xref: /openbmc/linux/include/linux/netdevice.h (revision dd2934a95701576203b2f61e8ded4e4a2f9183ea)
1 /*
2  * INET		An implementation of the TCP/IP protocol suite for the LINUX
3  *		operating system.  INET is implemented using the  BSD Socket
4  *		interface as the means of communication with the user level.
5  *
6  *		Definitions for the Interfaces handler.
7  *
8  * Version:	@(#)dev.h	1.0.10	08/12/93
9  *
10  * Authors:	Ross Biro
11  *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12  *		Corey Minyard <wf-rch!minyard@relay.EU.net>
13  *		Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
14  *		Alan Cox, <alan@lxorguk.ukuu.org.uk>
15  *		Bjorn Ekwall. <bj0rn@blox.se>
16  *              Pekka Riikonen <priikone@poseidon.pspt.fi>
17  *
18  *		This program is free software; you can redistribute it and/or
19  *		modify it under the terms of the GNU General Public License
20  *		as published by the Free Software Foundation; either version
21  *		2 of the License, or (at your option) any later version.
22  *
23  *		Moved to /usr/include/linux for NET3
24  */
25 #ifndef _LINUX_NETDEVICE_H
26 #define _LINUX_NETDEVICE_H
27 
28 #include <linux/timer.h>
29 #include <linux/bug.h>
30 #include <linux/delay.h>
31 #include <linux/atomic.h>
32 #include <linux/prefetch.h>
33 #include <asm/cache.h>
34 #include <asm/byteorder.h>
35 
36 #include <linux/percpu.h>
37 #include <linux/rculist.h>
38 #include <linux/workqueue.h>
39 #include <linux/dynamic_queue_limits.h>
40 
41 #include <linux/ethtool.h>
42 #include <net/net_namespace.h>
43 #ifdef CONFIG_DCB
44 #include <net/dcbnl.h>
45 #endif
46 #include <net/netprio_cgroup.h>
47 #include <net/xdp.h>
48 
49 #include <linux/netdev_features.h>
50 #include <linux/neighbour.h>
51 #include <uapi/linux/netdevice.h>
52 #include <uapi/linux/if_bonding.h>
53 #include <uapi/linux/pkt_cls.h>
54 #include <linux/hashtable.h>
55 
56 struct netpoll_info;
57 struct device;
58 struct phy_device;
59 struct dsa_port;
60 
61 struct sfp_bus;
62 /* 802.11 specific */
63 struct wireless_dev;
64 /* 802.15.4 specific */
65 struct wpan_dev;
66 struct mpls_dev;
67 /* UDP Tunnel offloads */
68 struct udp_tunnel_info;
69 struct bpf_prog;
70 struct xdp_buff;
71 
72 void netdev_set_default_ethtool_ops(struct net_device *dev,
73 				    const struct ethtool_ops *ops);
74 
75 /* Backlog congestion levels */
76 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
77 #define NET_RX_DROP		1	/* packet dropped */
78 
79 /*
80  * Transmit return codes: transmit return codes originate from three different
81  * namespaces:
82  *
83  * - qdisc return codes
84  * - driver transmit return codes
85  * - errno values
86  *
87  * Drivers are allowed to return any one of those in their hard_start_xmit()
88  * function. Real network devices commonly used with qdiscs should only return
89  * the driver transmit return codes though - when qdiscs are used, the actual
90  * transmission happens asynchronously, so the value is not propagated to
91  * higher layers. Virtual network devices transmit synchronously; in this case
92  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
93  * others are propagated to higher layers.
94  */
95 
96 /* qdisc ->enqueue() return codes. */
97 #define NET_XMIT_SUCCESS	0x00
98 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
99 #define NET_XMIT_CN		0x02	/* congestion notification	*/
100 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
101 
102 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
103  * indicates that the device will soon be dropping packets, or already drops
104  * some packets of the same priority; prompting us to send less aggressively. */
105 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
106 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
107 
108 /* Driver transmit return codes */
109 #define NETDEV_TX_MASK		0xf0
110 
111 enum netdev_tx {
112 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
113 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
114 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
115 };
116 typedef enum netdev_tx netdev_tx_t;
117 
118 /*
119  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
120  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
121  */
122 static inline bool dev_xmit_complete(int rc)
123 {
124 	/*
125 	 * Positive cases with an skb consumed by a driver:
126 	 * - successful transmission (rc == NETDEV_TX_OK)
127 	 * - error while transmitting (rc < 0)
128 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
129 	 */
130 	if (likely(rc < NET_XMIT_MASK))
131 		return true;
132 
133 	return false;
134 }
135 
136 /*
137  *	Compute the worst-case header length according to the protocols
138  *	used.
139  */
140 
141 #if defined(CONFIG_HYPERV_NET)
142 # define LL_MAX_HEADER 128
143 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
144 # if defined(CONFIG_MAC80211_MESH)
145 #  define LL_MAX_HEADER 128
146 # else
147 #  define LL_MAX_HEADER 96
148 # endif
149 #else
150 # define LL_MAX_HEADER 32
151 #endif
152 
153 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
154     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
155 #define MAX_HEADER LL_MAX_HEADER
156 #else
157 #define MAX_HEADER (LL_MAX_HEADER + 48)
158 #endif
159 
160 /*
161  *	Old network device statistics. Fields are native words
162  *	(unsigned long) so they can be read and written atomically.
163  */
164 
165 struct net_device_stats {
166 	unsigned long	rx_packets;
167 	unsigned long	tx_packets;
168 	unsigned long	rx_bytes;
169 	unsigned long	tx_bytes;
170 	unsigned long	rx_errors;
171 	unsigned long	tx_errors;
172 	unsigned long	rx_dropped;
173 	unsigned long	tx_dropped;
174 	unsigned long	multicast;
175 	unsigned long	collisions;
176 	unsigned long	rx_length_errors;
177 	unsigned long	rx_over_errors;
178 	unsigned long	rx_crc_errors;
179 	unsigned long	rx_frame_errors;
180 	unsigned long	rx_fifo_errors;
181 	unsigned long	rx_missed_errors;
182 	unsigned long	tx_aborted_errors;
183 	unsigned long	tx_carrier_errors;
184 	unsigned long	tx_fifo_errors;
185 	unsigned long	tx_heartbeat_errors;
186 	unsigned long	tx_window_errors;
187 	unsigned long	rx_compressed;
188 	unsigned long	tx_compressed;
189 };
190 
191 
192 #include <linux/cache.h>
193 #include <linux/skbuff.h>
194 
195 #ifdef CONFIG_RPS
196 #include <linux/static_key.h>
197 extern struct static_key rps_needed;
198 extern struct static_key rfs_needed;
199 #endif
200 
201 struct neighbour;
202 struct neigh_parms;
203 struct sk_buff;
204 
205 struct netdev_hw_addr {
206 	struct list_head	list;
207 	unsigned char		addr[MAX_ADDR_LEN];
208 	unsigned char		type;
209 #define NETDEV_HW_ADDR_T_LAN		1
210 #define NETDEV_HW_ADDR_T_SAN		2
211 #define NETDEV_HW_ADDR_T_SLAVE		3
212 #define NETDEV_HW_ADDR_T_UNICAST	4
213 #define NETDEV_HW_ADDR_T_MULTICAST	5
214 	bool			global_use;
215 	int			sync_cnt;
216 	int			refcount;
217 	int			synced;
218 	struct rcu_head		rcu_head;
219 };
220 
221 struct netdev_hw_addr_list {
222 	struct list_head	list;
223 	int			count;
224 };
225 
226 #define netdev_hw_addr_list_count(l) ((l)->count)
227 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
228 #define netdev_hw_addr_list_for_each(ha, l) \
229 	list_for_each_entry(ha, &(l)->list, list)
230 
231 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
232 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
233 #define netdev_for_each_uc_addr(ha, dev) \
234 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
235 
236 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
237 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
238 #define netdev_for_each_mc_addr(ha, dev) \
239 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
240 
241 struct hh_cache {
242 	unsigned int	hh_len;
243 	seqlock_t	hh_lock;
244 
245 	/* cached hardware header; allow for machine alignment needs.        */
246 #define HH_DATA_MOD	16
247 #define HH_DATA_OFF(__len) \
248 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
249 #define HH_DATA_ALIGN(__len) \
250 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
251 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
252 };
253 
254 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
255  * Alternative is:
256  *   dev->hard_header_len ? (dev->hard_header_len +
257  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
258  *
259  * We could use other alignment values, but we must maintain the
260  * relationship HH alignment <= LL alignment.
261  */
262 #define LL_RESERVED_SPACE(dev) \
263 	((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
264 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
265 	((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
266 
267 struct header_ops {
268 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
269 			   unsigned short type, const void *daddr,
270 			   const void *saddr, unsigned int len);
271 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
272 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
273 	void	(*cache_update)(struct hh_cache *hh,
274 				const struct net_device *dev,
275 				const unsigned char *haddr);
276 	bool	(*validate)(const char *ll_header, unsigned int len);
277 };
278 
279 /* These flag bits are private to the generic network queueing
280  * layer; they may not be explicitly referenced by any other
281  * code.
282  */
283 
284 enum netdev_state_t {
285 	__LINK_STATE_START,
286 	__LINK_STATE_PRESENT,
287 	__LINK_STATE_NOCARRIER,
288 	__LINK_STATE_LINKWATCH_PENDING,
289 	__LINK_STATE_DORMANT,
290 };
291 
292 
293 /*
294  * This structure holds boot-time configured netdevice settings. They
295  * are then used in the device probing.
296  */
297 struct netdev_boot_setup {
298 	char name[IFNAMSIZ];
299 	struct ifmap map;
300 };
301 #define NETDEV_BOOT_SETUP_MAX 8
302 
303 int __init netdev_boot_setup(char *str);
304 
305 struct gro_list {
306 	struct list_head	list;
307 	int			count;
308 };
309 
310 /*
311  * size of gro hash buckets, must less than bit number of
312  * napi_struct::gro_bitmask
313  */
314 #define GRO_HASH_BUCKETS	8
315 
316 /*
317  * Structure for NAPI scheduling similar to tasklet but with weighting
318  */
319 struct napi_struct {
320 	/* The poll_list must only be managed by the entity which
321 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
322 	 * whoever atomically sets that bit can add this napi_struct
323 	 * to the per-CPU poll_list, and whoever clears that bit
324 	 * can remove from the list right before clearing the bit.
325 	 */
326 	struct list_head	poll_list;
327 
328 	unsigned long		state;
329 	int			weight;
330 	unsigned long		gro_bitmask;
331 	int			(*poll)(struct napi_struct *, int);
332 #ifdef CONFIG_NETPOLL
333 	int			poll_owner;
334 #endif
335 	struct net_device	*dev;
336 	struct gro_list		gro_hash[GRO_HASH_BUCKETS];
337 	struct sk_buff		*skb;
338 	struct hrtimer		timer;
339 	struct list_head	dev_list;
340 	struct hlist_node	napi_hash_node;
341 	unsigned int		napi_id;
342 };
343 
344 enum {
345 	NAPI_STATE_SCHED,	/* Poll is scheduled */
346 	NAPI_STATE_MISSED,	/* reschedule a napi */
347 	NAPI_STATE_DISABLE,	/* Disable pending */
348 	NAPI_STATE_NPSVC,	/* Netpoll - don't dequeue from poll_list */
349 	NAPI_STATE_HASHED,	/* In NAPI hash (busy polling possible) */
350 	NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
351 	NAPI_STATE_IN_BUSY_POLL,/* sk_busy_loop() owns this NAPI */
352 };
353 
354 enum {
355 	NAPIF_STATE_SCHED	 = BIT(NAPI_STATE_SCHED),
356 	NAPIF_STATE_MISSED	 = BIT(NAPI_STATE_MISSED),
357 	NAPIF_STATE_DISABLE	 = BIT(NAPI_STATE_DISABLE),
358 	NAPIF_STATE_NPSVC	 = BIT(NAPI_STATE_NPSVC),
359 	NAPIF_STATE_HASHED	 = BIT(NAPI_STATE_HASHED),
360 	NAPIF_STATE_NO_BUSY_POLL = BIT(NAPI_STATE_NO_BUSY_POLL),
361 	NAPIF_STATE_IN_BUSY_POLL = BIT(NAPI_STATE_IN_BUSY_POLL),
362 };
363 
364 enum gro_result {
365 	GRO_MERGED,
366 	GRO_MERGED_FREE,
367 	GRO_HELD,
368 	GRO_NORMAL,
369 	GRO_DROP,
370 	GRO_CONSUMED,
371 };
372 typedef enum gro_result gro_result_t;
373 
374 /*
375  * enum rx_handler_result - Possible return values for rx_handlers.
376  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
377  * further.
378  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
379  * case skb->dev was changed by rx_handler.
380  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
381  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
382  *
383  * rx_handlers are functions called from inside __netif_receive_skb(), to do
384  * special processing of the skb, prior to delivery to protocol handlers.
385  *
386  * Currently, a net_device can only have a single rx_handler registered. Trying
387  * to register a second rx_handler will return -EBUSY.
388  *
389  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
390  * To unregister a rx_handler on a net_device, use
391  * netdev_rx_handler_unregister().
392  *
393  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
394  * do with the skb.
395  *
396  * If the rx_handler consumed the skb in some way, it should return
397  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
398  * the skb to be delivered in some other way.
399  *
400  * If the rx_handler changed skb->dev, to divert the skb to another
401  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
402  * new device will be called if it exists.
403  *
404  * If the rx_handler decides the skb should be ignored, it should return
405  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
406  * are registered on exact device (ptype->dev == skb->dev).
407  *
408  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
409  * delivered, it should return RX_HANDLER_PASS.
410  *
411  * A device without a registered rx_handler will behave as if rx_handler
412  * returned RX_HANDLER_PASS.
413  */
414 
415 enum rx_handler_result {
416 	RX_HANDLER_CONSUMED,
417 	RX_HANDLER_ANOTHER,
418 	RX_HANDLER_EXACT,
419 	RX_HANDLER_PASS,
420 };
421 typedef enum rx_handler_result rx_handler_result_t;
422 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
423 
424 void __napi_schedule(struct napi_struct *n);
425 void __napi_schedule_irqoff(struct napi_struct *n);
426 
427 static inline bool napi_disable_pending(struct napi_struct *n)
428 {
429 	return test_bit(NAPI_STATE_DISABLE, &n->state);
430 }
431 
432 bool napi_schedule_prep(struct napi_struct *n);
433 
434 /**
435  *	napi_schedule - schedule NAPI poll
436  *	@n: NAPI context
437  *
438  * Schedule NAPI poll routine to be called if it is not already
439  * running.
440  */
441 static inline void napi_schedule(struct napi_struct *n)
442 {
443 	if (napi_schedule_prep(n))
444 		__napi_schedule(n);
445 }
446 
447 /**
448  *	napi_schedule_irqoff - schedule NAPI poll
449  *	@n: NAPI context
450  *
451  * Variant of napi_schedule(), assuming hard irqs are masked.
452  */
453 static inline void napi_schedule_irqoff(struct napi_struct *n)
454 {
455 	if (napi_schedule_prep(n))
456 		__napi_schedule_irqoff(n);
457 }
458 
459 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
460 static inline bool napi_reschedule(struct napi_struct *napi)
461 {
462 	if (napi_schedule_prep(napi)) {
463 		__napi_schedule(napi);
464 		return true;
465 	}
466 	return false;
467 }
468 
469 bool napi_complete_done(struct napi_struct *n, int work_done);
470 /**
471  *	napi_complete - NAPI processing complete
472  *	@n: NAPI context
473  *
474  * Mark NAPI processing as complete.
475  * Consider using napi_complete_done() instead.
476  * Return false if device should avoid rearming interrupts.
477  */
478 static inline bool napi_complete(struct napi_struct *n)
479 {
480 	return napi_complete_done(n, 0);
481 }
482 
483 /**
484  *	napi_hash_del - remove a NAPI from global table
485  *	@napi: NAPI context
486  *
487  * Warning: caller must observe RCU grace period
488  * before freeing memory containing @napi, if
489  * this function returns true.
490  * Note: core networking stack automatically calls it
491  * from netif_napi_del().
492  * Drivers might want to call this helper to combine all
493  * the needed RCU grace periods into a single one.
494  */
495 bool napi_hash_del(struct napi_struct *napi);
496 
497 /**
498  *	napi_disable - prevent NAPI from scheduling
499  *	@n: NAPI context
500  *
501  * Stop NAPI from being scheduled on this context.
502  * Waits till any outstanding processing completes.
503  */
504 void napi_disable(struct napi_struct *n);
505 
506 /**
507  *	napi_enable - enable NAPI scheduling
508  *	@n: NAPI context
509  *
510  * Resume NAPI from being scheduled on this context.
511  * Must be paired with napi_disable.
512  */
513 static inline void napi_enable(struct napi_struct *n)
514 {
515 	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
516 	smp_mb__before_atomic();
517 	clear_bit(NAPI_STATE_SCHED, &n->state);
518 	clear_bit(NAPI_STATE_NPSVC, &n->state);
519 }
520 
521 /**
522  *	napi_synchronize - wait until NAPI is not running
523  *	@n: NAPI context
524  *
525  * Wait until NAPI is done being scheduled on this context.
526  * Waits till any outstanding processing completes but
527  * does not disable future activations.
528  */
529 static inline void napi_synchronize(const struct napi_struct *n)
530 {
531 	if (IS_ENABLED(CONFIG_SMP))
532 		while (test_bit(NAPI_STATE_SCHED, &n->state))
533 			msleep(1);
534 	else
535 		barrier();
536 }
537 
538 /**
539  *	napi_if_scheduled_mark_missed - if napi is running, set the
540  *	NAPIF_STATE_MISSED
541  *	@n: NAPI context
542  *
543  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
544  * NAPI is scheduled.
545  **/
546 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
547 {
548 	unsigned long val, new;
549 
550 	do {
551 		val = READ_ONCE(n->state);
552 		if (val & NAPIF_STATE_DISABLE)
553 			return true;
554 
555 		if (!(val & NAPIF_STATE_SCHED))
556 			return false;
557 
558 		new = val | NAPIF_STATE_MISSED;
559 	} while (cmpxchg(&n->state, val, new) != val);
560 
561 	return true;
562 }
563 
564 enum netdev_queue_state_t {
565 	__QUEUE_STATE_DRV_XOFF,
566 	__QUEUE_STATE_STACK_XOFF,
567 	__QUEUE_STATE_FROZEN,
568 };
569 
570 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
571 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
572 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
573 
574 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
575 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
576 					QUEUE_STATE_FROZEN)
577 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
578 					QUEUE_STATE_FROZEN)
579 
580 /*
581  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
582  * netif_tx_* functions below are used to manipulate this flag.  The
583  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
584  * queue independently.  The netif_xmit_*stopped functions below are called
585  * to check if the queue has been stopped by the driver or stack (either
586  * of the XOFF bits are set in the state).  Drivers should not need to call
587  * netif_xmit*stopped functions, they should only be using netif_tx_*.
588  */
589 
590 struct netdev_queue {
591 /*
592  * read-mostly part
593  */
594 	struct net_device	*dev;
595 	struct Qdisc __rcu	*qdisc;
596 	struct Qdisc		*qdisc_sleeping;
597 #ifdef CONFIG_SYSFS
598 	struct kobject		kobj;
599 #endif
600 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
601 	int			numa_node;
602 #endif
603 	unsigned long		tx_maxrate;
604 	/*
605 	 * Number of TX timeouts for this queue
606 	 * (/sys/class/net/DEV/Q/trans_timeout)
607 	 */
608 	unsigned long		trans_timeout;
609 
610 	/* Subordinate device that the queue has been assigned to */
611 	struct net_device	*sb_dev;
612 /*
613  * write-mostly part
614  */
615 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
616 	int			xmit_lock_owner;
617 	/*
618 	 * Time (in jiffies) of last Tx
619 	 */
620 	unsigned long		trans_start;
621 
622 	unsigned long		state;
623 
624 #ifdef CONFIG_BQL
625 	struct dql		dql;
626 #endif
627 } ____cacheline_aligned_in_smp;
628 
629 extern int sysctl_fb_tunnels_only_for_init_net;
630 
631 static inline bool net_has_fallback_tunnels(const struct net *net)
632 {
633 	return net == &init_net ||
634 	       !IS_ENABLED(CONFIG_SYSCTL) ||
635 	       !sysctl_fb_tunnels_only_for_init_net;
636 }
637 
638 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
639 {
640 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
641 	return q->numa_node;
642 #else
643 	return NUMA_NO_NODE;
644 #endif
645 }
646 
647 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
648 {
649 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
650 	q->numa_node = node;
651 #endif
652 }
653 
654 #ifdef CONFIG_RPS
655 /*
656  * This structure holds an RPS map which can be of variable length.  The
657  * map is an array of CPUs.
