xref: /linux-6.15/include/linux/netdevice.h (revision f7b5bd72)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		Definitions for the Interfaces handler.
8  *
9  * Version:	@(#)dev.h	1.0.10	08/12/93
10  *
11  * Authors:	Ross Biro
12  *		Fred N. van Kempen, <[email protected]>
13  *		Corey Minyard <[email protected]>
14  *		Donald J. Becker, <[email protected]>
15  *		Alan Cox, <[email protected]>
16  *		Bjorn Ekwall. <[email protected]>
17  *              Pekka Riikonen <[email protected]>
18  *
19  *		Moved to /usr/include/linux for NET3
20  */
21 #ifndef _LINUX_NETDEVICE_H
22 #define _LINUX_NETDEVICE_H
23 
24 #include <linux/timer.h>
25 #include <linux/bug.h>
26 #include <linux/delay.h>
27 #include <linux/atomic.h>
28 #include <linux/prefetch.h>
29 #include <asm/cache.h>
30 #include <asm/byteorder.h>
31 #include <asm/local.h>
32 
33 #include <linux/percpu.h>
34 #include <linux/rculist.h>
35 #include <linux/workqueue.h>
36 #include <linux/dynamic_queue_limits.h>
37 
38 #include <net/net_namespace.h>
39 #ifdef CONFIG_DCB
40 #include <net/dcbnl.h>
41 #endif
42 #include <net/netprio_cgroup.h>
43 
44 #include <linux/netdev_features.h>
45 #include <linux/neighbour.h>
46 #include <uapi/linux/netdevice.h>
47 #include <uapi/linux/if_bonding.h>
48 #include <uapi/linux/pkt_cls.h>
49 #include <uapi/linux/netdev.h>
50 #include <linux/hashtable.h>
51 #include <linux/rbtree.h>
52 #include <net/net_trackers.h>
53 #include <net/net_debug.h>
54 #include <net/dropreason-core.h>
55 
56 struct netpoll_info;
57 struct device;
58 struct ethtool_ops;
59 struct kernel_hwtstamp_config;
60 struct phy_device;
61 struct dsa_port;
62 struct ip_tunnel_parm;
63 struct macsec_context;
64 struct macsec_ops;
65 struct netdev_name_node;
66 struct sd_flow_limit;
67 struct sfp_bus;
68 /* 802.11 specific */
69 struct wireless_dev;
70 /* 802.15.4 specific */
71 struct wpan_dev;
72 struct mpls_dev;
73 /* UDP Tunnel offloads */
74 struct udp_tunnel_info;
75 struct udp_tunnel_nic_info;
76 struct udp_tunnel_nic;
77 struct bpf_prog;
78 struct xdp_buff;
79 struct xdp_frame;
80 struct xdp_metadata_ops;
81 struct xdp_md;
82 /* DPLL specific */
83 struct dpll_pin;
84 
85 typedef u32 xdp_features_t;
86 
87 void synchronize_net(void);
88 void netdev_set_default_ethtool_ops(struct net_device *dev,
89 				    const struct ethtool_ops *ops);
90 void netdev_sw_irq_coalesce_default_on(struct net_device *dev);
91 
92 /* Backlog congestion levels */
93 #define NET_RX_SUCCESS		0	/* keep 'em coming, baby */
94 #define NET_RX_DROP		1	/* packet dropped */
95 
96 #define MAX_NEST_DEV 8
97 
98 /*
99  * Transmit return codes: transmit return codes originate from three different
100  * namespaces:
101  *
102  * - qdisc return codes
103  * - driver transmit return codes
104  * - errno values
105  *
106  * Drivers are allowed to return any one of those in their hard_start_xmit()
107  * function. Real network devices commonly used with qdiscs should only return
108  * the driver transmit return codes though - when qdiscs are used, the actual
109  * transmission happens asynchronously, so the value is not propagated to
110  * higher layers. Virtual network devices transmit synchronously; in this case
111  * the driver transmit return codes are consumed by dev_queue_xmit(), and all
112  * others are propagated to higher layers.
113  */
114 
115 /* qdisc ->enqueue() return codes. */
116 #define NET_XMIT_SUCCESS	0x00
117 #define NET_XMIT_DROP		0x01	/* skb dropped			*/
118 #define NET_XMIT_CN		0x02	/* congestion notification	*/
119 #define NET_XMIT_MASK		0x0f	/* qdisc flags in net/sch_generic.h */
120 
121 /* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
122  * indicates that the device will soon be dropping packets, or already drops
123  * some packets of the same priority; prompting us to send less aggressively. */
124 #define net_xmit_eval(e)	((e) == NET_XMIT_CN ? 0 : (e))
125 #define net_xmit_errno(e)	((e) != NET_XMIT_CN ? -ENOBUFS : 0)
126 
127 /* Driver transmit return codes */
128 #define NETDEV_TX_MASK		0xf0
129 
130 enum netdev_tx {
131 	__NETDEV_TX_MIN	 = INT_MIN,	/* make sure enum is signed */
132 	NETDEV_TX_OK	 = 0x00,	/* driver took care of packet */
133 	NETDEV_TX_BUSY	 = 0x10,	/* driver tx path was busy*/
134 };
135 typedef enum netdev_tx netdev_tx_t;
136 
137 /*
138  * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
139  * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
140  */
141 static inline bool dev_xmit_complete(int rc)
142 {
143 	/*
144 	 * Positive cases with an skb consumed by a driver:
145 	 * - successful transmission (rc == NETDEV_TX_OK)
146 	 * - error while transmitting (rc < 0)
147 	 * - error while queueing to a different device (rc & NET_XMIT_MASK)
148 	 */
149 	if (likely(rc < NET_XMIT_MASK))
150 		return true;
151 
152 	return false;
153 }
154 
155 /*
156  *	Compute the worst-case header length according to the protocols
157  *	used.
158  */
159 
160 #if defined(CONFIG_HYPERV_NET)
161 # define LL_MAX_HEADER 128
162 #elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
163 # if defined(CONFIG_MAC80211_MESH)
164 #  define LL_MAX_HEADER 128
165 # else
166 #  define LL_MAX_HEADER 96
167 # endif
168 #else
169 # define LL_MAX_HEADER 32
170 #endif
171 
172 #if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
173     !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
174 #define MAX_HEADER LL_MAX_HEADER
175 #else
176 #define MAX_HEADER (LL_MAX_HEADER + 48)
177 #endif
178 
179 /*
180  *	Old network device statistics. Fields are native words
181  *	(unsigned long) so they can be read and written atomically.
182  */
183 
184 #define NET_DEV_STAT(FIELD)			\
185 	union {					\
186 		unsigned long FIELD;		\
187 		atomic_long_t __##FIELD;	\
188 	}
189 
190 struct net_device_stats {
191 	NET_DEV_STAT(rx_packets);
192 	NET_DEV_STAT(tx_packets);
193 	NET_DEV_STAT(rx_bytes);
194 	NET_DEV_STAT(tx_bytes);
195 	NET_DEV_STAT(rx_errors);
196 	NET_DEV_STAT(tx_errors);
197 	NET_DEV_STAT(rx_dropped);
198 	NET_DEV_STAT(tx_dropped);
199 	NET_DEV_STAT(multicast);
200 	NET_DEV_STAT(collisions);
201 	NET_DEV_STAT(rx_length_errors);
202 	NET_DEV_STAT(rx_over_errors);
203 	NET_DEV_STAT(rx_crc_errors);
204 	NET_DEV_STAT(rx_frame_errors);
205 	NET_DEV_STAT(rx_fifo_errors);
206 	NET_DEV_STAT(rx_missed_errors);
207 	NET_DEV_STAT(tx_aborted_errors);
208 	NET_DEV_STAT(tx_carrier_errors);
209 	NET_DEV_STAT(tx_fifo_errors);
210 	NET_DEV_STAT(tx_heartbeat_errors);
211 	NET_DEV_STAT(tx_window_errors);
212 	NET_DEV_STAT(rx_compressed);
213 	NET_DEV_STAT(tx_compressed);
214 };
215 #undef NET_DEV_STAT
216 
217 /* per-cpu stats, allocated on demand.
218  * Try to fit them in a single cache line, for dev_get_stats() sake.
219  */
220 struct net_device_core_stats {
221 	unsigned long	rx_dropped;
222 	unsigned long	tx_dropped;
223 	unsigned long	rx_nohandler;
224 	unsigned long	rx_otherhost_dropped;
225 } __aligned(4 * sizeof(unsigned long));
226 
227 #include <linux/cache.h>
228 #include <linux/skbuff.h>
229 
230 #ifdef CONFIG_RPS
231 #include <linux/static_key.h>
232 extern struct static_key_false rps_needed;
233 extern struct static_key_false rfs_needed;
234 #endif
235 
236 struct neighbour;
237 struct neigh_parms;
238 struct sk_buff;
239 
240 struct netdev_hw_addr {
241 	struct list_head	list;
242 	struct rb_node		node;
243 	unsigned char		addr[MAX_ADDR_LEN];
244 	unsigned char		type;
245 #define NETDEV_HW_ADDR_T_LAN		1
246 #define NETDEV_HW_ADDR_T_SAN		2
247 #define NETDEV_HW_ADDR_T_UNICAST	3
248 #define NETDEV_HW_ADDR_T_MULTICAST	4
249 	bool			global_use;
250 	int			sync_cnt;
251 	int			refcount;
252 	int			synced;
253 	struct rcu_head		rcu_head;
254 };
255 
256 struct netdev_hw_addr_list {
257 	struct list_head	list;
258 	int			count;
259 
260 	/* Auxiliary tree for faster lookup on addition and deletion */
261 	struct rb_root		tree;
262 };
263 
264 #define netdev_hw_addr_list_count(l) ((l)->count)
265 #define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
266 #define netdev_hw_addr_list_for_each(ha, l) \
267 	list_for_each_entry(ha, &(l)->list, list)
268 
269 #define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
270 #define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
271 #define netdev_for_each_uc_addr(ha, dev) \
272 	netdev_hw_addr_list_for_each(ha, &(dev)->uc)
273 #define netdev_for_each_synced_uc_addr(_ha, _dev) \
274 	netdev_for_each_uc_addr((_ha), (_dev)) \
275 		if ((_ha)->sync_cnt)
276 
277 #define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
278 #define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
279 #define netdev_for_each_mc_addr(ha, dev) \
280 	netdev_hw_addr_list_for_each(ha, &(dev)->mc)
281 #define netdev_for_each_synced_mc_addr(_ha, _dev) \
282 	netdev_for_each_mc_addr((_ha), (_dev)) \
283 		if ((_ha)->sync_cnt)
284 
285 struct hh_cache {
286 	unsigned int	hh_len;
287 	seqlock_t	hh_lock;
288 
289 	/* cached hardware header; allow for machine alignment needs.        */
290 #define HH_DATA_MOD	16
291 #define HH_DATA_OFF(__len) \
292 	(HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
293 #define HH_DATA_ALIGN(__len) \
294 	(((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
295 	unsigned long	hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
296 };
297 
298 /* Reserve HH_DATA_MOD byte-aligned hard_header_len, but at least that much.
299  * Alternative is:
300  *   dev->hard_header_len ? (dev->hard_header_len +
301  *                           (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
302  *
303  * We could use other alignment values, but we must maintain the
304  * relationship HH alignment <= LL alignment.
305  */
306 #define LL_RESERVED_SPACE(dev) \
307 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom)) \
308 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
309 #define LL_RESERVED_SPACE_EXTRA(dev,extra) \
310 	((((dev)->hard_header_len + READ_ONCE((dev)->needed_headroom) + (extra)) \
311 	  & ~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
312 
313 struct header_ops {
314 	int	(*create) (struct sk_buff *skb, struct net_device *dev,
315 			   unsigned short type, const void *daddr,
316 			   const void *saddr, unsigned int len);
317 	int	(*parse)(const struct sk_buff *skb, unsigned char *haddr);
318 	int	(*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
319 	void	(*cache_update)(struct hh_cache *hh,
320 				const struct net_device *dev,
321 				const unsigned char *haddr);
322 	bool	(*validate)(const char *ll_header, unsigned int len);
323 	__be16	(*parse_protocol)(const struct sk_buff *skb);
324 };
325 
326 /* These flag bits are private to the generic network queueing
327  * layer; they may not be explicitly referenced by any other
328  * code.
329  */
330 
331 enum netdev_state_t {
332 	__LINK_STATE_START,
333 	__LINK_STATE_PRESENT,
334 	__LINK_STATE_NOCARRIER,
335 	__LINK_STATE_LINKWATCH_PENDING,
336 	__LINK_STATE_DORMANT,
337 	__LINK_STATE_TESTING,
338 };
339 
340 struct gro_list {
341 	struct list_head	list;
342 	int			count;
343 };
344 
345 /*
346  * size of gro hash buckets, must less than bit number of
347  * napi_struct::gro_bitmask
348  */
349 #define GRO_HASH_BUCKETS	8
350 
351 /*
352  * Structure for NAPI scheduling similar to tasklet but with weighting
353  */
354 struct napi_struct {
355 	/* The poll_list must only be managed by the entity which
356 	 * changes the state of the NAPI_STATE_SCHED bit.  This means
357 	 * whoever atomically sets that bit can add this napi_struct
358 	 * to the per-CPU poll_list, and whoever clears that bit
359 	 * can remove from the list right before clearing the bit.
360 	 */
361 	struct list_head	poll_list;
362 
363 	unsigned long		state;
364 	int			weight;
365 	int			defer_hard_irqs_count;
366 	unsigned long		gro_bitmask;
367 	int			(*poll)(struct napi_struct *, int);
368 #ifdef CONFIG_NETPOLL
369 	/* CPU actively polling if netpoll is configured */
370 	int			poll_owner;
371 #endif
372 	/* CPU on which NAPI has been scheduled for processing */
373 	int			list_owner;
374 	struct net_device	*dev;
375 	struct gro_list		gro_hash[GRO_HASH_BUCKETS];
376 	struct sk_buff		*skb;
377 	struct list_head	rx_list; /* Pending GRO_NORMAL skbs */
378 	int			rx_count; /* length of rx_list */
379 	unsigned int		napi_id;
380 	struct hrtimer		timer;
381 	struct task_struct	*thread;
382 	/* control-path-only fields follow */
383 	struct list_head	dev_list;
384 	struct hlist_node	napi_hash_node;
385 };
386 
387 enum {
388 	NAPI_STATE_SCHED,		/* Poll is scheduled */
389 	NAPI_STATE_MISSED,		/* reschedule a napi */
390 	NAPI_STATE_DISABLE,		/* Disable pending */
391 	NAPI_STATE_NPSVC,		/* Netpoll - don't dequeue from poll_list */
392 	NAPI_STATE_LISTED,		/* NAPI added to system lists */
393 	NAPI_STATE_NO_BUSY_POLL,	/* Do not add in napi_hash, no busy polling */
394 	NAPI_STATE_IN_BUSY_POLL,	/* sk_busy_loop() owns this NAPI */
395 	NAPI_STATE_PREFER_BUSY_POLL,	/* prefer busy-polling over softirq processing*/
396 	NAPI_STATE_THREADED,		/* The poll is performed inside its own thread*/
397 	NAPI_STATE_SCHED_THREADED,	/* Napi is currently scheduled in threaded mode */
398 };
399 
400 enum {
401 	NAPIF_STATE_SCHED		= BIT(NAPI_STATE_SCHED),
402 	NAPIF_STATE_MISSED		= BIT(NAPI_STATE_MISSED),
403 	NAPIF_STATE_DISABLE		= BIT(NAPI_STATE_DISABLE),
404 	NAPIF_STATE_NPSVC		= BIT(NAPI_STATE_NPSVC),
405 	NAPIF_STATE_LISTED		= BIT(NAPI_STATE_LISTED),
406 	NAPIF_STATE_NO_BUSY_POLL	= BIT(NAPI_STATE_NO_BUSY_POLL),
407 	NAPIF_STATE_IN_BUSY_POLL	= BIT(NAPI_STATE_IN_BUSY_POLL),
408 	NAPIF_STATE_PREFER_BUSY_POLL	= BIT(NAPI_STATE_PREFER_BUSY_POLL),
409 	NAPIF_STATE_THREADED		= BIT(NAPI_STATE_THREADED),
410 	NAPIF_STATE_SCHED_THREADED	= BIT(NAPI_STATE_SCHED_THREADED),
411 };
412 
413 enum gro_result {
414 	GRO_MERGED,
415 	GRO_MERGED_FREE,
416 	GRO_HELD,
417 	GRO_NORMAL,
418 	GRO_CONSUMED,
419 };
420 typedef enum gro_result gro_result_t;
421 
422 /*
423  * enum rx_handler_result - Possible return values for rx_handlers.
424  * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
425  * further.
426  * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
427  * case skb->dev was changed by rx_handler.
428  * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
429  * @RX_HANDLER_PASS: Do nothing, pass the skb as if no rx_handler was called.
430  *
431  * rx_handlers are functions called from inside __netif_receive_skb(), to do
432  * special processing of the skb, prior to delivery to protocol handlers.
433  *
434  * Currently, a net_device can only have a single rx_handler registered. Trying
435  * to register a second rx_handler will return -EBUSY.
436  *
437  * To register a rx_handler on a net_device, use netdev_rx_handler_register().
438  * To unregister a rx_handler on a net_device, use
439  * netdev_rx_handler_unregister().
440  *
441  * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
442  * do with the skb.
443  *
444  * If the rx_handler consumed the skb in some way, it should return
445  * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
446  * the skb to be delivered in some other way.
447  *
448  * If the rx_handler changed skb->dev, to divert the skb to another
449  * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
450  * new device will be called if it exists.
451  *
452  * If the rx_handler decides the skb should be ignored, it should return
453  * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
454  * are registered on exact device (ptype->dev == skb->dev).
455  *
456  * If the rx_handler didn't change skb->dev, but wants the skb to be normally
457  * delivered, it should return RX_HANDLER_PASS.
458  *
459  * A device without a registered rx_handler will behave as if rx_handler
460  * returned RX_HANDLER_PASS.
461  */
462 
463 enum rx_handler_result {
464 	RX_HANDLER_CONSUMED,
465 	RX_HANDLER_ANOTHER,
466 	RX_HANDLER_EXACT,
467 	RX_HANDLER_PASS,
468 };
469 typedef enum rx_handler_result rx_handler_result_t;
470 typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
471 
472 void __napi_schedule(struct napi_struct *n);
473 void __napi_schedule_irqoff(struct napi_struct *n);
474 
475 static inline bool napi_disable_pending(struct napi_struct *n)
476 {
477 	return test_bit(NAPI_STATE_DISABLE, &n->state);
478 }
479 
480 static inline bool napi_prefer_busy_poll(struct napi_struct *n)
481 {
482 	return test_bit(NAPI_STATE_PREFER_BUSY_POLL, &n->state);
483 }
484 
485 bool napi_schedule_prep(struct napi_struct *n);
486 
487 /**
488  *	napi_schedule - schedule NAPI poll
489  *	@n: NAPI context
490  *
491  * Schedule NAPI poll routine to be called if it is not already
492  * running.
493  */
494 static inline void napi_schedule(struct napi_struct *n)
495 {
496 	if (napi_schedule_prep(n))
497 		__napi_schedule(n);
498 }
499 
500 /**
501  *	napi_schedule_irqoff - schedule NAPI poll
502  *	@n: NAPI context
503  *
504  * Variant of napi_schedule(), assuming hard irqs are masked.
505  */
506 static inline void napi_schedule_irqoff(struct napi_struct *n)
507 {
508 	if (napi_schedule_prep(n))
509 		__napi_schedule_irqoff(n);
510 }
511 
512 /* Try to reschedule poll. Called by dev->poll() after napi_complete().  */
513 static inline bool napi_reschedule(struct napi_struct *napi)
514 {
515 	if (napi_schedule_prep(napi)) {
516 		__napi_schedule(napi);
517 		return true;
518 	}
519 	return false;
520 }
521 
522 /**
523  * napi_complete_done - NAPI processing complete
524  * @n: NAPI context
525  * @work_done: number of packets processed
526  *
527  * Mark NAPI processing as complete. Should only be called if poll budget
528  * has not been completely consumed.
529  * Prefer over napi_complete().
530  * Return false if device should avoid rearming interrupts.
531  */
532 bool napi_complete_done(struct napi_struct *n, int work_done);
533 
534 static inline bool napi_complete(struct napi_struct *n)
535 {
536 	return napi_complete_done(n, 0);
537 }
538 
539 int dev_set_threaded(struct net_device *dev, bool threaded);
540 
541 /**
542  *	napi_disable - prevent NAPI from scheduling
543  *	@n: NAPI context
544  *
545  * Stop NAPI from being scheduled on this context.
546  * Waits till any outstanding processing completes.
547  */
548 void napi_disable(struct napi_struct *n);
549 
550 void napi_enable(struct napi_struct *n);
551 
552 /**
553  *	napi_synchronize - wait until NAPI is not running
554  *	@n: NAPI context
555  *
556  * Wait until NAPI is done being scheduled on this context.
557  * Waits till any outstanding processing completes but
558  * does not disable future activations.
559  */
560 static inline void napi_synchronize(const struct napi_struct *n)
561 {
562 	if (IS_ENABLED(CONFIG_SMP))
563 		while (test_bit(NAPI_STATE_SCHED, &n->state))
564 			msleep(1);
565 	else
566 		barrier();
567 }
568 
569 /**
570  *	napi_if_scheduled_mark_missed - if napi is running, set the
571  *	NAPIF_STATE_MISSED
572  *	@n: NAPI context
573  *
574  * If napi is running, set the NAPIF_STATE_MISSED, and return true if
575  * NAPI is scheduled.
576  **/
577 static inline bool napi_if_scheduled_mark_missed(struct napi_struct *n)
578 {
579 	unsigned long val, new;
580 
581 	val = READ_ONCE(n->state);
582 	do {
583 		if (val & NAPIF_STATE_DISABLE)
584 			return true;
585 
586 		if (!(val & NAPIF_STATE_SCHED))
587 			return false;
588 
589 		new = val | NAPIF_STATE_MISSED;
590 	} while (!try_cmpxchg(&n->state, &val, new));
591 
592 	return true;
593 }
594 
595 enum netdev_queue_state_t {
596 	__QUEUE_STATE_DRV_XOFF,
597 	__QUEUE_STATE_STACK_XOFF,
598 	__QUEUE_STATE_FROZEN,
599 };
600 
601 #define QUEUE_STATE_DRV_XOFF	(1 << __QUEUE_STATE_DRV_XOFF)
602 #define QUEUE_STATE_STACK_XOFF	(1 << __QUEUE_STATE_STACK_XOFF)
603 #define QUEUE_STATE_FROZEN	(1 << __QUEUE_STATE_FROZEN)
604 
605 #define QUEUE_STATE_ANY_XOFF	(QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
606 #define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
607 					QUEUE_STATE_FROZEN)
608 #define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
609 					QUEUE_STATE_FROZEN)
610 
611 /*
612  * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue.  The
613  * netif_tx_* functions below are used to manipulate this flag.  The
614  * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
615  * queue independently.  The netif_xmit_*stopped functions below are called
616  * to check if the queue has been stopped by the driver or stack (either
617  * of the XOFF bits are set in the state).  Drivers should not need to call
618  * netif_xmit*stopped functions, they should only be using netif_tx_*.
619  */
620 
621 struct netdev_queue {
622 /*
623  * read-mostly part
624  */
625 	struct net_device	*dev;
626 	netdevice_tracker	dev_tracker;
627 
628 	struct Qdisc __rcu	*qdisc;
629 	struct Qdisc __rcu	*qdisc_sleeping;
630 #ifdef CONFIG_SYSFS
631 	struct kobject		kobj;
632 #endif
633 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
634 	int			numa_node;
635 #endif
636 	unsigned long		tx_maxrate;
637 	/*
638 	 * Number of TX timeouts for this queue
639 	 * (/sys/class/net/DEV/Q/trans_timeout)
640 	 */
641 	atomic_long_t		trans_timeout;
642 
643 	/* Subordinate device that the queue has been assigned to */
644 	struct net_device	*sb_dev;
645 #ifdef CONFIG_XDP_SOCKETS
646 	struct xsk_buff_pool    *pool;
647 #endif
648 /*
649  * write-mostly part
650  */
651 	spinlock_t		_xmit_lock ____cacheline_aligned_in_smp;
652 	int			xmit_lock_owner;
653 	/*
654 	 * Time (in jiffies) of last Tx
655 	 */
656 	unsigned long		trans_start;
657 
658 	unsigned long		state;
659 
660 #ifdef CONFIG_BQL
661 	struct dql		dql;
662 #endif
663 } ____cacheline_aligned_in_smp;
664 
665 extern int sysctl_fb_tunnels_only_for_init_net;
666 extern int sysctl_devconf_inherit_init_net;
667 
668 /*
669  * sysctl_fb_tunnels_only_for_init_net == 0 : For all netns
670  *                                     == 1 : For initns only
671  *                                     == 2 : For none.
672  */
673 static inline bool net_has_fallback_tunnels(const struct net *net)
674 {
675 #if IS_ENABLED(CONFIG_SYSCTL)
676 	int fb_tunnels_only_for_init_net = READ_ONCE(sysctl_fb_tunnels_only_for_init_net);
677 
678 	return !fb_tunnels_only_for_init_net ||
679 		(net_eq(net, &init_net) && fb_tunnels_only_for_init_net == 1);
680 #else
681 	return true;
682 #endif
683 }
684 
685 static inline int net_inherit_devconf(void)
686 {
687 #if IS_ENABLED(CONFIG_SYSCTL)
688 	return READ_ONCE(sysctl_devconf_inherit_init_net);
689 #else
690 	return 0;
691 #endif
692 }
693 
694 static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
695 {
696 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
697 	return q->numa_node;
698 #else
699 	return NUMA_NO_NODE;
700 #endif
701 }
702 
703 static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
704 {
705 #if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
706 	q->numa_node = node;
707 #endif
708 }
709 
710 #ifdef CONFIG_RPS
711 /*
712  * This structure holds an RPS map which can be of variable length.  The
713  * map is an array of CPUs.
