
2436 	NETDEV_LAG_HASH_E34,
2437 	NETDEV_LAG_HASH_UNKNOWN,
2438 };
2439 
2440 struct netdev_lag_upper_info {
2441 	enum netdev_lag_tx_type tx_type;
2442 	enum netdev_lag_hash hash_type;
2443 };
2444 
2445 struct netdev_lag_lower_state_info {
2446 	u8 link_up : 1,
2447 	   tx_enabled : 1;
2448 };
2449 
2450 #include <linux/notifier.h>
2451 
2452 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2453  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2454  * adding new types.
2455  */
2456 enum netdev_cmd {
2457 	NETDEV_UP	= 1,	/* For now you can't veto a device up/down */
2458 	NETDEV_DOWN,
2459 	NETDEV_REBOOT,		/* Tell a protocol stack a network interface
2460 				   detected a hardware crash and restarted
2461 				   - we can use this eg to kick tcp sessions
2462 				   once done */
2463 	NETDEV_CHANGE,		/* Notify device state change */
2464 	NETDEV_REGISTER,
2465 	NETDEV_UNREGISTER,
2466 	NETDEV_CHANGEMTU,	/* notify after mtu change happened */
2467 	NETDEV_CHANGEADDR,	/* notify after the address change */
2468 	NETDEV_PRE_CHANGEADDR,	/* notify before the address change */
2469 	NETDEV_GOING_DOWN,
2470 	NETDEV_CHANGENAME,
2471 	NETDEV_FEAT_CHANGE,
2472 	NETDEV_BONDING_FAILOVER,
2473 	NETDEV_PRE_UP,
2474 	NETDEV_PRE_TYPE_CHANGE,
2475 	NETDEV_POST_TYPE_CHANGE,
2476 	NETDEV_POST_INIT,
2477 	NETDEV_RELEASE,
2478 	NETDEV_NOTIFY_PEERS,
2479 	NETDEV_JOIN,
2480 	NETDEV_CHANGEUPPER,
2481 	NETDEV_RESEND_IGMP,
2482 	NETDEV_PRECHANGEMTU,	/* notify before mtu change happened */
2483 	NETDEV_CHANGEINFODATA,
2484 	NETDEV_BONDING_INFO,
2485 	NETDEV_PRECHANGEUPPER,
2486 	NETDEV_CHANGELOWERSTATE,
2487 	NETDEV_UDP_TUNNEL_PUSH_INFO,
2488 	NETDEV_UDP_TUNNEL_DROP_INFO,
2489 	NETDEV_CHANGE_TX_QUEUE_LEN,
2490 	NETDEV_CVLAN_FILTER_PUSH_INFO,
2491 	NETDEV_CVLAN_FILTER_DROP_INFO,
2492 	NETDEV_SVLAN_FILTER_PUSH_INFO,
2493 	NETDEV_SVLAN_FILTER_DROP_INFO,
2494 };
2495 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2496 
2497 int register_netdevice_notifier(struct notifier_block *nb);
2498 int unregister_netdevice_notifier(struct notifier_block *nb);
2499 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2500 int unregister_netdevice_notifier_net(struct net *net,
2501 				      struct notifier_block *nb);
2502 
2503 struct netdev_notifier_info {
2504 	struct net_device	*dev;
2505 	struct netlink_ext_ack	*extack;
2506 };
2507 
2508 struct netdev_notifier_info_ext {
2509 	struct netdev_notifier_info info; /* must be first */
2510 	union {
2511 		u32 mtu;
2512 	} ext;
2513 };
2514 
2515 struct netdev_notifier_change_info {
2516 	struct netdev_notifier_info info; /* must be first */
2517 	unsigned int flags_changed;
2518 };
2519 
2520 struct netdev_notifier_changeupper_info {
2521 	struct netdev_notifier_info info; /* must be first */
2522 	struct net_device *upper_dev; /* new upper dev */
2523 	bool master; /* is upper dev master */
2524 	bool linking; /* is the notification for link or unlink */
2525 	void *upper_info; /* upper dev info */
2526 };
2527 
2528 struct netdev_notifier_changelowerstate_info {
2529 	struct netdev_notifier_info info; /* must be first */
2530 	void *lower_state_info; /* is lower dev state */
2531 };
2532 
2533 struct netdev_notifier_pre_changeaddr_info {
2534 	struct netdev_notifier_info info; /* must be first */
2535 	const unsigned char *dev_addr;
2536 };
2537 
2538 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2539 					     struct net_device *dev)
2540 {
2541 	info->dev = dev;
2542 	info->extack = NULL;
2543 }
2544 
2545 static inline struct net_device *
2546 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2547 {
2548 	return info->dev;
2549 }
2550 
2551 static inline struct netlink_ext_ack *
2552 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2553 {
2554 	return info->extack;
2555 }
2556 
2557 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2558 
2559 
2560 extern rwlock_t				dev_base_lock;		/* Device list lock */
2561 
2562 #define for_each_netdev(net, d)		\
2563 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2564 #define for_each_netdev_reverse(net, d)	\
2565 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2566 #define for_each_netdev_rcu(net, d)		\
2567 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2568 #define for_each_netdev_safe(net, d, n)	\
2569 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2570 #define for_each_netdev_continue(net, d)		\
2571 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2572 #define for_each_netdev_continue_reverse(net, d)		\
2573 		list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
2574 						     dev_list)
2575 #define for_each_netdev_continue_rcu(net, d)		\
2576 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2577 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2578 		for_each_netdev_rcu(&init_net, slave)	\
2579 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2580 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2581 
2582 static inline struct net_device *next_net_device(struct net_device *dev)
2583 {
2584 	struct list_head *lh;
2585 	struct net *net;
2586 
2587 	net = dev_net(dev);
2588 	lh = dev->dev_list.next;
2589 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2590 }
2591 
2592 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2593 {
2594 	struct list_head *lh;
2595 	struct net *net;
2596 
2597 	net = dev_net(dev);
2598 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2599 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2600 }
2601 
2602 static inline struct net_device *first_net_device(struct net *net)
2603 {
2604 	return list_empty(&net->dev_base_head) ? NULL :
2605 		net_device_entry(net->dev_base_head.next);
2606 }
2607 
2608 static inline struct net_device *first_net_device_rcu(struct net *net)
2609 {
2610 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2611 
2612 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2613 }
2614 
2615 int netdev_boot_setup_check(struct net_device *dev);
2616 unsigned long netdev_boot_base(const char *prefix, int unit);
2617 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2618 				       const char *hwaddr);
2619 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2620 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2621 void dev_add_pack(struct packet_type *pt);
2622 void dev_remove_pack(struct packet_type *pt);
2623 void __dev_remove_pack(struct packet_type *pt);
2624 void dev_add_offload(struct packet_offload *po);
2625 void dev_remove_offload(struct packet_offload *po);
2626 
2627 int dev_get_iflink(const struct net_device *dev);
2628 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2629 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2630 				      unsigned short mask);
2631 struct net_device *dev_get_by_name(struct net *net, const char *name);
2632 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2633 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2634 int dev_alloc_name(struct net_device *dev, const char *name);
2635 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
2636 void dev_close(struct net_device *dev);
2637 void dev_close_many(struct list_head *head, bool unlink);
2638 void dev_disable_lro(struct net_device *dev);
2639 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2640 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2641 		     struct net_device *sb_dev);
2642 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2643 		       struct net_device *sb_dev);
2644 int dev_queue_xmit(struct sk_buff *skb);
2645 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2646 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2647 int register_netdevice(struct net_device *dev);
2648 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2649 void unregister_netdevice_many(struct list_head *head);
2650 static inline void unregister_netdevice(struct net_device *dev)
2651 {
2652 	unregister_netdevice_queue(dev, NULL);
2653 }
2654 
2655 int netdev_refcnt_read(const struct net_device *dev);
2656 void free_netdev(struct net_device *dev);
2657 void netdev_freemem(struct net_device *dev);
2658 void synchronize_net(void);
2659 int init_dummy_netdev(struct net_device *dev);
2660 
2661 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2662 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2663 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2664 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2665 int netdev_get_name(struct net *net, char *name, int ifindex);
2666 int dev_restart(struct net_device *dev);
2667 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2668 
2669 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2670 {
2671 	return NAPI_GRO_CB(skb)->data_offset;
2672 }
2673 
2674 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2675 {
2676 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2677 }
2678 
2679 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2680 {
2681 	NAPI_GRO_CB(skb)->data_offset += len;
2682 }
2683 
2684 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2685 					unsigned int offset)
2686 {
2687 	return NAPI_GRO_CB(skb)->frag0 + offset;
2688 }
2689 
2690 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2691 {
2692 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2693 }
2694 
2695 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2696 {
2697 	NAPI_GRO_CB(skb)->frag0 = NULL;
2698 	NAPI_GRO_CB(skb)->frag0_len = 0;
2699 }
2700 
2701 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2702 					unsigned int offset)
2703 {
2704 	if (!pskb_may_pull(skb, hlen))
2705 		return NULL;
2706 
2707 	skb_gro_frag0_invalidate(skb);
2708 	return skb->data + offset;
2709 }
2710 
2711 static inline void *skb_gro_network_header(struct sk_buff *skb)
2712 {
2713 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2714 	       skb_network_offset(skb);
2715 }
2716 
2717 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2718 					const void *start, unsigned int len)
2719 {
2720 	if (NAPI_GRO_CB(skb)->csum_valid)
2721 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2722 						  csum_partial(start, len, 0));
2723 }
2724 
2725 /* GRO checksum functions. These are logical equivalents of the normal
2726  * checksum functions (in skbuff.h) except that they operate on the GRO
2727  * offsets and fields in sk_buff.