658  */
659 struct rps_map {
660 	unsigned int len;
661 	struct rcu_head rcu;
662 	u16 cpus[0];
663 };
664 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
665 
666 /*
667  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
668  * tail pointer for that CPU's input queue at the time of last enqueue, and
669  * a hardware filter index.
670  */
671 struct rps_dev_flow {
672 	u16 cpu;
673 	u16 filter;
674 	unsigned int last_qtail;
675 };
676 #define RPS_NO_FILTER 0xffff
677 
678 /*
679  * The rps_dev_flow_table structure contains a table of flow mappings.
680  */
681 struct rps_dev_flow_table {
682 	unsigned int mask;
683 	struct rcu_head rcu;
684 	struct rps_dev_flow flows[0];
685 };
686 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
687     ((_num) * sizeof(struct rps_dev_flow)))
688 
689 /*
690  * The rps_sock_flow_table contains mappings of flows to the last CPU
691  * on which they were processed by the application (set in recvmsg).
692  * Each entry is a 32bit value. Upper part is the high-order bits
693  * of flow hash, lower part is CPU number.
694  * rps_cpu_mask is used to partition the space, depending on number of
695  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
696  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
697  * meaning we use 32-6=26 bits for the hash.
698  */
699 struct rps_sock_flow_table {
700 	u32	mask;
701 
702 	u32	ents[0] ____cacheline_aligned_in_smp;
703 };
704 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
705 
706 #define RPS_NO_CPU 0xffff
707 
708 extern u32 rps_cpu_mask;
709 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
710 
711 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
712 					u32 hash)
713 {
714 	if (table && hash) {
715 		unsigned int index = hash & table->mask;
716 		u32 val = hash & ~rps_cpu_mask;
717 
718 		/* We only give a hint, preemption can change CPU under us */
719 		val |= raw_smp_processor_id();
720 
721 		if (table->ents[index] != val)
722 			table->ents[index] = val;
723 	}
724 }
725 
726 #ifdef CONFIG_RFS_ACCEL
727 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
728 			 u16 filter_id);
729 #endif
730 #endif /* CONFIG_RPS */
731 
732 /* This structure contains an instance of an RX queue. */
733 struct netdev_rx_queue {
734 #ifdef CONFIG_RPS
735 	struct rps_map __rcu		*rps_map;
736 	struct rps_dev_flow_table __rcu	*rps_flow_table;
737 #endif
738 	struct kobject			kobj;
739 	struct net_device		*dev;
740 	struct xdp_rxq_info		xdp_rxq;
741 } ____cacheline_aligned_in_smp;
742 
743 /*
744  * RX queue sysfs structures and functions.
745  */
746 struct rx_queue_attribute {
747 	struct attribute attr;
748 	ssize_t (*show)(struct netdev_rx_queue *queue, char *buf);
749 	ssize_t (*store)(struct netdev_rx_queue *queue,
750 			 const char *buf, size_t len);
751 };
752 
753 #ifdef CONFIG_XPS
754 /*
755  * This structure holds an XPS map which can be of variable length.  The
756  * map is an array of queues.
757  */
758 struct xps_map {
759 	unsigned int len;
760 	unsigned int alloc_len;
761 	struct rcu_head rcu;
762 	u16 queues[0];
763 };
764 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
765 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
766        - sizeof(struct xps_map)) / sizeof(u16))
767 
768 /*
769  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
770  */
771 struct xps_dev_maps {
772 	struct rcu_head rcu;
773 	struct xps_map __rcu *attr_map[0]; /* Either CPUs map or RXQs map */
774 };
775 
776 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +	\
777 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
778 
779 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
780 	(_rxqs * (_tcs) * sizeof(struct xps_map *)))
781 
782 #endif /* CONFIG_XPS */
783 
784 #define TC_MAX_QUEUE	16
785 #define TC_BITMASK	15
786 /* HW offloaded queuing disciplines txq count and offset maps */
787 struct netdev_tc_txq {
788 	u16 count;
789 	u16 offset;
790 };
791 
792 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
793 /*
794  * This structure is to hold information about the device
795  * configured to run FCoE protocol stack.
796  */
797 struct netdev_fcoe_hbainfo {
798 	char	manufacturer[64];
799 	char	serial_number[64];
800 	char	hardware_version[64];
801 	char	driver_version[64];
802 	char	optionrom_version[64];
803 	char	firmware_version[64];
804 	char	model[256];
805 	char	model_description[256];
806 };
807 #endif
808 
809 #define MAX_PHYS_ITEM_ID_LEN 32
810 
811 /* This structure holds a unique identifier to identify some
812  * physical item (port for example) used by a netdevice.
813  */
814 struct netdev_phys_item_id {
815 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
816 	unsigned char id_len;
817 };
818 
819 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
820 					    struct netdev_phys_item_id *b)
821 {
822 	return a->id_len == b->id_len &&
823 	       memcmp(a->id, b->id, a->id_len) == 0;
824 }
825 
826 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
827 				       struct sk_buff *skb,
828 				       struct net_device *sb_dev);
829 
830 enum tc_setup_type {
831 	TC_SETUP_QDISC_MQPRIO,
832 	TC_SETUP_CLSU32,
833 	TC_SETUP_CLSFLOWER,
834 	TC_SETUP_CLSMATCHALL,
835 	TC_SETUP_CLSBPF,
836 	TC_SETUP_BLOCK,
837 	TC_SETUP_QDISC_CBS,
838 	TC_SETUP_QDISC_RED,
839 	TC_SETUP_QDISC_PRIO,
840 	TC_SETUP_QDISC_MQ,
841 	TC_SETUP_QDISC_ETF,
842 };
843 
844 /* These structures hold the attributes of bpf state that are being passed
845  * to the netdevice through the bpf op.
846  */
847 enum bpf_netdev_command {
848 	/* Set or clear a bpf program used in the earliest stages of packet
849 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
850 	 * is responsible for calling bpf_prog_put on any old progs that are
851 	 * stored. In case of error, the callee need not release the new prog
852 	 * reference, but on success it takes ownership and must bpf_prog_put
853 	 * when it is no longer used.
854 	 */
855 	XDP_SETUP_PROG,
856 	XDP_SETUP_PROG_HW,
857 	XDP_QUERY_PROG,
858 	XDP_QUERY_PROG_HW,
859 	/* BPF program for offload callbacks, invoked at program load time. */
860 	BPF_OFFLOAD_VERIFIER_PREP,
861 	BPF_OFFLOAD_TRANSLATE,
862 	BPF_OFFLOAD_DESTROY,
863 	BPF_OFFLOAD_MAP_ALLOC,
864 	BPF_OFFLOAD_MAP_FREE,
865 	XDP_QUERY_XSK_UMEM,
866 	XDP_SETUP_XSK_UMEM,
867 };
868 
869 struct bpf_prog_offload_ops;
870 struct netlink_ext_ack;
871 struct xdp_umem;
872 
873 struct netdev_bpf {
874 	enum bpf_netdev_command command;
875 	union {
876 		/* XDP_SETUP_PROG */
877 		struct {
878 			u32 flags;
879 			struct bpf_prog *prog;
880 			struct netlink_ext_ack *extack;
881 		};
882 		/* XDP_QUERY_PROG, XDP_QUERY_PROG_HW */
883 		struct {
884 			u32 prog_id;
885 			/* flags with which program was installed */
886 			u32 prog_flags;
887 		};
888 		/* BPF_OFFLOAD_VERIFIER_PREP */
889 		struct {
890 			struct bpf_prog *prog;
891 			const struct bpf_prog_offload_ops *ops; /* callee set */
892 		} verifier;
893 		/* BPF_OFFLOAD_TRANSLATE, BPF_OFFLOAD_DESTROY */
894 		struct {
895 			struct bpf_prog *prog;
896 		} offload;
897 		/* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
898 		struct {
899 			struct bpf_offloaded_map *offmap;
900 		};
901 		/* XDP_QUERY_XSK_UMEM, XDP_SETUP_XSK_UMEM */
902 		struct {
903 			struct xdp_umem *umem; /* out for query*/
904 			u16 queue_id; /* in for query */
905 		} xsk;
906 	};
907 };
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 #if IS_ENABLED(CONFIG_TLS_DEVICE)
921 enum tls_offload_ctx_dir {
922 	TLS_OFFLOAD_CTX_DIR_RX,
923 	TLS_OFFLOAD_CTX_DIR_TX,
924 };
925 
926 struct tls_crypto_info;
927 struct tls_context;
928 
929 struct tlsdev_ops {
930 	int (*tls_dev_add)(struct net_device *netdev, struct sock *sk,
931 			   enum tls_offload_ctx_dir direction,
932 			   struct tls_crypto_info *crypto_info,
933 			   u32 start_offload_tcp_sn);
934 	void (*tls_dev_del)(struct net_device *netdev,
935 			    struct tls_context *ctx,
936 			    enum tls_offload_ctx_dir direction);
937 	void (*tls_dev_resync_rx)(struct net_device *netdev,
938 				  struct sock *sk, u32 seq, u64 rcd_sn);
939 };
940 #endif
941 
942 struct dev_ifalias {
943 	struct rcu_head rcuhead;
944 	char ifalias[];
945 };
946 
947 /*
948  * This structure defines the management hooks for network devices.
949  * The following hooks can be defined; unless noted otherwise, they are
950  * optional and can be filled with a null pointer.
951  *
952  * int (*ndo_init)(struct net_device *dev);
953  *     This function is called once when a network device is registered.
954  *     The network device can use this for any late stage initialization
955  *     or semantic validation. It can fail with an error code which will
956  *     be propagated back to register_netdev.
957  *
958  * void (*ndo_uninit)(struct net_device *dev);
959  *     This function is called when device is unregistered or when registration
960  *     fails. It is not called if init fails.
961  *
962  * int (*ndo_open)(struct net_device *dev);
963  *     This function is called when a network device transitions to the up
964  *     state.
965  *
966  * int (*ndo_stop)(struct net_device *dev);
967  *     This function is called when a network device transitions to the down
968  *     state.
969  *
970  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
971  *                               struct net_device *dev);
972  *	Called when a packet needs to be transmitted.
973  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
974  *	the queue before that can happen; it's for obsolete devices and weird
975  *	corner cases, but the stack really does a non-trivial amount
976  *	of useless work if you return NETDEV_TX_BUSY.
977  *	Required; cannot be NULL.
978  *
979  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
980  *					   struct net_device *dev
981  *					   netdev_features_t features);
982  *	Called by core transmit path to determine if device is capable of
983  *	performing offload operations on a given packet. This is to give
984  *	the device an opportunity to implement any restrictions that cannot
985  *	be otherwise expressed by feature flags. The check is called with
986  *	the set of features that the stack has calculated and it returns
987  *	those the driver believes to be appropriate.
988  *
989  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
990  *                         struct net_device *sb_dev,
991  *                         select_queue_fallback_t fallback);
992  *	Called to decide which queue to use when device supports multiple
993  *	transmit queues.
994  *
995  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
996  *	This function is called to allow device receiver to make
997  *	changes to configuration when multicast or promiscuous is enabled.
998  *
999  * void (*ndo_set_rx_mode)(struct net_device *dev);
1000  *	This function is called device changes address list filtering.
1001  *	If driver handles unicast address filtering, it should set
1002  *	IFF_UNICAST_FLT in its priv_flags.
1003  *
1004  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1005  *	This function  is called when the Media Access Control address
1006  *	needs to be changed. If this interface is not defined, the
1007  *	MAC address can not be changed.
1008  *
1009  * int (*ndo_validate_addr)(struct net_device *dev);
1010  *	Test if Media Access Control address is valid for the device.
1011  *
1012  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1013  *	Called when a user requests an ioctl which can't be handled by
1014  *	the generic interface code. If not defined ioctls return
1015  *	not supported error code.
1016  *
1017  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1018  *	Used to set network devices bus interface parameters. This interface
1019  *	is retained for legacy reasons; new devices should use the bus
1020  *	interface (PCI) for low level management.
1021  *
1022  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1023  *	Called when a user wants to change the Maximum Transfer Unit
1024  *	of a device.
1025  *
1026  * void (*ndo_tx_timeout)(struct net_device *dev);
1027  *	Callback used when the transmitter has not made any progress
1028  *	for dev->watchdog ticks.
1029  *
1030  * void (*ndo_get_stats64)(struct net_device *dev,
1031  *                         struct rtnl_link_stats64 *storage);
1032  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1033  *	Called when a user wants to get the network device usage
1034  *	statistics. Drivers must do one of the following:
1035  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
1036  *	   rtnl_link_stats64 structure passed by the caller.
1037  *	2. Define @ndo_get_stats to update a net_device_stats structure
1038  *	   (which should normally be dev->stats) and return a pointer to
1039  *	   it. The structure may be changed asynchronously only if each
1040  *	   field is written atomically.
1041  *	3. Update dev->stats asynchronously and atomically, and define
1042  *	   neither operation.
1043  *
1044  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1045  *	Return true if this device supports offload stats of this attr_id.
1046  *
1047  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1048  *	void *attr_data)
1049  *	Get statistics for offload operations by attr_id. Write it into the
1050  *	attr_data pointer.
1051  *
1052  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1053  *	If device supports VLAN filtering this function is called when a
1054  *	VLAN id is registered.
1055  *
1056  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1057  *	If device supports VLAN filtering this function is called when a
1058  *	VLAN id is unregistered.
1059  *
1060  * void (*ndo_poll_controller)(struct net_device *dev);
1061  *
1062  *	SR-IOV management functions.
1063  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1064  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1065  *			  u8 qos, __be16 proto);
1066  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1067  *			  int max_tx_rate);
1068  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1069  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1070  * int (*ndo_get_vf_config)(struct net_device *dev,
1071  *			    int vf, struct ifla_vf_info *ivf);
1072  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1073  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1074  *			  struct nlattr *port[]);
1075  *
1076  *      Enable or disable the VF ability to query its RSS Redirection Table and
1077  *      Hash Key. This is needed since on some devices VF share this information
1078  *      with PF and querying it may introduce a theoretical security risk.
1079  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1080  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1081  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1082  *		       void *type_data);
1083  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
1084  *	This is always called from the stack with the rtnl lock held and netif
1085  *	tx queues stopped. This allows the netdevice to perform queue
1086  *	management safely.
1087  *
1088  *	Fiber Channel over Ethernet (FCoE) offload functions.
1089  * int (*ndo_fcoe_enable)(struct net_device *dev);
1090  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
1091  *	so the underlying device can perform whatever needed configuration or
1092  *	initialization to support acceleration of FCoE traffic.
1093  *
1094  * int (*ndo_fcoe_disable)(struct net_device *dev);
1095  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
1096  *	so the underlying device can perform whatever needed clean-ups to
1097  *	stop supporting acceleration of FCoE traffic.
1098  *
1099  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1100  *			     struct scatterlist *sgl, unsigned int sgc);
1101  *	Called when the FCoE Initiator wants to initialize an I/O that
1102  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1103  *	perform necessary setup and returns 1 to indicate the device is set up
1104  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1105  *
1106  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1107  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1108  *	indicated by the FC exchange id 'xid', so the underlying device can
1109  *	clean up and reuse resources for later DDP requests.
1110  *
1111  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1112  *			      struct scatterlist *sgl, unsigned int sgc);
1113  *	Called when the FCoE Target wants to initialize an I/O that
1114  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1115  *	perform necessary setup and returns 1 to indicate the device is set up
1116  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1117  *
1118  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1119  *			       struct netdev_fcoe_hbainfo *hbainfo);
1120  *	Called when the FCoE Protocol stack wants information on the underlying
1121  *	device. This information is utilized by the FCoE protocol stack to
1122  *	register attributes with Fiber Channel management service as per the
1123  *	FC-GS Fabric Device Management Information(FDMI) specification.
1124  *
1125  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1126  *	Called when the underlying device wants to override default World Wide
1127  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1128  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1129  *	protocol stack to use.
1130  *
1131  *	RFS acceleration.
1132  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1133  *			    u16 rxq_index, u32 flow_id);
1134  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1135  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1136  *	Return the filter ID on success, or a negative error code.
1137  *
1138  *	Slave management functions (for bridge, bonding, etc).
1139  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1140  *	Called to make another netdev an underling.
1141  *
1142  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1143  *	Called to release previously enslaved netdev.
1144  *
1145  *      Feature/offload setting functions.
1146  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1147  *		netdev_features_t features);
1148  *	Adjusts the requested feature flags according to device-specific
1149  *	constraints, and returns the resulting flags. Must not modify
1150  *	the device state.
1151  *
1152  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1153  *	Called to update device configuration to new features. Passed
1154  *	feature set might be less than what was returned by ndo_fix_features()).
1155  *	Must return >0 or -errno if it changed dev->features itself.
1156  *
1157  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1158  *		      struct net_device *dev,
1159  *		      const unsigned char *addr, u16 vid, u16 flags)
1160  *	Adds an FDB entry to dev for addr.
1161  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1162  *		      struct net_device *dev,
1163  *		      const unsigned char *addr, u16 vid)
1164  *	Deletes the FDB entry from dev coresponding to addr.
1165  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1166  *		       struct net_device *dev, struct net_device *filter_dev,
1167  *		       int *idx)
1168  *	Used to add FDB entries to dump requests. Implementers should add
1169  *	entries to skb and update idx with the number of entries.
1170  *
1171  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1172  *			     u16 flags)
1173  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1174  *			     struct net_device *dev, u32 filter_mask,
1175  *			     int nlflags)
1176  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1177  *			     u16 flags);
1178  *
1179  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1180  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1181  *	which do not represent real hardware may define this to allow their
1182  *	userspace components to manage their virtual carrier state. Devices
1183  *	that determine carrier state from physical hardware properties (eg
1184  *	network cables) or protocol-dependent mechanisms (eg
1185  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1186  *
1187  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1188  *			       struct netdev_phys_item_id *ppid);
1189  *	Called to get ID of physical port of this device. If driver does
1190  *	not implement this, it is assumed that the hw is not able to have
1191  *	multiple net devices on single physical port.
1192  *
1193  * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1194  *			      struct udp_tunnel_info *ti);
1195  *	Called by UDP tunnel to notify a driver about the UDP port and socket
1196  *	address family that a UDP tunnel is listnening to. It is called only
1197  *	when a new port starts listening. The operation is protected by the
1198  *	RTNL.
1199  *
1200  * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1201  *			      struct udp_tunnel_info *ti);
1202  *	Called by UDP tunnel to notify the driver about a UDP port and socket
1203  *	address family that the UDP tunnel is not listening to anymore. The
1204  *	operation is protected by the RTNL.
1205  *
1206  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1207  *				 struct net_device *dev)
1208  *	Called by upper layer devices to accelerate switching or other
1209  *	station functionality into hardware. 'pdev is the lowerdev
1210  *	to use for the offload and 'dev' is the net device that will
1211  *	back the offload. Returns a pointer to the private structure
1212  *	the upper layer will maintain.
1213  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1214  *	Called by upper layer device to delete the station created
1215  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1216  *	the station and priv is the structure returned by the add
1217  *	operation.