714  */
715 struct rps_map {
716 	unsigned int len;
717 	struct rcu_head rcu;
718 	u16 cpus[];
719 };
720 #define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
721 
722 /*
723  * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
724  * tail pointer for that CPU's input queue at the time of last enqueue, and
725  * a hardware filter index.
726  */
727 struct rps_dev_flow {
728 	u16 cpu;
729 	u16 filter;
730 	unsigned int last_qtail;
731 };
732 #define RPS_NO_FILTER 0xffff
733 
734 /*
735  * The rps_dev_flow_table structure contains a table of flow mappings.
736  */
737 struct rps_dev_flow_table {
738 	unsigned int mask;
739 	struct rcu_head rcu;
740 	struct rps_dev_flow flows[];
741 };
742 #define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
743     ((_num) * sizeof(struct rps_dev_flow)))
744 
745 /*
746  * The rps_sock_flow_table contains mappings of flows to the last CPU
747  * on which they were processed by the application (set in recvmsg).
748  * Each entry is a 32bit value. Upper part is the high-order bits
749  * of flow hash, lower part is CPU number.
750  * rps_cpu_mask is used to partition the space, depending on number of
751  * possible CPUs : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
752  * For example, if 64 CPUs are possible, rps_cpu_mask = 0x3f,
753  * meaning we use 32-6=26 bits for the hash.
754  */
755 struct rps_sock_flow_table {
756 	u32	mask;
757 
758 	u32	ents[] ____cacheline_aligned_in_smp;
759 };
760 #define	RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
761 
762 #define RPS_NO_CPU 0xffff
763 
764 extern u32 rps_cpu_mask;
765 extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
766 
767 static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
768 					u32 hash)
769 {
770 	if (table && hash) {
771 		unsigned int index = hash & table->mask;
772 		u32 val = hash & ~rps_cpu_mask;
773 
774 		/* We only give a hint, preemption can change CPU under us */
775 		val |= raw_smp_processor_id();
776 
777 		/* The following WRITE_ONCE() is paired with the READ_ONCE()
778 		 * here, and another one in get_rps_cpu().
779 		 */
780 		if (READ_ONCE(table->ents[index]) != val)
781 			WRITE_ONCE(table->ents[index], val);
782 	}
783 }
784 
785 #ifdef CONFIG_RFS_ACCEL
786 bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
787 			 u16 filter_id);
788 #endif
789 #endif /* CONFIG_RPS */
790 
791 /* XPS map type and offset of the xps map within net_device->xps_maps[]. */
792 enum xps_map_type {
793 	XPS_CPUS = 0,
794 	XPS_RXQS,
795 	XPS_MAPS_MAX,
796 };
797 
798 #ifdef CONFIG_XPS
799 /*
800  * This structure holds an XPS map which can be of variable length.  The
801  * map is an array of queues.
802  */
803 struct xps_map {
804 	unsigned int len;
805 	unsigned int alloc_len;
806 	struct rcu_head rcu;
807 	u16 queues[];
808 };
809 #define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
810 #define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
811        - sizeof(struct xps_map)) / sizeof(u16))
812 
813 /*
814  * This structure holds all XPS maps for device.  Maps are indexed by CPU.
815  *
816  * We keep track of the number of cpus/rxqs used when the struct is allocated,
817  * in nr_ids. This will help not accessing out-of-bound memory.
818  *
819  * We keep track of the number of traffic classes used when the struct is
820  * allocated, in num_tc. This will be used to navigate the maps, to ensure we're
821  * not crossing its upper bound, as the original dev->num_tc can be updated in
822  * the meantime.
823  */
824 struct xps_dev_maps {
825 	struct rcu_head rcu;
826 	unsigned int nr_ids;
827 	s16 num_tc;
828 	struct xps_map __rcu *attr_map[]; /* Either CPUs map or RXQs map */
829 };
830 
831 #define XPS_CPU_DEV_MAPS_SIZE(_tcs) (sizeof(struct xps_dev_maps) +	\
832 	(nr_cpu_ids * (_tcs) * sizeof(struct xps_map *)))
833 
834 #define XPS_RXQ_DEV_MAPS_SIZE(_tcs, _rxqs) (sizeof(struct xps_dev_maps) +\
835 	(_rxqs * (_tcs) * sizeof(struct xps_map *)))
836 
837 #endif /* CONFIG_XPS */
838 
839 #define TC_MAX_QUEUE	16
840 #define TC_BITMASK	15
841 /* HW offloaded queuing disciplines txq count and offset maps */
842 struct netdev_tc_txq {
843 	u16 count;
844 	u16 offset;
845 };
846 
847 #if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
848 /*
849  * This structure is to hold information about the device
850  * configured to run FCoE protocol stack.
851  */
852 struct netdev_fcoe_hbainfo {
853 	char	manufacturer[64];
854 	char	serial_number[64];
855 	char	hardware_version[64];
856 	char	driver_version[64];
857 	char	optionrom_version[64];
858 	char	firmware_version[64];
859 	char	model[256];
860 	char	model_description[256];
861 };
862 #endif
863 
864 #define MAX_PHYS_ITEM_ID_LEN 32
865 
866 /* This structure holds a unique identifier to identify some
867  * physical item (port for example) used by a netdevice.
868  */
869 struct netdev_phys_item_id {
870 	unsigned char id[MAX_PHYS_ITEM_ID_LEN];
871 	unsigned char id_len;
872 };
873 
874 static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
875 					    struct netdev_phys_item_id *b)
876 {
877 	return a->id_len == b->id_len &&
878 	       memcmp(a->id, b->id, a->id_len) == 0;
879 }
880 
881 typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
882 				       struct sk_buff *skb,
883 				       struct net_device *sb_dev);
884 
885 enum net_device_path_type {
886 	DEV_PATH_ETHERNET = 0,
887 	DEV_PATH_VLAN,
888 	DEV_PATH_BRIDGE,
889 	DEV_PATH_PPPOE,
890 	DEV_PATH_DSA,
891 	DEV_PATH_MTK_WDMA,
892 };
893 
894 struct net_device_path {
895 	enum net_device_path_type	type;
896 	const struct net_device		*dev;
897 	union {
898 		struct {
899 			u16		id;
900 			__be16		proto;
901 			u8		h_dest[ETH_ALEN];
902 		} encap;
903 		struct {
904 			enum {
905 				DEV_PATH_BR_VLAN_KEEP,
906 				DEV_PATH_BR_VLAN_TAG,
907 				DEV_PATH_BR_VLAN_UNTAG,
908 				DEV_PATH_BR_VLAN_UNTAG_HW,
909 			}		vlan_mode;
910 			u16		vlan_id;
911 			__be16		vlan_proto;
912 		} bridge;
913 		struct {
914 			int port;
915 			u16 proto;
916 		} dsa;
917 		struct {
918 			u8 wdma_idx;
919 			u8 queue;
920 			u16 wcid;
921 			u8 bss;
922 		} mtk_wdma;
923 	};
924 };
925 
926 #define NET_DEVICE_PATH_STACK_MAX	5
927 #define NET_DEVICE_PATH_VLAN_MAX	2
928 
929 struct net_device_path_stack {
930 	int			num_paths;
931 	struct net_device_path	path[NET_DEVICE_PATH_STACK_MAX];
932 };
933 
934 struct net_device_path_ctx {
935 	const struct net_device *dev;
936 	u8			daddr[ETH_ALEN];
937 
938 	int			num_vlans;
939 	struct {
940 		u16		id;
941 		__be16		proto;
942 	} vlan[NET_DEVICE_PATH_VLAN_MAX];
943 };
944 
945 enum tc_setup_type {
946 	TC_QUERY_CAPS,
947 	TC_SETUP_QDISC_MQPRIO,
948 	TC_SETUP_CLSU32,
949 	TC_SETUP_CLSFLOWER,
950 	TC_SETUP_CLSMATCHALL,
951 	TC_SETUP_CLSBPF,
952 	TC_SETUP_BLOCK,
953 	TC_SETUP_QDISC_CBS,
954 	TC_SETUP_QDISC_RED,
955 	TC_SETUP_QDISC_PRIO,
956 	TC_SETUP_QDISC_MQ,
957 	TC_SETUP_QDISC_ETF,
958 	TC_SETUP_ROOT_QDISC,
959 	TC_SETUP_QDISC_GRED,
960 	TC_SETUP_QDISC_TAPRIO,
961 	TC_SETUP_FT,
962 	TC_SETUP_QDISC_ETS,
963 	TC_SETUP_QDISC_TBF,
964 	TC_SETUP_QDISC_FIFO,
965 	TC_SETUP_QDISC_HTB,
966 	TC_SETUP_ACT,
967 };
968 
969 /* These structures hold the attributes of bpf state that are being passed
970  * to the netdevice through the bpf op.
971  */
972 enum bpf_netdev_command {
973 	/* Set or clear a bpf program used in the earliest stages of packet
974 	 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
975 	 * is responsible for calling bpf_prog_put on any old progs that are
976 	 * stored. In case of error, the callee need not release the new prog
977 	 * reference, but on success it takes ownership and must bpf_prog_put
978 	 * when it is no longer used.
979 	 */
980 	XDP_SETUP_PROG,
981 	XDP_SETUP_PROG_HW,
982 	/* BPF program for offload callbacks, invoked at program load time. */
983 	BPF_OFFLOAD_MAP_ALLOC,
984 	BPF_OFFLOAD_MAP_FREE,
985 	XDP_SETUP_XSK_POOL,
986 };
987 
988 struct bpf_prog_offload_ops;
989 struct netlink_ext_ack;
990 struct xdp_umem;
991 struct xdp_dev_bulk_queue;
992 struct bpf_xdp_link;
993 
994 enum bpf_xdp_mode {
995 	XDP_MODE_SKB = 0,
996 	XDP_MODE_DRV = 1,
997 	XDP_MODE_HW = 2,
998 	__MAX_XDP_MODE
999 };
1000 
1001 struct bpf_xdp_entity {
1002 	struct bpf_prog *prog;
1003 	struct bpf_xdp_link *link;
1004 };
1005 
1006 struct netdev_bpf {
1007 	enum bpf_netdev_command command;
1008 	union {
1009 		/* XDP_SETUP_PROG */
1010 		struct {
1011 			u32 flags;
1012 			struct bpf_prog *prog;
1013 			struct netlink_ext_ack *extack;
1014 		};
1015 		/* BPF_OFFLOAD_MAP_ALLOC, BPF_OFFLOAD_MAP_FREE */
1016 		struct {
1017 			struct bpf_offloaded_map *offmap;
1018 		};
1019 		/* XDP_SETUP_XSK_POOL */
1020 		struct {
1021 			struct xsk_buff_pool *pool;
1022 			u16 queue_id;
1023 		} xsk;
1024 	};
1025 };
1026 
1027 /* Flags for ndo_xsk_wakeup. */
1028 #define XDP_WAKEUP_RX (1 << 0)
1029 #define XDP_WAKEUP_TX (1 << 1)
1030 
1031 #ifdef CONFIG_XFRM_OFFLOAD
1032 struct xfrmdev_ops {
1033 	int	(*xdo_dev_state_add) (struct xfrm_state *x, struct netlink_ext_ack *extack);
1034 	void	(*xdo_dev_state_delete) (struct xfrm_state *x);
1035 	void	(*xdo_dev_state_free) (struct xfrm_state *x);
1036 	bool	(*xdo_dev_offload_ok) (struct sk_buff *skb,
1037 				       struct xfrm_state *x);
1038 	void	(*xdo_dev_state_advance_esn) (struct xfrm_state *x);
1039 	void	(*xdo_dev_state_update_curlft) (struct xfrm_state *x);
1040 	int	(*xdo_dev_policy_add) (struct xfrm_policy *x, struct netlink_ext_ack *extack);
1041 	void	(*xdo_dev_policy_delete) (struct xfrm_policy *x);
1042 	void	(*xdo_dev_policy_free) (struct xfrm_policy *x);
1043 };
1044 #endif
1045 
1046 struct dev_ifalias {
1047 	struct rcu_head rcuhead;
1048 	char ifalias[];
1049 };
1050 
1051 struct devlink;
1052 struct tlsdev_ops;
1053 
1054 struct netdev_net_notifier {
1055 	struct list_head list;
1056 	struct notifier_block *nb;
1057 };
1058 
1059 /*
1060  * This structure defines the management hooks for network devices.
1061  * The following hooks can be defined; unless noted otherwise, they are
1062  * optional and can be filled with a null pointer.
1063  *
1064  * int (*ndo_init)(struct net_device *dev);
1065  *     This function is called once when a network device is registered.
1066  *     The network device can use this for any late stage initialization
1067  *     or semantic validation. It can fail with an error code which will
1068  *     be propagated back to register_netdev.
1069  *
1070  * void (*ndo_uninit)(struct net_device *dev);
1071  *     This function is called when device is unregistered or when registration
1072  *     fails. It is not called if init fails.
1073  *
1074  * int (*ndo_open)(struct net_device *dev);
1075  *     This function is called when a network device transitions to the up
1076  *     state.
1077  *
1078  * int (*ndo_stop)(struct net_device *dev);
1079  *     This function is called when a network device transitions to the down
1080  *     state.
1081  *
1082  * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1083  *                               struct net_device *dev);
1084  *	Called when a packet needs to be transmitted.
1085  *	Returns NETDEV_TX_OK.  Can return NETDEV_TX_BUSY, but you should stop
1086  *	the queue before that can happen; it's for obsolete devices and weird
1087  *	corner cases, but the stack really does a non-trivial amount
1088  *	of useless work if you return NETDEV_TX_BUSY.
1089  *	Required; cannot be NULL.
1090  *
1091  * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1092  *					   struct net_device *dev
1093  *					   netdev_features_t features);
1094  *	Called by core transmit path to determine if device is capable of
1095  *	performing offload operations on a given packet. This is to give
1096  *	the device an opportunity to implement any restrictions that cannot
1097  *	be otherwise expressed by feature flags. The check is called with
1098  *	the set of features that the stack has calculated and it returns
1099  *	those the driver believes to be appropriate.
1100  *
1101  * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
1102  *                         struct net_device *sb_dev);
1103  *	Called to decide which queue to use when device supports multiple
1104  *	transmit queues.
1105  *
1106  * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
1107  *	This function is called to allow device receiver to make
1108  *	changes to configuration when multicast or promiscuous is enabled.
1109  *
1110  * void (*ndo_set_rx_mode)(struct net_device *dev);
1111  *	This function is called device changes address list filtering.
1112  *	If driver handles unicast address filtering, it should set
1113  *	IFF_UNICAST_FLT in its priv_flags.
1114  *
1115  * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
1116  *	This function  is called when the Media Access Control address
1117  *	needs to be changed. If this interface is not defined, the
1118  *	MAC address can not be changed.
1119  *
1120  * int (*ndo_validate_addr)(struct net_device *dev);
1121  *	Test if Media Access Control address is valid for the device.
1122  *
1123  * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1124  *	Old-style ioctl entry point. This is used internally by the
1125  *	appletalk and ieee802154 subsystems but is no longer called by
1126  *	the device ioctl handler.
1127  *
1128  * int (*ndo_siocbond)(struct net_device *dev, struct ifreq *ifr, int cmd);
1129  *	Used by the bonding driver for its device specific ioctls:
1130  *	SIOCBONDENSLAVE, SIOCBONDRELEASE, SIOCBONDSETHWADDR, SIOCBONDCHANGEACTIVE,
1131  *	SIOCBONDSLAVEINFOQUERY, and SIOCBONDINFOQUERY
1132  *
1133  * * int (*ndo_eth_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
1134  *	Called for ethernet specific ioctls: SIOCGMIIPHY, SIOCGMIIREG,
1135  *	SIOCSMIIREG, SIOCSHWTSTAMP and SIOCGHWTSTAMP.
1136  *
1137  * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
1138  *	Used to set network devices bus interface parameters. This interface
1139  *	is retained for legacy reasons; new devices should use the bus
1140  *	interface (PCI) for low level management.
1141  *
1142  * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
1143  *	Called when a user wants to change the Maximum Transfer Unit
1144  *	of a device.
1145  *
1146  * void (*ndo_tx_timeout)(struct net_device *dev, unsigned int txqueue);
1147  *	Callback used when the transmitter has not made any progress
1148  *	for dev->watchdog ticks.
1149  *
1150  * void (*ndo_get_stats64)(struct net_device *dev,
1151  *                         struct rtnl_link_stats64 *storage);
1152  * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1153  *	Called when a user wants to get the network device usage
1154  *	statistics. Drivers must do one of the following:
1155  *	1. Define @ndo_get_stats64 to fill in a zero-initialised
1156  *	   rtnl_link_stats64 structure passed by the caller.
1157  *	2. Define @ndo_get_stats to update a net_device_stats structure
1158  *	   (which should normally be dev->stats) and return a pointer to
1159  *	   it. The structure may be changed asynchronously only if each
1160  *	   field is written atomically.
1161  *	3. Update dev->stats asynchronously and atomically, and define
1162  *	   neither operation.
1163  *
1164  * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
1165  *	Return true if this device supports offload stats of this attr_id.
1166  *
1167  * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
1168  *	void *attr_data)
1169  *	Get statistics for offload operations by attr_id. Write it into the
1170  *	attr_data pointer.
1171  *
1172  * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
1173  *	If device supports VLAN filtering this function is called when a
1174  *	VLAN id is registered.
1175  *
1176  * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
1177  *	If device supports VLAN filtering this function is called when a
1178  *	VLAN id is unregistered.
1179  *
1180  * void (*ndo_poll_controller)(struct net_device *dev);
1181  *
1182  *	SR-IOV management functions.
1183  * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
1184  * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
1185  *			  u8 qos, __be16 proto);
1186  * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
1187  *			  int max_tx_rate);
1188  * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
1189  * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
1190  * int (*ndo_get_vf_config)(struct net_device *dev,
1191  *			    int vf, struct ifla_vf_info *ivf);
1192  * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
1193  * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
1194  *			  struct nlattr *port[]);
1195  *
1196  *      Enable or disable the VF ability to query its RSS Redirection Table and
1197  *      Hash Key. This is needed since on some devices VF share this information
1198  *      with PF and querying it may introduce a theoretical security risk.
1199  * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
1200  * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
1201  * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
1202  *		       void *type_data);
1203  *	Called to setup any 'tc' scheduler, classifier or action on @dev.
1204  *	This is always called from the stack with the rtnl lock held and netif
1205  *	tx queues stopped. This allows the netdevice to perform queue
1206  *	management safely.
1207  *
1208  *	Fiber Channel over Ethernet (FCoE) offload functions.
1209  * int (*ndo_fcoe_enable)(struct net_device *dev);
1210  *	Called when the FCoE protocol stack wants to start using LLD for FCoE
1211  *	so the underlying device can perform whatever needed configuration or
1212  *	initialization to support acceleration of FCoE traffic.
1213  *
1214  * int (*ndo_fcoe_disable)(struct net_device *dev);
1215  *	Called when the FCoE protocol stack wants to stop using LLD for FCoE
1216  *	so the underlying device can perform whatever needed clean-ups to
1217  *	stop supporting acceleration of FCoE traffic.
1218  *
1219  * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
1220  *			     struct scatterlist *sgl, unsigned int sgc);
1221  *	Called when the FCoE Initiator wants to initialize an I/O that
1222  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1223  *	perform necessary setup and returns 1 to indicate the device is set up
1224  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1225  *
1226  * int (*ndo_fcoe_ddp_done)(struct net_device *dev,  u16 xid);
1227  *	Called when the FCoE Initiator/Target is done with the DDPed I/O as
1228  *	indicated by the FC exchange id 'xid', so the underlying device can
1229  *	clean up and reuse resources for later DDP requests.
1230  *
1231  * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
1232  *			      struct scatterlist *sgl, unsigned int sgc);
1233  *	Called when the FCoE Target wants to initialize an I/O that
1234  *	is a possible candidate for Direct Data Placement (DDP). The LLD can
1235  *	perform necessary setup and returns 1 to indicate the device is set up
1236  *	successfully to perform DDP on this I/O, otherwise this returns 0.
1237  *
1238  * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1239  *			       struct netdev_fcoe_hbainfo *hbainfo);
1240  *	Called when the FCoE Protocol stack wants information on the underlying
1241  *	device. This information is utilized by the FCoE protocol stack to
1242  *	register attributes with Fiber Channel management service as per the
1243  *	FC-GS Fabric Device Management Information(FDMI) specification.
1244  *
1245  * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1246  *	Called when the underlying device wants to override default World Wide
1247  *	Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1248  *	World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1249  *	protocol stack to use.
1250  *
1251  *	RFS acceleration.
1252  * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1253  *			    u16 rxq_index, u32 flow_id);
1254  *	Set hardware filter for RFS.  rxq_index is the target queue index;
1255  *	flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1256  *	Return the filter ID on success, or a negative error code.
1257  *
1258  *	Slave management functions (for bridge, bonding, etc).
1259  * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1260  *	Called to make another netdev an underling.
1261  *
1262  * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1263  *	Called to release previously enslaved netdev.
1264  *
1265  * struct net_device *(*ndo_get_xmit_slave)(struct net_device *dev,
1266  *					    struct sk_buff *skb,
1267  *					    bool all_slaves);
1268  *	Get the xmit slave of master device. If all_slaves is true, function
1269  *	assume all the slaves can transmit.
1270  *
1271  *      Feature/offload setting functions.
1272  * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1273  *		netdev_features_t features);
1274  *	Adjusts the requested feature flags according to device-specific
1275  *	constraints, and returns the resulting flags. Must not modify
1276  *	the device state.
1277  *
1278  * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
1279  *	Called to update device configuration to new features. Passed
1280  *	feature set might be less than what was returned by ndo_fix_features()).
1281  *	Must return >0 or -errno if it changed dev->features itself.
1282  *
1283  * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1284  *		      struct net_device *dev,
1285  *		      const unsigned char *addr, u16 vid, u16 flags,
1286  *		      struct netlink_ext_ack *extack);
1287  *	Adds an FDB entry to dev for addr.
1288  * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1289  *		      struct net_device *dev,
1290  *		      const unsigned char *addr, u16 vid)
1291  *	Deletes the FDB entry from dev coresponding to addr.
1292  * int (*ndo_fdb_del_bulk)(struct ndmsg *ndm, struct nlattr *tb[],
1293  *			   struct net_device *dev,
1294  *			   u16 vid,
1295  *			   struct netlink_ext_ack *extack);
1296  * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
1297  *		       struct net_device *dev, struct net_device *filter_dev,
1298  *		       int *idx)
1299  *	Used to add FDB entries to dump requests. Implementers should add
1300  *	entries to skb and update idx with the number of entries.
1301  *
1302  * int (*ndo_mdb_add)(struct net_device *dev, struct nlattr *tb[],
1303  *		      u16 nlmsg_flags, struct netlink_ext_ack *extack);
1304  *	Adds an MDB entry to dev.
1305  * int (*ndo_mdb_del)(struct net_device *dev, struct nlattr *tb[],
1306  *		      struct netlink_ext_ack *extack);
1307  *	Deletes the MDB entry from dev.
1308  * int (*ndo_mdb_dump)(struct net_device *dev, struct sk_buff *skb,
1309  *		       struct netlink_callback *cb);
1310  *	Dumps MDB entries from dev. The first argument (marker) in the netlink
1311  *	callback is used by core rtnetlink code.
1312  *
1313  * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1314  *			     u16 flags, struct netlink_ext_ack *extack)
1315  * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
1316  *			     struct net_device *dev, u32 filter_mask,
1317  *			     int nlflags)
1318  * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1319  *			     u16 flags);
1320  *
1321  * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1322  *	Called to change device carrier. Soft-devices (like dummy, team, etc)
1323  *	which do not represent real hardware may define this to allow their
1324  *	userspace components to manage their virtual carrier state. Devices
1325  *	that determine carrier state from physical hardware properties (eg
1326  *	network cables) or protocol-dependent mechanisms (eg
1327  *	USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
1328  *
1329  * int (*ndo_get_phys_port_id)(struct net_device *dev,
1330  *			       struct netdev_phys_item_id *ppid);
1331  *	Called to get ID of physical port of this device. If driver does
1332  *	not implement this, it is assumed that the hw is not able to have
1333  *	multiple net devices on single physical port.
1334  *
1335  * int (*ndo_get_port_parent_id)(struct net_device *dev,
1336  *				 struct netdev_phys_item_id *ppid)
1337  *	Called to get the parent ID of the physical port of this device.
1338  *
1339  * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1340  *				 struct net_device *dev)
1341  *	Called by upper layer devices to accelerate switching or other
1342  *	station functionality into hardware. 'pdev is the lowerdev
1343  *	to use for the offload and 'dev' is the net device that will
1344  *	back the offload. Returns a pointer to the private structure
1345  *	the upper layer will maintain.
1346  * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1347  *	Called by upper layer device to delete the station created
1348  *	by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1349  *	the station and priv is the structure returned by the add
1350  *	operation.
1351  * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1352  *			     int queue_index, u32 maxrate);
1353  *	Called when a user wants to set a max-rate limitation of specific
1354  *	TX queue.
1355  * int (*ndo_get_iflink)(const struct net_device *dev);
1356  *	Called to get the iflink value of this device.
1357  * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1358  *	This function is used to get egress tunnel information for given skb.
1359  *	This is useful for retrieving outer tunnel header parameters while
1360  *	sampling packet.
1361  * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1362  *	This function is used to specify the headroom that the skb must
1363  *	consider when allocation skb during packet reception. Setting
1364  *	appropriate rx headroom value allows avoiding skb head copy on
1365  *	forward. Setting a negative value resets the rx headroom to the
1366  *	default value.
1367  * int (*ndo_bpf)(struct net_device *dev, struct netdev_bpf *bpf);
1368  *	This function is used to set or query state related to XDP on the
1369  *	netdevice and manage BPF offload. See definition of
1370  *	enum bpf_netdev_command for details.
1371  * int (*ndo_xdp_xmit)(struct net_device *dev, int n, struct xdp_frame **xdp,
1372  *			u32 flags);
1373  *	This function is used to submit @n XDP packets for transmit on a
1374  *	netdevice. Returns number of frames successfully transmitted, frames
1375  *	that got dropped are freed/returned via xdp_return_frame().