2728  */
2729 
2730 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2731 
2732 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2733 {
2734 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2735 }
2736 
2737 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2738 						      bool zero_okay,
2739 						      __sum16 check)
2740 {
2741 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2742 		skb_checksum_start_offset(skb) <
2743 		 skb_gro_offset(skb)) &&
2744 		!skb_at_gro_remcsum_start(skb) &&
2745 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2746 		(!zero_okay || check));
2747 }
2748 
2749 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2750 							   __wsum psum)
2751 {
2752 	if (NAPI_GRO_CB(skb)->csum_valid &&
2753 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2754 		return 0;
2755 
2756 	NAPI_GRO_CB(skb)->csum = psum;
2757 
2758 	return __skb_gro_checksum_complete(skb);
2759 }
2760 
2761 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2762 {
2763 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2764 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2765 		NAPI_GRO_CB(skb)->csum_cnt--;
2766 	} else {
2767 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2768 		 * verified a new top level checksum or an encapsulated one
2769 		 * during GRO. This saves work if we fallback to normal path.
2770 		 */
2771 		__skb_incr_checksum_unnecessary(skb);
2772 	}
2773 }
2774 
2775 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2776 				    compute_pseudo)			\
2777 ({									\
2778 	__sum16 __ret = 0;						\
2779 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2780 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2781 				compute_pseudo(skb, proto));		\
2782 	if (!__ret)							\
2783 		skb_gro_incr_csum_unnecessary(skb);			\
2784 	__ret;								\
2785 })
2786 
2787 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2788 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2789 
2790 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2791 					     compute_pseudo)		\
2792 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2793 
2794 #define skb_gro_checksum_simple_validate(skb)				\
2795 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2796 
2797 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2798 {
2799 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2800 		!NAPI_GRO_CB(skb)->csum_valid);
2801 }
2802 
2803 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2804 					      __sum16 check, __wsum pseudo)
2805 {
2806 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2807 	NAPI_GRO_CB(skb)->csum_valid = 1;
2808 }
2809 
2810 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2811 do {									\
2812 	if (__skb_gro_checksum_convert_check(skb))			\
2813 		__skb_gro_checksum_convert(skb, check,			\
2814 					   compute_pseudo(skb, proto));	\
2815 } while (0)
2816 
2817 struct gro_remcsum {
2818 	int offset;
2819 	__wsum delta;
2820 };
2821 
2822 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2823 {
2824 	grc->offset = 0;
2825 	grc->delta = 0;
2826 }
2827 
2828 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2829 					    unsigned int off, size_t hdrlen,
2830 					    int start, int offset,
2831 					    struct gro_remcsum *grc,
2832 					    bool nopartial)
2833 {
2834 	__wsum delta;
2835 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2836 
2837 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2838 
2839 	if (!nopartial) {
2840 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2841 		return ptr;
2842 	}
2843 
2844 	ptr = skb_gro_header_fast(skb, off);
2845 	if (skb_gro_header_hard(skb, off + plen)) {
2846 		ptr = skb_gro_header_slow(skb, off + plen, off);
2847 		if (!ptr)
2848 			return NULL;
2849 	}
2850 
2851 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2852 			       start, offset);
2853 
2854 	/* Adjust skb->csum since we changed the packet */
2855 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2856 
2857 	grc->offset = off + hdrlen + offset;
2858 	grc->delta = delta;
2859 
2860 	return ptr;
2861 }
2862 
2863 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2864 					   struct gro_remcsum *grc)
2865 {
2866 	void *ptr;
2867 	size_t plen = grc->offset + sizeof(u16);
2868 
2869 	if (!grc->delta)
2870 		return;
2871 
2872 	ptr = skb_gro_header_fast(skb, grc->offset);
2873 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2874 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2875 		if (!ptr)
2876 			return;
2877 	}
2878 
2879 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2880 }
2881 
2882 #ifdef CONFIG_XFRM_OFFLOAD
2883 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2884 {
2885 	if (PTR_ERR(pp) != -EINPROGRESS)
2886 		NAPI_GRO_CB(skb)->flush |= flush;
2887 }
2888 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2889 					       struct sk_buff *pp,
2890 					       int flush,
2891 					       struct gro_remcsum *grc)
2892 {
2893 	if (PTR_ERR(pp) != -EINPROGRESS) {
2894 		NAPI_GRO_CB(skb)->flush |= flush;
2895 		skb_gro_remcsum_cleanup(skb, grc);
2896 		skb->remcsum_offload = 0;
2897 	}
2898 }
2899 #else
2900 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2901 {
2902 	NAPI_GRO_CB(skb)->flush |= flush;
2903 }
2904 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2905 					       struct sk_buff *pp,
2906 					       int flush,
2907 					       struct gro_remcsum *grc)
2908 {
2909 	NAPI_GRO_CB(skb)->flush |= flush;
2910 	skb_gro_remcsum_cleanup(skb, grc);
2911 	skb->remcsum_offload = 0;
2912 }
2913 #endif
2914 
2915 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2916 				  unsigned short type,
2917 				  const void *daddr, const void *saddr,
2918 				  unsigned int len)
2919 {
2920 	if (!dev->header_ops || !dev->header_ops->create)
2921 		return 0;
2922 
2923 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2924 }
2925 
2926 static inline int dev_parse_header(const struct sk_buff *skb,
2927 				   unsigned char *haddr)
2928 {
2929 	const struct net_device *dev = skb->dev;
2930 
2931 	if (!dev->header_ops || !dev->header_ops->parse)
2932 		return 0;
2933 	return dev->header_ops->parse(skb, haddr);
2934 }
2935 
2936 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
2937 {
2938 	const struct net_device *dev = skb->dev;
2939 
2940 	if (!dev->header_ops || !dev->header_ops->parse_protocol)
2941 		return 0;
2942 	return dev->header_ops->parse_protocol(skb);
2943 }
2944 
2945 /* ll_header must have at least hard_header_len allocated */
2946 static inline bool dev_validate_header(const struct net_device *dev,
2947 				       char *ll_header, int len)
2948 {
2949 	if (likely(len >= dev->hard_header_len))
2950 		return true;
2951 	if (len < dev->min_header_len)
2952 		return false;
2953 
2954 	if (capable(CAP_SYS_RAWIO)) {
2955 		memset(ll_header + len, 0, dev->hard_header_len - len);
2956 		return true;
2957 	}
2958 
2959 	if (dev->header_ops && dev->header_ops->validate)
2960 		return dev->header_ops->validate(ll_header, len);
2961 
2962 	return false;
2963 }
2964 
2965 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2966 			   int len, int size);
2967 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2968 static inline int unregister_gifconf(unsigned int family)
2969 {
2970 	return register_gifconf(family, NULL);
2971 }
2972 
2973 #ifdef CONFIG_NET_FLOW_LIMIT
2974 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2975 struct sd_flow_limit {
2976 	u64			count;
2977 	unsigned int		num_buckets;
2978 	unsigned int		history_head;
2979 	u16			history[FLOW_LIMIT_HISTORY];
2980 	u8			buckets[];
2981 };
2982 
2983 extern int netdev_flow_limit_table_len;
2984 #endif /* CONFIG_NET_FLOW_LIMIT */
2985 
2986 /*
2987  * Incoming packets are placed on per-CPU queues
2988  */
2989 struct softnet_data {
2990 	struct list_head	poll_list;
2991 	struct sk_buff_head	process_queue;
2992 
2993 	/* stats */
2994 	unsigned int		processed;
2995 	unsigned int		time_squeeze;
2996 	unsigned int		received_rps;
2997 #ifdef CONFIG_RPS
2998 	struct softnet_data	*rps_ipi_list;
2999 #endif
3000 #ifdef CONFIG_NET_FLOW_LIMIT
3001 	struct sd_flow_limit __rcu *flow_limit;
3002 #endif
3003 	struct Qdisc		*output_queue;
3004 	struct Qdisc		**output_queue_tailp;
3005 	struct sk_buff		*completion_queue;
3006 #ifdef CONFIG_XFRM_OFFLOAD
3007 	struct sk_buff_head	xfrm_backlog;
3008 #endif
3009 	/* written and read only by owning cpu: */
3010 	struct {
3011 		u16 recursion;
3012 		u8  more;
3013 	} xmit;
3014 #ifdef CONFIG_RPS
3015 	/* input_queue_head should be written by cpu owning this struct,
3016 	 * and only read by other cpus. Worth using a cache line.