1218  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1219  *			     int queue_index, u32 maxrate);
1220  *	Called when a user wants to set a max-rate limitation of specific
1221  *	TX queue.
1222  * int (*ndo_get_iflink)(const struct net_device *dev);
1223  *	Called to get the iflink value of this device.
1224  * void (*ndo_change_proto_down)(struct net_device *dev,
1225  *				 bool proto_down);
1226  *	This function is used to pass protocol port error state information
1227  *	to the switch driver. The switch driver can react to the proto_down
1228  *      by doing a phys down on the associated switch port.
1229  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1230  *	This function is used to get egress tunnel information for given skb.
1231  *	This is useful for retrieving outer tunnel header parameters while
1232  *	sampling packet.
1233  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1234  *	This function is used to specify the headroom that the skb must
1235  *	consider when allocation skb during packet reception. Setting
1236  *	appropriate rx headroom value allows avoiding skb head copy on
1237  *	forward. Setting a negative value resets the rx headroom to the
1238  *	default value.
1239  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1240  *	This function is used to set or query state related to XDP on the
1241  *	netdevice and manage BPF offload. See definition of
1242  *	enum bpf_netdev_command for details.
1243  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1244  *			u32 flags);
1245  *	This function is used to submit @n XDP packets for transmit on a
1246  *	netdevice. Returns number of frames successfully transmitted, frames
1247  *	that got dropped are freed/returned via xdp_return_frame().
1248  *	Returns negative number, means general error invoking ndo, meaning
1249  *	no frames were xmit'ed and core-caller will free all frames.
1250  */
1251 struct net_device_ops {
1252 	int			(*ndo_init)(struct net_device *dev);
1253 	void			(*ndo_uninit)(struct net_device *dev);
1254 	int			(*ndo_open)(struct net_device *dev);
1255 	int			(*ndo_stop)(struct net_device *dev);
1256 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1257 						  struct net_device *dev);
1258 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1259 						      struct net_device *dev,
1260 						      netdev_features_t features);
1261 	u16			(*ndo_select_queue)(struct net_device *dev,
1262 						    struct sk_buff *skb,
1263 						    struct net_device *sb_dev,
1264 						    select_queue_fallback_t fallback);
1265 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1266 						       int flags);
1267 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1268 	int			(*ndo_set_mac_address)(struct net_device *dev,
1269 						       void *addr);
1270 	int			(*ndo_validate_addr)(struct net_device *dev);
1271 	int			(*ndo_do_ioctl)(struct net_device *dev,
1272 					        struct ifreq *ifr, int cmd);
1273 	int			(*ndo_set_config)(struct net_device *dev,
1274 					          struct ifmap *map);
1275 	int			(*ndo_change_mtu)(struct net_device *dev,
1276 						  int new_mtu);
1277 	int			(*ndo_neigh_setup)(struct net_device *dev,
1278 						   struct neigh_parms *);
1279 	void			(*ndo_tx_timeout) (struct net_device *dev);
1280 
1281 	void			(*ndo_get_stats64)(struct net_device *dev,
1282 						   struct rtnl_link_stats64 *storage);
1283 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1284 	int			(*ndo_get_offload_stats)(int attr_id,
1285 							 const struct net_device *dev,
1286 							 void *attr_data);
1287 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1288 
1289 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1290 						       __be16 proto, u16 vid);
1291 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1292 						        __be16 proto, u16 vid);
1293 #ifdef CONFIG_NET_POLL_CONTROLLER
1294 	void                    (*ndo_poll_controller)(struct net_device *dev);
1295 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1296 						     struct netpoll_info *info);
1297 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1298 #endif
1299 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1300 						  int queue, u8 *mac);
1301 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1302 						   int queue, u16 vlan,
1303 						   u8 qos, __be16 proto);
1304 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1305 						   int vf, int min_tx_rate,
1306 						   int max_tx_rate);
1307 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1308 						       int vf, bool setting);
1309 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1310 						    int vf, bool setting);
1311 	int			(*ndo_get_vf_config)(struct net_device *dev,
1312 						     int vf,
1313 						     struct ifla_vf_info *ivf);
1314 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1315 							 int vf, int link_state);
1316 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1317 						    int vf,
1318 						    struct ifla_vf_stats
1319 						    *vf_stats);
1320 	int			(*ndo_set_vf_port)(struct net_device *dev,
1321 						   int vf,
1322 						   struct nlattr *port[]);
1323 	int			(*ndo_get_vf_port)(struct net_device *dev,
1324 						   int vf, struct sk_buff *skb);
1325 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1326 						   int vf, u64 guid,
1327 						   int guid_type);
1328 	int			(*ndo_set_vf_rss_query_en)(
1329 						   struct net_device *dev,
1330 						   int vf, bool setting);
1331 	int			(*ndo_setup_tc)(struct net_device *dev,
1332 						enum tc_setup_type type,
1333 						void *type_data);
1334 #if IS_ENABLED(CONFIG_FCOE)
1335 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1336 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1337 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1338 						      u16 xid,
1339 						      struct scatterlist *sgl,
1340 						      unsigned int sgc);
1341 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1342 						     u16 xid);
1343 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1344 						       u16 xid,
1345 						       struct scatterlist *sgl,
1346 						       unsigned int sgc);
1347 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1348 							struct netdev_fcoe_hbainfo *hbainfo);
1349 #endif
1350 
1351 #if IS_ENABLED(CONFIG_LIBFCOE)
1352 #define NETDEV_FCOE_WWNN 0
1353 #define NETDEV_FCOE_WWPN 1
1354 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1355 						    u64 *wwn, int type);
1356 #endif
1357 
1358 #ifdef CONFIG_RFS_ACCEL
1359 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1360 						     const struct sk_buff *skb,
1361 						     u16 rxq_index,
1362 						     u32 flow_id);
1363 #endif
1364 	int			(*ndo_add_slave)(struct net_device *dev,
1365 						 struct net_device *slave_dev,
1366 						 struct netlink_ext_ack *extack);
1367 	int			(*ndo_del_slave)(struct net_device *dev,
1368 						 struct net_device *slave_dev);
1369 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1370 						    netdev_features_t features);
1371 	int			(*ndo_set_features)(struct net_device *dev,
1372 						    netdev_features_t features);
1373 	int			(*ndo_neigh_construct)(struct net_device *dev,
1374 						       struct neighbour *n);
1375 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1376 						     struct neighbour *n);
1377 
1378 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1379 					       struct nlattr *tb[],
1380 					       struct net_device *dev,
1381 					       const unsigned char *addr,
1382 					       u16 vid,
1383 					       u16 flags);
1384 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1385 					       struct nlattr *tb[],
1386 					       struct net_device *dev,
1387 					       const unsigned char *addr,
1388 					       u16 vid);
1389 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1390 						struct netlink_callback *cb,
1391 						struct net_device *dev,
1392 						struct net_device *filter_dev,
1393 						int *idx);
1394 
1395 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1396 						      struct nlmsghdr *nlh,
1397 						      u16 flags);
1398 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1399 						      u32 pid, u32 seq,
1400 						      struct net_device *dev,
1401 						      u32 filter_mask,
1402 						      int nlflags);
1403 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1404 						      struct nlmsghdr *nlh,
1405 						      u16 flags);
1406 	int			(*ndo_change_carrier)(struct net_device *dev,
1407 						      bool new_carrier);
1408 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1409 							struct netdev_phys_item_id *ppid);
1410 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1411 							  char *name, size_t len);
1412 	void			(*ndo_udp_tunnel_add)(struct net_device *dev,
1413 						      struct udp_tunnel_info *ti);
1414 	void			(*ndo_udp_tunnel_del)(struct net_device *dev,
1415 						      struct udp_tunnel_info *ti);
1416 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1417 							struct net_device *dev);
1418 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1419 							void *priv);
1420 
1421 	int			(*ndo_get_lock_subclass)(struct net_device *dev);
1422 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1423 						      int queue_index,
1424 						      u32 maxrate);
1425 	int			(*ndo_get_iflink)(const struct net_device *dev);
1426 	int			(*ndo_change_proto_down)(struct net_device *dev,
1427 							 bool proto_down);
1428 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1429 						       struct sk_buff *skb);
1430 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1431 						       int needed_headroom);
1432 	int			(*ndo_bpf)(struct net_device *dev,
1433 					   struct netdev_bpf *bpf);
1434 	int			(*ndo_xdp_xmit)(struct net_device *dev, int n,
1435 						struct xdp_frame **xdp,
1436 						u32 flags);
1437 	int			(*ndo_xsk_async_xmit)(struct net_device *dev,
1438 						      u32 queue_id);
1439 };
1440 
1441 /**
1442  * enum net_device_priv_flags - &struct net_device priv_flags
1443  *
1444  * These are the &struct net_device, they are only set internally
1445  * by drivers and used in the kernel. These flags are invisible to
1446  * userspace; this means that the order of these flags can change
1447  * during any kernel release.
1448  *
1449  * You should have a pretty good reason to be extending these flags.
1450  *
1451  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1452  * @IFF_EBRIDGE: Ethernet bridging device
1453  * @IFF_BONDING: bonding master or slave
1454  * @IFF_ISATAP: ISATAP interface (RFC4214)
1455  * @IFF_WAN_HDLC: WAN HDLC device
1456  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1457  *	release skb->dst
1458  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1459  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1460  * @IFF_MACVLAN_PORT: device used as macvlan port
1461  * @IFF_BRIDGE_PORT: device used as bridge port
1462  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1463  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1464  * @IFF_UNICAST_FLT: Supports unicast filtering
1465  * @IFF_TEAM_PORT: device used as team port
1466  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1467  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1468  *	change when it's running
1469  * @IFF_MACVLAN: Macvlan device
1470  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1471  *	underlying stacked devices
1472  * @IFF_L3MDEV_MASTER: device is an L3 master device
1473  * @IFF_NO_QUEUE: device can run without qdisc attached
1474  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1475  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1476  * @IFF_TEAM: device is a team device
1477  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1478  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1479  *	entity (i.e. the master device for bridged veth)
1480  * @IFF_MACSEC: device is a MACsec device
1481  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1482  * @IFF_FAILOVER: device is a failover master device
1483  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1484  */
1485 enum netdev_priv_flags {
1486 	IFF_802_1Q_VLAN			= 1<<0,
1487 	IFF_EBRIDGE			= 1<<1,
1488 	IFF_BONDING			= 1<<2,
1489 	IFF_ISATAP			= 1<<3,
1490 	IFF_WAN_HDLC			= 1<<4,
1491 	IFF_XMIT_DST_RELEASE		= 1<<5,
1492 	IFF_DONT_BRIDGE			= 1<<6,
1493 	IFF_DISABLE_NETPOLL		= 1<<7,
1494 	IFF_MACVLAN_PORT		= 1<<8,
1495 	IFF_BRIDGE_PORT			= 1<<9,
1496 	IFF_OVS_DATAPATH		= 1<<10,
1497 	IFF_TX_SKB_SHARING		= 1<<11,
1498 	IFF_UNICAST_FLT			= 1<<12,
1499 	IFF_TEAM_PORT			= 1<<13,
1500 	IFF_SUPP_NOFCS			= 1<<14,
1501 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1502 	IFF_MACVLAN			= 1<<16,
1503 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1504 	IFF_L3MDEV_MASTER		= 1<<18,
1505 	IFF_NO_QUEUE			= 1<<19,
1506 	IFF_OPENVSWITCH			= 1<<20,
1507 	IFF_L3MDEV_SLAVE		= 1<<21,
1508 	IFF_TEAM			= 1<<22,
1509 	IFF_RXFH_CONFIGURED		= 1<<23,
1510 	IFF_PHONY_HEADROOM		= 1<<24,
1511 	IFF_MACSEC			= 1<<25,
1512 	IFF_NO_RX_HANDLER		= 1<<26,
1513 	IFF_FAILOVER			= 1<<27,
1514 	IFF_FAILOVER_SLAVE		= 1<<28,
1515 };
1516 
1517 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1518 #define IFF_EBRIDGE			IFF_EBRIDGE
1519 #define IFF_BONDING			IFF_BONDING
1520 #define IFF_ISATAP			IFF_ISATAP
1521 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1522 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1523 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1524 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1525 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1526 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1527 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1528 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1529 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1530 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1531 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1532 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1533 #define IFF_MACVLAN			IFF_MACVLAN
1534 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1535 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1536 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1537 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1538 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1539 #define IFF_TEAM			IFF_TEAM
1540 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1541 #define IFF_MACSEC			IFF_MACSEC
1542 #define IFF_NO_RX_HANDLER		IFF_NO_RX_HANDLER
1543 #define IFF_FAILOVER			IFF_FAILOVER
1544 #define IFF_FAILOVER_SLAVE		IFF_FAILOVER_SLAVE
1545 
1546 /**
1547  *	struct net_device - The DEVICE structure.
1548  *
1549  *	Actually, this whole structure is a big mistake.  It mixes I/O
1550  *	data with strictly "high-level" data, and it has to know about
1551  *	almost every data structure used in the INET module.
1552  *
1553  *	@name:	This is the first field of the "visible" part of this structure
1554  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1555  *		of the interface.
1556  *
1557  *	@name_hlist: 	Device name hash chain, please keep it close to name[]
1558  *	@ifalias:	SNMP alias
1559  *	@mem_end:	Shared memory end
1560  *	@mem_start:	Shared memory start
1561  *	@base_addr:	Device I/O address
1562  *	@irq:		Device IRQ number
1563  *
1564  *	@state:		Generic network queuing layer state, see netdev_state_t
1565  *	@dev_list:	The global list of network devices
1566  *	@napi_list:	List entry used for polling NAPI devices
1567  *	@unreg_list:	List entry  when we are unregistering the
1568  *			device; see the function unregister_netdev
1569  *	@close_list:	List entry used when we are closing the device
1570  *	@ptype_all:     Device-specific packet handlers for all protocols
1571  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1572  *
1573  *	@adj_list:	Directly linked devices, like slaves for bonding
1574  *	@features:	Currently active device features
1575  *	@hw_features:	User-changeable features
1576  *
1577  *	@wanted_features:	User-requested features
1578  *	@vlan_features:		Mask of features inheritable by VLAN devices
1579  *
1580  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1581  *				This field indicates what encapsulation
1582  *				offloads the hardware is capable of doing,
1583  *				and drivers will need to set them appropriately.
1584  *
1585  *	@mpls_features:	Mask of features inheritable by MPLS
1586  *
1587  *	@ifindex:	interface index
1588  *	@group:		The group the device belongs to
1589  *
1590  *	@stats:		Statistics struct, which was left as a legacy, use
1591  *			rtnl_link_stats64 instead
1592  *
1593  *	@rx_dropped:	Dropped packets by core network,
1594  *			do not use this in drivers
1595  *	@tx_dropped:	Dropped packets by core network,
1596  *			do not use this in drivers
1597  *	@rx_nohandler:	nohandler dropped packets by core network on
1598  *			inactive devices, do not use this in drivers
1599  *	@carrier_up_count:	Number of times the carrier has been up
1600  *	@carrier_down_count:	Number of times the carrier has been down
1601  *
1602  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1603  *				instead of ioctl,
1604  *				see <net/iw_handler.h> for details.
1605  *	@wireless_data:	Instance data managed by the core of wireless extensions
1606  *
1607  *	@netdev_ops:	Includes several pointers to callbacks,
1608  *			if one wants to override the ndo_*() functions
1609  *	@ethtool_ops:	Management operations
1610  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1611  *			discovery handling. Necessary for e.g. 6LoWPAN.
1612  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1613  *			of Layer 2 headers.
1614  *
1615  *	@flags:		Interface flags (a la BSD)
1616  *	@priv_flags:	Like 'flags' but invisible to userspace,
1617  *			see if.h for the definitions
1618  *	@gflags:	Global flags ( kept as legacy )
1619  *	@padded:	How much padding added by alloc_netdev()
1620  *	@operstate:	RFC2863 operstate
1621  *	@link_mode:	Mapping policy to operstate
1622  *	@if_port:	Selectable AUI, TP, ...
1623  *	@dma:		DMA channel
1624  *	@mtu:		Interface MTU value
1625  *	@min_mtu:	Interface Minimum MTU value
1626  *	@max_mtu:	Interface Maximum MTU value
1627  *	@type:		Interface hardware type
1628  *	@hard_header_len: Maximum hardware header length.
1629  *	@min_header_len:  Minimum hardware header length
1630  *
1631  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1632  *			  cases can this be guaranteed
1633  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1634  *			  cases can this be guaranteed. Some cases also use
1635  *			  LL_MAX_HEADER instead to allocate the skb
1636  *
1637  *	interface address info:
1638  *
1639  * 	@perm_addr:		Permanent hw address
1640  * 	@addr_assign_type:	Hw address assignment type
1641  * 	@addr_len:		Hardware address length
1642  *	@neigh_priv_len:	Used in neigh_alloc()
1643  * 	@dev_id:		Used to differentiate devices that share
1644  * 				the same link layer address
1645  * 	@dev_port:		Used to differentiate devices that share
1646  * 				the same function
1647  *	@addr_list_lock:	XXX: need comments on this one
1648  *	@uc_promisc:		Counter that indicates promiscuous mode
1649  *				has been enabled due to the need to listen to
1650  *				additional unicast addresses in a device that
1651  *				does not implement ndo_set_rx_mode()
1652  *	@uc:			unicast mac addresses
1653  *	@mc:			multicast mac addresses
1654  *	@dev_addrs:		list of device hw addresses
1655  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1656  *	@promiscuity:		Number of times the NIC is told to work in
1657  *				promiscuous mode; if it becomes 0 the NIC will
1658  *				exit promiscuous mode
1659  *	@allmulti:		Counter, enables or disables allmulticast mode
1660  *
1661  *	@vlan_info:	VLAN info
1662  *	@dsa_ptr:	dsa specific data
1663  *	@tipc_ptr:	TIPC specific data
1664  *	@atalk_ptr:	AppleTalk link
1665  *	@ip_ptr:	IPv4 specific data
1666  *	@dn_ptr:	DECnet specific data
1667  *	@ip6_ptr:	IPv6 specific data
1668  *	@ax25_ptr:	AX.25 specific data
1669  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1670  *
1671  *	@dev_addr:	Hw address (before bcast,
1672  *			because most packets are unicast)
1673  *
1674  *	@_rx:			Array of RX queues
1675  *	@num_rx_queues:		Number of RX queues
1676  *				allocated at register_netdev() time
1677  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1678  *
1679  *	@rx_handler:		handler for received packets
1680  *	@rx_handler_data: 	XXX: need comments on this one
1681  *	@miniq_ingress:		ingress/clsact qdisc specific data for
1682  *				ingress processing
1683  *	@ingress_queue:		XXX: need comments on this one
1684  *	@broadcast:		hw bcast address
1685  *
1686  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1687  *			indexed by RX queue number. Assigned by driver.
1688  *			This must only be set if the ndo_rx_flow_steer
1689  *			operation is defined
1690  *	@index_hlist:		Device index hash chain
1691  *
1692  *	@_tx:			Array of TX queues
1693  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1694  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1695  *	@qdisc:			Root qdisc from userspace point of view
1696  *	@tx_queue_len:		Max frames per queue allowed
1697  *	@tx_global_lock: 	XXX: need comments on this one
1698  *
1699  *	@xps_maps:	XXX: need comments on this one
1700  *	@miniq_egress:		clsact qdisc specific data for
1701  *				egress processing
1702  *	@watchdog_timeo:	Represents the timeout that is used by
1703  *				the watchdog (see dev_watchdog())
1704  *	@watchdog_timer:	List of timers
1705  *
1706  *	@pcpu_refcnt:		Number of references to this device
1707  *	@todo_list:		Delayed register/unregister
1708  *	@link_watch_list:	XXX: need comments on this one
1709  *
1710  *	@reg_state:		Register/unregister state machine
1711  *	@dismantle:		Device is going to be freed
1712  *	@rtnl_link_state:	This enum represents the phases of creating
1713  *				a new link
1714  *
1715  *	@needs_free_netdev:	Should unregister perform free_netdev?