1376  *	Returns negative number, means general error invoking ndo, meaning
1377  *	no frames were xmit'ed and core-caller will free all frames.
1378  * struct net_device *(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1379  *					        struct xdp_buff *xdp);
1380  *      Get the xmit slave of master device based on the xdp_buff.
1381  * int (*ndo_xsk_wakeup)(struct net_device *dev, u32 queue_id, u32 flags);
1382  *      This function is used to wake up the softirq, ksoftirqd or kthread
1383  *	responsible for sending and/or receiving packets on a specific
1384  *	queue id bound to an AF_XDP socket. The flags field specifies if
1385  *	only RX, only Tx, or both should be woken up using the flags
1386  *	XDP_WAKEUP_RX and XDP_WAKEUP_TX.
1387  * int (*ndo_tunnel_ctl)(struct net_device *dev, struct ip_tunnel_parm *p,
1388  *			 int cmd);
1389  *	Add, change, delete or get information on an IPv4 tunnel.
1390  * struct net_device *(*ndo_get_peer_dev)(struct net_device *dev);
1391  *	If a device is paired with a peer device, return the peer instance.
1392  *	The caller must be under RCU read context.
1393  * int (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx, struct net_device_path *path);
1394  *     Get the forwarding path to reach the real device from the HW destination address
1395  * ktime_t (*ndo_get_tstamp)(struct net_device *dev,
1396  *			     const struct skb_shared_hwtstamps *hwtstamps,
1397  *			     bool cycles);
1398  *	Get hardware timestamp based on normal/adjustable time or free running
1399  *	cycle counter. This function is required if physical clock supports a
1400  *	free running cycle counter.
1401  *
1402  * int (*ndo_hwtstamp_get)(struct net_device *dev,
1403  *			   struct kernel_hwtstamp_config *kernel_config);
1404  *	Get the currently configured hardware timestamping parameters for the
1405  *	NIC device.
1406  *
1407  * int (*ndo_hwtstamp_set)(struct net_device *dev,
1408  *			   struct kernel_hwtstamp_config *kernel_config,
1409  *			   struct netlink_ext_ack *extack);
1410  *	Change the hardware timestamping parameters for NIC device.
1411  */
1412 struct net_device_ops {
1413 	int			(*ndo_init)(struct net_device *dev);
1414 	void			(*ndo_uninit)(struct net_device *dev);
1415 	int			(*ndo_open)(struct net_device *dev);
1416 	int			(*ndo_stop)(struct net_device *dev);
1417 	netdev_tx_t		(*ndo_start_xmit)(struct sk_buff *skb,
1418 						  struct net_device *dev);
1419 	netdev_features_t	(*ndo_features_check)(struct sk_buff *skb,
1420 						      struct net_device *dev,
1421 						      netdev_features_t features);
1422 	u16			(*ndo_select_queue)(struct net_device *dev,
1423 						    struct sk_buff *skb,
1424 						    struct net_device *sb_dev);
1425 	void			(*ndo_change_rx_flags)(struct net_device *dev,
1426 						       int flags);
1427 	void			(*ndo_set_rx_mode)(struct net_device *dev);
1428 	int			(*ndo_set_mac_address)(struct net_device *dev,
1429 						       void *addr);
1430 	int			(*ndo_validate_addr)(struct net_device *dev);
1431 	int			(*ndo_do_ioctl)(struct net_device *dev,
1432 					        struct ifreq *ifr, int cmd);
1433 	int			(*ndo_eth_ioctl)(struct net_device *dev,
1434 						 struct ifreq *ifr, int cmd);
1435 	int			(*ndo_siocbond)(struct net_device *dev,
1436 						struct ifreq *ifr, int cmd);
1437 	int			(*ndo_siocwandev)(struct net_device *dev,
1438 						  struct if_settings *ifs);
1439 	int			(*ndo_siocdevprivate)(struct net_device *dev,
1440 						      struct ifreq *ifr,
1441 						      void __user *data, int cmd);
1442 	int			(*ndo_set_config)(struct net_device *dev,
1443 					          struct ifmap *map);
1444 	int			(*ndo_change_mtu)(struct net_device *dev,
1445 						  int new_mtu);
1446 	int			(*ndo_neigh_setup)(struct net_device *dev,
1447 						   struct neigh_parms *);
1448 	void			(*ndo_tx_timeout) (struct net_device *dev,
1449 						   unsigned int txqueue);
1450 
1451 	void			(*ndo_get_stats64)(struct net_device *dev,
1452 						   struct rtnl_link_stats64 *storage);
1453 	bool			(*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
1454 	int			(*ndo_get_offload_stats)(int attr_id,
1455 							 const struct net_device *dev,
1456 							 void *attr_data);
1457 	struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1458 
1459 	int			(*ndo_vlan_rx_add_vid)(struct net_device *dev,
1460 						       __be16 proto, u16 vid);
1461 	int			(*ndo_vlan_rx_kill_vid)(struct net_device *dev,
1462 						        __be16 proto, u16 vid);
1463 #ifdef CONFIG_NET_POLL_CONTROLLER
1464 	void                    (*ndo_poll_controller)(struct net_device *dev);
1465 	int			(*ndo_netpoll_setup)(struct net_device *dev,
1466 						     struct netpoll_info *info);
1467 	void			(*ndo_netpoll_cleanup)(struct net_device *dev);
1468 #endif
1469 	int			(*ndo_set_vf_mac)(struct net_device *dev,
1470 						  int queue, u8 *mac);
1471 	int			(*ndo_set_vf_vlan)(struct net_device *dev,
1472 						   int queue, u16 vlan,
1473 						   u8 qos, __be16 proto);
1474 	int			(*ndo_set_vf_rate)(struct net_device *dev,
1475 						   int vf, int min_tx_rate,
1476 						   int max_tx_rate);
1477 	int			(*ndo_set_vf_spoofchk)(struct net_device *dev,
1478 						       int vf, bool setting);
1479 	int			(*ndo_set_vf_trust)(struct net_device *dev,
1480 						    int vf, bool setting);
1481 	int			(*ndo_get_vf_config)(struct net_device *dev,
1482 						     int vf,
1483 						     struct ifla_vf_info *ivf);
1484 	int			(*ndo_set_vf_link_state)(struct net_device *dev,
1485 							 int vf, int link_state);
1486 	int			(*ndo_get_vf_stats)(struct net_device *dev,
1487 						    int vf,
1488 						    struct ifla_vf_stats
1489 						    *vf_stats);
1490 	int			(*ndo_set_vf_port)(struct net_device *dev,
1491 						   int vf,
1492 						   struct nlattr *port[]);
1493 	int			(*ndo_get_vf_port)(struct net_device *dev,
1494 						   int vf, struct sk_buff *skb);
1495 	int			(*ndo_get_vf_guid)(struct net_device *dev,
1496 						   int vf,
1497 						   struct ifla_vf_guid *node_guid,
1498 						   struct ifla_vf_guid *port_guid);
1499 	int			(*ndo_set_vf_guid)(struct net_device *dev,
1500 						   int vf, u64 guid,
1501 						   int guid_type);
1502 	int			(*ndo_set_vf_rss_query_en)(
1503 						   struct net_device *dev,
1504 						   int vf, bool setting);
1505 	int			(*ndo_setup_tc)(struct net_device *dev,
1506 						enum tc_setup_type type,
1507 						void *type_data);
1508 #if IS_ENABLED(CONFIG_FCOE)
1509 	int			(*ndo_fcoe_enable)(struct net_device *dev);
1510 	int			(*ndo_fcoe_disable)(struct net_device *dev);
1511 	int			(*ndo_fcoe_ddp_setup)(struct net_device *dev,
1512 						      u16 xid,
1513 						      struct scatterlist *sgl,
1514 						      unsigned int sgc);
1515 	int			(*ndo_fcoe_ddp_done)(struct net_device *dev,
1516 						     u16 xid);
1517 	int			(*ndo_fcoe_ddp_target)(struct net_device *dev,
1518 						       u16 xid,
1519 						       struct scatterlist *sgl,
1520 						       unsigned int sgc);
1521 	int			(*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1522 							struct netdev_fcoe_hbainfo *hbainfo);
1523 #endif
1524 
1525 #if IS_ENABLED(CONFIG_LIBFCOE)
1526 #define NETDEV_FCOE_WWNN 0
1527 #define NETDEV_FCOE_WWPN 1
1528 	int			(*ndo_fcoe_get_wwn)(struct net_device *dev,
1529 						    u64 *wwn, int type);
1530 #endif
1531 
1532 #ifdef CONFIG_RFS_ACCEL
1533 	int			(*ndo_rx_flow_steer)(struct net_device *dev,
1534 						     const struct sk_buff *skb,
1535 						     u16 rxq_index,
1536 						     u32 flow_id);
1537 #endif
1538 	int			(*ndo_add_slave)(struct net_device *dev,
1539 						 struct net_device *slave_dev,
1540 						 struct netlink_ext_ack *extack);
1541 	int			(*ndo_del_slave)(struct net_device *dev,
1542 						 struct net_device *slave_dev);
1543 	struct net_device*	(*ndo_get_xmit_slave)(struct net_device *dev,
1544 						      struct sk_buff *skb,
1545 						      bool all_slaves);
1546 	struct net_device*	(*ndo_sk_get_lower_dev)(struct net_device *dev,
1547 							struct sock *sk);
1548 	netdev_features_t	(*ndo_fix_features)(struct net_device *dev,
1549 						    netdev_features_t features);
1550 	int			(*ndo_set_features)(struct net_device *dev,
1551 						    netdev_features_t features);
1552 	int			(*ndo_neigh_construct)(struct net_device *dev,
1553 						       struct neighbour *n);
1554 	void			(*ndo_neigh_destroy)(struct net_device *dev,
1555 						     struct neighbour *n);
1556 
1557 	int			(*ndo_fdb_add)(struct ndmsg *ndm,
1558 					       struct nlattr *tb[],
1559 					       struct net_device *dev,
1560 					       const unsigned char *addr,
1561 					       u16 vid,
1562 					       u16 flags,
1563 					       struct netlink_ext_ack *extack);
1564 	int			(*ndo_fdb_del)(struct ndmsg *ndm,
1565 					       struct nlattr *tb[],
1566 					       struct net_device *dev,
1567 					       const unsigned char *addr,
1568 					       u16 vid, struct netlink_ext_ack *extack);
1569 	int			(*ndo_fdb_del_bulk)(struct ndmsg *ndm,
1570 						    struct nlattr *tb[],
1571 						    struct net_device *dev,
1572 						    u16 vid,
1573 						    struct netlink_ext_ack *extack);
1574 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
1575 						struct netlink_callback *cb,
1576 						struct net_device *dev,
1577 						struct net_device *filter_dev,
1578 						int *idx);
1579 	int			(*ndo_fdb_get)(struct sk_buff *skb,
1580 					       struct nlattr *tb[],
1581 					       struct net_device *dev,
1582 					       const unsigned char *addr,
1583 					       u16 vid, u32 portid, u32 seq,
1584 					       struct netlink_ext_ack *extack);
1585 	int			(*ndo_mdb_add)(struct net_device *dev,
1586 					       struct nlattr *tb[],
1587 					       u16 nlmsg_flags,
1588 					       struct netlink_ext_ack *extack);
1589 	int			(*ndo_mdb_del)(struct net_device *dev,
1590 					       struct nlattr *tb[],
1591 					       struct netlink_ext_ack *extack);
1592 	int			(*ndo_mdb_dump)(struct net_device *dev,
1593 						struct sk_buff *skb,
1594 						struct netlink_callback *cb);
1595 	int			(*ndo_bridge_setlink)(struct net_device *dev,
1596 						      struct nlmsghdr *nlh,
1597 						      u16 flags,
1598 						      struct netlink_ext_ack *extack);
1599 	int			(*ndo_bridge_getlink)(struct sk_buff *skb,
1600 						      u32 pid, u32 seq,
1601 						      struct net_device *dev,
1602 						      u32 filter_mask,
1603 						      int nlflags);
1604 	int			(*ndo_bridge_dellink)(struct net_device *dev,
1605 						      struct nlmsghdr *nlh,
1606 						      u16 flags);
1607 	int			(*ndo_change_carrier)(struct net_device *dev,
1608 						      bool new_carrier);
1609 	int			(*ndo_get_phys_port_id)(struct net_device *dev,
1610 							struct netdev_phys_item_id *ppid);
1611 	int			(*ndo_get_port_parent_id)(struct net_device *dev,
1612 							  struct netdev_phys_item_id *ppid);
1613 	int			(*ndo_get_phys_port_name)(struct net_device *dev,
1614 							  char *name, size_t len);
1615 	void*			(*ndo_dfwd_add_station)(struct net_device *pdev,
1616 							struct net_device *dev);
1617 	void			(*ndo_dfwd_del_station)(struct net_device *pdev,
1618 							void *priv);
1619 
1620 	int			(*ndo_set_tx_maxrate)(struct net_device *dev,
1621 						      int queue_index,
1622 						      u32 maxrate);
1623 	int			(*ndo_get_iflink)(const struct net_device *dev);
1624 	int			(*ndo_fill_metadata_dst)(struct net_device *dev,
1625 						       struct sk_buff *skb);
1626 	void			(*ndo_set_rx_headroom)(struct net_device *dev,
1627 						       int needed_headroom);
1628 	int			(*ndo_bpf)(struct net_device *dev,
1629 					   struct netdev_bpf *bpf);
1630 	int			(*ndo_xdp_xmit)(struct net_device *dev, int n,
1631 						struct xdp_frame **xdp,
1632 						u32 flags);
1633 	struct net_device *	(*ndo_xdp_get_xmit_slave)(struct net_device *dev,
1634 							  struct xdp_buff *xdp);
1635 	int			(*ndo_xsk_wakeup)(struct net_device *dev,
1636 						  u32 queue_id, u32 flags);
1637 	int			(*ndo_tunnel_ctl)(struct net_device *dev,
1638 						  struct ip_tunnel_parm *p, int cmd);
1639 	struct net_device *	(*ndo_get_peer_dev)(struct net_device *dev);
1640 	int                     (*ndo_fill_forward_path)(struct net_device_path_ctx *ctx,
1641                                                          struct net_device_path *path);
1642 	ktime_t			(*ndo_get_tstamp)(struct net_device *dev,
1643 						  const struct skb_shared_hwtstamps *hwtstamps,
1644 						  bool cycles);
1645 	int			(*ndo_hwtstamp_get)(struct net_device *dev,
1646 						    struct kernel_hwtstamp_config *kernel_config);
1647 	int			(*ndo_hwtstamp_set)(struct net_device *dev,
1648 						    struct kernel_hwtstamp_config *kernel_config,
1649 						    struct netlink_ext_ack *extack);
1650 };
1651 
1652 /**
1653  * enum netdev_priv_flags - &struct net_device priv_flags
1654  *
1655  * These are the &struct net_device, they are only set internally
1656  * by drivers and used in the kernel. These flags are invisible to
1657  * userspace; this means that the order of these flags can change
1658  * during any kernel release.
1659  *
1660  * You should have a pretty good reason to be extending these flags.
1661  *
1662  * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1663  * @IFF_EBRIDGE: Ethernet bridging device
1664  * @IFF_BONDING: bonding master or slave
1665  * @IFF_ISATAP: ISATAP interface (RFC4214)
1666  * @IFF_WAN_HDLC: WAN HDLC device
1667  * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1668  *	release skb->dst
1669  * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1670  * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1671  * @IFF_MACVLAN_PORT: device used as macvlan port
1672  * @IFF_BRIDGE_PORT: device used as bridge port
1673  * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1674  * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1675  * @IFF_UNICAST_FLT: Supports unicast filtering
1676  * @IFF_TEAM_PORT: device used as team port
1677  * @IFF_SUPP_NOFCS: device supports sending custom FCS
1678  * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1679  *	change when it's running
1680  * @IFF_MACVLAN: Macvlan device
1681  * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1682  *	underlying stacked devices
1683  * @IFF_L3MDEV_MASTER: device is an L3 master device
1684  * @IFF_NO_QUEUE: device can run without qdisc attached
1685  * @IFF_OPENVSWITCH: device is a Open vSwitch master
1686  * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
1687  * @IFF_TEAM: device is a team device
1688  * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
1689  * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1690  *	entity (i.e. the master device for bridged veth)
1691  * @IFF_MACSEC: device is a MACsec device
1692  * @IFF_NO_RX_HANDLER: device doesn't support the rx_handler hook
1693  * @IFF_FAILOVER: device is a failover master device
1694  * @IFF_FAILOVER_SLAVE: device is lower dev of a failover master device
1695  * @IFF_L3MDEV_RX_HANDLER: only invoke the rx handler of L3 master device
1696  * @IFF_NO_ADDRCONF: prevent ipv6 addrconf
1697  * @IFF_TX_SKB_NO_LINEAR: device/driver is capable of xmitting frames with
1698  *	skb_headlen(skb) == 0 (data starts from frag0)
1699  * @IFF_CHANGE_PROTO_DOWN: device supports setting carrier via IFLA_PROTO_DOWN
1700  * @IFF_SEE_ALL_HWTSTAMP_REQUESTS: device wants to see calls to
1701  *	ndo_hwtstamp_set() for all timestamp requests regardless of source,
1702  *	even if those aren't HWTSTAMP_SOURCE_NETDEV.
1703  */
1704 enum netdev_priv_flags {
1705 	IFF_802_1Q_VLAN			= 1<<0,
1706 	IFF_EBRIDGE			= 1<<1,
1707 	IFF_BONDING			= 1<<2,
1708 	IFF_ISATAP			= 1<<3,
1709 	IFF_WAN_HDLC			= 1<<4,
1710 	IFF_XMIT_DST_RELEASE		= 1<<5,
1711 	IFF_DONT_BRIDGE			= 1<<6,
1712 	IFF_DISABLE_NETPOLL		= 1<<7,
1713 	IFF_MACVLAN_PORT		= 1<<8,
1714 	IFF_BRIDGE_PORT			= 1<<9,
1715 	IFF_OVS_DATAPATH		= 1<<10,
1716 	IFF_TX_SKB_SHARING		= 1<<11,
1717 	IFF_UNICAST_FLT			= 1<<12,
1718 	IFF_TEAM_PORT			= 1<<13,
1719 	IFF_SUPP_NOFCS			= 1<<14,
1720 	IFF_LIVE_ADDR_CHANGE		= 1<<15,
1721 	IFF_MACVLAN			= 1<<16,
1722 	IFF_XMIT_DST_RELEASE_PERM	= 1<<17,
1723 	IFF_L3MDEV_MASTER		= 1<<18,
1724 	IFF_NO_QUEUE			= 1<<19,
1725 	IFF_OPENVSWITCH			= 1<<20,
1726 	IFF_L3MDEV_SLAVE		= 1<<21,
1727 	IFF_TEAM			= 1<<22,
1728 	IFF_RXFH_CONFIGURED		= 1<<23,
1729 	IFF_PHONY_HEADROOM		= 1<<24,
1730 	IFF_MACSEC			= 1<<25,
1731 	IFF_NO_RX_HANDLER		= 1<<26,
1732 	IFF_FAILOVER			= 1<<27,
1733 	IFF_FAILOVER_SLAVE		= 1<<28,
1734 	IFF_L3MDEV_RX_HANDLER		= 1<<29,
1735 	IFF_NO_ADDRCONF			= BIT_ULL(30),
1736 	IFF_TX_SKB_NO_LINEAR		= BIT_ULL(31),
1737 	IFF_CHANGE_PROTO_DOWN		= BIT_ULL(32),
1738 	IFF_SEE_ALL_HWTSTAMP_REQUESTS	= BIT_ULL(33),
1739 };
1740 
1741 #define IFF_802_1Q_VLAN			IFF_802_1Q_VLAN
1742 #define IFF_EBRIDGE			IFF_EBRIDGE
1743 #define IFF_BONDING			IFF_BONDING
1744 #define IFF_ISATAP			IFF_ISATAP
1745 #define IFF_WAN_HDLC			IFF_WAN_HDLC
1746 #define IFF_XMIT_DST_RELEASE		IFF_XMIT_DST_RELEASE
1747 #define IFF_DONT_BRIDGE			IFF_DONT_BRIDGE
1748 #define IFF_DISABLE_NETPOLL		IFF_DISABLE_NETPOLL
1749 #define IFF_MACVLAN_PORT		IFF_MACVLAN_PORT
1750 #define IFF_BRIDGE_PORT			IFF_BRIDGE_PORT
1751 #define IFF_OVS_DATAPATH		IFF_OVS_DATAPATH
1752 #define IFF_TX_SKB_SHARING		IFF_TX_SKB_SHARING
1753 #define IFF_UNICAST_FLT			IFF_UNICAST_FLT
1754 #define IFF_TEAM_PORT			IFF_TEAM_PORT
1755 #define IFF_SUPP_NOFCS			IFF_SUPP_NOFCS
1756 #define IFF_LIVE_ADDR_CHANGE		IFF_LIVE_ADDR_CHANGE
1757 #define IFF_MACVLAN			IFF_MACVLAN
1758 #define IFF_XMIT_DST_RELEASE_PERM	IFF_XMIT_DST_RELEASE_PERM
1759 #define IFF_L3MDEV_MASTER		IFF_L3MDEV_MASTER
1760 #define IFF_NO_QUEUE			IFF_NO_QUEUE
1761 #define IFF_OPENVSWITCH			IFF_OPENVSWITCH
1762 #define IFF_L3MDEV_SLAVE		IFF_L3MDEV_SLAVE
1763 #define IFF_TEAM			IFF_TEAM
1764 #define IFF_RXFH_CONFIGURED		IFF_RXFH_CONFIGURED
1765 #define IFF_PHONY_HEADROOM		IFF_PHONY_HEADROOM
1766 #define IFF_MACSEC			IFF_MACSEC
1767 #define IFF_NO_RX_HANDLER		IFF_NO_RX_HANDLER
1768 #define IFF_FAILOVER			IFF_FAILOVER
1769 #define IFF_FAILOVER_SLAVE		IFF_FAILOVER_SLAVE
1770 #define IFF_L3MDEV_RX_HANDLER		IFF_L3MDEV_RX_HANDLER
1771 #define IFF_TX_SKB_NO_LINEAR		IFF_TX_SKB_NO_LINEAR
1772 
1773 /* Specifies the type of the struct net_device::ml_priv pointer */
1774 enum netdev_ml_priv_type {
1775 	ML_PRIV_NONE,
1776 	ML_PRIV_CAN,
1777 };
1778 
1779 /**
1780  *	struct net_device - The DEVICE structure.
1781  *
1782  *	Actually, this whole structure is a big mistake.  It mixes I/O
1783  *	data with strictly "high-level" data, and it has to know about
1784  *	almost every data structure used in the INET module.
1785  *
1786  *	@name:	This is the first field of the "visible" part of this structure
1787  *		(i.e. as seen by users in the "Space.c" file).  It is the name
1788  *		of the interface.
1789  *
1790  *	@name_node:	Name hashlist node
1791  *	@ifalias:	SNMP alias
1792  *	@mem_end:	Shared memory end
1793  *	@mem_start:	Shared memory start
1794  *	@base_addr:	Device I/O address
1795  *	@irq:		Device IRQ number
1796  *
1797  *	@state:		Generic network queuing layer state, see netdev_state_t
1798  *	@dev_list:	The global list of network devices
1799  *	@napi_list:	List entry used for polling NAPI devices
1800  *	@unreg_list:	List entry  when we are unregistering the
1801  *			device; see the function unregister_netdev
1802  *	@close_list:	List entry used when we are closing the device
1803  *	@ptype_all:     Device-specific packet handlers for all protocols
1804  *	@ptype_specific: Device-specific, protocol-specific packet handlers
1805  *
1806  *	@adj_list:	Directly linked devices, like slaves for bonding
1807  *	@features:	Currently active device features
1808  *	@hw_features:	User-changeable features
1809  *
1810  *	@wanted_features:	User-requested features
1811  *	@vlan_features:		Mask of features inheritable by VLAN devices
1812  *
1813  *	@hw_enc_features:	Mask of features inherited by encapsulating devices
1814  *				This field indicates what encapsulation
1815  *				offloads the hardware is capable of doing,
1816  *				and drivers will need to set them appropriately.
1817  *
1818  *	@mpls_features:	Mask of features inheritable by MPLS
1819  *	@gso_partial_features: value(s) from NETIF_F_GSO\*
1820  *
1821  *	@ifindex:	interface index
1822  *	@group:		The group the device belongs to
1823  *
1824  *	@stats:		Statistics struct, which was left as a legacy, use
1825  *			rtnl_link_stats64 instead
1826  *
1827  *	@core_stats:	core networking counters,
1828  *			do not use this in drivers
1829  *	@carrier_up_count:	Number of times the carrier has been up
1830  *	@carrier_down_count:	Number of times the carrier has been down
1831  *
1832  *	@wireless_handlers:	List of functions to handle Wireless Extensions,
1833  *				instead of ioctl,
1834  *				see <net/iw_handler.h> for details.
1835  *	@wireless_data:	Instance data managed by the core of wireless extensions
1836  *
1837  *	@netdev_ops:	Includes several pointers to callbacks,
1838  *			if one wants to override the ndo_*() functions
1839  *	@xdp_metadata_ops:	Includes pointers to XDP metadata callbacks.
1840  *	@ethtool_ops:	Management operations
1841  *	@l3mdev_ops:	Layer 3 master device operations
1842  *	@ndisc_ops:	Includes callbacks for different IPv6 neighbour
1843  *			discovery handling. Necessary for e.g. 6LoWPAN.
1844  *	@xfrmdev_ops:	Transformation offload operations
1845  *	@tlsdev_ops:	Transport Layer Security offload operations
1846  *	@header_ops:	Includes callbacks for creating,parsing,caching,etc
1847  *			of Layer 2 headers.