3017 	 */
3018 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
3019 
3020 	/* Elements below can be accessed between CPUs for RPS/RFS */
3021 	call_single_data_t	csd ____cacheline_aligned_in_smp;
3022 	struct softnet_data	*rps_ipi_next;
3023 	unsigned int		cpu;
3024 	unsigned int		input_queue_tail;
3025 #endif
3026 	unsigned int		dropped;
3027 	struct sk_buff_head	input_pkt_queue;
3028 	struct napi_struct	backlog;
3029 
3030 };
3031 
3032 static inline void input_queue_head_incr(struct softnet_data *sd)
3033 {
3034 #ifdef CONFIG_RPS
3035 	sd->input_queue_head++;
3036 #endif
3037 }
3038 
3039 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3040 					      unsigned int *qtail)
3041 {
3042 #ifdef CONFIG_RPS
3043 	*qtail = ++sd->input_queue_tail;
3044 #endif
3045 }
3046 
3047 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3048 
3049 static inline int dev_recursion_level(void)
3050 {
3051 	return this_cpu_read(softnet_data.xmit.recursion);
3052 }
3053 
3054 #define XMIT_RECURSION_LIMIT	10
3055 static inline bool dev_xmit_recursion(void)
3056 {
3057 	return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3058 			XMIT_RECURSION_LIMIT);
3059 }
3060 
3061 static inline void dev_xmit_recursion_inc(void)
3062 {
3063 	__this_cpu_inc(softnet_data.xmit.recursion);
3064 }
3065 
3066 static inline void dev_xmit_recursion_dec(void)
3067 {
3068 	__this_cpu_dec(softnet_data.xmit.recursion);
3069 }
3070 
3071 void __netif_schedule(struct Qdisc *q);
3072 void netif_schedule_queue(struct netdev_queue *txq);
3073 
3074 static inline void netif_tx_schedule_all(struct net_device *dev)
3075 {
3076 	unsigned int i;
3077 
3078 	for (i = 0; i < dev->num_tx_queues; i++)
3079 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
3080 }
3081 
3082 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3083 {
3084 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3085 }
3086 
3087 /**
3088  *	netif_start_queue - allow transmit
3089  *	@dev: network device
3090  *
3091  *	Allow upper layers to call the device hard_start_xmit routine.
3092  */
3093 static inline void netif_start_queue(struct net_device *dev)
3094 {
3095 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3096 }
3097 
3098 static inline void netif_tx_start_all_queues(struct net_device *dev)
3099 {
3100 	unsigned int i;
3101 
3102 	for (i = 0; i < dev->num_tx_queues; i++) {
3103 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3104 		netif_tx_start_queue(txq);
3105 	}
3106 }
3107 
3108 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3109 
3110 /**
3111  *	netif_wake_queue - restart transmit
3112  *	@dev: network device
3113  *
3114  *	Allow upper layers to call the device hard_start_xmit routine.
3115  *	Used for flow control when transmit resources are available.
3116  */
3117 static inline void netif_wake_queue(struct net_device *dev)
3118 {
3119 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3120 }
3121 
3122 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3123 {
3124 	unsigned int i;
3125 
3126 	for (i = 0; i < dev->num_tx_queues; i++) {
3127 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3128 		netif_tx_wake_queue(txq);
3129 	}
3130 }
3131 
3132 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3133 {
3134 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3135 }
3136 
3137 /**
3138  *	netif_stop_queue - stop transmitted packets
3139  *	@dev: network device
3140  *
3141  *	Stop upper layers calling the device hard_start_xmit routine.
3142  *	Used for flow control when transmit resources are unavailable.
3143  */
3144 static inline void netif_stop_queue(struct net_device *dev)
3145 {
3146 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3147 }
3148 
3149 void netif_tx_stop_all_queues(struct net_device *dev);
3150 void netdev_update_lockdep_key(struct net_device *dev);
3151 
3152 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3153 {
3154 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3155 }
3156 
3157 /**
3158  *	netif_queue_stopped - test if transmit queue is flowblocked
3159  *	@dev: network device
3160  *
3161  *	Test if transmit queue on device is currently unable to send.
3162  */
3163 static inline bool netif_queue_stopped(const struct net_device *dev)
3164 {
3165 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3166 }
3167 
3168 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3169 {
3170 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3171 }
3172 
3173 static inline bool
3174 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3175 {
3176 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3177 }
3178 
3179 static inline bool
3180 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3181 {
3182 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3183 }
3184 
3185 /**
3186  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3187  *	@dev_queue: pointer to transmit queue
3188  *
3189  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3190  * to give appropriate hint to the CPU.
3191  */
3192 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3193 {
3194 #ifdef CONFIG_BQL
3195 	prefetchw(&dev_queue->dql.num_queued);
3196 #endif
3197 }
3198 
3199 /**
3200  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3201  *	@dev_queue: pointer to transmit queue
3202  *
3203  * BQL enabled drivers might use this helper in their TX completion path,
3204  * to give appropriate hint to the CPU.
3205  */
3206 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3207 {
3208 #ifdef CONFIG_BQL
3209 	prefetchw(&dev_queue->dql.limit);
3210 #endif
3211 }
3212 
3213 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3214 					unsigned int bytes)
3215 {
3216 #ifdef CONFIG_BQL
3217 	dql_queued(&dev_queue->dql, bytes);
3218 
3219 	if (likely(dql_avail(&dev_queue->dql) >= 0))
3220 		return;
3221 
3222 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3223 
3224 	/*
3225 	 * The XOFF flag must be set before checking the dql_avail below,
3226 	 * because in netdev_tx_completed_queue we update the dql_completed
3227 	 * before checking the XOFF flag.
3228 	 */
3229 	smp_mb();
3230 
3231 	/* check again in case another CPU has just made room avail */
3232 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3233 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3234 #endif
3235 }
3236 
3237 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3238  * that they should not test BQL status themselves.
3239  * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3240  * skb of a batch.
3241  * Returns true if the doorbell must be used to kick the NIC.
3242  */
3243 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3244 					  unsigned int bytes,
3245 					  bool xmit_more)
3246 {
3247 	if (xmit_more) {
3248 #ifdef CONFIG_BQL
3249 		dql_queued(&dev_queue->dql, bytes);
3250 #endif
3251 		return netif_tx_queue_stopped(dev_queue);
3252 	}
3253 	netdev_tx_sent_queue(dev_queue, bytes);
3254 	return true;
3255 }
3256 
3257 /**
3258  * 	netdev_sent_queue - report the number of bytes queued to hardware
3259  * 	@dev: network device
3260  * 	@bytes: number of bytes queued to the hardware device queue
3261  *
3262  * 	Report the number of bytes queued for sending/completion to the network
3263  * 	device hardware queue. @bytes should be a good approximation and should
3264  * 	exactly match netdev_completed_queue() @bytes
3265  */
3266 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3267 {
3268 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3269 }
3270 
3271 static inline bool __netdev_sent_queue(struct net_device *dev,
3272 				       unsigned int bytes,
3273 				       bool xmit_more)
3274 {
3275 	return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3276 				      xmit_more);
3277 }
3278 
3279 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3280 					     unsigned int pkts, unsigned int bytes)
3281 {
3282 #ifdef CONFIG_BQL
3283 	if (unlikely(!bytes))
3284 		return;
3285 
3286 	dql_completed(&dev_queue->dql, bytes);
3287 
3288 	/*
3289 	 * Without the memory barrier there is a small possiblity that
3290 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3291 	 * be stopped forever
3292 	 */
3293 	smp_mb();
3294 
3295 	if (unlikely(dql_avail(&dev_queue->dql) < 0))
3296 		return;
3297 
3298 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3299 		netif_schedule_queue(dev_queue);
3300 #endif
3301 }
3302 
3303 /**
3304  * 	netdev_completed_queue - report bytes and packets completed by device
3305  * 	@dev: network device
3306  * 	@pkts: actual number of packets sent over the medium
3307  * 	@bytes: actual number of bytes sent over the medium
3308  *
3309  * 	Report the number of bytes and packets transmitted by the network device
3310  * 	hardware queue over the physical medium, @bytes must exactly match the
3311  * 	@bytes amount passed to netdev_sent_queue()
3312  */
3313 static inline void netdev_completed_queue(struct net_device *dev,
3314 					  unsigned int pkts, unsigned int bytes)
3315 {
3316 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3317 }
3318 
3319 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3320 {
3321 #ifdef CONFIG_BQL
3322 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3323 	dql_reset(&q->dql);
3324 #endif
3325 }
3326 
3327 /**
3328  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3329  * 	@dev_queue: network device
3330  *
3331  * 	Reset the bytes and packet count of a network device and clear the
3332  * 	software flow control OFF bit for this network device
3333  */
3334 static inline void netdev_reset_queue(struct net_device *dev_queue)
3335 {
3336 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3337 }
3338 
3339 /**
3340  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3341  * 	@dev: network device
3342  * 	@queue_index: given tx queue index
3343  *
3344  * 	Returns 0 if given tx queue index >= number of device tx queues,
3345  * 	otherwise returns the originally passed tx queue index.
3346  */
3347 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3348 {
3349 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3350 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3351 				     dev->name, queue_index,
3352 				     dev->real_num_tx_queues);
3353 		return 0;
3354 	}
3355 
3356 	return queue_index;
3357 }
3358 
3359 /**
3360  *	netif_running - test if up
3361  *	@dev: network device
3362  *
3363  *	Test if the device has been brought up.
3364  */
3365 static inline bool netif_running(const struct net_device *dev)
3366 {
3367 	return test_bit(__LINK_STATE_START, &dev->state);
3368 }
3369 
3370 /*
3371  * Routines to manage the subqueues on a device.  We only need start,
3372  * stop, and a check if it's stopped.  All other device management is
3373  * done at the overall netdevice level.
3374  * Also test the device if we're multiqueue.
3375  */
3376 
3377 /**
3378  *	netif_start_subqueue - allow sending packets on subqueue
3379  *	@dev: network device
3380  *	@queue_index: sub queue index
3381  *
3382  * Start individual transmit queue of a device with multiple transmit queues.