1716  *	@priv_destructor:	Called from unregister
1717  *	@npinfo:		XXX: need comments on this one
1718  * 	@nd_net:		Network namespace this network device is inside
1719  *
1720  * 	@ml_priv:	Mid-layer private
1721  * 	@lstats:	Loopback statistics
1722  * 	@tstats:	Tunnel statistics
1723  * 	@dstats:	Dummy statistics
1724  * 	@vstats:	Virtual ethernet statistics
1725  *
1726  *	@garp_port:	GARP
1727  *	@mrp_port:	MRP
1728  *
1729  *	@dev:		Class/net/name entry
1730  *	@sysfs_groups:	Space for optional device, statistics and wireless
1731  *			sysfs groups
1732  *
1733  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1734  *	@rtnl_link_ops:	Rtnl_link_ops
1735  *
1736  *	@gso_max_size:	Maximum size of generic segmentation offload
1737  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1738  *			NIC for GSO
1739  *
1740  *	@dcbnl_ops:	Data Center Bridging netlink ops
1741  *	@num_tc:	Number of traffic classes in the net device
1742  *	@tc_to_txq:	XXX: need comments on this one
1743  *	@prio_tc_map:	XXX: need comments on this one
1744  *
1745  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
1746  *
1747  *	@priomap:	XXX: need comments on this one
1748  *	@phydev:	Physical device may attach itself
1749  *			for hardware timestamping
1750  *	@sfp_bus:	attached &struct sfp_bus structure.
1751  *
1752  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1753  *	@qdisc_running_key: lockdep class annotating Qdisc->running seqcount
1754  *
1755  *	@proto_down:	protocol port state information can be sent to the
1756  *			switch driver and used to set the phys state of the
1757  *			switch port.
1758  *
1759  *	FIXME: cleanup struct net_device such that network protocol info
1760  *	moves out.
1761  */
1762 
1763 struct net_device {
1764 	char			name[IFNAMSIZ];
1765 	struct hlist_node	name_hlist;
1766 	struct dev_ifalias	__rcu *ifalias;
1767 	/*
1768 	 *	I/O specific fields
1769 	 *	FIXME: Merge these and struct ifmap into one
1770 	 */
1771 	unsigned long		mem_end;
1772 	unsigned long		mem_start;
1773 	unsigned long		base_addr;
1774 	int			irq;
1775 
1776 	/*
1777 	 *	Some hardware also needs these fields (state,dev_list,
1778 	 *	napi_list,unreg_list,close_list) but they are not
1779 	 *	part of the usual set specified in Space.c.
1780 	 */
1781 
1782 	unsigned long		state;
1783 
1784 	struct list_head	dev_list;
1785 	struct list_head	napi_list;
1786 	struct list_head	unreg_list;
1787 	struct list_head	close_list;
1788 	struct list_head	ptype_all;
1789 	struct list_head	ptype_specific;
1790 
1791 	struct {
1792 		struct list_head upper;
1793 		struct list_head lower;
1794 	} adj_list;
1795 
1796 	netdev_features_t	features;
1797 	netdev_features_t	hw_features;
1798 	netdev_features_t	wanted_features;
1799 	netdev_features_t	vlan_features;
1800 	netdev_features_t	hw_enc_features;
1801 	netdev_features_t	mpls_features;
1802 	netdev_features_t	gso_partial_features;
1803 
1804 	int			ifindex;
1805 	int			group;
1806 
1807 	struct net_device_stats	stats;
1808 
1809 	atomic_long_t		rx_dropped;
1810 	atomic_long_t		tx_dropped;
1811 	atomic_long_t		rx_nohandler;
1812 
1813 	/* Stats to monitor link on/off, flapping */
1814 	atomic_t		carrier_up_count;
1815 	atomic_t		carrier_down_count;
1816 
1817 #ifdef CONFIG_WIRELESS_EXT
1818 	const struct iw_handler_def *wireless_handlers;
1819 	struct iw_public_data	*wireless_data;
1820 #endif
1821 	const struct net_device_ops *netdev_ops;
1822 	const struct ethtool_ops *ethtool_ops;
1823 #ifdef CONFIG_NET_SWITCHDEV
1824 	const struct switchdev_ops *switchdev_ops;
1825 #endif
1826 #ifdef CONFIG_NET_L3_MASTER_DEV
1827 	const struct l3mdev_ops	*l3mdev_ops;
1828 #endif
1829 #if IS_ENABLED(CONFIG_IPV6)
1830 	const struct ndisc_ops *ndisc_ops;
1831 #endif
1832 
1833 #ifdef CONFIG_XFRM_OFFLOAD
1834 	const struct xfrmdev_ops *xfrmdev_ops;
1835 #endif
1836 
1837 #if IS_ENABLED(CONFIG_TLS_DEVICE)
1838 	const struct tlsdev_ops *tlsdev_ops;
1839 #endif
1840 
1841 	const struct header_ops *header_ops;
1842 
1843 	unsigned int		flags;
1844 	unsigned int		priv_flags;
1845 
1846 	unsigned short		gflags;
1847 	unsigned short		padded;
1848 
1849 	unsigned char		operstate;
1850 	unsigned char		link_mode;
1851 
1852 	unsigned char		if_port;
1853 	unsigned char		dma;
1854 
1855 	unsigned int		mtu;
1856 	unsigned int		min_mtu;
1857 	unsigned int		max_mtu;
1858 	unsigned short		type;
1859 	unsigned short		hard_header_len;
1860 	unsigned char		min_header_len;
1861 
1862 	unsigned short		needed_headroom;
1863 	unsigned short		needed_tailroom;
1864 
1865 	/* Interface address info. */
1866 	unsigned char		perm_addr[MAX_ADDR_LEN];
1867 	unsigned char		addr_assign_type;
1868 	unsigned char		addr_len;
1869 	unsigned short		neigh_priv_len;
1870 	unsigned short          dev_id;
1871 	unsigned short          dev_port;
1872 	spinlock_t		addr_list_lock;
1873 	unsigned char		name_assign_type;
1874 	bool			uc_promisc;
1875 	struct netdev_hw_addr_list	uc;
1876 	struct netdev_hw_addr_list	mc;
1877 	struct netdev_hw_addr_list	dev_addrs;
1878 
1879 #ifdef CONFIG_SYSFS
1880 	struct kset		*queues_kset;
1881 #endif
1882 	unsigned int		promiscuity;
1883 	unsigned int		allmulti;
1884 
1885 
1886 	/* Protocol-specific pointers */
1887 
1888 #if IS_ENABLED(CONFIG_VLAN_8021Q)
1889 	struct vlan_info __rcu	*vlan_info;
1890 #endif
1891 #if IS_ENABLED(CONFIG_NET_DSA)
1892 	struct dsa_port		*dsa_ptr;
1893 #endif
1894 #if IS_ENABLED(CONFIG_TIPC)
1895 	struct tipc_bearer __rcu *tipc_ptr;
1896 #endif
1897 #if IS_ENABLED(CONFIG_IRDA) || IS_ENABLED(CONFIG_ATALK)
1898 	void 			*atalk_ptr;
1899 #endif
1900 	struct in_device __rcu	*ip_ptr;
1901 #if IS_ENABLED(CONFIG_DECNET)
1902 	struct dn_dev __rcu     *dn_ptr;
1903 #endif
1904 	struct inet6_dev __rcu	*ip6_ptr;
1905 #if IS_ENABLED(CONFIG_AX25)
1906 	void			*ax25_ptr;
1907 #endif
1908 	struct wireless_dev	*ieee80211_ptr;
1909 	struct wpan_dev		*ieee802154_ptr;
1910 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
1911 	struct mpls_dev __rcu	*mpls_ptr;
1912 #endif
1913 
1914 /*
1915  * Cache lines mostly used on receive path (including eth_type_trans())
1916  */
1917 	/* Interface address info used in eth_type_trans() */
1918 	unsigned char		*dev_addr;
1919 
1920 	struct netdev_rx_queue	*_rx;
1921 	unsigned int		num_rx_queues;
1922 	unsigned int		real_num_rx_queues;
1923 
1924 	struct bpf_prog __rcu	*xdp_prog;
1925 	unsigned long		gro_flush_timeout;
1926 	rx_handler_func_t __rcu	*rx_handler;
1927 	void __rcu		*rx_handler_data;
1928 
1929 #ifdef CONFIG_NET_CLS_ACT
1930 	struct mini_Qdisc __rcu	*miniq_ingress;
1931 #endif
1932 	struct netdev_queue __rcu *ingress_queue;
1933 #ifdef CONFIG_NETFILTER_INGRESS
1934 	struct nf_hook_entries __rcu *nf_hooks_ingress;
1935 #endif
1936 
1937 	unsigned char		broadcast[MAX_ADDR_LEN];
1938 #ifdef CONFIG_RFS_ACCEL
1939 	struct cpu_rmap		*rx_cpu_rmap;
1940 #endif
1941 	struct hlist_node	index_hlist;
1942 
1943 /*
1944  * Cache lines mostly used on transmit path
1945  */
1946 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
1947 	unsigned int		num_tx_queues;
1948 	unsigned int		real_num_tx_queues;
1949 	struct Qdisc		*qdisc;
1950 #ifdef CONFIG_NET_SCHED
1951 	DECLARE_HASHTABLE	(qdisc_hash, 4);
1952 #endif
1953 	unsigned int		tx_queue_len;
1954 	spinlock_t		tx_global_lock;
1955 	int			watchdog_timeo;
1956 
1957 #ifdef CONFIG_XPS
1958 	struct xps_dev_maps __rcu *xps_cpus_map;
1959 	struct xps_dev_maps __rcu *xps_rxqs_map;
1960 #endif
1961 #ifdef CONFIG_NET_CLS_ACT
1962 	struct mini_Qdisc __rcu	*miniq_egress;
1963 #endif
1964 
1965 	/* These may be needed for future network-power-down code. */
1966 	struct timer_list	watchdog_timer;
1967 
1968 	int __percpu		*pcpu_refcnt;
1969 	struct list_head	todo_list;
1970 
1971 	struct list_head	link_watch_list;
1972 
1973 	enum { NETREG_UNINITIALIZED=0,
1974 	       NETREG_REGISTERED,	/* completed register_netdevice */
1975 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
1976 	       NETREG_UNREGISTERED,	/* completed unregister todo */
1977 	       NETREG_RELEASED,		/* called free_netdev */
1978 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
1979 	} reg_state:8;
1980 
1981 	bool dismantle;
1982 
1983 	enum {
1984 		RTNL_LINK_INITIALIZED,
1985 		RTNL_LINK_INITIALIZING,
1986 	} rtnl_link_state:16;
1987 
1988 	bool needs_free_netdev;
1989 	void (*priv_destructor)(struct net_device *dev);
1990 
1991 #ifdef CONFIG_NETPOLL
1992 	struct netpoll_info __rcu	*npinfo;
1993 #endif
1994 
1995 	possible_net_t			nd_net;
1996 
1997 	/* mid-layer private */
1998 	union {
1999 		void					*ml_priv;
2000 		struct pcpu_lstats __percpu		*lstats;
2001 		struct pcpu_sw_netstats __percpu	*tstats;
2002 		struct pcpu_dstats __percpu		*dstats;
2003 		struct pcpu_vstats __percpu		*vstats;
2004 	};
2005 
2006 #if IS_ENABLED(CONFIG_GARP)
2007 	struct garp_port __rcu	*garp_port;
2008 #endif
2009 #if IS_ENABLED(CONFIG_MRP)
2010 	struct mrp_port __rcu	*mrp_port;
2011 #endif
2012 
2013 	struct device		dev;
2014 	const struct attribute_group *sysfs_groups[4];
2015 	const struct attribute_group *sysfs_rx_queue_group;
2016 
2017 	const struct rtnl_link_ops *rtnl_link_ops;
2018 
2019 	/* for setting kernel sock attribute on TCP connection setup */
2020 #define GSO_MAX_SIZE		65536
2021 	unsigned int		gso_max_size;
2022 #define GSO_MAX_SEGS		65535
2023 	u16			gso_max_segs;
2024 
2025 #ifdef CONFIG_DCB
2026 	const struct dcbnl_rtnl_ops *dcbnl_ops;
2027 #endif
2028 	s16			num_tc;
2029 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
2030 	u8			prio_tc_map[TC_BITMASK + 1];
2031 
2032 #if IS_ENABLED(CONFIG_FCOE)
2033 	unsigned int		fcoe_ddp_xid;
2034 #endif
2035 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2036 	struct netprio_map __rcu *priomap;
2037 #endif
2038 	struct phy_device	*phydev;
2039 	struct sfp_bus		*sfp_bus;
2040 	struct lock_class_key	*qdisc_tx_busylock;
2041 	struct lock_class_key	*qdisc_running_key;
2042 	bool			proto_down;
2043 };
2044 #define to_net_dev(d) container_of(d, struct net_device, dev)
2045 
2046 static inline bool netif_elide_gro(const struct net_device *dev)
2047 {
2048 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2049 		return true;
2050 	return false;
2051 }
2052 
2053 #define	NETDEV_ALIGN		32
2054 
2055 static inline
2056 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2057 {
2058 	return dev->prio_tc_map[prio & TC_BITMASK];
2059 }
2060 
2061 static inline
2062 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2063 {
2064 	if (tc >= dev->num_tc)
2065 		return -EINVAL;
2066 
2067 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2068 	return 0;
2069 }
2070 
2071 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2072 void netdev_reset_tc(struct net_device *dev);
2073 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2074 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2075 
2076 static inline
2077 int netdev_get_num_tc(struct net_device *dev)
2078 {
2079 	return dev->num_tc;
2080 }
2081 
2082 void netdev_unbind_sb_channel(struct net_device *dev,
2083 			      struct net_device *sb_dev);
2084 int netdev_bind_sb_channel_queue(struct net_device *dev,
2085 				 struct net_device *sb_dev,
2086 				 u8 tc, u16 count, u16 offset);
2087 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2088 static inline int netdev_get_sb_channel(struct net_device *dev)
2089 {
2090 	return max_t(int, -dev->num_tc, 0);
2091 }
2092 
2093 static inline
2094 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2095 					 unsigned int index)
2096 {
2097 	return &dev->_tx[index];
2098 }
2099 
2100 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2101 						    const struct sk_buff *skb)
2102 {
2103 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2104 }
2105 
2106 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2107 					    void (*f)(struct net_device *,
2108 						      struct netdev_queue *,
2109 						      void *),
2110 					    void *arg)
2111 {
2112 	unsigned int i;
2113 
2114 	for (i = 0; i < dev->num_tx_queues; i++)
2115 		f(dev, &dev->_tx[i], arg);
2116 }
2117 
2118 #define netdev_lockdep_set_classes(dev)				\
2119 {								\
2120 	static struct lock_class_key qdisc_tx_busylock_key;	\
2121 	static struct lock_class_key qdisc_running_key;		\
2122 	static struct lock_class_key qdisc_xmit_lock_key;	\
2123 	static struct lock_class_key dev_addr_list_lock_key;	\
2124 	unsigned int i;						\
2125 								\
2126 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2127 	(dev)->qdisc_running_key = &qdisc_running_key;		\
2128 	lockdep_set_class(&(dev)->addr_list_lock,		\
2129 			  &dev_addr_list_lock_key); 		\
2130 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2131 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2132 				  &qdisc_xmit_lock_key);	\
2133 }
2134 
2135 struct netdev_queue *netdev_pick_tx(struct net_device *dev,
2136 				    struct sk_buff *skb,
2137 				    struct net_device *sb_dev);
2138 
2139 /* returns the headroom that the master device needs to take in account
2140  * when forwarding to this dev
2141  */
2142 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2143 {
2144 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2145 }
2146 
2147 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2148 {
2149 	if (dev->netdev_ops->ndo_set_rx_headroom)
2150 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2151 }
2152 
2153 /* set the device rx headroom to the dev's default */
2154 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2155 {
2156 	netdev_set_rx_headroom(dev, -1);
2157 }
2158 
2159 /*
2160  * Net namespace inlines
2161  */
2162 static inline
2163 struct net *dev_net(const struct net_device *dev)
2164 {
2165 	return read_pnet(&dev->nd_net);
2166 }
2167 
2168 static inline
2169 void dev_net_set(struct net_device *dev, struct net *net)
2170 {
2171 	write_pnet(&dev->nd_net, net);
2172 }
2173 
2174 /**
2175  *	netdev_priv - access network device private data
2176  *	@dev: network device
2177  *
2178  * Get network device private data
2179  */
2180 static inline void *netdev_priv(const struct net_device *dev)
2181 {
2182 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2183 }
2184 
2185 /* Set the sysfs physical device reference for the network logical device
2186  * if set prior to registration will cause a symlink during initialization.
2187  */
2188 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2189 
2190 /* Set the sysfs device type for the network logical device to allow
2191  * fine-grained identification of different network device types. For
2192  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2193  */
2194 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2195 
2196 /* Default NAPI poll() weight
2197  * Device drivers are strongly advised to not use bigger value
2198  */
2199 #define NAPI_POLL_WEIGHT 64
2200 
2201 /**
2202  *	netif_napi_add - initialize a NAPI context
2203  *	@dev:  network device
2204  *	@napi: NAPI context
2205  *	@poll: polling function
2206  *	@weight: default weight
2207  *
2208  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2209  * *any* of the other NAPI-related functions.
2210  */
2211 void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2212 		    int (*poll)(struct napi_struct *, int), int weight);
2213 
2214 /**
2215  *	netif_tx_napi_add - initialize a NAPI context
2216  *	@dev:  network device
2217  *	@napi: NAPI context
2218  *	@poll: polling function
2219  *	@weight: default weight
2220  *
2221  * This variant of netif_napi_add() should be used from drivers using NAPI
2222  * to exclusively poll a TX queue.