1848  *
1849  *	@flags:		Interface flags (a la BSD)
1850  *	@xdp_features:	XDP capability supported by the device
1851  *	@priv_flags:	Like 'flags' but invisible to userspace,
1852  *			see if.h for the definitions
1853  *	@gflags:	Global flags ( kept as legacy )
1854  *	@padded:	How much padding added by alloc_netdev()
1855  *	@operstate:	RFC2863 operstate
1856  *	@link_mode:	Mapping policy to operstate
1857  *	@if_port:	Selectable AUI, TP, ...
1858  *	@dma:		DMA channel
1859  *	@mtu:		Interface MTU value
1860  *	@min_mtu:	Interface Minimum MTU value
1861  *	@max_mtu:	Interface Maximum MTU value
1862  *	@type:		Interface hardware type
1863  *	@hard_header_len: Maximum hardware header length.
1864  *	@min_header_len:  Minimum hardware header length
1865  *
1866  *	@needed_headroom: Extra headroom the hardware may need, but not in all
1867  *			  cases can this be guaranteed
1868  *	@needed_tailroom: Extra tailroom the hardware may need, but not in all
1869  *			  cases can this be guaranteed. Some cases also use
1870  *			  LL_MAX_HEADER instead to allocate the skb
1871  *
1872  *	interface address info:
1873  *
1874  * 	@perm_addr:		Permanent hw address
1875  * 	@addr_assign_type:	Hw address assignment type
1876  * 	@addr_len:		Hardware address length
1877  *	@upper_level:		Maximum depth level of upper devices.
1878  *	@lower_level:		Maximum depth level of lower devices.
1879  *	@neigh_priv_len:	Used in neigh_alloc()
1880  * 	@dev_id:		Used to differentiate devices that share
1881  * 				the same link layer address
1882  * 	@dev_port:		Used to differentiate devices that share
1883  * 				the same function
1884  *	@addr_list_lock:	XXX: need comments on this one
1885  *	@name_assign_type:	network interface name assignment type
1886  *	@uc_promisc:		Counter that indicates promiscuous mode
1887  *				has been enabled due to the need to listen to
1888  *				additional unicast addresses in a device that
1889  *				does not implement ndo_set_rx_mode()
1890  *	@uc:			unicast mac addresses
1891  *	@mc:			multicast mac addresses
1892  *	@dev_addrs:		list of device hw addresses
1893  *	@queues_kset:		Group of all Kobjects in the Tx and RX queues
1894  *	@promiscuity:		Number of times the NIC is told to work in
1895  *				promiscuous mode; if it becomes 0 the NIC will
1896  *				exit promiscuous mode
1897  *	@allmulti:		Counter, enables or disables allmulticast mode
1898  *
1899  *	@vlan_info:	VLAN info
1900  *	@dsa_ptr:	dsa specific data
1901  *	@tipc_ptr:	TIPC specific data
1902  *	@atalk_ptr:	AppleTalk link
1903  *	@ip_ptr:	IPv4 specific data
1904  *	@ip6_ptr:	IPv6 specific data
1905  *	@ax25_ptr:	AX.25 specific data
1906  *	@ieee80211_ptr:	IEEE 802.11 specific data, assign before registering
1907  *	@ieee802154_ptr: IEEE 802.15.4 low-rate Wireless Personal Area Network
1908  *			 device struct
1909  *	@mpls_ptr:	mpls_dev struct pointer
1910  *	@mctp_ptr:	MCTP specific data
1911  *
1912  *	@dev_addr:	Hw address (before bcast,
1913  *			because most packets are unicast)
1914  *
1915  *	@_rx:			Array of RX queues
1916  *	@num_rx_queues:		Number of RX queues
1917  *				allocated at register_netdev() time
1918  *	@real_num_rx_queues: 	Number of RX queues currently active in device
1919  *	@xdp_prog:		XDP sockets filter program pointer
1920  *	@gro_flush_timeout:	timeout for GRO layer in NAPI
1921  *	@napi_defer_hard_irqs:	If not zero, provides a counter that would
1922  *				allow to avoid NIC hard IRQ, on busy queues.
1923  *
1924  *	@rx_handler:		handler for received packets
1925  *	@rx_handler_data: 	XXX: need comments on this one
1926  *	@tcx_ingress:		BPF & clsact qdisc specific data for ingress processing
1927  *	@ingress_queue:		XXX: need comments on this one
1928  *	@nf_hooks_ingress:	netfilter hooks executed for ingress packets
1929  *	@broadcast:		hw bcast address
1930  *
1931  *	@rx_cpu_rmap:	CPU reverse-mapping for RX completion interrupts,
1932  *			indexed by RX queue number. Assigned by driver.
1933  *			This must only be set if the ndo_rx_flow_steer
1934  *			operation is defined
1935  *	@index_hlist:		Device index hash chain
1936  *
1937  *	@_tx:			Array of TX queues
1938  *	@num_tx_queues:		Number of TX queues allocated at alloc_netdev_mq() time
1939  *	@real_num_tx_queues: 	Number of TX queues currently active in device
1940  *	@qdisc:			Root qdisc from userspace point of view
1941  *	@tx_queue_len:		Max frames per queue allowed
1942  *	@tx_global_lock: 	XXX: need comments on this one
1943  *	@xdp_bulkq:		XDP device bulk queue
1944  *	@xps_maps:		all CPUs/RXQs maps for XPS device
1945  *
1946  *	@xps_maps:	XXX: need comments on this one
1947  *	@tcx_egress:		BPF & clsact qdisc specific data for egress processing
1948  *	@nf_hooks_egress:	netfilter hooks executed for egress packets
1949  *	@qdisc_hash:		qdisc hash table
1950  *	@watchdog_timeo:	Represents the timeout that is used by
1951  *				the watchdog (see dev_watchdog())
1952  *	@watchdog_timer:	List of timers
1953  *
1954  *	@proto_down_reason:	reason a netdev interface is held down
1955  *	@pcpu_refcnt:		Number of references to this device
1956  *	@dev_refcnt:		Number of references to this device
1957  *	@refcnt_tracker:	Tracker directory for tracked references to this device
1958  *	@todo_list:		Delayed register/unregister
1959  *	@link_watch_list:	XXX: need comments on this one
1960  *
1961  *	@reg_state:		Register/unregister state machine
1962  *	@dismantle:		Device is going to be freed
1963  *	@rtnl_link_state:	This enum represents the phases of creating
1964  *				a new link
1965  *
1966  *	@needs_free_netdev:	Should unregister perform free_netdev?
1967  *	@priv_destructor:	Called from unregister
1968  *	@npinfo:		XXX: need comments on this one
1969  * 	@nd_net:		Network namespace this network device is inside
1970  *
1971  * 	@ml_priv:	Mid-layer private
1972  *	@ml_priv_type:  Mid-layer private type
1973  * 	@lstats:	Loopback statistics
1974  * 	@tstats:	Tunnel statistics
1975  * 	@dstats:	Dummy statistics
1976  * 	@vstats:	Virtual ethernet statistics
1977  *
1978  *	@garp_port:	GARP
1979  *	@mrp_port:	MRP
1980  *
1981  *	@dm_private:	Drop monitor private
1982  *
1983  *	@dev:		Class/net/name entry
1984  *	@sysfs_groups:	Space for optional device, statistics and wireless
1985  *			sysfs groups
1986  *
1987  *	@sysfs_rx_queue_group:	Space for optional per-rx queue attributes
1988  *	@rtnl_link_ops:	Rtnl_link_ops
1989  *
1990  *	@gso_max_size:	Maximum size of generic segmentation offload
1991  *	@tso_max_size:	Device (as in HW) limit on the max TSO request size
1992  *	@gso_max_segs:	Maximum number of segments that can be passed to the
1993  *			NIC for GSO
1994  *	@tso_max_segs:	Device (as in HW) limit on the max TSO segment count
1995  * 	@gso_ipv4_max_size:	Maximum size of generic segmentation offload,
1996  * 				for IPv4.
1997  *
1998  *	@dcbnl_ops:	Data Center Bridging netlink ops
1999  *	@num_tc:	Number of traffic classes in the net device
2000  *	@tc_to_txq:	XXX: need comments on this one
2001  *	@prio_tc_map:	XXX: need comments on this one
2002  *
2003  *	@fcoe_ddp_xid:	Max exchange id for FCoE LRO by ddp
2004  *
2005  *	@priomap:	XXX: need comments on this one
2006  *	@phydev:	Physical device may attach itself
2007  *			for hardware timestamping
2008  *	@sfp_bus:	attached &struct sfp_bus structure.
2009  *
2010  *	@qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
2011  *
2012  *	@proto_down:	protocol port state information can be sent to the
2013  *			switch driver and used to set the phys state of the
2014  *			switch port.
2015  *
2016  *	@wol_enabled:	Wake-on-LAN is enabled
2017  *
2018  *	@threaded:	napi threaded mode is enabled
2019  *
2020  *	@net_notifier_list:	List of per-net netdev notifier block
2021  *				that follow this device when it is moved
2022  *				to another network namespace.
2023  *
2024  *	@macsec_ops:    MACsec offloading ops
2025  *
2026  *	@udp_tunnel_nic_info:	static structure describing the UDP tunnel
2027  *				offload capabilities of the device
2028  *	@udp_tunnel_nic:	UDP tunnel offload state
2029  *	@xdp_state:		stores info on attached XDP BPF programs
2030  *
2031  *	@nested_level:	Used as a parameter of spin_lock_nested() of
2032  *			dev->addr_list_lock.
2033  *	@unlink_list:	As netif_addr_lock() can be called recursively,
2034  *			keep a list of interfaces to be deleted.
2035  *	@gro_max_size:	Maximum size of aggregated packet in generic
2036  *			receive offload (GRO)
2037  * 	@gro_ipv4_max_size:	Maximum size of aggregated packet in generic
2038  * 				receive offload (GRO), for IPv4.
2039  *	@xdp_zc_max_segs:	Maximum number of segments supported by AF_XDP
2040  *				zero copy driver
2041  *
2042  *	@dev_addr_shadow:	Copy of @dev_addr to catch direct writes.
2043  *	@linkwatch_dev_tracker:	refcount tracker used by linkwatch.
2044  *	@watchdog_dev_tracker:	refcount tracker used by watchdog.
2045  *	@dev_registered_tracker:	tracker for reference held while
2046  *					registered
2047  *	@offload_xstats_l3:	L3 HW stats for this netdevice.
2048  *
2049  *	@devlink_port:	Pointer to related devlink port structure.
2050  *			Assigned by a driver before netdev registration using
2051  *			SET_NETDEV_DEVLINK_PORT macro. This pointer is static
2052  *			during the time netdevice is registered.
2053  *
2054  *	@dpll_pin: Pointer to the SyncE source pin of a DPLL subsystem,
2055  *		   where the clock is recovered.
2056  *
2057  *	FIXME: cleanup struct net_device such that network protocol info
2058  *	moves out.
2059  */
2060 
2061 struct net_device {
2062 	char			name[IFNAMSIZ];
2063 	struct netdev_name_node	*name_node;
2064 	struct dev_ifalias	__rcu *ifalias;
2065 	/*
2066 	 *	I/O specific fields
2067 	 *	FIXME: Merge these and struct ifmap into one
2068 	 */
2069 	unsigned long		mem_end;
2070 	unsigned long		mem_start;
2071 	unsigned long		base_addr;
2072 
2073 	/*
2074 	 *	Some hardware also needs these fields (state,dev_list,
2075 	 *	napi_list,unreg_list,close_list) but they are not
2076 	 *	part of the usual set specified in Space.c.
2077 	 */
2078 
2079 	unsigned long		state;
2080 
2081 	struct list_head	dev_list;
2082 	struct list_head	napi_list;
2083 	struct list_head	unreg_list;
2084 	struct list_head	close_list;
2085 	struct list_head	ptype_all;
2086 	struct list_head	ptype_specific;
2087 
2088 	struct {
2089 		struct list_head upper;
2090 		struct list_head lower;
2091 	} adj_list;
2092 
2093 	/* Read-mostly cache-line for fast-path access */
2094 	unsigned int		flags;
2095 	xdp_features_t		xdp_features;
2096 	unsigned long long	priv_flags;
2097 	const struct net_device_ops *netdev_ops;
2098 	const struct xdp_metadata_ops *xdp_metadata_ops;
2099 	int			ifindex;
2100 	unsigned short		gflags;
2101 	unsigned short		hard_header_len;
2102 
2103 	/* Note : dev->mtu is often read without holding a lock.
2104 	 * Writers usually hold RTNL.
2105 	 * It is recommended to use READ_ONCE() to annotate the reads,
2106 	 * and to use WRITE_ONCE() to annotate the writes.
2107 	 */
2108 	unsigned int		mtu;
2109 	unsigned short		needed_headroom;
2110 	unsigned short		needed_tailroom;
2111 
2112 	netdev_features_t	features;
2113 	netdev_features_t	hw_features;
2114 	netdev_features_t	wanted_features;
2115 	netdev_features_t	vlan_features;
2116 	netdev_features_t	hw_enc_features;
2117 	netdev_features_t	mpls_features;
2118 	netdev_features_t	gso_partial_features;
2119 
2120 	unsigned int		min_mtu;
2121 	unsigned int		max_mtu;
2122 	unsigned short		type;
2123 	unsigned char		min_header_len;
2124 	unsigned char		name_assign_type;
2125 
2126 	int			group;
2127 
2128 	struct net_device_stats	stats; /* not used by modern drivers */
2129 
2130 	struct net_device_core_stats __percpu *core_stats;
2131 
2132 	/* Stats to monitor link on/off, flapping */
2133 	atomic_t		carrier_up_count;
2134 	atomic_t		carrier_down_count;
2135 
2136 #ifdef CONFIG_WIRELESS_EXT
2137 	const struct iw_handler_def *wireless_handlers;
2138 	struct iw_public_data	*wireless_data;
2139 #endif
2140 	const struct ethtool_ops *ethtool_ops;
2141 #ifdef CONFIG_NET_L3_MASTER_DEV
2142 	const struct l3mdev_ops	*l3mdev_ops;
2143 #endif
2144 #if IS_ENABLED(CONFIG_IPV6)
2145 	const struct ndisc_ops *ndisc_ops;
2146 #endif
2147 
2148 #ifdef CONFIG_XFRM_OFFLOAD
2149 	const struct xfrmdev_ops *xfrmdev_ops;
2150 #endif
2151 
2152 #if IS_ENABLED(CONFIG_TLS_DEVICE)
2153 	const struct tlsdev_ops *tlsdev_ops;
2154 #endif
2155 
2156 	const struct header_ops *header_ops;
2157 
2158 	unsigned char		operstate;
2159 	unsigned char		link_mode;
2160 
2161 	unsigned char		if_port;
2162 	unsigned char		dma;
2163 
2164 	/* Interface address info. */
2165 	unsigned char		perm_addr[MAX_ADDR_LEN];
2166 	unsigned char		addr_assign_type;
2167 	unsigned char		addr_len;
2168 	unsigned char		upper_level;
2169 	unsigned char		lower_level;
2170 
2171 	unsigned short		neigh_priv_len;
2172 	unsigned short          dev_id;
2173 	unsigned short          dev_port;
2174 	unsigned short		padded;
2175 
2176 	spinlock_t		addr_list_lock;
2177 	int			irq;
2178 
2179 	struct netdev_hw_addr_list	uc;
2180 	struct netdev_hw_addr_list	mc;
2181 	struct netdev_hw_addr_list	dev_addrs;
2182 
2183 #ifdef CONFIG_SYSFS
2184 	struct kset		*queues_kset;
2185 #endif
2186 #ifdef CONFIG_LOCKDEP
2187 	struct list_head	unlink_list;
2188 #endif
2189 	unsigned int		promiscuity;
2190 	unsigned int		allmulti;
2191 	bool			uc_promisc;
2192 #ifdef CONFIG_LOCKDEP
2193 	unsigned char		nested_level;
2194 #endif
2195 
2196 
2197 	/* Protocol-specific pointers */
2198 
2199 	struct in_device __rcu	*ip_ptr;
2200 	struct inet6_dev __rcu	*ip6_ptr;
2201 #if IS_ENABLED(CONFIG_VLAN_8021Q)
2202 	struct vlan_info __rcu	*vlan_info;
2203 #endif
2204 #if IS_ENABLED(CONFIG_NET_DSA)
2205 	struct dsa_port		*dsa_ptr;
2206 #endif
2207 #if IS_ENABLED(CONFIG_TIPC)
2208 	struct tipc_bearer __rcu *tipc_ptr;
2209 #endif
2210 #if IS_ENABLED(CONFIG_ATALK)
2211 	void 			*atalk_ptr;
2212 #endif
2213 #if IS_ENABLED(CONFIG_AX25)
2214 	void			*ax25_ptr;
2215 #endif
2216 #if IS_ENABLED(CONFIG_CFG80211)
2217 	struct wireless_dev	*ieee80211_ptr;
2218 #endif
2219 #if IS_ENABLED(CONFIG_IEEE802154) || IS_ENABLED(CONFIG_6LOWPAN)
2220 	struct wpan_dev		*ieee802154_ptr;
2221 #endif
2222 #if IS_ENABLED(CONFIG_MPLS_ROUTING)
2223 	struct mpls_dev __rcu	*mpls_ptr;
2224 #endif
2225 #if IS_ENABLED(CONFIG_MCTP)
2226 	struct mctp_dev __rcu	*mctp_ptr;
2227 #endif
2228 
2229 /*
2230  * Cache lines mostly used on receive path (including eth_type_trans())
2231  */
2232 	/* Interface address info used in eth_type_trans() */
2233 	const unsigned char	*dev_addr;
2234 
2235 	struct netdev_rx_queue	*_rx;
2236 	unsigned int		num_rx_queues;
2237 	unsigned int		real_num_rx_queues;
2238 
2239 	struct bpf_prog __rcu	*xdp_prog;
2240 	unsigned long		gro_flush_timeout;
2241 	int			napi_defer_hard_irqs;
2242 #define GRO_LEGACY_MAX_SIZE	65536u
2243 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2244  * and shinfo->gso_segs is a 16bit field.
2245  */
2246 #define GRO_MAX_SIZE		(8 * 65535u)
2247 	unsigned int		gro_max_size;
2248 	unsigned int		gro_ipv4_max_size;
2249 	unsigned int		xdp_zc_max_segs;
2250 	rx_handler_func_t __rcu	*rx_handler;
2251 	void __rcu		*rx_handler_data;
2252 #ifdef CONFIG_NET_XGRESS
2253 	struct bpf_mprog_entry __rcu *tcx_ingress;
2254 #endif
2255 	struct netdev_queue __rcu *ingress_queue;
2256 #ifdef CONFIG_NETFILTER_INGRESS
2257 	struct nf_hook_entries __rcu *nf_hooks_ingress;
2258 #endif
2259 
2260 	unsigned char		broadcast[MAX_ADDR_LEN];
2261 #ifdef CONFIG_RFS_ACCEL
2262 	struct cpu_rmap		*rx_cpu_rmap;
2263 #endif
2264 	struct hlist_node	index_hlist;
2265 
2266 /*
2267  * Cache lines mostly used on transmit path
2268  */
2269 	struct netdev_queue	*_tx ____cacheline_aligned_in_smp;
2270 	unsigned int		num_tx_queues;
2271 	unsigned int		real_num_tx_queues;
2272 	struct Qdisc __rcu	*qdisc;
2273 	unsigned int		tx_queue_len;
2274 	spinlock_t		tx_global_lock;
2275 
2276 	struct xdp_dev_bulk_queue __percpu *xdp_bulkq;
2277 
2278 #ifdef CONFIG_XPS
2279 	struct xps_dev_maps __rcu *xps_maps[XPS_MAPS_MAX];
2280 #endif
2281 #ifdef CONFIG_NET_XGRESS
2282 	struct bpf_mprog_entry __rcu *tcx_egress;
2283 #endif
2284 #ifdef CONFIG_NETFILTER_EGRESS
2285 	struct nf_hook_entries __rcu *nf_hooks_egress;
2286 #endif
2287 
2288 #ifdef CONFIG_NET_SCHED
2289 	DECLARE_HASHTABLE	(qdisc_hash, 4);
2290 #endif
2291 	/* These may be needed for future network-power-down code. */
2292 	struct timer_list	watchdog_timer;
2293 	int			watchdog_timeo;
2294 
2295 	u32                     proto_down_reason;
2296 
2297 	struct list_head	todo_list;
2298 
2299 #ifdef CONFIG_PCPU_DEV_REFCNT
2300 	int __percpu		*pcpu_refcnt;
2301 #else
2302 	refcount_t		dev_refcnt;
2303 #endif
2304 	struct ref_tracker_dir	refcnt_tracker;
2305 
2306 	struct list_head	link_watch_list;
2307 
2308 	enum { NETREG_UNINITIALIZED=0,
2309 	       NETREG_REGISTERED,	/* completed register_netdevice */
2310 	       NETREG_UNREGISTERING,	/* called unregister_netdevice */
2311 	       NETREG_UNREGISTERED,	/* completed unregister todo */
2312 	       NETREG_RELEASED,		/* called free_netdev */
2313 	       NETREG_DUMMY,		/* dummy device for NAPI poll */
2314 	} reg_state:8;
2315 
2316 	bool dismantle;
2317 
2318 	enum {
2319 		RTNL_LINK_INITIALIZED,
2320 		RTNL_LINK_INITIALIZING,
2321 	} rtnl_link_state:16;
2322 
2323 	bool needs_free_netdev;
2324 	void (*priv_destructor)(struct net_device *dev);
2325 
2326 #ifdef CONFIG_NETPOLL
2327 	struct netpoll_info __rcu	*npinfo;
2328 #endif
2329 
2330 	possible_net_t			nd_net;
2331 
2332 	/* mid-layer private */
2333 	void				*ml_priv;
2334 	enum netdev_ml_priv_type	ml_priv_type;
2335 
2336 	union {
2337 		struct pcpu_lstats __percpu		*lstats;
2338 		struct pcpu_sw_netstats __percpu	*tstats;
2339 		struct pcpu_dstats __percpu		*dstats;
2340 	};
2341 
2342 #if IS_ENABLED(CONFIG_GARP)
2343 	struct garp_port __rcu	*garp_port;
2344 #endif
2345 #if IS_ENABLED(CONFIG_MRP)
2346 	struct mrp_port __rcu	*mrp_port;
2347 #endif
2348 #if IS_ENABLED(CONFIG_NET_DROP_MONITOR)
2349 	struct dm_hw_stat_delta __rcu *dm_private;
2350 #endif
2351 	struct device		dev;
2352 	const struct attribute_group *sysfs_groups[4];
2353 	const struct attribute_group *sysfs_rx_queue_group;
2354 
2355 	const struct rtnl_link_ops *rtnl_link_ops;
2356 
2357 	/* for setting kernel sock attribute on TCP connection setup */
2358 #define GSO_MAX_SEGS		65535u
2359 #define GSO_LEGACY_MAX_SIZE	65536u
2360 /* TCP minimal MSS is 8 (TCP_MIN_GSO_SIZE),
2361  * and shinfo->gso_segs is a 16bit field.
2362  */
2363 #define GSO_MAX_SIZE		(8 * GSO_MAX_SEGS)
2364 
2365 	unsigned int		gso_max_size;
2366 #define TSO_LEGACY_MAX_SIZE	65536
2367 #define TSO_MAX_SIZE		UINT_MAX
2368 	unsigned int		tso_max_size;
2369 	u16			gso_max_segs;
2370 #define TSO_MAX_SEGS		U16_MAX
2371 	u16			tso_max_segs;
2372 	unsigned int		gso_ipv4_max_size;
2373 
2374 #ifdef CONFIG_DCB
2375 	const struct dcbnl_rtnl_ops *dcbnl_ops;
2376 #endif
2377 	s16			num_tc;
2378 	struct netdev_tc_txq	tc_to_txq[TC_MAX_QUEUE];
2379 	u8			prio_tc_map[TC_BITMASK + 1];
2380 
2381 #if IS_ENABLED(CONFIG_FCOE)
2382 	unsigned int		fcoe_ddp_xid;
2383 #endif
2384 #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
2385 	struct netprio_map __rcu *priomap;
2386 #endif
2387 	struct phy_device	*phydev;
2388 	struct sfp_bus		*sfp_bus;
2389 	struct lock_class_key	*qdisc_tx_busylock;
2390 	bool			proto_down;
2391 	unsigned		wol_enabled:1;
2392 	unsigned		threaded:1;
2393 
2394 	struct list_head	net_notifier_list;
2395 
2396 #if IS_ENABLED(CONFIG_MACSEC)
2397 	/* MACsec management functions */
2398 	const struct macsec_ops *macsec_ops;
2399 #endif
2400 	const struct udp_tunnel_nic_info	*udp_tunnel_nic_info;
2401 	struct udp_tunnel_nic	*udp_tunnel_nic;
2402 
2403 	/* protected by rtnl_lock */
2404 	struct bpf_xdp_entity	xdp_state[__MAX_XDP_MODE];
2405 
2406 	u8 dev_addr_shadow[MAX_ADDR_LEN];
2407 	netdevice_tracker	linkwatch_dev_tracker;
2408 	netdevice_tracker	watchdog_dev_tracker;
2409 	netdevice_tracker	dev_registered_tracker;
2410 	struct rtnl_hw_stats64	*offload_xstats_l3;
2411 
2412 	struct devlink_port	*devlink_port;
2413 
2414 #if IS_ENABLED(CONFIG_DPLL)
2415 	struct dpll_pin		*dpll_pin;
2416 #endif
2417 };
2418 #define to_net_dev(d) container_of(d, struct net_device, dev)
2419 
2420 /*
2421  * Driver should use this to assign devlink port instance to a netdevice
2422  * before it registers the netdevice. Therefore devlink_port is static
2423  * during the netdev lifetime after it is registered.