3383  */
3384 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3385 {
3386 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3387 
3388 	netif_tx_start_queue(txq);
3389 }
3390 
3391 /**
3392  *	netif_stop_subqueue - stop sending packets on subqueue
3393  *	@dev: network device
3394  *	@queue_index: sub queue index
3395  *
3396  * Stop individual transmit queue of a device with multiple transmit queues.
3397  */
3398 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3399 {
3400 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3401 	netif_tx_stop_queue(txq);
3402 }
3403 
3404 /**
3405  *	netif_subqueue_stopped - test status of subqueue
3406  *	@dev: network device
3407  *	@queue_index: sub queue index
3408  *
3409  * Check individual transmit queue of a device with multiple transmit queues.
3410  */
3411 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3412 					    u16 queue_index)
3413 {
3414 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3415 
3416 	return netif_tx_queue_stopped(txq);
3417 }
3418 
3419 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3420 					  struct sk_buff *skb)
3421 {
3422 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3423 }
3424 
3425 /**
3426  *	netif_wake_subqueue - allow sending packets on subqueue
3427  *	@dev: network device
3428  *	@queue_index: sub queue index
3429  *
3430  * Resume individual transmit queue of a device with multiple transmit queues.
3431  */
3432 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3433 {
3434 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3435 
3436 	netif_tx_wake_queue(txq);
3437 }
3438 
3439 #ifdef CONFIG_XPS
3440 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3441 			u16 index);
3442 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3443 			  u16 index, bool is_rxqs_map);
3444 
3445 /**
3446  *	netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3447  *	@j: CPU/Rx queue index
3448  *	@mask: bitmask of all cpus/rx queues
3449  *	@nr_bits: number of bits in the bitmask
3450  *
3451  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3452  */
3453 static inline bool netif_attr_test_mask(unsigned long j,
3454 					const unsigned long *mask,
3455 					unsigned int nr_bits)
3456 {
3457 	cpu_max_bits_warn(j, nr_bits);
3458 	return test_bit(j, mask);
3459 }
3460 
3461 /**
3462  *	netif_attr_test_online - Test for online CPU/Rx queue
3463  *	@j: CPU/Rx queue index
3464  *	@online_mask: bitmask for CPUs/Rx queues that are online
3465  *	@nr_bits: number of bits in the bitmask
3466  *
3467  * Returns true if a CPU/Rx queue is online.
3468  */
3469 static inline bool netif_attr_test_online(unsigned long j,
3470 					  const unsigned long *online_mask,
3471 					  unsigned int nr_bits)
3472 {
3473 	cpu_max_bits_warn(j, nr_bits);
3474 
3475 	if (online_mask)
3476 		return test_bit(j, online_mask);
3477 
3478 	return (j < nr_bits);
3479 }
3480 
3481 /**
3482  *	netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3483  *	@n: CPU/Rx queue index
3484  *	@srcp: the cpumask/Rx queue mask pointer
3485  *	@nr_bits: number of bits in the bitmask
3486  *
3487  * Returns >= nr_bits if no further CPUs/Rx queues set.
3488  */
3489 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3490 					       unsigned int nr_bits)
3491 {
3492 	/* -1 is a legal arg here. */
3493 	if (n != -1)
3494 		cpu_max_bits_warn(n, nr_bits);
3495 
3496 	if (srcp)
3497 		return find_next_bit(srcp, nr_bits, n + 1);
3498 
3499 	return n + 1;
3500 }
3501 
3502 /**
3503  *	netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3504  *	@n: CPU/Rx queue index
3505  *	@src1p: the first CPUs/Rx queues mask pointer
3506  *	@src2p: the second CPUs/Rx queues mask pointer
3507  *	@nr_bits: number of bits in the bitmask
3508  *
3509  * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3510  */
3511 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3512 					  const unsigned long *src2p,
3513 					  unsigned int nr_bits)
3514 {
3515 	/* -1 is a legal arg here. */
3516 	if (n != -1)
3517 		cpu_max_bits_warn(n, nr_bits);
3518 
3519 	if (src1p && src2p)
3520 		return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3521 	else if (src1p)
3522 		return find_next_bit(src1p, nr_bits, n + 1);
3523 	else if (src2p)
3524 		return find_next_bit(src2p, nr_bits, n + 1);
3525 
3526 	return n + 1;
3527 }
3528 #else
3529 static inline int netif_set_xps_queue(struct net_device *dev,
3530 				      const struct cpumask *mask,
3531 				      u16 index)
3532 {
3533 	return 0;
3534 }
3535 
3536 static inline int __netif_set_xps_queue(struct net_device *dev,
3537 					const unsigned long *mask,
3538 					u16 index, bool is_rxqs_map)
3539 {
3540 	return 0;
3541 }
3542 #endif
3543 
3544 /**
3545  *	netif_is_multiqueue - test if device has multiple transmit queues
3546  *	@dev: network device
3547  *
3548  * Check if device has multiple transmit queues
3549  */
3550 static inline bool netif_is_multiqueue(const struct net_device *dev)
3551 {
3552 	return dev->num_tx_queues > 1;
3553 }
3554 
3555 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3556 
3557 #ifdef CONFIG_SYSFS
3558 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3559 #else
3560 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3561 						unsigned int rxqs)
3562 {
3563 	dev->real_num_rx_queues = rxqs;
3564 	return 0;
3565 }
3566 #endif
3567 
3568 static inline struct netdev_rx_queue *
3569 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3570 {
3571 	return dev->_rx + rxq;
3572 }
3573 
3574 #ifdef CONFIG_SYSFS
3575 static inline unsigned int get_netdev_rx_queue_index(
3576 		struct netdev_rx_queue *queue)
3577 {
3578 	struct net_device *dev = queue->dev;
3579 	int index = queue - dev->_rx;
3580 
3581 	BUG_ON(index >= dev->num_rx_queues);
3582 	return index;
3583 }
3584 #endif
3585 
3586 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3587 int netif_get_num_default_rss_queues(void);
3588 
3589 enum skb_free_reason {
3590 	SKB_REASON_CONSUMED,
3591 	SKB_REASON_DROPPED,
3592 };
3593 
3594 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3595 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3596 
3597 /*
3598  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3599  * interrupt context or with hardware interrupts being disabled.
3600  * (in_irq() || irqs_disabled())
3601  *
3602  * We provide four helpers that can be used in following contexts :
3603  *
3604  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3605  *  replacing kfree_skb(skb)
3606  *
3607  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3608  *  Typically used in place of consume_skb(skb) in TX completion path
3609  *
3610  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3611  *  replacing kfree_skb(skb)
3612  *
3613  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3614  *  and consumed a packet. Used in place of consume_skb(skb)
3615  */
3616 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3617 {
3618 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3619 }
3620 
3621 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3622 {
3623 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3624 }
3625 
3626 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3627 {
3628 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3629 }
3630 
3631 static inline void dev_consume_skb_any(struct sk_buff *skb)
3632 {
3633 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3634 }
3635 
3636 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3637 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3638 int netif_rx(struct sk_buff *skb);
3639 int netif_rx_ni(struct sk_buff *skb);
3640 int netif_receive_skb(struct sk_buff *skb);
3641 int netif_receive_skb_core(struct sk_buff *skb);
3642 void netif_receive_skb_list(struct list_head *head);
3643 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3644 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3645 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3646 gro_result_t napi_gro_frags(struct napi_struct *napi);
3647 struct packet_offload *gro_find_receive_by_type(__be16 type);
3648 struct packet_offload *gro_find_complete_by_type(__be16 type);
3649 
3650 static inline void napi_free_frags(struct napi_struct *napi)
3651 {
3652 	kfree_skb(napi->skb);
3653 	napi->skb = NULL;
3654 }
3655 
3656 bool netdev_is_rx_handler_busy(struct net_device *dev);
3657 int netdev_rx_handler_register(struct net_device *dev,
3658 			       rx_handler_func_t *rx_handler,
3659 			       void *rx_handler_data);
3660 void netdev_rx_handler_unregister(struct net_device *dev);
3661 
3662 bool dev_valid_name(const char *name);
3663 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3664 		bool *need_copyout);
3665 int dev_ifconf(struct net *net, struct ifconf *, int);
3666 int dev_ethtool(struct net *net, struct ifreq *);
3667 unsigned int dev_get_flags(const struct net_device *);
3668 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3669 		       struct netlink_ext_ack *extack);
3670 int dev_change_flags(struct net_device *dev, unsigned int flags,
3671 		     struct netlink_ext_ack *extack);
3672 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3673 			unsigned int gchanges);
3674 int dev_change_name(struct net_device *, const char *);
3675 int dev_set_alias(struct net_device *, const char *, size_t);
3676 int dev_get_alias(const struct net_device *, char *, size_t);
3677 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3678 int __dev_set_mtu(struct net_device *, int);
3679 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3680 		    struct netlink_ext_ack *extack);
3681 int dev_set_mtu(struct net_device *, int);
3682 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3683 void dev_set_group(struct net_device *, int);
3684 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3685 			      struct netlink_ext_ack *extack);
3686 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3687 			struct netlink_ext_ack *extack);
3688 int dev_change_carrier(struct net_device *, bool new_carrier);
3689 int dev_get_phys_port_id(struct net_device *dev,
3690 			 struct netdev_phys_item_id *ppid);
3691 int dev_get_phys_port_name(struct net_device *dev,
3692 			   char *name, size_t len);
3693 int dev_get_port_parent_id(struct net_device *dev,
3694 			   struct netdev_phys_item_id *ppid, bool recurse);
3695 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3696 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3697 int dev_change_proto_down_generic(struct net_device *dev, bool proto_down);
3698 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3699 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3700 				    struct netdev_queue *txq, int *ret);
3701 
3702 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3703 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3704 		      int fd, u32 flags);
3705 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3706 		    enum bpf_netdev_command cmd);
3707 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3708 
3709 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3710 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3711 bool is_skb_forwardable(const struct net_device *dev,
3712 			const struct sk_buff *skb);
3713 
3714 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3715 					       struct sk_buff *skb)
3716 {
3717 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3718 	    unlikely(!is_skb_forwardable(dev, skb))) {
3719 		atomic_long_inc(&dev->rx_dropped);
3720 		kfree_skb(skb);
3721 		return NET_RX_DROP;
3722 	}
3723 
3724 	skb_scrub_packet(skb, true);
3725 	skb->priority = 0;
3726 	return 0;
3727 }
3728 
3729 bool dev_nit_active(struct net_device *dev);
3730 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3731 
3732 extern int		netdev_budget;
3733 extern unsigned int	netdev_budget_usecs;
3734 
3735 /* Called by rtnetlink.c:rtnl_unlock() */
3736 void netdev_run_todo(void);
3737 
3738 /**
3739  *	dev_put - release reference to device
3740  *	@dev: network device
3741  *
3742  * Release reference to device to allow it to be freed.