2223  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2224  */
2225 static inline void netif_tx_napi_add(struct net_device *dev,
2226 				     struct napi_struct *napi,
2227 				     int (*poll)(struct napi_struct *, int),
2228 				     int weight)
2229 {
2230 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2231 	netif_napi_add(dev, napi, poll, weight);
2232 }
2233 
2234 /**
2235  *  netif_napi_del - remove a NAPI context
2236  *  @napi: NAPI context
2237  *
2238  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2239  */
2240 void netif_napi_del(struct napi_struct *napi);
2241 
2242 struct napi_gro_cb {
2243 	/* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2244 	void	*frag0;
2245 
2246 	/* Length of frag0. */
2247 	unsigned int frag0_len;
2248 
2249 	/* This indicates where we are processing relative to skb->data. */
2250 	int	data_offset;
2251 
2252 	/* This is non-zero if the packet cannot be merged with the new skb. */
2253 	u16	flush;
2254 
2255 	/* Save the IP ID here and check when we get to the transport layer */
2256 	u16	flush_id;
2257 
2258 	/* Number of segments aggregated. */
2259 	u16	count;
2260 
2261 	/* Start offset for remote checksum offload */
2262 	u16	gro_remcsum_start;
2263 
2264 	/* jiffies when first packet was created/queued */
2265 	unsigned long age;
2266 
2267 	/* Used in ipv6_gro_receive() and foo-over-udp */
2268 	u16	proto;
2269 
2270 	/* This is non-zero if the packet may be of the same flow. */
2271 	u8	same_flow:1;
2272 
2273 	/* Used in tunnel GRO receive */
2274 	u8	encap_mark:1;
2275 
2276 	/* GRO checksum is valid */
2277 	u8	csum_valid:1;
2278 
2279 	/* Number of checksums via CHECKSUM_UNNECESSARY */
2280 	u8	csum_cnt:3;
2281 
2282 	/* Free the skb? */
2283 	u8	free:2;
2284 #define NAPI_GRO_FREE		  1
2285 #define NAPI_GRO_FREE_STOLEN_HEAD 2
2286 
2287 	/* Used in foo-over-udp, set in udp[46]_gro_receive */
2288 	u8	is_ipv6:1;
2289 
2290 	/* Used in GRE, set in fou/gue_gro_receive */
2291 	u8	is_fou:1;
2292 
2293 	/* Used to determine if flush_id can be ignored */
2294 	u8	is_atomic:1;
2295 
2296 	/* Number of gro_receive callbacks this packet already went through */
2297 	u8 recursion_counter:4;
2298 
2299 	/* 1 bit hole */
2300 
2301 	/* used to support CHECKSUM_COMPLETE for tunneling protocols */
2302 	__wsum	csum;
2303 
2304 	/* used in skb_gro_receive() slow path */
2305 	struct sk_buff *last;
2306 };
2307 
2308 #define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
2309 
2310 #define GRO_RECURSION_LIMIT 15
2311 static inline int gro_recursion_inc_test(struct sk_buff *skb)
2312 {
2313 	return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2314 }
2315 
2316 typedef struct sk_buff *(*gro_receive_t)(struct list_head *, struct sk_buff *);
2317 static inline struct sk_buff *call_gro_receive(gro_receive_t cb,
2318 					       struct list_head *head,
2319 					       struct sk_buff *skb)
2320 {
2321 	if (unlikely(gro_recursion_inc_test(skb))) {
2322 		NAPI_GRO_CB(skb)->flush |= 1;
2323 		return NULL;
2324 	}
2325 
2326 	return cb(head, skb);
2327 }
2328 
2329 typedef struct sk_buff *(*gro_receive_sk_t)(struct sock *, struct list_head *,
2330 					    struct sk_buff *);
2331 static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
2332 						  struct sock *sk,
2333 						  struct list_head *head,
2334 						  struct sk_buff *skb)
2335 {
2336 	if (unlikely(gro_recursion_inc_test(skb))) {
2337 		NAPI_GRO_CB(skb)->flush |= 1;
2338 		return NULL;
2339 	}
2340 
2341 	return cb(sk, head, skb);
2342 }
2343 
2344 struct packet_type {
2345 	__be16			type;	/* This is really htons(ether_type). */
2346 	bool			ignore_outgoing;
2347 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2348 	int			(*func) (struct sk_buff *,
2349 					 struct net_device *,
2350 					 struct packet_type *,
2351 					 struct net_device *);
2352 	void			(*list_func) (struct list_head *,
2353 					      struct packet_type *,
2354 					      struct net_device *);
2355 	bool			(*id_match)(struct packet_type *ptype,
2356 					    struct sock *sk);
2357 	void			*af_packet_priv;
2358 	struct list_head	list;
2359 };
2360 
2361 struct offload_callbacks {
2362 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2363 						netdev_features_t features);
2364 	struct sk_buff		*(*gro_receive)(struct list_head *head,
2365 						struct sk_buff *skb);
2366 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2367 };
2368 
2369 struct packet_offload {
2370 	__be16			 type;	/* This is really htons(ether_type). */
2371 	u16			 priority;
2372 	struct offload_callbacks callbacks;
2373 	struct list_head	 list;
2374 };
2375 
2376 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2377 struct pcpu_sw_netstats {
2378 	u64     rx_packets;
2379 	u64     rx_bytes;
2380 	u64     tx_packets;
2381 	u64     tx_bytes;
2382 	struct u64_stats_sync   syncp;
2383 };
2384 
2385 struct pcpu_lstats {
2386 	u64 packets;
2387 	u64 bytes;
2388 	struct u64_stats_sync syncp;
2389 };
2390 
2391 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2392 ({									\
2393 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2394 	if (pcpu_stats)	{						\
2395 		int __cpu;						\
2396 		for_each_possible_cpu(__cpu) {				\
2397 			typeof(type) *stat;				\
2398 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2399 			u64_stats_init(&stat->syncp);			\
2400 		}							\
2401 	}								\
2402 	pcpu_stats;							\
2403 })
2404 
2405 #define netdev_alloc_pcpu_stats(type)					\
2406 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2407 
2408 enum netdev_lag_tx_type {
2409 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2410 	NETDEV_LAG_TX_TYPE_RANDOM,
2411 	NETDEV_LAG_TX_TYPE_BROADCAST,
2412 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2413 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2414 	NETDEV_LAG_TX_TYPE_HASH,
2415 };
2416 
2417 enum netdev_lag_hash {
2418 	NETDEV_LAG_HASH_NONE,
2419 	NETDEV_LAG_HASH_L2,
2420 	NETDEV_LAG_HASH_L34,
2421 	NETDEV_LAG_HASH_L23,
2422 	NETDEV_LAG_HASH_E23,
2423 	NETDEV_LAG_HASH_E34,
2424 	NETDEV_LAG_HASH_UNKNOWN,
2425 };
2426 
2427 struct netdev_lag_upper_info {
2428 	enum netdev_lag_tx_type tx_type;
2429 	enum netdev_lag_hash hash_type;
2430 };
2431 
2432 struct netdev_lag_lower_state_info {
2433 	u8 link_up : 1,
2434 	   tx_enabled : 1;
2435 };
2436 
2437 #include <linux/notifier.h>
2438 
2439 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2440  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2441  * adding new types.
2442  */
2443 enum netdev_cmd {
2444 	NETDEV_UP	= 1,	/* For now you can't veto a device up/down */
2445 	NETDEV_DOWN,
2446 	NETDEV_REBOOT,		/* Tell a protocol stack a network interface
2447 				   detected a hardware crash and restarted
2448 				   - we can use this eg to kick tcp sessions
2449 				   once done */
2450 	NETDEV_CHANGE,		/* Notify device state change */
2451 	NETDEV_REGISTER,
2452 	NETDEV_UNREGISTER,
2453 	NETDEV_CHANGEMTU,	/* notify after mtu change happened */
2454 	NETDEV_CHANGEADDR,
2455 	NETDEV_GOING_DOWN,
2456 	NETDEV_CHANGENAME,
2457 	NETDEV_FEAT_CHANGE,
2458 	NETDEV_BONDING_FAILOVER,
2459 	NETDEV_PRE_UP,
2460 	NETDEV_PRE_TYPE_CHANGE,
2461 	NETDEV_POST_TYPE_CHANGE,
2462 	NETDEV_POST_INIT,
2463 	NETDEV_RELEASE,
2464 	NETDEV_NOTIFY_PEERS,
2465 	NETDEV_JOIN,
2466 	NETDEV_CHANGEUPPER,
2467 	NETDEV_RESEND_IGMP,
2468 	NETDEV_PRECHANGEMTU,	/* notify before mtu change happened */
2469 	NETDEV_CHANGEINFODATA,
2470 	NETDEV_BONDING_INFO,
2471 	NETDEV_PRECHANGEUPPER,
2472 	NETDEV_CHANGELOWERSTATE,
2473 	NETDEV_UDP_TUNNEL_PUSH_INFO,
2474 	NETDEV_UDP_TUNNEL_DROP_INFO,
2475 	NETDEV_CHANGE_TX_QUEUE_LEN,
2476 	NETDEV_CVLAN_FILTER_PUSH_INFO,
2477 	NETDEV_CVLAN_FILTER_DROP_INFO,
2478 	NETDEV_SVLAN_FILTER_PUSH_INFO,
2479 	NETDEV_SVLAN_FILTER_DROP_INFO,
2480 };
2481 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2482 
2483 int register_netdevice_notifier(struct notifier_block *nb);
2484 int unregister_netdevice_notifier(struct notifier_block *nb);
2485 
2486 struct netdev_notifier_info {
2487 	struct net_device	*dev;
2488 	struct netlink_ext_ack	*extack;
2489 };
2490 
2491 struct netdev_notifier_change_info {
2492 	struct netdev_notifier_info info; /* must be first */
2493 	unsigned int flags_changed;
2494 };
2495 
2496 struct netdev_notifier_changeupper_info {
2497 	struct netdev_notifier_info info; /* must be first */
2498 	struct net_device *upper_dev; /* new upper dev */
2499 	bool master; /* is upper dev master */
2500 	bool linking; /* is the notification for link or unlink */
2501 	void *upper_info; /* upper dev info */
2502 };
2503 
2504 struct netdev_notifier_changelowerstate_info {
2505 	struct netdev_notifier_info info; /* must be first */
2506 	void *lower_state_info; /* is lower dev state */
2507 };
2508 
2509 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2510 					     struct net_device *dev)
2511 {
2512 	info->dev = dev;
2513 	info->extack = NULL;
2514 }
2515 
2516 static inline struct net_device *
2517 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2518 {
2519 	return info->dev;
2520 }
2521 
2522 static inline struct netlink_ext_ack *
2523 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2524 {
2525 	return info->extack;
2526 }
2527 
2528 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2529 
2530 
2531 extern rwlock_t				dev_base_lock;		/* Device list lock */
2532 
2533 #define for_each_netdev(net, d)		\
2534 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
2535 #define for_each_netdev_reverse(net, d)	\
2536 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
2537 #define for_each_netdev_rcu(net, d)		\
2538 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
2539 #define for_each_netdev_safe(net, d, n)	\
2540 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2541 #define for_each_netdev_continue(net, d)		\
2542 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
2543 #define for_each_netdev_continue_rcu(net, d)		\
2544 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
2545 #define for_each_netdev_in_bond_rcu(bond, slave)	\
2546 		for_each_netdev_rcu(&init_net, slave)	\
2547 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
2548 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
2549 
2550 static inline struct net_device *next_net_device(struct net_device *dev)
2551 {
2552 	struct list_head *lh;
2553 	struct net *net;
2554 
2555 	net = dev_net(dev);
2556 	lh = dev->dev_list.next;
2557 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2558 }
2559 
2560 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2561 {
2562 	struct list_head *lh;
2563 	struct net *net;
2564 
2565 	net = dev_net(dev);
2566 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
2567 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2568 }
2569 
2570 static inline struct net_device *first_net_device(struct net *net)
2571 {
2572 	return list_empty(&net->dev_base_head) ? NULL :
2573 		net_device_entry(net->dev_base_head.next);
2574 }
2575 
2576 static inline struct net_device *first_net_device_rcu(struct net *net)
2577 {
2578 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2579 
2580 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2581 }
2582 
2583 int netdev_boot_setup_check(struct net_device *dev);
2584 unsigned long netdev_boot_base(const char *prefix, int unit);
2585 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2586 				       const char *hwaddr);
2587 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2588 struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2589 void dev_add_pack(struct packet_type *pt);
2590 void dev_remove_pack(struct packet_type *pt);
2591 void __dev_remove_pack(struct packet_type *pt);
2592 void dev_add_offload(struct packet_offload *po);
2593 void dev_remove_offload(struct packet_offload *po);
2594 
2595 int dev_get_iflink(const struct net_device *dev);
2596 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
2597 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2598 				      unsigned short mask);
2599 struct net_device *dev_get_by_name(struct net *net, const char *name);
2600 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2601 struct net_device *__dev_get_by_name(struct net *net, const char *name);
2602 int dev_alloc_name(struct net_device *dev, const char *name);
2603 int dev_open(struct net_device *dev);
2604 void dev_close(struct net_device *dev);
2605 void dev_close_many(struct list_head *head, bool unlink);
2606 void dev_disable_lro(struct net_device *dev);
2607 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
2608 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
2609 		     struct net_device *sb_dev,
2610 		     select_queue_fallback_t fallback);
2611 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
2612 		       struct net_device *sb_dev,
2613 		       select_queue_fallback_t fallback);
2614 int dev_queue_xmit(struct sk_buff *skb);
2615 int dev_queue_xmit_accel(struct sk_buff *skb, struct net_device *sb_dev);
2616 int dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
2617 int register_netdevice(struct net_device *dev);
2618 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2619 void unregister_netdevice_many(struct list_head *head);
2620 static inline void unregister_netdevice(struct net_device *dev)
2621 {
2622 	unregister_netdevice_queue(dev, NULL);
2623 }
2624 
2625 int netdev_refcnt_read(const struct net_device *dev);
2626 void free_netdev(struct net_device *dev);
2627 void netdev_freemem(struct net_device *dev);
2628 void synchronize_net(void);
2629 int init_dummy_netdev(struct net_device *dev);
2630 
2631 DECLARE_PER_CPU(int, xmit_recursion);
2632 #define XMIT_RECURSION_LIMIT	10
2633 
2634 static inline int dev_recursion_level(void)
2635 {
2636 	return this_cpu_read(xmit_recursion);
2637 }
2638 
2639 struct net_device *dev_get_by_index(struct net *net, int ifindex);
2640 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2641 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2642 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
2643 int netdev_get_name(struct net *net, char *name, int ifindex);
2644 int dev_restart(struct net_device *dev);
2645 int skb_gro_receive(struct sk_buff *p, struct sk_buff *skb);
2646 
2647 static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2648 {
2649 	return NAPI_GRO_CB(skb)->data_offset;
2650 }
2651 
2652 static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2653 {
2654 	return skb->len - NAPI_GRO_CB(skb)->data_offset;
2655 }
2656 
2657 static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2658 {
2659 	NAPI_GRO_CB(skb)->data_offset += len;
2660 }
2661 
2662 static inline void *skb_gro_header_fast(struct sk_buff *skb,
2663 					unsigned int offset)
2664 {
2665 	return NAPI_GRO_CB(skb)->frag0 + offset;
2666 }
2667 
2668 static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2669 {
2670 	return NAPI_GRO_CB(skb)->frag0_len < hlen;
2671 }
2672 
2673 static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2674 {
2675 	NAPI_GRO_CB(skb)->frag0 = NULL;
2676 	NAPI_GRO_CB(skb)->frag0_len = 0;
2677 }
2678 
2679 static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2680 					unsigned int offset)
2681 {
2682 	if (!pskb_may_pull(skb, hlen))
2683 		return NULL;
2684 
2685 	skb_gro_frag0_invalidate(skb);
2686 	return skb->data + offset;
2687 }
2688 
2689 static inline void *skb_gro_network_header(struct sk_buff *skb)
2690 {
2691 	return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2692 	       skb_network_offset(skb);
2693 }
2694 
2695 static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2696 					const void *start, unsigned int len)
2697 {
2698 	if (NAPI_GRO_CB(skb)->csum_valid)
2699 		NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2700 						  csum_partial(start, len, 0));
2701 }
2702 
2703 /* GRO checksum functions. These are logical equivalents of the normal
2704  * checksum functions (in skbuff.h) except that they operate on the GRO
2705  * offsets and fields in sk_buff.
2706  */
2707 
2708 __sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2709 
2710 static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2711 {
2712 	return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
2713 }
2714 
2715 static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2716 						      bool zero_okay,
2717 						      __sum16 check)
2718 {
2719 	return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2720 		skb_checksum_start_offset(skb) <
2721 		 skb_gro_offset(skb)) &&
2722 		!skb_at_gro_remcsum_start(skb) &&
2723 		NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2724 		(!zero_okay || check));
2725 }
2726 
2727 static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2728 							   __wsum psum)
2729 {
2730 	if (NAPI_GRO_CB(skb)->csum_valid &&
2731 	    !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2732 		return 0;
2733 
2734 	NAPI_GRO_CB(skb)->csum = psum;
2735 
2736 	return __skb_gro_checksum_complete(skb);
2737 }
2738 
2739 static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2740 {
2741 	if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2742 		/* Consume a checksum from CHECKSUM_UNNECESSARY */
2743 		NAPI_GRO_CB(skb)->csum_cnt--;
2744 	} else {
2745 		/* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2746 		 * verified a new top level checksum or an encapsulated one
2747 		 * during GRO. This saves work if we fallback to normal path.
2748 		 */
2749 		__skb_incr_checksum_unnecessary(skb);
2750 	}
2751 }
2752 
2753 #define __skb_gro_checksum_validate(skb, proto, zero_okay, check,	\
2754 				    compute_pseudo)			\
2755 ({									\
2756 	__sum16 __ret = 0;						\
2757 	if (__skb_gro_checksum_validate_needed(skb, zero_okay, check))	\
2758 		__ret = __skb_gro_checksum_validate_complete(skb,	\
2759 				compute_pseudo(skb, proto));		\
2760 	if (!__ret)							\
2761 		skb_gro_incr_csum_unnecessary(skb);			\
2762 	__ret;								\
2763 })
2764 
2765 #define skb_gro_checksum_validate(skb, proto, compute_pseudo)		\
2766 	__skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2767 
2768 #define skb_gro_checksum_validate_zero_check(skb, proto, check,		\
2769 					     compute_pseudo)		\
2770 	__skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2771 
2772 #define skb_gro_checksum_simple_validate(skb)				\
2773 	__skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2774 
2775 static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2776 {
2777 	return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2778 		!NAPI_GRO_CB(skb)->csum_valid);
2779 }
2780 
2781 static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2782 					      __sum16 check, __wsum pseudo)
2783 {
2784 	NAPI_GRO_CB(skb)->csum = ~pseudo;
2785 	NAPI_GRO_CB(skb)->csum_valid = 1;
2786 }
2787 
2788 #define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo)	\
2789 do {									\
2790 	if (__skb_gro_checksum_convert_check(skb))			\
2791 		__skb_gro_checksum_convert(skb, check,			\
2792 					   compute_pseudo(skb, proto));	\
2793 } while (0)
2794 
2795 struct gro_remcsum {
2796 	int offset;
2797 	__wsum delta;
2798 };
2799 
2800 static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2801 {
2802 	grc->offset = 0;
2803 	grc->delta = 0;
2804 }
2805 
2806 static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2807 					    unsigned int off, size_t hdrlen,
2808 					    int start, int offset,
2809 					    struct gro_remcsum *grc,
2810 					    bool nopartial)
2811 {
2812 	__wsum delta;
2813 	size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
2814 
2815 	BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2816 
2817 	if (!nopartial) {
2818 		NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2819 		return ptr;
2820 	}
2821 
2822 	ptr = skb_gro_header_fast(skb, off);
2823 	if (skb_gro_header_hard(skb, off + plen)) {
2824 		ptr = skb_gro_header_slow(skb, off + plen, off);
2825 		if (!ptr)
2826 			return NULL;
2827 	}
2828 
2829 	delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2830 			       start, offset);
2831 
2832 	/* Adjust skb->csum since we changed the packet */
2833 	NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
2834 
2835 	grc->offset = off + hdrlen + offset;
2836 	grc->delta = delta;
2837 
2838 	return ptr;
2839 }
2840 
2841 static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2842 					   struct gro_remcsum *grc)
2843 {
2844 	void *ptr;
2845 	size_t plen = grc->offset + sizeof(u16);
2846 
2847 	if (!grc->delta)
2848 		return;
2849 
2850 	ptr = skb_gro_header_fast(skb, grc->offset);
2851 	if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2852 		ptr = skb_gro_header_slow(skb, plen, grc->offset);
2853 		if (!ptr)
2854 			return;
2855 	}
2856 
2857 	remcsum_unadjust((__sum16 *)ptr, grc->delta);
2858 }
2859 
2860 #ifdef CONFIG_XFRM_OFFLOAD
2861 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2862 {
2863 	if (PTR_ERR(pp) != -EINPROGRESS)
2864 		NAPI_GRO_CB(skb)->flush |= flush;
2865 }
2866 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2867 					       struct sk_buff *pp,
2868 					       int flush,
2869 					       struct gro_remcsum *grc)
2870 {
2871 	if (PTR_ERR(pp) != -EINPROGRESS) {
2872 		NAPI_GRO_CB(skb)->flush |= flush;
2873 		skb_gro_remcsum_cleanup(skb, grc);
2874 		skb->remcsum_offload = 0;
2875 	}
2876 }
2877 #else
2878 static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff *pp, int flush)
2879 {
2880 	NAPI_GRO_CB(skb)->flush |= flush;
2881 }
2882 static inline void skb_gro_flush_final_remcsum(struct sk_buff *skb,
2883 					       struct sk_buff *pp,
2884 					       int flush,
2885 					       struct gro_remcsum *grc)
2886 {
2887 	NAPI_GRO_CB(skb)->flush |= flush;
2888 	skb_gro_remcsum_cleanup(skb, grc);
2889 	skb->remcsum_offload = 0;
2890 }
2891 #endif
2892 
2893 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2894 				  unsigned short type,
2895 				  const void *daddr, const void *saddr,
2896 				  unsigned int len)
2897 {
2898 	if (!dev->header_ops || !dev->header_ops->create)
2899 		return 0;
2900 
2901 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
2902 }
2903 
2904 static inline int dev_parse_header(const struct sk_buff *skb,
2905 				   unsigned char *haddr)
2906 {
2907 	const struct net_device *dev = skb->dev;
2908 
2909 	if (!dev->header_ops || !dev->header_ops->parse)
2910 		return 0;
2911 	return dev->header_ops->parse(skb, haddr);
2912 }
2913 
2914 /* ll_header must have at least hard_header_len allocated */
2915 static inline bool dev_validate_header(const struct net_device *dev,
2916 				       char *ll_header, int len)
2917 {
2918 	if (likely(len >= dev->hard_header_len))
2919 		return true;
2920 	if (len < dev->min_header_len)
2921 		return false;
2922 
2923 	if (capable(CAP_SYS_RAWIO)) {
2924 		memset(ll_header + len, 0, dev->hard_header_len - len);
2925 		return true;
2926 	}
2927 
2928 	if (dev->header_ops && dev->header_ops->validate)
2929 		return dev->header_ops->validate(ll_header, len);
2930 
2931 	return false;
2932 }
2933 
2934 typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr,
2935 			   int len, int size);
2936 int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
2937 static inline int unregister_gifconf(unsigned int family)
2938 {
2939 	return register_gifconf(family, NULL);
2940 }
2941 
2942 #ifdef CONFIG_NET_FLOW_LIMIT
2943 #define FLOW_LIMIT_HISTORY	(1 << 7)  /* must be ^2 and !overflow buckets */
2944 struct sd_flow_limit {
2945 	u64			count;
2946 	unsigned int		num_buckets;
2947 	unsigned int		history_head;
2948 	u16			history[FLOW_LIMIT_HISTORY];
2949 	u8			buckets[];
2950 };
2951 
2952 extern int netdev_flow_limit_table_len;
2953 #endif /* CONFIG_NET_FLOW_LIMIT */
2954 
2955 /*
2956  * Incoming packets are placed on per-CPU queues
2957  */
2958 struct softnet_data {
2959 	struct list_head	poll_list;
2960 	struct sk_buff_head	process_queue;
2961 
2962 	/* stats */
2963 	unsigned int		processed;
2964 	unsigned int		time_squeeze;
2965 	unsigned int		received_rps;
2966 #ifdef CONFIG_RPS
2967 	struct softnet_data	*rps_ipi_list;
2968 #endif
2969 #ifdef CONFIG_NET_FLOW_LIMIT
2970 	struct sd_flow_limit __rcu *flow_limit;
2971 #endif
2972 	struct Qdisc		*output_queue;
2973 	struct Qdisc		**output_queue_tailp;
2974 	struct sk_buff		*completion_queue;
2975 #ifdef CONFIG_XFRM_OFFLOAD
2976 	struct sk_buff_head	xfrm_backlog;
2977 #endif
2978 #ifdef CONFIG_RPS
2979 	/* input_queue_head should be written by cpu owning this struct,
2980 	 * and only read by other cpus. Worth using a cache line.