2424  */
2425 #define SET_NETDEV_DEVLINK_PORT(dev, port)			\
2426 ({								\
2427 	WARN_ON((dev)->reg_state != NETREG_UNINITIALIZED);	\
2428 	((dev)->devlink_port = (port));				\
2429 })
2430 
2431 static inline bool netif_elide_gro(const struct net_device *dev)
2432 {
2433 	if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
2434 		return true;
2435 	return false;
2436 }
2437 
2438 #define	NETDEV_ALIGN		32
2439 
2440 static inline
2441 int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
2442 {
2443 	return dev->prio_tc_map[prio & TC_BITMASK];
2444 }
2445 
2446 static inline
2447 int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
2448 {
2449 	if (tc >= dev->num_tc)
2450 		return -EINVAL;
2451 
2452 	dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
2453 	return 0;
2454 }
2455 
2456 int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
2457 void netdev_reset_tc(struct net_device *dev);
2458 int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
2459 int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
2460 
2461 static inline
2462 int netdev_get_num_tc(struct net_device *dev)
2463 {
2464 	return dev->num_tc;
2465 }
2466 
2467 static inline void net_prefetch(void *p)
2468 {
2469 	prefetch(p);
2470 #if L1_CACHE_BYTES < 128
2471 	prefetch((u8 *)p + L1_CACHE_BYTES);
2472 #endif
2473 }
2474 
2475 static inline void net_prefetchw(void *p)
2476 {
2477 	prefetchw(p);
2478 #if L1_CACHE_BYTES < 128
2479 	prefetchw((u8 *)p + L1_CACHE_BYTES);
2480 #endif
2481 }
2482 
2483 void netdev_unbind_sb_channel(struct net_device *dev,
2484 			      struct net_device *sb_dev);
2485 int netdev_bind_sb_channel_queue(struct net_device *dev,
2486 				 struct net_device *sb_dev,
2487 				 u8 tc, u16 count, u16 offset);
2488 int netdev_set_sb_channel(struct net_device *dev, u16 channel);
2489 static inline int netdev_get_sb_channel(struct net_device *dev)
2490 {
2491 	return max_t(int, -dev->num_tc, 0);
2492 }
2493 
2494 static inline
2495 struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
2496 					 unsigned int index)
2497 {
2498 	DEBUG_NET_WARN_ON_ONCE(index >= dev->num_tx_queues);
2499 	return &dev->_tx[index];
2500 }
2501 
2502 static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
2503 						    const struct sk_buff *skb)
2504 {
2505 	return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
2506 }
2507 
2508 static inline void netdev_for_each_tx_queue(struct net_device *dev,
2509 					    void (*f)(struct net_device *,
2510 						      struct netdev_queue *,
2511 						      void *),
2512 					    void *arg)
2513 {
2514 	unsigned int i;
2515 
2516 	for (i = 0; i < dev->num_tx_queues; i++)
2517 		f(dev, &dev->_tx[i], arg);
2518 }
2519 
2520 #define netdev_lockdep_set_classes(dev)				\
2521 {								\
2522 	static struct lock_class_key qdisc_tx_busylock_key;	\
2523 	static struct lock_class_key qdisc_xmit_lock_key;	\
2524 	static struct lock_class_key dev_addr_list_lock_key;	\
2525 	unsigned int i;						\
2526 								\
2527 	(dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key;	\
2528 	lockdep_set_class(&(dev)->addr_list_lock,		\
2529 			  &dev_addr_list_lock_key);		\
2530 	for (i = 0; i < (dev)->num_tx_queues; i++)		\
2531 		lockdep_set_class(&(dev)->_tx[i]._xmit_lock,	\
2532 				  &qdisc_xmit_lock_key);	\
2533 }
2534 
2535 u16 netdev_pick_tx(struct net_device *dev, struct sk_buff *skb,
2536 		     struct net_device *sb_dev);
2537 struct netdev_queue *netdev_core_pick_tx(struct net_device *dev,
2538 					 struct sk_buff *skb,
2539 					 struct net_device *sb_dev);
2540 
2541 /* returns the headroom that the master device needs to take in account
2542  * when forwarding to this dev
2543  */
2544 static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
2545 {
2546 	return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
2547 }
2548 
2549 static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
2550 {
2551 	if (dev->netdev_ops->ndo_set_rx_headroom)
2552 		dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2553 }
2554 
2555 /* set the device rx headroom to the dev's default */
2556 static inline void netdev_reset_rx_headroom(struct net_device *dev)
2557 {
2558 	netdev_set_rx_headroom(dev, -1);
2559 }
2560 
2561 static inline void *netdev_get_ml_priv(struct net_device *dev,
2562 				       enum netdev_ml_priv_type type)
2563 {
2564 	if (dev->ml_priv_type != type)
2565 		return NULL;
2566 
2567 	return dev->ml_priv;
2568 }
2569 
2570 static inline void netdev_set_ml_priv(struct net_device *dev,
2571 				      void *ml_priv,
2572 				      enum netdev_ml_priv_type type)
2573 {
2574 	WARN(dev->ml_priv_type && dev->ml_priv_type != type,
2575 	     "Overwriting already set ml_priv_type (%u) with different ml_priv_type (%u)!\n",
2576 	     dev->ml_priv_type, type);
2577 	WARN(!dev->ml_priv_type && dev->ml_priv,
2578 	     "Overwriting already set ml_priv and ml_priv_type is ML_PRIV_NONE!\n");
2579 
2580 	dev->ml_priv = ml_priv;
2581 	dev->ml_priv_type = type;
2582 }
2583 
2584 /*
2585  * Net namespace inlines
2586  */
2587 static inline
2588 struct net *dev_net(const struct net_device *dev)
2589 {
2590 	return read_pnet(&dev->nd_net);
2591 }
2592 
2593 static inline
2594 void dev_net_set(struct net_device *dev, struct net *net)
2595 {
2596 	write_pnet(&dev->nd_net, net);
2597 }
2598 
2599 /**
2600  *	netdev_priv - access network device private data
2601  *	@dev: network device
2602  *
2603  * Get network device private data
2604  */
2605 static inline void *netdev_priv(const struct net_device *dev)
2606 {
2607 	return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
2608 }
2609 
2610 /* Set the sysfs physical device reference for the network logical device
2611  * if set prior to registration will cause a symlink during initialization.
2612  */
2613 #define SET_NETDEV_DEV(net, pdev)	((net)->dev.parent = (pdev))
2614 
2615 /* Set the sysfs device type for the network logical device to allow
2616  * fine-grained identification of different network device types. For
2617  * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
2618  */
2619 #define SET_NETDEV_DEVTYPE(net, devtype)	((net)->dev.type = (devtype))
2620 
2621 /* Default NAPI poll() weight
2622  * Device drivers are strongly advised to not use bigger value
2623  */
2624 #define NAPI_POLL_WEIGHT 64
2625 
2626 void netif_napi_add_weight(struct net_device *dev, struct napi_struct *napi,
2627 			   int (*poll)(struct napi_struct *, int), int weight);
2628 
2629 /**
2630  * netif_napi_add() - initialize a NAPI context
2631  * @dev:  network device
2632  * @napi: NAPI context
2633  * @poll: polling function
2634  *
2635  * netif_napi_add() must be used to initialize a NAPI context prior to calling
2636  * *any* of the other NAPI-related functions.
2637  */
2638 static inline void
2639 netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2640 	       int (*poll)(struct napi_struct *, int))
2641 {
2642 	netif_napi_add_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2643 }
2644 
2645 static inline void
2646 netif_napi_add_tx_weight(struct net_device *dev,
2647 			 struct napi_struct *napi,
2648 			 int (*poll)(struct napi_struct *, int),
2649 			 int weight)
2650 {
2651 	set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2652 	netif_napi_add_weight(dev, napi, poll, weight);
2653 }
2654 
2655 /**
2656  * netif_napi_add_tx() - initialize a NAPI context to be used for Tx only
2657  * @dev:  network device
2658  * @napi: NAPI context
2659  * @poll: polling function
2660  *
2661  * This variant of netif_napi_add() should be used from drivers using NAPI
2662  * to exclusively poll a TX queue.
2663  * This will avoid we add it into napi_hash[], thus polluting this hash table.
2664  */
2665 static inline void netif_napi_add_tx(struct net_device *dev,
2666 				     struct napi_struct *napi,
2667 				     int (*poll)(struct napi_struct *, int))
2668 {
2669 	netif_napi_add_tx_weight(dev, napi, poll, NAPI_POLL_WEIGHT);
2670 }
2671 
2672 /**
2673  *  __netif_napi_del - remove a NAPI context
2674  *  @napi: NAPI context
2675  *
2676  * Warning: caller must observe RCU grace period before freeing memory
2677  * containing @napi. Drivers might want to call this helper to combine
2678  * all the needed RCU grace periods into a single one.
2679  */
2680 void __netif_napi_del(struct napi_struct *napi);
2681 
2682 /**
2683  *  netif_napi_del - remove a NAPI context
2684  *  @napi: NAPI context
2685  *
2686  *  netif_napi_del() removes a NAPI context from the network device NAPI list
2687  */
2688 static inline void netif_napi_del(struct napi_struct *napi)
2689 {
2690 	__netif_napi_del(napi);
2691 	synchronize_net();
2692 }
2693 
2694 struct packet_type {
2695 	__be16			type;	/* This is really htons(ether_type). */
2696 	bool			ignore_outgoing;
2697 	struct net_device	*dev;	/* NULL is wildcarded here	     */
2698 	netdevice_tracker	dev_tracker;
2699 	int			(*func) (struct sk_buff *,
2700 					 struct net_device *,
2701 					 struct packet_type *,
2702 					 struct net_device *);
2703 	void			(*list_func) (struct list_head *,
2704 					      struct packet_type *,
2705 					      struct net_device *);
2706 	bool			(*id_match)(struct packet_type *ptype,
2707 					    struct sock *sk);
2708 	struct net		*af_packet_net;
2709 	void			*af_packet_priv;
2710 	struct list_head	list;
2711 };
2712 
2713 struct offload_callbacks {
2714 	struct sk_buff		*(*gso_segment)(struct sk_buff *skb,
2715 						netdev_features_t features);
2716 	struct sk_buff		*(*gro_receive)(struct list_head *head,
2717 						struct sk_buff *skb);
2718 	int			(*gro_complete)(struct sk_buff *skb, int nhoff);
2719 };
2720 
2721 struct packet_offload {
2722 	__be16			 type;	/* This is really htons(ether_type). */
2723 	u16			 priority;
2724 	struct offload_callbacks callbacks;
2725 	struct list_head	 list;
2726 };
2727 
2728 /* often modified stats are per-CPU, other are shared (netdev->stats) */
2729 struct pcpu_sw_netstats {
2730 	u64_stats_t		rx_packets;
2731 	u64_stats_t		rx_bytes;
2732 	u64_stats_t		tx_packets;
2733 	u64_stats_t		tx_bytes;
2734 	struct u64_stats_sync   syncp;
2735 } __aligned(4 * sizeof(u64));
2736 
2737 struct pcpu_lstats {
2738 	u64_stats_t packets;
2739 	u64_stats_t bytes;
2740 	struct u64_stats_sync syncp;
2741 } __aligned(2 * sizeof(u64));
2742 
2743 void dev_lstats_read(struct net_device *dev, u64 *packets, u64 *bytes);
2744 
2745 static inline void dev_sw_netstats_rx_add(struct net_device *dev, unsigned int len)
2746 {
2747 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2748 
2749 	u64_stats_update_begin(&tstats->syncp);
2750 	u64_stats_add(&tstats->rx_bytes, len);
2751 	u64_stats_inc(&tstats->rx_packets);
2752 	u64_stats_update_end(&tstats->syncp);
2753 }
2754 
2755 static inline void dev_sw_netstats_tx_add(struct net_device *dev,
2756 					  unsigned int packets,
2757 					  unsigned int len)
2758 {
2759 	struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
2760 
2761 	u64_stats_update_begin(&tstats->syncp);
2762 	u64_stats_add(&tstats->tx_bytes, len);
2763 	u64_stats_add(&tstats->tx_packets, packets);
2764 	u64_stats_update_end(&tstats->syncp);
2765 }
2766 
2767 static inline void dev_lstats_add(struct net_device *dev, unsigned int len)
2768 {
2769 	struct pcpu_lstats *lstats = this_cpu_ptr(dev->lstats);
2770 
2771 	u64_stats_update_begin(&lstats->syncp);
2772 	u64_stats_add(&lstats->bytes, len);
2773 	u64_stats_inc(&lstats->packets);
2774 	u64_stats_update_end(&lstats->syncp);
2775 }
2776 
2777 #define __netdev_alloc_pcpu_stats(type, gfp)				\
2778 ({									\
2779 	typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2780 	if (pcpu_stats)	{						\
2781 		int __cpu;						\
2782 		for_each_possible_cpu(__cpu) {				\
2783 			typeof(type) *stat;				\
2784 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2785 			u64_stats_init(&stat->syncp);			\
2786 		}							\
2787 	}								\
2788 	pcpu_stats;							\
2789 })
2790 
2791 #define netdev_alloc_pcpu_stats(type)					\
2792 	__netdev_alloc_pcpu_stats(type, GFP_KERNEL)
2793 
2794 #define devm_netdev_alloc_pcpu_stats(dev, type)				\
2795 ({									\
2796 	typeof(type) __percpu *pcpu_stats = devm_alloc_percpu(dev, type);\
2797 	if (pcpu_stats) {						\
2798 		int __cpu;						\
2799 		for_each_possible_cpu(__cpu) {				\
2800 			typeof(type) *stat;				\
2801 			stat = per_cpu_ptr(pcpu_stats, __cpu);		\
2802 			u64_stats_init(&stat->syncp);			\
2803 		}							\
2804 	}								\
2805 	pcpu_stats;							\
2806 })
2807 
2808 enum netdev_lag_tx_type {
2809 	NETDEV_LAG_TX_TYPE_UNKNOWN,
2810 	NETDEV_LAG_TX_TYPE_RANDOM,
2811 	NETDEV_LAG_TX_TYPE_BROADCAST,
2812 	NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2813 	NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2814 	NETDEV_LAG_TX_TYPE_HASH,
2815 };
2816 
2817 enum netdev_lag_hash {
2818 	NETDEV_LAG_HASH_NONE,
2819 	NETDEV_LAG_HASH_L2,
2820 	NETDEV_LAG_HASH_L34,
2821 	NETDEV_LAG_HASH_L23,
2822 	NETDEV_LAG_HASH_E23,
2823 	NETDEV_LAG_HASH_E34,
2824 	NETDEV_LAG_HASH_VLAN_SRCMAC,
2825 	NETDEV_LAG_HASH_UNKNOWN,
2826 };
2827 
2828 struct netdev_lag_upper_info {
2829 	enum netdev_lag_tx_type tx_type;
2830 	enum netdev_lag_hash hash_type;
2831 };
2832 
2833 struct netdev_lag_lower_state_info {
2834 	u8 link_up : 1,
2835 	   tx_enabled : 1;
2836 };
2837 
2838 #include <linux/notifier.h>
2839 
2840 /* netdevice notifier chain. Please remember to update netdev_cmd_to_name()
2841  * and the rtnetlink notification exclusion list in rtnetlink_event() when
2842  * adding new types.
2843  */
2844 enum netdev_cmd {
2845 	NETDEV_UP	= 1,	/* For now you can't veto a device up/down */
2846 	NETDEV_DOWN,
2847 	NETDEV_REBOOT,		/* Tell a protocol stack a network interface
2848 				   detected a hardware crash and restarted
2849 				   - we can use this eg to kick tcp sessions
2850 				   once done */
2851 	NETDEV_CHANGE,		/* Notify device state change */
2852 	NETDEV_REGISTER,
2853 	NETDEV_UNREGISTER,
2854 	NETDEV_CHANGEMTU,	/* notify after mtu change happened */
2855 	NETDEV_CHANGEADDR,	/* notify after the address change */
2856 	NETDEV_PRE_CHANGEADDR,	/* notify before the address change */
2857 	NETDEV_GOING_DOWN,
2858 	NETDEV_CHANGENAME,
2859 	NETDEV_FEAT_CHANGE,
2860 	NETDEV_BONDING_FAILOVER,
2861 	NETDEV_PRE_UP,
2862 	NETDEV_PRE_TYPE_CHANGE,
2863 	NETDEV_POST_TYPE_CHANGE,
2864 	NETDEV_POST_INIT,
2865 	NETDEV_PRE_UNINIT,
2866 	NETDEV_RELEASE,
2867 	NETDEV_NOTIFY_PEERS,
2868 	NETDEV_JOIN,
2869 	NETDEV_CHANGEUPPER,
2870 	NETDEV_RESEND_IGMP,
2871 	NETDEV_PRECHANGEMTU,	/* notify before mtu change happened */
2872 	NETDEV_CHANGEINFODATA,
2873 	NETDEV_BONDING_INFO,
2874 	NETDEV_PRECHANGEUPPER,
2875 	NETDEV_CHANGELOWERSTATE,
2876 	NETDEV_UDP_TUNNEL_PUSH_INFO,
2877 	NETDEV_UDP_TUNNEL_DROP_INFO,
2878 	NETDEV_CHANGE_TX_QUEUE_LEN,
2879 	NETDEV_CVLAN_FILTER_PUSH_INFO,
2880 	NETDEV_CVLAN_FILTER_DROP_INFO,
2881 	NETDEV_SVLAN_FILTER_PUSH_INFO,
2882 	NETDEV_SVLAN_FILTER_DROP_INFO,
2883 	NETDEV_OFFLOAD_XSTATS_ENABLE,
2884 	NETDEV_OFFLOAD_XSTATS_DISABLE,
2885 	NETDEV_OFFLOAD_XSTATS_REPORT_USED,
2886 	NETDEV_OFFLOAD_XSTATS_REPORT_DELTA,
2887 	NETDEV_XDP_FEAT_CHANGE,
2888 };
2889 const char *netdev_cmd_to_name(enum netdev_cmd cmd);
2890 
2891 int register_netdevice_notifier(struct notifier_block *nb);
2892 int unregister_netdevice_notifier(struct notifier_block *nb);
2893 int register_netdevice_notifier_net(struct net *net, struct notifier_block *nb);
2894 int unregister_netdevice_notifier_net(struct net *net,
2895 				      struct notifier_block *nb);
2896 int register_netdevice_notifier_dev_net(struct net_device *dev,
2897 					struct notifier_block *nb,
2898 					struct netdev_net_notifier *nn);
2899 int unregister_netdevice_notifier_dev_net(struct net_device *dev,
2900 					  struct notifier_block *nb,
2901 					  struct netdev_net_notifier *nn);
2902 
2903 struct netdev_notifier_info {
2904 	struct net_device	*dev;
2905 	struct netlink_ext_ack	*extack;
2906 };
2907 
2908 struct netdev_notifier_info_ext {
2909 	struct netdev_notifier_info info; /* must be first */
2910 	union {
2911 		u32 mtu;
2912 	} ext;
2913 };
2914 
2915 struct netdev_notifier_change_info {
2916 	struct netdev_notifier_info info; /* must be first */
2917 	unsigned int flags_changed;
2918 };
2919 
2920 struct netdev_notifier_changeupper_info {
2921 	struct netdev_notifier_info info; /* must be first */
2922 	struct net_device *upper_dev; /* new upper dev */
2923 	bool master; /* is upper dev master */
2924 	bool linking; /* is the notification for link or unlink */
2925 	void *upper_info; /* upper dev info */
2926 };
2927 
2928 struct netdev_notifier_changelowerstate_info {
2929 	struct netdev_notifier_info info; /* must be first */
2930 	void *lower_state_info; /* is lower dev state */
2931 };
2932 
2933 struct netdev_notifier_pre_changeaddr_info {
2934 	struct netdev_notifier_info info; /* must be first */
2935 	const unsigned char *dev_addr;
2936 };
2937 
2938 enum netdev_offload_xstats_type {
2939 	NETDEV_OFFLOAD_XSTATS_TYPE_L3 = 1,
2940 };
2941 
2942 struct netdev_notifier_offload_xstats_info {
2943 	struct netdev_notifier_info info; /* must be first */
2944 	enum netdev_offload_xstats_type type;
2945 
2946 	union {
2947 		/* NETDEV_OFFLOAD_XSTATS_REPORT_DELTA */
2948 		struct netdev_notifier_offload_xstats_rd *report_delta;
2949 		/* NETDEV_OFFLOAD_XSTATS_REPORT_USED */
2950 		struct netdev_notifier_offload_xstats_ru *report_used;
2951 	};
2952 };
2953 
2954 int netdev_offload_xstats_enable(struct net_device *dev,
2955 				 enum netdev_offload_xstats_type type,
2956 				 struct netlink_ext_ack *extack);
2957 int netdev_offload_xstats_disable(struct net_device *dev,
2958 				  enum netdev_offload_xstats_type type);
2959 bool netdev_offload_xstats_enabled(const struct net_device *dev,
2960 				   enum netdev_offload_xstats_type type);
2961 int netdev_offload_xstats_get(struct net_device *dev,
2962 			      enum netdev_offload_xstats_type type,
2963 			      struct rtnl_hw_stats64 *stats, bool *used,
2964 			      struct netlink_ext_ack *extack);
2965 void
2966 netdev_offload_xstats_report_delta(struct netdev_notifier_offload_xstats_rd *rd,
2967 				   const struct rtnl_hw_stats64 *stats);
2968 void
2969 netdev_offload_xstats_report_used(struct netdev_notifier_offload_xstats_ru *ru);
2970 void netdev_offload_xstats_push_delta(struct net_device *dev,
2971 				      enum netdev_offload_xstats_type type,
2972 				      const struct rtnl_hw_stats64 *stats);
2973 
2974 static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2975 					     struct net_device *dev)
2976 {
2977 	info->dev = dev;
2978 	info->extack = NULL;
2979 }
2980 
2981 static inline struct net_device *
2982 netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2983 {
2984 	return info->dev;
2985 }
2986 
2987 static inline struct netlink_ext_ack *
2988 netdev_notifier_info_to_extack(const struct netdev_notifier_info *info)
2989 {
2990 	return info->extack;
2991 }
2992 
2993 int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
2994 int call_netdevice_notifiers_info(unsigned long val,
2995 				  struct netdev_notifier_info *info);
2996 
2997 extern rwlock_t				dev_base_lock;		/* Device list lock */
2998 
2999 #define for_each_netdev(net, d)		\
3000 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
3001 #define for_each_netdev_reverse(net, d)	\
3002 		list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
3003 #define for_each_netdev_rcu(net, d)		\
3004 		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
3005 #define for_each_netdev_safe(net, d, n)	\
3006 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
3007 #define for_each_netdev_continue(net, d)		\
3008 		list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
3009 #define for_each_netdev_continue_reverse(net, d)		\
3010 		list_for_each_entry_continue_reverse(d, &(net)->dev_base_head, \
3011 						     dev_list)
3012 #define for_each_netdev_continue_rcu(net, d)		\
3013 	list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
3014 #define for_each_netdev_in_bond_rcu(bond, slave)	\
3015 		for_each_netdev_rcu(&init_net, slave)	\
3016 			if (netdev_master_upper_dev_get_rcu(slave) == (bond))
3017 #define net_device_entry(lh)	list_entry(lh, struct net_device, dev_list)
3018 
3019 #define for_each_netdev_dump(net, d, ifindex)				\
3020 	xa_for_each_start(&(net)->dev_by_index, (ifindex), (d), (ifindex))
3021 
3022 static inline struct net_device *next_net_device(struct net_device *dev)
3023 {
3024 	struct list_head *lh;
3025 	struct net *net;
3026 
3027 	net = dev_net(dev);
3028 	lh = dev->dev_list.next;
3029 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3030 }
3031 
3032 static inline struct net_device *next_net_device_rcu(struct net_device *dev)
3033 {
3034 	struct list_head *lh;
3035 	struct net *net;
3036 
3037 	net = dev_net(dev);
3038 	lh = rcu_dereference(list_next_rcu(&dev->dev_list));
3039 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3040 }
3041 
3042 static inline struct net_device *first_net_device(struct net *net)
3043 {
3044 	return list_empty(&net->dev_base_head) ? NULL :
3045 		net_device_entry(net->dev_base_head.next);
3046 }
3047 
3048 static inline struct net_device *first_net_device_rcu(struct net *net)
3049 {
3050 	struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
3051 
3052 	return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
3053 }
3054 
3055 int netdev_boot_setup_check(struct net_device *dev);
3056 struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
3057 				       const char *hwaddr);
3058 struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
3059 void dev_add_pack(struct packet_type *pt);
3060 void dev_remove_pack(struct packet_type *pt);
3061 void __dev_remove_pack(struct packet_type *pt);
3062 void dev_add_offload(struct packet_offload *po);
3063 void dev_remove_offload(struct packet_offload *po);
3064 
3065 int dev_get_iflink(const struct net_device *dev);
3066 int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
3067 int dev_fill_forward_path(const struct net_device *dev, const u8 *daddr,
3068 			  struct net_device_path_stack *stack);
3069 struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
3070 				      unsigned short mask);
3071 struct net_device *dev_get_by_name(struct net *net, const char *name);
3072 struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
3073 struct net_device *__dev_get_by_name(struct net *net, const char *name);
3074 bool netdev_name_in_use(struct net *net, const char *name);
3075 int dev_alloc_name(struct net_device *dev, const char *name);
3076 int dev_open(struct net_device *dev, struct netlink_ext_ack *extack);
3077 void dev_close(struct net_device *dev);
3078 void dev_close_many(struct list_head *head, bool unlink);
3079 void dev_disable_lro(struct net_device *dev);
3080 int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
3081 u16 dev_pick_tx_zero(struct net_device *dev, struct sk_buff *skb,
3082 		     struct net_device *sb_dev);
3083 u16 dev_pick_tx_cpu_id(struct net_device *dev, struct sk_buff *skb,
3084 		       struct net_device *sb_dev);
3085 
3086 int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev);
3087 int __dev_direct_xmit(struct sk_buff *skb, u16 queue_id);
3088 
3089 static inline int dev_queue_xmit(struct sk_buff *skb)
3090 {
3091 	return __dev_queue_xmit(skb, NULL);
3092 }
3093 
3094 static inline int dev_queue_xmit_accel(struct sk_buff *skb,
3095 				       struct net_device *sb_dev)
3096 {
3097 	return __dev_queue_xmit(skb, sb_dev);
3098 }
3099 
3100 static inline int dev_direct_xmit(struct sk_buff *skb, u16 queue_id)
3101 {
3102 	int ret;
3103 
3104 	ret = __dev_direct_xmit(skb, queue_id);
3105 	if (!dev_xmit_complete(ret))
3106 		kfree_skb(skb);
3107 	return ret;
3108 }
3109 
3110 int register_netdevice(struct net_device *dev);
3111 void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
3112 void unregister_netdevice_many(struct list_head *head);
3113 static inline void unregister_netdevice(struct net_device *dev)
3114 {
3115 	unregister_netdevice_queue(dev, NULL);
3116 }
3117 
3118 int netdev_refcnt_read(const struct net_device *dev);
3119 void free_netdev(struct net_device *dev);
3120 void netdev_freemem(struct net_device *dev);
3121 int init_dummy_netdev(struct net_device *dev);
3122 
3123 struct net_device *netdev_get_xmit_slave(struct net_device *dev,
3124 					 struct sk_buff *skb,
3125 					 bool all_slaves);
3126 struct net_device *netdev_sk_get_lowest_dev(struct net_device *dev,
3127 					    struct sock *sk);
3128 struct net_device *dev_get_by_index(struct net *net, int ifindex);
3129 struct net_device *__dev_get_by_index(struct net *net, int ifindex);
3130 struct net_device *netdev_get_by_index(struct net *net, int ifindex,
3131 				       netdevice_tracker *tracker, gfp_t gfp);
3132 struct net_device *netdev_get_by_name(struct net *net, const char *name,
3133 				      netdevice_tracker *tracker, gfp_t gfp);
3134 struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
3135 struct net_device *dev_get_by_napi_id(unsigned int napi_id);
3136 
3137 static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
3138 				  unsigned short type,
3139 				  const void *daddr, const void *saddr,
3140 				  unsigned int len)
3141 {
3142 	if (!dev->header_ops || !dev->header_ops->create)
3143 		return 0;
3144 
3145 	return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
3146 }
3147 
3148 static inline int dev_parse_header(const struct sk_buff *skb,
3149 				   unsigned char *haddr)
3150 {
3151 	const struct net_device *dev = skb->dev;
3152 
3153 	if (!dev->header_ops || !dev->header_ops->parse)
3154 		return 0;
3155 	return dev->header_ops->parse(skb, haddr);
3156 }
3157 
3158 static inline __be16 dev_parse_header_protocol(const struct sk_buff *skb)
3159 {
3160 	const struct net_device *dev = skb->dev;
3161 
3162 	if (!dev->header_ops || !dev->header_ops->parse_protocol)
3163 		return 0;
3164 	return dev->header_ops->parse_protocol(skb);
3165 }
3166 
3167 /* ll_header must have at least hard_header_len allocated */
3168 static inline bool dev_validate_header(const struct net_device *dev,
3169 				       char *ll_header, int len)
3170 {
3171 	if (likely(len >= dev->hard_header_len))
3172 		return true;
3173 	if (len < dev->min_header_len)
3174 		return false;
3175 
3176 	if (capable(CAP_SYS_RAWIO)) {
3177 		memset(ll_header + len, 0, dev->hard_header_len - len);
3178 		return true;
3179 	}
3180 
3181 	if (dev->header_ops && dev->header_ops->validate)
3182 		return dev->header_ops->validate(ll_header, len);
3183 
3184 	return false;
3185 }
3186 
3187 static inline bool dev_has_header(const struct net_device *dev)
3188 {
3189 	return dev->header_ops && dev->header_ops->create;
3190 }
3191 
3192 /*
3193  * Incoming packets are placed on per-CPU queues
3194  */
3195 struct softnet_data {
3196 	struct list_head	poll_list;
3197 	struct sk_buff_head	process_queue;
3198 
3199 	/* stats */
3200 	unsigned int		processed;
3201 	unsigned int		time_squeeze;
3202 #ifdef CONFIG_RPS
3203 	struct softnet_data	*rps_ipi_list;
3204 #endif
3205 
3206 	bool			in_net_rx_action;
3207 	bool			in_napi_threaded_poll;
3208 
3209 #ifdef CONFIG_NET_FLOW_LIMIT
3210 	struct sd_flow_limit __rcu *flow_limit;
3211 #endif
3212 	struct Qdisc		*output_queue;
3213 	struct Qdisc		**output_queue_tailp;
3214 	struct sk_buff		*completion_queue;
3215 #ifdef CONFIG_XFRM_OFFLOAD
3216 	struct sk_buff_head	xfrm_backlog;
3217 #endif
3218 	/* written and read only by owning cpu: */
3219 	struct {
3220 		u16 recursion;
3221 		u8  more;
3222 #ifdef CONFIG_NET_EGRESS
3223 		u8  skip_txqueue;
3224 #endif
3225 	} xmit;
3226 #ifdef CONFIG_RPS
3227 	/* input_queue_head should be written by cpu owning this struct,
3228 	 * and only read by other cpus. Worth using a cache line.