3743  */
3744 static inline void dev_put(struct net_device *dev)
3745 {
3746 	this_cpu_dec(*dev->pcpu_refcnt);
3747 }
3748 
3749 /**
3750  *	dev_hold - get reference to device
3751  *	@dev: network device
3752  *
3753  * Hold reference to device to keep it from being freed.
3754  */
3755 static inline void dev_hold(struct net_device *dev)
3756 {
3757 	this_cpu_inc(*dev->pcpu_refcnt);
3758 }
3759 
3760 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3761  * and _off may be called from IRQ context, but it is caller
3762  * who is responsible for serialization of these calls.
3763  *
3764  * The name carrier is inappropriate, these functions should really be
3765  * called netif_lowerlayer_*() because they represent the state of any
3766  * kind of lower layer not just hardware media.
3767  */
3768 
3769 void linkwatch_init_dev(struct net_device *dev);
3770 void linkwatch_fire_event(struct net_device *dev);
3771 void linkwatch_forget_dev(struct net_device *dev);
3772 
3773 /**
3774  *	netif_carrier_ok - test if carrier present
3775  *	@dev: network device
3776  *
3777  * Check if carrier is present on device
3778  */
3779 static inline bool netif_carrier_ok(const struct net_device *dev)
3780 {
3781 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3782 }
3783 
3784 unsigned long dev_trans_start(struct net_device *dev);
3785 
3786 void __netdev_watchdog_up(struct net_device *dev);
3787 
3788 void netif_carrier_on(struct net_device *dev);
3789 
3790 void netif_carrier_off(struct net_device *dev);
3791 
3792 /**
3793  *	netif_dormant_on - mark device as dormant.
3794  *	@dev: network device
3795  *
3796  * Mark device as dormant (as per RFC2863).
3797  *
3798  * The dormant state indicates that the relevant interface is not
3799  * actually in a condition to pass packets (i.e., it is not 'up') but is
3800  * in a "pending" state, waiting for some external event.  For "on-
3801  * demand" interfaces, this new state identifies the situation where the
3802  * interface is waiting for events to place it in the up state.
3803  */
3804 static inline void netif_dormant_on(struct net_device *dev)
3805 {
3806 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3807 		linkwatch_fire_event(dev);
3808 }
3809 
3810 /**
3811  *	netif_dormant_off - set device as not dormant.
3812  *	@dev: network device
3813  *
3814  * Device is not in dormant state.
3815  */
3816 static inline void netif_dormant_off(struct net_device *dev)
3817 {
3818 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3819 		linkwatch_fire_event(dev);
3820 }
3821 
3822 /**
3823  *	netif_dormant - test if device is dormant
3824  *	@dev: network device
3825  *
3826  * Check if device is dormant.
3827  */
3828 static inline bool netif_dormant(const struct net_device *dev)
3829 {
3830 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3831 }
3832 
3833 
3834 /**
3835  *	netif_oper_up - test if device is operational
3836  *	@dev: network device
3837  *
3838  * Check if carrier is operational
3839  */
3840 static inline bool netif_oper_up(const struct net_device *dev)
3841 {
3842 	return (dev->operstate == IF_OPER_UP ||
3843 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3844 }
3845 
3846 /**
3847  *	netif_device_present - is device available or removed
3848  *	@dev: network device
3849  *
3850  * Check if device has not been removed from system.
3851  */
3852 static inline bool netif_device_present(struct net_device *dev)
3853 {
3854 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3855 }
3856 
3857 void netif_device_detach(struct net_device *dev);
3858 
3859 void netif_device_attach(struct net_device *dev);
3860 
3861 /*
3862  * Network interface message level settings
3863  */
3864 
3865 enum {
3866 	NETIF_MSG_DRV		= 0x0001,
3867 	NETIF_MSG_PROBE		= 0x0002,
3868 	NETIF_MSG_LINK		= 0x0004,
3869 	NETIF_MSG_TIMER		= 0x0008,
3870 	NETIF_MSG_IFDOWN	= 0x0010,
3871 	NETIF_MSG_IFUP		= 0x0020,
3872 	NETIF_MSG_RX_ERR	= 0x0040,
3873 	NETIF_MSG_TX_ERR	= 0x0080,
3874 	NETIF_MSG_TX_QUEUED	= 0x0100,
3875 	NETIF_MSG_INTR		= 0x0200,
3876 	NETIF_MSG_TX_DONE	= 0x0400,
3877 	NETIF_MSG_RX_STATUS	= 0x0800,
3878 	NETIF_MSG_PKTDATA	= 0x1000,
3879 	NETIF_MSG_HW		= 0x2000,
3880 	NETIF_MSG_WOL		= 0x4000,
3881 };
3882 
3883 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3884 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3885 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3886 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3887 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3888 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3889 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3890 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3891 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3892 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3893 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3894 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3895 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3896 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3897 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3898 
3899 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3900 {
3901 	/* use default */
3902 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3903 		return default_msg_enable_bits;
3904 	if (debug_value == 0)	/* no output */
3905 		return 0;
3906 	/* set low N bits */
3907 	return (1U << debug_value) - 1;
3908 }
3909 
3910 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3911 {
3912 	spin_lock(&txq->_xmit_lock);
3913 	txq->xmit_lock_owner = cpu;
3914 }
3915 
3916 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3917 {
3918 	__acquire(&txq->_xmit_lock);
3919 	return true;
3920 }
3921 
3922 static inline void __netif_tx_release(struct netdev_queue *txq)
3923 {
3924 	__release(&txq->_xmit_lock);
3925 }
3926 
3927 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3928 {
3929 	spin_lock_bh(&txq->_xmit_lock);
3930 	txq->xmit_lock_owner = smp_processor_id();
3931 }
3932 
3933 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3934 {
3935 	bool ok = spin_trylock(&txq->_xmit_lock);
3936 	if (likely(ok))
3937 		txq->xmit_lock_owner = smp_processor_id();
3938 	return ok;
3939 }
3940 
3941 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3942 {
3943 	txq->xmit_lock_owner = -1;
3944 	spin_unlock(&txq->_xmit_lock);
3945 }
3946 
3947 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3948 {
3949 	txq->xmit_lock_owner = -1;
3950 	spin_unlock_bh(&txq->_xmit_lock);
3951 }
3952 
3953 static inline void txq_trans_update(struct netdev_queue *txq)
3954 {
3955 	if (txq->xmit_lock_owner != -1)
3956 		txq->trans_start = jiffies;
3957 }
3958 
3959 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3960 static inline void netif_trans_update(struct net_device *dev)
3961 {
3962 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3963 
3964 	if (txq->trans_start != jiffies)
3965 		txq->trans_start = jiffies;
3966 }
3967 
3968 /**
3969  *	netif_tx_lock - grab network device transmit lock
3970  *	@dev: network device
3971  *
3972  * Get network device transmit lock
3973  */
3974 static inline void netif_tx_lock(struct net_device *dev)
3975 {
3976 	unsigned int i;
3977 	int cpu;
3978 
3979 	spin_lock(&dev->tx_global_lock);
3980 	cpu = smp_processor_id();
3981 	for (i = 0; i < dev->num_tx_queues; i++) {
3982 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3983 
3984 		/* We are the only thread of execution doing a
3985 		 * freeze, but we have to grab the _xmit_lock in
3986 		 * order to synchronize with threads which are in
3987 		 * the ->hard_start_xmit() handler and already
3988 		 * checked the frozen bit.
3989 		 */
3990 		__netif_tx_lock(txq, cpu);
3991 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3992 		__netif_tx_unlock(txq);
3993 	}
3994 }
3995 
3996 static inline void netif_tx_lock_bh(struct net_device *dev)
3997 {
3998 	local_bh_disable();
3999 	netif_tx_lock(dev);
4000 }
4001 
4002 static inline void netif_tx_unlock(struct net_device *dev)
4003 {
4004 	unsigned int i;
4005 
4006 	for (i = 0; i < dev->num_tx_queues; i++) {
4007 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4008 
4009 		/* No need to grab the _xmit_lock here.  If the
4010 		 * queue is not stopped for another reason, we
4011 		 * force a schedule.