2981 	 */
2982 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
2983 
2984 	/* Elements below can be accessed between CPUs for RPS/RFS */
2985 	call_single_data_t	csd ____cacheline_aligned_in_smp;
2986 	struct softnet_data	*rps_ipi_next;
2987 	unsigned int		cpu;
2988 	unsigned int		input_queue_tail;
2989 #endif
2990 	unsigned int		dropped;
2991 	struct sk_buff_head	input_pkt_queue;
2992 	struct napi_struct	backlog;
2993 
2994 };
2995 
2996 static inline void input_queue_head_incr(struct softnet_data *sd)
2997 {
2998 #ifdef CONFIG_RPS
2999 	sd->input_queue_head++;
3000 #endif
3001 }
3002 
3003 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3004 					      unsigned int *qtail)
3005 {
3006 #ifdef CONFIG_RPS
3007 	*qtail = ++sd->input_queue_tail;
3008 #endif
3009 }
3010 
3011 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3012 
3013 void __netif_schedule(struct Qdisc *q);
3014 void netif_schedule_queue(struct netdev_queue *txq);
3015 
3016 static inline void netif_tx_schedule_all(struct net_device *dev)
3017 {
3018 	unsigned int i;
3019 
3020 	for (i = 0; i < dev->num_tx_queues; i++)
3021 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
3022 }
3023 
3024 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3025 {
3026 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3027 }
3028 
3029 /**
3030  *	netif_start_queue - allow transmit
3031  *	@dev: network device
3032  *
3033  *	Allow upper layers to call the device hard_start_xmit routine.
3034  */
3035 static inline void netif_start_queue(struct net_device *dev)
3036 {
3037 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3038 }
3039 
3040 static inline void netif_tx_start_all_queues(struct net_device *dev)
3041 {
3042 	unsigned int i;
3043 
3044 	for (i = 0; i < dev->num_tx_queues; i++) {
3045 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3046 		netif_tx_start_queue(txq);
3047 	}
3048 }
3049 
3050 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3051 
3052 /**
3053  *	netif_wake_queue - restart transmit
3054  *	@dev: network device
3055  *
3056  *	Allow upper layers to call the device hard_start_xmit routine.
3057  *	Used for flow control when transmit resources are available.
3058  */
3059 static inline void netif_wake_queue(struct net_device *dev)
3060 {
3061 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3062 }
3063 
3064 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3065 {
3066 	unsigned int i;
3067 
3068 	for (i = 0; i < dev->num_tx_queues; i++) {
3069 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3070 		netif_tx_wake_queue(txq);
3071 	}
3072 }
3073 
3074 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3075 {
3076 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3077 }
3078 
3079 /**
3080  *	netif_stop_queue - stop transmitted packets
3081  *	@dev: network device
3082  *
3083  *	Stop upper layers calling the device hard_start_xmit routine.
3084  *	Used for flow control when transmit resources are unavailable.
3085  */
3086 static inline void netif_stop_queue(struct net_device *dev)
3087 {
3088 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3089 }
3090 
3091 void netif_tx_stop_all_queues(struct net_device *dev);
3092 
3093 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3094 {
3095 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3096 }
3097 
3098 /**
3099  *	netif_queue_stopped - test if transmit queue is flowblocked
3100  *	@dev: network device
3101  *
3102  *	Test if transmit queue on device is currently unable to send.
3103  */
3104 static inline bool netif_queue_stopped(const struct net_device *dev)
3105 {
3106 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3107 }
3108 
3109 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3110 {
3111 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3112 }
3113 
3114 static inline bool
3115 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3116 {
3117 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3118 }
3119 
3120 static inline bool
3121 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3122 {
3123 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3124 }
3125 
3126 /**
3127  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3128  *	@dev_queue: pointer to transmit queue
3129  *
3130  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3131  * to give appropriate hint to the CPU.
3132  */
3133 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3134 {
3135 #ifdef CONFIG_BQL
3136 	prefetchw(&dev_queue->dql.num_queued);
3137 #endif
3138 }
3139 
3140 /**
3141  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3142  *	@dev_queue: pointer to transmit queue
3143  *
3144  * BQL enabled drivers might use this helper in their TX completion path,
3145  * to give appropriate hint to the CPU.
3146  */
3147 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3148 {
3149 #ifdef CONFIG_BQL
3150 	prefetchw(&dev_queue->dql.limit);
3151 #endif
3152 }
3153 
3154 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3155 					unsigned int bytes)
3156 {
3157 #ifdef CONFIG_BQL
3158 	dql_queued(&dev_queue->dql, bytes);
3159 
3160 	if (likely(dql_avail(&dev_queue->dql) >= 0))
3161 		return;
3162 
3163 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3164 
3165 	/*
3166 	 * The XOFF flag must be set before checking the dql_avail below,
3167 	 * because in netdev_tx_completed_queue we update the dql_completed
3168 	 * before checking the XOFF flag.
3169 	 */
3170 	smp_mb();
3171 
3172 	/* check again in case another CPU has just made room avail */
3173 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3174 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3175 #endif
3176 }
3177 
3178 /**
3179  * 	netdev_sent_queue - report the number of bytes queued to hardware
3180  * 	@dev: network device
3181  * 	@bytes: number of bytes queued to the hardware device queue
3182  *
3183  * 	Report the number of bytes queued for sending/completion to the network
3184  * 	device hardware queue. @bytes should be a good approximation and should
3185  * 	exactly match netdev_completed_queue() @bytes
3186  */
3187 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3188 {
3189 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3190 }
3191 
3192 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3193 					     unsigned int pkts, unsigned int bytes)
3194 {
3195 #ifdef CONFIG_BQL
3196 	if (unlikely(!bytes))
3197 		return;
3198 
3199 	dql_completed(&dev_queue->dql, bytes);
3200 
3201 	/*
3202 	 * Without the memory barrier there is a small possiblity that
3203 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3204 	 * be stopped forever
3205 	 */
3206 	smp_mb();
3207 
3208 	if (dql_avail(&dev_queue->dql) < 0)
3209 		return;
3210 
3211 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3212 		netif_schedule_queue(dev_queue);
3213 #endif
3214 }
3215 
3216 /**
3217  * 	netdev_completed_queue - report bytes and packets completed by device
3218  * 	@dev: network device
3219  * 	@pkts: actual number of packets sent over the medium
3220  * 	@bytes: actual number of bytes sent over the medium
3221  *
3222  * 	Report the number of bytes and packets transmitted by the network device
3223  * 	hardware queue over the physical medium, @bytes must exactly match the
3224  * 	@bytes amount passed to netdev_sent_queue()
3225  */
3226 static inline void netdev_completed_queue(struct net_device *dev,
3227 					  unsigned int pkts, unsigned int bytes)
3228 {
3229 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3230 }
3231 
3232 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3233 {
3234 #ifdef CONFIG_BQL
3235 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3236 	dql_reset(&q->dql);
3237 #endif
3238 }
3239 
3240 /**
3241  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3242  * 	@dev_queue: network device
3243  *
3244  * 	Reset the bytes and packet count of a network device and clear the
3245  * 	software flow control OFF bit for this network device
3246  */
3247 static inline void netdev_reset_queue(struct net_device *dev_queue)
3248 {
3249 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3250 }
3251 
3252 /**
3253  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3254  * 	@dev: network device
3255  * 	@queue_index: given tx queue index
3256  *
3257  * 	Returns 0 if given tx queue index >= number of device tx queues,
3258  * 	otherwise returns the originally passed tx queue index.
3259  */
3260 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3261 {
3262 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3263 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3264 				     dev->name, queue_index,
3265 				     dev->real_num_tx_queues);
3266 		return 0;
3267 	}
3268 
3269 	return queue_index;
3270 }
3271 
3272 /**
3273  *	netif_running - test if up
3274  *	@dev: network device
3275  *
3276  *	Test if the device has been brought up.
3277  */
3278 static inline bool netif_running(const struct net_device *dev)
3279 {
3280 	return test_bit(__LINK_STATE_START, &dev->state);
3281 }
3282 
3283 /*
3284  * Routines to manage the subqueues on a device.  We only need start,
3285  * stop, and a check if it's stopped.  All other device management is
3286  * done at the overall netdevice level.
3287  * Also test the device if we're multiqueue.
3288  */
3289 
3290 /**
3291  *	netif_start_subqueue - allow sending packets on subqueue
3292  *	@dev: network device
3293  *	@queue_index: sub queue index
3294  *
3295  * Start individual transmit queue of a device with multiple transmit queues.
3296  */
3297 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3298 {
3299 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3300 
3301 	netif_tx_start_queue(txq);
3302 }
3303 
3304 /**
3305  *	netif_stop_subqueue - stop sending packets on subqueue
3306  *	@dev: network device
3307  *	@queue_index: sub queue index
3308  *
3309  * Stop individual transmit queue of a device with multiple transmit queues.
3310  */
3311 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3312 {
3313 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3314 	netif_tx_stop_queue(txq);
3315 }
3316 
3317 /**
3318  *	netif_subqueue_stopped - test status of subqueue
3319  *	@dev: network device
3320  *	@queue_index: sub queue index
3321  *
3322  * Check individual transmit queue of a device with multiple transmit queues.
3323  */
3324 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3325 					    u16 queue_index)
3326 {
3327 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3328 
3329 	return netif_tx_queue_stopped(txq);
3330 }
3331 
3332 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3333 					  struct sk_buff *skb)
3334 {
3335 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3336 }
3337 
3338 /**
3339  *	netif_wake_subqueue - allow sending packets on subqueue
3340  *	@dev: network device
3341  *	@queue_index: sub queue index
3342  *
3343  * Resume individual transmit queue of a device with multiple transmit queues.
3344  */
3345 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3346 {
3347 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3348 
3349 	netif_tx_wake_queue(txq);
3350 }
3351 
3352 #ifdef CONFIG_XPS
3353 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3354 			u16 index);
3355 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3356 			  u16 index, bool is_rxqs_map);
3357 
3358 /**
3359  *	netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3360  *	@j: CPU/Rx queue index
3361  *	@mask: bitmask of all cpus/rx queues
3362  *	@nr_bits: number of bits in the bitmask
3363  *
3364  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3365  */
3366 static inline bool netif_attr_test_mask(unsigned long j,
3367 					const unsigned long *mask,
3368 					unsigned int nr_bits)
3369 {
3370 	cpu_max_bits_warn(j, nr_bits);
3371 	return test_bit(j, mask);
3372 }
3373 
3374 /**
3375  *	netif_attr_test_online - Test for online CPU/Rx queue
3376  *	@j: CPU/Rx queue index
3377  *	@online_mask: bitmask for CPUs/Rx queues that are online
3378  *	@nr_bits: number of bits in the bitmask
3379  *
3380  * Returns true if a CPU/Rx queue is online.
3381  */
3382 static inline bool netif_attr_test_online(unsigned long j,
3383 					  const unsigned long *online_mask,
3384 					  unsigned int nr_bits)
3385 {
3386 	cpu_max_bits_warn(j, nr_bits);
3387 
3388 	if (online_mask)
3389 		return test_bit(j, online_mask);
3390 
3391 	return (j < nr_bits);
3392 }
3393 
3394 /**
3395  *	netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3396  *	@n: CPU/Rx queue index
3397  *	@srcp: the cpumask/Rx queue mask pointer
3398  *	@nr_bits: number of bits in the bitmask
3399  *
3400  * Returns >= nr_bits if no further CPUs/Rx queues set.
3401  */
3402 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3403 					       unsigned int nr_bits)
3404 {
3405 	/* -1 is a legal arg here. */
3406 	if (n != -1)
3407 		cpu_max_bits_warn(n, nr_bits);
3408 
3409 	if (srcp)
3410 		return find_next_bit(srcp, nr_bits, n + 1);
3411 
3412 	return n + 1;
3413 }
3414 
3415 /**
3416  *	netif_attrmask_next_and - get the next CPU/Rx queue in *src1p & *src2p
3417  *	@n: CPU/Rx queue index
3418  *	@src1p: the first CPUs/Rx queues mask pointer
3419  *	@src2p: the second CPUs/Rx queues mask pointer
3420  *	@nr_bits: number of bits in the bitmask
3421  *
3422  * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3423  */
3424 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3425 					  const unsigned long *src2p,
3426 					  unsigned int nr_bits)
3427 {
3428 	/* -1 is a legal arg here. */
3429 	if (n != -1)
3430 		cpu_max_bits_warn(n, nr_bits);
3431 
3432 	if (src1p && src2p)
3433 		return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3434 	else if (src1p)
3435 		return find_next_bit(src1p, nr_bits, n + 1);
3436 	else if (src2p)
3437 		return find_next_bit(src2p, nr_bits, n + 1);
3438 
3439 	return n + 1;
3440 }
3441 #else
3442 static inline int netif_set_xps_queue(struct net_device *dev,
3443 				      const struct cpumask *mask,
3444 				      u16 index)
3445 {
3446 	return 0;
3447 }
3448 
3449 static inline int __netif_set_xps_queue(struct net_device *dev,
3450 					const unsigned long *mask,
3451 					u16 index, bool is_rxqs_map)
3452 {
3453 	return 0;
3454 }
3455 #endif
3456 
3457 /**
3458  *	netif_is_multiqueue - test if device has multiple transmit queues
3459  *	@dev: network device
3460  *
3461  * Check if device has multiple transmit queues
3462  */
3463 static inline bool netif_is_multiqueue(const struct net_device *dev)
3464 {
3465 	return dev->num_tx_queues > 1;
3466 }
3467 
3468 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3469 
3470 #ifdef CONFIG_SYSFS
3471 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3472 #else
3473 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3474 						unsigned int rxqs)
3475 {
3476 	dev->real_num_rx_queues = rxqs;
3477 	return 0;
3478 }
3479 #endif
3480 
3481 static inline struct netdev_rx_queue *
3482 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
3483 {
3484 	return dev->_rx + rxq;
3485 }
3486 
3487 #ifdef CONFIG_SYSFS
3488 static inline unsigned int get_netdev_rx_queue_index(
3489 		struct netdev_rx_queue *queue)
3490 {
3491 	struct net_device *dev = queue->dev;
3492 	int index = queue - dev->_rx;
3493 
3494 	BUG_ON(index >= dev->num_rx_queues);
3495 	return index;
3496 }
3497 #endif
3498 
3499 #define DEFAULT_MAX_NUM_RSS_QUEUES	(8)
3500 int netif_get_num_default_rss_queues(void);
3501 
3502 enum skb_free_reason {
3503 	SKB_REASON_CONSUMED,
3504 	SKB_REASON_DROPPED,
3505 };
3506 
3507 void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3508 void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3509 
3510 /*
3511  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3512  * interrupt context or with hardware interrupts being disabled.