3229 	 */
3230 	unsigned int		input_queue_head ____cacheline_aligned_in_smp;
3231 
3232 	/* Elements below can be accessed between CPUs for RPS/RFS */
3233 	call_single_data_t	csd ____cacheline_aligned_in_smp;
3234 	struct softnet_data	*rps_ipi_next;
3235 	unsigned int		cpu;
3236 	unsigned int		input_queue_tail;
3237 #endif
3238 	unsigned int		received_rps;
3239 	unsigned int		dropped;
3240 	struct sk_buff_head	input_pkt_queue;
3241 	struct napi_struct	backlog;
3242 
3243 	/* Another possibly contended cache line */
3244 	spinlock_t		defer_lock ____cacheline_aligned_in_smp;
3245 	int			defer_count;
3246 	int			defer_ipi_scheduled;
3247 	struct sk_buff		*defer_list;
3248 	call_single_data_t	defer_csd;
3249 };
3250 
3251 static inline void input_queue_head_incr(struct softnet_data *sd)
3252 {
3253 #ifdef CONFIG_RPS
3254 	sd->input_queue_head++;
3255 #endif
3256 }
3257 
3258 static inline void input_queue_tail_incr_save(struct softnet_data *sd,
3259 					      unsigned int *qtail)
3260 {
3261 #ifdef CONFIG_RPS
3262 	*qtail = ++sd->input_queue_tail;
3263 #endif
3264 }
3265 
3266 DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
3267 
3268 static inline int dev_recursion_level(void)
3269 {
3270 	return this_cpu_read(softnet_data.xmit.recursion);
3271 }
3272 
3273 #define XMIT_RECURSION_LIMIT	8
3274 static inline bool dev_xmit_recursion(void)
3275 {
3276 	return unlikely(__this_cpu_read(softnet_data.xmit.recursion) >
3277 			XMIT_RECURSION_LIMIT);
3278 }
3279 
3280 static inline void dev_xmit_recursion_inc(void)
3281 {
3282 	__this_cpu_inc(softnet_data.xmit.recursion);
3283 }
3284 
3285 static inline void dev_xmit_recursion_dec(void)
3286 {
3287 	__this_cpu_dec(softnet_data.xmit.recursion);
3288 }
3289 
3290 void __netif_schedule(struct Qdisc *q);
3291 void netif_schedule_queue(struct netdev_queue *txq);
3292 
3293 static inline void netif_tx_schedule_all(struct net_device *dev)
3294 {
3295 	unsigned int i;
3296 
3297 	for (i = 0; i < dev->num_tx_queues; i++)
3298 		netif_schedule_queue(netdev_get_tx_queue(dev, i));
3299 }
3300 
3301 static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
3302 {
3303 	clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3304 }
3305 
3306 /**
3307  *	netif_start_queue - allow transmit
3308  *	@dev: network device
3309  *
3310  *	Allow upper layers to call the device hard_start_xmit routine.
3311  */
3312 static inline void netif_start_queue(struct net_device *dev)
3313 {
3314 	netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
3315 }
3316 
3317 static inline void netif_tx_start_all_queues(struct net_device *dev)
3318 {
3319 	unsigned int i;
3320 
3321 	for (i = 0; i < dev->num_tx_queues; i++) {
3322 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3323 		netif_tx_start_queue(txq);
3324 	}
3325 }
3326 
3327 void netif_tx_wake_queue(struct netdev_queue *dev_queue);
3328 
3329 /**
3330  *	netif_wake_queue - restart transmit
3331  *	@dev: network device
3332  *
3333  *	Allow upper layers to call the device hard_start_xmit routine.
3334  *	Used for flow control when transmit resources are available.
3335  */
3336 static inline void netif_wake_queue(struct net_device *dev)
3337 {
3338 	netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
3339 }
3340 
3341 static inline void netif_tx_wake_all_queues(struct net_device *dev)
3342 {
3343 	unsigned int i;
3344 
3345 	for (i = 0; i < dev->num_tx_queues; i++) {
3346 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3347 		netif_tx_wake_queue(txq);
3348 	}
3349 }
3350 
3351 static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
3352 {
3353 	/* Must be an atomic op see netif_txq_try_stop() */
3354 	set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3355 }
3356 
3357 /**
3358  *	netif_stop_queue - stop transmitted packets
3359  *	@dev: network device
3360  *
3361  *	Stop upper layers calling the device hard_start_xmit routine.
3362  *	Used for flow control when transmit resources are unavailable.
3363  */
3364 static inline void netif_stop_queue(struct net_device *dev)
3365 {
3366 	netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
3367 }
3368 
3369 void netif_tx_stop_all_queues(struct net_device *dev);
3370 
3371 static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
3372 {
3373 	return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
3374 }
3375 
3376 /**
3377  *	netif_queue_stopped - test if transmit queue is flowblocked
3378  *	@dev: network device
3379  *
3380  *	Test if transmit queue on device is currently unable to send.
3381  */
3382 static inline bool netif_queue_stopped(const struct net_device *dev)
3383 {
3384 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
3385 }
3386 
3387 static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
3388 {
3389 	return dev_queue->state & QUEUE_STATE_ANY_XOFF;
3390 }
3391 
3392 static inline bool
3393 netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
3394 {
3395 	return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
3396 }
3397 
3398 static inline bool
3399 netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
3400 {
3401 	return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
3402 }
3403 
3404 /**
3405  *	netdev_queue_set_dql_min_limit - set dql minimum limit
3406  *	@dev_queue: pointer to transmit queue
3407  *	@min_limit: dql minimum limit
3408  *
3409  * Forces xmit_more() to return true until the minimum threshold
3410  * defined by @min_limit is reached (or until the tx queue is
3411  * empty). Warning: to be use with care, misuse will impact the
3412  * latency.
3413  */
3414 static inline void netdev_queue_set_dql_min_limit(struct netdev_queue *dev_queue,
3415 						  unsigned int min_limit)
3416 {
3417 #ifdef CONFIG_BQL
3418 	dev_queue->dql.min_limit = min_limit;
3419 #endif
3420 }
3421 
3422 /**
3423  *	netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
3424  *	@dev_queue: pointer to transmit queue
3425  *
3426  * BQL enabled drivers might use this helper in their ndo_start_xmit(),
3427  * to give appropriate hint to the CPU.
3428  */
3429 static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
3430 {
3431 #ifdef CONFIG_BQL
3432 	prefetchw(&dev_queue->dql.num_queued);
3433 #endif
3434 }
3435 
3436 /**
3437  *	netdev_txq_bql_complete_prefetchw - prefetch bql data for write
3438  *	@dev_queue: pointer to transmit queue
3439  *
3440  * BQL enabled drivers might use this helper in their TX completion path,
3441  * to give appropriate hint to the CPU.
3442  */
3443 static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
3444 {
3445 #ifdef CONFIG_BQL
3446 	prefetchw(&dev_queue->dql.limit);
3447 #endif
3448 }
3449 
3450 /**
3451  *	netdev_tx_sent_queue - report the number of bytes queued to a given tx queue
3452  *	@dev_queue: network device queue
3453  *	@bytes: number of bytes queued to the device queue
3454  *
3455  *	Report the number of bytes queued for sending/completion to the network
3456  *	device hardware queue. @bytes should be a good approximation and should
3457  *	exactly match netdev_completed_queue() @bytes.
3458  *	This is typically called once per packet, from ndo_start_xmit().
3459  */
3460 static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3461 					unsigned int bytes)
3462 {
3463 #ifdef CONFIG_BQL
3464 	dql_queued(&dev_queue->dql, bytes);
3465 
3466 	if (likely(dql_avail(&dev_queue->dql) >= 0))
3467 		return;
3468 
3469 	set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3470 
3471 	/*
3472 	 * The XOFF flag must be set before checking the dql_avail below,
3473 	 * because in netdev_tx_completed_queue we update the dql_completed
3474 	 * before checking the XOFF flag.
3475 	 */
3476 	smp_mb();
3477 
3478 	/* check again in case another CPU has just made room avail */
3479 	if (unlikely(dql_avail(&dev_queue->dql) >= 0))
3480 		clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
3481 #endif
3482 }
3483 
3484 /* Variant of netdev_tx_sent_queue() for drivers that are aware
3485  * that they should not test BQL status themselves.
3486  * We do want to change __QUEUE_STATE_STACK_XOFF only for the last
3487  * skb of a batch.
3488  * Returns true if the doorbell must be used to kick the NIC.
3489  */
3490 static inline bool __netdev_tx_sent_queue(struct netdev_queue *dev_queue,
3491 					  unsigned int bytes,
3492 					  bool xmit_more)
3493 {
3494 	if (xmit_more) {
3495 #ifdef CONFIG_BQL
3496 		dql_queued(&dev_queue->dql, bytes);
3497 #endif
3498 		return netif_tx_queue_stopped(dev_queue);
3499 	}
3500 	netdev_tx_sent_queue(dev_queue, bytes);
3501 	return true;
3502 }
3503 
3504 /**
3505  *	netdev_sent_queue - report the number of bytes queued to hardware
3506  *	@dev: network device
3507  *	@bytes: number of bytes queued to the hardware device queue
3508  *
3509  *	Report the number of bytes queued for sending/completion to the network
3510  *	device hardware queue#0. @bytes should be a good approximation and should
3511  *	exactly match netdev_completed_queue() @bytes.
3512  *	This is typically called once per packet, from ndo_start_xmit().
3513  */
3514 static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
3515 {
3516 	netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
3517 }
3518 
3519 static inline bool __netdev_sent_queue(struct net_device *dev,
3520 				       unsigned int bytes,
3521 				       bool xmit_more)
3522 {
3523 	return __netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes,
3524 				      xmit_more);
3525 }
3526 
3527 /**
3528  *	netdev_tx_completed_queue - report number of packets/bytes at TX completion.
3529  *	@dev_queue: network device queue
3530  *	@pkts: number of packets (currently ignored)
3531  *	@bytes: number of bytes dequeued from the device queue
3532  *
3533  *	Must be called at most once per TX completion round (and not per
3534  *	individual packet), so that BQL can adjust its limits appropriately.
3535  */
3536 static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
3537 					     unsigned int pkts, unsigned int bytes)
3538 {
3539 #ifdef CONFIG_BQL
3540 	if (unlikely(!bytes))
3541 		return;
3542 
3543 	dql_completed(&dev_queue->dql, bytes);
3544 
3545 	/*
3546 	 * Without the memory barrier there is a small possiblity that
3547 	 * netdev_tx_sent_queue will miss the update and cause the queue to
3548 	 * be stopped forever
3549 	 */
3550 	smp_mb(); /* NOTE: netdev_txq_completed_mb() assumes this exists */
3551 
3552 	if (unlikely(dql_avail(&dev_queue->dql) < 0))
3553 		return;
3554 
3555 	if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
3556 		netif_schedule_queue(dev_queue);
3557 #endif
3558 }
3559 
3560 /**
3561  * 	netdev_completed_queue - report bytes and packets completed by device
3562  * 	@dev: network device
3563  * 	@pkts: actual number of packets sent over the medium
3564  * 	@bytes: actual number of bytes sent over the medium
3565  *
3566  * 	Report the number of bytes and packets transmitted by the network device
3567  * 	hardware queue over the physical medium, @bytes must exactly match the
3568  * 	@bytes amount passed to netdev_sent_queue()
3569  */
3570 static inline void netdev_completed_queue(struct net_device *dev,
3571 					  unsigned int pkts, unsigned int bytes)
3572 {
3573 	netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3574 }
3575 
3576 static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3577 {
3578 #ifdef CONFIG_BQL
3579 	clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
3580 	dql_reset(&q->dql);
3581 #endif
3582 }
3583 
3584 /**
3585  * 	netdev_reset_queue - reset the packets and bytes count of a network device
3586  * 	@dev_queue: network device
3587  *
3588  * 	Reset the bytes and packet count of a network device and clear the
3589  * 	software flow control OFF bit for this network device
3590  */
3591 static inline void netdev_reset_queue(struct net_device *dev_queue)
3592 {
3593 	netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
3594 }
3595 
3596 /**
3597  * 	netdev_cap_txqueue - check if selected tx queue exceeds device queues
3598  * 	@dev: network device
3599  * 	@queue_index: given tx queue index
3600  *
3601  * 	Returns 0 if given tx queue index >= number of device tx queues,
3602  * 	otherwise returns the originally passed tx queue index.
3603  */
3604 static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3605 {
3606 	if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3607 		net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3608 				     dev->name, queue_index,
3609 				     dev->real_num_tx_queues);
3610 		return 0;
3611 	}
3612 
3613 	return queue_index;
3614 }
3615 
3616 /**
3617  *	netif_running - test if up
3618  *	@dev: network device
3619  *
3620  *	Test if the device has been brought up.
3621  */
3622 static inline bool netif_running(const struct net_device *dev)
3623 {
3624 	return test_bit(__LINK_STATE_START, &dev->state);
3625 }
3626 
3627 /*
3628  * Routines to manage the subqueues on a device.  We only need start,
3629  * stop, and a check if it's stopped.  All other device management is
3630  * done at the overall netdevice level.
3631  * Also test the device if we're multiqueue.
3632  */
3633 
3634 /**
3635  *	netif_start_subqueue - allow sending packets on subqueue
3636  *	@dev: network device
3637  *	@queue_index: sub queue index
3638  *
3639  * Start individual transmit queue of a device with multiple transmit queues.
3640  */
3641 static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3642 {
3643 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3644 
3645 	netif_tx_start_queue(txq);
3646 }
3647 
3648 /**
3649  *	netif_stop_subqueue - stop sending packets on subqueue
3650  *	@dev: network device
3651  *	@queue_index: sub queue index
3652  *
3653  * Stop individual transmit queue of a device with multiple transmit queues.
3654  */
3655 static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3656 {
3657 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3658 	netif_tx_stop_queue(txq);
3659 }
3660 
3661 /**
3662  *	__netif_subqueue_stopped - test status of subqueue
3663  *	@dev: network device
3664  *	@queue_index: sub queue index
3665  *
3666  * Check individual transmit queue of a device with multiple transmit queues.
3667  */
3668 static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3669 					    u16 queue_index)
3670 {
3671 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3672 
3673 	return netif_tx_queue_stopped(txq);
3674 }
3675 
3676 /**
3677  *	netif_subqueue_stopped - test status of subqueue
3678  *	@dev: network device
3679  *	@skb: sub queue buffer pointer
3680  *
3681  * Check individual transmit queue of a device with multiple transmit queues.
3682  */
3683 static inline bool netif_subqueue_stopped(const struct net_device *dev,
3684 					  struct sk_buff *skb)
3685 {
3686 	return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3687 }
3688 
3689 /**
3690  *	netif_wake_subqueue - allow sending packets on subqueue
3691  *	@dev: network device
3692  *	@queue_index: sub queue index
3693  *
3694  * Resume individual transmit queue of a device with multiple transmit queues.
3695  */
3696 static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3697 {
3698 	struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3699 
3700 	netif_tx_wake_queue(txq);
3701 }
3702 
3703 #ifdef CONFIG_XPS
3704 int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
3705 			u16 index);
3706 int __netif_set_xps_queue(struct net_device *dev, const unsigned long *mask,
3707 			  u16 index, enum xps_map_type type);
3708 
3709 /**
3710  *	netif_attr_test_mask - Test a CPU or Rx queue set in a mask
3711  *	@j: CPU/Rx queue index
3712  *	@mask: bitmask of all cpus/rx queues
3713  *	@nr_bits: number of bits in the bitmask
3714  *
3715  * Test if a CPU or Rx queue index is set in a mask of all CPU/Rx queues.
3716  */
3717 static inline bool netif_attr_test_mask(unsigned long j,
3718 					const unsigned long *mask,
3719 					unsigned int nr_bits)
3720 {
3721 	cpu_max_bits_warn(j, nr_bits);
3722 	return test_bit(j, mask);
3723 }
3724 
3725 /**
3726  *	netif_attr_test_online - Test for online CPU/Rx queue
3727  *	@j: CPU/Rx queue index
3728  *	@online_mask: bitmask for CPUs/Rx queues that are online
3729  *	@nr_bits: number of bits in the bitmask
3730  *
3731  * Returns true if a CPU/Rx queue is online.
3732  */
3733 static inline bool netif_attr_test_online(unsigned long j,
3734 					  const unsigned long *online_mask,
3735 					  unsigned int nr_bits)
3736 {
3737 	cpu_max_bits_warn(j, nr_bits);
3738 
3739 	if (online_mask)
3740 		return test_bit(j, online_mask);
3741 
3742 	return (j < nr_bits);
3743 }
3744 
3745 /**
3746  *	netif_attrmask_next - get the next CPU/Rx queue in a cpu/Rx queues mask
3747  *	@n: CPU/Rx queue index
3748  *	@srcp: the cpumask/Rx queue mask pointer
3749  *	@nr_bits: number of bits in the bitmask
3750  *
3751  * Returns >= nr_bits if no further CPUs/Rx queues set.
3752  */
3753 static inline unsigned int netif_attrmask_next(int n, const unsigned long *srcp,
3754 					       unsigned int nr_bits)
3755 {
3756 	/* -1 is a legal arg here. */
3757 	if (n != -1)
3758 		cpu_max_bits_warn(n, nr_bits);
3759 
3760 	if (srcp)
3761 		return find_next_bit(srcp, nr_bits, n + 1);
3762 
3763 	return n + 1;
3764 }
3765 
3766 /**
3767  *	netif_attrmask_next_and - get the next CPU/Rx queue in \*src1p & \*src2p
3768  *	@n: CPU/Rx queue index
3769  *	@src1p: the first CPUs/Rx queues mask pointer
3770  *	@src2p: the second CPUs/Rx queues mask pointer
3771  *	@nr_bits: number of bits in the bitmask
3772  *
3773  * Returns >= nr_bits if no further CPUs/Rx queues set in both.
3774  */
3775 static inline int netif_attrmask_next_and(int n, const unsigned long *src1p,
3776 					  const unsigned long *src2p,
3777 					  unsigned int nr_bits)
3778 {
3779 	/* -1 is a legal arg here. */
3780 	if (n != -1)
3781 		cpu_max_bits_warn(n, nr_bits);
3782 
3783 	if (src1p && src2p)
3784 		return find_next_and_bit(src1p, src2p, nr_bits, n + 1);
3785 	else if (src1p)
3786 		return find_next_bit(src1p, nr_bits, n + 1);
3787 	else if (src2p)
3788 		return find_next_bit(src2p, nr_bits, n + 1);
3789 
3790 	return n + 1;
3791 }
3792 #else
3793 static inline int netif_set_xps_queue(struct net_device *dev,
3794 				      const struct cpumask *mask,
3795 				      u16 index)
3796 {
3797 	return 0;
3798 }
3799 
3800 static inline int __netif_set_xps_queue(struct net_device *dev,
3801 					const unsigned long *mask,
3802 					u16 index, enum xps_map_type type)
3803 {
3804 	return 0;
3805 }
3806 #endif
3807 
3808 /**
3809  *	netif_is_multiqueue - test if device has multiple transmit queues
3810  *	@dev: network device
3811  *
3812  * Check if device has multiple transmit queues
3813  */
3814 static inline bool netif_is_multiqueue(const struct net_device *dev)
3815 {
3816 	return dev->num_tx_queues > 1;
3817 }
3818 
3819 int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
3820 
3821 #ifdef CONFIG_SYSFS
3822 int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
3823 #else
3824 static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3825 						unsigned int rxqs)
3826 {
3827 	dev->real_num_rx_queues = rxqs;
3828 	return 0;
3829 }
3830 #endif
3831 int netif_set_real_num_queues(struct net_device *dev,
3832 			      unsigned int txq, unsigned int rxq);
3833 
3834 int netif_get_num_default_rss_queues(void);
3835 
3836 void dev_kfree_skb_irq_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3837 void dev_kfree_skb_any_reason(struct sk_buff *skb, enum skb_drop_reason reason);
3838 
3839 /*
3840  * It is not allowed to call kfree_skb() or consume_skb() from hardware
3841  * interrupt context or with hardware interrupts being disabled.