4012 		 */
4013 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
4014 		netif_schedule_queue(txq);
4015 	}
4016 	spin_unlock(&dev->tx_global_lock);
4017 }
4018 
4019 static inline void netif_tx_unlock_bh(struct net_device *dev)
4020 {
4021 	netif_tx_unlock(dev);
4022 	local_bh_enable();
4023 }
4024 
4025 #define HARD_TX_LOCK(dev, txq, cpu) {			\
4026 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4027 		__netif_tx_lock(txq, cpu);		\
4028 	} else {					\
4029 		__netif_tx_acquire(txq);		\
4030 	}						\
4031 }
4032 
4033 #define HARD_TX_TRYLOCK(dev, txq)			\
4034 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
4035 		__netif_tx_trylock(txq) :		\
4036 		__netif_tx_acquire(txq))
4037 
4038 #define HARD_TX_UNLOCK(dev, txq) {			\
4039 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4040 		__netif_tx_unlock(txq);			\
4041 	} else {					\
4042 		__netif_tx_release(txq);		\
4043 	}						\
4044 }
4045 
4046 static inline void netif_tx_disable(struct net_device *dev)
4047 {
4048 	unsigned int i;
4049 	int cpu;
4050 
4051 	local_bh_disable();
4052 	cpu = smp_processor_id();
4053 	for (i = 0; i < dev->num_tx_queues; i++) {
4054 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4055 
4056 		__netif_tx_lock(txq, cpu);
4057 		netif_tx_stop_queue(txq);
4058 		__netif_tx_unlock(txq);
4059 	}
4060 	local_bh_enable();
4061 }
4062 
4063 static inline void netif_addr_lock(struct net_device *dev)
4064 {
4065 	spin_lock(&dev->addr_list_lock);
4066 }
4067 
4068 static inline void netif_addr_lock_bh(struct net_device *dev)
4069 {
4070 	spin_lock_bh(&dev->addr_list_lock);
4071 }
4072 
4073 static inline void netif_addr_unlock(struct net_device *dev)
4074 {
4075 	spin_unlock(&dev->addr_list_lock);
4076 }
4077 
4078 static inline void netif_addr_unlock_bh(struct net_device *dev)
4079 {
4080 	spin_unlock_bh(&dev->addr_list_lock);
4081 }
4082 
4083 /*
4084  * dev_addrs walker. Should be used only for read access. Call with
4085  * rcu_read_lock held.
4086  */
4087 #define for_each_dev_addr(dev, ha) \
4088 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4089 
4090 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4091 
4092 void ether_setup(struct net_device *dev);
4093 
4094 /* Support for loadable net-drivers */
4095 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4096 				    unsigned char name_assign_type,
4097 				    void (*setup)(struct net_device *),
4098 				    unsigned int txqs, unsigned int rxqs);
4099 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4100 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4101 
4102 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4103 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4104 			 count)
4105 
4106 int register_netdev(struct net_device *dev);
4107 void unregister_netdev(struct net_device *dev);
4108 
4109 /* General hardware address lists handling functions */
4110 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4111 		   struct netdev_hw_addr_list *from_list, int addr_len);
4112 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4113 		      struct netdev_hw_addr_list *from_list, int addr_len);
4114 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4115 		       struct net_device *dev,
4116 		       int (*sync)(struct net_device *, const unsigned char *),
4117 		       int (*unsync)(struct net_device *,
4118 				     const unsigned char *));
4119 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4120 			   struct net_device *dev,
4121 			   int (*sync)(struct net_device *,
4122 				       const unsigned char *, int),
4123 			   int (*unsync)(struct net_device *,
4124 					 const unsigned char *, int));
4125 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4126 			      struct net_device *dev,
4127 			      int (*unsync)(struct net_device *,
4128 					    const unsigned char *, int));
4129 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4130 			  struct net_device *dev,
4131 			  int (*unsync)(struct net_device *,
4132 					const unsigned char *));
4133 void __hw_addr_init(struct netdev_hw_addr_list *list);
4134 
4135 /* Functions used for device addresses handling */
4136 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4137 		 unsigned char addr_type);
4138 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4139 		 unsigned char addr_type);
4140 void dev_addr_flush(struct net_device *dev);
4141 int dev_addr_init(struct net_device *dev);
4142 
4143 /* Functions used for unicast addresses handling */
4144 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4145 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4146 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4147 int dev_uc_sync(struct net_device *to, struct net_device *from);
4148 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4149 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4150 void dev_uc_flush(struct net_device *dev);
4151 void dev_uc_init(struct net_device *dev);
4152 
4153 /**
4154  *  __dev_uc_sync - Synchonize device's unicast list
4155  *  @dev:  device to sync
4156  *  @sync: function to call if address should be added
4157  *  @unsync: function to call if address should be removed
4158  *
4159  *  Add newly added addresses to the interface, and release
4160  *  addresses that have been deleted.
4161  */
4162 static inline int __dev_uc_sync(struct net_device *dev,
4163 				int (*sync)(struct net_device *,
4164 					    const unsigned char *),
4165 				int (*unsync)(struct net_device *,
4166 					      const unsigned char *))
4167 {
4168 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4169 }
4170 
4171 /**
4172  *  __dev_uc_unsync - Remove synchronized addresses from device
4173  *  @dev:  device to sync
4174  *  @unsync: function to call if address should be removed
4175  *
4176  *  Remove all addresses that were added to the device by dev_uc_sync().
4177  */
4178 static inline void __dev_uc_unsync(struct net_device *dev,
4179 				   int (*unsync)(struct net_device *,
4180 						 const unsigned char *))
4181 {
4182 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
4183 }
4184 
4185 /* Functions used for multicast addresses handling */
4186 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4187 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4188 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4189 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4190 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4191 int dev_mc_sync(struct net_device *to, struct net_device *from);
4192 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4193 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4194 void dev_mc_flush(struct net_device *dev);
4195 void dev_mc_init(struct net_device *dev);
4196 
4197 /**
4198  *  __dev_mc_sync - Synchonize device's multicast list
4199  *  @dev:  device to sync
4200  *  @sync: function to call if address should be added
4201  *  @unsync: function to call if address should be removed
4202  *
4203  *  Add newly added addresses to the interface, and release
4204  *  addresses that have been deleted.
4205  */
4206 static inline int __dev_mc_sync(struct net_device *dev,
4207 				int (*sync)(struct net_device *,
4208 					    const unsigned char *),
4209 				int (*unsync)(struct net_device *,
4210 					      const unsigned char *))
4211 {
4212 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4213 }
4214 
4215 /**
4216  *  __dev_mc_unsync - Remove synchronized addresses from device
4217  *  @dev:  device to sync
4218  *  @unsync: function to call if address should be removed
4219  *
4220  *  Remove all addresses that were added to the device by dev_mc_sync().