3513  * (in_irq() || irqs_disabled())
3514  *
3515  * We provide four helpers that can be used in following contexts :
3516  *
3517  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3518  *  replacing kfree_skb(skb)
3519  *
3520  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3521  *  Typically used in place of consume_skb(skb) in TX completion path
3522  *
3523  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3524  *  replacing kfree_skb(skb)
3525  *
3526  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3527  *  and consumed a packet. Used in place of consume_skb(skb)
3528  */
3529 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3530 {
3531 	__dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3532 }
3533 
3534 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3535 {
3536 	__dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3537 }
3538 
3539 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3540 {
3541 	__dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3542 }
3543 
3544 static inline void dev_consume_skb_any(struct sk_buff *skb)
3545 {
3546 	__dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3547 }
3548 
3549 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3550 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3551 int netif_rx(struct sk_buff *skb);
3552 int netif_rx_ni(struct sk_buff *skb);
3553 int netif_receive_skb(struct sk_buff *skb);
3554 int netif_receive_skb_core(struct sk_buff *skb);
3555 void netif_receive_skb_list(struct list_head *head);
3556 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3557 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3558 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3559 gro_result_t napi_gro_frags(struct napi_struct *napi);
3560 struct packet_offload *gro_find_receive_by_type(__be16 type);
3561 struct packet_offload *gro_find_complete_by_type(__be16 type);
3562 
3563 static inline void napi_free_frags(struct napi_struct *napi)
3564 {
3565 	kfree_skb(napi->skb);
3566 	napi->skb = NULL;
3567 }
3568 
3569 bool netdev_is_rx_handler_busy(struct net_device *dev);
3570 int netdev_rx_handler_register(struct net_device *dev,
3571 			       rx_handler_func_t *rx_handler,
3572 			       void *rx_handler_data);
3573 void netdev_rx_handler_unregister(struct net_device *dev);
3574 
3575 bool dev_valid_name(const char *name);
3576 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3577 		bool *need_copyout);
3578 int dev_ifconf(struct net *net, struct ifconf *, int);
3579 int dev_ethtool(struct net *net, struct ifreq *);
3580 unsigned int dev_get_flags(const struct net_device *);
3581 int __dev_change_flags(struct net_device *, unsigned int flags);
3582 int dev_change_flags(struct net_device *, unsigned int);
3583 void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3584 			unsigned int gchanges);
3585 int dev_change_name(struct net_device *, const char *);
3586 int dev_set_alias(struct net_device *, const char *, size_t);
3587 int dev_get_alias(const struct net_device *, char *, size_t);
3588 int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3589 int __dev_set_mtu(struct net_device *, int);
3590 int dev_set_mtu_ext(struct net_device *dev, int mtu,
3591 		    struct netlink_ext_ack *extack);
3592 int dev_set_mtu(struct net_device *, int);
3593 int dev_change_tx_queue_len(struct net_device *, unsigned long);
3594 void dev_set_group(struct net_device *, int);
3595 int dev_set_mac_address(struct net_device *, struct sockaddr *);
3596 int dev_change_carrier(struct net_device *, bool new_carrier);
3597 int dev_get_phys_port_id(struct net_device *dev,
3598 			 struct netdev_phys_item_id *ppid);
3599 int dev_get_phys_port_name(struct net_device *dev,
3600 			   char *name, size_t len);
3601 int dev_change_proto_down(struct net_device *dev, bool proto_down);
3602 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3603 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3604 				    struct netdev_queue *txq, int *ret);
3605 
3606 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf);
3607 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3608 		      int fd, u32 flags);
3609 u32 __dev_xdp_query(struct net_device *dev, bpf_op_t xdp_op,
3610 		    enum bpf_netdev_command cmd);
3611 int xdp_umem_query(struct net_device *dev, u16 queue_id);
3612 
3613 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3614 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3615 bool is_skb_forwardable(const struct net_device *dev,
3616 			const struct sk_buff *skb);
3617 
3618 static __always_inline int ____dev_forward_skb(struct net_device *dev,
3619 					       struct sk_buff *skb)
3620 {
3621 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3622 	    unlikely(!is_skb_forwardable(dev, skb))) {
3623 		atomic_long_inc(&dev->rx_dropped);
3624 		kfree_skb(skb);
3625 		return NET_RX_DROP;
3626 	}
3627 
3628 	skb_scrub_packet(skb, true);
3629 	skb->priority = 0;
3630 	return 0;
3631 }
3632 
3633 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3634 
3635 extern int		netdev_budget;
3636 extern unsigned int	netdev_budget_usecs;
3637 
3638 /* Called by rtnetlink.c:rtnl_unlock() */
3639 void netdev_run_todo(void);
3640 
3641 /**
3642  *	dev_put - release reference to device
3643  *	@dev: network device
3644  *
3645  * Release reference to device to allow it to be freed.
3646  */
3647 static inline void dev_put(struct net_device *dev)
3648 {
3649 	this_cpu_dec(*dev->pcpu_refcnt);
3650 }
3651 
3652 /**
3653  *	dev_hold - get reference to device
3654  *	@dev: network device
3655  *
3656  * Hold reference to device to keep it from being freed.
3657  */
3658 static inline void dev_hold(struct net_device *dev)
3659 {
3660 	this_cpu_inc(*dev->pcpu_refcnt);
3661 }
3662 
3663 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
3664  * and _off may be called from IRQ context, but it is caller
3665  * who is responsible for serialization of these calls.
3666  *
3667  * The name carrier is inappropriate, these functions should really be
3668  * called netif_lowerlayer_*() because they represent the state of any
3669  * kind of lower layer not just hardware media.
3670  */
3671 
3672 void linkwatch_init_dev(struct net_device *dev);
3673 void linkwatch_fire_event(struct net_device *dev);
3674 void linkwatch_forget_dev(struct net_device *dev);
3675 
3676 /**
3677  *	netif_carrier_ok - test if carrier present
3678  *	@dev: network device
3679  *
3680  * Check if carrier is present on device
3681  */
3682 static inline bool netif_carrier_ok(const struct net_device *dev)
3683 {
3684 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3685 }
3686 
3687 unsigned long dev_trans_start(struct net_device *dev);
3688 
3689 void __netdev_watchdog_up(struct net_device *dev);
3690 
3691 void netif_carrier_on(struct net_device *dev);
3692 
3693 void netif_carrier_off(struct net_device *dev);
3694 
3695 /**
3696  *	netif_dormant_on - mark device as dormant.
3697  *	@dev: network device
3698  *
3699  * Mark device as dormant (as per RFC2863).
3700  *
3701  * The dormant state indicates that the relevant interface is not
3702  * actually in a condition to pass packets (i.e., it is not 'up') but is
3703  * in a "pending" state, waiting for some external event.  For "on-
3704  * demand" interfaces, this new state identifies the situation where the
3705  * interface is waiting for events to place it in the up state.
3706  */
3707 static inline void netif_dormant_on(struct net_device *dev)
3708 {
3709 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3710 		linkwatch_fire_event(dev);
3711 }
3712 
3713 /**
3714  *	netif_dormant_off - set device as not dormant.
3715  *	@dev: network device
3716  *
3717  * Device is not in dormant state.
3718  */
3719 static inline void netif_dormant_off(struct net_device *dev)
3720 {
3721 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3722 		linkwatch_fire_event(dev);
3723 }
3724 
3725 /**
3726  *	netif_dormant - test if device is dormant
3727  *	@dev: network device
3728  *
3729  * Check if device is dormant.
3730  */
3731 static inline bool netif_dormant(const struct net_device *dev)
3732 {
3733 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
3734 }
3735 
3736 
3737 /**
3738  *	netif_oper_up - test if device is operational
3739  *	@dev: network device
3740  *
3741  * Check if carrier is operational
3742  */
3743 static inline bool netif_oper_up(const struct net_device *dev)
3744 {
3745 	return (dev->operstate == IF_OPER_UP ||
3746 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3747 }
3748 
3749 /**
3750  *	netif_device_present - is device available or removed
3751  *	@dev: network device
3752  *
3753  * Check if device has not been removed from system.
3754  */
3755 static inline bool netif_device_present(struct net_device *dev)
3756 {
3757 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
3758 }
3759 
3760 void netif_device_detach(struct net_device *dev);
3761 
3762 void netif_device_attach(struct net_device *dev);
3763 
3764 /*
3765  * Network interface message level settings
3766  */
3767 
3768 enum {
3769 	NETIF_MSG_DRV		= 0x0001,
3770 	NETIF_MSG_PROBE		= 0x0002,
3771 	NETIF_MSG_LINK		= 0x0004,
3772 	NETIF_MSG_TIMER		= 0x0008,
3773 	NETIF_MSG_IFDOWN	= 0x0010,
3774 	NETIF_MSG_IFUP		= 0x0020,
3775 	NETIF_MSG_RX_ERR	= 0x0040,
3776 	NETIF_MSG_TX_ERR	= 0x0080,
3777 	NETIF_MSG_TX_QUEUED	= 0x0100,
3778 	NETIF_MSG_INTR		= 0x0200,
3779 	NETIF_MSG_TX_DONE	= 0x0400,
3780 	NETIF_MSG_RX_STATUS	= 0x0800,
3781 	NETIF_MSG_PKTDATA	= 0x1000,
3782 	NETIF_MSG_HW		= 0x2000,
3783 	NETIF_MSG_WOL		= 0x4000,
3784 };
3785 
3786 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
3787 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
3788 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
3789 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
3790 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
3791 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
3792 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
3793 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
3794 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3795 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
3796 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
3797 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
3798 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
3799 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
3800 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
3801 
3802 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3803 {
3804 	/* use default */
3805 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3806 		return default_msg_enable_bits;
3807 	if (debug_value == 0)	/* no output */
3808 		return 0;
3809 	/* set low N bits */
3810 	return (1 << debug_value) - 1;
3811 }
3812 
3813 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
3814 {
3815 	spin_lock(&txq->_xmit_lock);
3816 	txq->xmit_lock_owner = cpu;
3817 }
3818 
3819 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3820 {
3821 	__acquire(&txq->_xmit_lock);
3822 	return true;
3823 }
3824 
3825 static inline void __netif_tx_release(struct netdev_queue *txq)
3826 {
3827 	__release(&txq->_xmit_lock);
3828 }
3829 
3830 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3831 {
3832 	spin_lock_bh(&txq->_xmit_lock);
3833 	txq->xmit_lock_owner = smp_processor_id();
3834 }
3835 
3836 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
3837 {
3838 	bool ok = spin_trylock(&txq->_xmit_lock);
3839 	if (likely(ok))
3840 		txq->xmit_lock_owner = smp_processor_id();
3841 	return ok;
3842 }
3843 
3844 static inline void __netif_tx_unlock(struct netdev_queue *txq)
3845 {
3846 	txq->xmit_lock_owner = -1;
3847 	spin_unlock(&txq->_xmit_lock);
3848 }
3849 
3850 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3851 {
3852 	txq->xmit_lock_owner = -1;
3853 	spin_unlock_bh(&txq->_xmit_lock);
3854 }
3855 
3856 static inline void txq_trans_update(struct netdev_queue *txq)
3857 {
3858 	if (txq->xmit_lock_owner != -1)
3859 		txq->trans_start = jiffies;
3860 }
3861 
3862 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3863 static inline void netif_trans_update(struct net_device *dev)
3864 {
3865 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3866 
3867 	if (txq->trans_start != jiffies)
3868 		txq->trans_start = jiffies;
3869 }
3870 
3871 /**
3872  *	netif_tx_lock - grab network device transmit lock
3873  *	@dev: network device
3874  *
3875  * Get network device transmit lock
3876  */
3877 static inline void netif_tx_lock(struct net_device *dev)
3878 {
3879 	unsigned int i;
3880 	int cpu;
3881 
3882 	spin_lock(&dev->tx_global_lock);
3883 	cpu = smp_processor_id();
3884 	for (i = 0; i < dev->num_tx_queues; i++) {
3885 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3886 
3887 		/* We are the only thread of execution doing a
3888 		 * freeze, but we have to grab the _xmit_lock in
3889 		 * order to synchronize with threads which are in
3890 		 * the ->hard_start_xmit() handler and already
3891 		 * checked the frozen bit.
3892 		 */
3893 		__netif_tx_lock(txq, cpu);
3894 		set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3895 		__netif_tx_unlock(txq);
3896 	}
3897 }
3898 
3899 static inline void netif_tx_lock_bh(struct net_device *dev)
3900 {
3901 	local_bh_disable();
3902 	netif_tx_lock(dev);
3903 }
3904 
3905 static inline void netif_tx_unlock(struct net_device *dev)
3906 {
3907 	unsigned int i;
3908 
3909 	for (i = 0; i < dev->num_tx_queues; i++) {
3910 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3911 
3912 		/* No need to grab the _xmit_lock here.  If the
3913 		 * queue is not stopped for another reason, we
3914 		 * force a schedule.
3915 		 */
3916 		clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
3917 		netif_schedule_queue(txq);
3918 	}
3919 	spin_unlock(&dev->tx_global_lock);
3920 }
3921 
3922 static inline void netif_tx_unlock_bh(struct net_device *dev)
3923 {
3924 	netif_tx_unlock(dev);
3925 	local_bh_enable();
3926 }
3927 
3928 #define HARD_TX_LOCK(dev, txq, cpu) {			\
3929 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3930 		__netif_tx_lock(txq, cpu);		\
3931 	} else {					\
3932 		__netif_tx_acquire(txq);		\
3933 	}						\
3934 }
3935 
3936 #define HARD_TX_TRYLOCK(dev, txq)			\
3937 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
3938 		__netif_tx_trylock(txq) :		\
3939 		__netif_tx_acquire(txq))
3940 
3941 #define HARD_TX_UNLOCK(dev, txq) {			\
3942 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
3943 		__netif_tx_unlock(txq);			\
3944 	} else {					\
3945 		__netif_tx_release(txq);		\
3946 	}						\
3947 }
3948 
3949 static inline void netif_tx_disable(struct net_device *dev)
3950 {
3951 	unsigned int i;
3952 	int cpu;
3953 
3954 	local_bh_disable();
3955 	cpu = smp_processor_id();
3956 	for (i = 0; i < dev->num_tx_queues; i++) {
3957 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3958 
3959 		__netif_tx_lock(txq, cpu);
3960 		netif_tx_stop_queue(txq);
3961 		__netif_tx_unlock(txq);
3962 	}
3963 	local_bh_enable();
3964 }
3965 
3966 static inline void netif_addr_lock(struct net_device *dev)
3967 {
3968 	spin_lock(&dev->addr_list_lock);
3969 }
3970 
3971 static inline void netif_addr_lock_nested(struct net_device *dev)
3972 {
3973 	int subclass = SINGLE_DEPTH_NESTING;
3974 
3975 	if (dev->netdev_ops->ndo_get_lock_subclass)
3976 		subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3977 
3978 	spin_lock_nested(&dev->addr_list_lock, subclass);
3979 }
3980 
3981 static inline void netif_addr_lock_bh(struct net_device *dev)
3982 {
3983 	spin_lock_bh(&dev->addr_list_lock);
3984 }
3985 
3986 static inline void netif_addr_unlock(struct net_device *dev)
3987 {
3988 	spin_unlock(&dev->addr_list_lock);
3989 }
3990 
3991 static inline void netif_addr_unlock_bh(struct net_device *dev)
3992 {
3993 	spin_unlock_bh(&dev->addr_list_lock);
3994 }
3995 
3996 /*
3997  * dev_addrs walker. Should be used only for read access. Call with
3998  * rcu_read_lock held.
3999  */
4000 #define for_each_dev_addr(dev, ha) \
4001 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4002 
4003 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4004 
4005 void ether_setup(struct net_device *dev);
4006 
4007 /* Support for loadable net-drivers */
4008 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4009 				    unsigned char name_assign_type,
4010 				    void (*setup)(struct net_device *),
4011 				    unsigned int txqs, unsigned int rxqs);
4012 int dev_get_valid_name(struct net *net, struct net_device *dev,
4013 		       const char *name);
4014 
4015 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4016 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4017 
4018 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4019 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4020 			 count)
4021 
4022 int register_netdev(struct net_device *dev);
4023 void unregister_netdev(struct net_device *dev);
4024 
4025 /* General hardware address lists handling functions */
4026 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4027 		   struct netdev_hw_addr_list *from_list, int addr_len);
4028 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4029 		      struct netdev_hw_addr_list *from_list, int addr_len);
4030 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4031 		       struct net_device *dev,
4032 		       int (*sync)(struct net_device *, const unsigned char *),
4033 		       int (*unsync)(struct net_device *,
4034 				     const unsigned char *));
4035 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4036 			  struct net_device *dev,
4037 			  int (*unsync)(struct net_device *,
4038 					const unsigned char *));
4039 void __hw_addr_init(struct netdev_hw_addr_list *list);
4040 
4041 /* Functions used for device addresses handling */
4042 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4043 		 unsigned char addr_type);
4044 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4045 		 unsigned char addr_type);
4046 void dev_addr_flush(struct net_device *dev);
4047 int dev_addr_init(struct net_device *dev);
4048 
4049 /* Functions used for unicast addresses handling */
4050 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4051 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4052 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4053 int dev_uc_sync(struct net_device *to, struct net_device *from);
4054 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4055 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4056 void dev_uc_flush(struct net_device *dev);
4057 void dev_uc_init(struct net_device *dev);
4058 
4059 /**
4060  *  __dev_uc_sync - Synchonize device's unicast list
4061  *  @dev:  device to sync
4062  *  @sync: function to call if address should be added
4063  *  @unsync: function to call if address should be removed
4064  *
4065  *  Add newly added addresses to the interface, and release
4066  *  addresses that have been deleted.
4067  */
4068 static inline int __dev_uc_sync(struct net_device *dev,
4069 				int (*sync)(struct net_device *,
4070 					    const unsigned char *),
4071 				int (*unsync)(struct net_device *,
4072 					      const unsigned char *))
4073 {
4074 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4075 }
4076 
4077 /**
4078  *  __dev_uc_unsync - Remove synchronized addresses from device
4079  *  @dev:  device to sync
4080  *  @unsync: function to call if address should be removed
4081  *
4082  *  Remove all addresses that were added to the device by dev_uc_sync().
4083  */
4084 static inline void __dev_uc_unsync(struct net_device *dev,
4085 				   int (*unsync)(struct net_device *,
4086 						 const unsigned char *))
4087 {
4088 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
4089 }
4090 
4091 /* Functions used for multicast addresses handling */
4092 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4093 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4094 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4095 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4096 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4097 int dev_mc_sync(struct net_device *to, struct net_device *from);
4098 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4099 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4100 void dev_mc_flush(struct net_device *dev);
4101 void dev_mc_init(struct net_device *dev);
4102 
4103 /**
4104  *  __dev_mc_sync - Synchonize device's multicast list
4105  *  @dev:  device to sync
4106  *  @sync: function to call if address should be added
4107  *  @unsync: function to call if address should be removed
4108  *
4109  *  Add newly added addresses to the interface, and release
4110  *  addresses that have been deleted.
4111  */
4112 static inline int __dev_mc_sync(struct net_device *dev,
4113 				int (*sync)(struct net_device *,
4114 					    const unsigned char *),
4115 				int (*unsync)(struct net_device *,
4116 					      const unsigned char *))
4117 {
4118 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4119 }
4120 
4121 /**
4122  *  __dev_mc_unsync - Remove synchronized addresses from device
4123  *  @dev:  device to sync
4124  *  @unsync: function to call if address should be removed
4125  *
4126  *  Remove all addresses that were added to the device by dev_mc_sync().