3842  * (in_hardirq() || irqs_disabled())
3843  *
3844  * We provide four helpers that can be used in following contexts :
3845  *
3846  * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3847  *  replacing kfree_skb(skb)
3848  *
3849  * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3850  *  Typically used in place of consume_skb(skb) in TX completion path
3851  *
3852  * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3853  *  replacing kfree_skb(skb)
3854  *
3855  * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3856  *  and consumed a packet. Used in place of consume_skb(skb)
3857  */
3858 static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3859 {
3860 	dev_kfree_skb_irq_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3861 }
3862 
3863 static inline void dev_consume_skb_irq(struct sk_buff *skb)
3864 {
3865 	dev_kfree_skb_irq_reason(skb, SKB_CONSUMED);
3866 }
3867 
3868 static inline void dev_kfree_skb_any(struct sk_buff *skb)
3869 {
3870 	dev_kfree_skb_any_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED);
3871 }
3872 
3873 static inline void dev_consume_skb_any(struct sk_buff *skb)
3874 {
3875 	dev_kfree_skb_any_reason(skb, SKB_CONSUMED);
3876 }
3877 
3878 u32 bpf_prog_run_generic_xdp(struct sk_buff *skb, struct xdp_buff *xdp,
3879 			     struct bpf_prog *xdp_prog);
3880 void generic_xdp_tx(struct sk_buff *skb, struct bpf_prog *xdp_prog);
3881 int do_xdp_generic(struct bpf_prog *xdp_prog, struct sk_buff *skb);
3882 int netif_rx(struct sk_buff *skb);
3883 int __netif_rx(struct sk_buff *skb);
3884 
3885 int netif_receive_skb(struct sk_buff *skb);
3886 int netif_receive_skb_core(struct sk_buff *skb);
3887 void netif_receive_skb_list_internal(struct list_head *head);
3888 void netif_receive_skb_list(struct list_head *head);
3889 gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3890 void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3891 struct sk_buff *napi_get_frags(struct napi_struct *napi);
3892 void napi_get_frags_check(struct napi_struct *napi);
3893 gro_result_t napi_gro_frags(struct napi_struct *napi);
3894 struct packet_offload *gro_find_receive_by_type(__be16 type);
3895 struct packet_offload *gro_find_complete_by_type(__be16 type);
3896 
3897 static inline void napi_free_frags(struct napi_struct *napi)
3898 {
3899 	kfree_skb(napi->skb);
3900 	napi->skb = NULL;
3901 }
3902 
3903 bool netdev_is_rx_handler_busy(struct net_device *dev);
3904 int netdev_rx_handler_register(struct net_device *dev,
3905 			       rx_handler_func_t *rx_handler,
3906 			       void *rx_handler_data);
3907 void netdev_rx_handler_unregister(struct net_device *dev);
3908 
3909 bool dev_valid_name(const char *name);
3910 static inline bool is_socket_ioctl_cmd(unsigned int cmd)
3911 {
3912 	return _IOC_TYPE(cmd) == SOCK_IOC_TYPE;
3913 }
3914 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg);
3915 int put_user_ifreq(struct ifreq *ifr, void __user *arg);
3916 int dev_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr,
3917 		void __user *data, bool *need_copyout);
3918 int dev_ifconf(struct net *net, struct ifconf __user *ifc);
3919 int generic_hwtstamp_get_lower(struct net_device *dev,
3920 			       struct kernel_hwtstamp_config *kernel_cfg);
3921 int generic_hwtstamp_set_lower(struct net_device *dev,
3922 			       struct kernel_hwtstamp_config *kernel_cfg,
3923 			       struct netlink_ext_ack *extack);
3924 int dev_ethtool(struct net *net, struct ifreq *ifr, void __user *userdata);
3925 unsigned int dev_get_flags(const struct net_device *);
3926 int __dev_change_flags(struct net_device *dev, unsigned int flags,
3927 		       struct netlink_ext_ack *extack);
3928 int dev_change_flags(struct net_device *dev, unsigned int flags,
3929 		     struct netlink_ext_ack *extack);
3930 int dev_set_alias(struct net_device *, const char *, size_t);
3931 int dev_get_alias(const struct net_device *, char *, size_t);
3932 int __dev_change_net_namespace(struct net_device *dev, struct net *net,
3933 			       const char *pat, int new_ifindex);
3934 static inline
3935 int dev_change_net_namespace(struct net_device *dev, struct net *net,
3936 			     const char *pat)
3937 {
3938 	return __dev_change_net_namespace(dev, net, pat, 0);
3939 }
3940 int __dev_set_mtu(struct net_device *, int);
3941 int dev_set_mtu(struct net_device *, int);
3942 int dev_pre_changeaddr_notify(struct net_device *dev, const char *addr,
3943 			      struct netlink_ext_ack *extack);
3944 int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa,
3945 			struct netlink_ext_ack *extack);
3946 int dev_set_mac_address_user(struct net_device *dev, struct sockaddr *sa,
3947 			     struct netlink_ext_ack *extack);
3948 int dev_get_mac_address(struct sockaddr *sa, struct net *net, char *dev_name);
3949 int dev_get_port_parent_id(struct net_device *dev,
3950 			   struct netdev_phys_item_id *ppid, bool recurse);
3951 bool netdev_port_same_parent_id(struct net_device *a, struct net_device *b);
3952 void netdev_dpll_pin_set(struct net_device *dev, struct dpll_pin *dpll_pin);
3953 void netdev_dpll_pin_clear(struct net_device *dev);
3954 
3955 static inline struct dpll_pin *netdev_dpll_pin(const struct net_device *dev)
3956 {
3957 #if IS_ENABLED(CONFIG_DPLL)
3958 	return dev->dpll_pin;
3959 #else
3960 	return NULL;
3961 #endif
3962 }
3963 
3964 struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev, bool *again);
3965 struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3966 				    struct netdev_queue *txq, int *ret);
3967 
3968 int bpf_xdp_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
3969 u8 dev_xdp_prog_count(struct net_device *dev);
3970 u32 dev_xdp_prog_id(struct net_device *dev, enum bpf_xdp_mode mode);
3971 
3972 int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3973 int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
3974 int dev_forward_skb_nomtu(struct net_device *dev, struct sk_buff *skb);
3975 bool is_skb_forwardable(const struct net_device *dev,
3976 			const struct sk_buff *skb);
3977 
3978 static __always_inline bool __is_skb_forwardable(const struct net_device *dev,
3979 						 const struct sk_buff *skb,
3980 						 const bool check_mtu)
3981 {
3982 	const u32 vlan_hdr_len = 4; /* VLAN_HLEN */
3983 	unsigned int len;
3984 
3985 	if (!(dev->flags & IFF_UP))
3986 		return false;
3987 
3988 	if (!check_mtu)
3989 		return true;
3990 
3991 	len = dev->mtu + dev->hard_header_len + vlan_hdr_len;
3992 	if (skb->len <= len)
3993 		return true;
3994 
3995 	/* if TSO is enabled, we don't care about the length as the packet
3996 	 * could be forwarded without being segmented before
3997 	 */
3998 	if (skb_is_gso(skb))
3999 		return true;
4000 
4001 	return false;
4002 }
4003 
4004 struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
4005 
4006 static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
4007 {
4008 	/* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
4009 	struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
4010 
4011 	if (likely(p))
4012 		return p;
4013 
4014 	return netdev_core_stats_alloc(dev);
4015 }
4016 
4017 #define DEV_CORE_STATS_INC(FIELD)						\
4018 static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev)		\
4019 {										\
4020 	struct net_device_core_stats __percpu *p;				\
4021 										\
4022 	p = dev_core_stats(dev);						\
4023 	if (p)									\
4024 		this_cpu_inc(p->FIELD);						\
4025 }
4026 DEV_CORE_STATS_INC(rx_dropped)
4027 DEV_CORE_STATS_INC(tx_dropped)
4028 DEV_CORE_STATS_INC(rx_nohandler)
4029 DEV_CORE_STATS_INC(rx_otherhost_dropped)
4030 
4031 static __always_inline int ____dev_forward_skb(struct net_device *dev,
4032 					       struct sk_buff *skb,
4033 					       const bool check_mtu)
4034 {
4035 	if (skb_orphan_frags(skb, GFP_ATOMIC) ||
4036 	    unlikely(!__is_skb_forwardable(dev, skb, check_mtu))) {
4037 		dev_core_stats_rx_dropped_inc(dev);
4038 		kfree_skb(skb);
4039 		return NET_RX_DROP;
4040 	}
4041 
4042 	skb_scrub_packet(skb, !net_eq(dev_net(dev), dev_net(skb->dev)));
4043 	skb->priority = 0;
4044 	return 0;
4045 }
4046 
4047 bool dev_nit_active(struct net_device *dev);
4048 void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
4049 
4050 static inline void __dev_put(struct net_device *dev)
4051 {
4052 	if (dev) {
4053 #ifdef CONFIG_PCPU_DEV_REFCNT
4054 		this_cpu_dec(*dev->pcpu_refcnt);
4055 #else
4056 		refcount_dec(&dev->dev_refcnt);
4057 #endif
4058 	}
4059 }
4060 
4061 static inline void __dev_hold(struct net_device *dev)
4062 {
4063 	if (dev) {
4064 #ifdef CONFIG_PCPU_DEV_REFCNT
4065 		this_cpu_inc(*dev->pcpu_refcnt);
4066 #else
4067 		refcount_inc(&dev->dev_refcnt);
4068 #endif
4069 	}
4070 }
4071 
4072 static inline void __netdev_tracker_alloc(struct net_device *dev,
4073 					  netdevice_tracker *tracker,
4074 					  gfp_t gfp)
4075 {
4076 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4077 	ref_tracker_alloc(&dev->refcnt_tracker, tracker, gfp);
4078 #endif
4079 }
4080 
4081 /* netdev_tracker_alloc() can upgrade a prior untracked reference
4082  * taken by dev_get_by_name()/dev_get_by_index() to a tracked one.
4083  */
4084 static inline void netdev_tracker_alloc(struct net_device *dev,
4085 					netdevice_tracker *tracker, gfp_t gfp)
4086 {
4087 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4088 	refcount_dec(&dev->refcnt_tracker.no_tracker);
4089 	__netdev_tracker_alloc(dev, tracker, gfp);
4090 #endif
4091 }
4092 
4093 static inline void netdev_tracker_free(struct net_device *dev,
4094 				       netdevice_tracker *tracker)
4095 {
4096 #ifdef CONFIG_NET_DEV_REFCNT_TRACKER
4097 	ref_tracker_free(&dev->refcnt_tracker, tracker);
4098 #endif
4099 }
4100 
4101 static inline void netdev_hold(struct net_device *dev,
4102 			       netdevice_tracker *tracker, gfp_t gfp)
4103 {
4104 	if (dev) {
4105 		__dev_hold(dev);
4106 		__netdev_tracker_alloc(dev, tracker, gfp);
4107 	}
4108 }
4109 
4110 static inline void netdev_put(struct net_device *dev,
4111 			      netdevice_tracker *tracker)
4112 {
4113 	if (dev) {
4114 		netdev_tracker_free(dev, tracker);
4115 		__dev_put(dev);
4116 	}
4117 }
4118 
4119 /**
4120  *	dev_hold - get reference to device
4121  *	@dev: network device
4122  *
4123  * Hold reference to device to keep it from being freed.
4124  * Try using netdev_hold() instead.
4125  */
4126 static inline void dev_hold(struct net_device *dev)
4127 {
4128 	netdev_hold(dev, NULL, GFP_ATOMIC);
4129 }
4130 
4131 /**
4132  *	dev_put - release reference to device
4133  *	@dev: network device
4134  *
4135  * Release reference to device to allow it to be freed.
4136  * Try using netdev_put() instead.
4137  */
4138 static inline void dev_put(struct net_device *dev)
4139 {
4140 	netdev_put(dev, NULL);
4141 }
4142 
4143 static inline void netdev_ref_replace(struct net_device *odev,
4144 				      struct net_device *ndev,
4145 				      netdevice_tracker *tracker,
4146 				      gfp_t gfp)
4147 {
4148 	if (odev)
4149 		netdev_tracker_free(odev, tracker);
4150 
4151 	__dev_hold(ndev);
4152 	__dev_put(odev);
4153 
4154 	if (ndev)
4155 		__netdev_tracker_alloc(ndev, tracker, gfp);
4156 }
4157 
4158 /* Carrier loss detection, dial on demand. The functions netif_carrier_on
4159  * and _off may be called from IRQ context, but it is caller
4160  * who is responsible for serialization of these calls.
4161  *
4162  * The name carrier is inappropriate, these functions should really be
4163  * called netif_lowerlayer_*() because they represent the state of any
4164  * kind of lower layer not just hardware media.
4165  */
4166 void linkwatch_fire_event(struct net_device *dev);
4167 
4168 /**
4169  *	netif_carrier_ok - test if carrier present
4170  *	@dev: network device
4171  *
4172  * Check if carrier is present on device
4173  */
4174 static inline bool netif_carrier_ok(const struct net_device *dev)
4175 {
4176 	return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
4177 }
4178 
4179 unsigned long dev_trans_start(struct net_device *dev);
4180 
4181 void __netdev_watchdog_up(struct net_device *dev);
4182 
4183 void netif_carrier_on(struct net_device *dev);
4184 void netif_carrier_off(struct net_device *dev);
4185 void netif_carrier_event(struct net_device *dev);
4186 
4187 /**
4188  *	netif_dormant_on - mark device as dormant.
4189  *	@dev: network device
4190  *
4191  * Mark device as dormant (as per RFC2863).
4192  *
4193  * The dormant state indicates that the relevant interface is not
4194  * actually in a condition to pass packets (i.e., it is not 'up') but is
4195  * in a "pending" state, waiting for some external event.  For "on-
4196  * demand" interfaces, this new state identifies the situation where the
4197  * interface is waiting for events to place it in the up state.
4198  */
4199 static inline void netif_dormant_on(struct net_device *dev)
4200 {
4201 	if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
4202 		linkwatch_fire_event(dev);
4203 }
4204 
4205 /**
4206  *	netif_dormant_off - set device as not dormant.
4207  *	@dev: network device
4208  *
4209  * Device is not in dormant state.
4210  */
4211 static inline void netif_dormant_off(struct net_device *dev)
4212 {
4213 	if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
4214 		linkwatch_fire_event(dev);
4215 }
4216 
4217 /**
4218  *	netif_dormant - test if device is dormant
4219  *	@dev: network device
4220  *
4221  * Check if device is dormant.
4222  */
4223 static inline bool netif_dormant(const struct net_device *dev)
4224 {
4225 	return test_bit(__LINK_STATE_DORMANT, &dev->state);
4226 }
4227 
4228 
4229 /**
4230  *	netif_testing_on - mark device as under test.
4231  *	@dev: network device
4232  *
4233  * Mark device as under test (as per RFC2863).
4234  *
4235  * The testing state indicates that some test(s) must be performed on
4236  * the interface. After completion, of the test, the interface state
4237  * will change to up, dormant, or down, as appropriate.
4238  */
4239 static inline void netif_testing_on(struct net_device *dev)
4240 {
4241 	if (!test_and_set_bit(__LINK_STATE_TESTING, &dev->state))
4242 		linkwatch_fire_event(dev);
4243 }
4244 
4245 /**
4246  *	netif_testing_off - set device as not under test.
4247  *	@dev: network device
4248  *
4249  * Device is not in testing state.
4250  */
4251 static inline void netif_testing_off(struct net_device *dev)
4252 {
4253 	if (test_and_clear_bit(__LINK_STATE_TESTING, &dev->state))
4254 		linkwatch_fire_event(dev);
4255 }
4256 
4257 /**
4258  *	netif_testing - test if device is under test
4259  *	@dev: network device
4260  *
4261  * Check if device is under test
4262  */
4263 static inline bool netif_testing(const struct net_device *dev)
4264 {
4265 	return test_bit(__LINK_STATE_TESTING, &dev->state);
4266 }
4267 
4268 
4269 /**
4270  *	netif_oper_up - test if device is operational
4271  *	@dev: network device
4272  *
4273  * Check if carrier is operational
4274  */
4275 static inline bool netif_oper_up(const struct net_device *dev)
4276 {
4277 	return (dev->operstate == IF_OPER_UP ||
4278 		dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
4279 }
4280 
4281 /**
4282  *	netif_device_present - is device available or removed
4283  *	@dev: network device
4284  *
4285  * Check if device has not been removed from system.
4286  */
4287 static inline bool netif_device_present(const struct net_device *dev)
4288 {
4289 	return test_bit(__LINK_STATE_PRESENT, &dev->state);
4290 }
4291 
4292 void netif_device_detach(struct net_device *dev);
4293 
4294 void netif_device_attach(struct net_device *dev);
4295 
4296 /*
4297  * Network interface message level settings
4298  */
4299 
4300 enum {
4301 	NETIF_MSG_DRV_BIT,
4302 	NETIF_MSG_PROBE_BIT,
4303 	NETIF_MSG_LINK_BIT,
4304 	NETIF_MSG_TIMER_BIT,
4305 	NETIF_MSG_IFDOWN_BIT,
4306 	NETIF_MSG_IFUP_BIT,
4307 	NETIF_MSG_RX_ERR_BIT,
4308 	NETIF_MSG_TX_ERR_BIT,
4309 	NETIF_MSG_TX_QUEUED_BIT,
4310 	NETIF_MSG_INTR_BIT,
4311 	NETIF_MSG_TX_DONE_BIT,
4312 	NETIF_MSG_RX_STATUS_BIT,
4313 	NETIF_MSG_PKTDATA_BIT,
4314 	NETIF_MSG_HW_BIT,
4315 	NETIF_MSG_WOL_BIT,
4316 
4317 	/* When you add a new bit above, update netif_msg_class_names array
4318 	 * in net/ethtool/common.c
4319 	 */
4320 	NETIF_MSG_CLASS_COUNT,
4321 };
4322 /* Both ethtool_ops interface and internal driver implementation use u32 */
4323 static_assert(NETIF_MSG_CLASS_COUNT <= 32);
4324 
4325 #define __NETIF_MSG_BIT(bit)	((u32)1 << (bit))
4326 #define __NETIF_MSG(name)	__NETIF_MSG_BIT(NETIF_MSG_ ## name ## _BIT)
4327 
4328 #define NETIF_MSG_DRV		__NETIF_MSG(DRV)
4329 #define NETIF_MSG_PROBE		__NETIF_MSG(PROBE)
4330 #define NETIF_MSG_LINK		__NETIF_MSG(LINK)
4331 #define NETIF_MSG_TIMER		__NETIF_MSG(TIMER)
4332 #define NETIF_MSG_IFDOWN	__NETIF_MSG(IFDOWN)
4333 #define NETIF_MSG_IFUP		__NETIF_MSG(IFUP)
4334 #define NETIF_MSG_RX_ERR	__NETIF_MSG(RX_ERR)
4335 #define NETIF_MSG_TX_ERR	__NETIF_MSG(TX_ERR)
4336 #define NETIF_MSG_TX_QUEUED	__NETIF_MSG(TX_QUEUED)
4337 #define NETIF_MSG_INTR		__NETIF_MSG(INTR)
4338 #define NETIF_MSG_TX_DONE	__NETIF_MSG(TX_DONE)
4339 #define NETIF_MSG_RX_STATUS	__NETIF_MSG(RX_STATUS)
4340 #define NETIF_MSG_PKTDATA	__NETIF_MSG(PKTDATA)
4341 #define NETIF_MSG_HW		__NETIF_MSG(HW)
4342 #define NETIF_MSG_WOL		__NETIF_MSG(WOL)
4343 
4344 #define netif_msg_drv(p)	((p)->msg_enable & NETIF_MSG_DRV)
4345 #define netif_msg_probe(p)	((p)->msg_enable & NETIF_MSG_PROBE)
4346 #define netif_msg_link(p)	((p)->msg_enable & NETIF_MSG_LINK)
4347 #define netif_msg_timer(p)	((p)->msg_enable & NETIF_MSG_TIMER)
4348 #define netif_msg_ifdown(p)	((p)->msg_enable & NETIF_MSG_IFDOWN)
4349 #define netif_msg_ifup(p)	((p)->msg_enable & NETIF_MSG_IFUP)
4350 #define netif_msg_rx_err(p)	((p)->msg_enable & NETIF_MSG_RX_ERR)
4351 #define netif_msg_tx_err(p)	((p)->msg_enable & NETIF_MSG_TX_ERR)
4352 #define netif_msg_tx_queued(p)	((p)->msg_enable & NETIF_MSG_TX_QUEUED)
4353 #define netif_msg_intr(p)	((p)->msg_enable & NETIF_MSG_INTR)
4354 #define netif_msg_tx_done(p)	((p)->msg_enable & NETIF_MSG_TX_DONE)
4355 #define netif_msg_rx_status(p)	((p)->msg_enable & NETIF_MSG_RX_STATUS)
4356 #define netif_msg_pktdata(p)	((p)->msg_enable & NETIF_MSG_PKTDATA)
4357 #define netif_msg_hw(p)		((p)->msg_enable & NETIF_MSG_HW)
4358 #define netif_msg_wol(p)	((p)->msg_enable & NETIF_MSG_WOL)
4359 
4360 static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
4361 {
4362 	/* use default */
4363 	if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
4364 		return default_msg_enable_bits;
4365 	if (debug_value == 0)	/* no output */
4366 		return 0;
4367 	/* set low N bits */
4368 	return (1U << debug_value) - 1;
4369 }
4370 
4371 static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
4372 {
4373 	spin_lock(&txq->_xmit_lock);
4374 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4375 	WRITE_ONCE(txq->xmit_lock_owner, cpu);
4376 }
4377 
4378 static inline bool __netif_tx_acquire(struct netdev_queue *txq)
4379 {
4380 	__acquire(&txq->_xmit_lock);
4381 	return true;
4382 }
4383 
4384 static inline void __netif_tx_release(struct netdev_queue *txq)
4385 {
4386 	__release(&txq->_xmit_lock);
4387 }
4388 
4389 static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
4390 {
4391 	spin_lock_bh(&txq->_xmit_lock);
4392 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4393 	WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4394 }
4395 
4396 static inline bool __netif_tx_trylock(struct netdev_queue *txq)
4397 {
4398 	bool ok = spin_trylock(&txq->_xmit_lock);
4399 
4400 	if (likely(ok)) {
4401 		/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4402 		WRITE_ONCE(txq->xmit_lock_owner, smp_processor_id());
4403 	}
4404 	return ok;
4405 }
4406 
4407 static inline void __netif_tx_unlock(struct netdev_queue *txq)
4408 {
4409 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4410 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4411 	spin_unlock(&txq->_xmit_lock);
4412 }
4413 
4414 static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
4415 {
4416 	/* Pairs with READ_ONCE() in __dev_queue_xmit() */
4417 	WRITE_ONCE(txq->xmit_lock_owner, -1);
4418 	spin_unlock_bh(&txq->_xmit_lock);
4419 }
4420 
4421 /*
4422  * txq->trans_start can be read locklessly from dev_watchdog()
4423  */
4424 static inline void txq_trans_update(struct netdev_queue *txq)
4425 {
4426 	if (txq->xmit_lock_owner != -1)
4427 		WRITE_ONCE(txq->trans_start, jiffies);
4428 }
4429 
4430 static inline void txq_trans_cond_update(struct netdev_queue *txq)
4431 {
4432 	unsigned long now = jiffies;
4433 
4434 	if (READ_ONCE(txq->trans_start) != now)
4435 		WRITE_ONCE(txq->trans_start, now);
4436 }
4437 
4438 /* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
4439 static inline void netif_trans_update(struct net_device *dev)
4440 {
4441 	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
4442 
4443 	txq_trans_cond_update(txq);
4444 }
4445 
4446 /**
4447  *	netif_tx_lock - grab network device transmit lock
4448  *	@dev: network device
4449  *
4450  * Get network device transmit lock
4451  */
4452 void netif_tx_lock(struct net_device *dev);
4453 
4454 static inline void netif_tx_lock_bh(struct net_device *dev)
4455 {
4456 	local_bh_disable();
4457 	netif_tx_lock(dev);
4458 }
4459 
4460 void netif_tx_unlock(struct net_device *dev);
4461 
4462 static inline void netif_tx_unlock_bh(struct net_device *dev)
4463 {
4464 	netif_tx_unlock(dev);
4465 	local_bh_enable();
4466 }
4467 
4468 #define HARD_TX_LOCK(dev, txq, cpu) {			\
4469 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4470 		__netif_tx_lock(txq, cpu);		\
4471 	} else {					\
4472 		__netif_tx_acquire(txq);		\
4473 	}						\
4474 }
4475 
4476 #define HARD_TX_TRYLOCK(dev, txq)			\
4477 	(((dev->features & NETIF_F_LLTX) == 0) ?	\
4478 		__netif_tx_trylock(txq) :		\
4479 		__netif_tx_acquire(txq))
4480 
4481 #define HARD_TX_UNLOCK(dev, txq) {			\
4482 	if ((dev->features & NETIF_F_LLTX) == 0) {	\
4483 		__netif_tx_unlock(txq);			\
4484 	} else {					\
4485 		__netif_tx_release(txq);		\
4486 	}						\
4487 }
4488 
4489 static inline void netif_tx_disable(struct net_device *dev)
4490 {
4491 	unsigned int i;
4492 	int cpu;
4493 
4494 	local_bh_disable();
4495 	cpu = smp_processor_id();
4496 	spin_lock(&dev->tx_global_lock);
4497 	for (i = 0; i < dev->num_tx_queues; i++) {
4498 		struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
4499 
4500 		__netif_tx_lock(txq, cpu);
4501 		netif_tx_stop_queue(txq);
4502 		__netif_tx_unlock(txq);
4503 	}
4504 	spin_unlock(&dev->tx_global_lock);
4505 	local_bh_enable();
4506 }
4507 
4508 static inline void netif_addr_lock(struct net_device *dev)
4509 {
4510 	unsigned char nest_level = 0;
4511 
4512 #ifdef CONFIG_LOCKDEP
4513 	nest_level = dev->nested_level;
4514 #endif
4515 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4516 }
4517 
4518 static inline void netif_addr_lock_bh(struct net_device *dev)
4519 {
4520 	unsigned char nest_level = 0;
4521 
4522 #ifdef CONFIG_LOCKDEP
4523 	nest_level = dev->nested_level;
4524 #endif
4525 	local_bh_disable();
4526 	spin_lock_nested(&dev->addr_list_lock, nest_level);
4527 }
4528 
4529 static inline void netif_addr_unlock(struct net_device *dev)
4530 {
4531 	spin_unlock(&dev->addr_list_lock);
4532 }
4533 
4534 static inline void netif_addr_unlock_bh(struct net_device *dev)
4535 {
4536 	spin_unlock_bh(&dev->addr_list_lock);
4537 }
4538 
4539 /*
4540  * dev_addrs walker. Should be used only for read access. Call with
4541  * rcu_read_lock held.