4221  */
4222 static inline void __dev_mc_unsync(struct net_device *dev,
4223 				   int (*unsync)(struct net_device *,
4224 						 const unsigned char *))
4225 {
4226 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
4227 }
4228 
4229 /* Functions used for secondary unicast and multicast support */
4230 void dev_set_rx_mode(struct net_device *dev);
4231 void __dev_set_rx_mode(struct net_device *dev);
4232 int dev_set_promiscuity(struct net_device *dev, int inc);
4233 int dev_set_allmulti(struct net_device *dev, int inc);
4234 void netdev_state_change(struct net_device *dev);
4235 void netdev_notify_peers(struct net_device *dev);
4236 void netdev_features_change(struct net_device *dev);
4237 /* Load a device via the kmod */
4238 void dev_load(struct net *net, const char *name);
4239 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4240 					struct rtnl_link_stats64 *storage);
4241 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4242 			     const struct net_device_stats *netdev_stats);
4243 
4244 extern int		netdev_max_backlog;
4245 extern int		netdev_tstamp_prequeue;
4246 extern int		weight_p;
4247 extern int		dev_weight_rx_bias;
4248 extern int		dev_weight_tx_bias;
4249 extern int		dev_rx_weight;
4250 extern int		dev_tx_weight;
4251 extern int		gro_normal_batch;
4252 
4253 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4254 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4255 						     struct list_head **iter);
4256 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4257 						     struct list_head **iter);
4258 
4259 /* iterate through upper list, must be called under RCU read lock */
4260 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4261 	for (iter = &(dev)->adj_list.upper, \
4262 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4263 	     updev; \
4264 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4265 
4266 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4267 				  int (*fn)(struct net_device *upper_dev,
4268 					    void *data),
4269 				  void *data);
4270 
4271 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4272 				  struct net_device *upper_dev);
4273 
4274 bool netdev_has_any_upper_dev(struct net_device *dev);
4275 
4276 void *netdev_lower_get_next_private(struct net_device *dev,
4277 				    struct list_head **iter);
4278 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4279 					struct list_head **iter);
4280 
4281 #define netdev_for_each_lower_private(dev, priv, iter) \
4282 	for (iter = (dev)->adj_list.lower.next, \
4283 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
4284 	     priv; \
4285 	     priv = netdev_lower_get_next_private(dev, &(iter)))
4286 
4287 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4288 	for (iter = &(dev)->adj_list.lower, \
4289 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4290 	     priv; \
4291 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4292 
4293 void *netdev_lower_get_next(struct net_device *dev,
4294 				struct list_head **iter);
4295 
4296 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4297 	for (iter = (dev)->adj_list.lower.next, \
4298 	     ldev = netdev_lower_get_next(dev, &(iter)); \
4299 	     ldev; \
4300 	     ldev = netdev_lower_get_next(dev, &(iter)))
4301 
4302 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4303 					     struct list_head **iter);
4304 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4305 						 struct list_head **iter);
4306 
4307 int netdev_walk_all_lower_dev(struct net_device *dev,
4308 			      int (*fn)(struct net_device *lower_dev,
4309 					void *data),
4310 			      void *data);
4311 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4312 				  int (*fn)(struct net_device *lower_dev,
4313 					    void *data),
4314 				  void *data);
4315 
4316 void *netdev_adjacent_get_private(struct list_head *adj_list);
4317 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4318 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4319 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4320 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4321 			  struct netlink_ext_ack *extack);
4322 int netdev_master_upper_dev_link(struct net_device *dev,
4323 				 struct net_device *upper_dev,
4324 				 void *upper_priv, void *upper_info,
4325 				 struct netlink_ext_ack *extack);
4326 void netdev_upper_dev_unlink(struct net_device *dev,
4327 			     struct net_device *upper_dev);
4328 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4329 				   struct net_device *new_dev,
4330 				   struct net_device *dev,
4331 				   struct netlink_ext_ack *extack);
4332 void netdev_adjacent_change_commit(struct net_device *old_dev,
4333 				   struct net_device *new_dev,
4334 				   struct net_device *dev);
4335 void netdev_adjacent_change_abort(struct net_device *old_dev,
4336 				  struct net_device *new_dev,
4337 				  struct net_device *dev);
4338 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4339 void *netdev_lower_dev_get_private(struct net_device *dev,
4340 				   struct net_device *lower_dev);
4341 void netdev_lower_state_changed(struct net_device *lower_dev,
4342 				void *lower_state_info);
4343 
4344 /* RSS keys are 40 or 52 bytes long */
4345 #define NETDEV_RSS_KEY_LEN 52
4346 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4347 void netdev_rss_key_fill(void *buffer, size_t len);
4348 
4349 int skb_checksum_help(struct sk_buff *skb);
4350 int skb_crc32c_csum_help(struct sk_buff *skb);
4351 int skb_csum_hwoffload_help(struct sk_buff *skb,
4352 			    const netdev_features_t features);
4353 
4354 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4355 				  netdev_features_t features, bool tx_path);
4356 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4357 				    netdev_features_t features);
4358 
4359 struct netdev_bonding_info {
4360 	ifslave	slave;
4361 	ifbond	master;
4362 };
4363 
4364 struct netdev_notifier_bonding_info {
4365 	struct netdev_notifier_info info; /* must be first */
4366 	struct netdev_bonding_info  bonding_info;
4367 };
4368 
4369 void netdev_bonding_info_change(struct net_device *dev,
4370 				struct netdev_bonding_info *bonding_info);
4371 
4372 static inline
4373 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4374 {
4375 	return __skb_gso_segment(skb, features, true);
4376 }
4377 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4378 
4379 static inline bool can_checksum_protocol(netdev_features_t features,
4380 					 __be16 protocol)
4381 {
4382 	if (protocol == htons(ETH_P_FCOE))
4383 		return !!(features & NETIF_F_FCOE_CRC);
4384 
4385 	/* Assume this is an IP checksum (not SCTP CRC) */
4386 
4387 	if (features & NETIF_F_HW_CSUM) {
4388 		/* Can checksum everything */
4389 		return true;
4390 	}
4391 
4392 	switch (protocol) {
4393 	case htons(ETH_P_IP):
4394 		return !!(features & NETIF_F_IP_CSUM);
4395 	case htons(ETH_P_IPV6):
4396 		return !!(features & NETIF_F_IPV6_CSUM);
4397 	default:
4398 		return false;
4399 	}
4400 }
4401 
4402 #ifdef CONFIG_BUG
4403 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4404 #else
4405 static inline void netdev_rx_csum_fault(struct net_device *dev,
4406 					struct sk_buff *skb)
4407 {
4408 }
4409 #endif
4410 /* rx skb timestamps */
4411 void net_enable_timestamp(void);
4412 void net_disable_timestamp(void);
4413 
4414 #ifdef CONFIG_PROC_FS
4415 int __init dev_proc_init(void);
4416 #else
4417 #define dev_proc_init() 0
4418 #endif
4419 
4420 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4421 					      struct sk_buff *skb, struct net_device *dev,
4422 					      bool more)
4423 {
4424 	__this_cpu_write(softnet_data.xmit.more, more);
4425 	return ops->ndo_start_xmit(skb, dev);
4426 }
4427 
4428 static inline bool netdev_xmit_more(void)
4429 {
4430 	return __this_cpu_read(softnet_data.xmit.more);
4431 }
4432 
4433 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4434 					    struct netdev_queue *txq, bool more)
4435 {
4436 	const struct net_device_ops *ops = dev->netdev_ops;
4437 	netdev_tx_t rc;
4438 
4439 	rc = __netdev_start_xmit(ops, skb, dev, more);
4440 	if (rc == NETDEV_TX_OK)
4441 		txq_trans_update(txq);
4442 
4443 	return rc;
4444 }
4445 
4446 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4447 				const void *ns);
4448 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4449 				 const void *ns);
4450 
4451 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4452 {
4453 	return netdev_class_create_file_ns(class_attr, NULL);
4454 }
4455 
4456 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4457 {
4458 	netdev_class_remove_file_ns(class_attr, NULL);
4459 }
4460 
4461 extern const struct kobj_ns_type_operations net_ns_type_operations;
4462 
4463 const char *netdev_drivername(const struct net_device *dev);
4464 
4465 void linkwatch_run_queue(void);
4466 
4467 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4468 							  netdev_features_t f2)
4469 {
4470 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4471 		if (f1 & NETIF_F_HW_CSUM)
4472 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4473 		else
4474 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4475 	}
4476 
4477 	return f1 & f2;
4478 }
4479 
4480 static inline netdev_features_t netdev_get_wanted_features(
4481 	struct net_device *dev)
4482 {
4483 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4484 }
4485 netdev_features_t netdev_increment_features(netdev_features_t all,
4486 	netdev_features_t one, netdev_features_t mask);
4487 
4488 /* Allow TSO being used on stacked device :
4489  * Performing the GSO segmentation before last device
4490  * is a performance improvement.
4491  */
4492 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4493 							netdev_features_t mask)
4494 {
4495 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4496 }
4497 
4498 int __netdev_update_features(struct net_device *dev);
4499 void netdev_update_features(struct net_device *dev);
4500 void netdev_change_features(struct net_device *dev);
4501 
4502 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4503 					struct net_device *dev);
4504 
4505 netdev_features_t passthru_features_check(struct sk_buff *skb,
4506 					  struct net_device *dev,
4507 					  netdev_features_t features);
4508 netdev_features_t netif_skb_features(struct sk_buff *skb);
4509 
4510 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4511 {
4512 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4513 
4514 	/* check flags correspondence */
4515 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4516 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4517 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4518 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4519 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4520 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4521 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4522 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4523 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4524 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4525 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4526 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4527 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4528 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4529 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4530 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4531 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4532 	BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4533 
4534 	return (features & feature) == feature;
4535 }
4536 
4537 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4538 {
4539 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4540 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4541 }
4542 