4127  */
4128 static inline void __dev_mc_unsync(struct net_device *dev,
4129 				   int (*unsync)(struct net_device *,
4130 						 const unsigned char *))
4131 {
4132 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
4133 }
4134 
4135 /* Functions used for secondary unicast and multicast support */
4136 void dev_set_rx_mode(struct net_device *dev);
4137 void __dev_set_rx_mode(struct net_device *dev);
4138 int dev_set_promiscuity(struct net_device *dev, int inc);
4139 int dev_set_allmulti(struct net_device *dev, int inc);
4140 void netdev_state_change(struct net_device *dev);
4141 void netdev_notify_peers(struct net_device *dev);
4142 void netdev_features_change(struct net_device *dev);
4143 /* Load a device via the kmod */
4144 void dev_load(struct net *net, const char *name);
4145 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4146 					struct rtnl_link_stats64 *storage);
4147 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4148 			     const struct net_device_stats *netdev_stats);
4149 
4150 extern int		netdev_max_backlog;
4151 extern int		netdev_tstamp_prequeue;
4152 extern int		weight_p;
4153 extern int		dev_weight_rx_bias;
4154 extern int		dev_weight_tx_bias;
4155 extern int		dev_rx_weight;
4156 extern int		dev_tx_weight;
4157 
4158 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4159 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4160 						     struct list_head **iter);
4161 struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
4162 						     struct list_head **iter);
4163 
4164 /* iterate through upper list, must be called under RCU read lock */
4165 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4166 	for (iter = &(dev)->adj_list.upper, \
4167 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4168 	     updev; \
4169 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4170 
4171 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4172 				  int (*fn)(struct net_device *upper_dev,
4173 					    void *data),
4174 				  void *data);
4175 
4176 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4177 				  struct net_device *upper_dev);
4178 
4179 bool netdev_has_any_upper_dev(struct net_device *dev);
4180 
4181 void *netdev_lower_get_next_private(struct net_device *dev,
4182 				    struct list_head **iter);
4183 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4184 					struct list_head **iter);
4185 
4186 #define netdev_for_each_lower_private(dev, priv, iter) \
4187 	for (iter = (dev)->adj_list.lower.next, \
4188 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
4189 	     priv; \
4190 	     priv = netdev_lower_get_next_private(dev, &(iter)))
4191 
4192 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4193 	for (iter = &(dev)->adj_list.lower, \
4194 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4195 	     priv; \
4196 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4197 
4198 void *netdev_lower_get_next(struct net_device *dev,
4199 				struct list_head **iter);
4200 
4201 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4202 	for (iter = (dev)->adj_list.lower.next, \
4203 	     ldev = netdev_lower_get_next(dev, &(iter)); \
4204 	     ldev; \
4205 	     ldev = netdev_lower_get_next(dev, &(iter)))
4206 
4207 struct net_device *netdev_all_lower_get_next(struct net_device *dev,
4208 					     struct list_head **iter);
4209 struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
4210 						 struct list_head **iter);
4211 
4212 int netdev_walk_all_lower_dev(struct net_device *dev,
4213 			      int (*fn)(struct net_device *lower_dev,
4214 					void *data),
4215 			      void *data);
4216 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4217 				  int (*fn)(struct net_device *lower_dev,
4218 					    void *data),
4219 				  void *data);
4220 
4221 void *netdev_adjacent_get_private(struct list_head *adj_list);
4222 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4223 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4224 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4225 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4226 			  struct netlink_ext_ack *extack);
4227 int netdev_master_upper_dev_link(struct net_device *dev,
4228 				 struct net_device *upper_dev,
4229 				 void *upper_priv, void *upper_info,
4230 				 struct netlink_ext_ack *extack);
4231 void netdev_upper_dev_unlink(struct net_device *dev,
4232 			     struct net_device *upper_dev);
4233 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4234 void *netdev_lower_dev_get_private(struct net_device *dev,
4235 				   struct net_device *lower_dev);
4236 void netdev_lower_state_changed(struct net_device *lower_dev,
4237 				void *lower_state_info);
4238 
4239 /* RSS keys are 40 or 52 bytes long */
4240 #define NETDEV_RSS_KEY_LEN 52
4241 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4242 void netdev_rss_key_fill(void *buffer, size_t len);
4243 
4244 int dev_get_nest_level(struct net_device *dev);
4245 int skb_checksum_help(struct sk_buff *skb);
4246 int skb_crc32c_csum_help(struct sk_buff *skb);
4247 int skb_csum_hwoffload_help(struct sk_buff *skb,
4248 			    const netdev_features_t features);
4249 
4250 struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
4251 				  netdev_features_t features, bool tx_path);
4252 struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
4253 				    netdev_features_t features);
4254 
4255 struct netdev_bonding_info {
4256 	ifslave	slave;
4257 	ifbond	master;
4258 };
4259 
4260 struct netdev_notifier_bonding_info {
4261 	struct netdev_notifier_info info; /* must be first */
4262 	struct netdev_bonding_info  bonding_info;
4263 };
4264 
4265 void netdev_bonding_info_change(struct net_device *dev,
4266 				struct netdev_bonding_info *bonding_info);
4267 
4268 static inline
4269 struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
4270 {
4271 	return __skb_gso_segment(skb, features, true);
4272 }
4273 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4274 
4275 static inline bool can_checksum_protocol(netdev_features_t features,
4276 					 __be16 protocol)
4277 {
4278 	if (protocol == htons(ETH_P_FCOE))
4279 		return !!(features & NETIF_F_FCOE_CRC);
4280 
4281 	/* Assume this is an IP checksum (not SCTP CRC) */
4282 
4283 	if (features & NETIF_F_HW_CSUM) {
4284 		/* Can checksum everything */
4285 		return true;
4286 	}
4287 
4288 	switch (protocol) {
4289 	case htons(ETH_P_IP):
4290 		return !!(features & NETIF_F_IP_CSUM);
4291 	case htons(ETH_P_IPV6):
4292 		return !!(features & NETIF_F_IPV6_CSUM);
4293 	default:
4294 		return false;
4295 	}
4296 }
4297 
4298 #ifdef CONFIG_BUG
4299 void netdev_rx_csum_fault(struct net_device *dev);
4300 #else
4301 static inline void netdev_rx_csum_fault(struct net_device *dev)
4302 {
4303 }
4304 #endif
4305 /* rx skb timestamps */
4306 void net_enable_timestamp(void);
4307 void net_disable_timestamp(void);
4308 
4309 #ifdef CONFIG_PROC_FS
4310 int __init dev_proc_init(void);
4311 #else
4312 #define dev_proc_init() 0
4313 #endif
4314 
4315 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4316 					      struct sk_buff *skb, struct net_device *dev,
4317 					      bool more)
4318 {
4319 	skb->xmit_more = more ? 1 : 0;
4320 	return ops->ndo_start_xmit(skb, dev);
4321 }
4322 
4323 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4324 					    struct netdev_queue *txq, bool more)
4325 {
4326 	const struct net_device_ops *ops = dev->netdev_ops;
4327 	int rc;
4328 
4329 	rc = __netdev_start_xmit(ops, skb, dev, more);
4330 	if (rc == NETDEV_TX_OK)
4331 		txq_trans_update(txq);
4332 
4333 	return rc;
4334 }
4335 
4336 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4337 				const void *ns);
4338 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4339 				 const void *ns);
4340 
4341 static inline int netdev_class_create_file(const struct class_attribute *class_attr)
4342 {
4343 	return netdev_class_create_file_ns(class_attr, NULL);
4344 }
4345 
4346 static inline void netdev_class_remove_file(const struct class_attribute *class_attr)
4347 {
4348 	netdev_class_remove_file_ns(class_attr, NULL);
4349 }
4350 
4351 extern const struct kobj_ns_type_operations net_ns_type_operations;
4352 
4353 const char *netdev_drivername(const struct net_device *dev);
4354 
4355 void linkwatch_run_queue(void);
4356 
4357 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4358 							  netdev_features_t f2)
4359 {
4360 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4361 		if (f1 & NETIF_F_HW_CSUM)
4362 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4363 		else
4364 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4365 	}
4366 
4367 	return f1 & f2;
4368 }
4369 
4370 static inline netdev_features_t netdev_get_wanted_features(
4371 	struct net_device *dev)
4372 {
4373 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4374 }
4375 netdev_features_t netdev_increment_features(netdev_features_t all,
4376 	netdev_features_t one, netdev_features_t mask);
4377 
4378 /* Allow TSO being used on stacked device :
4379  * Performing the GSO segmentation before last device
4380  * is a performance improvement.
4381  */
4382 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4383 							netdev_features_t mask)
4384 {
4385 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4386 }
4387 
4388 int __netdev_update_features(struct net_device *dev);
4389 void netdev_update_features(struct net_device *dev);
4390 void netdev_change_features(struct net_device *dev);
4391 
4392 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4393 					struct net_device *dev);
4394 
4395 netdev_features_t passthru_features_check(struct sk_buff *skb,
4396 					  struct net_device *dev,
4397 					  netdev_features_t features);
4398 netdev_features_t netif_skb_features(struct sk_buff *skb);
4399 
4400 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4401 {
4402 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4403 
4404 	/* check flags correspondence */
4405 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4406 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4407 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4408 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4409 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4410 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4411 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4412 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4413 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4414 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4415 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4416 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
4417 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
4418 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
4419 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
4420 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
4421 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
4422 	BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
4423 
4424 	return (features & feature) == feature;
4425 }
4426 
4427 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
4428 {
4429 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
4430 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
4431 }
4432 
4433 static inline bool netif_needs_gso(struct sk_buff *skb,
4434 				   netdev_features_t features)
4435 {
4436 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
4437 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4438 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
4439 }
4440 
4441 static inline void netif_set_gso_max_size(struct net_device *dev,
4442 					  unsigned int size)
4443 {
4444 	dev->gso_max_size = size;
4445 }
4446 
4447 static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4448 					int pulled_hlen, u16 mac_offset,
4449 					int mac_len)
4450 {
4451 	skb->protocol = protocol;
4452 	skb->encapsulation = 1;
4453 	skb_push(skb, pulled_hlen);
4454 	skb_reset_transport_header(skb);
4455 	skb->mac_header = mac_offset;
4456 	skb->network_header = skb->mac_header + mac_len;
4457 	skb->mac_len = mac_len;
4458 }
4459 
4460 static inline bool netif_is_macsec(const struct net_device *dev)
4461 {
4462 	return dev->priv_flags & IFF_MACSEC;
4463 }
4464 
4465 static inline bool netif_is_macvlan(const struct net_device *dev)
4466 {
4467 	return dev->priv_flags & IFF_MACVLAN;
4468 }
4469 
4470 static inline bool netif_is_macvlan_port(const struct net_device *dev)
4471 {
4472 	return dev->priv_flags & IFF_MACVLAN_PORT;
4473 }
4474 
4475 static inline bool netif_is_bond_master(const struct net_device *dev)
4476 {
4477 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4478 }
4479 
4480 static inline bool netif_is_bond_slave(const struct net_device *dev)
4481 {
4482 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4483 }
4484 
4485 static inline bool netif_supports_nofcs(struct net_device *dev)
4486 {
4487 	return dev->priv_flags & IFF_SUPP_NOFCS;
4488 }
4489 
4490 static inline bool netif_is_l3_master(const struct net_device *dev)
4491 {
4492 	return dev->priv_flags & IFF_L3MDEV_MASTER;
4493 }
4494 
4495 static inline bool netif_is_l3_slave(const struct net_device *dev)
4496 {
4497 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
4498 }
4499 
4500 static inline bool netif_is_bridge_master(const struct net_device *dev)
4501 {
4502 	return dev->priv_flags & IFF_EBRIDGE;
4503 }
4504 
4505 static inline bool netif_is_bridge_port(const struct net_device *dev)
4506 {
4507 	return dev->priv_flags & IFF_BRIDGE_PORT;
4508 }
4509 
4510 static inline bool netif_is_ovs_master(const struct net_device *dev)
4511 {
4512 	return dev->priv_flags & IFF_OPENVSWITCH;
4513 }
4514 
4515 static inline bool netif_is_ovs_port(const struct net_device *dev)
4516 {
4517 	return dev->priv_flags & IFF_OVS_DATAPATH;
4518 }
4519 
4520 static inline bool netif_is_team_master(const struct net_device *dev)
4521 {
4522 	return dev->priv_flags & IFF_TEAM;
4523 }
4524 
4525 static inline bool netif_is_team_port(const struct net_device *dev)
4526 {
4527 	return dev->priv_flags & IFF_TEAM_PORT;
4528 }
4529 
4530 static inline bool netif_is_lag_master(const struct net_device *dev)
4531 {
4532 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
4533 }
4534 
4535 static inline bool netif_is_lag_port(const struct net_device *dev)
4536 {
4537 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4538 }
4539 
4540 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4541 {
4542 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
4543 }
4544 
4545 static inline bool netif_is_failover(const struct net_device *dev)
4546 {
4547 	return dev->priv_flags & IFF_FAILOVER;
4548 }
4549 
4550 static inline bool netif_is_failover_slave(const struct net_device *dev)
4551 {
4552 	return dev->priv_flags & IFF_FAILOVER_SLAVE;
4553 }
4554 
4555 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4556 static inline void netif_keep_dst(struct net_device *dev)
4557 {
4558 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4559 }
4560 
4561 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
4562 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4563 {
4564 	/* TODO: reserve and use an additional IFF bit, if we get more users */
4565 	return dev->priv_flags & IFF_MACSEC;
4566 }
4567 
4568 extern struct pernet_operations __net_initdata loopback_net_ops;
4569 
4570 /* Logging, debugging and troubleshooting/diagnostic helpers. */
4571 
4572 /* netdev_printk helpers, similar to dev_printk */
4573 
4574 static inline const char *netdev_name(const struct net_device *dev)
4575 {
4576 	if (!dev->name[0] || strchr(dev->name, '%'))
4577 		return "(unnamed net_device)";
4578 	return dev->name;
4579 }
4580 
4581 static inline bool netdev_unregistering(const struct net_device *dev)
4582 {
4583 	return dev->reg_state == NETREG_UNREGISTERING;
4584 }
4585 
4586 static inline const char *netdev_reg_state(const struct net_device *dev)
4587 {
4588 	switch (dev->reg_state) {
4589 	case NETREG_UNINITIALIZED: return " (uninitialized)";
4590 	case NETREG_REGISTERED: return "";
4591 	case NETREG_UNREGISTERING: return " (unregistering)";
4592 	case NETREG_UNREGISTERED: return " (unregistered)";
4593 	case NETREG_RELEASED: return " (released)";
4594 	case NETREG_DUMMY: return " (dummy)";
4595 	}
4596 
4597 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4598 	return " (unknown)";
4599 }
4600 
4601 __printf(3, 4)
4602 void netdev_printk(const char *level, const struct net_device *dev,
4603 		   const char *format, ...);
4604 __printf(2, 3)
4605 void netdev_emerg(const struct net_device *dev, const char *format, ...);
4606 __printf(2, 3)
4607 void netdev_alert(const struct net_device *dev, const char *format, ...);
4608 __printf(2, 3)
4609 void netdev_crit(const struct net_device *dev, const char *format, ...);
4610 __printf(2, 3)
4611 void netdev_err(const struct net_device *dev, const char *format, ...);
4612 __printf(2, 3)
4613 void netdev_warn(const struct net_device *dev, const char *format, ...);
4614 __printf(2, 3)
4615 void netdev_notice(const struct net_device *dev, const char *format, ...);
4616 __printf(2, 3)
4617 void netdev_info(const struct net_device *dev, const char *format, ...);
4618 
4619 #define netdev_level_once(level, dev, fmt, ...)			\
4620 do {								\
4621 	static bool __print_once __read_mostly;			\
4622 								\
4623 	if (!__print_once) {					\
4624 		__print_once = true;				\
4625 		netdev_printk(level, dev, fmt, ##__VA_ARGS__);	\
4626 	}							\
4627 } while (0)
4628 
4629 #define netdev_emerg_once(dev, fmt, ...) \
4630 	netdev_level_once(KERN_EMERG, dev, fmt, ##__VA_ARGS__)
4631 #define netdev_alert_once(dev, fmt, ...) \
4632 	netdev_level_once(KERN_ALERT, dev, fmt, ##__VA_ARGS__)
4633 #define netdev_crit_once(dev, fmt, ...) \
4634 	netdev_level_once(KERN_CRIT, dev, fmt, ##__VA_ARGS__)
4635 #define netdev_err_once(dev, fmt, ...) \
4636 	netdev_level_once(KERN_ERR, dev, fmt, ##__VA_ARGS__)
4637 #define netdev_warn_once(dev, fmt, ...) \
4638 	netdev_level_once(KERN_WARNING, dev, fmt, ##__VA_ARGS__)
4639 #define netdev_notice_once(dev, fmt, ...) \
4640 	netdev_level_once(KERN_NOTICE, dev, fmt, ##__VA_ARGS__)
4641 #define netdev_info_once(dev, fmt, ...) \
4642 	netdev_level_once(KERN_INFO, dev, fmt, ##__VA_ARGS__)
4643 
4644 #define MODULE_ALIAS_NETDEV(device) \
4645 	MODULE_ALIAS("netdev-" device)
4646 
4647 #if defined(CONFIG_DYNAMIC_DEBUG)
4648 #define netdev_dbg(__dev, format, args...)			\
4649 do {								\
4650 	dynamic_netdev_dbg(__dev, format, ##args);		\
4651 } while (0)
4652 #elif defined(DEBUG)
4653 #define netdev_dbg(__dev, format, args...)			\
4654 	netdev_printk(KERN_DEBUG, __dev, format, ##args)
4655 #else
4656 #define netdev_dbg(__dev, format, args...)			\
4657 ({								\
4658 	if (0)							\
4659 		netdev_printk(KERN_DEBUG, __dev, format, ##args); \
4660 })
4661 #endif
4662 
4663 #if defined(VERBOSE_DEBUG)
4664 #define netdev_vdbg	netdev_dbg
4665 #else
4666 
4667 #define netdev_vdbg(dev, format, args...)			\
4668 ({								\
4669 	if (0)							\
4670 		netdev_printk(KERN_DEBUG, dev, format, ##args);	\
4671 	0;							\
4672 })
4673 #endif
4674 
4675 /*
4676  * netdev_WARN() acts like dev_printk(), but with the key difference
4677  * of using a WARN/WARN_ON to get the message out, including the
4678  * file/line information and a backtrace.
4679  */
4680 #define netdev_WARN(dev, format, args...)			\
4681 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
4682 	     netdev_reg_state(dev), ##args)
4683 
4684 #define netdev_WARN_ONCE(dev, format, args...)				\
4685 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
4686 		  netdev_reg_state(dev), ##args)
4687 
4688 /* netif printk helpers, similar to netdev_printk */
4689 
4690 #define netif_printk(priv, type, level, dev, fmt, args...)	\
4691 do {					  			\
4692 	if (netif_msg_##type(priv))				\
4693 		netdev_printk(level, (dev), fmt, ##args);	\
4694 } while (0)
4695 
4696 #define netif_level(level, priv, type, dev, fmt, args...)	\
4697 do {								\
4698 	if (netif_msg_##type(priv))				\
4699 		netdev_##level(dev, fmt, ##args);		\
4700 } while (0)
4701 
4702 #define netif_emerg(priv, type, dev, fmt, args...)		\
4703 	netif_level(emerg, priv, type, dev, fmt, ##args)
4704 #define netif_alert(priv, type, dev, fmt, args...)		\
4705 	netif_level(alert, priv, type, dev, fmt, ##args)
4706 #define netif_crit(priv, type, dev, fmt, args...)		\
4707 	netif_level(crit, priv, type, dev, fmt, ##args)
4708 #define netif_err(priv, type, dev, fmt, args...)		\
4709 	netif_level(err, priv, type, dev, fmt, ##args)
4710 #define netif_warn(priv, type, dev, fmt, args...)		\
4711 	netif_level(warn, priv, type, dev, fmt, ##args)
4712 #define netif_notice(priv, type, dev, fmt, args...)		\
4713 	netif_level(notice, priv, type, dev, fmt, ##args)
4714 #define netif_info(priv, type, dev, fmt, args...)		\
4715 	netif_level(info, priv, type, dev, fmt, ##args)
4716 
4717 #if defined(CONFIG_DYNAMIC_DEBUG)
4718 #define netif_dbg(priv, type, netdev, format, args...)		\
4719 do {								\
4720 	if (netif_msg_##type(priv))				\
4721 		dynamic_netdev_dbg(netdev, format, ##args);	\
4722 } while (0)
4723 #elif defined(DEBUG)
4724 #define netif_dbg(priv, type, dev, format, args...)		\
4725 	netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
4726 #else
4727 #define netif_dbg(priv, type, dev, format, args...)			\
4728 ({									\
4729 	if (0)								\
4730 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4731 	0;								\
4732 })
4733 #endif
4734 
4735 /* if @cond then downgrade to debug, else print at @level */
4736 #define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...)     \
4737 	do {                                                              \
4738 		if (cond)                                                 \
4739 			netif_dbg(priv, type, netdev, fmt, ##args);       \
4740 		else                                                      \
4741 			netif_ ## level(priv, type, netdev, fmt, ##args); \
4742 	} while (0)
4743 
4744 #if defined(VERBOSE_DEBUG)
4745 #define netif_vdbg	netif_dbg
4746 #else
4747 #define netif_vdbg(priv, type, dev, format, args...)		\
4748 ({								\
4749 	if (0)							\
4750 		netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4751 	0;							\
4752 })
4753 #endif
4754 
4755 /*
4756  *	The list of packet types we will receive (as opposed to discard)
4757  *	and the routines to invoke.
4758  *
4759  *	Why 16. Because with 16 the only overlap we get on a hash of the
4760  *	low nibble of the protocol value is RARP/SNAP/X.25.
4761  *
4762  *		0800	IP
4763  *		0001	802.3
4764  *		0002	AX.25
4765  *		0004	802.2
4766  *		8035	RARP
4767  *		0005	SNAP
4768  *		0805	X.25
4769  *		0806	ARP
4770  *		8137	IPX
4771  *		0009	Localtalk
4772  *		86DD	IPv6
4773  */
4774 #define PTYPE_HASH_SIZE	(16)
4775 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
4776 
4777 #endif	/* _LINUX_NETDEVICE_H */
4778