4542  */
4543 #define for_each_dev_addr(dev, ha) \
4544 		list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
4545 
4546 /* These functions live elsewhere (drivers/net/net_init.c, but related) */
4547 
4548 void ether_setup(struct net_device *dev);
4549 
4550 /* Support for loadable net-drivers */
4551 struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
4552 				    unsigned char name_assign_type,
4553 				    void (*setup)(struct net_device *),
4554 				    unsigned int txqs, unsigned int rxqs);
4555 #define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
4556 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
4557 
4558 #define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
4559 	alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
4560 			 count)
4561 
4562 int register_netdev(struct net_device *dev);
4563 void unregister_netdev(struct net_device *dev);
4564 
4565 int devm_register_netdev(struct device *dev, struct net_device *ndev);
4566 
4567 /* General hardware address lists handling functions */
4568 int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
4569 		   struct netdev_hw_addr_list *from_list, int addr_len);
4570 void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
4571 		      struct netdev_hw_addr_list *from_list, int addr_len);
4572 int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
4573 		       struct net_device *dev,
4574 		       int (*sync)(struct net_device *, const unsigned char *),
4575 		       int (*unsync)(struct net_device *,
4576 				     const unsigned char *));
4577 int __hw_addr_ref_sync_dev(struct netdev_hw_addr_list *list,
4578 			   struct net_device *dev,
4579 			   int (*sync)(struct net_device *,
4580 				       const unsigned char *, int),
4581 			   int (*unsync)(struct net_device *,
4582 					 const unsigned char *, int));
4583 void __hw_addr_ref_unsync_dev(struct netdev_hw_addr_list *list,
4584 			      struct net_device *dev,
4585 			      int (*unsync)(struct net_device *,
4586 					    const unsigned char *, int));
4587 void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
4588 			  struct net_device *dev,
4589 			  int (*unsync)(struct net_device *,
4590 					const unsigned char *));
4591 void __hw_addr_init(struct netdev_hw_addr_list *list);
4592 
4593 /* Functions used for device addresses handling */
4594 void dev_addr_mod(struct net_device *dev, unsigned int offset,
4595 		  const void *addr, size_t len);
4596 
4597 static inline void
4598 __dev_addr_set(struct net_device *dev, const void *addr, size_t len)
4599 {
4600 	dev_addr_mod(dev, 0, addr, len);
4601 }
4602 
4603 static inline void dev_addr_set(struct net_device *dev, const u8 *addr)
4604 {
4605 	__dev_addr_set(dev, addr, dev->addr_len);
4606 }
4607 
4608 int dev_addr_add(struct net_device *dev, const unsigned char *addr,
4609 		 unsigned char addr_type);
4610 int dev_addr_del(struct net_device *dev, const unsigned char *addr,
4611 		 unsigned char addr_type);
4612 
4613 /* Functions used for unicast addresses handling */
4614 int dev_uc_add(struct net_device *dev, const unsigned char *addr);
4615 int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
4616 int dev_uc_del(struct net_device *dev, const unsigned char *addr);
4617 int dev_uc_sync(struct net_device *to, struct net_device *from);
4618 int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
4619 void dev_uc_unsync(struct net_device *to, struct net_device *from);
4620 void dev_uc_flush(struct net_device *dev);
4621 void dev_uc_init(struct net_device *dev);
4622 
4623 /**
4624  *  __dev_uc_sync - Synchonize device's unicast list
4625  *  @dev:  device to sync
4626  *  @sync: function to call if address should be added
4627  *  @unsync: function to call if address should be removed
4628  *
4629  *  Add newly added addresses to the interface, and release
4630  *  addresses that have been deleted.
4631  */
4632 static inline int __dev_uc_sync(struct net_device *dev,
4633 				int (*sync)(struct net_device *,
4634 					    const unsigned char *),
4635 				int (*unsync)(struct net_device *,
4636 					      const unsigned char *))
4637 {
4638 	return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
4639 }
4640 
4641 /**
4642  *  __dev_uc_unsync - Remove synchronized addresses from device
4643  *  @dev:  device to sync
4644  *  @unsync: function to call if address should be removed
4645  *
4646  *  Remove all addresses that were added to the device by dev_uc_sync().
4647  */
4648 static inline void __dev_uc_unsync(struct net_device *dev,
4649 				   int (*unsync)(struct net_device *,
4650 						 const unsigned char *))
4651 {
4652 	__hw_addr_unsync_dev(&dev->uc, dev, unsync);
4653 }
4654 
4655 /* Functions used for multicast addresses handling */
4656 int dev_mc_add(struct net_device *dev, const unsigned char *addr);
4657 int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
4658 int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
4659 int dev_mc_del(struct net_device *dev, const unsigned char *addr);
4660 int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
4661 int dev_mc_sync(struct net_device *to, struct net_device *from);
4662 int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
4663 void dev_mc_unsync(struct net_device *to, struct net_device *from);
4664 void dev_mc_flush(struct net_device *dev);
4665 void dev_mc_init(struct net_device *dev);
4666 
4667 /**
4668  *  __dev_mc_sync - Synchonize device's multicast list
4669  *  @dev:  device to sync
4670  *  @sync: function to call if address should be added
4671  *  @unsync: function to call if address should be removed
4672  *
4673  *  Add newly added addresses to the interface, and release
4674  *  addresses that have been deleted.
4675  */
4676 static inline int __dev_mc_sync(struct net_device *dev,
4677 				int (*sync)(struct net_device *,
4678 					    const unsigned char *),
4679 				int (*unsync)(struct net_device *,
4680 					      const unsigned char *))
4681 {
4682 	return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
4683 }
4684 
4685 /**
4686  *  __dev_mc_unsync - Remove synchronized addresses from device
4687  *  @dev:  device to sync
4688  *  @unsync: function to call if address should be removed
4689  *
4690  *  Remove all addresses that were added to the device by dev_mc_sync().
4691  */
4692 static inline void __dev_mc_unsync(struct net_device *dev,
4693 				   int (*unsync)(struct net_device *,
4694 						 const unsigned char *))
4695 {
4696 	__hw_addr_unsync_dev(&dev->mc, dev, unsync);
4697 }
4698 
4699 /* Functions used for secondary unicast and multicast support */
4700 void dev_set_rx_mode(struct net_device *dev);
4701 int dev_set_promiscuity(struct net_device *dev, int inc);
4702 int dev_set_allmulti(struct net_device *dev, int inc);
4703 void netdev_state_change(struct net_device *dev);
4704 void __netdev_notify_peers(struct net_device *dev);
4705 void netdev_notify_peers(struct net_device *dev);
4706 void netdev_features_change(struct net_device *dev);
4707 /* Load a device via the kmod */
4708 void dev_load(struct net *net, const char *name);
4709 struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
4710 					struct rtnl_link_stats64 *storage);
4711 void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
4712 			     const struct net_device_stats *netdev_stats);
4713 void dev_fetch_sw_netstats(struct rtnl_link_stats64 *s,
4714 			   const struct pcpu_sw_netstats __percpu *netstats);
4715 void dev_get_tstats64(struct net_device *dev, struct rtnl_link_stats64 *s);
4716 
4717 extern int		netdev_max_backlog;
4718 extern int		dev_rx_weight;
4719 extern int		dev_tx_weight;
4720 extern int		gro_normal_batch;
4721 
4722 enum {
4723 	NESTED_SYNC_IMM_BIT,
4724 	NESTED_SYNC_TODO_BIT,
4725 };
4726 
4727 #define __NESTED_SYNC_BIT(bit)	((u32)1 << (bit))
4728 #define __NESTED_SYNC(name)	__NESTED_SYNC_BIT(NESTED_SYNC_ ## name ## _BIT)
4729 
4730 #define NESTED_SYNC_IMM		__NESTED_SYNC(IMM)
4731 #define NESTED_SYNC_TODO	__NESTED_SYNC(TODO)
4732 
4733 struct netdev_nested_priv {
4734 	unsigned char flags;
4735 	void *data;
4736 };
4737 
4738 bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
4739 struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
4740 						     struct list_head **iter);
4741 
4742 /* iterate through upper list, must be called under RCU read lock */
4743 #define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
4744 	for (iter = &(dev)->adj_list.upper, \
4745 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
4746 	     updev; \
4747 	     updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
4748 
4749 int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
4750 				  int (*fn)(struct net_device *upper_dev,
4751 					    struct netdev_nested_priv *priv),
4752 				  struct netdev_nested_priv *priv);
4753 
4754 bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
4755 				  struct net_device *upper_dev);
4756 
4757 bool netdev_has_any_upper_dev(struct net_device *dev);
4758 
4759 void *netdev_lower_get_next_private(struct net_device *dev,
4760 				    struct list_head **iter);
4761 void *netdev_lower_get_next_private_rcu(struct net_device *dev,
4762 					struct list_head **iter);
4763 
4764 #define netdev_for_each_lower_private(dev, priv, iter) \
4765 	for (iter = (dev)->adj_list.lower.next, \
4766 	     priv = netdev_lower_get_next_private(dev, &(iter)); \
4767 	     priv; \
4768 	     priv = netdev_lower_get_next_private(dev, &(iter)))
4769 
4770 #define netdev_for_each_lower_private_rcu(dev, priv, iter) \
4771 	for (iter = &(dev)->adj_list.lower, \
4772 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
4773 	     priv; \
4774 	     priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
4775 
4776 void *netdev_lower_get_next(struct net_device *dev,
4777 				struct list_head **iter);
4778 
4779 #define netdev_for_each_lower_dev(dev, ldev, iter) \
4780 	for (iter = (dev)->adj_list.lower.next, \
4781 	     ldev = netdev_lower_get_next(dev, &(iter)); \
4782 	     ldev; \
4783 	     ldev = netdev_lower_get_next(dev, &(iter)))
4784 
4785 struct net_device *netdev_next_lower_dev_rcu(struct net_device *dev,
4786 					     struct list_head **iter);
4787 int netdev_walk_all_lower_dev(struct net_device *dev,
4788 			      int (*fn)(struct net_device *lower_dev,
4789 					struct netdev_nested_priv *priv),
4790 			      struct netdev_nested_priv *priv);
4791 int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
4792 				  int (*fn)(struct net_device *lower_dev,
4793 					    struct netdev_nested_priv *priv),
4794 				  struct netdev_nested_priv *priv);
4795 
4796 void *netdev_adjacent_get_private(struct list_head *adj_list);
4797 void *netdev_lower_get_first_private_rcu(struct net_device *dev);
4798 struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
4799 struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
4800 int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev,
4801 			  struct netlink_ext_ack *extack);
4802 int netdev_master_upper_dev_link(struct net_device *dev,
4803 				 struct net_device *upper_dev,
4804 				 void *upper_priv, void *upper_info,
4805 				 struct netlink_ext_ack *extack);
4806 void netdev_upper_dev_unlink(struct net_device *dev,
4807 			     struct net_device *upper_dev);
4808 int netdev_adjacent_change_prepare(struct net_device *old_dev,
4809 				   struct net_device *new_dev,
4810 				   struct net_device *dev,
4811 				   struct netlink_ext_ack *extack);
4812 void netdev_adjacent_change_commit(struct net_device *old_dev,
4813 				   struct net_device *new_dev,
4814 				   struct net_device *dev);
4815 void netdev_adjacent_change_abort(struct net_device *old_dev,
4816 				  struct net_device *new_dev,
4817 				  struct net_device *dev);
4818 void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
4819 void *netdev_lower_dev_get_private(struct net_device *dev,
4820 				   struct net_device *lower_dev);
4821 void netdev_lower_state_changed(struct net_device *lower_dev,
4822 				void *lower_state_info);
4823 
4824 /* RSS keys are 40 or 52 bytes long */
4825 #define NETDEV_RSS_KEY_LEN 52
4826 extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
4827 void netdev_rss_key_fill(void *buffer, size_t len);
4828 
4829 int skb_checksum_help(struct sk_buff *skb);
4830 int skb_crc32c_csum_help(struct sk_buff *skb);
4831 int skb_csum_hwoffload_help(struct sk_buff *skb,
4832 			    const netdev_features_t features);
4833 
4834 struct netdev_bonding_info {
4835 	ifslave	slave;
4836 	ifbond	master;
4837 };
4838 
4839 struct netdev_notifier_bonding_info {
4840 	struct netdev_notifier_info info; /* must be first */
4841 	struct netdev_bonding_info  bonding_info;
4842 };
4843 
4844 void netdev_bonding_info_change(struct net_device *dev,
4845 				struct netdev_bonding_info *bonding_info);
4846 
4847 #if IS_ENABLED(CONFIG_ETHTOOL_NETLINK)
4848 void ethtool_notify(struct net_device *dev, unsigned int cmd, const void *data);
4849 #else
4850 static inline void ethtool_notify(struct net_device *dev, unsigned int cmd,
4851 				  const void *data)
4852 {
4853 }
4854 #endif
4855 
4856 __be16 skb_network_protocol(struct sk_buff *skb, int *depth);
4857 
4858 static inline bool can_checksum_protocol(netdev_features_t features,
4859 					 __be16 protocol)
4860 {
4861 	if (protocol == htons(ETH_P_FCOE))
4862 		return !!(features & NETIF_F_FCOE_CRC);
4863 
4864 	/* Assume this is an IP checksum (not SCTP CRC) */
4865 
4866 	if (features & NETIF_F_HW_CSUM) {
4867 		/* Can checksum everything */
4868 		return true;
4869 	}
4870 
4871 	switch (protocol) {
4872 	case htons(ETH_P_IP):
4873 		return !!(features & NETIF_F_IP_CSUM);
4874 	case htons(ETH_P_IPV6):
4875 		return !!(features & NETIF_F_IPV6_CSUM);
4876 	default:
4877 		return false;
4878 	}
4879 }
4880 
4881 #ifdef CONFIG_BUG
4882 void netdev_rx_csum_fault(struct net_device *dev, struct sk_buff *skb);
4883 #else
4884 static inline void netdev_rx_csum_fault(struct net_device *dev,
4885 					struct sk_buff *skb)
4886 {
4887 }
4888 #endif
4889 /* rx skb timestamps */
4890 void net_enable_timestamp(void);
4891 void net_disable_timestamp(void);
4892 
4893 static inline ktime_t netdev_get_tstamp(struct net_device *dev,
4894 					const struct skb_shared_hwtstamps *hwtstamps,
4895 					bool cycles)
4896 {
4897 	const struct net_device_ops *ops = dev->netdev_ops;
4898 
4899 	if (ops->ndo_get_tstamp)
4900 		return ops->ndo_get_tstamp(dev, hwtstamps, cycles);
4901 
4902 	return hwtstamps->hwtstamp;
4903 }
4904 
4905 static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
4906 					      struct sk_buff *skb, struct net_device *dev,
4907 					      bool more)
4908 {
4909 	__this_cpu_write(softnet_data.xmit.more, more);
4910 	return ops->ndo_start_xmit(skb, dev);
4911 }
4912 
4913 static inline bool netdev_xmit_more(void)
4914 {
4915 	return __this_cpu_read(softnet_data.xmit.more);
4916 }
4917 
4918 static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
4919 					    struct netdev_queue *txq, bool more)
4920 {
4921 	const struct net_device_ops *ops = dev->netdev_ops;
4922 	netdev_tx_t rc;
4923 
4924 	rc = __netdev_start_xmit(ops, skb, dev, more);
4925 	if (rc == NETDEV_TX_OK)
4926 		txq_trans_update(txq);
4927 
4928 	return rc;
4929 }
4930 
4931 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
4932 				const void *ns);
4933 void netdev_class_remove_file_ns(const struct class_attribute *class_attr,
4934 				 const void *ns);
4935 
4936 extern const struct kobj_ns_type_operations net_ns_type_operations;
4937 
4938 const char *netdev_drivername(const struct net_device *dev);
4939 
4940 static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4941 							  netdev_features_t f2)
4942 {
4943 	if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4944 		if (f1 & NETIF_F_HW_CSUM)
4945 			f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4946 		else
4947 			f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
4948 	}
4949 
4950 	return f1 & f2;
4951 }
4952 
4953 static inline netdev_features_t netdev_get_wanted_features(
4954 	struct net_device *dev)
4955 {
4956 	return (dev->features & ~dev->hw_features) | dev->wanted_features;
4957 }
4958 netdev_features_t netdev_increment_features(netdev_features_t all,
4959 	netdev_features_t one, netdev_features_t mask);
4960 
4961 /* Allow TSO being used on stacked device :
4962  * Performing the GSO segmentation before last device
4963  * is a performance improvement.
4964  */
4965 static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4966 							netdev_features_t mask)
4967 {
4968 	return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4969 }
4970 
4971 int __netdev_update_features(struct net_device *dev);
4972 void netdev_update_features(struct net_device *dev);
4973 void netdev_change_features(struct net_device *dev);
4974 
4975 void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4976 					struct net_device *dev);
4977 
4978 netdev_features_t passthru_features_check(struct sk_buff *skb,
4979 					  struct net_device *dev,
4980 					  netdev_features_t features);
4981 netdev_features_t netif_skb_features(struct sk_buff *skb);
4982 void skb_warn_bad_offload(const struct sk_buff *skb);
4983 
4984 static inline bool net_gso_ok(netdev_features_t features, int gso_type)
4985 {
4986 	netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
4987 
4988 	/* check flags correspondence */
4989 	BUILD_BUG_ON(SKB_GSO_TCPV4   != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4990 	BUILD_BUG_ON(SKB_GSO_DODGY   != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4991 	BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4992 	BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
4993 	BUILD_BUG_ON(SKB_GSO_TCPV6   != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4994 	BUILD_BUG_ON(SKB_GSO_FCOE    != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
4995 	BUILD_BUG_ON(SKB_GSO_GRE     != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4996 	BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4997 	BUILD_BUG_ON(SKB_GSO_IPXIP4  != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4998 	BUILD_BUG_ON(SKB_GSO_IPXIP6  != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
4999 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
5000 	BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
5001 	BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
5002 	BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
5003 	BUILD_BUG_ON(SKB_GSO_SCTP    != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
5004 	BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
5005 	BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_GSO_UDP >> NETIF_F_GSO_SHIFT));
5006 	BUILD_BUG_ON(SKB_GSO_UDP_L4 != (NETIF_F_GSO_UDP_L4 >> NETIF_F_GSO_SHIFT));
5007 	BUILD_BUG_ON(SKB_GSO_FRAGLIST != (NETIF_F_GSO_FRAGLIST >> NETIF_F_GSO_SHIFT));
5008 
5009 	return (features & feature) == feature;
5010 }
5011 
5012 static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
5013 {
5014 	return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
5015 	       (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
5016 }
5017 
5018 static inline bool netif_needs_gso(struct sk_buff *skb,
5019 				   netdev_features_t features)
5020 {
5021 	return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
5022 		unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
5023 			 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
5024 }
5025 
5026 void netif_set_tso_max_size(struct net_device *dev, unsigned int size);
5027 void netif_set_tso_max_segs(struct net_device *dev, unsigned int segs);
5028 void netif_inherit_tso_max(struct net_device *to,
5029 			   const struct net_device *from);
5030 
5031 static inline bool netif_is_macsec(const struct net_device *dev)
5032 {
5033 	return dev->priv_flags & IFF_MACSEC;
5034 }
5035 
5036 static inline bool netif_is_macvlan(const struct net_device *dev)
5037 {
5038 	return dev->priv_flags & IFF_MACVLAN;
5039 }
5040 
5041 static inline bool netif_is_macvlan_port(const struct net_device *dev)
5042 {
5043 	return dev->priv_flags & IFF_MACVLAN_PORT;
5044 }
5045 
5046 static inline bool netif_is_bond_master(const struct net_device *dev)
5047 {
5048 	return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
5049 }
5050 
5051 static inline bool netif_is_bond_slave(const struct net_device *dev)
5052 {
5053 	return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
5054 }
5055 
5056 static inline bool netif_supports_nofcs(struct net_device *dev)
5057 {
5058 	return dev->priv_flags & IFF_SUPP_NOFCS;
5059 }
5060 
5061 static inline bool netif_has_l3_rx_handler(const struct net_device *dev)
5062 {
5063 	return dev->priv_flags & IFF_L3MDEV_RX_HANDLER;
5064 }
5065 
5066 static inline bool netif_is_l3_master(const struct net_device *dev)
5067 {
5068 	return dev->priv_flags & IFF_L3MDEV_MASTER;
5069 }
5070 
5071 static inline bool netif_is_l3_slave(const struct net_device *dev)
5072 {
5073 	return dev->priv_flags & IFF_L3MDEV_SLAVE;
5074 }
5075 
5076 static inline int dev_sdif(const struct net_device *dev)
5077 {
5078 #ifdef CONFIG_NET_L3_MASTER_DEV
5079 	if (netif_is_l3_slave(dev))
5080 		return dev->ifindex;
5081 #endif
5082 	return 0;
5083 }
5084 
5085 static inline bool netif_is_bridge_master(const struct net_device *dev)
5086 {
5087 	return dev->priv_flags & IFF_EBRIDGE;
5088 }
5089 
5090 static inline bool netif_is_bridge_port(const struct net_device *dev)
5091 {
5092 	return dev->priv_flags & IFF_BRIDGE_PORT;
5093 }
5094 
5095 static inline bool netif_is_ovs_master(const struct net_device *dev)
5096 {
5097 	return dev->priv_flags & IFF_OPENVSWITCH;
5098 }
5099 
5100 static inline bool netif_is_ovs_port(const struct net_device *dev)
5101 {
5102 	return dev->priv_flags & IFF_OVS_DATAPATH;
5103 }
5104 
5105 static inline bool netif_is_any_bridge_master(const struct net_device *dev)
5106 {
5107 	return netif_is_bridge_master(dev) || netif_is_ovs_master(dev);
5108 }
5109 
5110 static inline bool netif_is_any_bridge_port(const struct net_device *dev)
5111 {
5112 	return netif_is_bridge_port(dev) || netif_is_ovs_port(dev);
5113 }
5114 
5115 static inline bool netif_is_team_master(const struct net_device *dev)
5116 {
5117 	return dev->priv_flags & IFF_TEAM;
5118 }
5119 
5120 static inline bool netif_is_team_port(const struct net_device *dev)
5121 {
5122 	return dev->priv_flags & IFF_TEAM_PORT;
5123 }
5124 
5125 static inline bool netif_is_lag_master(const struct net_device *dev)
5126 {
5127 	return netif_is_bond_master(dev) || netif_is_team_master(dev);
5128 }
5129 
5130 static inline bool netif_is_lag_port(const struct net_device *dev)
5131 {
5132 	return netif_is_bond_slave(dev) || netif_is_team_port(dev);
5133 }
5134 
5135 static inline bool netif_is_rxfh_configured(const struct net_device *dev)
5136 {
5137 	return dev->priv_flags & IFF_RXFH_CONFIGURED;
5138 }
5139 
5140 static inline bool netif_is_failover(const struct net_device *dev)
5141 {
5142 	return dev->priv_flags & IFF_FAILOVER;
5143 }
5144 
5145 static inline bool netif_is_failover_slave(const struct net_device *dev)
5146 {
5147 	return dev->priv_flags & IFF_FAILOVER_SLAVE;
5148 }
5149 
5150 /* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
5151 static inline void netif_keep_dst(struct net_device *dev)
5152 {
5153 	dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
5154 }
5155 
5156 /* return true if dev can't cope with mtu frames that need vlan tag insertion */
5157 static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
5158 {
5159 	/* TODO: reserve and use an additional IFF bit, if we get more users */
5160 	return netif_is_macsec(dev);
5161 }
5162 
5163 extern struct pernet_operations __net_initdata loopback_net_ops;
5164 
5165 /* Logging, debugging and troubleshooting/diagnostic helpers. */
5166 
5167 /* netdev_printk helpers, similar to dev_printk */
5168 
5169 static inline const char *netdev_name(const struct net_device *dev)
5170 {
5171 	if (!dev->name[0] || strchr(dev->name, '%'))
5172 		return "(unnamed net_device)";
5173 	return dev->name;
5174 }
5175 
5176 static inline const char *netdev_reg_state(const struct net_device *dev)
5177 {
5178 	switch (dev->reg_state) {
5179 	case NETREG_UNINITIALIZED: return " (uninitialized)";
5180 	case NETREG_REGISTERED: return "";
5181 	case NETREG_UNREGISTERING: return " (unregistering)";
5182 	case NETREG_UNREGISTERED: return " (unregistered)";
5183 	case NETREG_RELEASED: return " (released)";
5184 	case NETREG_DUMMY: return " (dummy)";
5185 	}
5186 
5187 	WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
5188 	return " (unknown)";
5189 }
5190 
5191 #define MODULE_ALIAS_NETDEV(device) \
5192 	MODULE_ALIAS("netdev-" device)
5193 
5194 /*
5195  * netdev_WARN() acts like dev_printk(), but with the key difference
5196  * of using a WARN/WARN_ON to get the message out, including the
5197  * file/line information and a backtrace.
5198  */
5199 #define netdev_WARN(dev, format, args...)			\
5200 	WARN(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5201 	     netdev_reg_state(dev), ##args)
5202 
5203 #define netdev_WARN_ONCE(dev, format, args...)				\
5204 	WARN_ONCE(1, "netdevice: %s%s: " format, netdev_name(dev),	\
5205 		  netdev_reg_state(dev), ##args)
5206 
5207 /*
5208  *	The list of packet types we will receive (as opposed to discard)
5209  *	and the routines to invoke.
5210  *
5211  *	Why 16. Because with 16 the only overlap we get on a hash of the
5212  *	low nibble of the protocol value is RARP/SNAP/X.25.
5213  *
5214  *		0800	IP
5215  *		0001	802.3
5216  *		0002	AX.25
5217  *		0004	802.2
5218  *		8035	RARP
5219  *		0005	SNAP
5220  *		0805	X.25
5221  *		0806	ARP
5222  *		8137	IPX
5223  *		0009	Localtalk
5224  *		86DD	IPv6
5225  */
5226 #define PTYPE_HASH_SIZE	(16)
5227 #define PTYPE_HASH_MASK	(PTYPE_HASH_SIZE - 1)
5228 
5229 extern struct list_head ptype_all __read_mostly;
5230 extern struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
5231 
5232 extern struct net_device *blackhole_netdev;
5233 
5234 /* Note: Avoid these macros in fast path, prefer per-cpu or per-queue counters. */
5235 #define DEV_STATS_INC(DEV, FIELD) atomic_long_inc(&(DEV)->stats.__##FIELD)
5236 #define DEV_STATS_ADD(DEV, FIELD, VAL) 	\
5237 		atomic_long_add((VAL), &(DEV)->stats.__##FIELD)
5238 
5239 #endif	/* _LINUX_NETDEVICE_H */
5240