4543 static inline bool netif_needs_gso(struct sk_buff *skb,
4544 				   netdev_features_t features)
4545 {
4546 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4547 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4548 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4549 }
4550 
4551 static inline void netif_set_gso_max_size(struct net_device *dev,
4552 					  unsigned int size)
4553 {
4554 	dev->gso_max_size = size;
4555 }
4556 
4557 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4558 					int pulled_hlen, u16 mac_offset,
4559 					int mac_len)
4560 {
4561 	skb->protocol = protocol;
4562 	skb->encapsulation = 1;
4563 	skb_push(skb, pulled_hlen);
4564 	skb_reset_transport_header(skb);
4565 	skb->mac_header = mac_offset;
4566 	skb->network_header = skb->mac_header + mac_len;
4567 	skb->mac_len = mac_len;
4568 }
4569 
4570 static inline bool netif_is_macsec(const struct net_device *dev)
4571 {
4572 	return dev->priv_flags & IFF_MACSEC;
4573 }
4574 
4575 static inline bool netif_is_macvlan(const struct net_device *dev)
4576 {
4577 	return dev->priv_flags & IFF_MACVLAN;
4578 }
4579 
4580 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4581 {
4582 	return dev->priv_flags & IFF_MACVLAN_PORT;
4583 }
4584 
4585 static inline bool netif_is_bond_master(const struct net_device *dev)
4586 {
4587 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4588 }
4589 
4590 static inline bool netif_is_bond_slave(const struct net_device *dev)
4591 {
4592 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4593 }
4594 
4595 static inline bool netif_supports_nofcs(struct net_device *dev)
4596 {
4597 	return dev->priv_flags & IFF_SUPP_NOFCS;
4598 }
4599 
4600 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
4601 {
4602 	return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
4603 }
4604 
4605 static inline bool netif_is_l3_master(const struct net_device *dev)
4606 {
4607 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4608 }
4609 
4610 static inline bool netif_is_l3_slave(const struct net_device *dev)
4611 {
4612 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4613 }
4614 
4615 static inline bool netif_is_bridge_master(const struct net_device *dev)
4616 {
4617 	return dev->priv_flags & IFF_EBRIDGE;
4618 }
4619 
4620 static inline bool netif_is_bridge_port(const struct net_device *dev)
4621 {
4622 	return dev->priv_flags & IFF_BRIDGE_PORT;
4623 }
4624 
4625 static inline bool netif_is_ovs_master(const struct net_device *dev)
4626 {
4627 	return dev->priv_flags & IFF_OPENVSWITCH;
4628 }
4629 
4630 static inline bool netif_is_ovs_port(const struct net_device *dev)
4631 {
4632 	return dev->priv_flags & IFF_OVS_DATAPATH;
4633 }
4634 
4635 static inline bool netif_is_team_master(const struct net_device *dev)
4636 {
4637 	return dev->priv_flags & IFF_TEAM;
4638 }
4639 
4640 static inline bool netif_is_team_port(const struct net_device *dev)
4641 {
4642 	return dev->priv_flags & IFF_TEAM_PORT;
4643 }
4644 
4645 static inline bool netif_is_lag_master(const struct net_device *dev)
4646 {
4647 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4648 }
4649 
4650 static inline bool netif_is_lag_port(const struct net_device *dev)
4651 {
4652 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4653 }
4654 
4655 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4656 {
4657 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4658 }
4659 
4660 static inline bool netif_is_failover(const struct net_device *dev)
4661 {
4662 	return dev->priv_flags & IFF_FAILOVER;
4663 }
4664 
4665 static inline bool netif_is_failover_slave(const struct net_device *dev)
4666 {
4667 	return dev->priv_flags & IFF_FAILOVER_SLAVE;
4668 }
4669 
4670 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4671 static inline void netif_keep_dst(struct net_device *dev)
4672 {
4673 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4674 }
4675 
4676 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4677 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4678 {
4679 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4680 	return dev->priv_flags & IFF_MACSEC;
4681 }
4682 
4683 extern struct pernet_operations __net_initdata loopback_net_ops;
4684 
4685 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4686 
4687 /* netdev_printk helpers, similar to dev_printk */
4688 
4689 static inline const char *netdev_name(const struct net_device *dev)
4690 {
4691 	if (!dev->name[0] || strchr(dev->name, '%'))
4692 		return "(unnamed net_device)";
4693 	return dev->name;
4694 }
4695 
4696 static inline bool netdev_unregistering(const struct net_device *dev)
4697 {
4698 	return dev->reg_state == NETREG_UNREGISTERING;
4699 }
4700 
4701 static inline const char *netdev_reg_state(const struct net_device *dev)
4702 {
4703 	switch (dev->reg_state) {
4704 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4705 	case NETREG_REGISTERED: return "";
4706 	case NETREG_UNREGISTERING: return " (unregistering)";
4707 	case NETREG_UNREGISTERED: return " (unregistered)";
4708 	case NETREG_RELEASED: return " (released)";
4709 	case NETREG_DUMMY: return " (dummy)";
4710 	}
4711 
4712 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4713 	return " (unknown)";
4714 }
4715 
4716 __printf(3, 4) __cold
4717 void netdev_printk(const char *level, const struct net_device *dev,
4718 		   const char *format, ...);
4719 __printf(2, 3) __cold
4720 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4721 __printf(2, 3) __cold
4722 void netdev_alert(const struct net_device *dev, const char *format, ...);
4723 __printf(2, 3) __cold
4724 void netdev_crit(const struct net_device *dev, const char *format, ...);
4725 __printf(2, 3) __cold
4726 void netdev_err(const struct net_device *dev, const char *format, ...);
4727 __printf(2, 3) __cold
4728 void netdev_warn(const struct net_device *dev, const char *format, ...);
4729 __printf(2, 3) __cold
4730 void netdev_notice(const struct net_device *dev, const char *format, ...);
4731 __printf(2, 3) __cold
4732 void netdev_info(const struct net_device *dev, const char *format, ...);
4733 
4734 #define netdev_level_once(level, dev, fmt, ...)			\
4735 do {								\
4736 	static bool __print_once __read_mostly;			\
4737 								\
4738 	if (!__print_once) {					\
4739 		__print_once = true;				\
4740 		netdev_printk(level, dev, fmt, ##__VA_ARGS__);	\
4741 	}							\
4742 } while (0)
4743 
4744 #define netdev_emerg_once(dev, fmt, ...) \
4745 	netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4746 #define netdev_alert_once(dev, fmt, ...) \
4747 	netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4748 #define netdev_crit_once(dev, fmt, ...) \
4749 	netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4750 #define netdev_err_once(dev, fmt, ...) \
4751 	netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4752 #define netdev_warn_once(dev, fmt, ...) \
4753 	netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4754 #define netdev_notice_once(dev, fmt, ...) \
4755 	netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4756 #define netdev_info_once(dev, fmt, ...) \
4757 	netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4758 
4759 #define MODULE_ALIAS_NETDEV(device) \
4760 	MODULE_ALIAS("netdev-" device)
4761 
4762 #if defined(CONFIG_DYNAMIC_DEBUG)
4763 #define netdev_dbg(__dev, format, args...)			\
4764 do {								\
4765 	dynamic_netdev_dbg(__dev, format, ##args);		\
4766 } while (0)
4767 #elif defined(DEBUG)
4768 #define netdev_dbg(__dev, format, args...)			\
4769 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4770 #else
4771 #define netdev_dbg(__dev, format, args...)			\
4772 ({								\
4773 	if (0)							\
4774 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4775 })
4776 #endif
4777 
4778 #if defined(VERBOSE_DEBUG)
4779 #define netdev_vdbg	netdev_dbg
4780 #else
4781 
4782 #define netdev_vdbg(dev, format, args...)			\
4783 ({								\
4784 	if (0)							\
4785 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4786 	0;							\
4787 })
4788 #endif
4789 
4790 /*
4791  * netdev_WARN() acts like dev_printk(), but with the key difference
4792  * of using a WARN/WARN_ON to get the message out, including the
4793  * file/line information and a backtrace.
4794  */
4795 #define netdev_WARN(dev, format, args...)			\
4796 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
4797 	     netdev_reg_state(dev), ##args)
4798 
4799 #define netdev_WARN_ONCE(dev, format, args...)				\
4800 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
4801 		  netdev_reg_state(dev), ##args)
4802 
4803 /* netif printk helpers, similar to netdev_printk */
4804 
4805 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4806 do {					  			\
4807 	if (netif_msg_##type(priv))				\
4808 		netdev_printk(level, (dev), fmt, ##args);	\
4809 } while (0)
4810 
4811 #define netif_level(level, priv, type, dev, fmt, args...)	\
4812 do {								\
4813 	if (netif_msg_##type(priv))				\
4814 		netdev_##level(dev, fmt, ##args);		\
4815 } while (0)
4816 
4817 #define netif_emerg(priv, type, dev, fmt, args...)		\
4818 	netif_level(emerg, priv, type, dev, fmt, ##args)
4819 #define netif_alert(priv, type, dev, fmt, args...)		\
4820 	netif_level(alert, priv, type, dev, fmt, ##args)
4821 #define netif_crit(priv, type, dev, fmt, args...)		\
4822 	netif_level(crit, priv, type, dev, fmt, ##args)
4823 #define netif_err(priv, type, dev, fmt, args...)		\
4824 	netif_level(err, priv, type, dev, fmt, ##args)
4825 #define netif_warn(priv, type, dev, fmt, args...)		\
4826 	netif_level(warn, priv, type, dev, fmt, ##args)
4827 #define netif_notice(priv, type, dev, fmt, args...)		\
4828 	netif_level(notice, priv, type, dev, fmt, ##args)
4829 #define netif_info(priv, type, dev, fmt, args...)		\
4830 	netif_level(info, priv, type, dev, fmt, ##args)
4831 
4832 #if defined(CONFIG_DYNAMIC_DEBUG)
4833 #define netif_dbg(priv, type, netdev, format, args...)		\
4834 do {								\
4835 	if (netif_msg_##type(priv))				\
4836 		dynamic_netdev_dbg(netdev, format, ##args);	\
4837 } while (0)
4838 #elif defined(DEBUG)
4839 #define netif_dbg(priv, type, dev, format, args...)		\
4840 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4841 #else
4842 #define netif_dbg(priv, type, dev, format, args...)			\
4843 ({									\
4844 	if (0)								\
4845 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4846 	0;								\
4847 })
4848 #endif
4849 
4850 /* if @cond then downgrade to debug, else print at @level */
4851 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4852 	do {                                                              \
4853 		if (cond)                                                 \
4854 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4855 		else                                                      \
4856 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4857 	} while (0)
4858 
4859 #if defined(VERBOSE_DEBUG)
4860 #define netif_vdbg	netif_dbg
4861 #else
4862 #define netif_vdbg(priv, type, dev, format, args...)		\
4863 ({								\
4864 	if (0)							\
4865 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4866 	0;							\
4867 })
4868 #endif
4869 
4870 /*
4871  *	The list of packet types we will receive (as opposed to discard)
4872  *	and the routines to invoke.
4873  *
4874  *	Why 16. Because with 16 the only overlap we get on a hash of the
4875  *	low nibble of the protocol value is RARP/SNAP/X.25.
4876  *
4877  *		0800	IP
4878  *		0001	802.3
4879  *		0002	AX.25
4880  *		0004	802.2
4881  *		8035	RARP
4882  *		0005	SNAP
4883  *		0805	X.25
4884  *		0806	ARP
4885  *		8137	IPX
4886  *		0009	Localtalk
4887  *		86DD	IPv6
4888  */
4889 #define PTYPE_HASH_SIZE	(16)
4890 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4891 
4892 extern struct net_device *blackhole_netdev;
4893 
4894 #endif	/* _LINUX_NETDEVICE_H */
4895